build: sync glusterfs.spec.in with Fedora glusterfs.spec

* add glusterfs-api and glusterfs-api-devel sub-packages * add extras/systemd/glusterd.service. Use it (or extras/init.d/glusterd) instead of the Fedora versions. * restores ability to build glusterfs RPMs from the dist tarball with 'rpmbuild {-ts,-tb,-ta} glusterfs-XXX.tar.gz' * other minor cleanups mostly to sync with fedora .spec and to build from the dist tarball. Any differences will be resolved in the fedora .spec in the next release (i.e. beta4 or GA). (still considering whether to add fedora glusterfsd.{init,service} files to glusterfs or remove them from fedora.) BUG: 950083 Change-Id: Ibda3cd57b24ef5c2a27446703e718e5044ec546c Signed-off-by: Kaleb S. KEITHLEY <kkeithle@redhat.com> Reviewed-on: http://review.gluster.org/5230 Tested-by: Gluster Build System <jenkins@build.gluster.com> Reviewed-by: Bala FA <barumuga@redhat.com> Reviewed-by: Niels de Vos <ndevos@redhat.com>
author: Kaleb S. KEITHLEY <kkeithle@redhat.com> 2013-06-17 13:44:10 -0400
committer: Vijay Bellur <vbellur@redhat.com> 2013-06-26 04:36:47 -0700
commit: f7b1245ca221fd7d4cee626ae358a2ebc933ffdd (patch)
tree: 565477a07be9aaf891fbc858f74300d93bc1fdef /extras
parent: 183546aa2dbfe3371cf155800e2f70057e95e2bc (diff)
5 files changed, 38 insertions, 13 deletions
diff --git a/extras/LinuxRPM/Makefile.am b/extras/LinuxRPM/Makefile.am
index 09d2f9e4201..1dafa982bd6 100644
--- a/extras/LinuxRPM/Makefile.am
+++ b/extras/LinuxRPM/Makefile.am
@@ -1,6 +1,5 @@
 
-GFS_TAR = ../../glusterfs-*git.tar.gz
-GFS_SPEC = ../../glusterfs.spec
+GFS_TAR = ../../glusterfs-$(VERSION).tar.gz
 
 .PHONY: all
 
@@ -12,9 +11,9 @@ all:
 glusterrpms: prep srcrpm rpms
 	-rm -rf rpmbuild
 
-prep::
-	if [ ! -e $(GFS_TAR) -a ! -e $(GFS_SPEC) ]; then \
-		$(MAKE) -c ../.. dist; \
+prep:
+	if [ ! -e $(GFS_TAR) ]; then \
+		$(MAKE) -C ../.. dist; \
 	fi
 	-mkdir -p rpmbuild/SPECS
 	-mkdir -p rpmbuild/RPMS
diff --git a/extras/Makefile.am b/extras/Makefile.am
index 7b55ef2ee7e..63ad7101655 100644
--- a/extras/Makefile.am
+++ b/extras/Makefile.am
@@ -2,7 +2,7 @@
 EditorModedir = $(docdir)
 EditorMode_DATA = glusterfs-mode.el glusterfs.vim
 
-SUBDIRS = init.d benchmarking hook-scripts $(OCF_SUBDIR) LinuxRPM
+SUBDIRS = init.d systemd benchmarking hook-scripts $(OCF_SUBDIR) LinuxRPM
 
 confdir = $(sysconfdir)/glusterfs
 conf_DATA = glusterfs-logrotate
@@ -10,5 +10,5 @@ conf_DATA = glusterfs-logrotate
 EXTRA_DIST = specgen.scm MacOSX/Portfile glusterfs-mode.el glusterfs.vim  \
 	migrate-unify-to-distribute.sh backend-xattr-sanitize.sh          \
 	backend-cleanup.sh disk_usage_sync.sh quota-remove-xattr.sh       \
-	quota-metadata-cleanup.sh glusterfs-logrotate clear_xattrs.sh	\
+	quota-metadata-cleanup.sh glusterfs-logrotate clear_xattrs.sh     \
 	group-virt.example glusterd-sysconfig
diff --git a/extras/init.d/Makefile.am b/extras/init.d/Makefile.am
index 7c26a0e8e2b..38898fddd2e 100644
--- a/extras/init.d/Makefile.am
+++ b/extras/init.d/Makefile.am
@@ -3,17 +3,20 @@ EXTRA_DIST = glusterd-Debian glusterd-Redhat glusterd-SuSE glusterd.plist rhel5-
 
 CLEANFILES = 
 
-initdir = @initdir@
-launchddir = @launchddir@
+INIT_DIR = @initdir@
+SYSTEMD_DIR = @systemddir@
+LAUNCHD_DIR = @launchddir@
 
 $(GF_DISTRIBUTION): 
-	$(mkdir_p) $(DESTDIR)$(initdir)	
-	$(INSTALL_PROGRAM) glusterd-$(GF_DISTRIBUTION) $(DESTDIR)$(initdir)/glusterd
+	@if [ ! -d $(SYSTEMD_DIR) ]; then \
+		$(mkdir_p) $(DESTDIR)$(INIT_DIR); \
+		$(INSTALL_PROGRAM) glusterd-$(GF_DISTRIBUTION) $(DESTDIR)$(INIT_DIR)/glusterd; \
+	fi
 
 install-exec-local: $(GF_DISTRIBUTION)
 
 install-data-local: 
 if GF_DARWIN_HOST_OS
-	$(mkdir_p) $(DESTDIR)$(launchddir)
-	$(INSTALL_PROGRAM) glusterd.plist $(DESTDIR)$(launchddir)/com.gluster.glusterd.plist
+	$(mkdir_p) $(DESTDIR)$(LAUNCHD_DIR)
+	$(INSTALL_PROGRAM) glusterd.plist $(DESTDIR)$(LAUNCHD_DIR)/com.gluster.glusterd.plist
 endif
diff --git a/extras/systemd/Makefile.am b/extras/systemd/Makefile.am
new file mode 100644
index 00000000000..3fc656b8262
--- /dev/null
+++ b/extras/systemd/Makefile.am
@@ -0,0 +1,11 @@
+
+CLEANFILES =
+
+SYSTEMD_DIR = @systemddir@
+
+install-exec-local:
+	@if [ -d $(SYSTEMD_DIR) ]; then \
+		$(mkdir_p) $(DESTDIR)$(SYSTEMD_DIR); \
+		$(INSTALL_PROGRAM) glusterd.service $(DESTDIR)$(SYSTEMD_DIR)/; \
+	fi
+
diff --git a/extras/systemd/glusterd.service.in b/extras/systemd/glusterd.service.in
new file mode 100644
index 00000000000..0b24aea3bec
--- /dev/null
+++ b/extras/systemd/glusterd.service.in
@@ -0,0 +1,12 @@
+[Unit]
+Description=GlusterFS an clustered file-system server
+After=network.target rpcbind.service
+
+[Service]
+Type=forking
+PIDFile=/run/glusterd.pid
+LimitNOFILE=65536
+ExecStart=/usr/sbin/glusterd -p /run/glusterd.pid
+
+[Install]
+WantedBy=multi-user.target
diff --git a/xlators/Makefile.am b/xlators/Makefile.am
index 4c94f5e44..f60fa85ce 100644
--- a/xlators/Makefile.am
+++ b/xlators/Makefile.am
@@ -1,3 +1,4 @@
-SUBDIRS = cluster storage protocol performance debug features encryption mount nfs mgmt
+SUBDIRS = cluster storage protocol performance debug features encryption mount nfs mgmt system \
+          playground
 
 CLEANFILES =
diff --git a/xlators/bindings/python/src/Makefile.am b/xlators/bindings/python/src/Makefile.am
index c0b9141c6..90370d861 100644
--- a/xlators/bindings/python/src/Makefile.am
+++ b/xlators/bindings/python/src/Makefile.am
@@ -9,7 +9,7 @@ pythondir = $(xlatordir)/python
 
 python_so_SOURCES = python.c
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall \
+AM_CFLAGS = -fPIC $(GF_CPPFLAGS) -Wall \
 	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles \
 	$(PYTHON_CPPLAGS) -DGLUSTER_PYTHON_PATH=\"$(pythondir)\"
 
diff --git a/xlators/bindings/python/src/gluster.py b/xlators/bindings/python/src/gluster.py
index ee0eb1310..337c983ec 100644
--- a/xlators/bindings/python/src/gluster.py
+++ b/xlators/bindings/python/src/gluster.py
@@ -1,19 +1,12 @@
-#   Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-#   This file is part of GlusterFS.
-#
-#   GlusterFS is free software; you can redistribute it and/or modify
-#   it under the terms of the GNU General Public License as published
-#   by the Free Software Foundation; either version 3 of the License,
-#   or (at your option) any later version.
-#
-#   GlusterFS is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-#   General Public License for more details.
+
+#  Copyright (c) 2007-2012 Red Hat, Inc. <http://www.redhat.com>
+#  This file is part of GlusterFS.
 #
-#   You should have received a copy of the GNU General Public License
-#   along with this program.  If not, see
-#   <http://www.gnu.org/licenses/>.
+#  This file is licensed to you under your choice of the GNU Lesser
+#  General Public License, version 3 or any later version (LGPLv3 or
+#  later), or the GNU General Public License, version 2 (GPLv2), in all
+#  cases as published by the Free Software Foundation.
+
 from ctypes import *
 from glustertypes import *
 from glusterstack import *
diff --git a/xlators/bindings/python/src/glusterstack.py b/xlators/bindings/python/src/glusterstack.py
index ba24c8165..0c071ae98 100644
--- a/xlators/bindings/python/src/glusterstack.py
+++ b/xlators/bindings/python/src/glusterstack.py
@@ -1,19 +1,12 @@
-#   Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-#   This file is part of GlusterFS.
-#
-#   GlusterFS is free software; you can redistribute it and/or modify
-#   it under the terms of the GNU General Public License as published
-#   by the Free Software Foundation; either version 3 of the License,
-#   or (at your option) any later version.
-#
-#   GlusterFS is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-#   General Public License for more details.
+
+#  Copyright (c) 2007-2012 Red Hat, Inc. <http://www.redhat.com>
+#  This file is part of GlusterFS.
 #
-#   You should have received a copy of the GNU General Public License
-#   along with this program.  If not, see
-#   <http://www.gnu.org/licenses/>.
+#  This file is licensed to you under your choice of the GNU Lesser
+#  General Public License, version 3 or any later version (LGPLv3 or
+#  later), or the GNU General Public License, version 2 (GPLv2), in all
+#  cases as published by the Free Software Foundation.
+
 from ctypes import *
 from glustertypes import *
 
diff --git a/xlators/bindings/python/src/glustertypes.py b/xlators/bindings/python/src/glustertypes.py
index e9069d07c..98437d22e 100644
--- a/xlators/bindings/python/src/glustertypes.py
+++ b/xlators/bindings/python/src/glustertypes.py
@@ -1,19 +1,12 @@
-#   Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-#   This file is part of GlusterFS.
-#
-#   GlusterFS is free software; you can redistribute it and/or modify
-#   it under the terms of the GNU General Public License as published
-#   by the Free Software Foundation; either version 3 of the License,
-#   or (at your option) any later version.
-#
-#   GlusterFS is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-#   General Public License for more details.
+
+#  Copyright (c) 2007-2012 Red Hat, Inc. <http://www.redhat.com>
+#  This file is part of GlusterFS.
 #
-#   You should have received a copy of the GNU General Public License
-#   along with this program.  If not, see
-#   <http://www.gnu.org/licenses/>.
+#  This file is licensed to you under your choice of the GNU Lesser
+#  General Public License, version 3 or any later version (LGPLv3 or
+#  later), or the GNU General Public License, version 2 (GPLv2), in all
+#  cases as published by the Free Software Foundation.
+
 from ctypes import *
 import collections
 
diff --git a/xlators/bindings/python/src/python.c b/xlators/bindings/python/src/python.c
index c11323cda..9b96790de 100644
--- a/xlators/bindings/python/src/python.c
+++ b/xlators/bindings/python/src/python.c
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2007-2009 Chris AtLee <chris@atlee.ca>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #include <Python.h>
 
 #ifndef _CONFIG_H
@@ -45,7 +35,7 @@ python_writev (call_frame_t *frame,
               xlator_t *this,
               fd_t *fd,
               struct iovec *vector,
-              int32_t count, 
+              int32_t count,
               off_t offset)
 {
   python_private_t *priv = (python_private_t *)this->private;
@@ -148,7 +138,7 @@ init (xlator_t *this)
   Py_InitializeEx(0);
 
   if (!this->children) {
-    gf_log ("python", GF_LOG_ERROR, 
+    gf_log ("python", GF_LOG_ERROR,
             "FATAL: python should have exactly one child");
     return -1;
   }
@@ -166,7 +156,7 @@ init (xlator_t *this)
   }
 
   priv->pInterp = Py_NewInterpreter();
-    
+
   // Adjust python's path
   PyObject *syspath = PySys_GetObject("path");
   PyObject *path = PyString_FromString(GLUSTER_PYTHON_PATH);
@@ -188,7 +178,7 @@ init (xlator_t *this)
   priv->pVectorType = PyObject_GetAttrString(priv->pGlusterModule, "iovec");
 
   gf_log("python", GF_LOG_DEBUG, "Loading script...%s", priv->scriptname);
-  
+
   priv->pScriptModule = AnonModule_FromFile(priv->scriptname);
   if (!priv->pScriptModule || PyErr_Occurred())
   {
@@ -217,7 +207,7 @@ init (xlator_t *this)
   return 0;
 }
 
-void 
+void
 fini (xlator_t *this)
 {
   python_private_t *priv = (python_private_t*)(this->private);
diff --git a/xlators/bindings/python/src/testxlator.py b/xlators/bindings/python/src/testxlator.py
index 507455c85..59a991dca 100644
--- a/xlators/bindings/python/src/testxlator.py
+++ b/xlators/bindings/python/src/testxlator.py
@@ -1,19 +1,12 @@
-#   Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-#   This file is part of GlusterFS.
-#
-#   GlusterFS is free software; you can redistribute it and/or modify
-#   it under the terms of the GNU General Public License as published
-#   by the Free Software Foundation; either version 3 of the License,
-#   or (at your option) any later version.
-#
-#   GlusterFS is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-#   General Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License
-#   along with this program.  If not, see
-#   <http://www.gnu.org/licenses/>.
+"""
+  Copyright (c) 2007-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+"""
 
 """
 This is a test translator written in python.
diff --git a/xlators/cluster/afr/src/Makefile.am b/xlators/cluster/afr/src/Makefile.am
index 4e4b4c752..35d18a6c0 100644
--- a/xlators/cluster/afr/src/Makefile.am
+++ b/xlators/cluster/afr/src/Makefile.am
@@ -1,20 +1,31 @@
 xlator_LTLIBRARIES = afr.la pump.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
 
-afr_common_source = afr-dir-read.c afr-dir-write.c afr-inode-read.c afr-inode-write.c afr-open.c afr-transaction.c afr-self-heal-data.c afr-self-heal-common.c afr-self-heal-metadata.c afr-self-heal-entry.c afr-self-heal-algorithm.c afr-lk-common.c
+afr_common_source = afr-dir-read.c afr-dir-write.c afr-inode-read.c \
+	afr-inode-write.c afr-open.c afr-transaction.c afr-self-heal-data.c \
+	afr-self-heal-common.c afr-self-heal-metadata.c afr-self-heal-entry.c \
+	afr-self-heal-algorithm.c afr-lk-common.c afr-self-heald.c \
+	$(top_builddir)/xlators/lib/src/libxlator.c
 
-afr_la_LDFLAGS = -module -avoidversion
+afr_la_LDFLAGS = -module -avoid-version
 afr_la_SOURCES = $(afr_common_source) afr.c
 afr_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-pump_la_LDFLAGS = -module -avoidversion
+pump_la_LDFLAGS = -module -avoid-version
 pump_la_SOURCES =  $(afr_common_source) pump.c
 pump_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = afr.h afr-transaction.h afr-inode-write.h afr-inode-read.h afr-dir-read.h afr-dir-write.h afr-self-heal.h afr-self-heal-common.h afr-self-heal-algorithm.h pump.h afr-mem-types.h afr-common.c
+noinst_HEADERS = afr.h afr-transaction.h afr-inode-write.h afr-inode-read.h \
+	afr-dir-read.h afr-dir-write.h afr-self-heal.h afr-self-heal-common.h \
+	afr-self-heal-algorithm.h pump.h afr-mem-types.h afr-common.c \
+	afr-self-heald.h $(top_builddir)/xlators/lib/src/libxlator.h \
+	$(top_builddir)/glusterfsd/src/glusterfsd.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	    -I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/contrib/md5 -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) \
+	-I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/xlators/lib/src \
+	-I$(top_srcdir)/rpc/rpc-lib/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES =
 
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index 9135e84ed..af01f2ef2 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <libgen.h>
@@ -44,6 +35,7 @@
 #include "compat.h"
 #include "byte-order.h"
 #include "statedump.h"
+#include "inode.h"
 
 #include "fd.h"
 
@@ -54,524 +46,1199 @@
 #include "afr-transaction.h"
 #include "afr-self-heal.h"
 #include "afr-self-heal-common.h"
+#include "afr-self-heald.h"
 #include "pump.h"
 
-#define AFR_ICTX_OPENDIR_DONE_MASK     0x0000000200000000ULL
-#define AFR_ICTX_SPLIT_BRAIN_MASK      0x0000000100000000ULL
+#define AFR_ICTX_OPENDIR_DONE_MASK     0x0000000100000000ULL
 #define AFR_ICTX_READ_CHILD_MASK       0x00000000FFFFFFFFULL
+#define AFR_STATISTICS_HISTORY_SIZE    50
+int
+afr_lookup_done_success_action (call_frame_t *frame, xlator_t *this,
+                                gf_boolean_t fail_conflict);
+void
+afr_children_copy (int32_t *dst, int32_t *src, unsigned int child_count)
+{
+        int     i = 0;
+
+        for (i = 0; i < child_count; i++)
+                dst[i] = src[i];
+}
+
+void
+afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req, const char *path)
+{
+        int             i           = 0;
+        afr_private_t   *priv       = NULL;
+        int             ret         = 0;
+
+        priv   = this->private;
+
+        for (i = 0; i < priv->child_count; i++) {
+                ret = dict_set_uint64 (xattr_req, priv->pending_key[i],
+                                       3 * sizeof(int32_t));
+                if (ret < 0)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: Unable to set dict value for %s",
+                                path, priv->pending_key[i]);
+                /* 3 = data+metadata+entry */
+        }
+        ret = dict_set_int32 (xattr_req, GF_GFIDLESS_LOOKUP, 1);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "%s: failed to set gfidless "
+                        "lookup", path);
+        }
+}
+
+int
+afr_lookup_xattr_req_prepare (afr_local_t *local, xlator_t *this,
+                              dict_t *xattr_req, loc_t *loc, void **gfid_req)
+{
+        int     ret = -ENOMEM;
+
+        GF_ASSERT (gfid_req);
+
+        *gfid_req = NULL;
+        local->xattr_req = dict_new ();
+        if (!local->xattr_req)
+                goto out;
+        if (xattr_req)
+                dict_copy (xattr_req, local->xattr_req);
+
+        afr_xattr_req_prepare (this, local->xattr_req, loc->path);
+        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: Unable to set dict value for %s",
+                        loc->path, GLUSTERFS_INODELK_COUNT);
+        }
+        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: Unable to set dict value for %s",
+                        loc->path, GLUSTERFS_ENTRYLK_COUNT);
+        }
+
+        ret = dict_set_uint32 (local->xattr_req, GLUSTERFS_PARENT_ENTRYLK, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: Unable to set dict value for %s",
+                        loc->path, GLUSTERFS_PARENT_ENTRYLK);
+        }
+
+        ret = dict_get_ptr (local->xattr_req, "gfid-req", gfid_req);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "%s: failed to get the gfid from dict", loc->path);
+                *gfid_req = NULL;
+        } else {
+                if (loc->parent != NULL)
+                        dict_del (local->xattr_req, "gfid-req");
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+void
+afr_lookup_save_gfid (uuid_t dst, void* new, const loc_t *loc)
+{
+        inode_t  *inode = NULL;
+
+        inode = loc->inode;
+        if (inode && !uuid_is_null (inode->gfid))
+                uuid_copy (dst, inode->gfid);
+        else if (!uuid_is_null (loc->gfid))
+                uuid_copy (dst, loc->gfid);
+        else if (new && !uuid_is_null (new))
+                uuid_copy (dst, new);
+}
+
+int
+afr_errno_count (int32_t *children, int *child_errno,
+                 unsigned int child_count, int32_t op_errno)
+{
+        int i = 0;
+        int errno_count = 0;
+        int child = 0;
+
+        for (i = 0; i < child_count; i++) {
+                if (children) {
+                        child = children[i];
+                        if (child == -1)
+                                break;
+                } else {
+                        child = i;
+                }
+                if (child_errno[child] == op_errno)
+                        errno_count++;
+        }
+        return errno_count;
+}
 
 int32_t
 afr_set_dict_gfid (dict_t *dict, uuid_t gfid)
 {
-        int ret = 0;
+        int ret       = 0;
+        uuid_t *pgfid = NULL;
 
         GF_ASSERT (gfid);
 
-        ret = dict_set_static_bin (dict, "gfid-req", gfid, 16);
+        pgfid = GF_CALLOC (1, sizeof (uuid_t), gf_common_mt_char);
+        if (!pgfid) {
+                ret = -1;
+                goto out;
+        }
+
+        uuid_copy (*pgfid, gfid);
+
+        ret = dict_set_dynptr (dict, "gfid-req", pgfid, sizeof (uuid_t));
         if (ret)
-                gf_log (THIS->name, GF_LOG_DEBUG, "gfid set failed");
+                gf_log (THIS->name, GF_LOG_ERROR, "gfid set failed");
+
+out:
+        if (ret && pgfid)
+                GF_FREE (pgfid);
 
         return ret;
 }
 
-uint64_t
-afr_is_split_brain (xlator_t *this, inode_t *inode)
+void
+afr_inode_ctx_destroy (afr_inode_ctx_t *ctx)
 {
-        int ret = 0;
+        if (!ctx)
+                return;
+        GF_FREE (ctx->fresh_children);
+        GF_FREE (ctx);
+}
+
+afr_inode_ctx_t*
+__afr_inode_ctx_get (inode_t *inode, xlator_t *this)
+{
+        int             ret      = 0;
+        uint64_t        ctx_addr = 0;
+        afr_inode_ctx_t *ctx     = NULL;
+        afr_private_t   *priv    = NULL;
+
+        priv = this->private;
+        ret = __inode_ctx_get (inode, this, &ctx_addr);
+        if (ret < 0)
+                ctx_addr = 0;
+        if (ctx_addr != 0) {
+                ctx = (afr_inode_ctx_t*) (long) ctx_addr;
+                goto out;
+        }
+        ctx = GF_CALLOC (1, sizeof (*ctx),
+                         gf_afr_mt_inode_ctx_t);
+        if (!ctx)
+                goto fail;
+        ctx->fresh_children = GF_CALLOC (priv->child_count,
+                                         sizeof (*ctx->fresh_children),
+                                         gf_afr_mt_int32_t);
+        if (!ctx->fresh_children)
+                goto fail;
+        ret = __inode_ctx_put (inode, this, (uint64_t)ctx);
+        if (ret) {
+                gf_log_callingfn (this->name, GF_LOG_ERROR, "failed to "
+                                  "set the inode ctx (%s)",
+                                  uuid_utoa (inode->gfid));
+                goto fail;
+        }
 
-        uint64_t ctx         = 0;
-        uint64_t split_brain = 0;
+out:
+        return ctx;
+
+fail:
+        afr_inode_ctx_destroy (ctx);
+        return NULL;
+}
 
-        VALIDATE_OR_GOTO (inode, out);
+afr_inode_ctx_t*
+afr_inode_ctx_get (inode_t *inode, xlator_t *this)
+{
+        afr_inode_ctx_t *ctx = NULL;
 
         LOCK (&inode->lock);
         {
-                ret = __inode_ctx_get (inode, this, &ctx);
+                ctx = __afr_inode_ctx_get (inode, this);
+        }
+        UNLOCK (&inode->lock);
+        return ctx;
+}
 
-                if (ret < 0)
-                        goto unlock;
+void
+afr_inode_get_ctx_params (xlator_t *this, inode_t *inode,
+                          afr_inode_params_t *params)
+{
+        GF_ASSERT (inode);
+        GF_ASSERT (params);
+
+        afr_inode_ctx_t *ctx = NULL;
+        afr_private_t   *priv = NULL;
+        int             i = 0;
+        int32_t         read_child = -1;
+        int32_t         *fresh_children = NULL;
 
-                split_brain = ctx & AFR_ICTX_SPLIT_BRAIN_MASK;
+        priv = this->private;
+        LOCK (&inode->lock);
+        {
+                ctx = __afr_inode_ctx_get (inode, this);
+                if (!ctx)
+                        goto unlock;
+                switch (params->op) {
+                case AFR_INODE_GET_READ_CTX:
+                        fresh_children = params->u.read_ctx.children;
+                        read_child = (int32_t)(ctx->masks &
+                                               AFR_ICTX_READ_CHILD_MASK);
+                        params->u.read_ctx.read_child = read_child;
+                        if (!fresh_children)
+                                goto unlock;
+                        for (i = 0; i < priv->child_count; i++)
+                                fresh_children[i] = ctx->fresh_children[i];
+                        break;
+                case AFR_INODE_GET_OPENDIR_DONE:
+                        params->u.value = _gf_false;
+                        if (ctx->masks & AFR_ICTX_OPENDIR_DONE_MASK)
+                                params->u.value = _gf_true;
+                        break;
+                default:
+                        GF_ASSERT (0);
+                        break;
+                }
         }
 unlock:
         UNLOCK (&inode->lock);
+}
 
+gf_boolean_t
+afr_is_split_brain (xlator_t *this, inode_t *inode)
+{
+        afr_inode_ctx_t *ctx = NULL;
+        gf_boolean_t    spb  = _gf_false;
+
+        ctx = afr_inode_ctx_get (inode, this);
+        if (!ctx)
+                goto out;
+        if ((ctx->mdata_spb == SPB) || (ctx->data_spb == SPB))
+                spb = _gf_true;
 out:
-        return split_brain;
+        return spb;
 }
 
+gf_boolean_t
+afr_is_opendir_done (xlator_t *this, inode_t *inode)
+{
+        afr_inode_params_t params = {0};
 
-void
-afr_set_split_brain (xlator_t *this, inode_t *inode, gf_boolean_t set)
+        params.op = AFR_INODE_GET_OPENDIR_DONE;
+        afr_inode_get_ctx_params (this, inode, &params);
+        return params.u.value;
+}
+
+int32_t
+afr_inode_get_read_ctx (xlator_t *this, inode_t *inode, int32_t *fresh_children)
 {
-        uint64_t ctx = 0;
-        int      ret = 0;
+        afr_inode_params_t      params = {0};
 
-        VALIDATE_OR_GOTO (inode, out);
+        params.op = AFR_INODE_GET_READ_CTX;
+        params.u.read_ctx.children = fresh_children;
+        afr_inode_get_ctx_params (this, inode, &params);
+        return params.u.read_ctx.read_child;
+}
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+void
+afr_inode_ctx_set_read_child (afr_inode_ctx_t *ctx, int32_t read_child)
+{
+        uint64_t        remaining_mask = 0;
+        uint64_t        mask         = 0;
 
-                if (ret < 0) {
-                        ctx = 0;
-                }
+        remaining_mask = (~AFR_ICTX_READ_CHILD_MASK & ctx->masks);
+        mask = (AFR_ICTX_READ_CHILD_MASK & read_child);
+        ctx->masks = remaining_mask | mask;
+}
 
-                if (set) {
-			ctx = (~AFR_ICTX_SPLIT_BRAIN_MASK & ctx)
-                                | (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_SPLIT_BRAIN_MASK);
-		} else {
-			ctx = (~AFR_ICTX_SPLIT_BRAIN_MASK & ctx);
-		}
-                __inode_ctx_put (inode, this, ctx);
+void
+afr_inode_ctx_set_read_ctx (afr_inode_ctx_t *ctx, int32_t read_child,
+                            int32_t *fresh_children, int32_t child_count)
+{
+        int             i            = 0;
+
+        afr_inode_ctx_set_read_child (ctx, read_child);
+        for (i = 0; i < child_count; i++) {
+                if (fresh_children)
+                        ctx->fresh_children[i] = fresh_children[i];
+                else
+                        ctx->fresh_children[i] = -1;
         }
-        UNLOCK (&inode->lock);
-out:
-        return;
 }
 
+void
+afr_inode_ctx_rm_stale_children (afr_inode_ctx_t *ctx, int32_t *stale_children,
+                                 int32_t child_count)
+{
+        int             i            = 0;
+        int32_t         read_child   = -1;
+
+        GF_ASSERT (stale_children);
+        for (i = 0; i < child_count; i++) {
+                if (stale_children[i] == -1)
+                        break;
+                afr_children_rm_child (ctx->fresh_children,
+                                       stale_children[i], child_count);
+        }
+        read_child = (int32_t)(ctx->masks & AFR_ICTX_READ_CHILD_MASK);
+        if (!afr_is_child_present (ctx->fresh_children, child_count,
+                                   read_child))
+                afr_inode_ctx_set_read_child (ctx, ctx->fresh_children[0]);
+}
 
-uint64_t
-afr_is_opendir_done (xlator_t *this, inode_t *inode)
+void
+afr_inode_ctx_set_opendir_done (afr_inode_ctx_t *ctx)
 {
-        int ret = 0;
+        uint64_t        remaining_mask = 0;
+        uint64_t        mask = 0;
+
+        remaining_mask = (~AFR_ICTX_OPENDIR_DONE_MASK & ctx->masks);
+        mask = (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_OPENDIR_DONE_MASK);
+        ctx->masks = remaining_mask | mask;
+}
 
-        uint64_t ctx          = 0;
-        uint64_t opendir_done = 0;
+void
+afr_inode_set_ctx_params (xlator_t *this, inode_t *inode,
+                          afr_inode_params_t *params)
+{
+        GF_ASSERT (inode);
+        GF_ASSERT (params);
 
-        VALIDATE_OR_GOTO (inode, out);
+        afr_inode_ctx_t *ctx            = NULL;
+        afr_private_t   *priv           = NULL;
+        int32_t         read_child      = -1;
+        int32_t         *fresh_children = NULL;
+        int32_t         *stale_children = NULL;
 
+        priv = this->private;
         LOCK (&inode->lock);
         {
-                ret = __inode_ctx_get (inode, this, &ctx);
-
-                if (ret < 0)
+                ctx = __afr_inode_ctx_get (inode, this);
+                if (!ctx)
                         goto unlock;
-
-                opendir_done = ctx & AFR_ICTX_OPENDIR_DONE_MASK;
+                switch (params->op) {
+                case AFR_INODE_SET_READ_CTX:
+                        read_child = params->u.read_ctx.read_child;
+                        fresh_children = params->u.read_ctx.children;
+                        afr_inode_ctx_set_read_ctx (ctx, read_child,
+                                                    fresh_children,
+                                                    priv->child_count);
+                        break;
+                case AFR_INODE_RM_STALE_CHILDREN:
+                        stale_children = params->u.read_ctx.children;
+                        afr_inode_ctx_rm_stale_children (ctx,
+                                                         stale_children,
+                                                         priv->child_count);
+                        break;
+                case AFR_INODE_SET_OPENDIR_DONE:
+                        afr_inode_ctx_set_opendir_done (ctx);
+                        break;
+                default:
+                        GF_ASSERT (0);
+                        break;
+                }
         }
 unlock:
         UNLOCK (&inode->lock);
-
-out:
-        return opendir_done;
 }
 
+void
+afr_set_split_brain (xlator_t *this, inode_t *inode, afr_spb_state_t mdata_spb,
+                     afr_spb_state_t data_spb)
+{
+        afr_inode_ctx_t *ctx = NULL;
+
+        ctx = afr_inode_ctx_get (inode, this);
+        if (mdata_spb != DONT_KNOW)
+                ctx->mdata_spb = mdata_spb;
+        if (data_spb != DONT_KNOW)
+                ctx->data_spb = data_spb;
+}
 
 void
 afr_set_opendir_done (xlator_t *this, inode_t *inode)
 {
-        uint64_t ctx = 0;
-        int      ret = 0;
+        afr_inode_params_t params = {0};
 
-        VALIDATE_OR_GOTO (inode, out);
+        params.op = AFR_INODE_SET_OPENDIR_DONE;
+        afr_inode_set_ctx_params (this, inode, &params);
+}
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+void
+afr_inode_set_read_ctx (xlator_t *this, inode_t *inode, int32_t read_child,
+                        int32_t *fresh_children)
+{
+        afr_inode_params_t params = {0};
+        afr_private_t      *priv  = NULL;
+
+        priv = this->private;
+        GF_ASSERT (read_child >= 0);
+        GF_ASSERT (fresh_children);
+        GF_ASSERT (afr_is_child_present (fresh_children, priv->child_count,
+                                         read_child));
+
+        params.op = AFR_INODE_SET_READ_CTX;
+        params.u.read_ctx.read_child     = read_child;
+        params.u.read_ctx.children = fresh_children;
+        afr_inode_set_ctx_params (this, inode, &params);
+}
+
+void
+afr_inode_rm_stale_children (xlator_t *this, inode_t *inode,
+                             int32_t *stale_children)
+{
+        afr_inode_params_t params = {0};
 
-                if (ret < 0) {
-                        ctx = 0;
+        GF_ASSERT (stale_children);
+
+        params.op = AFR_INODE_RM_STALE_CHILDREN;
+        params.u.read_ctx.children = stale_children;
+        afr_inode_set_ctx_params (this, inode, &params);
+}
+
+gf_boolean_t
+afr_is_source_child (int32_t *sources, int32_t child_count, int32_t child)
+{
+        gf_boolean_t             source_xattrs = _gf_false;
+
+        GF_ASSERT (child < child_count);
+
+        if ((child >= 0) && (child < child_count) &&
+             sources[child]) {
+                source_xattrs = _gf_true;
+        }
+        return source_xattrs;
+}
+
+gf_boolean_t
+afr_is_child_present (int32_t *success_children, int32_t child_count,
+                      int32_t child)
+{
+        gf_boolean_t             success_child = _gf_false;
+        int                      i = 0;
+
+        GF_ASSERT (child < child_count);
+
+        for (i = 0; i < child_count; i++) {
+                if (success_children[i] == -1)
+                        break;
+                if (child == success_children[i]) {
+                        success_child = _gf_true;
+                        break;
                 }
+        }
+        return success_child;
+}
 
-                ctx = (~AFR_ICTX_OPENDIR_DONE_MASK & ctx)
-                        | (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_OPENDIR_DONE_MASK);
+gf_boolean_t
+afr_is_read_child (int32_t *success_children, int32_t *sources,
+                   int32_t child_count, int32_t child)
+{
+        gf_boolean_t             success_child = _gf_false;
+        gf_boolean_t             source        = _gf_false;
 
-                __inode_ctx_put (inode, this, ctx);
+        if (child < 0) {
+                return _gf_false;
         }
-        UNLOCK (&inode->lock);
+
+        GF_ASSERT (success_children);
+        GF_ASSERT (child_count > 0);
+
+        success_child = afr_is_child_present (success_children, child_count,
+                                              child);
+        if (!success_child)
+                goto out;
+        if (NULL == sources) {
+                source = _gf_true;
+                goto out;
+        }
+        source = afr_is_source_child (sources, child_count, child);
 out:
-        return;
+        return (success_child && source);
 }
 
+int32_t
+afr_hash_child (int32_t *success_children, int32_t child_count,
+                unsigned int hmode, uuid_t gfid)
+{
+        uuid_t  gfid_copy = {0,};
+        pid_t pid;
 
-uint64_t
-afr_read_child (xlator_t *this, inode_t *inode)
+        if (!hmode) {
+                return -1;
+        }
+
+        if (gfid) {
+               uuid_copy(gfid_copy,gfid);
+        }
+        if (hmode > 1) {
+                /*
+                 * Why getpid?  Because it's one of the cheapest calls
+                 * available - faster than gethostname etc. - and returns a
+                 * constant-length value that's sure to be shorter than a UUID.
+                 * It's still very unlikely to be the same across clients, so
+                 * it still provides good mixing.  We're not trying for
+                 * perfection here.  All we need is a low probability that
+                 * multiple clients won't converge on the same subvolume.
+                 */
+                pid = getpid();
+                memcpy (gfid_copy, &pid, sizeof(pid));
+        }
+
+        return SuperFastHash((char *)gfid_copy,
+                             sizeof(gfid_copy)) % child_count;
+}
+
+/* If sources is NULL the xattrs are assumed to be of source for all
+ * success_children.
+ */
+int
+afr_select_read_child_from_policy (int32_t *success_children,
+                                   int32_t child_count, int32_t prev_read_child,
+                                   int32_t config_read_child, int32_t *sources,
+                                   unsigned int hmode, uuid_t gfid)
 {
-        int ret = 0;
+        int32_t                  read_child   = -1;
+        int                      i            = 0;
 
-        uint64_t ctx         = 0;
-        uint64_t read_child  = 0;
+        GF_ASSERT (success_children);
 
-        VALIDATE_OR_GOTO (inode, out);
+        read_child = config_read_child;
+        if (afr_is_read_child (success_children, sources, child_count,
+                               read_child))
+                goto out;
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+        read_child = prev_read_child;
+        if (afr_is_read_child (success_children, sources, child_count,
+                               read_child))
+                goto out;
 
-                if (ret < 0)
-                        goto unlock;
+        read_child = afr_hash_child (success_children, child_count,
+                                     hmode, gfid);
+        if (afr_is_read_child (success_children, sources, child_count,
+                               read_child)) {
+                goto out;
+        }
 
-                read_child = ctx & AFR_ICTX_READ_CHILD_MASK;
+        for (i = 0; i < child_count; i++) {
+                read_child = success_children[i];
+                if (read_child < 0)
+                        break;
+                if (afr_is_read_child (success_children, sources, child_count,
+                                       read_child))
+                        goto out;
         }
-unlock:
-        UNLOCK (&inode->lock);
+        read_child = -1;
 
 out:
         return read_child;
 }
 
-
+/* This function should be used when all the success_children are sources
+ */
 void
-afr_set_read_child (xlator_t *this, inode_t *inode, int32_t read_child)
+afr_set_read_ctx_from_policy (xlator_t *this, inode_t *inode,
+                              int32_t *fresh_children, int32_t prev_read_child,
+                              int32_t config_read_child, uuid_t gfid)
+{
+        int                      read_child = -1;
+        afr_private_t            *priv = NULL;
+
+        priv = this->private;
+        read_child = afr_select_read_child_from_policy (fresh_children,
+                                                        priv->child_count,
+                                                        prev_read_child,
+                                                        config_read_child,
+                                                        NULL,
+                                                        priv->hash_mode, gfid);
+        if (read_child >= 0)
+                afr_inode_set_read_ctx (this, inode, read_child,
+                                        fresh_children);
+}
+
+/* afr_next_call_child ()
+ * This is a common function used by all the read-type fops
+ * This function should not be called with the inode's read_children array.
+ * The fop's handler should make a copy of the inode's read_children,
+ * preferred read_child into the local vars, because while this function is
+ * in execution there is a chance for inode's read_ctx to change.
+ */
+int32_t
+afr_next_call_child (int32_t *fresh_children, unsigned char *child_up,
+                     size_t child_count, int32_t *last_index,
+                     int32_t read_child)
 {
-        uint64_t ctx = 0;
-        int      ret = 0;
+        int             next_index      = 0;
+        int32_t         next_call_child = -1;
 
-        VALIDATE_OR_GOTO (inode, out);
+        GF_ASSERT (last_index);
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+        next_index = *last_index;
+retry:
+        next_index++;
+        if ((next_index >= child_count) ||
+           (fresh_children[next_index] == -1))
+                goto out;
+        if ((fresh_children[next_index] == read_child) ||
+           (!child_up[fresh_children[next_index]]))
+                goto retry;
+        *last_index = next_index;
+        next_call_child = fresh_children[next_index];
+out:
+        return next_call_child;
+}
 
-                if (ret < 0) {
-                        ctx = 0;
+ /* This function should not be called with the inode's read_children array.
+ * The fop's handler should make a copy of the inode's read_children,
+ * preferred read_child into the local vars, because while this function is
+ * in execution there is a chance for inode's read_ctx to change.
+ */
+int32_t
+afr_get_call_child (xlator_t *this, unsigned char *child_up, int32_t read_child,
+                    int32_t *fresh_children,
+                    int32_t *call_child, int32_t *last_index)
+{
+        int             ret   = 0;
+        afr_private_t   *priv = NULL;
+        int             i     = 0;
+
+        GF_ASSERT (child_up);
+        GF_ASSERT (call_child);
+        GF_ASSERT (last_index);
+        GF_ASSERT (fresh_children);
+
+        if (read_child < 0) {
+                ret = -EIO;
+                goto out;
+        }
+        priv = this->private;
+        *call_child = -1;
+        *last_index = -1;
+
+        if (child_up[read_child]) {
+                *call_child = read_child;
+        } else {
+                for (i = 0; i < priv->child_count; i++) {
+                        if (fresh_children[i] == -1)
+                                break;
+                        if (child_up[fresh_children[i]]) {
+                                *call_child = fresh_children[i];
+                                ret = 0;
+                                break;
+                        }
                 }
 
-                ctx = (~AFR_ICTX_READ_CHILD_MASK & ctx)
-                        | (AFR_ICTX_READ_CHILD_MASK & read_child);
+                if (*call_child == -1) {
+                        ret = -ENOTCONN;
+                        goto out;
+                }
 
-                __inode_ctx_put (inode, this, ctx);
+                *last_index = i;
         }
-        UNLOCK (&inode->lock);
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d, call_child: %d, "
+                "last_index: %d", ret, *call_child, *last_index);
+        return ret;
+}
 
+void
+afr_reset_xattr (dict_t **xattr, unsigned int child_count)
+{
+        unsigned int i = 0;
+
+        if (!xattr)
+                goto out;
+        for (i = 0; i < child_count; i++) {
+                if (xattr[i]) {
+                        dict_unref (xattr[i]);
+                        xattr[i] = NULL;
+                }
+        }
 out:
         return;
 }
 
-
-/**
- * afr_local_cleanup - cleanup everything in frame->local
- */
+void
+afr_xattr_array_destroy (dict_t **xattr, unsigned int child_count)
+{
+        afr_reset_xattr (xattr, child_count);
+        GF_FREE (xattr);
+}
 
 void
 afr_local_sh_cleanup (afr_local_t *local, xlator_t *this)
 {
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
 
+        sh = &local->self_heal;
+        priv = this->private;
 
-	sh = &local->self_heal;
-	priv = this->private;
+        if (sh->data_sh_info && strcmp (sh->data_sh_info, ""))
+                GF_FREE (sh->data_sh_info);
 
-	if (sh->buf)
-		GF_FREE (sh->buf);
+        if (sh->metadata_sh_info && strcmp (sh->metadata_sh_info, ""))
+                GF_FREE (sh->metadata_sh_info);
 
-	if (sh->xattr) {
-		for (i = 0; i < priv->child_count; i++) {
-			if (sh->xattr[i]) {
-				dict_unref (sh->xattr[i]);
-				sh->xattr[i] = NULL;
-			}
-		}
-		GF_FREE (sh->xattr);
-	}
+        GF_FREE (sh->buf);
 
-	if (sh->child_errno)
-		GF_FREE (sh->child_errno);
+        GF_FREE (sh->parentbufs);
 
-	if (sh->pending_matrix) {
-		for (i = 0; i < priv->child_count; i++) {
-			GF_FREE (sh->pending_matrix[i]);
-		}
-		GF_FREE (sh->pending_matrix);
-	}
+        if (sh->inode)
+                inode_unref (sh->inode);
 
-	if (sh->delta_matrix) {
-		for (i = 0; i < priv->child_count; i++) {
-			GF_FREE (sh->delta_matrix[i]);
-		}
-		GF_FREE (sh->delta_matrix);
-	}
+        afr_xattr_array_destroy (sh->xattr, priv->child_count);
 
-	if (sh->sources)
-		GF_FREE (sh->sources);
+        GF_FREE (sh->child_errno);
 
-	if (sh->success)
-		GF_FREE (sh->success);
+        afr_matrix_cleanup (sh->pending_matrix, priv->child_count);
+        afr_matrix_cleanup (sh->delta_matrix, priv->child_count);
 
-	if (sh->locked_nodes)
-		GF_FREE (sh->locked_nodes);
+        GF_FREE (sh->sources);
 
-	if (sh->healing_fd && !sh->healing_fd_opened) {
-		fd_unref (sh->healing_fd);
-		sh->healing_fd = NULL;
-	}
+        GF_FREE (sh->success);
 
-        if (sh->linkname)
-                GF_FREE ((char *)sh->linkname);
+        GF_FREE (sh->locked_nodes);
 
-	loc_wipe (&sh->parent_loc);
+        if (sh->healing_fd) {
+                fd_unref (sh->healing_fd);
+                sh->healing_fd = NULL;
+        }
+
+        GF_FREE ((char *)sh->linkname);
+
+        GF_FREE (sh->success_children);
+
+        GF_FREE (sh->fresh_children);
+
+        GF_FREE (sh->fresh_parent_dirs);
+
+        loc_wipe (&sh->parent_loc);
+        loc_wipe (&sh->lookup_loc);
+
+        GF_FREE (sh->checksum);
+
+        GF_FREE (sh->write_needed);
+        if (sh->healing_fd)
+                fd_unref (sh->healing_fd);
 }
 
 
 void
 afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
 {
-        int             i = 0;
-        afr_private_t * priv = NULL;
+        afr_private_t *priv    = NULL;
+        int           i        = 0;
 
         priv = this->private;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->pending && local->pending[i])
-                        GF_FREE (local->pending[i]);
-        }
-
-        GF_FREE (local->pending);
+        afr_matrix_cleanup (local->pending, priv->child_count);
+        afr_matrix_cleanup (local->transaction.txn_changelog,
+                            priv->child_count);
 
-        if (local->internal_lock.locked_nodes)
-                GF_FREE (local->internal_lock.locked_nodes);
+        GF_FREE (local->internal_lock.locked_nodes);
 
-        if (local->internal_lock.inode_locked_nodes)
-                GF_FREE (local->internal_lock.inode_locked_nodes);
+        for (i = 0; local->internal_lock.inodelk[i].domain; i++) {
+                GF_FREE (local->internal_lock.inodelk[i].locked_nodes);
+        }
 
-        if (local->internal_lock.entry_locked_nodes)
-                GF_FREE (local->internal_lock.entry_locked_nodes);
+        GF_FREE (local->internal_lock.lower_locked_nodes);
 
-        if (local->internal_lock.lower_locked_nodes)
-                GF_FREE (local->internal_lock.lower_locked_nodes);
+        afr_entry_lockee_cleanup (&local->internal_lock);
 
+        GF_FREE (local->transaction.pre_op);
+        GF_FREE (local->transaction.eager_lock);
 
-	GF_FREE (local->transaction.child_errno);
-	GF_FREE (local->child_errno);
+        GF_FREE (local->transaction.basename);
+        GF_FREE (local->transaction.new_basename);
 
-	GF_FREE (local->transaction.basename);
-	GF_FREE (local->transaction.new_basename);
+        loc_wipe (&local->transaction.parent_loc);
+        loc_wipe (&local->transaction.new_parent_loc);
 
-	loc_wipe (&local->transaction.parent_loc);
-	loc_wipe (&local->transaction.new_parent_loc);
+        GF_FREE (local->transaction.postop_piggybacked);
 }
 
 
 void
 afr_local_cleanup (afr_local_t *local, xlator_t *this)
 {
-        int i;
         afr_private_t * priv = NULL;
 
-	if (!local)
-		return;
+        if (!local)
+                return;
 
-	afr_local_sh_cleanup (local, this);
+        afr_local_sh_cleanup (local, this);
 
         afr_local_transaction_cleanup (local, this);
 
         priv = this->private;
 
-	loc_wipe (&local->loc);
-	loc_wipe (&local->newloc);
+        loc_wipe (&local->loc);
+        loc_wipe (&local->newloc);
+
+        if (local->fd)
+                fd_unref (local->fd);
+
+        if (local->xattr_req)
+                dict_unref (local->xattr_req);
 
-	if (local->fd)
-		fd_unref (local->fd);
+        if (local->dict)
+                dict_unref (local->dict);
 
-	if (local->xattr_req)
-		dict_unref (local->xattr_req);
+	GF_FREE(local->replies);
 
-	GF_FREE (local->child_up);
+        GF_FREE (local->child_up);
 
-	{ /* lookup */
+        GF_FREE (local->child_errno);
+
+        GF_FREE (local->fresh_children);
+
+        { /* lookup */
                 if (local->cont.lookup.xattrs) {
-                        for (i = 0; i < priv->child_count; i++) {
-                                if (local->cont.lookup.xattrs[i]) {
-                                        dict_unref (local->cont.lookup.xattrs[i]);
-                                        local->cont.lookup.xattrs[i] = NULL;
-                                }
-                        }
+                        afr_reset_xattr (local->cont.lookup.xattrs,
+                                         priv->child_count);
                         GF_FREE (local->cont.lookup.xattrs);
                         local->cont.lookup.xattrs = NULL;
                 }
 
-		if (local->cont.lookup.xattr) {
+                if (local->cont.lookup.xattr) {
                         dict_unref (local->cont.lookup.xattr);
                 }
 
                 if (local->cont.lookup.inode) {
                         inode_unref (local->cont.lookup.inode);
                 }
-	}
 
-	{ /* getxattr */
-		if (local->cont.getxattr.name)
-			GF_FREE (local->cont.getxattr.name);
-	}
+                GF_FREE (local->cont.lookup.postparents);
 
-	{ /* lk */
-		if (local->cont.lk.locked_nodes)
-			GF_FREE (local->cont.lk.locked_nodes);
-	}
+                GF_FREE (local->cont.lookup.bufs);
+
+                GF_FREE (local->cont.lookup.success_children);
 
-	{ /* create */
-		if (local->cont.create.fd)
-			fd_unref (local->cont.create.fd);
+                GF_FREE (local->cont.lookup.sources);
+                afr_matrix_cleanup (local->cont.lookup.pending_matrix,
+                                    priv->child_count);
+        }
+
+        { /* getxattr */
+                GF_FREE (local->cont.getxattr.name);
+        }
+
+        { /* lk */
+                GF_FREE (local->cont.lk.locked_nodes);
+        }
+
+        { /* create */
+                if (local->cont.create.fd)
+                        fd_unref (local->cont.create.fd);
                 if (local->cont.create.params)
                         dict_unref (local->cont.create.params);
-	}
+        }
 
-	{ /* mknod */
+        { /* mknod */
                 if (local->cont.mknod.params)
                         dict_unref (local->cont.mknod.params);
-	}
+        }
 
-	{ /* mkdir */
+        { /* mkdir */
                 if (local->cont.mkdir.params)
                         dict_unref (local->cont.mkdir.params);
-	}
+        }
 
-	{ /* symlink */
+        { /* symlink */
                 if (local->cont.symlink.params)
                         dict_unref (local->cont.symlink.params);
-	}
+        }
 
-	{ /* writev */
-		GF_FREE (local->cont.writev.vector);
-	}
+        { /* writev */
+                GF_FREE (local->cont.writev.vector);
+        }
 
-	{ /* setxattr */
-		if (local->cont.setxattr.dict)
-			dict_unref (local->cont.setxattr.dict);
-	}
+        { /* setxattr */
+                if (local->cont.setxattr.dict)
+                        dict_unref (local->cont.setxattr.dict);
+        }
 
-	{ /* removexattr */
-		GF_FREE (local->cont.removexattr.name);
-	}
+        { /* fsetxattr */
+                if (local->cont.fsetxattr.dict)
+                        dict_unref (local->cont.fsetxattr.dict);
+        }
 
-	{ /* symlink */
-		GF_FREE (local->cont.symlink.linkpath);
-	}
+        { /* removexattr */
+                GF_FREE (local->cont.removexattr.name);
+        }
+        { /* xattrop */
+                if (local->cont.xattrop.xattr)
+                        dict_unref (local->cont.xattrop.xattr);
+        }
+        { /* fxattrop */
+                if (local->cont.fxattrop.xattr)
+                        dict_unref (local->cont.fxattrop.xattr);
+        }
+        { /* symlink */
+                GF_FREE (local->cont.symlink.linkpath);
+        }
 
         { /* opendir */
-                if (local->cont.opendir.checksum)
-                        GF_FREE (local->cont.opendir.checksum);
+                GF_FREE (local->cont.opendir.checksum);
         }
+
+        { /* readdirp */
+                if (local->cont.readdir.dict)
+                        dict_unref (local->cont.readdir.dict);
+        }
+
+        if (local->xdata_req)
+                dict_unref (local->xdata_req);
+
+        if (local->xdata_rsp)
+                dict_unref (local->xdata_rsp);
 }
 
 
 int
 afr_frame_return (call_frame_t *frame)
 {
-	afr_local_t *local = NULL;
-	int          call_count = 0;
+        afr_local_t *local = NULL;
+        int          call_count = 0;
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		call_count = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
+        LOCK (&frame->lock);
+        {
+                call_count = --local->call_count;
+        }
+        UNLOCK (&frame->lock);
 
-	return call_count;
+        return call_count;
 }
 
+int
+afr_set_elem_count_get (unsigned char *elems, int child_count)
+{
+        int i   = 0;
+        int ret = 0;
+
+        for (i = 0; i < child_count; i++)
+                if (elems[i])
+                        ret++;
+        return ret;
+}
 
 /**
  * up_children_count - return the number of children that are up
  */
 
-int
-afr_up_children_count (int child_count, unsigned char *child_up)
+unsigned int
+afr_up_children_count (unsigned char *child_up, unsigned int child_count)
 {
-	int i   = 0;
-	int ret = 0;
-
-	for (i = 0; i < child_count; i++)
-		if (child_up[i])
-			ret++;
-	return ret;
+        return afr_set_elem_count_get (child_up, child_count);
 }
 
-
-ino64_t
-afr_itransform (ino64_t ino, int child_count, int child_index)
+unsigned int
+afr_locked_children_count (unsigned char *children, unsigned int child_count)
 {
-	ino64_t scaled_ino = -1;
-
-	if (ino == ((uint64_t) -1)) {
-		scaled_ino = ((uint64_t) -1);
-		goto out;
-	}
-
-	scaled_ino = (ino * child_count) + child_index;
-
-out:
-	return scaled_ino;
+        return afr_set_elem_count_get (children, child_count);
 }
 
+unsigned int
+afr_pre_op_done_children_count (unsigned char *pre_op,
+                                unsigned int child_count)
+{
+        return afr_set_elem_count_get (pre_op, child_count);
+}
 
-int
-afr_deitransform_orig (ino64_t ino, int child_count)
+gf_boolean_t
+afr_is_fresh_lookup (loc_t *loc, xlator_t *this)
 {
-	int index = -1;
+        uint64_t          ctx = 0;
+        int32_t           ret = 0;
 
-	index = ino % child_count;
+        GF_ASSERT (loc);
+        GF_ASSERT (this);
+        GF_ASSERT (loc->inode);
 
-	return index;
+        ret = inode_ctx_get (loc->inode, this, &ctx);
+        if (0 == ret)
+                return _gf_false;
+        return _gf_true;
 }
 
-
-int
-afr_deitransform (ino64_t ino, int child_count)
+void
+afr_update_loc_gfids (loc_t *loc, struct iatt *buf, struct iatt *postparent)
 {
-	return 0;
+        GF_ASSERT (loc);
+        GF_ASSERT (buf);
+
+        uuid_copy (loc->gfid, buf->ia_gfid);
+        if (postparent)
+                uuid_copy (loc->pargfid, postparent->ia_gfid);
 }
 
+/*
+ * Quota size xattrs are not maintained by afr. There is a
+ * possibility that they differ even when both the directory changelog xattrs
+ * suggest everything is fine. So if there is at least one 'source' check among
+ * the sources which has the maximum quota size. Otherwise check among all the
+ * available ones for maximum quota size. This way if there is a source and
+ * stale copies it always votes for the 'source'.
+ * */
 
-int
-afr_self_heal_lookup_unwind (call_frame_t *frame, xlator_t *this)
+static void
+afr_handle_quota_size (afr_local_t *local, xlator_t *this,
+                       dict_t *rsp_dict)
 {
-	afr_local_t *local = NULL;
+        int32_t       *sources       = NULL;
+        dict_t        *xattr         = NULL;
+        data_t        *max_data      = NULL;
+        int64_t       max_quota_size = -1;
+        data_t        *data          = NULL;
+        int64_t       *size          = NULL;
+        int64_t       quota_size     = -1;
+        afr_private_t *priv          = NULL;
+        int           i              = 0;
+        int           ret            = -1;
+        gf_boolean_t  source_present = _gf_false;
+
+        priv    = this->private;
+        sources = local->cont.lookup.sources;
+
+        if (rsp_dict == NULL) {
+                gf_log_callingfn (this->name, GF_LOG_ERROR, "%s: Invalid "
+                                  "response dictionary", local->loc.path);
+                return;
+        }
 
-	local = frame->local;
+        for (i = 0; i < priv->child_count; i++) {
+                if (sources[i]) {
+                        source_present = _gf_true;
+                        break;
+                }
+        }
 
-	if (local->govinda_gOvinda) {
-                afr_set_split_brain (this, local->cont.lookup.inode, _gf_true);
-	}
+        for (i = 0; i < priv->child_count; i++) {
+                /*
+                 * If there is at least one source lets check
+                 * for maximum quota sizes among sources, otherwise take the
+                 * maximum of the ones present to be on the safer side.
+                 */
+                if (source_present && !sources[i])
+                        continue;
+
+                xattr = local->cont.lookup.xattrs[i];
+                if (!xattr)
+                        continue;
+
+                data = dict_get (xattr, QUOTA_SIZE_KEY);
+                if (!data)
+                        continue;
+
+                size = (int64_t*)data->data;
+                quota_size = ntoh64(*size);
+                gf_log (this->name, GF_LOG_DEBUG, "%s: %d, size: %"PRId64,
+                        local->loc.path, i, quota_size);
+                if (quota_size > max_quota_size) {
+                        if (max_data)
+                                data_unref (max_data);
+
+                        max_quota_size = quota_size;
+                        max_data = data_ref (data);
+                }
+        }
 
-	AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-			  local->cont.lookup.inode,
-			  &local->cont.lookup.buf,
-			  local->cont.lookup.xattr,
-                          &local->cont.lookup.postparent);
+        if (max_data) {
+                ret = dict_set (rsp_dict, QUOTA_SIZE_KEY, max_data);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s: Failed to set "
+                                "quota size", local->loc.path);
+                }
 
-	return 0;
+                data_unref (max_data);
+        }
 }
 
-
-static void
-afr_lookup_collect_xattr (afr_local_t *local, xlator_t *this,
-                          int child_index, dict_t *xattr)
+int
+afr_lookup_build_response_params (afr_local_t *local, xlator_t *this)
 {
-	uint32_t        open_fd_count = 0;
-        uint32_t        inodelk_count = 0;
-        uint32_t        entrylk_count = 0;
+        struct iatt     *buf = NULL;
+        struct iatt     *postparent = NULL;
+        dict_t          **xattr = NULL;
+        int32_t         *success_children = NULL;
+        int32_t         *sources = NULL;
+        afr_private_t   *priv = NULL;
+        int32_t         read_child = -1;
+        int             ret = 0;
+        int             i = 0;
 
-        int ret = 0;
+        GF_ASSERT (local);
 
-        if (afr_sh_has_metadata_pending (xattr, child_index, this)) {
-                local->self_heal.need_metadata_self_heal = _gf_true;
-                gf_log(this->name, GF_LOG_DEBUG,
-                       "metadata self-heal is pending for %s.",
-                       local->loc.path);
-        }
+        buf = &local->cont.lookup.buf;
+        postparent = &local->cont.lookup.postparent;
+        xattr = &local->cont.lookup.xattr;
+        priv = this->private;
 
-        if (afr_sh_has_entry_pending (xattr, child_index, this)) {
-                local->self_heal.need_entry_self_heal = _gf_true;
-                gf_log(this->name, GF_LOG_DEBUG,
-                       "entry self-heal is pending for %s.", local->loc.path);
+        read_child = afr_inode_get_read_ctx (this, local->cont.lookup.inode,
+                                             local->fresh_children);
+        if (read_child < 0) {
+                ret = -1;
+                goto out;
+        }
+        success_children = local->cont.lookup.success_children;
+        sources = local->cont.lookup.sources;
+        memset (sources, 0, sizeof (*sources) * priv->child_count);
+        afr_children_intersection_get (local->fresh_children, success_children,
+                                       sources, priv->child_count);
+        if (!sources[read_child]) {
+                read_child = -1;
+                for (i = 0; i < priv->child_count; i++) {
+                        if (sources[i]) {
+                                read_child = i;
+                                break;
+                        }
+                }
+        }
+        if (read_child < 0) {
+                ret = -1;
+                goto out;
         }
 
-        if (afr_sh_has_data_pending (xattr, child_index, this)) {
-                local->self_heal.need_data_self_heal = _gf_true;
-                gf_log(this->name, GF_LOG_DEBUG,
-                       "data self-heal is pending for %s.", local->loc.path);
+        gf_log (this->name, GF_LOG_DEBUG, "Building lookup response from %d",
+                read_child);
+        if (!*xattr)
+                *xattr = dict_ref (local->cont.lookup.xattrs[read_child]);
+
+        *buf = local->cont.lookup.bufs[read_child];
+        *postparent = local->cont.lookup.postparents[read_child];
+
+        if (dict_get (local->xattr_req, QUOTA_SIZE_KEY))
+                afr_handle_quota_size (local, this, *xattr);
+
+        if (IA_INVAL == local->cont.lookup.inode->ia_type) {
+                /* fix for RT #602 */
+                local->cont.lookup.inode->ia_type = buf->ia_type;
         }
+out:
+        return ret;
+}
 
-        ret = dict_get_uint32 (xattr, GLUSTERFS_OPEN_FD_COUNT,
-                               &open_fd_count);
-        if (ret == 0)
-                local->open_fd_count += open_fd_count;
+static void
+afr_lookup_update_lk_counts (afr_local_t *local, xlator_t *this,
+                            int child_index, dict_t *xattr)
+{
+        uint32_t inodelk_count = 0;
+        uint32_t entrylk_count = 0;
+        int      ret           = -1;
+        uint32_t parent_entrylk = 0;
+
+        GF_ASSERT (local);
+        GF_ASSERT (this);
+        GF_ASSERT (xattr);
+        GF_ASSERT (child_index >= 0);
 
         ret = dict_get_uint32 (xattr, GLUSTERFS_INODELK_COUNT,
                                &inodelk_count);
@@ -582,924 +1249,1505 @@ afr_lookup_collect_xattr (afr_local_t *local, xlator_t *this,
                                &entrylk_count);
         if (ret == 0)
                 local->entrylk_count += entrylk_count;
+        ret = dict_get_uint32 (xattr, GLUSTERFS_PARENT_ENTRYLK,
+                               &parent_entrylk);
+        if (!ret)
+                local->cont.lookup.parent_entrylk += parent_entrylk;
 }
 
-
+/*
+ * It's important to maintain a commutative property on do_*_self_heal and
+ * found*; once set, they must not be cleared by a subsequent iteration or
+ * call, so that they represent a logical OR of all iterations and calls
+ * regardless of child/key order.  That allows the caller to call us multiple
+ * times without having to use a separate variable as a "reduce" accumulator.
+ */
 static void
-afr_lookup_self_heal_check (xlator_t *this, afr_local_t *local,
-                            struct iatt *buf, struct iatt *lookup_buf)
+afr_lookup_set_self_heal_params_by_xattr (afr_local_t *local, xlator_t *this,
+                                          dict_t *xattr)
 {
-        if (FILETYPE_DIFFERS (buf, lookup_buf)) {
-                /* mismatching filetypes with same name
-                */
+        afr_private_t *priv        = NULL;
+        int            i           = 0;
+        int            ret         = -1;
+        void          *pending_raw = NULL;
+        int32_t       *pending     = NULL;
+
+        GF_ASSERT (local);
+        GF_ASSERT (this);
+        GF_ASSERT (xattr);
+
+        priv = this->private;
+
+        for (i = 0; i < priv->child_count; i++) {
+                ret = dict_get_ptr (xattr, priv->pending_key[i],
+                                    &pending_raw);
+                if (ret != 0) {
+                        continue;
+                }
+                pending = pending_raw;
+
+                if (pending[AFR_METADATA_TRANSACTION]) {
+                        gf_log(this->name, GF_LOG_DEBUG,
+                               "metadata self-heal is pending for %s.",
+                               local->loc.path);
+                        local->self_heal.do_metadata_self_heal = _gf_true;
+                }
 
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "filetype differs for %s ", local->loc.path);
+                if (pending[AFR_ENTRY_TRANSACTION]) {
+                        gf_log(this->name, GF_LOG_DEBUG,
+                               "entry self-heal is pending for %s.",
+                               local->loc.path);
+                        local->self_heal.do_entry_self_heal = _gf_true;
+                }
 
-                local->govinda_gOvinda = 1;
+                if (pending[AFR_DATA_TRANSACTION]) {
+                        gf_log(this->name, GF_LOG_DEBUG,
+                               "data self-heal is pending for %s.",
+                               local->loc.path);
+                        local->self_heal.do_data_self_heal = _gf_true;
+                }
         }
+}
+
+void
+afr_lookup_check_set_metadata_split_brain (afr_local_t *local, xlator_t *this)
+{
+        int32_t                  *sources = NULL;
+        afr_private_t            *priv = NULL;
+        int32_t                  subvol_status = 0;
+        int32_t                  *success_children   = NULL;
+        dict_t                   **xattrs = NULL;
+        struct iatt              *bufs = NULL;
+        int32_t                  **pending_matrix = NULL;
+
+        priv = this->private;
+
+        sources = GF_CALLOC (priv->child_count, sizeof (*sources),
+                             gf_afr_mt_int32_t);
+        if (NULL == sources)
+                goto out;
+        success_children = local->cont.lookup.success_children;
+        xattrs = local->cont.lookup.xattrs;
+        bufs = local->cont.lookup.bufs;
+        pending_matrix = local->cont.lookup.pending_matrix;
+        afr_build_sources (this, xattrs, bufs, pending_matrix,
+                           sources, success_children, AFR_METADATA_TRANSACTION,
+                           &subvol_status, _gf_false);
+        if (subvol_status & SPLIT_BRAIN)
+                local->cont.lookup.possible_spb = _gf_true;
+out:
+        GF_FREE (sources);
+}
 
+static void
+afr_detect_self_heal_by_iatt (afr_local_t *local, xlator_t *this,
+                            struct iatt *buf, struct iatt *lookup_buf)
+{
         if (PERMISSION_DIFFERS (buf, lookup_buf)) {
                 /* mismatching permissions */
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_DEBUG,
                         "permissions differ for %s ", local->loc.path);
-                local->self_heal.need_metadata_self_heal = _gf_true;
+                local->self_heal.do_metadata_self_heal = _gf_true;
         }
 
         if (OWNERSHIP_DIFFERS (buf, lookup_buf)) {
                 /* mismatching permissions */
-                local->self_heal.need_metadata_self_heal = _gf_true;
-                gf_log (this->name, GF_LOG_NORMAL,
+                local->self_heal.do_metadata_self_heal = _gf_true;
+                gf_log (this->name, GF_LOG_DEBUG,
                         "ownership differs for %s ", local->loc.path);
         }
 
         if (SIZE_DIFFERS (buf, lookup_buf)
             && IA_ISREG (buf->ia_type)) {
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_DEBUG,
                         "size differs for %s ", local->loc.path);
-                local->self_heal.need_data_self_heal = _gf_true;
+                local->self_heal.do_data_self_heal = _gf_true;
         }
 
+        if (uuid_compare (buf->ia_gfid, lookup_buf->ia_gfid)) {
+                /* mismatching gfid */
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "%s: gfid different on subvolume", local->loc.path);
+        }
 }
 
+static void
+afr_detect_self_heal_by_split_brain_status (afr_local_t *local, xlator_t *this)
+{
+        gf_boolean_t split_brain = _gf_false;
+        afr_self_heal_t *sh = NULL;
+
+        sh = &local->self_heal;
+
+        split_brain = afr_is_split_brain (this, local->cont.lookup.inode);
+        split_brain = split_brain || local->cont.lookup.possible_spb;
+        if ((local->success_count > 0) && split_brain &&
+            IA_ISREG (local->cont.lookup.inode->ia_type)) {
+                sh->force_confirm_spb = _gf_true;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "split brain detected during lookup of %s.",
+                        local->loc.path);
+        }
+}
 
 static void
-afr_lookup_done (call_frame_t *frame, xlator_t *this, struct iatt *lookup_buf)
+afr_detect_self_heal_by_lookup_status (afr_local_t *local, xlator_t *this)
 {
-        int  unwind = 1;
-        int  source = -1;
-        int  up_count = 0;
-        char sh_type_str[256] = {0,};
+        GF_ASSERT (local);
+        GF_ASSERT (this);
+
+        if ((local->success_count > 0) && (local->enoent_count > 0)) {
+                local->self_heal.do_metadata_self_heal = _gf_true;
+                local->self_heal.do_data_self_heal     = _gf_true;
+                local->self_heal.do_entry_self_heal    = _gf_true;
+                local->self_heal.do_gfid_self_heal    = _gf_true;
+                local->self_heal.do_missing_entry_self_heal    = _gf_true;
+                gf_log(this->name, GF_LOG_DEBUG,
+                       "entries are missing in lookup of %s.",
+                       local->loc.path);
+        }
 
-        afr_private_t *priv  = NULL;
-        afr_local_t   *local = NULL;
+        return;
+}
 
-        priv  = this->private;
-        local = frame->local;
+gf_boolean_t
+afr_can_self_heal_proceed (afr_self_heal_t *sh, afr_private_t *priv)
+{
+        GF_ASSERT (sh);
+        GF_ASSERT (priv);
+
+        if (sh->force_confirm_spb)
+                return _gf_true;
+        return (sh->do_gfid_self_heal
+                || sh->do_missing_entry_self_heal
+                || (afr_data_self_heal_enabled (priv->data_self_heal) &&
+                    sh->do_data_self_heal)
+                || (priv->metadata_self_heal && sh->do_metadata_self_heal)
+                || (priv->entry_self_heal && sh->do_entry_self_heal));
+}
+
+afr_transaction_type
+afr_transaction_type_get (ia_type_t ia_type)
+{
+        afr_transaction_type    type = AFR_METADATA_TRANSACTION;
 
-        local->cont.lookup.postparent.ia_ino  = local->cont.lookup.parent_ino;
+        GF_ASSERT (ia_type != IA_INVAL);
 
-        if (local->cont.lookup.ino) {
-                local->cont.lookup.buf.ia_ino = local->cont.lookup.ino;
+        if (IA_ISDIR (ia_type)) {
+                type = AFR_ENTRY_TRANSACTION;
+        } else if (IA_ISREG (ia_type)) {
+                type = AFR_DATA_TRANSACTION;
         }
+        return type;
+}
 
-        up_count = afr_up_children_count (priv->child_count, priv->child_up);
-        if (up_count == 1) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Only 1 child up - do not attempt to detect self heal");
-                goto unwind;
+int
+afr_lookup_select_read_child (afr_local_t *local, xlator_t *this,
+                              int32_t *read_child)
+{
+        ia_type_t               ia_type        = IA_INVAL;
+        int32_t                 source         = -1;
+        int                     ret            = -1;
+        dict_t                  **xattrs       = NULL;
+        int32_t                 *success_children = NULL;
+        afr_transaction_type    type           = AFR_METADATA_TRANSACTION;
+        uuid_t                  *gfid          = NULL;
+
+        GF_ASSERT (local);
+        GF_ASSERT (this);
+        GF_ASSERT (local->success_count > 0);
+
+        success_children = local->cont.lookup.success_children;
+        /*We can take the success_children[0] only because we already
+         *handle the conflicting children other wise, we could select the
+         *read_child based on wrong file type
+         */
+        ia_type = local->cont.lookup.bufs[success_children[0]].ia_type;
+        type = afr_transaction_type_get (ia_type);
+        xattrs = local->cont.lookup.xattrs;
+        gfid = &local->cont.lookup.buf.ia_gfid;
+        source = afr_lookup_select_read_child_by_txn_type (this, local, xattrs,
+                                                           type, *gfid);
+        if (source < 0) {
+                gf_log (this->name, GF_LOG_DEBUG, "failed to select source "
+                        "for %s", local->loc.path);
+                goto out;
         }
 
-        if (local->op_ret == 0) {
-                /* KLUDGE: assuming DHT will not itransform in
-                   revalidate */
-                if (local->cont.lookup.inode->ino) {
-                        local->cont.lookup.buf.ia_ino =
-                                local->cont.lookup.inode->ino;
+        gf_log (this->name, GF_LOG_DEBUG, "Source selected as %d for %s",
+                source, local->loc.path);
+        *read_child = source;
+        ret = 0;
+out:
+        return ret;
+}
+
+static inline gf_boolean_t
+afr_is_transaction_running (afr_local_t *local)
+{
+        GF_ASSERT (local->fop == GF_FOP_LOOKUP);
+        return ((local->inodelk_count > 0) || (local->entrylk_count > 0));
+}
+
+void
+afr_launch_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode,
+                      gf_boolean_t background, ia_type_t ia_type, char *reason,
+                      void (*gfid_sh_success_cbk) (call_frame_t *sh_frame,
+                                                   xlator_t *this),
+                      int (*unwind) (call_frame_t *frame, xlator_t *this,
+                                     int32_t op_ret, int32_t op_errno,
+                                     int32_t sh_failed))
+{
+        afr_local_t             *local = NULL;
+        char                    sh_type_str[256] = {0,};
+        char                    *bg = "";
+
+        GF_ASSERT (frame);
+        GF_ASSERT (this);
+        GF_ASSERT (inode);
+        GF_ASSERT (ia_type != IA_INVAL);
+
+        local = frame->local;
+        local->self_heal.background = background;
+        local->self_heal.type       = ia_type;
+        local->self_heal.unwind     = unwind;
+        local->self_heal.gfid_sh_success_cbk     = gfid_sh_success_cbk;
+
+        afr_self_heal_type_str_get (&local->self_heal,
+                                    sh_type_str,
+                                    sizeof (sh_type_str));
+
+        if (background)
+                bg = "background";
+        gf_log (this->name, GF_LOG_DEBUG,
+                "%s %s self-heal triggered. path: %s, reason: %s", bg,
+                sh_type_str, local->loc.path, reason);
+
+        afr_self_heal (frame, this, inode);
+}
+
+unsigned int
+afr_gfid_missing_count (const char *xlator_name, int32_t *success_children,
+                        struct iatt *bufs, unsigned int child_count,
+                        const char *path)
+{
+        unsigned int    gfid_miss_count   = 0;
+        int             i              = 0;
+        struct iatt     *child1        = NULL;
+
+        for (i = 0; i < child_count; i++) {
+                if (success_children[i] == -1)
+                        break;
+                child1 = &bufs[success_children[i]];
+                if (uuid_is_null (child1->ia_gfid)) {
+                        gf_log (xlator_name, GF_LOG_DEBUG, "%s: gfid is null"
+                                " on subvolume %d", path, success_children[i]);
+                        gfid_miss_count++;
                 }
         }
 
-        if (local->success_count && local->enoent_count) {
-                local->self_heal.need_metadata_self_heal = _gf_true;
-                local->self_heal.need_data_self_heal     = _gf_true;
-                local->self_heal.need_entry_self_heal    = _gf_true;
-                gf_log(this->name, GF_LOG_NORMAL,
-                       "entries are missing in lookup of %s.",
-                       local->loc.path);
+        return gfid_miss_count;
+}
+
+static int
+afr_lookup_gfid_missing_count (afr_local_t *local, xlator_t *this)
+{
+        int32_t         *success_children = NULL;
+        afr_private_t   *priv          = NULL;
+        struct iatt     *bufs          = NULL;
+        int             miss_count     = 0;
+
+        priv = this->private;
+        bufs = local->cont.lookup.bufs;
+        success_children = local->cont.lookup.success_children;
+
+        miss_count =  afr_gfid_missing_count (this->name, success_children,
+                                              bufs, priv->child_count,
+                                              local->loc.path);
+        return miss_count;
+}
+
+gf_boolean_t
+afr_conflicting_iattrs (struct iatt *bufs, int32_t *success_children,
+                        unsigned int child_count, const char *path,
+                        const char *xlator_name)
+{
+        gf_boolean_t    conflicting    = _gf_false;
+        int             i              = 0;
+        struct iatt     *child1        = NULL;
+        struct iatt     *child2        = NULL;
+        uuid_t          *gfid          = NULL;
+
+        for (i = 0; i < child_count; i++) {
+                if (success_children[i] == -1)
+                        break;
+                child1 = &bufs[success_children[i]];
+                if ((!gfid) && (!uuid_is_null (child1->ia_gfid)))
+                        gfid = &child1->ia_gfid;
+
+                if (i == 0)
+                        continue;
+
+                child2 = &bufs[success_children[i-1]];
+                if (FILETYPE_DIFFERS (child1, child2)) {
+                        gf_log (xlator_name, GF_LOG_DEBUG, "%s: filetype "
+                                "differs on subvolumes (%d, %d)", path,
+                                success_children[i-1], success_children[i]);
+                        conflicting = _gf_true;
+                        goto out;
+                }
+                if (!gfid || uuid_is_null (child1->ia_gfid))
+                        continue;
+                if (uuid_compare (*gfid, child1->ia_gfid)) {
+                       gf_log (xlator_name, GF_LOG_DEBUG, "%s: gfid differs"
+                               " on subvolume %d", path, success_children[i]);
+                       conflicting = _gf_true;
+                       goto out;
+                }
         }
+out:
+        return conflicting;
+}
 
-        if (local->success_count) {
-                /* check for split-brain case in previous lookup */
-                if (afr_is_split_brain (this,
-                                        local->cont.lookup.inode)) {
-                        local->self_heal.need_data_self_heal = _gf_true;
-                        gf_log(this->name, GF_LOG_NORMAL,
-                               "split brain detected during lookup of "
-                               "%s.", local->loc.path);
+/* afr_update_gfid_from_iatts: This function should be called only if the
+ * iatts are not conflicting.
+ */
+void
+afr_update_gfid_from_iatts (uuid_t uuid, struct iatt *bufs,
+                            int32_t *success_children, unsigned int child_count)
+{
+        uuid_t          *gfid = NULL;
+        int             i = 0;
+        int             child = 0;
+
+        for (i = 0; i < child_count; i++) {
+                child = success_children[i];
+                if (child == -1)
+                        break;
+                if ((!gfid) && (!uuid_is_null (bufs[child].ia_gfid))) {
+                        gfid = &bufs[child].ia_gfid;
+                } else if (gfid && (!uuid_is_null (bufs[child].ia_gfid))) {
+                        if (uuid_compare (*gfid, bufs[child].ia_gfid)) {
+                                GF_ASSERT (0);
+                                goto out;
+                        }
                 }
         }
+        if (gfid && (!uuid_is_null (*gfid)))
+                uuid_copy (uuid, *gfid);
+out:
+        return;
+}
 
-        if ((local->self_heal.need_metadata_self_heal
-             || local->self_heal.need_data_self_heal
-             || local->self_heal.need_entry_self_heal)
-            && ((!local->cont.lookup.is_revalidate)
-                || (local->op_ret != -1))) {
+static gf_boolean_t
+afr_lookup_conflicting_entries (afr_local_t *local, xlator_t *this)
+{
+        afr_private_t           *priv = NULL;
+        gf_boolean_t            conflict = _gf_false;
 
-                if (local->open_fd_count
-                    || local->inodelk_count
-                    || local->entrylk_count) {
+        priv = this->private;
+        conflict =  afr_conflicting_iattrs (local->cont.lookup.bufs,
+                                            local->cont.lookup.success_children,
+                                            priv->child_count, local->loc.path,
+                                            this->name);
+        return conflict;
+}
 
-                        /* Someone else is doing self-heal on this file.
-                           So just make a best effort to set the read-subvolume
-                           and return */
+gf_boolean_t
+afr_open_only_data_self_heal (char *data_self_heal)
+{
+        return !strcmp (data_self_heal, "open");
+}
 
-                        if (IA_ISREG (local->cont.lookup.inode->ia_type)) {
-                                source = afr_self_heal_get_source (this, local, local->cont.lookup.xattrs);
+gf_boolean_t
+afr_data_self_heal_enabled (char *data_self_heal)
+{
+        gf_boolean_t    enabled = _gf_false;
 
-                                if (source >= 0) {
-                                        afr_set_read_child (this,
-                                                            local->cont.lookup.inode,
-                                                            source);
-                                }
-                        }
-                } else {
-                        if (!local->cont.lookup.inode->ia_type) {
-                                /* fix for RT #602 */
-                                local->cont.lookup.inode->ia_type =
-                                        lookup_buf->ia_type;
-                        }
+        if (gf_string2boolean (data_self_heal, &enabled) == -1) {
+                enabled = !strcmp (data_self_heal, "open");
+                GF_ASSERT (enabled);
+        }
 
-                        local->self_heal.background = _gf_true;
-                        local->self_heal.type       = local->cont.lookup.buf.ia_type;
-                        local->self_heal.unwind     = afr_self_heal_lookup_unwind;
+        return enabled;
+}
 
-                        unwind = 0;
+static void
+afr_lookup_set_self_heal_params (afr_local_t *local, xlator_t *this)
+{
+        int                     i = 0;
+        struct iatt             *bufs = NULL;
+        dict_t                  **xattr = NULL;
+        afr_private_t           *priv = NULL;
+        int32_t                 child1 = -1;
+        int32_t                 child2 = -1;
+        afr_self_heal_t         *sh = NULL;
 
-                        afr_self_heal_type_str_get(&local->self_heal,
-                                                   sh_type_str,
-                                                   sizeof(sh_type_str));
+        priv  = this->private;
+        sh = &local->self_heal;
+
+        afr_detect_self_heal_by_lookup_status (local, this);
+
+        if (afr_lookup_gfid_missing_count (local, this))
+                local->self_heal.do_gfid_self_heal    = _gf_true;
+
+        if (_gf_true == afr_lookup_conflicting_entries (local, this))
+                local->self_heal.do_missing_entry_self_heal    = _gf_true;
+        else
+                afr_update_gfid_from_iatts (local->self_heal.sh_gfid_req,
+                                            local->cont.lookup.bufs,
+                                            local->cont.lookup.success_children,
+                                            priv->child_count);
+
+        bufs = local->cont.lookup.bufs;
+        for (i = 1; i < local->success_count; i++) {
+                child1 = local->cont.lookup.success_children[i-1];
+                child2 = local->cont.lookup.success_children[i];
+                afr_detect_self_heal_by_iatt (local, this,
+                                              &bufs[child1], &bufs[child2]);
+        }
 
-                        gf_log (this->name, GF_LOG_NORMAL, "background %s "
-                                "self-heal triggered. path: %s",
-                                sh_type_str, local->loc.path);
+        xattr = local->cont.lookup.xattrs;
+        for (i = 0; i < local->success_count; i++) {
+                child1 = local->cont.lookup.success_children[i];
+                afr_lookup_set_self_heal_params_by_xattr (local, this,
+                                                          xattr[child1]);
+        }
+        if (afr_open_only_data_self_heal (priv->data_self_heal))
+                sh->do_data_self_heal = _gf_false;
+        if (sh->do_metadata_self_heal)
+                afr_lookup_check_set_metadata_split_brain (local, this);
+        afr_detect_self_heal_by_split_brain_status (local, this);
+}
 
-                        afr_self_heal (frame, this);
-                }
+int
+afr_self_heal_lookup_unwind (call_frame_t *frame, xlator_t *this,
+                             int32_t op_ret, int32_t op_errno,
+                             int32_t sh_failed)
+{
+        afr_local_t *local = NULL;
+        int         ret    = -1;
+        dict_t      *xattr = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                local->op_ret = -1;
+		local->op_errno = afr_most_important_error(local->op_errno,
+							   op_errno, _gf_true);
+
+                goto out;
+        } else {
+                local->op_ret = 0;
         }
 
-unwind:
-        if (unwind) {
-                AFR_STACK_UNWIND (lookup, frame, local->op_ret,
-                                  local->op_errno,
-                                  local->cont.lookup.inode,
-                                  &local->cont.lookup.buf,
-                                  local->cont.lookup.xattr,
-                                  &local->cont.lookup.postparent);
+        afr_lookup_done_success_action (frame, this, _gf_true);
+        xattr = local->cont.lookup.xattr;
+        if (xattr) {
+                ret = dict_set_int32 (xattr, "sh-failed", sh_failed);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR, "%s: Failed to set "
+                                "sh-failed to %d", local->loc.path, sh_failed);
+
+                if (local->self_heal.actual_sh_started == _gf_true &&
+                    sh_failed == 0) {
+                        ret = dict_set_int32 (xattr, "actual-sh-done", 1);
+                        if (ret)
+                                gf_log(this->name, GF_LOG_ERROR, "%s: Failed to"
+                                       " set actual-sh-done to %d",
+                                       local->loc.path,
+                                       local->self_heal.actual_sh_started);
+                }
         }
+out:
+        AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+                          local->cont.lookup.inode, &local->cont.lookup.buf,
+                          local->cont.lookup.xattr,
+                          &local->cont.lookup.postparent);
+
+        return 0;
 }
 
+//TODO: At the moment only lookup needs this, so not doing any checks, in the
+// future we will have to do fop specific operations
+void
+afr_post_gfid_sh_success (call_frame_t *sh_frame, xlator_t *this)
+{
+        afr_local_t             *local = NULL;
+        afr_local_t             *sh_local = NULL;
+        afr_private_t           *priv = NULL;
+        afr_self_heal_t         *sh = NULL;
+        int                     i = 0;
+        struct iatt             *lookup_bufs = NULL;
+        struct iatt             *lookup_parentbufs = NULL;
+
+        sh_local = sh_frame->local;
+        sh       = &sh_local->self_heal;
+        local = sh->orig_frame->local;
+        lookup_bufs = local->cont.lookup.bufs;
+        lookup_parentbufs = local->cont.lookup.postparents;
+        priv = this->private;
 
-/*
- * During a lookup, some errors are more "important" than
- * others in that they must be given higher priority while
- * returning to the user.
- *
- * The hierarchy is ESTALE > ENOENT > others
- *
- */
+        memcpy (lookup_bufs, sh->buf, priv->child_count * sizeof (*sh->buf));
+        memcpy (lookup_parentbufs, sh->parentbufs,
+                priv->child_count * sizeof (*sh->parentbufs));
 
-static gf_boolean_t
-__error_more_important (int32_t old_errno, int32_t new_errno)
+        afr_reset_xattr (local->cont.lookup.xattrs, priv->child_count);
+        if (local->cont.lookup.xattr) {
+                dict_unref (local->cont.lookup.xattr);
+                local->cont.lookup.xattr = NULL;
+        }
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (sh->xattr[i])
+                        local->cont.lookup.xattrs[i] = dict_ref (sh->xattr[i]);
+        }
+
+        afr_reset_children (local->cont.lookup.success_children,
+                            priv->child_count);
+        afr_children_copy (local->cont.lookup.success_children,
+                           sh->fresh_children, priv->child_count);
+}
+
+static void
+afr_lookup_perform_self_heal (call_frame_t *frame, xlator_t *this,
+                              gf_boolean_t *sh_launched)
 {
-        gf_boolean_t ret = _gf_true;
+        unsigned int         up_count = 0;
+        afr_private_t       *priv    = NULL;
+        afr_local_t         *local   = NULL;
+        char                *reason  = NULL;
 
-        /* Nothing should ever overwrite ESTALE */
-        if (old_errno == ESTALE)
-                ret = _gf_false;
+        GF_ASSERT (sh_launched);
+        *sh_launched = _gf_false;
+        priv         = this->private;
+        local        = frame->local;
 
-        /* Nothing should overwrite ENOENT, except ESTALE */
-        else if ((old_errno == ENOENT) && (new_errno != ESTALE))
-                ret = _gf_false;
+        up_count  = afr_up_children_count (local->child_up, priv->child_count);
+        if (up_count == 1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Only 1 child up - do not attempt to detect self heal");
+                goto out;
+        }
 
-        return ret;
+        afr_lookup_set_self_heal_params (local, this);
+        if (afr_can_self_heal_proceed (&local->self_heal, priv)) {
+                if  (afr_is_transaction_running (local) &&
+                     (!local->allow_sh_for_running_transaction))
+                        goto out;
+
+                reason = "lookup detected pending operations";
+                afr_launch_self_heal (frame, this, local->cont.lookup.inode,
+                                      _gf_true, local->cont.lookup.buf.ia_type,
+                                      reason, afr_post_gfid_sh_success,
+                                      afr_self_heal_lookup_unwind);
+                *sh_launched = _gf_true;
+        }
+out:
+        return;
 }
 
+void
+afr_get_fresh_children (int32_t *success_children, int32_t *sources,
+                        int32_t *fresh_children, unsigned int child_count)
+{
+        unsigned int i = 0;
+        unsigned int j = 0;
+
+        GF_ASSERT (success_children);
+        GF_ASSERT (sources);
+        GF_ASSERT (fresh_children);
+
+        afr_reset_children (fresh_children, child_count);
+        for (i = 0; i < child_count; i++) {
+                if (success_children[i] == -1)
+                        break;
+                if (afr_is_read_child (success_children, sources, child_count,
+                                       success_children[i])) {
+                        fresh_children[j] = success_children[i];
+                        j++;
+                }
+        }
+}
 
-int
-afr_fresh_lookup_cbk (call_frame_t *frame, void *cookie,
-                      xlator_t *this,  int32_t op_ret,	int32_t op_errno,
-                      inode_t *inode,	struct iatt *buf, dict_t *xattr,
-                      struct iatt *postparent)
+static int
+afr_lookup_set_read_ctx (afr_local_t *local, xlator_t *this, int32_t read_child)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	struct iatt *   lookup_buf = NULL;
+        afr_private_t           *priv = NULL;
 
-	int             call_count      = -1;
-	int             child_index     = -1;
-        int             first_up_child  = -1;
+        GF_ASSERT (read_child >= 0);
 
-	child_index = (long) cookie;
-	priv = this->private;
+        priv = this->private;
+        afr_get_fresh_children (local->cont.lookup.success_children,
+                                local->cont.lookup.sources,
+                                local->fresh_children, priv->child_count);
+        afr_inode_set_read_ctx (this, local->cont.lookup.inode, read_child,
+                                local->fresh_children);
 
-	LOCK (&frame->lock);
-	{
-		local = frame->local;
+        return 0;
+}
 
-                lookup_buf = &local->cont.lookup.buf;
+int
+afr_lookup_done_success_action (call_frame_t *frame, xlator_t *this,
+                                gf_boolean_t fail_conflict)
+{
+        int32_t             read_child = -1;
+        int32_t             ret        = -1;
+        afr_local_t         *local     = NULL;
+        gf_boolean_t        fresh_lookup = _gf_false;
 
-		if (op_ret == -1) {
-			if (op_errno == ENOENT)
-				local->enoent_count++;
+        local   = frame->local;
+        fresh_lookup = local->cont.lookup.fresh_lookup;
 
-                        if (__error_more_important (local->op_errno, op_errno))
-                                local->op_errno = op_errno;
+        if (local->loc.parent == NULL)
+                fail_conflict = _gf_true;
 
-                        if (local->op_errno == ESTALE) {
-                                local->op_ret = -1;
-                        }
+        if (afr_lookup_conflicting_entries (local, this)) {
+                if (fail_conflict == _gf_false)
+                        ret = 0;
+                goto out;
+        }
 
-                        goto unlock;
-		}
+        ret = afr_lookup_select_read_child (local, this, &read_child);
+        if (!afr_is_transaction_running (local) || fresh_lookup) {
+                if (read_child < 0)
+                        goto out;
 
-                afr_lookup_collect_xattr (local, this, child_index, xattr);
+                ret = afr_lookup_set_read_ctx (local, this, read_child);
+                if (ret)
+                        goto out;
+        }
 
-                first_up_child = afr_first_up_child (priv);
+        ret = afr_lookup_build_response_params (local, this);
+        if (ret)
+                goto out;
+        afr_update_loc_gfids (&local->loc,
+                              &local->cont.lookup.buf,
+                              &local->cont.lookup.postparent);
+
+        ret = 0;
+out:
+        if (ret) {
+                local->op_ret = -1;
+                local->op_errno = EIO;
+        }
+        return ret;
+}
+
+int
+afr_lookup_get_latest_subvol (afr_local_t *local, xlator_t *this)
+{
+        afr_private_t *priv = NULL;
+        int32_t       *success_children = NULL;
+        struct iatt   *bufs = NULL;
+        int           i = 0;
+        int           child = 0;
+        int           lsubvol = -1;
 
-                if (child_index == first_up_child) {
-                        local->cont.lookup.ino =
-                                afr_itransform (buf->ia_ino,
-                                                priv->child_count,
-                                                first_up_child);
+        priv = this->private;
+        success_children = local->cont.lookup.success_children;
+        bufs = local->cont.lookup.bufs;
+        for (i = 0; i < priv->child_count; i++) {
+                child = success_children[i];
+                if (child == -1)
+                        break;
+                if (uuid_is_null (bufs[child].ia_gfid))
+                        continue;
+                if (lsubvol < 0) {
+                        lsubvol = child;
+                } else if (bufs[lsubvol].ia_ctime < bufs[child].ia_ctime) {
+                        lsubvol = child;
+                } else if ((bufs[lsubvol].ia_ctime == bufs[child].ia_ctime) &&
+                  (bufs[lsubvol].ia_ctime_nsec < bufs[child].ia_ctime_nsec)) {
+                        lsubvol = child;
                 }
+        }
+        return lsubvol;
+}
 
-		if (local->success_count == 0) {
-                        if (local->op_errno != ESTALE)
-                                local->op_ret = op_ret;
+void
+afr_lookup_mark_other_entries_stale (afr_local_t *local, xlator_t *this,
+                                     int subvol)
+{
+        afr_private_t *priv = NULL;
+        int32_t       *success_children = NULL;
+        struct iatt   *bufs = NULL;
+        int           i = 0;
+        int           child = 0;
 
-			local->cont.lookup.inode               = inode_ref (inode);
-			local->cont.lookup.xattr               = dict_ref (xattr);
-			local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
-                        local->cont.lookup.postparent          = *postparent;
+        priv = this->private;
+        success_children = local->cont.lookup.success_children;
+        bufs = local->cont.lookup.bufs;
+        memcpy (local->fresh_children, success_children,
+                sizeof (*success_children) * priv->child_count);
+        for (i = 0; i < priv->child_count; i++) {
+                child = local->fresh_children[i];
+                if (child == -1)
+                        break;
+                if (child == subvol)
+                        continue;
+                if (uuid_is_null (bufs[child].ia_gfid) &&
+                    (bufs[child].ia_type == bufs[subvol].ia_type))
+                        continue;
+                afr_children_rm_child (success_children, child,
+                                       priv->child_count);
+                local->success_count--;
+        }
+        afr_reset_children (local->fresh_children, priv->child_count);
+}
 
-                        if (priv->first_lookup && inode->ino == 1) {
-                                gf_log (this->name, GF_LOG_TRACE,
-                                        "added root inode");
-                                priv->root_inode = inode_ref (inode);
-                                priv->first_lookup = 0;
-                        }
+void
+afr_succeed_lookup_on_latest_iatt (afr_local_t *local, xlator_t *this)
+{
+        int    lsubvol = 0;
 
-                        *lookup_buf = *buf;
+        if (!afr_lookup_conflicting_entries (local, this))
+                goto out;
 
-                        lookup_buf->ia_ino = afr_itransform (buf->ia_ino,
-                                                             priv->child_count,
-                                                             child_index);
-                        if (priv->read_child >= 0) {
-                                afr_set_read_child (this,
-                                                    local->cont.lookup.inode,
-                                                    priv->read_child);
-                        } else {
-                                afr_set_read_child (this,
-                                                    local->cont.lookup.inode,
-                                                    child_index);
-                        }
+        lsubvol = afr_lookup_get_latest_subvol (local, this);
+        if (lsubvol < 0)
+                goto out;
+        afr_lookup_mark_other_entries_stale (local, this, lsubvol);
+out:
+        return;
+}
 
-		} else {
-                        afr_lookup_self_heal_check (this, local, buf, lookup_buf);
-
-                        if (child_index == local->read_child_index) {
-                                /*
-                                   lookup has succeeded on the read child.
-                                   So use its inode number
-                                */
-                                if (local->cont.lookup.xattr)
-                                        dict_unref (local->cont.lookup.xattr);
-
-                                local->cont.lookup.xattr = dict_ref (xattr);
-                                local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
-                                local->cont.lookup.postparent          = *postparent;
-
-                                *lookup_buf = *buf;
-
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this,
-                                                            local->cont.lookup.inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this,
-                                                            local->cont.lookup.inode,
-                                                            local->read_child_index);
-                                }
-                        }
+gf_boolean_t
+afr_is_entry_possibly_under_creation (afr_local_t *local, xlator_t *this)
+{
+        /*
+         * We need to perform this test in lookup done and treat on going
+         * create/DELETE as ENOENT.
+         * Reason:
+        Multiple clients A, B and C are attempting 'mkdir -p /mnt/a/b/c'
+
+        1 Client A is in the middle of mkdir(/a). It has acquired lock.
+          It has performed mkdir(/a) on one subvol, and second one is still
+          in progress
+        2 Client B performs a lookup, sees directory /a on one,
+          ENOENT on the other, succeeds lookup.
+        3 Client B performs lookup on /a/b on both subvols, both return ENOENT
+          (one subvol because /a/b does not exist, another because /a
+          itself does not exist)
+        4 Client B proceeds to mkdir /a/b. It obtains entrylk on inode=/a with
+          basename=b on one subvol, but fails on other subvol as /a is yet to
+          be created by Client A.
+        5 Client A finishes mkdir of /a on other subvol
+        6 Client C also attempts to create /a/b, lookup returns ENOENT on
+          both subvols.
+        7 Client C tries to obtain entrylk on on inode=/a with basename=b,
+          obtains on one subvol (where B had failed), and waits for B to unlock
+          on other subvol.
+        8 Client B finishes mkdir() on one subvol with GFID-1 and completes
+          transaction and unlocks
+        9 Client C gets the lock on the second subvol, At this stage second
+          subvol already has /a/b created from Client B, but Client C does not
+          check that in the middle of mkdir transaction
+        10 Client C attempts mkdir /a/b on both subvols. It succeeds on
+           ONLY ONE (where Client B could not get lock because of
+           missing parent /a dir) with GFID-2, and gets EEXIST from ONE subvol.
+        This way we have /a/b in GFID mismatch. One subvol got GFID-1 because
+        Client B performed transaction on only one subvol (because entrylk()
+        could not be obtained on second subvol because of missing parent dir --
+        caused by premature/speculative succeeding of lookup() on /a when locks
+        are detected). Other subvol gets GFID-2 from Client C because while
+        it was waiting for entrylk() on both subvols, Client B was in the
+        middle of creating mkdir() on only one subvol, and Client C does not
+        "expect" this when it is between lock() and pre-op()/op() phase of the
+        transaction.
+         */
+	if (local->cont.lookup.parent_entrylk && local->enoent_count)
+		return _gf_true;
+
+	return _gf_false;
+}
 
-		}
 
-		local->success_count++;
-	}
-unlock:
-	UNLOCK (&frame->lock);
+static void
+afr_lookup_done (call_frame_t *frame, xlator_t *this)
+{
+        int                 unwind = 1;
+        afr_private_t       *priv  = NULL;
+        afr_local_t         *local = NULL;
+        int                 ret = -1;
+        gf_boolean_t        sh_launched = _gf_false;
+        gf_boolean_t        fail_conflict = _gf_false;
+        int                 gfid_miss_count = 0;
+        int                 enotconn_count = 0;
+        int                 up_children_count = 0;
 
-	call_count = afr_frame_return (frame);
+        priv  = this->private;
+        local = frame->local;
 
-	if (call_count == 0) {
-                afr_lookup_done (frame, this, lookup_buf);
+	if (afr_is_entry_possibly_under_creation (local, this)) {
+		local->op_ret = -1;
+		local->op_errno = ENOENT;
+		goto unwind;
 	}
 
-	return 0;
-}
+        if (local->op_ret < 0)
+                goto unwind;
 
+        if (local->cont.lookup.parent_entrylk && local->success_count > 1)
+                afr_succeed_lookup_on_latest_iatt (local, this);
+
+        gfid_miss_count = afr_lookup_gfid_missing_count (local, this);
+        up_children_count = afr_up_children_count (local->child_up,
+                                                   priv->child_count);
+        enotconn_count = priv->child_count - up_children_count;
+        if ((gfid_miss_count == local->success_count) &&
+            (enotconn_count > 0)) {
+                local->op_ret = -1;
+                local->op_errno = EIO;
+                gf_log (this->name, GF_LOG_ERROR, "Failing lookup for %s, "
+                        "LOOKUP on a file without gfid is not allowed when "
+                        "some of the children are down", local->loc.path);
+                goto unwind;
+        }
 
-int
-afr_revalidate_lookup_cbk (call_frame_t *frame, void *cookie,
-                           xlator_t *this, int32_t op_ret, int32_t op_errno,
-                           inode_t *inode, struct iatt *buf, dict_t *xattr,
-                           struct iatt *postparent)
+        if ((gfid_miss_count == local->success_count) &&
+            uuid_is_null (local->cont.lookup.gfid_req)) {
+                local->op_ret = -1;
+                local->op_errno = ENODATA;
+                gf_log (this->name, GF_LOG_ERROR, "%s: No gfid present",
+                        local->loc.path);
+                goto unwind;
+        }
+
+        if (gfid_miss_count && uuid_is_null (local->cont.lookup.gfid_req))
+                fail_conflict = _gf_true;
+        ret = afr_lookup_done_success_action (frame, this, fail_conflict);
+        if (ret)
+                goto unwind;
+        uuid_copy (local->self_heal.sh_gfid_req, local->cont.lookup.gfid_req);
+
+        afr_lookup_perform_self_heal (frame, this, &sh_launched);
+        if (sh_launched) {
+                unwind = 0;
+                goto unwind;
+        }
+
+ unwind:
+         if (unwind) {
+                 AFR_STACK_UNWIND (lookup, frame, local->op_ret,
+                                   local->op_errno, local->cont.lookup.inode,
+                                   &local->cont.lookup.buf,
+                                   local->cont.lookup.xattr,
+                                   &local->cont.lookup.postparent);
+        }
+}
+
+/*
+ * During a lookup, some errors are more "important" than
+ * others in that they must be given higher priority while
+ * returning to the user.
+ *
+ * The hierarchy is ESTALE > EIO > ENOENT > others
+ */
+int32_t
+afr_most_important_error(int32_t old_errno, int32_t new_errno,
+			 gf_boolean_t eio)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	struct iatt *   lookup_buf = NULL;
+	if (old_errno == ESTALE || new_errno == ESTALE)
+		return ESTALE;
+	if (eio && (old_errno == EIO || new_errno == EIO))
+		return EIO;
+	if (old_errno == ENOENT || new_errno == ENOENT)
+		return ENOENT;
+
+	return new_errno;
+}
 
-	int             call_count      = -1;
-	int             child_index     = -1;
-        int             first_up_child  = -1;
+int32_t
+afr_resultant_errno_get (int32_t *children,
+                         int *child_errno, unsigned int child_count)
+{
+        int     i = 0;
+        int32_t op_errno = 0;
+        int     child = 0;
+
+        for (i = 0; i < child_count; i++) {
+                if (children) {
+                        child = children[i];
+                        if (child == -1)
+                                break;
+                } else {
+                        child = i;
+                }
+		op_errno = afr_most_important_error(op_errno,
+						    child_errno[child],
+						    _gf_false);
+        }
+        return op_errno;
+}
 
-	child_index = (long) cookie;
-	priv = this->private;
+static void
+afr_lookup_handle_error (afr_local_t *local, int32_t op_ret,  int32_t op_errno)
+{
+        GF_ASSERT (local);
+        if (op_errno == ENOENT)
+                local->enoent_count++;
 
-	LOCK (&frame->lock);
-	{
-		local = frame->local;
+	local->op_errno = afr_most_important_error(local->op_errno, op_errno,
+						   _gf_false);
 
-		lookup_buf = &local->cont.lookup.buf;
+        if (local->op_errno == ESTALE) {
+                local->op_ret = -1;
+        }
+}
 
-		if (op_ret == -1) {
-			if (op_errno == ENOENT)
-				local->enoent_count++;
+static void
+afr_set_root_inode_on_first_lookup (afr_local_t *local, xlator_t *this,
+                                    inode_t *inode)
+{
+        afr_private_t           *priv = NULL;
+        GF_ASSERT (inode);
 
-                        if (__error_more_important (local->op_errno, op_errno))
-                            local->op_errno = op_errno;
+        if (!__is_root_gfid (inode->gfid))
+                goto out;
+        if (!afr_is_fresh_lookup (&local->loc, this))
+                goto out;
+        priv = this->private;
+        if ((priv->first_lookup)) {
+                gf_log (this->name, GF_LOG_INFO, "added root inode");
+                priv->root_inode = inode_ref (inode);
+                priv->first_lookup = 0;
+        }
+out:
+        return;
+}
 
-                            if (local->op_errno == ESTALE) {
-                                    local->op_ret = -1;
-                            }
+static void
+afr_lookup_cache_args (afr_local_t *local, int child_index, dict_t *xattr,
+                       struct iatt *buf, struct iatt *postparent)
+{
+        GF_ASSERT (child_index >= 0);
+        local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
+        local->cont.lookup.postparents[child_index] = *postparent;
+        local->cont.lookup.bufs[child_index] = *buf;
+}
 
-                            goto unlock;
-                }
+static void
+afr_lookup_handle_first_success (afr_local_t *local, xlator_t *this,
+                                 inode_t *inode, struct iatt *buf)
+{
+        local->cont.lookup.inode      = inode_ref (inode);
+        local->cont.lookup.buf        = *buf;
+        afr_set_root_inode_on_first_lookup (local, this, inode);
+}
 
-                afr_lookup_collect_xattr (local, this, child_index, xattr);
+static int32_t
+afr_discovery_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, dict_t *dict,
+                   dict_t *xdata)
+{
+        int              ret            = 0;
+        char            *pathinfo       = NULL;
+        gf_boolean_t     is_local        = _gf_false;
+        afr_private_t   *priv           = NULL;
+        int32_t          child_index    = -1;
 
-                first_up_child = afr_first_up_child (priv);
+        if (op_ret != 0) {
+                goto out;
+        }
 
-                if (child_index == first_up_child) {
-                        local->cont.lookup.ino =
-                                afr_itransform (buf->ia_ino,
-                                                priv->child_count,
-                                                first_up_child);
-                }
+        ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, &pathinfo);
+        if (ret != 0) {
+                goto out;
+        }
 
-		/* in case of revalidate, we need to send stat of the
-		 * child whose stat was sent during the first lookup.
-		 * (so that time stamp does not vary with revalidate.
-		 * in case it is down, stat of the fist success will
-		 * be replied */
+        ret = afr_local_pathinfo (pathinfo, &is_local);
+        if (ret) {
+                goto out;
+        }
 
-		/* inode number should be preserved across revalidates */
+        priv = this->private;
+        /*
+         * Note that one local subvolume will override another here.  The only
+         * way to avoid that would be to retain extra information about whether
+         * the previous read_child is local, and it's just not worth it.  Even
+         * the slowest local subvolume is far preferable to a remote one.
+         */
+        if (is_local) {
+                child_index = (int32_t)(long)cookie;
+                gf_log (this->name, GF_LOG_INFO,
+                        "selecting local read_child %s",
+                        priv->children[child_index]->name);
+                priv->read_child = child_index;
+        }
 
-		if (local->success_count == 0) {
-                        if (local->op_errno != ESTALE)
-                                local->op_ret = op_ret;
+out:
+        STACK_DESTROY(frame->root);
+        return 0;
+}
 
-			local->cont.lookup.inode               = inode_ref (inode);
-			local->cont.lookup.xattr               = dict_ref (xattr);
-			local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
-                        local->cont.lookup.postparent          = *postparent;
+static void
+afr_attempt_local_discovery (xlator_t *this, int32_t child_index)
+{
+        call_frame_t    *newframe = NULL;
+        loc_t            tmploc = {0,};
+        afr_private_t   *priv = this->private;
 
-			*lookup_buf = *buf;
+        newframe = create_frame(this,this->ctx->pool);
+        if (!newframe) {
+                return;
+        }
 
-                        lookup_buf->ia_ino = afr_itransform (buf->ia_ino,
-                                                             priv->child_count,
-                                                             child_index);
+        tmploc.gfid[sizeof(tmploc.gfid)-1] = 1;
+        STACK_WIND_COOKIE (newframe, afr_discovery_cbk,
+                           (void *)(long)child_index,
+                           priv->children[child_index],
+                           priv->children[child_index]->fops->getxattr,
+                           &tmploc, GF_XATTR_PATHINFO_KEY, NULL);
+}
 
-                        if (priv->read_child >= 0) {
-                                afr_set_read_child (this,
-                                                    local->cont.lookup.inode,
-                                                    priv->read_child);
-                        } else {
-                                afr_set_read_child (this,
-                                                    local->cont.lookup.inode,
-                                                    child_index);
-                        }
+static void
+afr_lookup_handle_success (afr_local_t *local, xlator_t *this, int32_t child_index,
+                           int32_t op_ret, int32_t op_errno, inode_t *inode,
+                           struct iatt *buf, dict_t *xattr,
+                           struct iatt *postparent)
+{
+        afr_private_t   *priv   = this->private;
 
-		} else {
-                        afr_lookup_self_heal_check (this, local, buf, lookup_buf);
+        if (local->success_count == 0) {
+                if (local->op_errno != ESTALE) {
+                        local->op_ret = op_ret;
+                        local->op_errno = 0;
+                }
+                afr_lookup_handle_first_success (local, this, inode, buf);
+        }
+        afr_lookup_update_lk_counts (local, this,
+                                     child_index, xattr);
 
-                        if (child_index == local->read_child_index) {
+        afr_lookup_cache_args (local, child_index, xattr,
+                               buf, postparent);
 
-                                /*
-                                   lookup has succeeded on the read child.
-                                   So use its inode number
-                                */
+        if (local->do_discovery && (priv->read_child == (-1))) {
+                afr_attempt_local_discovery(this,child_index);
+        }
 
-                                if (local->cont.lookup.xattr)
-                                        dict_unref (local->cont.lookup.xattr);
+        local->cont.lookup.success_children[local->success_count] = child_index;
+        local->success_count++;
+}
 
-                                local->cont.lookup.xattr               = dict_ref (xattr);
-                                local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
-                                local->cont.lookup.postparent          = *postparent;
+int
+afr_lookup_cbk (call_frame_t *frame, void *cookie,
+                xlator_t *this,  int32_t op_ret,  int32_t op_errno,
+                inode_t *inode,   struct iatt *buf, dict_t *xattr,
+                struct iatt *postparent)
+{
+        afr_local_t *   local = NULL;
+        int             call_count      = -1;
+        int             child_index     = -1;
 
-                                *lookup_buf = *buf;
+         child_index = (long) cookie;
 
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this,
-                                                            local->cont.lookup.inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this,
-                                                            local->cont.lookup.inode,
-                                                            local->read_child_index);
-                                }
-                        }
+        LOCK (&frame->lock);
+        {
+                local = frame->local;
 
-		}
+                if (op_ret == -1) {
+                        afr_lookup_handle_error (local, op_ret, op_errno);
+                        goto unlock;
+                }
+                afr_lookup_handle_success (local, this, child_index, op_ret,
+                                           op_errno, inode, buf, xattr,
+                                           postparent);
 
-		local->success_count++;
-	}
+         }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-                afr_lookup_done (frame, this, lookup_buf);
-	}
+        call_count = afr_frame_return (frame);
+        if (call_count == 0) {
+               afr_lookup_done (frame, this);
+        }
 
-	return 0;
+         return 0;
 }
 
-
 int
-afr_lookup (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, dict_t *xattr_req)
+afr_lookup_cont_init (afr_local_t *local, unsigned int child_count)
 {
-	afr_private_t *priv = NULL;
-	afr_local_t   *local = NULL;
-	int            ret = -1;
-	int            i = 0;
+        int               ret            = -ENOMEM;
+        struct iatt       *iatts         = NULL;
+        int32_t           *success_children = NULL;
+        int32_t           *sources       = NULL;
+        int32_t           **pending_matrix = NULL;
+
+        GF_ASSERT (local);
+        local->cont.lookup.xattrs = GF_CALLOC (child_count,
+                                               sizeof (*local->cont.lookup.xattr),
+                                               gf_afr_mt_dict_t);
+        if (NULL == local->cont.lookup.xattrs)
+                goto out;
 
-        fop_lookup_cbk_t callback;
+        iatts = GF_CALLOC (child_count, sizeof (*iatts), gf_afr_mt_iatt);
+        if (NULL == iatts)
+                goto out;
+        local->cont.lookup.postparents = iatts;
 
-        int call_count = 0;
+        iatts = GF_CALLOC (child_count, sizeof (*iatts), gf_afr_mt_iatt);
+        if (NULL == iatts)
+                goto out;
+        local->cont.lookup.bufs = iatts;
 
-        uint64_t       ctx;
+        success_children = afr_children_create (child_count);
+        if (NULL == success_children)
+                goto out;
+        local->cont.lookup.success_children = success_children;
 
-	int32_t        op_errno = 0;
+        local->fresh_children = afr_children_create (child_count);
+        if (NULL == local->fresh_children)
+                goto out;
 
-	priv = this->private;
+        sources = GF_CALLOC (sizeof (*sources), child_count, gf_afr_mt_int32_t);
+        if (NULL == sources)
+                goto out;
+        local->cont.lookup.sources = sources;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        pending_matrix = afr_matrix_create (child_count, child_count);
+        if (NULL == pending_matrix)
+                goto out;
+        local->cont.lookup.pending_matrix = pending_matrix;
 
-	local->op_ret = -1;
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+afr_lookup (call_frame_t *frame, xlator_t *this,
+            loc_t *loc, dict_t *xattr_req)
+{
+        afr_private_t    *priv           = NULL;
+        afr_local_t      *local          = NULL;
+        void              *gfid_req      = NULL;
+        int               ret            = -1;
+        int               i              = 0;
+        int               call_count     = 0;
+        uint64_t          ctx            = 0;
+        int32_t           op_errno       = 0;
+        int               allow_sh       = 0;
+        priv = this->private;
+
+        AFR_LOCAL_ALLOC_OR_GOTO (local, out);
 
-	frame->local = local;
+        local->op_ret = -1;
 
-        if (!strcmp (loc->path, "/" GF_REPLICATE_TRASH_DIR)) {
-                op_errno = ENOENT;
+        frame->local = local;
+        local->fop = GF_FOP_LOOKUP;
+
+        loc_copy (&local->loc, loc);
+        ret = loc_path (&local->loc, NULL);
+        if (ret < 0) {
+                op_errno = EINVAL;
                 goto out;
         }
 
-	loc_copy (&local->loc, loc);
+        if (local->loc.path &&
+            (strcmp (local->loc.path, "/" GF_REPLICATE_TRASH_DIR) == 0)) {
+                op_errno = EPERM;
+                ret = -1;
+                goto out;
+        }
 
-        ret = inode_ctx_get (loc->inode, this, &ctx);
+        ret = inode_ctx_get (local->loc.inode, this, &ctx);
         if (ret == 0) {
                 /* lookup is a revalidate */
 
-                callback = afr_revalidate_lookup_cbk;
-
-                local->cont.lookup.is_revalidate = _gf_true;
-                local->read_child_index          = afr_read_child (this,
-                                                                   loc->inode);
+                local->read_child_index = afr_inode_get_read_ctx (this,
+                                                               local->loc.inode,
+                                                               NULL);
         } else {
-                callback = afr_fresh_lookup_cbk;
-
                 LOCK (&priv->read_child_lock);
                 {
-                        local->read_child_index = (++priv->read_child_rr)
-                                % (priv->child_count);
+                        if (priv->hash_mode) {
+                                local->read_child_index = -1;
+                        }
+                        else {
+                                local->read_child_index =
+                                        (++priv->read_child_rr) %
+                                        (priv->child_count);
+                        }
                 }
                 UNLOCK (&priv->read_child_lock);
+                local->cont.lookup.fresh_lookup = _gf_true;
         }
 
-        if (loc->parent)
-                local->cont.lookup.parent_ino = loc->parent->ino;
-
-	local->child_up = memdup (priv->child_up, priv->child_count);
+        local->child_up = memdup (priv->child_up,
+                                  sizeof (*local->child_up) * priv->child_count);
+        if (NULL == local->child_up) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-        local->cont.lookup.xattrs = GF_CALLOC (priv->child_count,
-                                    sizeof (*local->cont.lookup.xattr),
-                                    gf_afr_mt_dict_t);
+        ret = afr_lookup_cont_init (local, priv->child_count);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto out;
+        }
 
-	local->call_count = afr_up_children_count (priv->child_count,
-                                                   local->child_up);
+        local->call_count = afr_up_children_count (local->child_up,
+                                                   priv->child_count);
         call_count = local->call_count;
-
         if (local->call_count == 0) {
                 ret      = -1;
                 op_errno = ENOTCONN;
                 goto out;
         }
 
-	/* By default assume ENOTCONN. On success it will be set to 0. */
-	local->op_errno = ENOTCONN;
+        /* By default assume ENOTCONN. On success it will be set to 0. */
+        local->op_errno = ENOTCONN;
 
-	if (xattr_req == NULL)
-		local->xattr_req = dict_new ();
-	else
-		local->xattr_req = dict_ref (xattr_req);
+        ret = dict_get_int32 (xattr_req, "allow-sh-for-running-transaction",
+                              &allow_sh);
+        dict_del (xattr_req, "allow-sh-for-running-transaction");
+        local->allow_sh_for_running_transaction = allow_sh;
 
-        for (i = 0; i < priv->child_count; i++) {
-		ret = dict_set_uint64 (local->xattr_req, priv->pending_key[i],
-				       3 * sizeof(int32_t));
-                if (ret < 0)
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "Unable to set dict value.");
-                /* 3 = data+metadata+entry */
-        }
-
-	ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_OPEN_FD_COUNT, 0);
-        if (ret < 0) {
-                gf_log (this->name, GF_LOG_WARNING,
-                        "Unable to set dict value.");
-        }
-
-        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
-        if (ret < 0) {
-                gf_log (this->name, GF_LOG_WARNING,
-                        "Unable to set dict value.");
+        ret = afr_lookup_xattr_req_prepare (local, this, xattr_req, &local->loc,
+                                            &gfid_req);
+        if (ret) {
+                local->op_errno = -ret;
+                goto out;
         }
-
-        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
-        if (ret < 0) {
-                gf_log (this->name, GF_LOG_WARNING,
-                        "Unable to set dict value.");
+        afr_lookup_save_gfid (local->cont.lookup.gfid_req, gfid_req,
+                              &local->loc);
+        local->fop = GF_FOP_LOOKUP;
+        if (priv->choose_local && !priv->did_discovery) {
+                if (gfid_req && __is_root_gfid(gfid_req)) {
+                        local->do_discovery = _gf_true;
+                        priv->did_discovery = _gf_true;
+                }
         }
-
-	for (i = 0; i < priv->child_count; i++) {
+        for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, callback, (void *) (long) i,
+                        STACK_WIND_COOKIE (frame, afr_lookup_cbk,
+                                           (void *) (long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->lookup,
-                                           loc, local->xattr_req);
+                                           &local->loc, local->xattr_req);
                         if (!--call_count)
                                 break;
                 }
-	}
+        }
 
-	ret = 0;
+        ret = 0;
 out:
-	if (ret == -1)
-		AFR_STACK_UNWIND (lookup, frame, -1, op_errno,
+        if (ret)
+                AFR_STACK_UNWIND (lookup, frame, -1, op_errno,
                                   NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 /* {{{ open */
 
 int
-afr_fd_ctx_set (xlator_t *this, fd_t *fd)
+__afr_fd_ctx_set (xlator_t *this, fd_t *fd)
 {
-        afr_private_t * priv = NULL;
-
-        int ret    = 0;
-
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
+        afr_private_t * priv   = NULL;
+        int             ret    = -1;
+        uint64_t        ctx    = 0;
+        afr_fd_ctx_t *  fd_ctx = NULL;
 
         VALIDATE_OR_GOTO (this->private, out);
         VALIDATE_OR_GOTO (fd, out);
 
         priv = this->private;
 
-        LOCK (&fd->lock);
-        {
-                ret = __fd_ctx_get (fd, this, &ctx);
-
-                if (ret == 0)
-                        goto unlock;
-
-                fd_ctx = GF_CALLOC (1, sizeof (afr_fd_ctx_t),
-                                    gf_afr_mt_afr_fd_ctx_t);
-                if (!fd_ctx) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
+        ret = __fd_ctx_get (fd, this, &ctx);
 
-                        ret = -ENOMEM;
-                        goto unlock;
-                }
-
-                fd_ctx->pre_op_done = GF_CALLOC (sizeof (*fd_ctx->pre_op_done),
-                                                 priv->child_count,
-                                                 gf_afr_mt_char);
-                if (!fd_ctx->pre_op_done) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        ret = -ENOMEM;
-                        goto unlock;
-                }
+        if (ret == 0)
+                goto out;
 
-                fd_ctx->opened_on = GF_CALLOC (sizeof (*fd_ctx->opened_on),
-                                               priv->child_count,
-                                               gf_afr_mt_char);
-                if (!fd_ctx->opened_on) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        ret = -ENOMEM;
-                        goto unlock;
-                }
+        fd_ctx = GF_CALLOC (1, sizeof (afr_fd_ctx_t),
+                            gf_afr_mt_afr_fd_ctx_t);
+        if (!fd_ctx) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-                fd_ctx->child_failed = GF_CALLOC (
-                                         sizeof (*fd_ctx->child_failed),
+        fd_ctx->pre_op_done = GF_CALLOC (sizeof (*fd_ctx->pre_op_done),
                                          priv->child_count,
                                          gf_afr_mt_char);
-
-                if (!fd_ctx->child_failed) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-
-                        ret = -ENOMEM;
-                        goto unlock;
-                }
-
-                fd_ctx->up_count   = priv->up_count;
-                fd_ctx->down_count = priv->down_count;
-
-                ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
-
-                INIT_LIST_HEAD (&fd_ctx->entries);
+        if (!fd_ctx->pre_op_done) {
+                ret = -ENOMEM;
+                goto out;
         }
-unlock:
-        UNLOCK (&fd->lock);
-out:
-        return ret;
-}
-
-/* {{{ flush */
-
-int
-afr_flush_unwind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *   local = NULL;
-	call_frame_t   *main_frame = NULL;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame)
-			main_frame = local->transaction.main_frame;
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-		AFR_STACK_UNWIND (flush, main_frame,
-                                  local->op_ret, local->op_errno);
-	}
-
-	return 0;
-}
-
-
-int
-afr_flush_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		      int32_t op_ret, int32_t op_errno)
-{
-        afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
-	int need_unwind = 0;
-
-	local = frame->local;
-	priv  = this->private;
-
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-			}
-			local->success_count++;
-
-			if (local->success_count == priv->wait_count) {
-				need_unwind = 1;
-			}
-		}
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	if (need_unwind)
-		afr_flush_unwind (frame, this);
-
-	call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
-
-	return 0;
-}
-
-
-int
-afr_flush_wind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
+        fd_ctx->pre_op_piggyback = GF_CALLOC (sizeof (*fd_ctx->pre_op_piggyback),
+                                              priv->child_count,
+                                              gf_afr_mt_char);
+        if (!fd_ctx->pre_op_piggyback) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-	int i = 0;
-	int call_count = -1;
+        fd_ctx->opened_on = GF_CALLOC (sizeof (*fd_ctx->opened_on),
+                                       priv->child_count,
+                                       gf_afr_mt_int32_t);
+        if (!fd_ctx->opened_on) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-	local = frame->local;
-	priv = this->private;
+        fd_ctx->lock_piggyback = GF_CALLOC (sizeof (*fd_ctx->lock_piggyback),
+                                            priv->child_count,
+                                            gf_afr_mt_char);
+        if (!fd_ctx->lock_piggyback) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+        fd_ctx->lock_acquired = GF_CALLOC (sizeof (*fd_ctx->lock_acquired),
+                                           priv->child_count,
+                                           gf_afr_mt_char);
+        if (!fd_ctx->lock_acquired) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+        fd_ctx->up_count   = priv->up_count;
+        fd_ctx->down_count = priv->down_count;
 
-	local->call_count = call_count;
+        fd_ctx->locked_on = GF_CALLOC (sizeof (*fd_ctx->locked_on),
+                                       priv->child_count,
+                                       gf_afr_mt_char);
+        if (!fd_ctx->locked_on) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_flush_wind_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->flush,
-					   local->fd);
+	pthread_mutex_init (&fd_ctx->delay_lock, NULL);
+        INIT_LIST_HEAD (&fd_ctx->entries);
+        fd_ctx->call_child = -1;
 
-			if (!--call_count)
-				break;
-		}
-	}
+        INIT_LIST_HEAD (&fd_ctx->eager_locked);
 
-	return 0;
+        ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
+        if (ret)
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "failed to set fd ctx (%p)", fd);
+out:
+        return ret;
 }
 
 
 int
-afr_flush_done (call_frame_t *frame, xlator_t *this)
+afr_fd_ctx_set (xlator_t *this, fd_t *fd)
 {
-	afr_local_t *local = NULL;
+        int ret = -1;
 
-	local = frame->local;
-
-	local->transaction.unwind (frame, this);
-
-        AFR_STACK_DESTROY (frame);
+        LOCK (&fd->lock);
+        {
+                ret = __afr_fd_ctx_set (this, fd);
+        }
+        UNLOCK (&fd->lock);
 
-	return 0;
+        return ret;
 }
 
+/* {{{ flush */
 
 int
-afr_plain_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno)
-
+afr_flush_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_local_t *local = NULL;
-
-	int call_count = -1;
+        afr_local_t *   local = NULL;
+        int call_count  = -1;
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
+        LOCK (&frame->lock);
+        {
+                if (op_ret != -1) {
+                        if (local->success_count == 0) {
+                                local->op_ret = op_ret;
+                        }
+                        local->success_count++;
+                }
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
 
 	call_count = afr_frame_return (frame);
 
 	if (call_count == 0)
-		AFR_STACK_UNWIND (flush, frame, local->op_ret, local->op_errno);
+		AFR_STACK_UNWIND(flush, frame, local->op_ret,
+				 local->op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
-
 static int
-__no_pre_op_done (xlator_t *this, fd_t *fd)
+afr_flush_wrapper (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
-        int i      = 0;
-        int op_ret = 1;
-
-        int _ret = 0;
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
-
-        afr_private_t *priv = NULL;
+        int           i      = 0;
+        afr_local_t   *local = NULL;
+        afr_private_t *priv  = NULL;
+        int call_count       = -1;
 
         priv = this->private;
+        local = frame->local;
+        call_count = local->call_count;
 
-        LOCK (&fd->lock);
-        {
-                _ret = __fd_ctx_get (fd, this, &ctx);
-
-                if (_ret < 0) {
-                        goto out;
-                }
-
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                for (i = 0; i < priv->child_count; i++) {
-                        if (fd_ctx->pre_op_done[i]) {
-                                op_ret = 0;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, afr_flush_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->flush,
+                                           local->fd, NULL);
+                        if (!--call_count)
                                 break;
-                        }
+
                 }
         }
-out:
-        UNLOCK (&fd->lock);
 
-        return op_ret;
+        return 0;
 }
 
-
 int
-afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
+afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-
-        call_frame_t  * transaction_frame = NULL;
-
-	int ret        = -1;
-
-	int op_ret   = -1;
-	int op_errno = 0;
-
-        int i          = 0;
-        int call_count = 0;
+        afr_private_t *priv  = NULL;
+        afr_local_t   *local = NULL;
+        call_stub_t   *stub = NULL;
+        int            ret        = -1;
+        int            op_errno   = 0;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	priv = this->private;
+        priv = this->private;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+	AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+	local = frame->local;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
+	ret = afr_local_init(local, priv, &op_errno);
+	if (ret < 0)
 		goto out;
-	}
-
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-        if (__no_pre_op_done (this, fd)) {
-                frame->local = local;
-
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i]) {
-                                STACK_WIND_COOKIE (frame, afr_plain_flush_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->flush,
-                                                   fd);
-                                if (!--call_count)
-                                        break;
-                        }
-                }
-        } else {
-                transaction_frame = copy_frame (frame);
-                if (!transaction_frame) {
-                        op_errno = ENOMEM;
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory.");
-                        goto out;
-                }
 
-                transaction_frame->local = local;
-
-                local->op = GF_FOP_FLUSH;
-
-                local->transaction.fop    = afr_flush_wind;
-                local->transaction.done   = afr_flush_done;
-                local->transaction.unwind = afr_flush_unwind;
-
-                local->fd                 = fd_ref (fd);
-
-                local->transaction.main_frame = frame;
-                local->transaction.start  = 0;
-                local->transaction.len    = 0;
-
-                afr_transaction (transaction_frame, this, AFR_FLUSH_TRANSACTION);
+	local->fd = fd_ref(fd);
+        stub = fop_flush_stub (frame, afr_flush_wrapper, fd, xdata);
+        if (!stub) {
+                ret = -1;
+                op_errno = ENOMEM;
+                goto out;
         }
 
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
+        afr_delayed_changelog_wake_resume (this, fd, stub);
+	ret = 0;
 
-		AFR_STACK_UNWIND (flush, frame, op_ret, op_errno);
-	}
+out:
+	if (ret < 0)
+		AFR_STACK_UNWIND(flush, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 /* }}} */
@@ -1513,21 +2761,24 @@ afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd)
         int             ret = 0;
 
         ret = fd_ctx_get (fd, this, &ctx);
-
         if (ret < 0)
                 goto out;
 
         fd_ctx = (afr_fd_ctx_t *)(long) ctx;
 
         if (fd_ctx) {
-                if (fd_ctx->child_failed)
-                        GF_FREE (fd_ctx->child_failed);
+                GF_FREE (fd_ctx->pre_op_done);
+
+                GF_FREE (fd_ctx->opened_on);
 
-                if (fd_ctx->pre_op_done)
-                        GF_FREE (fd_ctx->pre_op_done);
+                GF_FREE (fd_ctx->locked_on);
 
-                if (fd_ctx->opened_on)
-                        GF_FREE (fd_ctx->opened_on);
+                GF_FREE (fd_ctx->pre_op_piggyback);
+                GF_FREE (fd_ctx->lock_piggyback);
+
+                GF_FREE (fd_ctx->lock_acquired);
+
+		pthread_mutex_destroy (&fd_ctx->delay_lock);
 
                 GF_FREE (fd_ctx);
         }
@@ -1540,8 +2791,24 @@ out:
 int
 afr_release (xlator_t *this, fd_t *fd)
 {
+        afr_locked_fd_t *locked_fd = NULL;
+        afr_locked_fd_t *tmp       = NULL;
+        afr_private_t   *priv      = NULL;
+
+        priv = this->private;
+
         afr_cleanup_fd_ctx (this, fd);
 
+        list_for_each_entry_safe (locked_fd, tmp, &priv->saved_fds,
+                                  list) {
+
+                if (locked_fd->fd == fd) {
+                        list_del_init (&locked_fd->list);
+                        GF_FREE (locked_fd);
+                }
+
+        }
+
         return 0;
 }
 
@@ -1549,114 +2816,140 @@ afr_release (xlator_t *this, fd_t *fd)
 /* {{{ fsync */
 
 int
+afr_fsync_unwind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                      struct iatt *postbuf, dict_t *xdata)
+{
+        AFR_STACK_UNWIND (fsync, frame, op_ret, op_errno, prebuf, postbuf,
+                          xdata);
+        return 0;
+}
+
+int
 afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-               struct iatt *postbuf)
+               struct iatt *postbuf, dict_t *xdata)
 {
-	afr_local_t *local = NULL;
-
-	int call_count = -1;
-
+        afr_local_t *local = NULL;
+        int call_count = -1;
         int child_index = (long) cookie;
         int read_child  = 0;
+	call_stub_t *stub = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-        read_child = afr_read_child (this, local->fd->inode);
+        read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
 
-	LOCK (&frame->lock);
-	{
+        LOCK (&frame->lock);
+        {
                 if (child_index == read_child) {
                         local->read_child_returned = _gf_true;
                 }
 
-		if (op_ret == 0) {
-			local->op_ret = 0;
+                if (op_ret == 0) {
+                        local->op_ret = 0;
 
-			if (local->success_count == 0) {
-				local->cont.fsync.prebuf  = *prebuf;
-				local->cont.fsync.postbuf = *postbuf;
-			}
+                        if (local->success_count == 0) {
+                                local->cont.inode_wfop.prebuf  = *prebuf;
+                                local->cont.inode_wfop.postbuf = *postbuf;
+                        }
 
                         if (child_index == read_child) {
-                                local->cont.fsync.prebuf  = *prebuf;
-                                local->cont.fsync.postbuf = *postbuf;
+                                local->cont.inode_wfop.prebuf  = *prebuf;
+                                local->cont.inode_wfop.postbuf = *postbuf;
                         }
 
-			local->success_count++;
+                        local->success_count++;
                 }
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-                local->cont.fsync.prebuf.ia_ino  = local->cont.fsync.ino;
-                local->cont.fsync.postbuf.ia_ino = local->cont.fsync.ino;
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+		/* Make a stub out of the frame, and register it
+		   with the waking up post-op. When the call-stub resumes,
+		   we are guaranteed that there was no post-op pending
+		   (i.e changelogs were unset in the server). This is an
+		   essential "guarantee", that fsync() returns only after
+		   completely finishing EVERYTHING, including the delayed
+		   post-op. This guarantee is expected by FUSE graph switching
+		   for example.
+		*/
+		stub = fop_fsync_cbk_stub (frame, afr_fsync_unwind_cbk,
+                                           local->op_ret, local->op_errno,
+                                           &local->cont.inode_wfop.prebuf,
+                                           &local->cont.inode_wfop.postbuf,
+                                           xdata);
+		if (!stub) {
+			AFR_STACK_UNWIND (fsync, frame, -1, ENOMEM, 0, 0, 0);
+			return 0;
+		}
 
-		AFR_STACK_UNWIND (fsync, frame, local->op_ret, local->op_errno,
-                                  &local->cont.fsync.prebuf,
-                                  &local->cont.fsync.postbuf);
+		/* If no new unstable writes happened between the
+		   time we cleared the unstable write witness flag in afr_fsync
+		   and now, calling afr_delayed_changelog_wake_up() should
+		   wake up and skip over the fsync phase and go straight to
+		   afr_changelog_post_op_now()
+		*/
+		afr_delayed_changelog_wake_resume (this, local->fd, stub);
         }
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
-	   int32_t datasync)
+           int32_t datasync, dict_t *xdata)
 {
-	afr_private_t *priv = NULL;
-	afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        int ret = -1;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	int ret = -1;
+        priv = this->private;
 
-	int i = 0;
-	int32_t call_count = 0;
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	priv = this->private;
+        call_count = local->call_count;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        local->fd             = fd_ref (fd);
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
+	if (afr_fd_has_witnessed_unstable_write (this, fd)) {
+		/* don't care. we only wanted to CLEAR the bit */
 	}
 
-	call_count = local->call_count;
-	frame->local = local;
-
-        local->fd             = fd_ref (fd);
-        local->cont.fsync.ino = fd->inode->ino;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_fsync_cbk,
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, afr_fsync_cbk,
                                            (void *) (long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->fsync,
-                                           fd, datasync);
-			if (!--call_count)
-				break;
-		}
-	}
+                                           fd, datasync, xdata);
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (fsync, frame, op_ret, op_errno, NULL, NULL);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
 }
 
 /* }}} */
@@ -1665,81 +2958,75 @@ out:
 
 int32_t
 afr_fsyncdir_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno)
+                  xlator_t *this, int32_t op_ret, int32_t op_errno,
+                  dict_t *xdata)
 {
-	afr_local_t *local = NULL;
-
-	int call_count = -1;
+        afr_local_t *local = NULL;
+        int call_count = -1;
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0)
+                        local->op_ret = 0;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0)
-		AFR_STACK_UNWIND (fsyncdir, frame, local->op_ret,
-                                  local->op_errno);
+        if (call_count == 0)
+                AFR_STACK_UNWIND (fsyncdir, frame, local->op_ret,
+                                  local->op_errno, xdata);
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
-	      int32_t datasync)
+              int32_t datasync, dict_t *xdata)
 {
-	afr_private_t *priv = NULL;
-	afr_local_t *local = NULL;
-
-	int ret = -1;
-
-	int i = 0;
-	int32_t call_count = 0;
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        int ret = -1;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        priv = this->private;
 
-	priv = this->private;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        call_count = local->call_count;
 
-	call_count = local->call_count;
-	frame->local = local;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_fsyncdir_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->fsyncdir,
-				    fd, datasync);
-			if (!--call_count)
-				break;
-		}
-	}
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND (frame, afr_fsyncdir_cbk,
+                                    priv->children[i],
+                                    priv->children[i]->fops->fsyncdir,
+                                    fd, datasync, xdata);
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (fsyncdir, frame, op_ret, op_errno);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
+        return 0;
 }
 
 /* }}} */
@@ -1748,82 +3035,78 @@ out:
 
 int32_t
 afr_xattrop_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno,
-		 dict_t *xattr)
+                 xlator_t *this, int32_t op_ret, int32_t op_errno,
+                 dict_t *xattr, dict_t *xdata)
 {
-	afr_local_t *local = NULL;
-
-	int call_count = -1;
+        afr_local_t *local = NULL;
+        int call_count = -1;
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0) {
+                        if (!local->cont.xattrop.xattr)
+                                local->cont.xattrop.xattr = dict_ref (xattr);
+                        local->op_ret = 0;
+                }
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0)
-		AFR_STACK_UNWIND (xattrop, frame, local->op_ret, local->op_errno,
-                                  xattr);
+        if (call_count == 0)
+                AFR_STACK_UNWIND (xattrop, frame, local->op_ret, local->op_errno,
+                local->cont.xattrop.xattr, xdata);
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
-	     gf_xattrop_flags_t optype, dict_t *xattr)
+             gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
 {
-	afr_private_t *priv = NULL;
-	afr_local_t *local  = NULL;
-
-	int ret = -1;
-
-	int i = 0;
-	int32_t call_count = 0;
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+        afr_private_t *priv = NULL;
+        afr_local_t *local  = NULL;
+        int ret = -1;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        priv = this->private;
 
-	priv = this->private;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        call_count = local->call_count;
 
-	call_count = local->call_count;
-	frame->local = local;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_xattrop_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->xattrop,
-				    loc, optype, xattr);
-			if (!--call_count)
-				break;
-		}
-	}
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND (frame, afr_xattrop_cbk,
+                                    priv->children[i],
+                                    priv->children[i]->fops->xattrop,
+                                    loc, optype, xattr, xdata);
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (xattrop, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
+        return 0;
 }
 
 /* }}} */
@@ -1832,82 +3115,80 @@ out:
 
 int32_t
 afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno,
-		  dict_t *xattr)
+                  xlator_t *this, int32_t op_ret, int32_t op_errno,
+                  dict_t *xattr, dict_t *xdata)
 {
-	afr_local_t *local = NULL;
+        afr_local_t *local = NULL;
 
-	int call_count = -1;
+        int call_count = -1;
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0) {
+                        if (!local->cont.fxattrop.xattr)
+                                local->cont.fxattrop.xattr = dict_ref (xattr);
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                        local->op_ret = 0;
+                }
 
-	call_count = afr_frame_return (frame);
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
 
-	if (call_count == 0)
-		AFR_STACK_UNWIND (fxattrop, frame, local->op_ret, local->op_errno,
-                                  xattr);
+        call_count = afr_frame_return (frame);
 
-	return 0;
+        if (call_count == 0)
+                AFR_STACK_UNWIND (fxattrop, frame, local->op_ret, local->op_errno,
+                                  local->cont.fxattrop.xattr, xdata);
+
+        return 0;
 }
 
 
 int32_t
 afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
-	      gf_xattrop_flags_t optype, dict_t *xattr)
+              gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
 {
-	afr_private_t *priv = NULL;
-	afr_local_t *local  = NULL;
-
-	int ret = -1;
-
-	int i = 0;
-	int32_t call_count = 0;
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+        afr_private_t *priv = NULL;
+        afr_local_t *local  = NULL;
+        int ret = -1;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        priv = this->private;
 
-	priv = this->private;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        call_count = local->call_count;
 
-	call_count = local->call_count;
-	frame->local = local;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_fxattrop_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->fxattrop,
-				    fd, optype, xattr);
-			if (!--call_count)
-				break;
-		}
-	}
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND (frame, afr_fxattrop_cbk,
+                                    priv->children[i],
+                                    priv->children[i]->fops->fxattrop,
+                                    fd, optype, xattr, xdata);
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
+        return 0;
 }
 
 /* }}} */
@@ -1915,643 +3196,623 @@ out:
 
 int32_t
 afr_inodelk_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno)
+                 xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
 
 {
-	afr_local_t *local = NULL;
-
-	int call_count = -1;
+        afr_local_t *local = NULL;
+        int call_count = -1;
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0)
+                        local->op_ret = 0;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0)
-		AFR_STACK_UNWIND (inodelk, frame, local->op_ret,
-                                  local->op_errno);
+        if (call_count == 0)
+                AFR_STACK_UNWIND (inodelk, frame, local->op_ret,
+                                  local->op_errno, xdata);
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_inodelk (call_frame_t *frame, xlator_t *this,
-             const char *volume, loc_t *loc, int32_t cmd, struct flock *flock)
+             const char *volume, loc_t *loc, int32_t cmd,
+             struct gf_flock *flock, dict_t *xdata)
 {
-	afr_private_t *priv = NULL;
-	afr_local_t *local  = NULL;
-
-	int ret = -1;
-
-	int i = 0;
-	int32_t call_count = 0;
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        afr_private_t *priv = NULL;
+        afr_local_t *local  = NULL;
+        int ret = -1;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	priv = this->private;
+        priv = this->private;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	call_count = local->call_count;
-	frame->local = local;
+        call_count = local->call_count;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_inodelk_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->inodelk,
-				    volume, loc, cmd, flock);
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND (frame, afr_inodelk_cbk,
+                                    priv->children[i],
+                                    priv->children[i]->fops->inodelk,
+                                    volume, loc, cmd, flock, xdata);
 
-			if (!--call_count)
-				break;
-		}
-	}
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (inodelk, frame, op_ret, op_errno);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (inodelk, frame, -1, op_errno, NULL);
+        return 0;
 }
 
 
 int32_t
 afr_finodelk_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno)
+                  xlator_t *this, int32_t op_ret, int32_t op_errno,
+                  dict_t *xdata)
 
 {
-	afr_local_t *local = NULL;
+        afr_local_t *local = NULL;
+        int call_count = -1;
 
-	int call_count = -1;
-
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0)
+                        local->op_ret = 0;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0)
-		AFR_STACK_UNWIND (finodelk, frame, local->op_ret,
-                                  local->op_errno);
+        if (call_count == 0)
+                AFR_STACK_UNWIND (finodelk, frame, local->op_ret,
+                                  local->op_errno, xdata);
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_finodelk (call_frame_t *frame, xlator_t *this,
-              const char *volume, fd_t *fd, int32_t cmd, struct flock *flock)
+              const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock,
+              dict_t *xdata)
 {
-	afr_private_t *priv = NULL;
-	afr_local_t *local  = NULL;
-
-	int ret = -1;
-
-	int i = 0;
-	int32_t call_count = 0;
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        afr_private_t *priv = NULL;
+        afr_local_t *local  = NULL;
+        int ret = -1;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	priv = this->private;
+        priv = this->private;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	call_count = local->call_count;
-	frame->local = local;
+        call_count = local->call_count;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_finodelk_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->finodelk,
-				    volume, fd, cmd, flock);
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND (frame, afr_finodelk_cbk,
+                                    priv->children[i],
+                                    priv->children[i]->fops->finodelk,
+                                    volume, fd, cmd, flock, xdata);
 
-			if (!--call_count)
-				break;
-		}
-	}
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (finodelk, frame, op_ret, op_errno);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (finodelk, frame, -1, op_errno, NULL);
+        return 0;
 }
 
 
 int32_t
-afr_entrylk_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno)
-
+afr_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_local_t *local = NULL;
+        afr_local_t *local = NULL;
+        int call_count = -1;
 
-	int call_count = -1;
-
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0)
+                        local->op_ret = 0;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0)
-		AFR_STACK_UNWIND (entrylk, frame, local->op_ret,
-                                  local->op_errno);
+        if (call_count == 0)
+                AFR_STACK_UNWIND (entrylk, frame, local->op_ret,
+                                  local->op_errno, xdata);
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_entrylk (call_frame_t *frame, xlator_t *this,
              const char *volume, loc_t *loc,
-	     const char *basename, entrylk_cmd cmd, entrylk_type type)
+             const char *basename, entrylk_cmd cmd, entrylk_type type,
+             dict_t *xdata)
 {
-	afr_private_t *priv = NULL;
-	afr_local_t *local  = NULL;
-
-	int ret = -1;
-
-	int i = 0;
-	int32_t call_count = 0;
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        afr_private_t *priv = NULL;
+        afr_local_t *local  = NULL;
+        int ret = -1;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	priv = this->private;
+        priv = this->private;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	call_count = local->call_count;
-	frame->local = local;
+        call_count = local->call_count;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_entrylk_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->entrylk,
-				    volume, loc, basename, cmd, type);
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND (frame, afr_entrylk_cbk,
+                                    priv->children[i],
+                                    priv->children[i]->fops->entrylk,
+                                    volume, loc, basename, cmd, type, xdata);
 
-			if (!--call_count)
-				break;
-		}
-	}
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (entrylk, frame, op_ret, op_errno);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (entrylk, frame, -1, op_errno, NULL);
+        return 0;
 }
 
 
 
 int32_t
 afr_fentrylk_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno)
+                  xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
 
 {
-	afr_local_t *local = NULL;
+        afr_local_t *local = NULL;
+        int call_count = -1;
 
-	int call_count = -1;
-
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0)
+                        local->op_ret = 0;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0)
-		AFR_STACK_UNWIND (fentrylk, frame, local->op_ret,
-                                  local->op_errno);
+        if (call_count == 0)
+                AFR_STACK_UNWIND (fentrylk, frame, local->op_ret,
+                                  local->op_errno, xdata);
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_fentrylk (call_frame_t *frame, xlator_t *this,
               const char *volume, fd_t *fd,
-	      const char *basename, entrylk_cmd cmd, entrylk_type type)
+              const char *basename, entrylk_cmd cmd,
+              entrylk_type type, dict_t *xdata)
 {
-	afr_private_t *priv = NULL;
-	afr_local_t *local  = NULL;
-
-	int ret = -1;
-
-	int i = 0;
-	int32_t call_count = 0;
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        afr_private_t *priv = NULL;
+        afr_local_t *local  = NULL;
+        int ret = -1;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	priv = this->private;
+        priv = this->private;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	call_count = local->call_count;
-	frame->local = local;
+        call_count = local->call_count;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_fentrylk_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->fentrylk,
-				    volume, fd, basename, cmd, type);
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND (frame, afr_fentrylk_cbk,
+                                    priv->children[i],
+                                    priv->children[i]->fops->fentrylk,
+                                    volume, fd, basename, cmd, type, xdata);
 
-			if (!--call_count)
-				break;
-		}
-	}
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (fentrylk, frame, op_ret, op_errno);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (fentrylk, frame, -1, op_errno, NULL);
+        return 0;
 }
 
 int32_t
 afr_statfs_cbk (call_frame_t *frame, void *cookie,
-		xlator_t *this, int32_t op_ret, int32_t op_errno,
-		struct statvfs *statvfs)
+                xlator_t *this, int32_t op_ret, int32_t op_errno,
+                struct statvfs *statvfs, dict_t *xdata)
 {
-	afr_local_t *local = NULL;
-
-	int call_count = 0;
+        afr_local_t *local = NULL;
+        int call_count = 0;
 
-	LOCK (&frame->lock);
-	{
-		local = frame->local;
+        LOCK (&frame->lock);
+        {
+                local = frame->local;
 
-		if (op_ret == 0) {
-			local->op_ret   = op_ret;
+                if (op_ret == 0) {
+                        local->op_ret   = op_ret;
 
-			if (local->cont.statfs.buf_set) {
-				if (statvfs->f_bavail < local->cont.statfs.buf.f_bavail)
-					local->cont.statfs.buf = *statvfs;
-			} else {
-				local->cont.statfs.buf = *statvfs;
-				local->cont.statfs.buf_set = 1;
-			}
-		}
+                        if (local->cont.statfs.buf_set) {
+                                if (statvfs->f_bavail < local->cont.statfs.buf.f_bavail)
+                                        local->cont.statfs.buf = *statvfs;
+                        } else {
+                                local->cont.statfs.buf = *statvfs;
+                                local->cont.statfs.buf_set = 1;
+                        }
+                }
 
-		if (op_ret == -1)
-			local->op_errno = op_errno;
+                if (op_ret == -1)
+                        local->op_errno = op_errno;
 
-	}
-	UNLOCK (&frame->lock);
+        }
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0)
-		AFR_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
-				  &local->cont.statfs.buf);
+        if (call_count == 0)
+                AFR_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
+                                  &local->cont.statfs.buf, xdata);
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_statfs (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc)
+            loc_t *loc, dict_t *xdata)
 {
-	afr_private_t *  priv        = NULL;
-	int              child_count = 0;
-	afr_local_t   *  local       = NULL;
-	int              i           = 0;
-
-	int ret = -1;
-	int              call_count = 0;
-	int32_t          op_ret      = -1;
-	int32_t          op_errno    = 0;
+        afr_private_t *  priv        = NULL;
+        int              child_count = 0;
+        afr_local_t   *  local       = NULL;
+        int              i           = 0;
+        int              ret = -1;
+        int              call_count = 0;
+        int32_t          op_errno    = 0;
+
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+        VALIDATE_OR_GOTO (loc, out);
 
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-	VALIDATE_OR_GOTO (loc, out);
+        priv = this->private;
+        child_count = priv->child_count;
 
-	priv = this->private;
-	child_count = priv->child_count;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        call_count = local->call_count;
 
-	frame->local = local;
-	call_count = local->call_count;
-
-	for (i = 0; i < child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_statfs_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->statfs,
-				    loc);
-			if (!--call_count)
-				break;
-		}
-	}
+        for (i = 0; i < child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND (frame, afr_statfs_cbk,
+                                    priv->children[i],
+                                    priv->children[i]->fops->statfs,
+                                    loc, xdata);
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (statfs, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
+        return 0;
 }
 
 
 int32_t
 afr_lk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno, struct flock *lock)
+                   int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+                   dict_t *xdata)
 {
-	afr_local_t * local = NULL;
-
-	int call_count = -1;
+        afr_local_t * local = NULL;
+        int call_count = -1;
 
-	local = frame->local;
-	call_count = afr_frame_return (frame);
+        local = frame->local;
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0)
-		AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
-				  lock);
+        if (call_count == 0)
+                AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
+                                  lock, xdata);
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_lk_unlock (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t   * local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int i;
-	int call_count = 0;
+        afr_local_t   * local = NULL;
+        afr_private_t * priv  = NULL;
+        int i = 0;
+        int call_count = 0;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
+        priv  = this->private;
 
-	call_count = afr_locked_nodes_count (local->cont.lk.locked_nodes,
-					     priv->child_count);
+        call_count = afr_locked_nodes_count (local->cont.lk.locked_nodes,
+                                             priv->child_count);
 
-	if (call_count == 0) {
-		AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
-                                  &local->cont.lk.ret_flock);
-		return 0;
-	}
+        if (call_count == 0) {
+                AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
+                                  &local->cont.lk.ret_flock, NULL);
+                return 0;
+        }
 
-	local->call_count = call_count;
+        local->call_count = call_count;
 
         local->cont.lk.user_flock.l_type = F_UNLCK;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->cont.lk.locked_nodes[i]) {
-			STACK_WIND (frame, afr_lk_unlock_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->lk,
-				    local->fd, F_SETLK,
-                                    &local->cont.lk.user_flock);
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->cont.lk.locked_nodes[i]) {
+                        STACK_WIND (frame, afr_lk_unlock_cbk,
+                                    priv->children[i],
+                                    priv->children[i]->fops->lk,
+                                    local->fd, F_SETLK,
+                                    &local->cont.lk.user_flock, NULL);
 
-			if (!--call_count)
-				break;
-		}
-	}
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	    int32_t op_ret, int32_t op_errno, struct flock *lock)
+            int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int child_index = -1;
+/*        int            ret  = 0; */
 
-	int child_index = -1;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
+        priv  = this->private;
 
-	child_index = (long) cookie;
+        child_index = (long) cookie;
 
-	if (!child_went_down (op_ret, op_errno) && (op_ret == -1)) {
-		local->op_ret   = -1;
-		local->op_errno = op_errno;
+        if (!child_went_down (op_ret, op_errno) && (op_ret == -1)) {
+                local->op_ret   = -1;
+                local->op_errno = op_errno;
 
-		afr_lk_unlock (frame, this);
-		return 0;
-	}
+                afr_lk_unlock (frame, this);
+                return 0;
+        }
 
-	if (op_ret == 0) {
-		local->op_ret        = 0;
-		local->op_errno      = 0;
-		local->cont.lk.locked_nodes[child_index] = 1;
+        if (op_ret == 0) {
+                local->op_ret        = 0;
+                local->op_errno      = 0;
+                local->cont.lk.locked_nodes[child_index] = 1;
                 local->cont.lk.ret_flock = *lock;
-	}
+        }
 
-	child_index++;
+        child_index++;
 
-	if (child_index < priv->child_count) {
-		STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) child_index,
-				   priv->children[child_index],
-				   priv->children[child_index]->fops->lk,
-				   local->fd, local->cont.lk.cmd,
-                                   &local->cont.lk.user_flock);
-	} else if (local->op_ret == -1) {
-		/* all nodes have gone down */
+        if (child_index < priv->child_count) {
+                STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) child_index,
+                                   priv->children[child_index],
+                                   priv->children[child_index]->fops->lk,
+                                   local->fd, local->cont.lk.cmd,
+                                   &local->cont.lk.user_flock, xdata);
+        } else if (local->op_ret == -1) {
+                /* all nodes have gone down */
 
                 AFR_STACK_UNWIND (lk, frame, -1, ENOTCONN,
-                                  &local->cont.lk.ret_flock);
-	} else {
-		/* locking has succeeded on all nodes that are up */
+                                  &local->cont.lk.ret_flock, NULL);
+        } else {
+                /* locking has succeeded on all nodes that are up */
 
-		AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
-                                  &local->cont.lk.ret_flock);
-	}
+                /* temporarily
+                   ret = afr_mark_locked_nodes (this, local->fd,
+                   local->cont.lk.locked_nodes);
+                   if (ret)
+                   gf_log (this->name, GF_LOG_DEBUG,
+                   "Could not save locked nodes info in fdctx");
 
-	return 0;
+                   ret = afr_save_locked_fd (this, local->fd);
+                   if (ret)
+                   gf_log (this->name, GF_LOG_DEBUG,
+                   "Could not save locked fd");
+
+                */
+                AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
+                                  &local->cont.lk.ret_flock, NULL);
+        }
+
+        return 0;
 }
 
 
 int
 afr_lk (call_frame_t *frame, xlator_t *this,
-	fd_t *fd, int32_t cmd,
-	struct flock *flock)
+        fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
-	afr_private_t *priv = NULL;
-	afr_local_t *local = NULL;
-
-	int i = 0;
-
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        int i = 0;
+        int32_t op_errno = 0;
+        int     ret      = -1;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	priv = this->private;
+        priv = this->private;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-	AFR_LOCAL_INIT (local, priv);
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	frame->local  = local;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	local->cont.lk.locked_nodes = GF_CALLOC (priv->child_count,
-					      sizeof (*local->cont.lk.locked_nodes),
-                                              gf_afr_mt_char);
+        local->cont.lk.locked_nodes = GF_CALLOC (priv->child_count,
+                                                 sizeof (*local->cont.lk.locked_nodes),
+                                                 gf_afr_mt_char);
 
-	if (!local->cont.lk.locked_nodes) {
-		gf_log (this->name, GF_LOG_ERROR, "Out of memory");
-		op_errno = ENOMEM;
-		goto out;
-	}
+        if (!local->cont.lk.locked_nodes) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	local->fd            = fd_ref (fd);
-	local->cont.lk.cmd   = cmd;
+        local->fd            = fd_ref (fd);
+        local->cont.lk.cmd   = cmd;
         local->cont.lk.user_flock = *flock;
         local->cont.lk.ret_flock = *flock;
 
-	STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) 0,
-			   priv->children[i],
-			   priv->children[i]->fops->lk,
-			   fd, cmd, flock);
+        STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) 0,
+                           priv->children[i],
+                           priv->children[i]->fops->lk,
+                           fd, cmd, flock, xdata);
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (lk, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
+        return 0;
+}
+
+int
+afr_forget (xlator_t *this, inode_t *inode)
+{
+        uint64_t        ctx_addr = 0;
+        afr_inode_ctx_t *ctx     = NULL;
+
+        inode_ctx_get (inode, this, &ctx_addr);
+
+        if (!ctx_addr)
+                goto out;
+
+        ctx = (afr_inode_ctx_t *)(long)ctx_addr;
+        GF_FREE (ctx->fresh_children);
+        GF_FREE (ctx);
+out:
+        return 0;
 }
 
 int
 afr_priv_dump (xlator_t *this)
 {
-	afr_private_t *priv = NULL;
+        afr_private_t *priv = NULL;
         char  key_prefix[GF_DUMP_MAX_BUF_LEN];
         char  key[GF_DUMP_MAX_BUF_LEN];
         int   i = 0;
 
 
-        assert(this);
-	priv = this->private;
+        GF_ASSERT (this);
+        priv = this->private;
 
-        assert(priv);
+        GF_ASSERT (priv);
         snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type, this->name);
         gf_proc_dump_add_section(key_prefix);
-        gf_proc_dump_build_key(key, key_prefix, "child_count");
-        gf_proc_dump_write(key, "%u", priv->child_count);
-        gf_proc_dump_build_key(key, key_prefix, "read_child_rr");
-        gf_proc_dump_write(key, "%u", priv->read_child_rr);
-	for (i = 0; i < priv->child_count; i++) {
-                gf_proc_dump_build_key(key, key_prefix, "child_up[%d]", i);
+        gf_proc_dump_write("child_count", "%u", priv->child_count);
+        gf_proc_dump_write("read_child_rr", "%u", priv->read_child_rr);
+        for (i = 0; i < priv->child_count; i++) {
+                sprintf (key, "child_up[%d]", i);
                 gf_proc_dump_write(key, "%d", priv->child_up[i]);
-                gf_proc_dump_build_key(key, key_prefix,
-                                        "pending_key[%d]", i);
+                sprintf (key, "pending_key[%d]", i);
                 gf_proc_dump_write(key, "%s", priv->pending_key[i]);
         }
-        gf_proc_dump_build_key(key, key_prefix, "data_self_heal");
-        gf_proc_dump_write(key, "%d", priv->data_self_heal);
-        gf_proc_dump_build_key(key, key_prefix, "metadata_self_heal");
-        gf_proc_dump_write(key, "%d", priv->metadata_self_heal);
-        gf_proc_dump_build_key(key, key_prefix, "entry_self_heal");
-        gf_proc_dump_write(key, "%d", priv->entry_self_heal);
-        gf_proc_dump_build_key(key, key_prefix, "data_change_log");
-        gf_proc_dump_write(key, "%d", priv->data_change_log);
-        gf_proc_dump_build_key(key, key_prefix, "metadata_change_log");
-        gf_proc_dump_write(key, "%d", priv->metadata_change_log);
-        gf_proc_dump_build_key(key, key_prefix, "entry_change_log");
-        gf_proc_dump_write(key, "%d", priv->entry_change_log);
-        gf_proc_dump_build_key(key, key_prefix, "read_child");
-        gf_proc_dump_write(key, "%d", priv->read_child);
-        gf_proc_dump_build_key(key, key_prefix, "favorite_child");
-        gf_proc_dump_write(key, "%u", priv->favorite_child);
-        gf_proc_dump_build_key(key, key_prefix, "data_lock_server_count");
-        gf_proc_dump_write(key, "%u", priv->data_lock_server_count);
-        gf_proc_dump_build_key(key, key_prefix, "metadata_lock_server_count");
-        gf_proc_dump_write(key, "%u", priv->metadata_lock_server_count);
-        gf_proc_dump_build_key(key, key_prefix, "entry_lock_server_count");
-        gf_proc_dump_write(key, "%u", priv->entry_lock_server_count);
-        gf_proc_dump_build_key(key, key_prefix, "wait_count");
-        gf_proc_dump_write(key, "%u", priv->wait_count);
+        gf_proc_dump_write("data_self_heal", "%s", priv->data_self_heal);
+        gf_proc_dump_write("metadata_self_heal", "%d", priv->metadata_self_heal);
+        gf_proc_dump_write("entry_self_heal", "%d", priv->entry_self_heal);
+        gf_proc_dump_write("data_change_log", "%d", priv->data_change_log);
+        gf_proc_dump_write("metadata_change_log", "%d", priv->metadata_change_log);
+        gf_proc_dump_write("entry-change_log", "%d", priv->entry_change_log);
+        gf_proc_dump_write("read_child", "%d", priv->read_child);
+        gf_proc_dump_write("favorite_child", "%d", priv->favorite_child);
+        gf_proc_dump_write("wait_count", "%u", priv->wait_count);
 
         return 0;
 }
@@ -2566,102 +3827,765 @@ afr_priv_dump (xlator_t *this)
 static int
 find_child_index (xlator_t *this, xlator_t *child)
 {
-	afr_private_t *priv = NULL;
-
-	int i = -1;
+        afr_private_t *priv = NULL;
+        int i = -1;
 
-	priv = this->private;
+        priv = this->private;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if ((xlator_t *) child == priv->children[i])
-			break;
-	}
+        for (i = 0; i < priv->child_count; i++) {
+                if ((xlator_t *) child == priv->children[i])
+                        break;
+        }
 
-	return i;
+        return i;
 }
 
 int32_t
 afr_notify (xlator_t *this, int32_t event,
-            void *data, ...)
+            void *data, void *data2)
 {
-	afr_private_t *     priv     = NULL;
-	unsigned char *     child_up = NULL;
-
-	int i           = -1;
-	int up_children = 0;
+        afr_private_t   *priv               = NULL;
+        int             i                   = -1;
+        int             up_children         = 0;
+        int             down_children       = 0;
+        int             propagate           = 0;
+        int             had_heard_from_all  = 0;
+        int             have_heard_from_all = 0;
+        int             idx                 = -1;
+        int             ret                 = -1;
+        int             call_psh            = 0;
+        int             up_child            = AFR_ALL_CHILDREN;
+        dict_t          *input              = NULL;
+        dict_t          *output             = NULL;
 
-	priv = this->private;
+        priv = this->private;
 
-	if (!priv)
-		return 0;
+        if (!priv)
+                return 0;
 
-	child_up = priv->child_up;
+        /*
+         * We need to reset this in case children come up in "staggered"
+         * fashion, so that we discover a late-arriving local subvolume.  Note
+         * that we could end up issuing N lookups to the first subvolume, and
+         * O(N^2) overall, but N is small for AFR so it shouldn't be an issue.
+         */
+        priv->did_discovery = _gf_false;
 
-	switch (event) {
-	case GF_EVENT_CHILD_UP:
-		i = find_child_index (this, data);
+        had_heard_from_all = 1;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!priv->last_event[i]) {
+                        had_heard_from_all = 0;
+                }
+        }
 
-		child_up[i] = 1;
+        /* parent xlators dont need to know about every child_up, child_down
+         * because of afr ha. If all subvolumes go down, child_down has
+         * to be triggered. In that state when 1 subvolume comes up child_up
+         * needs to be triggered. dht optimizes revalidate lookup by sending
+         * it only to one of its subvolumes. When child up/down happens
+         * for afr's subvolumes dht should be notified by child_modified. The
+         * subsequent revalidate lookup happens on all the dht's subvolumes
+         * which triggers afr self-heals if any.
+         */
+        idx = find_child_index (this, data);
+        if (idx < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Received child_up "
+                        "from invalid subvolume");
+                goto out;
+        }
 
+        switch (event) {
+        case GF_EVENT_CHILD_UP:
                 LOCK (&priv->lock);
                 {
-                        priv->up_count++;
+                        /*
+                         * This only really counts if the child was never up
+                         * (value = -1) or had been down (value = 0).  See
+                         * comment at GF_EVENT_CHILD_DOWN for a more detailed
+                         * explanation.
+                         */
+                        if (priv->child_up[idx] != 1) {
+                                priv->up_count++;
+                        }
+                        priv->child_up[idx] = 1;
+
+                        call_psh = 1;
+                        up_child = idx;
+                        for (i = 0; i < priv->child_count; i++)
+                                if (priv->child_up[i] == 1)
+                                        up_children++;
+                        if (up_children == 1) {
+                                gf_log (this->name, GF_LOG_INFO,
+                                        "Subvolume '%s' came back up; "
+                                        "going online.", ((xlator_t *)data)->name);
+                        } else {
+                                event = GF_EVENT_CHILD_MODIFIED;
+                        }
+
+                        priv->last_event[idx] = event;
                 }
                 UNLOCK (&priv->lock);
 
-		/*
-		   if all the children were down, and one child came up,
-		   send notify to parent
-		*/
+                break;
 
-		for (i = 0; i < priv->child_count; i++)
-			if (child_up[i])
-				up_children++;
+        case GF_EVENT_CHILD_DOWN:
+                LOCK (&priv->lock);
+                {
+                        /*
+                         * If a brick is down when we start, we'll get a
+                         * CHILD_DOWN to indicate its initial state.  There
+                         * was never a CHILD_UP in this case, so if we
+                         * increment "down_count" the difference between than
+                         * and "up_count" will no longer be the number of
+                         * children that are currently up.  This has serious
+                         * implications e.g. for quorum enforcement, so we
+                         * don't increment these values unless the event
+                         * represents an actual state transition between "up"
+                         * (value = 1) and anything else.
+                         */
+                        if (priv->child_up[idx] == 1) {
+                                priv->down_count++;
+                        }
+                        priv->child_up[idx] = 0;
+
+                        for (i = 0; i < priv->child_count; i++)
+                                if (priv->child_up[i] == 0)
+                                        down_children++;
+                        if (down_children == priv->child_count) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "All subvolumes are down. Going offline "
+                                        "until atleast one of them comes back up.");
+                        } else {
+                                event = GF_EVENT_CHILD_MODIFIED;
+                        }
 
-		if (up_children == 1) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "Subvolume '%s' came back up; "
-                                "going online.", ((xlator_t *)data)->name);
+                        priv->last_event[idx] = event;
+                }
+                UNLOCK (&priv->lock);
+
+                break;
 
-			default_notify (this, event, data);
+        case GF_EVENT_CHILD_CONNECTING:
+                LOCK (&priv->lock);
+                {
+                        priv->last_event[idx] = event;
                 }
+                UNLOCK (&priv->lock);
+
+                break;
 
-		break;
+        case GF_EVENT_TRANSLATOR_OP:
+                input = data;
+                output = data2;
+                ret = afr_xl_op (this, input, output);
+                goto out;
+                break;
+
+        default:
+                propagate = 1;
+                break;
+        }
 
-	case GF_EVENT_CHILD_DOWN:
-		i = find_child_index (this, data);
+        /* have all subvolumes reported status once by now? */
+        have_heard_from_all = 1;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!priv->last_event[i])
+                        have_heard_from_all = 0;
+        }
 
-		child_up[i] = 0;
+        /* if all subvols have reported status, no need to hide anything
+           or wait for anything else. Just propagate blindly */
+        if (have_heard_from_all)
+                propagate = 1;
+
+        if (!had_heard_from_all && have_heard_from_all) {
+                /* This is the first event which completes aggregation
+                   of events from all subvolumes. If at least one subvol
+                   had come up, propagate CHILD_UP, but only this time
+                */
+                event = GF_EVENT_CHILD_DOWN;
 
                 LOCK (&priv->lock);
                 {
-                        priv->down_count++;
+                        up_children = afr_up_children_count (priv->child_up,
+                                                             priv->child_count);
+                        for (i = 0; i < priv->child_count; i++) {
+                                if (priv->last_event[i] == GF_EVENT_CHILD_UP) {
+                                        event = GF_EVENT_CHILD_UP;
+                                        break;
+                                }
+
+                                if (priv->last_event[i] ==
+                                                GF_EVENT_CHILD_CONNECTING) {
+                                        event = GF_EVENT_CHILD_CONNECTING;
+                                        /* continue to check other events for CHILD_UP */
+                                }
+                        }
                 }
                 UNLOCK (&priv->lock);
+        }
 
-		/*
-		   if all children are down, and this was the last to go down,
-		   send notify to parent
-		*/
+        ret = 0;
+        if (propagate)
+                ret = default_notify (this, event, data);
+        if (call_psh && priv->shd.iamshd)
+                afr_proactive_self_heal ((void*) (long) up_child);
 
-		for (i = 0; i < priv->child_count; i++)
-			if (child_up[i])
-				up_children++;
+out:
+        return ret;
+}
 
-		if (up_children == 0) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "All subvolumes are down. Going offline "
-                                "until atleast one of them comes back up.");
+int
+afr_first_up_child (unsigned char *child_up, size_t child_count)
+{
+        int         ret      = -1;
+        int         i        = 0;
+
+        GF_ASSERT (child_up);
 
-			default_notify (this, event, data);
+        for (i = 0; i < child_count; i++) {
+                if (child_up[i]) {
+                        ret = i;
+                        break;
                 }
+        }
 
-		break;
+        return ret;
+}
 
-	default:
-		default_notify (this, event, data);
-	}
+int
+afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno)
+{
+        int     ret = -1;
+
+        local->op_ret = -1;
+        local->op_errno = EUCLEAN;
+
+        local->child_up = GF_CALLOC (priv->child_count,
+                                     sizeof (*local->child_up),
+                                     gf_afr_mt_char);
+        if (!local->child_up) {
+                if (op_errno)
+                        *op_errno = ENOMEM;
+                goto out;
+        }
 
-	return 0;
+        memcpy (local->child_up, priv->child_up,
+                sizeof (*local->child_up) * priv->child_count);
+        local->call_count = afr_up_children_count (local->child_up,
+                                                   priv->child_count);
+        if (local->call_count == 0) {
+                gf_log (THIS->name, GF_LOG_INFO, "no subvolumes up");
+                if (op_errno)
+                        *op_errno = ENOTCONN;
+                goto out;
+        }
+
+        local->child_errno = GF_CALLOC (priv->child_count,
+                                        sizeof (*local->child_errno),
+                                        gf_afr_mt_int32_t);
+        if (!local->child_errno) {
+                if (op_errno)
+                        *op_errno = ENOMEM;
+                goto out;
+        }
+
+        local->transaction.postop_piggybacked = GF_CALLOC (priv->child_count,
+							   sizeof (int),
+							   gf_afr_mt_int32_t);
+        if (!local->transaction.postop_piggybacked) {
+                if (op_errno)
+                        *op_errno = ENOMEM;
+                goto out;
+        }
+
+	local->append_write = _gf_false;
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count,
+                        transaction_lk_type_t lk_type)
+{
+        int             ret = -ENOMEM;
+
+        lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
+                                      child_count, gf_afr_mt_char);
+        if (NULL == lk->locked_nodes)
+                goto out;
+
+        lk->lower_locked_nodes = GF_CALLOC (sizeof (*lk->lower_locked_nodes),
+                                            child_count, gf_afr_mt_char);
+        if (NULL == lk->lower_locked_nodes)
+                goto out;
+
+        lk->lock_op_ret   = -1;
+        lk->lock_op_errno = EUCLEAN;
+        lk->transaction_lk_type = lk_type;
+
+        ret = 0;
+out:
+        return ret;
+}
+
+void
+afr_matrix_cleanup (int32_t **matrix, unsigned int m)
+{
+        int             i         = 0;
+
+        if (!matrix)
+                goto out;
+        for (i = 0; i < m; i++) {
+                GF_FREE (matrix[i]);
+        }
+
+        GF_FREE (matrix);
+out:
+        return;
+}
+
+int32_t**
+afr_matrix_create (unsigned int m, unsigned int n)
+{
+        int32_t         **matrix = NULL;
+        int             i       = 0;
+
+        matrix = GF_CALLOC (sizeof (*matrix), m, gf_afr_mt_int32_t);
+        if (!matrix)
+                goto out;
+
+        for (i = 0; i < m; i++) {
+                matrix[i] = GF_CALLOC (sizeof (*matrix[i]), n,
+                                       gf_afr_mt_int32_t);
+                if (!matrix[i])
+                        goto out;
+        }
+        return matrix;
+out:
+        afr_matrix_cleanup (matrix, m);
+        return NULL;
+}
+
+int
+afr_inodelk_init (afr_inodelk_t *lk, char *dom, size_t child_count)
+{
+        int             ret = -ENOMEM;
+
+        lk->domain = dom;
+        lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
+                                      child_count, gf_afr_mt_char);
+        if (NULL == lk->locked_nodes)
+                goto out;
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+afr_transaction_local_init (afr_local_t *local, xlator_t *this)
+{
+        int            child_up_count = 0;
+        int            ret = -ENOMEM;
+        afr_private_t *priv = NULL;
+
+        priv = this->private;
+        ret = afr_internal_lock_init (&local->internal_lock, priv->child_count,
+                                      AFR_TRANSACTION_LK);
+        if (ret < 0)
+                goto out;
+
+        if ((local->transaction.type == AFR_DATA_TRANSACTION) ||
+            (local->transaction.type == AFR_METADATA_TRANSACTION)) {
+                ret = afr_inodelk_init (&local->internal_lock.inodelk[0],
+                                        this->name, priv->child_count);
+                if (ret < 0)
+                        goto out;
+        }
+
+        ret = -ENOMEM;
+        child_up_count = afr_up_children_count (local->child_up,
+                                                priv->child_count);
+        if (priv->optimistic_change_log && child_up_count == priv->child_count)
+                local->optimistic_change_log = 1;
+
+        local->first_up_child = afr_first_up_child (local->child_up,
+                                                    priv->child_count);
+
+        local->transaction.eager_lock =
+                GF_CALLOC (sizeof (*local->transaction.eager_lock),
+                           priv->child_count,
+                           gf_afr_mt_int32_t);
+
+        if (!local->transaction.eager_lock)
+                goto out;
+
+        local->fresh_children = afr_children_create (priv->child_count);
+        if (!local->fresh_children)
+                goto out;
+
+        local->transaction.pre_op = GF_CALLOC (sizeof (*local->transaction.pre_op),
+                                               priv->child_count,
+                                               gf_afr_mt_char);
+        if (!local->transaction.pre_op)
+                goto out;
+
+        local->pending = afr_matrix_create (priv->child_count,
+                                            AFR_NUM_CHANGE_LOGS);
+        if (!local->pending)
+                goto out;
+
+        local->transaction.txn_changelog = afr_matrix_create (priv->child_count,
+                                                           AFR_NUM_CHANGE_LOGS);
+        if (!local->transaction.txn_changelog)
+                goto out;
+
+	INIT_LIST_HEAD (&local->transaction.eager_locked);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+void
+afr_reset_children (int32_t *fresh_children, int32_t child_count)
+{
+        unsigned int i = 0;
+        for (i = 0; i < child_count; i++)
+                fresh_children[i] = -1;
+}
+
+int32_t*
+afr_children_create (int32_t child_count)
+{
+        int32_t           *children = NULL;
+        int               i               = 0;
+
+        GF_ASSERT (child_count > 0);
+
+        children = GF_CALLOC (child_count, sizeof (*children),
+                              gf_afr_mt_int32_t);
+        if (NULL == children)
+                goto out;
+        for (i = 0; i < child_count; i++)
+                children[i] = -1;
+out:
+        return children;
+}
+
+void
+afr_children_add_child (int32_t *children, int32_t child,
+                        int32_t child_count)
+{
+        gf_boolean_t child_found = _gf_false;
+        int          i               = 0;
+
+        for (i = 0; i < child_count; i++) {
+                if (children[i] == -1)
+                        break;
+                if (children[i] == child) {
+                        child_found = _gf_true;
+                        break;
+                }
+        }
+
+        if (!child_found) {
+                GF_ASSERT (i < child_count);
+                children[i] = child;
+        }
+}
+
+void
+afr_children_rm_child (int32_t *children, int32_t child, int32_t child_count)
+{
+        int          i = 0;
+
+        GF_ASSERT ((child >= 0) && (child < child_count));
+        for (i = 0; i < child_count; i++) {
+                if (children[i] == -1)
+                        break;
+                if (children[i] == child) {
+                        if (i != (child_count - 1))
+                                memmove (children + i, children + i + 1,
+                                         sizeof (*children)*(child_count - i - 1));
+                        children[child_count - 1] = -1;
+                        break;
+                }
+        }
+}
+
+int
+afr_get_children_count (int32_t *children, unsigned int child_count)
+{
+        int count = 0;
+        int i = 0;
+
+        for (i = 0; i < child_count; i++) {
+                if (children[i] == -1)
+                        break;
+                count++;
+        }
+        return count;
+}
+
+void
+afr_set_low_priority (call_frame_t *frame)
+{
+        frame->root->pid = LOW_PRIO_PROC_PID;
+}
+
+int
+afr_child_fd_ctx_set (xlator_t *this, fd_t *fd, int32_t child,
+                      int flags)
+{
+        int             ret = 0;
+        uint64_t        ctx = 0;
+        afr_fd_ctx_t    *fd_ctx      = NULL;
+
+        GF_ASSERT (fd && fd->inode);
+        ret = afr_fd_ctx_set (this, fd);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "could not set fd ctx for fd=%p", fd);
+                goto out;
+        }
+
+        ret = fd_ctx_get (fd, this, &ctx);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "could not get fd ctx for fd=%p", fd);
+                goto out;
+        }
+
+        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        fd_ctx->opened_on[child] = AFR_FD_OPENED;
+        if (!IA_ISDIR (fd->inode->ia_type)) {
+                fd_ctx->flags            = flags;
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+gf_boolean_t
+afr_have_quorum (char *logname, afr_private_t *priv)
+{
+        unsigned int        quorum = 0;
+
+        GF_VALIDATE_OR_GOTO(logname,priv,out);
+
+        quorum = priv->quorum_count;
+        if (quorum != AFR_QUORUM_AUTO) {
+                return (priv->up_count >= (priv->down_count + quorum));
+        }
+
+        quorum = priv->child_count / 2 + 1;
+        if (priv->up_count >= (priv->down_count + quorum)) {
+                return _gf_true;
+        }
+
+        /*
+         * Special case for even numbers of nodes: if we have exactly half
+         * and that includes the first ("senior-most") node, then that counts
+         * as quorum even if it wouldn't otherwise.  This supports e.g. N=2
+         * while preserving the critical property that there can only be one
+         * such group.
+         */
+        if ((priv->child_count % 2) == 0) {
+                quorum = priv->child_count / 2;
+                if (priv->up_count >= (priv->down_count + quorum)) {
+                        if (priv->child_up[0]) {
+                                return _gf_true;
+                        }
+                }
+        }
+
+out:
+        return _gf_false;
+}
+
+void
+afr_priv_destroy (afr_private_t *priv)
+{
+        int            i           = 0;
+
+        if (!priv)
+                goto out;
+        inode_unref (priv->root_inode);
+        GF_FREE (priv->shd.pos);
+        GF_FREE (priv->shd.pending);
+        GF_FREE (priv->shd.inprogress);
+//        for (i = 0; i < priv->child_count; i++)
+//                if (priv->shd.timer && priv->shd.timer[i])
+//                        gf_timer_call_cancel (this->ctx, priv->shd.timer[i]);
+        GF_FREE (priv->shd.timer);
+
+        if (priv->shd.healed)
+                eh_destroy (priv->shd.healed);
+
+        if (priv->shd.heal_failed)
+                eh_destroy (priv->shd.heal_failed);
+
+        if (priv->shd.split_brain)
+                eh_destroy (priv->shd.split_brain);
+
+        for (i = 0; i < priv->child_count; i++)
+        {
+                if (priv->shd.statistics[i])
+                        eh_destroy (priv->shd.statistics[i]);
+        }
+
+        GF_FREE (priv->shd.statistics);
+
+        GF_FREE (priv->shd.crawl_events);
+
+        GF_FREE (priv->last_event);
+        if (priv->pending_key) {
+                for (i = 0; i < priv->child_count; i++)
+                        GF_FREE (priv->pending_key[i]);
+        }
+        GF_FREE (priv->pending_key);
+        GF_FREE (priv->children);
+        GF_FREE (priv->child_up);
+        LOCK_DESTROY (&priv->lock);
+        LOCK_DESTROY (&priv->read_child_lock);
+        pthread_mutex_destroy (&priv->mutex);
+        GF_FREE (priv);
+out:
+        return;
+}
+
+int
+xlator_subvolume_count (xlator_t *this)
+{
+        int i = 0;
+        xlator_list_t *list = NULL;
+
+        for (list = this->children; list; list = list->next)
+                i++;
+        return i;
+}
+
+inline gf_boolean_t
+afr_is_errno_set (int *child_errno, int child)
+{
+        return child_errno[child];
+}
+
+inline gf_boolean_t
+afr_is_errno_unset (int *child_errno, int child)
+{
+        return !afr_is_errno_set (child_errno, child);
+}
+
+void
+afr_prepare_new_entry_pending_matrix (int32_t **pending,
+                                      gf_boolean_t (*is_pending) (int *, int),
+                                      int *ctx, struct iatt *buf,
+                                      unsigned int child_count)
+{
+        int midx = 0;
+        int idx  = 0;
+        int i    = 0;
+
+        midx = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
+        if (IA_ISDIR (buf->ia_type))
+                idx = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
+        else if (IA_ISREG (buf->ia_type))
+                idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+        else
+                idx = -1;
+        for (i = 0; i < child_count; i++) {
+                if (is_pending (ctx, i)) {
+                        pending[i][midx] = hton32 (1);
+                        if (idx == -1)
+                                continue;
+                        pending[i][idx] = hton32 (1);
+                }
+        }
+}
+
+gf_boolean_t
+afr_is_fd_fixable (fd_t *fd)
+{
+        if (!fd || !fd->inode)
+                return _gf_false;
+        else if (fd_is_anonymous (fd))
+                return _gf_false;
+        else if (uuid_is_null (fd->inode->gfid))
+                return _gf_false;
+
+        return _gf_true;
+}
+
+void
+afr_handle_open_fd_count (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local = NULL;
+        inode_t         *inode = NULL;
+        afr_inode_ctx_t *ctx   = NULL;
+
+        local = frame->local;
+
+        if (local->fd)
+                inode = local->fd->inode;
+        else
+                inode = local->loc.inode;
+
+        if (!inode)
+                return;
+
+        LOCK (&inode->lock);
+        {
+                ctx = __afr_inode_ctx_get (inode, this);
+                ctx->open_fd_count = local->open_fd_count;
+        }
+        UNLOCK (&inode->lock);
+}
+
+int
+afr_initialise_statistics (xlator_t *this)
+{
+        afr_private_t       *priv = NULL;
+        int                 ret = -1;
+        int                 i = 0;
+        int                 child_count = 0;
+        eh_t                *stats_per_brick = NULL;
+        shd_crawl_event_t   ***shd_crawl_events = NULL;
+        priv = this->private;
+
+        priv->shd.statistics = GF_CALLOC (sizeof(eh_t *), priv->child_count,
+                                          gf_common_mt_eh_t);
+        if (!priv->shd.statistics) {
+                ret = -1;
+                goto out;
+        }
+        child_count = priv->child_count;
+        for (i=0; i < child_count ; i++) {
+                stats_per_brick = eh_new (AFR_STATISTICS_HISTORY_SIZE,
+                                          _gf_false,
+                                          _destroy_crawl_event_data);
+                if (!stats_per_brick) {
+                        ret = -1;
+                        goto out;
+                }
+                priv->shd.statistics[i] = stats_per_brick;
+
+        }
+
+        shd_crawl_events = (shd_crawl_event_t***)(&priv->shd.crawl_events);
+        *shd_crawl_events  = GF_CALLOC (sizeof(shd_crawl_event_t*),
+                                        priv->child_count,
+                                        gf_afr_mt_shd_crawl_event_t);
+
+        if (!priv->shd.crawl_events) {
+                ret = -1;
+                goto out;
+        }
+        ret = 0;
+out:
+        return ret;
 
 }
diff --git a/xlators/cluster/afr/src/afr-dir-read.c b/xlators/cluster/afr/src/afr-dir-read.c
index 1f29cc006..689dd84e6 100644
--- a/xlators/cluster/afr/src/afr-dir-read.c
+++ b/xlators/cluster/afr/src/afr-dir-read.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -49,9 +40,9 @@
 #include "afr-self-heal.h"
 #include "afr-self-heal-common.h"
 
-
 int
-afr_examine_dir_sh_unwind (call_frame_t *frame, xlator_t *this)
+afr_examine_dir_sh_unwind (call_frame_t *frame, xlator_t *this, int32_t op_ret,
+                           int32_t op_errno, int32_t sh_failed)
 {
         afr_local_t *local  = NULL;
 
@@ -60,7 +51,7 @@ afr_examine_dir_sh_unwind (call_frame_t *frame, xlator_t *this)
         afr_set_opendir_done (this, local->fd->inode);
 
         AFR_STACK_UNWIND (opendir, frame, local->op_ret,
-                          local->op_errno, local->fd);
+                          local->op_errno, local->fd, NULL);
 
         return 0;
 }
@@ -70,16 +61,19 @@ gf_boolean_t
 __checksums_differ (uint32_t *checksum, int child_count,
                     unsigned char *child_up)
 {
-        int ret = _gf_false;
-        int i = 0;
-
-        uint32_t cksum;
-
-        cksum = checksum[0];
+        int          ret            = _gf_false;
+        int          i              = 0;
+        uint32_t     cksum          = 0;
+        gf_boolean_t activate_check = _gf_false;
 
         for (i = 0; i < child_count; i++) {
                 if (!child_up[i])
                         continue;
+                if (_gf_false == activate_check) {
+                        cksum          = checksum[i];
+                        activate_check = _gf_true;
+                        continue;
+                }
 
                 if (cksum != checksum[i]) {
                         ret = _gf_true;
@@ -96,40 +90,45 @@ __checksums_differ (uint32_t *checksum, int child_count,
 int32_t
 afr_examine_dir_readdir_cbk (call_frame_t *frame, void *cookie,
                              xlator_t *this, int32_t op_ret, int32_t op_errno,
-                             gf_dirent_t *entries)
+                             gf_dirent_t *entries, dict_t *xdata)
 {
-	afr_private_t *    priv  = NULL;
-	afr_local_t *      local = NULL;
-        afr_self_heal_t *  sh    = NULL;
-
-        gf_dirent_t * entry = NULL;
-        gf_dirent_t * tmp   = NULL;
-
-        int child_index = 0;
-
-        uint32_t entry_cksum;
-
-        int   call_count    = 0;
-        off_t last_offset = 0;
-        char  sh_type_str[256] = {0,};
+        afr_private_t *   priv        = NULL;
+        afr_local_t *     local       = NULL;
+        afr_self_heal_t * sh          = NULL;
+        gf_dirent_t *     entry       = NULL;
+        gf_dirent_t *     tmp         = NULL;
+        char              *reason     = NULL;
+        int               child_index = 0;
+        uint32_t          entry_cksum = 0;
+        int               call_count  = 0;
+        off_t             last_offset = 0;
+        inode_t           *inode      = NULL;
 
         priv  = this->private;
         local = frame->local;
         sh    = &local->self_heal;
+        inode = local->fd->inode;
 
         child_index = (long) cookie;
 
         if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "%s: failed to do opendir on %s",
+                        local->loc.path, priv->children[child_index]->name);
                 local->op_ret = -1;
                 local->op_ret = op_errno;
                 goto out;
         }
 
-        if (op_ret == 0)
+        if (op_ret == 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "%s: no entries found in %s",
+                        local->loc.path, priv->children[child_index]->name);
                 goto out;
+        }
 
         list_for_each_entry_safe (entry, tmp, &entries->list, list) {
-                entry_cksum = gf_rsync_weak_checksum (entry->d_name,
+                entry_cksum = gf_rsync_weak_checksum ((unsigned char *)entry->d_name,
                                                       strlen (entry->d_name));
                 local->cont.opendir.checksum[child_index] ^= entry_cksum;
         }
@@ -144,39 +143,30 @@ afr_examine_dir_readdir_cbk (call_frame_t *frame, void *cookie,
                            (void *) (long) child_index,
                            priv->children[child_index],
                            priv->children[child_index]->fops->readdir,
-                           local->fd, 131072, last_offset);
+                           local->fd, 131072, last_offset, NULL);
+
+        return 0;
 
 out:
-        if ((op_ret == 0) || (op_ret == -1)) {
-                call_count = afr_frame_return (frame);
-
-                if (call_count == 0) {
-                        if (__checksums_differ (local->cont.opendir.checksum,
-                                                priv->child_count,
-                                                local->child_up)) {
-
-                                sh->need_entry_self_heal  = _gf_true;
-                                sh->forced_merge          = _gf_true;
-                                sh->type                  = local->fd->inode->ia_type;
-                                sh->background            = _gf_false;
-                                sh->unwind                = afr_examine_dir_sh_unwind;
-
-                                afr_self_heal_type_str_get(&local->self_heal,
-                                                           sh_type_str,
-                                                           sizeof(sh_type_str));
-                                gf_log (this->name, GF_LOG_NORMAL,
-                                        "%s self-heal triggered. path: %s, "
-                                         "reason: checksums of directory differ,"
-                                        " forced merge option set",
-                                        sh_type_str, local->loc.path);
-
-                                afr_self_heal (frame, this);
-                        } else {
-                                afr_set_opendir_done (this, local->fd->inode);
-
-                                AFR_STACK_UNWIND (opendir, frame, local->op_ret,
-                                                  local->op_errno, local->fd);
-                        }
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+                if (__checksums_differ (local->cont.opendir.checksum,
+                                        priv->child_count,
+                                        local->child_up)) {
+
+                        sh->do_entry_self_heal  = _gf_true;
+                        sh->forced_merge          = _gf_true;
+
+                        reason = "checksums of directory differ";
+                        afr_launch_self_heal (frame, this, inode, _gf_false,
+                                              inode->ia_type, reason, NULL,
+                                              afr_examine_dir_sh_unwind);
+                } else {
+                        afr_set_opendir_done (this, inode);
+
+                        AFR_STACK_UNWIND (opendir, frame, local->op_ret,
+                                          local->op_errno, local->fd, NULL);
                 }
         }
 
@@ -187,20 +177,19 @@ out:
 int
 afr_examine_dir (call_frame_t *frame, xlator_t *this)
 {
-        afr_private_t * priv  = NULL;
-        afr_local_t *   local = NULL;
-
-        int i;
-        int call_count = 0;
+        afr_private_t * priv       = NULL;
+        afr_local_t *   local      = NULL;
+        int             i          = 0;
+        int             call_count = 0;
 
         local = frame->local;
         priv  = this->private;
 
         local->cont.opendir.checksum = GF_CALLOC (priv->child_count,
-                                        sizeof (*local->cont.opendir.checksum),
-                                        gf_afr_mt_int32_t);
+                                                  sizeof (*local->cont.opendir.checksum),
+                                                  gf_afr_mt_int32_t);
 
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
+        call_count = afr_up_children_count (local->child_up, priv->child_count);
 
         local->call_count = call_count;
 
@@ -210,7 +199,7 @@ afr_examine_dir (call_frame_t *frame, xlator_t *this)
                                            (void *) (long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->readdir,
-                                           local->fd, 131072, 0);
+                                           local->fd, 131072, 0, NULL);
 
                         if (!--call_count)
                                 break;
@@ -223,132 +212,139 @@ afr_examine_dir (call_frame_t *frame, xlator_t *this)
 
 int32_t
 afr_opendir_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno,
-		 fd_t *fd)
+                 xlator_t *this, int32_t op_ret, int32_t op_errno,
+                 fd_t *fd, dict_t *xdata)
 {
         afr_private_t *priv              = NULL;
-	afr_local_t   *local             = NULL;
+        afr_local_t   *local             = NULL;
         int32_t        up_children_count = 0;
-
-	int call_count = -1;
+        int            ret               = -1;
+        int            call_count        = -1;
+        int32_t        child_index       = 0;
 
         priv  = this->private;
         local = frame->local;
+        child_index = (long) cookie;
 
-        up_children_count = afr_up_children_count (priv->child_count,
-                                                   local->child_up);
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret >= 0)
-			local->op_ret = op_ret;
+        up_children_count = afr_up_children_count (local->child_up,
+                                                   priv->child_count);
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret >= 0) {
+                        local->op_ret = op_ret;
+                        ret = afr_child_fd_ctx_set (this, fd, child_index, 0);
+                        if (ret) {
+                                local->op_ret = -1;
+                                local->op_errno = -ret;
+                                goto unlock;
+                        }
+                }
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                local->op_errno = op_errno;
+        }
+unlock:
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-                if (local->op_ret == 0) {
-                        afr_fd_ctx_set (this, local->fd);
+        if (call_count == 0) {
+                if (local->op_ret != 0)
+                        goto out;
 
-                        if (!afr_is_opendir_done (this, local->fd->inode) &&
-                            up_children_count > 1) {
+                if (!afr_is_opendir_done (this, local->fd->inode) &&
+                    up_children_count > 1 && priv->entry_self_heal) {
 
-                                /*
-                                 * This is the first opendir on this inode. We need
-                                 * to check if the directory's entries are the same
-                                 * on all subvolumes. This is needed in addition
-                                 * to regular entry self-heal because the readdir
-                                 * call is sent only to the first subvolume, and
-                                 * thus files that exist only there will never be healed
-                                 * otherwise (assuming changelog shows no anamolies).
-                                 */
+                        /*
+                         * This is the first opendir on this inode. We need
+                         * to check if the directory's entries are the same
+                         * on all subvolumes. This is needed in addition
+                         * to regular entry self-heal because the readdir
+                         * call is sent only to the first subvolume, and
+                         * thus files that exist only there will never be healed
+                         * otherwise (assuming changelog shows no anomalies).
+                         */
 
-                                gf_log (this->name, GF_LOG_TRACE,
-                                        "reading contents of directory %s looking for mismatch",
-                                        local->loc.path);
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "reading contents of directory %s looking for mismatch",
+                                local->loc.path);
 
-                                afr_examine_dir (frame, this);
+                        afr_examine_dir (frame, this);
 
-                        } else {
-                                AFR_STACK_UNWIND (opendir, frame, local->op_ret,
-                                                  local->op_errno, local->fd);
-                        }
                 } else {
-                        AFR_STACK_UNWIND (opendir, frame, local->op_ret,
-                                          local->op_errno, local->fd);
+                        /* do the unwind */
+                        goto out;
                 }
         }
 
-	return 0;
+        return 0;
+
+out:
+        AFR_STACK_UNWIND (opendir, frame, local->op_ret,
+                          local->op_errno, local->fd, NULL);
+
+        return 0;
 }
 
 
-int32_t 
+int32_t
 afr_opendir (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, fd_t *fd)
+             loc_t *loc, fd_t *fd)
 {
-	afr_private_t * priv        = NULL;
-	afr_local_t   * local       = NULL;
-
-	int             child_count = 0;
-	int             i           = 0;
-
-	int ret = -1;
-	int call_count = -1;
+        afr_private_t * priv        = NULL;
+        afr_local_t   * local       = NULL;
+        int             child_count = 0;
+        int             i           = 0;
+        int             ret         = -1;
+        int             call_count  = -1;
+        int32_t         op_errno    = 0;
 
-	int32_t         op_ret   = -1;
-	int32_t         op_errno = 0;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        priv = this->private;
 
-	priv = this->private;
+        child_count = priv->child_count;
 
-	child_count = priv->child_count;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
         loc_copy (&local->loc, loc);
 
-	frame->local = local;
-	local->fd    = fd_ref (fd);
+        local->fd    = fd_ref (fd);
 
-	call_count = local->call_count;
-	
-	for (i = 0; i < child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_opendir_cbk, 
-				    priv->children[i],
-				    priv->children[i]->fops->opendir,
-				    loc, fd);
+        call_count = local->call_count;
+
+        for (i = 0; i < child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, afr_opendir_cbk,
+                                           (void*) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->opendir,
+                                           loc, fd, NULL);
 
-			if (!--call_count)
-				break;
-		}
-	}
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (opendir, frame, op_ret, op_errno, fd);
-	}
+        if (ret < 0)
+                AFR_STACK_UNWIND (opendir, frame, -1, op_errno, fd, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 /**
  * Common algorithm for directory read calls:
- * 
+ *
  * - Try the fop on the first child that is up
  * - if we have failed due to ENOTCONN:
  *     try the next child
@@ -362,373 +358,179 @@ struct entry_name {
         struct list_head list;
 };
 
-
-static gf_boolean_t
-remembered_name (const char *name, struct list_head *entries)
-{
-        struct entry_name *e;
-        gf_boolean_t ret = _gf_false;
-
-        list_for_each_entry (e, entries, list) {
-                if (!strcmp (name, e->name)) {
-                        ret = _gf_true;
-                        goto out;
-                }
-        }
-
-out:
-        return ret;
-}
-
-
 static void
-afr_remember_entries (gf_dirent_t *entries, fd_t *fd)
-{
-	struct entry_name *n       = NULL;
-	gf_dirent_t *      entry   = NULL;
-
-	int ret = 0;
-
-	uint64_t      ctx;
-	afr_fd_ctx_t *fd_ctx;
-
-	ret = fd_ctx_get (fd, THIS, &ctx);
-	if (ret < 0) {
-		gf_log (THIS->name, GF_LOG_DEBUG,
-			"could not get fd ctx for fd=%p", fd);
-		return;
-	}
-
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-	list_for_each_entry (entry, &entries->list, list) {
-		n = GF_CALLOC (1, sizeof (*n), gf_afr_mt_entry_name);
-		n->name = gf_strdup (entry->d_name);
-		INIT_LIST_HEAD (&n->list);
-
-		list_add (&n->list, &fd_ctx->entries);
-	}
-}
-
-
-static off_t
-afr_filter_entries (gf_dirent_t *entries, fd_t *fd)
+afr_forget_entries (fd_t *fd)
 {
-	gf_dirent_t *entry, *tmp;
-	int ret = 0;
-
-	uint64_t      ctx;
-	afr_fd_ctx_t *fd_ctx;
-
-	off_t offset = 0;
-
-	ret = fd_ctx_get (fd, THIS, &ctx);
-	if (ret < 0) {
-		gf_log (THIS->name, GF_LOG_DEBUG,
-			"could not get fd ctx for fd=%p", fd);
-		return -1;
-	}
+        struct entry_name *entry  = NULL;
+        struct entry_name *tmp    = NULL;
+        int                ret    = 0;
+        uint64_t           ctx    = 0;
+        afr_fd_ctx_t      *fd_ctx = NULL;
+
+        ret = fd_ctx_get (fd, THIS, &ctx);
+        if (ret < 0) {
+                gf_log (THIS->name, GF_LOG_INFO,
+                        "could not get fd ctx for fd=%p", fd);
+                return;
+        }
 
         fd_ctx = (afr_fd_ctx_t *)(long) ctx;
 
-	list_for_each_entry_safe (entry, tmp, &entries->list, list) {
-		offset = entry->d_off;
-
-		if (remembered_name (entry->d_name, &fd_ctx->entries)) {
-			list_del (&entry->list);
-			GF_FREE (entry);
-		}
-	}
-
-	return offset;
+        list_for_each_entry_safe (entry, tmp, &fd_ctx->entries, list) {
+                GF_FREE (entry->name);
+                list_del (&entry->list);
+                GF_FREE (entry);
+        }
 }
 
-
 static void
-afr_forget_entries (fd_t *fd)
+afr_readdir_filter_trash_dir (gf_dirent_t *entries, fd_t *fd)
 {
-	struct entry_name *entry, *tmp;
-	int ret = 0;
-
-	uint64_t      ctx;
-	afr_fd_ctx_t *fd_ctx;
-
-	ret = fd_ctx_get (fd, THIS, &ctx);
-	if (ret < 0) {
-		gf_log (THIS->name, GF_LOG_DEBUG,
-			"could not get fd ctx for fd=%p", fd);
-		return;
-	}
+        gf_dirent_t *   entry       = NULL;
+        gf_dirent_t *   tmp         = NULL;
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-	list_for_each_entry_safe (entry, tmp, &fd_ctx->entries, list) {
-		GF_FREE (entry->name);
-		list_del (&entry->list);
-		GF_FREE (entry);
-	}
+        list_for_each_entry_safe (entry, tmp, &entries->list, list) {
+                if (__is_root_gfid (fd->inode->gfid) &&
+                    !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
+                        list_del_init (&entry->list);
+                        GF_FREE (entry);
+                }
+        }
 }
 
-
 int32_t
 afr_readdir_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno,
-		 gf_dirent_t *entries)
+                 xlator_t *this, int32_t op_ret, int32_t op_errno,
+                 gf_dirent_t *entries, dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-
-        gf_dirent_t * entry = NULL;
-        gf_dirent_t * tmp   = NULL;
+        afr_local_t     *local = NULL;
 
-        int child_index = -1;
-
-	priv     = this->private;
-	local = frame->local;
-        child_index = (long) cookie;
-
-	if (op_ret != -1) {
-                list_for_each_entry_safe (entry, tmp, &entries->list, list) {
-                        entry->d_ino = afr_itransform (entry->d_ino,
-                                                       priv->child_count,
-                                                       child_index);
-
-                        if ((local->fd->inode == local->fd->inode->table->root)
-                            && !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
-                                list_del_init (&entry->list);
-                                GF_FREE (entry);
-                        }
-                }
-    	}
+        if (op_ret == -1)
+                goto out;
 
-        AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, entries);
+        local = frame->local;
+        afr_readdir_filter_trash_dir (entries, local->fd);
 
-	return 0;
+out:
+        AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, entries, NULL);
+        return 0;
 }
 
 
 int32_t
 afr_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno, gf_dirent_t *entries)
+                  int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                  dict_t *xdata)
 {
-        afr_private_t * priv     = NULL;
-        afr_local_t *   local    = NULL;
-        xlator_t **     children = NULL;
-        ino_t           inum = 0;
+        afr_local_t     *local = NULL;
 
-        int call_child = 0;
-        int ret        = 0;
-
-        gf_dirent_t * entry = NULL;
-        gf_dirent_t * tmp   = NULL;
-
-        int child_index = -1;
-
-        uint64_t      ctx    = 0;
-        afr_fd_ctx_t *fd_ctx = NULL;
-
-	off_t offset = 0;
-
-        priv     = this->private;
-        children = priv->children;
+        if (op_ret == -1)
+                goto out;
 
         local = frame->local;
-
-        child_index = (long) cookie;
-
-	if (priv->strict_readdir) {
-		ret = fd_ctx_get (local->fd, this, &ctx);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not get fd ctx for fd=%p", local->fd);
-			op_ret   = -1;
-			op_errno = -ret;
-			goto out;
-		}
-
-		fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-		if (child_went_down (op_ret, op_errno)) {
-			if (all_tried (child_index, priv->child_count)) {
-				goto out;
-			}
-
-			call_child = ++child_index;
-
-			gf_log (this->name, GF_LOG_TRACE,
-				"starting readdir afresh on child %d, offset %"PRId64,
-				call_child, (uint64_t) 0);
-
-			fd_ctx->failed_over = _gf_true;
-
-			STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
-					   (void *) (long) call_child,
-					   children[call_child],
-					   children[call_child]->fops->readdirp, local->fd,
-					   local->cont.readdir.size, 0);
-			return 0;
-		}
-	}
-
-	if (op_ret != -1) {
-		list_for_each_entry_safe (entry, tmp, &entries->list, list) {
-			inum = afr_itransform (entry->d_ino, priv->child_count,
-					       child_index);
-			entry->d_ino = inum;
-			inum  = afr_itransform (entry->d_stat.ia_ino,
-						priv->child_count, child_index);
-			entry->d_stat.ia_ino = inum;
-
-			if ((local->fd->inode == local->fd->inode->table->root)
-			    && !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
-				list_del_init (&entry->list);
-				GF_FREE (entry);
-			}
-		}
-	}
-
-	if (priv->strict_readdir) {
-		if (fd_ctx->failed_over) {
-			if (list_empty (&entries->list)) {
-				goto out;
-			}
-
-			offset = afr_filter_entries (entries, local->fd);
-
-			afr_remember_entries (entries, local->fd);
-
-			if (list_empty (&entries->list)) {
-				/* All the entries we got were duplicate. We
-				   shouldn't send an empty list now, because
-				   that'll make the application stop reading. So
-				   try to get more entries */
-
-				gf_log (this->name, GF_LOG_TRACE,
-					"trying to fetch non-duplicate entries from offset %"PRId64", child %s",
-					offset, children[child_index]->name);
-
-				STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
-						   (void *) (long) child_index,
-						   children[child_index],
-						   children[child_index]->fops->readdirp,
-						   local->fd, local->cont.readdir.size, offset);
-				return 0;
-			}
-		} else {
-			afr_remember_entries (entries, local->fd);
-		}
-	}
+        afr_readdir_filter_trash_dir (entries, local->fd);
 
 out:
-        AFR_STACK_UNWIND (readdirp, frame, op_ret, op_errno, entries);
-
+        AFR_STACK_UNWIND (readdirp, frame, op_ret, op_errno, entries, NULL);
         return 0;
 }
 
-
 int32_t
 afr_do_readdir (call_frame_t *frame, xlator_t *this,
-	        fd_t *fd, size_t size, off_t offset, int whichop)
+                fd_t *fd, size_t size, off_t offset, int whichop, dict_t *dict)
 {
-	afr_private_t * priv       = NULL;
-	xlator_t **     children   = NULL;
-	int             call_child = 0;
-	afr_local_t     *local     = NULL;
+        afr_private_t *priv      = NULL;
+        xlator_t      **children = NULL;
+        int           call_child = 0;
+        afr_local_t   *local     = NULL;
+        afr_fd_ctx_t  *fd_ctx    = NULL;
+        int           ret        = -1;
+        int32_t       op_errno   = 0;
+        uint64_t      read_child = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-	int ret = -1;
+        priv     = this->private;
+        children = priv->children;
 
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	priv     = this->private;
-	children = priv->children;
+        local->fresh_children = afr_children_create (priv->child_count);
+        if (!local->fresh_children) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        read_child = afr_inode_get_read_ctx (this, fd->inode,
+                                             local->fresh_children);
+        ret = afr_get_call_child (this, local->child_up, read_child,
+                                  local->fresh_children,
+                                  &call_child,
+                                  &local->cont.readdir.last_index);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto out;
+        }
 
-	frame->local = local;
+        fd_ctx  = afr_fd_ctx_get (fd, this);
+        if (!fd_ctx) {
+                op_errno = EBADF;
+                goto out;
+        }
 
-	call_child = afr_first_up_child (priv);
-	if (call_child == -1) {
-		op_errno = ENOTCONN;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no child is up");
-		goto out;
-	}
+        if ((offset == 0) || (fd_ctx->call_child == -1)) {
+                fd_ctx->call_child = call_child;
+        } else if ((priv->readdir_failover == _gf_false) &&
+                   (call_child != fd_ctx->call_child)) {
+                op_errno = EBADF;
+                goto out;
+        }
 
         local->fd                  = fd_ref (fd);
         local->cont.readdir.size   = size;
-
-	if (priv->strict_readdir) {
-		ret = fd_ctx_get (fd, this, &ctx);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not get fd ctx for fd=%p", fd);
-			op_errno = -ret;
-			goto out;
-		}
-
-		fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-		if (fd_ctx->last_tried != call_child) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"first up child has changed from %d to %d, restarting readdir from offset 0",
-				fd_ctx->last_tried, call_child);
-
-			fd_ctx->failed_over = _gf_true;
-			offset = 0;
-		}
-
-		fd_ctx->last_tried = call_child;
-	}
+        local->cont.readdir.dict   = (dict)? dict_ref (dict) : NULL;
 
         if (whichop == GF_FOP_READDIR)
                 STACK_WIND_COOKIE (frame, afr_readdir_cbk,
                                    (void *) (long) call_child,
                                    children[call_child],
                                    children[call_child]->fops->readdir, fd,
-                                   size, offset);
+                                   size, offset, dict);
         else
                 STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
                                    (void *) (long) call_child,
                                    children[call_child],
                                    children[call_child]->fops->readdirp, fd,
-                                   size, offset);
+                                   size, offset, dict);
 
-	op_ret = 0;
+        return 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
+        AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
+        return 0;
 }
 
 
 int32_t
 afr_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-             off_t offset)
+             off_t offset, dict_t *xdata)
 {
-        afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR);
+        afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR, xdata);
         return 0;
 }
 
 
 int32_t
 afr_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-              off_t offset)
+              off_t offset, dict_t *dict)
 {
-        afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIRP);
+        afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIRP, dict);
         return 0;
 }
 
@@ -736,8 +538,8 @@ afr_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
 int32_t
 afr_releasedir (xlator_t *this, fd_t *fd)
 {
-	afr_forget_entries (fd);
+        afr_forget_entries (fd);
         afr_cleanup_fd_ctx (this, fd);
 
-	return 0;
+        return 0;
 }
diff --git a/xlators/cluster/afr/src/afr-dir-read.h b/xlators/cluster/afr/src/afr-dir-read.h
index abde2534d..09456d159 100644
--- a/xlators/cluster/afr/src/afr-dir-read.h
+++ b/xlators/cluster/afr/src/afr-dir-read.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __DIR_READ_H__
@@ -23,28 +14,23 @@
 
 int32_t
 afr_opendir (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, fd_t *fd);
+	     loc_t *loc, fd_t *fd, dict_t *xdata);
 
 int32_t
 afr_releasedir (xlator_t *this, fd_t *fd);
 
 int32_t
 afr_readdir (call_frame_t *frame, xlator_t *this,
-	     fd_t *fd, size_t size, off_t offset);
+	     fd_t *fd, size_t size, off_t offset, dict_t *xdata);
 
 
 int32_t
 afr_readdirp (call_frame_t *frame, xlator_t *this,
-	     fd_t *fd, size_t size, off_t offset);
-
-int32_t
-afr_getdents (call_frame_t *frame, xlator_t *this,
-	      fd_t *fd, size_t size, off_t offset, int32_t flag);
-
+              fd_t *fd, size_t size, off_t offset, dict_t *dict);
 
 int32_t
 afr_checksum (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, int32_t flags);
+	      loc_t *loc, int32_t flags, dict_t *xdata);
 
 
 #endif /* __DIR_READ_H__ */
diff --git a/xlators/cluster/afr/src/afr-dir-write.c b/xlators/cluster/afr/src/afr-dir-write.c
index 912dde327..1943b719b 100644
--- a/xlators/cluster/afr/src/afr-dir-write.c
+++ b/xlators/cluster/afr/src/afr-dir-write.c
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -47,28 +38,222 @@
 #include "afr.h"
 #include "afr-transaction.h"
 
+int
+afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno)
+{
+        int     ret = -1;
+        char    *child_path = NULL;
+
+        if (!child->parent) {
+                if (op_errno)
+                        *op_errno = EINVAL;
+                goto out;
+        }
+
+        child_path = gf_strdup (child->path);
+        if (!child_path) {
+                if (op_errno)
+                        *op_errno = ENOMEM;
+                goto out;
+        }
+        parent->path = gf_strdup( dirname (child_path) );
+	if (!parent->path) {
+                if (op_errno)
+                        *op_errno = ENOMEM;
+                goto out;
+        }
+        parent->inode  = inode_ref (child->parent);
+        uuid_copy (parent->gfid, child->pargfid);
+
+        ret = 0;
+out:
+	GF_FREE(child_path);
+
+        return ret;
+}
 
 void
-afr_build_parent_loc (loc_t *parent, loc_t *child)
+__dir_entry_fop_common_cbk (call_frame_t *frame, int child_index,
+                            xlator_t *this, int32_t op_ret,
+                            int32_t op_errno, inode_t *inode,
+                            struct iatt *buf, struct iatt *preparent,
+                            struct iatt *postparent, struct iatt *prenewparent,
+                            struct iatt *postnewparent)
 {
-	char *tmp = NULL;
+        afr_local_t     *local          = NULL;
+
+        local = frame->local;
+
+        if (afr_fop_failed (op_ret, op_errno))
+                afr_transaction_fop_failed (frame, this, child_index);
+
+        if (op_ret > -1) {
+                local->op_ret = op_ret;
+
+                if ((local->success_count == 0) ||
+                    (child_index == local->read_child_index)) {
+                        local->cont.dir_fop.preparent      = *preparent;
+                        local->cont.dir_fop.postparent     = *postparent;
+                        if (buf)
+                                local->cont.dir_fop.buf            = *buf;
+                        if (prenewparent)
+                             local->cont.dir_fop.prenewparent  = *prenewparent;
+                        if (postnewparent)
+                             local->cont.dir_fop.postnewparent = *postnewparent;
+                }
 
-	if (!child->parent) {
-		loc_copy (parent, child);
-		return;
-	}
+                local->cont.dir_fop.inode = inode;
 
-	tmp = gf_strdup (child->path);
-	parent->path   = gf_strdup (dirname (tmp));
-	GF_FREE (tmp);
+                local->fresh_children[local->success_count] = child_index;
+                local->success_count++;
+                local->child_errno[child_index] = 0;
+        } else {
+                local->child_errno[child_index] = op_errno;
+        }
 
-        parent->name   = strrchr (parent->path, '/');
-	if (parent->name)
-		parent->name++;
+        local->op_errno = op_errno;
+}
 
-	parent->inode  = inode_ref (child->parent);
-	parent->parent = inode_parent (parent->inode, 0, NULL);
-	parent->ino    = parent->inode->ino;
+int
+afr_mark_new_entry_changelog_cbk (call_frame_t *frame, void *cookie,
+                                  xlator_t *this,
+                                  int32_t op_ret, int32_t op_errno,
+                                  dict_t *xattr, dict_t *xdata)
+{
+        int     call_count = 0;
+
+        call_count = afr_frame_return (frame);
+        if (call_count == 0) {
+                AFR_STACK_DESTROY (frame);
+        }
+        return 0;
+}
+
+void
+afr_mark_new_entry_changelog (call_frame_t *frame, xlator_t *this)
+{
+        call_frame_t  *new_frame  = NULL;
+        afr_local_t   *local      = NULL;
+        afr_local_t   *new_local  = NULL;
+        afr_private_t *priv       = NULL;
+        dict_t        **xattr     = NULL;
+        int32_t       **changelog = NULL;
+        int           i           = 0;
+        GF_UNUSED int op_errno    = 0;
+
+        local = frame->local;
+        priv = this->private;
+
+        new_frame = copy_frame (frame);
+        if (!new_frame) {
+                goto out;
+        }
+
+        AFR_LOCAL_ALLOC_OR_GOTO (new_frame->local, out);
+        new_local = new_frame->local;
+        changelog = afr_matrix_create (priv->child_count, AFR_NUM_CHANGE_LOGS);
+        if (!changelog)
+                goto out;
+
+        xattr = GF_CALLOC (priv->child_count, sizeof (*xattr),
+                           gf_afr_mt_dict_t);
+        if (!xattr)
+                goto out;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_errno[i])
+                        continue;
+                xattr[i] = dict_new ();
+                if (!xattr[i])
+                        goto out;
+        }
+
+        afr_prepare_new_entry_pending_matrix (changelog,
+                                              afr_is_errno_set,
+                                              local->child_errno,
+                                              &local->cont.dir_fop.buf,
+                                              priv->child_count);
+
+        new_local->pending = changelog;
+        uuid_copy (new_local->loc.gfid, local->cont.dir_fop.buf.ia_gfid);
+        new_local->loc.inode = inode_ref (local->cont.dir_fop.inode);
+        new_local->call_count = local->success_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_errno[i])
+                        continue;
+
+                afr_set_pending_dict (priv, xattr[i], changelog, i, LOCAL_LAST);
+                STACK_WIND_COOKIE (new_frame, afr_mark_new_entry_changelog_cbk,
+                                   (void *) (long) i, priv->children[i],
+                                   priv->children[i]->fops->xattrop,
+                                   &new_local->loc, GF_XATTROP_ADD_ARRAY,
+                                   xattr[i], NULL);
+        }
+        new_frame = NULL;
+out:
+        if (new_frame)
+                AFR_STACK_DESTROY (new_frame);
+        afr_xattr_array_destroy (xattr, priv->child_count);
+        return;
+}
+
+gf_boolean_t
+afr_is_new_entry_changelog_needed (glusterfs_fop_t fop)
+{
+        glusterfs_fop_t fops[]   = {GF_FOP_CREATE, GF_FOP_MKNOD, GF_FOP_NULL};
+        int             i        = 0;
+
+        for (i = 0; fops[i] != GF_FOP_NULL; i++) {
+                if (fop == fops[i])
+                        return _gf_true;
+        }
+        return _gf_false;
+}
+
+void
+afr_dir_fop_mark_entry_pending_changelog (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t   *local      = NULL;
+        afr_private_t *priv       = NULL;
+
+        local = frame->local;
+        priv  = this->private;
+
+        if (local->op_ret < 0)
+                goto out;
+
+        if (local->success_count == priv->child_count)
+                goto out;
+
+        if (!afr_is_new_entry_changelog_needed (local->op))
+                goto out;
+
+        afr_mark_new_entry_changelog (frame, this);
+
+out:
+        return;
+}
+
+void
+afr_dir_fop_done (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local          = NULL;
+        afr_private_t   *priv           = NULL;
+
+        local = frame->local;
+        priv  = this->private;
+
+        if (local->cont.dir_fop.inode == NULL)
+                goto done;
+        afr_set_read_ctx_from_policy (this, local->cont.dir_fop.inode,
+                                      local->fresh_children,
+                                      local->read_child_index,
+                                      priv->read_child,
+                                      local->cont.dir_fop.buf.ia_gfid);
+done:
+        local->transaction.unwind (frame, this);
+        afr_dir_fop_mark_entry_pending_changelog (frame, this);
+        local->transaction.resume (frame, this);
 }
 
 /* {{{ create */
@@ -76,282 +261,247 @@ afr_build_parent_loc (loc_t *parent, loc_t *child)
 int
 afr_create_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-        struct iatt  *unwind_buf = NULL;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                if (local->cont.create.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.create.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.create.buf;
-                }
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-                unwind_buf->ia_ino = local->cont.create.ino;
+        local = frame->local;
 
-                local->cont.create.preparent.ia_ino  = local->cont.create.parent_ino;
-                local->cont.create.postparent.ia_ino = local->cont.create.parent_ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame) {
+                        main_frame = local->transaction.main_frame;
+                }
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-		AFR_STACK_UNWIND (create, main_frame,
+        if (main_frame) {
+                AFR_STACK_UNWIND (create, main_frame,
                                   local->op_ret, local->op_errno,
-				  local->cont.create.fd,
-				  local->cont.create.inode,
-				  unwind_buf, &local->cont.create.preparent,
-                                  &local->cont.create.postparent);
+                                  local->cont.create.fd,
+                                  local->cont.dir_fop.inode,
+                                  &local->cont.dir_fop.buf,
+                                  &local->cont.dir_fop.preparent,
+                                  &local->cont.dir_fop.postparent,
+                                  local->xdata_rsp);
         }
-        
-	return 0;
+
+        return 0;
 }
 
 
 int
-afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		     int32_t op_ret, int32_t op_errno, 
-		     fd_t *fd, inode_t *inode, struct iatt *buf,
-                     struct iatt *preparent, struct iatt *postparent)
+afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno,
+                     fd_t *fd, inode_t *inode, struct iatt *buf,
+                     struct iatt *preparent, struct iatt *postparent,
+                     dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-        int ret = 0;
-
-	int call_count = -1;
-	int child_index = -1;
-
-	local = frame->local;
-	priv  = this->private;
+        afr_local_t     *local = NULL;
+        uint64_t        ctx = 0;
+        afr_fd_ctx_t    *fd_ctx = NULL;
+        int             ret = 0;
+        int             call_count = -1;
+        int             child_index = -1;
 
-	child_index = (long) cookie;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-                
-		if (op_ret != -1) {
-			local->op_ret = op_ret;
+        child_index = (long) cookie;
 
+        LOCK (&frame->lock);
+        {
+                if (op_ret > -1) {
                         ret = afr_fd_ctx_set (this, fd);
-
                         if (ret < 0) {
-                                gf_log (this->name, GF_LOG_DEBUG,
+                                gf_log (this->name, GF_LOG_ERROR,
                                         "could not set ctx on fd=%p", fd);
 
                                 local->op_ret   = -1;
                                 local->op_errno = -ret;
+                                goto unlock;
                         }
 
                         ret = fd_ctx_get (fd, this, &ctx);
-
                         if (ret < 0) {
-                                gf_log (this->name, GF_LOG_DEBUG,
+                                gf_log (this->name, GF_LOG_ERROR,
                                         "could not get fd ctx for fd=%p", fd);
                                 local->op_ret   = -1;
                                 local->op_errno = -ret;
+                                goto unlock;
                         }
 
                         fd_ctx = (afr_fd_ctx_t *)(long) ctx;
 
-                        fd_ctx->opened_on[child_index] = 1;
+                        fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
                         fd_ctx->flags                  = local->cont.create.flags;
 
                         if (local->success_count == 0) {
-				local->cont.create.buf        = *buf;
-
-                                local->cont.create.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        child_index);
-
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this, inode, 
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this, inode, 
-                                                            local->read_child_index);
-                                }
+				if (xdata)
+					local->xdata_rsp = dict_ref(xdata);
 			}
-                        
-                        if (child_index == local->first_up_child) {
-                                local->cont.create.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                        }
-                        
-                        if (child_index == local->read_child_index) {
-                                local->cont.create.read_child_buf = *buf;
-                                local->cont.create.preparent      = *preparent;
-                                local->cont.create.postparent     = *postparent;
-                        }
-
-			local->cont.create.inode = inode;
-
-			local->success_count++;
-		}
+                }
+                __dir_entry_fop_common_cbk (frame, child_index, this,
+                                            op_ret, op_errno, inode, buf,
+                                            preparent, postparent, NULL, NULL);
+        }
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+unlock:
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+        if (call_count == 0)
+                afr_dir_fop_done (frame, this);
 
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int
 afr_create_wind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_create_wind_cbk,
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->create,
-					   &local->loc, 
-					   local->cont.create.flags, 
-					   local->cont.create.mode, 
-					   local->cont.create.fd,
-                                           local->cont.create.params);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
+
+        local = frame->local;
+        priv = this->private;
+
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
+
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_create_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->create,
+                                           &local->loc,
+                                           local->cont.create.flags,
+                                           local->cont.create.mode,
+                                           local->umask,
+                                           local->cont.create.fd,
+                                           local->xdata_req);
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
 }
 
 
 int
 afr_create_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t * local = NULL;
+        afr_local_t * local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
+        AFR_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_create (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t flags, mode_t mode,
-            fd_t *fd, dict_t *params)
+            loc_t *loc, int32_t flags, mode_t mode,
+            mode_t umask, fd_t *fd, dict_t *params)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
-
-	int ret = -1;
-
-	int op_ret   = -1;
-	int op_errno = 0;
+        afr_private_t           *priv                   = NULL;
+        afr_local_t             *local                  = NULL;
+        afr_internal_lock_t     *int_lock               = NULL;
+        call_frame_t            *transaction_frame      = NULL;
+        int                     ret                     = -1;
+        int                     op_errno                = 0;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	priv = this->private;
+        priv = this->private;
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
+        QUORUM_CHECK(create,out);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
 
-	transaction_frame->local = local;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	loc_copy (&local->loc, loc);
+        loc_copy (&local->loc, loc);
 
         LOCK (&priv->read_child_lock);
         {
-                local->read_child_index = (++priv->read_child_rr) 
+                local->read_child_index = (++priv->read_child_rr)
                         % (priv->child_count);
         }
         UNLOCK (&priv->read_child_lock);
 
-	local->cont.create.flags = flags;
-	local->cont.create.mode  = mode;
-	local->cont.create.fd    = fd_ref (fd);
+        local->op                = GF_FOP_CREATE;
+        local->cont.create.flags = flags;
+        local->cont.create.mode  = mode;
+        local->cont.create.fd    = fd_ref (fd);
+        local->umask  = umask;
         if (params)
-                local->cont.create.params = dict_ref (params);
-
-        if (loc->parent)
-                local->cont.create.parent_ino = loc->parent->ino;
-
-	local->transaction.fop    = afr_create_wind;
-	local->transaction.done   = afr_create_done;
-	local->transaction.unwind = afr_create_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
-
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+                local->xdata_req = dict_ref (params);
+
+        local->transaction.fop    = afr_create_wind;
+        local->transaction.done   = afr_create_done;
+        local->transaction.unwind = afr_create_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame = frame;
+        local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (create, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL, NULL);
-	}
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (create, frame, -1, op_errno,
+                                  NULL, NULL, NULL, NULL, NULL, NULL);
+        }
 
-	return 0;
+        return 0;
 }
 
 /* }}} */
@@ -361,42 +511,31 @@ out:
 int
 afr_mknod_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-
-        struct iatt *unwind_buf = NULL;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                if (local->cont.mknod.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.mknod.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.mknod.buf;
-                }
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-                unwind_buf->ia_ino = local->cont.mknod.ino;
+        local = frame->local;
 
-                local->cont.mknod.preparent.ia_ino  = local->cont.mknod.parent_ino;
-                local->cont.mknod.postparent.ia_ino = local->cont.mknod.parent_ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame) {
+                        main_frame = local->transaction.main_frame;
+                }
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-		AFR_STACK_UNWIND (mknod, main_frame,
+        if (main_frame) {
+                AFR_STACK_UNWIND (mknod, main_frame,
                                   local->op_ret, local->op_errno,
-				  local->cont.mknod.inode,
-				  unwind_buf, &local->cont.mknod.preparent,
-                                  &local->cont.mknod.postparent);
+                                  local->cont.dir_fop.inode,
+                                  &local->cont.dir_fop.buf,
+                                  &local->cont.dir_fop.preparent,
+                                  &local->cont.dir_fop.postparent,
+                                  NULL);
         }
 
-	return 0;
+        return 0;
 }
 
 
@@ -404,203 +543,169 @@ int
 afr_mknod_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno, inode_t *inode,
                     struct iatt *buf, struct iatt *preparent,
-                    struct iatt *postparent)
+                    struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count = -1;
-	int child_index = -1;
-
-	local = frame->local;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-		
-		if (op_ret != -1) {
-			local->op_ret = op_ret;
-
-			if (local->success_count == 0){
-				local->cont.mknod.buf   = *buf;
-				local->cont.mknod.ino   = 
-					afr_itransform (buf->ia_ino,
-							priv->child_count,
-							child_index);
-
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this, inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this, inode,
-                                                            local->read_child_index);
-                                }
-			}
-
-                        if (child_index == local->first_up_child) {
-                                local->cont.mknod.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                        }
-
-                        if (child_index == local->read_child_index) {
-                                local->cont.mknod.read_child_buf = *buf;
-                                local->cont.mknod.preparent      = *preparent;
-                                local->cont.mknod.postparent     = *postparent;
-                        }
-                        
-			local->cont.mknod.inode = inode;
+        int             call_count      = -1;
+        int             child_index     = -1;
 
-			local->success_count++;
-		}
+        child_index = (long) cookie;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        LOCK (&frame->lock);
+        {
+                __dir_entry_fop_common_cbk (frame, child_index, this,
+                                            op_ret, op_errno, inode, buf,
+                                            preparent, postparent, NULL, NULL);
+        }
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+        if (call_count == 0)
+                afr_dir_fop_done (frame, this);
 
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_mknod_wind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv  = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_mknod_wind_cbk, (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->mknod,
-					   &local->loc, local->cont.mknod.mode,
-					   local->cont.mknod.dev,
-                                           local->cont.mknod.params);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
+
+        local = frame->local;
+        priv  = this->private;
+
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
+
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_mknod_wind_cbk, (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->mknod,
+                                           &local->loc, local->cont.mknod.mode,
+                                           local->cont.mknod.dev,
+                                           local->umask,
+                                           local->xdata_req);
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
 }
 
 
 int
 afr_mknod_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t * local = NULL;
+        afr_local_t * local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	local->transaction.unwind (frame, this);
-	AFR_STACK_DESTROY (frame);
+        local->transaction.unwind (frame, this);
+        AFR_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
 
 int
-afr_mknod (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dev_t dev, dict_t *params)
+afr_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+           dev_t dev, mode_t umask, dict_t *params)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
+        afr_private_t           *priv                   = NULL;
+        afr_local_t             *local                  = NULL;
+        afr_internal_lock_t     *int_lock               = NULL;
+        call_frame_t            *transaction_frame      = NULL;
+        int                     ret                     = -1;
+        int                     op_errno                = 0;
 
-	int ret = -1;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        priv = this->private;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        QUORUM_CHECK(mknod,out);
 
-	priv = this->private;
-
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
 
-	transaction_frame->local = local;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	loc_copy (&local->loc, loc);
+        loc_copy (&local->loc, loc);
 
         LOCK (&priv->read_child_lock);
         {
-                local->read_child_index = (++priv->read_child_rr) 
+                local->read_child_index = (++priv->read_child_rr)
                         % (priv->child_count);
         }
         UNLOCK (&priv->read_child_lock);
 
-	local->cont.mknod.mode  = mode;
-	local->cont.mknod.dev   = dev;
+        local->op               = GF_FOP_MKNOD;
+        local->cont.mknod.mode  = mode;
+        local->cont.mknod.dev   = dev;
+        local->umask = umask;
         if (params)
-                local->cont.mknod.params = dict_ref (params);
-
-        if (loc->parent)
-                local->cont.mknod.parent_ino = loc->parent->ino;
-
-	local->transaction.fop    = afr_mknod_wind;
-	local->transaction.done   = afr_mknod_done;
-	local->transaction.unwind = afr_mknod_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
-
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+                local->xdata_req = dict_ref (params);
+
+        local->transaction.fop    = afr_mknod_wind;
+        local->transaction.done   = afr_mknod_done;
+        local->transaction.unwind = afr_mknod_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame = frame;
+        local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (mknod, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL);
-	}
-
-	return 0;
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (mknod, frame, -1, op_errno,
+                                  NULL, NULL, NULL, NULL, NULL);
+        }
+
+        return 0;
 }
 
 /* }}} */
@@ -611,42 +716,31 @@ out:
 int
 afr_mkdir_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-
-        struct iatt *unwind_buf = NULL;
-        
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                if (local->cont.mkdir.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.mkdir.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.mkdir.buf;
-                }
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-                unwind_buf->ia_ino = local->cont.mkdir.ino;
+        local = frame->local;
 
-                local->cont.mkdir.preparent.ia_ino  = local->cont.mkdir.parent_ino;
-                local->cont.mkdir.postparent.ia_ino = local->cont.mkdir.parent_ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame) {
+                        main_frame = local->transaction.main_frame;
+                }
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-		AFR_STACK_UNWIND (mkdir, main_frame,
+        if (main_frame) {
+                AFR_STACK_UNWIND (mkdir, main_frame,
                                   local->op_ret, local->op_errno,
-				  local->cont.mkdir.inode,
-				  unwind_buf, &local->cont.mkdir.preparent,
-                        &local->cont.mkdir.postparent);
+                                  local->cont.dir_fop.inode,
+                                  &local->cont.dir_fop.buf,
+                                  &local->cont.dir_fop.preparent,
+                                  &local->cont.dir_fop.postparent,
+                                  NULL);
         }
 
-	return 0;
+        return 0;
 }
 
 
@@ -654,205 +748,169 @@ int
 afr_mkdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno, inode_t *inode,
                     struct iatt *buf, struct iatt *preparent,
-                    struct iatt *postparent)
+                    struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count = -1;
-	int child_index = -1;
-
-	local = frame->local;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-
-		if (op_ret != -1) {
-			local->op_ret           = op_ret;
+        int             call_count      = -1;
+        int             child_index     = -1;
 
-			if (local->success_count == 0) {
-				local->cont.mkdir.buf   = *buf;
+        child_index = (long) cookie;
 
-                                local->cont.mkdir.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        child_index);
+        LOCK (&frame->lock);
+        {
+                __dir_entry_fop_common_cbk (frame, child_index, this,
+                                            op_ret, op_errno, inode, buf,
+                                            preparent, postparent, NULL, NULL);
+        }
+        UNLOCK (&frame->lock);
 
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this, inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this, inode,
-                                                            local->read_child_index);
-                                }
-			}
+        call_count = afr_frame_return (frame);
 
-                        if (child_index == local->first_up_child) {
-                                local->cont.mkdir.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                        }
-                        
-                        if (child_index == local->read_child_index) {
-                                local->cont.mkdir.read_child_buf = *buf;
-                                local->cont.mkdir.preparent      = *preparent;
-                                local->cont.mkdir.postparent     = *postparent;
-                        }
+        if (call_count == 0)
+                afr_dir_fop_done (frame, this);
 
-			local->cont.mkdir.inode = inode;
+        return 0;
+}
 
-			local->success_count++;
-		}
 
-		local->op_errno         = op_errno;
-	}
-	UNLOCK (&frame->lock);
+int
+afr_mkdir_wind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
-	call_count = afr_frame_return (frame);
+        local = frame->local;
+        priv  = this->private;
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
-}
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_mkdir_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->mkdir,
+                                           &local->loc, local->cont.mkdir.mode,
+                                           local->umask,
+                                           local->xdata_req);
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-int
-afr_mkdir_wind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-	
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv  = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_mkdir_wind_cbk,
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->mkdir,
-					   &local->loc, local->cont.mkdir.mode,
-                                           local->cont.mkdir.params);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int
 afr_mkdir_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t * local = NULL;
+        afr_local_t * local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
+        AFR_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
-
 int
 afr_mkdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dict_t *params)
+           loc_t *loc, mode_t mode, mode_t umask, dict_t *params)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
-
-	int ret = -1;
-
-	int op_ret   = -1;
-	int op_errno = 0;
+        afr_private_t           *priv                   = NULL;
+        afr_local_t             *local                  = NULL;
+        afr_internal_lock_t     *int_lock               = NULL;
+        call_frame_t            *transaction_frame      = NULL;
+        int                     ret                     = -1;
+        int                     op_errno                = 0;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	priv = this->private;
+        priv = this->private;
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
+        QUORUM_CHECK(mkdir,out);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
 
-	transaction_frame->local = local;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	loc_copy (&local->loc, loc);
+        loc_copy (&local->loc, loc);
 
         LOCK (&priv->read_child_lock);
         {
-                local->read_child_index = (++priv->read_child_rr) 
+                local->read_child_index = (++priv->read_child_rr)
                         % (priv->child_count);
         }
         UNLOCK (&priv->read_child_lock);
 
-	local->cont.mkdir.mode  = mode;
+        local->cont.mkdir.mode  = mode;
+        local->umask = umask;
         if (params)
-                local->cont.mkdir.params = dict_ref (params);
-
-        if (loc->parent)
-                local->cont.mkdir.parent_ino = loc->parent->ino;
-
-	local->transaction.fop    = afr_mkdir_wind;
-	local->transaction.done   = afr_mkdir_done;
-	local->transaction.unwind = afr_mkdir_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
-
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+                local->xdata_req = dict_ref (params);
+
+        local->op = GF_FOP_MKDIR;
+        local->transaction.fop    = afr_mkdir_wind;
+        local->transaction.done   = afr_mkdir_done;
+        local->transaction.unwind = afr_mkdir_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame = frame;
+        local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
 
-		AFR_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL);
-	}
+                AFR_STACK_UNWIND (mkdir, frame, -1, op_errno,
+                                  NULL, NULL, NULL, NULL, NULL);
+        }
 
-	return 0;
+        return 0;
 }
 
 /* }}} */
@@ -863,233 +921,196 @@ out:
 int
 afr_link_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-
-        struct iatt *unwind_buf = NULL;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                if (local->cont.link.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.link.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.link.buf;
-                }
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-		unwind_buf->ia_ino = local->cont.link.ino;
+        local = frame->local;
 
-                local->cont.link.preparent.ia_ino  = local->cont.link.parent_ino;
-                local->cont.link.postparent.ia_ino = local->cont.link.parent_ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame) {
+                        main_frame = local->transaction.main_frame;
+                }
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-		AFR_STACK_UNWIND (link, main_frame,
-                                  local->op_ret, local->op_errno, 
-				  local->cont.link.inode,
-				  unwind_buf, &local->cont.link.preparent,
-                                  &local->cont.link.postparent);
-	}
+        if (main_frame) {
+                AFR_STACK_UNWIND (link, main_frame,
+                                  local->op_ret, local->op_errno,
+                                  local->cont.dir_fop.inode,
+                                  &local->cont.dir_fop.buf,
+                                  &local->cont.dir_fop.preparent,
+                                  &local->cont.dir_fop.postparent,
+                                  NULL);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
-afr_link_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		   int32_t op_ret, int32_t op_errno, inode_t *inode,
+afr_link_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, inode_t *inode,
                    struct iatt *buf, struct iatt *preparent,
-                   struct iatt *postparent)
+                   struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count = -1;
-	int child_index = -1;
+        int             call_count      = -1;
+        int             child_index     = -1;
 
-	local = frame->local;
-	priv = this->private;
+        child_index = (long) cookie;
 
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
+        LOCK (&frame->lock);
+        {
+                __dir_entry_fop_common_cbk (frame, child_index, this,
+                                            op_ret, op_errno, inode, buf,
+                                            preparent, postparent, NULL, NULL);
+        }
+        UNLOCK (&frame->lock);
 
-		if (op_ret != -1) {
-			local->op_ret   = op_ret;
+        call_count = afr_frame_return (frame);
 
-			if (local->success_count == 0) {
-				local->cont.link.buf        = *buf;
+        if (call_count == 0)
+                afr_dir_fop_done (frame, this);
 
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this, inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this, inode,
-                                                            local->read_child_index);
-                                }
-			}
+        return 0;
+}
 
-                        if (child_index == local->read_child_index) {
-                                local->cont.link.read_child_buf = *buf;
-                                local->cont.link.preparent      = *preparent;
-                                local->cont.link.postparent     = *postparent;
-                        }
 
-			local->cont.link.inode    = inode;
+int
+afr_link_wind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
-			local->success_count++;
-		}
+        local = frame->local;
+        priv  = this->private;
 
-		local->op_errno = op_errno;		
-	}
-	UNLOCK (&frame->lock);
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-	call_count = afr_frame_return (frame);
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+        local->call_count = call_count;
 
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
-}
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_link_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->link,
+                                           &local->loc,
+                                           &local->newloc, local->xdata_req);
 
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-int
-afr_link_wind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv  = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_link_wind_cbk, (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->link,
-					   &local->loc,
-					   &local->newloc);
-			
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int
 afr_link_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t * local = frame->local;
+        afr_local_t * local = frame->local;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
+        AFR_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_link (call_frame_t *frame, xlator_t *this,
-	  loc_t *oldloc, loc_t *newloc)
+          loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
-
-	int ret = -1;
-
-	int op_ret   = -1;
-	int op_errno = 0;
+        afr_private_t           *priv                   = NULL;
+        afr_local_t             *local                  = NULL;
+        afr_internal_lock_t     *int_lock               = NULL;
+        call_frame_t            *transaction_frame      = NULL;
+        int                     ret                     = -1;
+        int                     op_errno                = 0;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	priv = this->private;
+        priv = this->private;
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
+        QUORUM_CHECK(link,out);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
 
-	transaction_frame->local = local;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	loc_copy (&local->loc,    oldloc);
-	loc_copy (&local->newloc, newloc);
+        loc_copy (&local->loc,    oldloc);
+        loc_copy (&local->newloc, newloc);
+        if (xdata)
+                local->xdata_req = dict_ref (xdata);
 
         LOCK (&priv->read_child_lock);
         {
-                local->read_child_index = (++priv->read_child_rr) 
+                local->read_child_index = (++priv->read_child_rr)
                         % (priv->child_count);
         }
         UNLOCK (&priv->read_child_lock);
 
-	local->cont.link.ino = oldloc->inode->ino;
-
-        if (oldloc->parent)
-                local->cont.link.parent_ino = newloc->parent->ino;
-
-	local->transaction.fop    = afr_link_wind;
-	local->transaction.done   = afr_link_done;
-	local->transaction.unwind = afr_link_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, oldloc);
-
-	local->transaction.main_frame   = frame;
-	local->transaction.basename     = AFR_BASENAME (oldloc->path);
-	local->transaction.new_basename = AFR_BASENAME (newloc->path);
-
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
-
-	op_ret = 0;
+        local->op = GF_FOP_LINK;
+        local->transaction.fop    = afr_link_wind;
+        local->transaction.done   = afr_link_done;
+        local->transaction.unwind = afr_link_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, newloc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame   = frame;
+        local->transaction.basename     = AFR_BASENAME (newloc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (link, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL);
-	}
-
-	return 0;
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (link, frame, -1, op_errno,
+                                  NULL, NULL, NULL, NULL, NULL);
+        }
+
+        return 0;
 }
 
 /* }}} */
@@ -1100,246 +1121,202 @@ out:
 int
 afr_symlink_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-        
-        struct iatt *unwind_buf = NULL;
-        
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                if (local->cont.symlink.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.symlink.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.symlink.buf;
-                }
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-                unwind_buf->ia_ino = local->cont.symlink.ino;
+        local = frame->local;
 
-                local->cont.symlink.preparent.ia_ino  = local->cont.symlink.parent_ino;
-                local->cont.symlink.postparent.ia_ino = local->cont.symlink.parent_ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame) {
+                        main_frame = local->transaction.main_frame;
+                }
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-		AFR_STACK_UNWIND (symlink, main_frame,
+        if (main_frame) {
+                AFR_STACK_UNWIND (symlink, main_frame,
                                   local->op_ret, local->op_errno,
-				  local->cont.symlink.inode,
-				  unwind_buf, &local->cont.symlink.preparent,
-                                  &local->cont.symlink.postparent);
+                                  local->cont.dir_fop.inode,
+                                  &local->cont.dir_fop.buf,
+                                  &local->cont.dir_fop.preparent,
+                                  &local->cont.dir_fop.postparent,
+                                  NULL);
         }
 
-	return 0;
+        return 0;
 }
 
 
 int
-afr_symlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		      int32_t op_ret, int32_t op_errno, inode_t *inode,
+afr_symlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, inode_t *inode,
                       struct iatt *buf, struct iatt *preparent,
-                      struct iatt *postparent)
+                      struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count = -1;
-	int child_index = -1;
-
-	local = frame->local;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-		
-		if (op_ret != -1) {
-			local->op_ret   = op_ret;
-
-			if (local->success_count == 0) {
-				local->cont.symlink.buf        = *buf;
-				local->cont.symlink.ino = 
-					afr_itransform (buf->ia_ino, priv->child_count,
-							child_index);
-
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this, inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this, inode,
-                                                            local->read_child_index);
-                                }
-			}
-
-                        if (child_index == local->first_up_child) {
-                                local->cont.symlink.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                        }
-
-                        if (child_index == local->read_child_index) {
-                                local->cont.symlink.read_child_buf = *buf;
-                                local->cont.symlink.preparent      = *preparent;
-                                local->cont.symlink.postparent     = *postparent;
-                        }
-
-			local->cont.symlink.inode    = inode;
+        int             call_count      = -1;
+        int             child_index     = -1;
 
-			local->success_count++;
-		}
+        child_index = (long) cookie;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        LOCK (&frame->lock);
+        {
+                __dir_entry_fop_common_cbk (frame, child_index, this,
+                                            op_ret, op_errno, inode, buf,
+                                            preparent, postparent, NULL, NULL);
+        }
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+        if (call_count == 0)
+                afr_dir_fop_done (frame, this);
 
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int
 afr_symlink_wind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
-	int call_count = -1;
-	int i = 0;
+        local = frame->local;
+        priv = this->private;
 
-	local = frame->local;
-	priv = this->private;
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+        local->call_count = call_count;
 
-	local->call_count = call_count;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_symlink_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->symlink,
+                                           local->cont.symlink.linkpath,
+                                           &local->loc,
+                                           local->umask,
+                                           local->xdata_req);
 
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_symlink_wind_cbk,
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->symlink,
-					   local->cont.symlink.linkpath,
-					   &local->loc,
-                                           local->cont.symlink.params);
+                        if (!--call_count)
+                                break;
 
-			if (!--call_count)
-				break;
+                }
+        }
 
-		}
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int
 afr_symlink_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t * local = frame->local;
+        afr_local_t * local = frame->local;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
-	
-	return 0;
+        AFR_STACK_DESTROY (frame);
+
+        return 0;
 }
 
 
 int
 afr_symlink (call_frame_t *frame, xlator_t *this,
-	     const char *linkpath, loc_t *loc, dict_t *params)
+             const char *linkpath, loc_t *loc, mode_t umask, dict_t *params)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
-
-	int ret = -1;
+        afr_private_t           *priv                   = NULL;
+        afr_local_t             *local                  = NULL;
+        afr_internal_lock_t     *int_lock               = NULL;
+        call_frame_t            *transaction_frame      = NULL;
+        int                     ret                     = -1;
+        int                     op_errno                = 0;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        priv = this->private;
 
-	priv = this->private;
+        QUORUM_CHECK(symlink,out);
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	transaction_frame->local = local;
-	
-	loc_copy (&local->loc, loc);
+        loc_copy (&local->loc, loc);
 
         LOCK (&priv->read_child_lock);
         {
-                local->read_child_index = (++priv->read_child_rr) 
+                local->read_child_index = (++priv->read_child_rr)
                         % (priv->child_count);
         }
         UNLOCK (&priv->read_child_lock);
 
-	local->cont.symlink.linkpath = gf_strdup (linkpath);
+        local->cont.symlink.linkpath = gf_strdup (linkpath);
+        local->umask = umask;
         if (params)
-                local->cont.symlink.params = dict_ref (params);
-
-        if (loc->parent)
-                local->cont.symlink.parent_ino = loc->parent->ino;
-
-	local->transaction.fop    = afr_symlink_wind;
-	local->transaction.done   = afr_symlink_done;
-	local->transaction.unwind = afr_symlink_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame   = frame;
-	local->transaction.basename     = AFR_BASENAME (loc->path);
-
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+                local->xdata_req = dict_ref (params);
+
+        local->op = GF_FOP_SYMLINK;
+        local->transaction.fop    = afr_symlink_wind;
+        local->transaction.done   = afr_symlink_done;
+        local->transaction.unwind = afr_symlink_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame   = frame;
+        local->transaction.basename     = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (symlink, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL);
-	}
-
-	return 0;
+        if (ret  < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (symlink, frame, -1, op_errno,
+                                  NULL, NULL, NULL, NULL, NULL);
+        }
+
+        return 0;
 }
 
 /* }}} */
@@ -1349,228 +1326,231 @@ out:
 int
 afr_rename_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-
-        struct iatt *unwind_buf = NULL;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                if (local->cont.rename.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.rename.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.rename.buf;
-                }
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-		unwind_buf->ia_ino = local->cont.rename.ino;
+        local = frame->local;
 
-                local->cont.rename.preoldparent.ia_ino  = local->cont.rename.oldparent_ino;
-                local->cont.rename.postoldparent.ia_ino = local->cont.rename.oldparent_ino;
-                local->cont.rename.prenewparent.ia_ino  = local->cont.rename.newparent_ino;
-                local->cont.rename.postnewparent.ia_ino = local->cont.rename.newparent_ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame) {
+                        main_frame = local->transaction.main_frame;
+                }
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-		AFR_STACK_UNWIND (rename, main_frame,
-                                  local->op_ret, local->op_errno, 
-				  unwind_buf,
-                                  &local->cont.rename.preoldparent,
-                                  &local->cont.rename.postoldparent,
-                                  &local->cont.rename.prenewparent,
-                                  &local->cont.rename.postnewparent);
-	}
+        if (main_frame) {
+                AFR_STACK_UNWIND (rename, main_frame,
+                                  local->op_ret, local->op_errno,
+                                  &local->cont.dir_fop.buf,
+                                  &local->cont.dir_fop.preparent,
+                                  &local->cont.dir_fop.postparent,
+                                  &local->cont.dir_fop.prenewparent,
+                                  &local->cont.dir_fop.postnewparent,
+                                  NULL);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
-afr_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		     int32_t op_ret, int32_t op_errno, struct iatt *buf,
+afr_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *buf,
                      struct iatt *preoldparent, struct iatt *postoldparent,
-                     struct iatt *prenewparent, struct iatt *postnewparent)
+                     struct iatt *prenewparent, struct iatt *postnewparent,
+                     dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-
-	int call_count = -1;
-	int child_index = -1;
-
-	local = frame->local;
-
-	child_index = (long) cookie;
+        afr_local_t *   local = NULL;
+        int call_count = -1;
+        int child_index = -1;
 
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno) && op_errno != ENOTEMPTY)
-			afr_transaction_fop_failed (frame, this, child_index);
+        local = frame->local;
 
-		if (op_ret != -1) {
-                        if (local->success_count == 0) {
-                                local->op_ret = op_ret;
-		
-                                if (buf) {
-                                        local->cont.rename.buf = *buf;
-                                }
+        child_index = (long) cookie;
 
-                                local->success_count++;
-                        }
+        LOCK (&frame->lock);
+        {
+                if (afr_fop_failed (op_ret, op_errno) && op_errno != ENOTEMPTY)
+                        afr_transaction_fop_failed (frame, this, child_index);
+                local->op_errno = op_errno;
+                local->child_errno[child_index] = op_errno;
 
-                        if (child_index == local->read_child_index) {
-                                local->cont.rename.read_child_buf = *buf;
+                if (op_ret > -1)
+                        __dir_entry_fop_common_cbk (frame, child_index, this,
+                                                   op_ret, op_errno, NULL, buf,
+                                                   preoldparent, postoldparent,
+                                                   prenewparent, postnewparent);
 
-                                local->cont.rename.preoldparent  = *preoldparent;
-                                local->cont.rename.postoldparent = *postoldparent;
-                                local->cont.rename.prenewparent  = *prenewparent;
-                                local->cont.rename.postnewparent = *postnewparent;
-                        }
-                }
+        }
+        UNLOCK (&frame->lock);
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        call_count = afr_frame_return (frame);
 
-	call_count = afr_frame_return (frame);
+        if (call_count == 0)
+                afr_dir_fop_done (frame, this);
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_rename_wind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_rename_wind_cbk, 
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->rename,
-					   &local->loc,
-					   &local->newloc);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
-}
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
+        local = frame->local;
+        priv = this->private;
 
-int
-afr_rename_done (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t * local = frame->local;
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-	local->transaction.unwind (frame, this);
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
-	AFR_STACK_DESTROY (frame);
-	
-	return 0;
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_rename_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->rename,
+                                           &local->loc,
+                                           &local->newloc, NULL);
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
 }
 
 
 int
-afr_rename (call_frame_t *frame, xlator_t *this,
-	    loc_t *oldloc, loc_t *newloc)
+afr_rename_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
-
-	int ret = -1;
-
-	int op_ret   = -1;
-	int op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-
-	priv = this->private;
-
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
-
-	transaction_frame->local = local;
-
-	loc_copy (&local->loc,    oldloc);
-	loc_copy (&local->newloc, newloc);
-
-        local->read_child_index = afr_read_child (this, oldloc->inode);
+        afr_local_t * local = frame->local;
 
-	local->cont.rename.ino = oldloc->inode->ino;
+        local->transaction.unwind (frame, this);
 
-        if (oldloc->parent)
-                local->cont.rename.oldparent_ino = oldloc->parent->ino;
-        if (newloc->parent)
-                local->cont.rename.newparent_ino = newloc->parent->ino;
+        AFR_STACK_DESTROY (frame);
 
-	local->transaction.fop    = afr_rename_wind;
-	local->transaction.done   = afr_rename_done;
-	local->transaction.unwind = afr_rename_unwind;
+        return 0;
+}
 
-	afr_build_parent_loc (&local->transaction.parent_loc, oldloc);
-	afr_build_parent_loc (&local->transaction.new_parent_loc, newloc);
 
-	local->transaction.main_frame   = frame;
-	local->transaction.basename     = AFR_BASENAME (oldloc->path);
-	local->transaction.new_basename = AFR_BASENAME (newloc->path);
+int
+afr_rename (call_frame_t *frame, xlator_t *this,
+            loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+        afr_private_t           *priv                   = NULL;
+        afr_local_t             *local                  = NULL;
+        afr_internal_lock_t     *int_lock               = NULL;
+        call_frame_t            *transaction_frame      = NULL;
+        int                     ret                     = -1;
+        int                     op_errno                = 0;
+        int                     nlockee                 = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        priv = this->private;
+
+        QUORUM_CHECK(rename,out);
+
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
+
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
+
+        loc_copy (&local->loc,    oldloc);
+        loc_copy (&local->newloc, newloc);
+
+        local->read_child_index = afr_inode_get_read_ctx (this, oldloc->inode, NULL);
+
+        local->op = GF_FOP_RENAME;
+        local->transaction.fop    = afr_rename_wind;
+        local->transaction.done   = afr_rename_done;
+        local->transaction.unwind = afr_rename_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, oldloc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+        ret = afr_build_parent_loc (&local->transaction.new_parent_loc, newloc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame   = frame;
+        local->transaction.basename     = AFR_BASENAME (oldloc->path);
+        local->transaction.new_basename = AFR_BASENAME (newloc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = nlockee = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+                                     &local->transaction.new_parent_loc,
+                                     local->transaction.new_basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        nlockee++;
+        ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        nlockee++;
+        if (local->newloc.inode && IA_ISDIR (local->newloc.inode->ia_type)) {
+                ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+                                             &local->newloc,
+                                             NULL,
+                                             priv->child_count);
+                if (ret)
+                        goto out;
+
+                nlockee++;
+        }
+        qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
+               afr_entry_lockee_cmp);
+        int_lock->lockee_count = nlockee;
+
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
 
-		AFR_STACK_UNWIND (rename, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL, NULL);
-	}
+                AFR_STACK_UNWIND (rename, frame, -1, op_errno,
+                                  NULL, NULL, NULL, NULL, NULL, NULL);
+        }
 
-	return 0;
+        return 0;
 }
 
 /* }}} */
@@ -1580,201 +1560,190 @@ out:
 int
 afr_unlink_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                local->cont.unlink.preparent.ia_ino  = local->cont.unlink.parent_ino;
-                local->cont.unlink.postparent.ia_ino = local->cont.unlink.parent_ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame) {
+                        main_frame = local->transaction.main_frame;
+                }
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-		AFR_STACK_UNWIND (unlink, main_frame,
+        if (main_frame) {
+                AFR_STACK_UNWIND (unlink, main_frame,
                                   local->op_ret, local->op_errno,
-                                  &local->cont.unlink.preparent,
-                                  &local->cont.unlink.postparent);
+                                  &local->cont.dir_fop.preparent,
+                                  &local->cont.dir_fop.postparent,
+                                  NULL);
         }
 
-	return 0;
+        return 0;
 }
 
 
 int
-afr_unlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		     int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                     struct iatt *postparent)
+afr_unlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
+        afr_local_t *   local = NULL;
+        int call_count  = -1;
+        int child_index = (long) cookie;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
+        LOCK (&frame->lock);
+        {
                 if (child_index == local->read_child_index) {
                         local->read_child_returned = _gf_true;
                 }
+                __dir_entry_fop_common_cbk (frame, child_index, this,
+                                            op_ret, op_errno, NULL, NULL,
+                                            preparent, postparent, NULL, NULL);
+        }
+        UNLOCK (&frame->lock);
 
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
+        call_count = afr_frame_return (frame);
+        if (call_count == 0)
+                afr_dir_fop_done (frame, this);
 
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret   = op_ret;
-                                local->cont.unlink.preparent  = *preparent;
-                                local->cont.unlink.postparent = *postparent;
-			}
+        return 0;
+}
 
-                        if (child_index == local->read_child_index) {
-                                local->cont.unlink.preparent  = *preparent;
-                                local->cont.unlink.postparent = *postparent;
-                        }
 
-			local->success_count++;
-		}
+int32_t
+afr_unlink_wind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
+
+        local = frame->local;
+        priv  = this->private;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-	call_count = afr_frame_return (frame);
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+        local->call_count = call_count;
 
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
-}
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_unlink_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->unlink,
+                                           &local->loc, local->xflag,
+                                           local->xdata_req);
 
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-int32_t
-afr_unlink_wind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv  = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_unlink_wind_cbk,	
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->unlink,
-					   &local->loc);
-			
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_unlink_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t * local = frame->local;
+        afr_local_t * local = frame->local;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
-	
-	return 0;
+        AFR_STACK_DESTROY (frame);
+
+        return 0;
 }
 
 
 int32_t
 afr_unlink (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc)
+            loc_t *loc, int xflag, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
-
-	int ret = -1;
+        afr_private_t           *priv                   = NULL;
+        afr_local_t             *local                  = NULL;
+        afr_internal_lock_t     *int_lock               = NULL;
+        call_frame_t            *transaction_frame      = NULL;
+        int                     ret                     = -1;
+        int                     op_errno                = 0;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        priv = this->private;
 
-	priv = this->private;
+        QUORUM_CHECK(unlink,out);
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
-
-	transaction_frame->local = local;
-
-	loc_copy (&local->loc, loc);
-
-        if (loc->parent)
-                local->cont.unlink.parent_ino = loc->parent->ino;
-
-	local->transaction.fop    = afr_unlink_wind;
-	local->transaction.done   = afr_unlink_done;
-	local->transaction.unwind = afr_unlink_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
+
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
+
+        loc_copy (&local->loc, loc);
+        local->xflag = xflag;
+        if (xdata)
+                local->xdata_req = dict_ref (xdata);
+
+        local->op = GF_FOP_UNLINK;
+        local->transaction.fop    = afr_unlink_wind;
+        local->transaction.done   = afr_unlink_done;
+        local->transaction.unwind = afr_unlink_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame = frame;
+        local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (unlink, frame, op_ret, op_errno,
-                                  NULL, NULL);
-	}
-
-	return 0;
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (unlink, frame, -1, op_errno,
+                                  NULL, NULL, NULL);
+        }
+
+        return 0;
 }
 
 /* }}} */
@@ -1786,203 +1755,208 @@ out:
 int
 afr_rmdir_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-
-	local = frame->local;
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        local = frame->local;
 
-	if (main_frame) {
-                local->cont.rmdir.preparent.ia_ino  = local->cont.rmdir.parent_ino;
-                local->cont.rmdir.postparent.ia_ino = local->cont.rmdir.parent_ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame) {
+                        main_frame = local->transaction.main_frame;
+                }
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-		AFR_STACK_UNWIND (rmdir, main_frame,
+        if (main_frame) {
+                AFR_STACK_UNWIND (rmdir, main_frame,
                                   local->op_ret, local->op_errno,
-                                  &local->cont.rmdir.preparent,
-                                  &local->cont.rmdir.postparent);
+                                  &local->cont.dir_fop.preparent,
+                                  &local->cont.dir_fop.postparent,
+                                  NULL);
         }
 
-	return 0;
+        return 0;
 }
 
 
 int
-afr_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		    int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                    struct iatt *postparent)
+afr_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                    struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
+        afr_local_t *   local = NULL;
+        int call_count  = -1;
+        int child_index = (long) cookie;
         int read_child  = 0;
 
-	local = frame->local;
-	priv = this->private;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
+        LOCK (&frame->lock);
+        {
                 if (child_index == read_child) {
                         local->read_child_returned = _gf_true;
                 }
+                if (afr_fop_failed (op_ret, op_errno) && (op_errno != ENOTEMPTY))
+                        afr_transaction_fop_failed (frame, this, child_index);
+                local->op_errno = op_errno;
+                local->child_errno[child_index] = op_errno;
+                if (op_ret > -1)
+                        __dir_entry_fop_common_cbk (frame, child_index, this,
+                                                   op_ret, op_errno, NULL, NULL,
+                                                   preparent, postparent, NULL,
+                                                   NULL);
 
-		if (afr_fop_failed (op_ret, op_errno) && (op_errno != ENOTEMPTY))
-			afr_transaction_fop_failed (frame, this, child_index);
+        }
+        UNLOCK (&frame->lock);
 
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-                                local->cont.rmdir.preparent  = *preparent;
-                                local->cont.rmdir.postparent = *postparent;
+        call_count = afr_frame_return (frame);
+        if (call_count == 0)
+                afr_dir_fop_done (frame, this);
 
-			}
+        return 0;
+}
 
-                        if (child_index == read_child) {
-                                local->cont.rmdir.preparent  = *preparent;
-                                local->cont.rmdir.postparent = *postparent;
-                        }
 
-			local->success_count++;
-		}
+int
+afr_rmdir_wind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        local = frame->local;
+        priv  = this->private;
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
-}
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
+        local->call_count = call_count;
 
-int
-afr_rmdir_wind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv  = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_rmdir_wind_cbk,	
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->rmdir,
-					   &local->loc);
-
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_rmdir_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->rmdir,
+                                           &local->loc, local->cont.rmdir.flags,
+                                           NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
 }
 
 
 int
 afr_rmdir_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t * local = frame->local;
+        afr_local_t * local = frame->local;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
-	
-	return 0;
+        AFR_STACK_DESTROY (frame);
+
+        return 0;
 }
 
 
 int
 afr_rmdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc)
+           loc_t *loc, int flags, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
-	
-	int ret = -1;
-
-	int op_ret   = -1;
-	int op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-
-	priv = this->private;
-
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
-
-	transaction_frame->local = local;
-
-	loc_copy (&local->loc, loc);
-
-        if (loc->parent)
-                local->cont.rmdir.parent_ino = loc->parent->ino;
-
-	local->transaction.fop    = afr_rmdir_wind;
-	local->transaction.done   = afr_rmdir_done;
-	local->transaction.unwind = afr_rmdir_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
+        afr_private_t           *priv                   = NULL;
+        afr_local_t             *local                  = NULL;
+        afr_internal_lock_t     *int_lock               = NULL;
+        call_frame_t            *transaction_frame      = NULL;
+        int                     ret                     = -1;
+        int                     op_errno                = 0;
+        int                     nlockee                 = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        priv = this->private;
+
+        QUORUM_CHECK(rmdir,out);
+
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
+
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
+
+        local->cont.rmdir.flags = flags;
+        loc_copy (&local->loc, loc);
+
+        local->op = GF_FOP_RMDIR;
+        local->transaction.fop    = afr_rmdir_wind;
+        local->transaction.done   = afr_rmdir_done;
+        local->transaction.unwind = afr_rmdir_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame = frame;
+        local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = nlockee = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        nlockee++;
+        ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+                                     &local->loc,
+                                     NULL,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        nlockee++;
+        qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
+               afr_entry_lockee_cmp);
+        int_lock->lockee_count = nlockee;
+
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (rmdir, frame, op_ret, op_errno,
-                                  NULL, NULL);
-	}
-
-	return 0;
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
+        }
+
+        return 0;
 }
 
 /* }}} */
-
diff --git a/xlators/cluster/afr/src/afr-dir-write.h b/xlators/cluster/afr/src/afr-dir-write.h
index e2ed8c759..02f0a3682 100644
--- a/xlators/cluster/afr/src/afr-dir-write.h
+++ b/xlators/cluster/afr/src/afr-dir-write.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __DIR_WRITE_H__
@@ -23,38 +14,34 @@
 int32_t
 afr_create (call_frame_t *frame, xlator_t *this,
             loc_t *loc, int32_t flags, mode_t mode,
-            fd_t *fd, dict_t *params);
+            mode_t umask, fd_t *fd, dict_t *xdata);
 
 int32_t
 afr_mknod (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dev_t dev, dict_t *params);
+	   loc_t *loc, mode_t mode, dev_t dev, mode_t umask, dict_t *xdata);
 
 int32_t
 afr_mkdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dict_t *params);
+	   loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata);
 
 int32_t
 afr_unlink (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc);
+	    loc_t *loc, int xflag, dict_t *xdata);
 
 int32_t
 afr_rmdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc);
+	   loc_t *loc, int flags, dict_t *xdata);
 
 int32_t
 afr_link (call_frame_t *frame, xlator_t *this,
-	  loc_t *oldloc, loc_t *newloc);
+	  loc_t *oldloc, loc_t *newloc, dict_t *xdata);
 
 int32_t
 afr_rename (call_frame_t *frame, xlator_t *this,
-	    loc_t *oldloc, loc_t *newloc);
+	    loc_t *oldloc, loc_t *newloc, dict_t *xdata);
 
 int
 afr_symlink (call_frame_t *frame, xlator_t *this,
-	     const char *linkpath, loc_t *oldloc, dict_t *params);
-
-int32_t
-afr_setdents (call_frame_t *frame, xlator_t *this,
-	      fd_t *fd, int32_t flags, dir_entry_t *entries, int32_t count);
+	     const char *linkpath, loc_t *oldloc, mode_t umask, dict_t *params);
 
 #endif /* __DIR_WRITE_H__ */
diff --git a/xlators/cluster/afr/src/afr-inode-read.c b/xlators/cluster/afr/src/afr-inode-read.c
index 8d6e12a93..e06e3b2f2 100644
--- a/xlators/cluster/afr/src/afr-inode-read.c
+++ b/xlators/cluster/afr/src/afr-inode-read.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -44,9 +35,6 @@
 #include "compat-errno.h"
 #include "compat.h"
 
-#include "afr.h"
-
-
 /**
  * Common algorithm for inode read calls:
  *
@@ -61,114 +49,115 @@
 
 int32_t
 afr_access_cbk (call_frame_t *frame, void *cookie,
-		xlator_t *this, int32_t op_ret, int32_t op_errno)
+                xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
-
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-        int read_child = -1;
+        afr_private_t * priv            = NULL;
+        afr_local_t *   local           = NULL;
+        xlator_t **     children        = NULL;
+        int             unwind          = 1;
+        int32_t         *last_index     = NULL;
+        int32_t         next_call_child = -1;
+        int32_t         read_child      = -1;
+        int32_t         *fresh_children  = NULL;
 
-	priv     = this->private;
-	children = priv->children;
+        priv     = this->private;
+        children = priv->children;
 
-	local = frame->local;
+        local = frame->local;
 
         read_child = (long) cookie;
 
-	if (op_ret == -1) {
-        retry:
-		last_tried = local->cont.access.last_tried;
-
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try    = ++local->cont.access.last_tried;
-
-                if (this_try == read_child) {
-                        goto retry;
-                }
+        if (op_ret == -1) {
+                last_index = &local->cont.access.last_index;
+                fresh_children = local->fresh_children;
+                next_call_child = afr_next_call_child (fresh_children,
+                                                       local->child_up,
+                                                       priv->child_count,
+                                                       last_index, read_child);
+                if (next_call_child < 0)
+                        goto out;
 
-		unwind = 0;
+                unwind = 0;
 
-		STACK_WIND_COOKIE (frame, afr_access_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->access,
-				   &local->loc, local->cont.access.mask);
-	}
+                STACK_WIND_COOKIE (frame, afr_access_cbk,
+                                   (void *) (long) read_child,
+                                   children[next_call_child],
+                                   children[next_call_child]->fops->access,
+                                   &local->loc, local->cont.access.mask,
+                                   NULL);
+        }
 
 out:
-	if (unwind) {
-		AFR_STACK_UNWIND (access, frame, op_ret, op_errno);
-	}
+        if (unwind) {
+                AFR_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
-afr_access (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t mask)
+afr_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+            dict_t *xdata)
 {
-	afr_private_t * priv       = NULL;
-	xlator_t **     children   = NULL;
-	int             call_child = 0;
-	afr_local_t     *local     = NULL;
-
+        afr_private_t   *priv      = NULL;
+        xlator_t        **children = NULL;
+        int             call_child = 0;
+        afr_local_t     *local     = NULL;
+        int32_t         op_errno   = 0;
         int32_t         read_child = -1;
+        int             ret        = -1;
 
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        priv     = this->private;
+        VALIDATE_OR_GOTO (priv->children, out);
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+        children = priv->children;
 
-	children = priv->children;
+        AFR_SBRAIN_CHECK_LOC (loc, out);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-        read_child = afr_read_child (this, loc->inode);
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-        if ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
+        local->fresh_children = afr_children_create (priv->child_count);
+        if (!local->fresh_children) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-                local->cont.access.last_tried = -1;
 
-        } else {
-                call_child = afr_first_up_child (priv);
-                if (call_child == -1) {
-                        op_errno = ENOTCONN;
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "no child is up");
-                        goto out;
-                }
-
-                local->cont.access.last_tried = call_child;
+        read_child = afr_inode_get_read_ctx (this, loc->inode,
+                                             local->fresh_children);
+        ret = afr_get_call_child (this, local->child_up, read_child,
+                                     local->fresh_children,
+                                     &call_child,
+                                     &local->cont.access.last_index);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto out;
         }
 
-	loc_copy (&local->loc, loc);
-	local->cont.access.mask       = mask;
+        loc_copy (&local->loc, loc);
+        local->cont.access.mask = mask;
 
-	STACK_WIND_COOKIE (frame, afr_access_cbk,
-			   (void *) (long) call_child,
-			   children[call_child], children[call_child]->fops->access,
-			   loc, mask);
+        STACK_WIND_COOKIE (frame, afr_access_cbk,
+                           (void *) (long) call_child,
+                           children[call_child],
+                           children[call_child]->fops->access,
+                           loc, mask, xdata);
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (access, frame, op_ret, op_errno);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (access, frame, -1, op_errno, NULL);
+        return 0;
 }
 
 
@@ -178,121 +167,111 @@ out:
 
 int32_t
 afr_stat_cbk (call_frame_t *frame, void *cookie,
-	      xlator_t *this, int32_t op_ret, int32_t op_errno,
-	      struct iatt *buf)
+              xlator_t *this, int32_t op_ret, int32_t op_errno,
+              struct iatt *buf, dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
-
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-	int read_child = -1;
-
-	priv     = this->private;
-	children = priv->children;
-
-	read_child = (long) cookie;
-
-	local = frame->local;
+        afr_private_t * priv            = NULL;
+        afr_local_t *   local           = NULL;
+        xlator_t **     children        = NULL;
+        int             unwind          = 1;
+        int32_t         *last_index     = NULL;
+        int32_t         next_call_child = -1;
+        int32_t         read_child      = -1;
+        int32_t         *fresh_children  = NULL;
+
+        priv     = this->private;
+        children = priv->children;
 
-	if (op_ret == -1) {
-	retry:
-		last_tried = local->cont.stat.last_tried;
+        read_child = (long) cookie;
 
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try = ++local->cont.stat.last_tried;
+        local = frame->local;
 
-		if (this_try == read_child) {
-			goto retry;
-		}
+        if (op_ret == -1) {
+                last_index = &local->cont.stat.last_index;
+                fresh_children = local->fresh_children;
+                next_call_child = afr_next_call_child (fresh_children,
+                                                       local->child_up,
+                                                       priv->child_count,
+                                                       last_index, read_child);
+                if (next_call_child < 0)
+                        goto out;
 
-		unwind = 0;
+                unwind = 0;
 
-		STACK_WIND_COOKIE (frame, afr_stat_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->stat,
-				   &local->loc);
-	}
+                STACK_WIND_COOKIE (frame, afr_stat_cbk,
+                                   (void *) (long) read_child,
+                                   children[next_call_child],
+                                   children[next_call_child]->fops->stat,
+                                   &local->loc, NULL);
+        }
 
 out:
-	if (unwind) {
-		if (buf)
-			buf->ia_ino = local->cont.stat.ino;
-
-		AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, buf);
-	}
+        if (unwind) {
+                AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
-afr_stat (call_frame_t *frame, xlator_t *this,
-	  loc_t *loc)
+afr_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-	afr_private_t * priv       = NULL;
-	afr_local_t   * local      = NULL;
-	xlator_t **     children   = NULL;
-
+        afr_private_t   *priv      = NULL;
+        afr_local_t     *local     = NULL;
+        xlator_t        **children = NULL;
+        int             call_child = 0;
+        int32_t         op_errno   = 0;
         int32_t         read_child = -1;
-	int             call_child = 0;
-
-	int32_t         op_ret     = -1;
-	int32_t         op_errno   = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
-
-	children = priv->children;
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        int             ret        = -1;
 
-	frame->local = local;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-        read_child = afr_read_child (this, loc->inode);
+        priv     = this->private;
+        VALIDATE_OR_GOTO (priv->children, out);
 
-        if ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
+        children = priv->children;
 
-                local->cont.stat.last_tried = -1;
+        AFR_SBRAIN_CHECK_LOC (loc, out);
 
-        } else {
-		call_child = afr_first_up_child (priv);
-		if (call_child == -1) {
-			op_errno = ENOTCONN;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"no child is up");
-			goto out;
-		}
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-		local->cont.stat.last_tried = call_child;
-	}
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	loc_copy (&local->loc, loc);
+        local->fresh_children = afr_children_create (priv->child_count);
+        if (!local->fresh_children) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	local->cont.stat.ino = loc->inode->ino;
+        read_child = afr_inode_get_read_ctx (this, loc->inode,
+                                             local->fresh_children);
+        ret = afr_get_call_child (this, local->child_up, read_child,
+                                     local->fresh_children,
+                                     &call_child,
+                                     &local->cont.stat.last_index);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto out;
+        }
+        loc_copy (&local->loc, loc);
 
-	STACK_WIND_COOKIE (frame, afr_stat_cbk, (void *) (long) call_child,
-			   children[call_child],
-			   children[call_child]->fops->stat,
-			   loc);
+        STACK_WIND_COOKIE (frame, afr_stat_cbk, (void *) (long) call_child,
+                           children[call_child],
+                           children[call_child]->fops->stat,
+                           loc, xdata);
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, NULL);
-	}
+        if (ret < 0)
+                AFR_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
@@ -301,124 +280,122 @@ out:
 /* {{{ fstat */
 
 int32_t
-afr_fstat_cbk (call_frame_t *frame, void *cookie,
-	       xlator_t *this, int32_t op_ret, int32_t op_errno,
-	       struct iatt *buf)
+afr_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, struct iatt *buf,
+               dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
-
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-        int read_child = -1;
-
-	priv     = this->private;
-	children = priv->children;
-
-	local = frame->local;
+        afr_private_t   *priv           = NULL;
+        afr_local_t     *local          = NULL;
+        xlator_t        **children      = NULL;
+        int             unwind          = 1;
+        int32_t         *last_index     = NULL;
+        int32_t         next_call_child = -1;
+        int32_t         read_child      = -1;
+        int32_t         *fresh_children  = NULL;
 
-	read_child = (long) cookie;
+        priv     = this->private;
+        children = priv->children;
 
-	if (op_ret == -1) {
-	retry:
-		last_tried = local->cont.fstat.last_tried;
+        local = frame->local;
 
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try   = ++local->cont.fstat.last_tried;
+        read_child = (long) cookie;
 
-		if (this_try == read_child) {
-			goto retry;
-		}
+        if (op_ret == -1) {
+                last_index = &local->cont.fstat.last_index;
+                fresh_children = local->fresh_children;
+                next_call_child = afr_next_call_child (fresh_children,
+                                                       local->child_up,
+                                                       priv->child_count,
+                                                       last_index, read_child);
+                if (next_call_child < 0)
+                        goto out;
 
-		unwind = 0;
+                unwind = 0;
 
-		STACK_WIND_COOKIE (frame, afr_fstat_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->fstat,
-				   local->fd);
-	}
+                STACK_WIND_COOKIE (frame, afr_fstat_cbk,
+                                   (void *) (long) read_child,
+                                   children[next_call_child],
+                                   children[next_call_child]->fops->fstat,
+                                   local->fd, NULL);
+        }
 
 out:
-	if (unwind) {
-		if (buf)
-			buf->ia_ino = local->cont.fstat.ino;
-
-		AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf);
-	}
+        if (unwind) {
+                AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_fstat (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd)
+           fd_t *fd, dict_t *xdata)
 {
-	afr_private_t * priv       = NULL;
-	afr_local_t   * local      = NULL;
-	xlator_t **     children   = NULL;
+        afr_private_t   *priv      = NULL;
+        afr_local_t     *local     = NULL;
+        xlator_t        **children = NULL;
+        int             call_child = 0;
+        int32_t         op_errno   = 0;
+        int32_t         read_child = 0;
+        int             ret        = -1;
 
-	int             call_child = 0;
-        int32_t         read_child = -1;
-
-	int32_t         op_ret     = -1;
-	int32_t         op_errno   = 0;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (fd, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (fd, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        priv     = this->private;
+        VALIDATE_OR_GOTO (priv->children, out);
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+        children = priv->children;
 
-	children = priv->children;
+        VALIDATE_OR_GOTO (fd->inode, out);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        AFR_SBRAIN_CHECK_FD (fd, out);
 
-	frame->local = local;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	VALIDATE_OR_GOTO (fd->inode, out);
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-        read_child = afr_read_child (this, fd->inode);
+        local->fresh_children = afr_children_create (priv->child_count);
+        if (!local->fresh_children) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-        if ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
+        read_child = afr_inode_get_read_ctx (this, fd->inode,
+                                             local->fresh_children);
 
-                local->cont.fstat.last_tried = -1;
-        } else {
-                call_child = afr_first_up_child (priv);
 
-                if (call_child == -1) {
-                        op_errno = ENOTCONN;
-                        gf_log (this->name, GF_LOG_DEBUG,
-				"no child is up");
-			goto out;
-		}
 
-                local->cont.fstat.last_tried = call_child;
+        ret = afr_get_call_child (this, local->child_up, read_child,
+                                     local->fresh_children,
+                                     &call_child,
+                                     &local->cont.fstat.last_index);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto out;
         }
 
-	local->cont.fstat.ino = fd->inode->ino;
-	local->fd = fd_ref (fd);
+        local->fd = fd_ref (fd);
+
+        afr_open_fd_fix (fd, this);
 
-	STACK_WIND_COOKIE (frame, afr_fstat_cbk, (void *) (long) call_child,
-			   children[call_child],
-			   children[call_child]->fops->fstat,
-			   fd);
+        STACK_WIND_COOKIE (frame, afr_fstat_cbk, (void *) (long) call_child,
+                           children[call_child],
+                           children[call_child]->fops->fstat,
+                           fd, xdata);
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, NULL);
-	}
+        if (ret < 0)
+                AFR_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 /* }}} */
@@ -427,122 +404,115 @@ out:
 
 int32_t
 afr_readlink_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno,
-		  const char *buf, struct iatt *sbuf)
+                  xlator_t *this, int32_t op_ret, int32_t op_errno,
+                  const char *buf, struct iatt *sbuf, dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
-
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-        int read_child = -1;
+        afr_private_t * priv                  = NULL;
+        afr_local_t *   local                 = NULL;
+        xlator_t **     children              = NULL;
+        int             unwind                = 1;
+        int32_t         *last_index           = NULL;
+        int32_t         next_call_child       = -1;
+        int32_t         read_child            = -1;
+        int32_t         *fresh_children        = NULL;
 
-	priv     = this->private;
-	children = priv->children;
+        priv     = this->private;
+        children = priv->children;
 
-	local = frame->local;
+        local = frame->local;
 
         read_child = (long) cookie;
 
-	if (op_ret == -1) {
-        retry:
-		last_tried = local->cont.readlink.last_tried;
-
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try = ++local->cont.readlink.last_tried;
-
-                if (this_try == read_child) {
-                        goto retry;
-                }
+        if (op_ret == -1) {
+                last_index = &local->cont.readlink.last_index;
+                fresh_children = local->fresh_children;
+                next_call_child = afr_next_call_child (fresh_children,
+                                                       local->child_up,
+                                                       priv->child_count,
+                                                       last_index, read_child);
+                if (next_call_child < 0)
+                        goto out;
 
-		unwind = 0;
-		STACK_WIND_COOKIE (frame, afr_readlink_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->readlink,
-				   &local->loc,
-				   local->cont.readlink.size);
-	}
+                unwind = 0;
+                STACK_WIND_COOKIE (frame, afr_readlink_cbk,
+                                   (void *) (long) read_child,
+                                   children[next_call_child],
+                                   children[next_call_child]->fops->readlink,
+                                   &local->loc,
+                                   local->cont.readlink.size, NULL);
+        }
 
 out:
-	if (unwind) {
-                if (sbuf)
-                        sbuf->ia_ino = local->cont.readlink.ino;
-
-		AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno, buf, sbuf);
-	}
+        if (unwind) {
+                AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno, buf, sbuf,
+                                  xdata);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_readlink (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, size_t size)
+              loc_t *loc, size_t size, dict_t *xdata)
 {
-	afr_private_t * priv       = NULL;
-	xlator_t **     children   = NULL;
-	int             call_child = 0;
-	afr_local_t     *local     = NULL;
-
+        afr_private_t   *priv      = NULL;
+        xlator_t        **children = NULL;
+        int             call_child = 0;
+        afr_local_t     *local     = NULL;
+        int32_t         op_errno   = 0;
         int32_t         read_child = -1;
+        int             ret        = -1;
 
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
-
-	children = priv->children;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        priv     = this->private;
+        VALIDATE_OR_GOTO (priv->children, out);
 
-	frame->local = local;
+        children = priv->children;
 
-        read_child = afr_read_child (this, loc->inode);
+        AFR_SBRAIN_CHECK_LOC (loc, out);
 
-        if ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-                local->cont.readlink.last_tried = -1;
-
-        } else {
-                call_child = afr_first_up_child (priv);
-
-                if (call_child == -1) {
-                        op_errno = ENOTCONN;
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "no child is up");
-                        goto out;
-                }
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-                local->cont.readlink.last_tried = call_child;
+        local->fresh_children = afr_children_create (priv->child_count);
+        if (!local->fresh_children) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+        read_child = afr_inode_get_read_ctx (this, loc->inode,
+                                             local->fresh_children);
+        ret = afr_get_call_child (this, local->child_up, read_child,
+                                     local->fresh_children,
+                                     &call_child,
+                                     &local->cont.readlink.last_index);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto out;
         }
 
-	loc_copy (&local->loc, loc);
+        loc_copy (&local->loc, loc);
 
-	local->cont.readlink.size       = size;
-        local->cont.readlink.ino        = loc->inode->ino;
+        local->cont.readlink.size       = size;
 
-	STACK_WIND_COOKIE (frame, afr_readlink_cbk,
-			   (void *) (long) call_child,
-			   children[call_child], children[call_child]->fops->readlink,
-			   loc, size);
+        STACK_WIND_COOKIE (frame, afr_readlink_cbk,
+                           (void *) (long) call_child,
+                           children[call_child],
+                           children[call_child]->fops->readlink,
+                           loc, size, xdata);
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno, NULL, NULL);
-	}
-	return 0;
+        if (ret < 0)
+                AFR_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
 }
 
 
@@ -556,7 +526,7 @@ struct _xattr_key {
 };
 
 
-void
+int
 __gather_xattr_keys (dict_t *dict, char *key, data_t *value,
                      void *data)
 {
@@ -568,23 +538,23 @@ __gather_xattr_keys (dict_t *dict, char *key, data_t *value,
 
                 xkey = GF_CALLOC (1, sizeof (*xkey), gf_afr_mt_xattr_key);
                 if (!xkey)
-                        return;
+                        return -1;
 
                 xkey->key = key;
                 INIT_LIST_HEAD (&xkey->list);
 
                 list_add_tail (&xkey->list, list);
         }
+        return 0;
 }
 
 
 void
 __filter_xattrs (dict_t *dict)
 {
-        struct list_head keys;
-
-        struct _xattr_key *key;
-        struct _xattr_key *tmp;
+        struct list_head   keys = {0,};
+        struct _xattr_key *key  = NULL;
+        struct _xattr_key *tmp  = NULL;
 
         INIT_LIST_HEAD (&keys);
 
@@ -604,129 +574,1222 @@ __filter_xattrs (dict_t *dict)
 
 int32_t
 afr_getxattr_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno,
-		  dict_t *dict)
+                  xlator_t *this, int32_t op_ret, int32_t op_errno,
+                  dict_t *dict, dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
+        afr_private_t * priv            = NULL;
+        afr_local_t *   local           = NULL;
+        xlator_t **     children        = NULL;
+        int             unwind          = 1;
+        int32_t         *last_index     = NULL;
+        int32_t         next_call_child = -1;
+        int32_t         read_child      = -1;
+        int32_t         *fresh_children  = NULL;
 
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-        int read_child = -1;
+        priv     = this->private;
+        children = priv->children;
 
-	priv     = this->private;
-	children = priv->children;
-
-	local = frame->local;
+        local = frame->local;
 
         read_child = (long) cookie;
 
-	if (op_ret == -1) {
-        retry:
-		last_tried = local->cont.getxattr.last_tried;
-
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try = ++local->cont.getxattr.last_tried;
-
-                if (this_try == read_child) {
-                        goto retry;
-                }
+        if (op_ret == -1) {
+                last_index = &local->cont.getxattr.last_index;
+                fresh_children = local->fresh_children;
+                next_call_child = afr_next_call_child (fresh_children,
+                                                       local->child_up,
+                                                       priv->child_count,
+                                                       last_index, read_child);
+                if (next_call_child < 0)
+                        goto out;
 
-		unwind = 0;
-		STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->getxattr,
-				   &local->loc,
-				   local->cont.getxattr.name);
-	}
+                unwind = 0;
+                STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
+                                   (void *) (long) read_child,
+                                   children[next_call_child],
+                                   children[next_call_child]->fops->getxattr,
+                                   &local->loc,
+                                   local->cont.getxattr.name,
+                                   NULL);
+        }
 
 out:
-	if (unwind) {
+        if (unwind) {
                 if (op_ret >= 0 && dict)
                         __filter_xattrs (dict);
 
-		AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
-	}
+                AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+        }
 
-	return 0;
+        return 0;
 }
 
 int32_t
-afr_getxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, const char *name)
+afr_getxattr_unwind (call_frame_t *frame, int op_ret, int op_errno,
+                     dict_t *dict, dict_t *xdata)
+
+{
+        AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+int32_t
+afr_fgetxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
+                         xlator_t *this, int32_t op_ret, int32_t op_errno,
+                         dict_t *dict, dict_t *xdata)
 {
-	afr_private_t *   priv       = NULL;
-	xlator_t **       children   = NULL;
-	int               call_child = 0;
-	afr_local_t     * local      = NULL;
+        afr_local_t     *local                  = NULL;
+        afr_private_t   *priv                   = NULL;
+        xlator_t        **children              = NULL;
+        dict_t          *xattr                  = NULL;
+        char            *tmp_report             = NULL;
+        char            lk_summary[1024]        = {0,};
+        int             serz_len                = 0;
+        int32_t         callcnt                 = 0;
+        long int        cky                     = 0;
+        int             ret                     = 0;
+
+        priv     = this->private;
+        children = priv->children;
+
+        local = frame->local;
+        cky = (long) cookie;
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
+                if (op_ret == -1)
+                        local->child_errno[cky] = op_errno;
+
+                if (!local->dict)
+                        local->dict = dict_new ();
+                if (local->dict) {
+                        ret = dict_get_str (dict, local->cont.getxattr.name,
+                                            &tmp_report);
+                        if (ret)
+                                goto unlock;
+                        ret = dict_set_dynstr (local->dict,
+                                               children[cky]->name,
+                                               gf_strdup (tmp_report));
+                        if (ret)
+                                goto unlock;
+                }
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                xattr = dict_new ();
+                if (!xattr) {
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        goto unwind;
+                }
+                ret = dict_serialize_value_with_delim (local->dict,
+                                                       lk_summary,
+                                                       &serz_len, '\n');
+                if (ret) {
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Error serializing dictionary");
+                        goto unwind;
+                }
+                if (serz_len == -1)
+                        snprintf (lk_summary, sizeof (lk_summary),
+                                  "No locks cleared.");
+                ret = dict_set_dynstr (xattr, local->cont.getxattr.name,
+                                       gf_strdup (lk_summary));
+                if (ret) {
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Error setting dictionary");
+                        goto unwind;
+                }
+
+        unwind:
+                // Updating child_errno with more recent 'events'
+                local->child_errno[cky] = op_errno;
+                op_errno = afr_resultant_errno_get (NULL, local->child_errno,
+                                                    priv->child_count);
+                AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, xattr,
+                                  xdata);
+
+                if (xattr)
+                        dict_unref (xattr);
+        }
+
+        return ret;
+}
+
+int32_t
+afr_getxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
+                        xlator_t *this, int32_t op_ret, int32_t op_errno,
+                        dict_t *dict, dict_t *xdata)
+{
+        afr_local_t     *local                  = NULL;
+        afr_private_t   *priv                   = NULL;
+        xlator_t        **children              = NULL;
+        dict_t          *xattr                  = NULL;
+        char            *tmp_report             = NULL;
+        char            lk_summary[1024]        = {0,};
+        int             serz_len                = 0;
+        int32_t         callcnt                 = 0;
+        long int        cky                     = 0;
+        int             ret                     = 0;
+
+        priv     = this->private;
+        children = priv->children;
+
+        local = frame->local;
+        cky = (long) cookie;
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
+                if (op_ret == -1)
+                        local->child_errno[cky] = op_errno;
+
+                if (!local->dict)
+                        local->dict = dict_new ();
+                if (local->dict) {
+                        ret = dict_get_str (dict, local->cont.getxattr.name,
+                                            &tmp_report);
+                        if (ret)
+                                goto unlock;
+                        ret = dict_set_dynstr (local->dict,
+                                               children[cky]->name,
+                                               gf_strdup (tmp_report));
+                        if (ret)
+                                goto unlock;
+                }
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                xattr = dict_new ();
+                if (!xattr) {
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        goto unwind;
+                }
+                ret = dict_serialize_value_with_delim (local->dict,
+                                                       lk_summary,
+                                                       &serz_len, '\n');
+                if (ret) {
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Error serializing dictionary");
+                        goto unwind;
+                }
+                if (serz_len == -1)
+                        snprintf (lk_summary, sizeof (lk_summary),
+                                  "No locks cleared.");
+                ret = dict_set_dynstr (xattr, local->cont.getxattr.name,
+                                       gf_strdup (lk_summary));
+                if (ret) {
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Error setting dictionary");
+                        goto unwind;
+                }
+
+        unwind:
+                // Updating child_errno with more recent 'events'
+                local->child_errno[cky] = op_errno;
+                op_errno = afr_resultant_errno_get (NULL, local->child_errno,
+                                                    priv->child_count);
+                AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
+
+                if (xattr)
+                        dict_unref (xattr);
+        }
+
+        return ret;
+}
+
+/**
+ * node-uuid cbk uses next child querying mechanism
+ */
+int32_t
+afr_getxattr_node_uuid_cbk (call_frame_t *frame, void *cookie,
+                            xlator_t *this, int32_t op_ret, int32_t op_errno,
+                            dict_t *dict, dict_t *xdata)
+{
+        afr_private_t  *priv            = NULL;
+        afr_local_t    *local           = NULL;
+        xlator_t      **children        = NULL;
+        int             unwind          = 1;
+        int             curr_call_child = 0;
+
+        priv = this->private;
+        children = priv->children;
+
+        local = frame->local;
+
+        if (op_ret == -1) { /** query the _next_ child */
+
+                /**
+                 * _current_ becomes _next_
+                 * If done with all childs and yet no success; give up !
+                 */
+                curr_call_child = (int) ((long)cookie);
+                if (++curr_call_child == priv->child_count)
+                        goto unwind;
+
+                gf_log (this->name, GF_LOG_WARNING,
+                        "op_ret (-1): Re-querying afr-child (%d/%d)",
+                        curr_call_child, priv->child_count);
+
+                unwind = 0;
+                STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
+                                   (void *) (long) curr_call_child,
+                                   children[curr_call_child],
+                                   children[curr_call_child]->fops->getxattr,
+                                   &local->loc,
+                                   local->cont.getxattr.name,
+                                   NULL);
+        }
+
+ unwind:
+        if (unwind)
+                AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict,
+                                  NULL);
+
+        return 0;
+}
+
+int32_t
+afr_getxattr_lockinfo_cbk (call_frame_t *frame, void *cookie,
+                           xlator_t *this, int32_t op_ret, int32_t op_errno,
+                           dict_t *dict, dict_t *xdata)
+{
+        int          call_cnt     = 0, len = 0;
+        char        *lockinfo_buf = NULL;
+        dict_t      *lockinfo     = NULL, *newdict = NULL;
+        afr_local_t *local        = NULL;
+
+        LOCK (&frame->lock);
+        {
+                local = frame->local;
+
+                call_cnt = --local->call_count;
+
+                if ((op_ret < 0) || (!dict && !xdata)) {
+                        goto unlock;
+                }
+
+                if (xdata) {
+                        if (!local->xdata_rsp) {
+                                local->xdata_rsp = dict_new ();
+                                if (!local->xdata_rsp) {
+                                        local->op_ret = -1;
+                                        local->op_errno = ENOMEM;
+                                        goto unlock;
+                                }
+                        }
+                }
+
+                if (!dict) {
+                        goto unlock;
+                }
+
+                op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
+                                               (void **)&lockinfo_buf, &len);
+
+                if (!lockinfo_buf) {
+                        goto unlock;
+                }
+
+                if (!local->dict) {
+                        local->dict = dict_new ();
+                        if (!local->dict) {
+                                local->op_ret = -1;
+                                local->op_errno = ENOMEM;
+                                goto unlock;
+                        }
+                }
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        if (lockinfo_buf != NULL) {
+                lockinfo = dict_new ();
+                if (lockinfo == NULL) {
+                        local->op_ret = -1;
+                        local->op_errno = ENOMEM;
+                } else {
+                        op_ret = dict_unserialize (lockinfo_buf, len,
+                                                   &lockinfo);
+
+                        if (lockinfo && local->dict) {
+                                dict_copy (lockinfo, local->dict);
+                        }
+                }
+        }
+
+        if (xdata && local->xdata_rsp) {
+                dict_copy (xdata, local->xdata_rsp);
+        }
+
+        if (!call_cnt) {
+                newdict = dict_new ();
+                if (!newdict) {
+                        local->op_ret = -1;
+                        local->op_errno = ENOMEM;
+                        goto unwind;
+                }
+
+                len = dict_serialized_length (local->dict);
+                if (len == 0) {
+                        goto unwind;
+                }
+
+                lockinfo_buf = GF_CALLOC (1, len, gf_common_mt_char);
+                if (!lockinfo_buf) {
+                        local->op_ret = -1;
+                        local->op_errno = ENOMEM;
+                        goto unwind;
+                }
+
+                op_ret = dict_serialize (local->dict, lockinfo_buf);
+                if (op_ret < 0) {
+                        local->op_ret = -1;
+                        local->op_errno = -op_ret;
+                }
+
+                op_ret = dict_set_dynptr (newdict, GF_XATTR_LOCKINFO_KEY,
+                                          (void *)lockinfo_buf, len);
+                if (op_ret < 0) {
+                        local->op_ret = -1;
+                        local->op_errno = -op_ret;
+                        goto unwind;
+                }
+
+        unwind:
+                AFR_STACK_UNWIND (getxattr, frame, op_ret,
+                                  op_errno, newdict,
+                                  local->xdata_rsp);
+        }
+
+        dict_unref (lockinfo);
+
+        return 0;
+}
+
+int32_t
+afr_fgetxattr_lockinfo_cbk (call_frame_t *frame, void *cookie,
+                            xlator_t *this, int32_t op_ret, int32_t op_errno,
+                            dict_t *dict, dict_t *xdata)
+{
+        int          call_cnt     = 0, len = 0;
+        char        *lockinfo_buf = NULL;
+        dict_t      *lockinfo     = NULL, *newdict = NULL;
+        afr_local_t *local        = NULL;
+
+        LOCK (&frame->lock);
+        {
+                local = frame->local;
+
+                call_cnt = --local->call_count;
+
+                if ((op_ret < 0) || (!dict && !xdata)) {
+                        goto unlock;
+                }
+
+                if (xdata) {
+                        if (!local->xdata_rsp) {
+                                local->xdata_rsp = dict_new ();
+                                if (!local->xdata_rsp) {
+                                        local->op_ret = -1;
+                                        local->op_errno = ENOMEM;
+                                        goto unlock;
+                                }
+                        }
+                }
+
+                if (!dict) {
+                        goto unlock;
+                }
+
+                op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
+                                               (void **)&lockinfo_buf, &len);
+
+                if (!lockinfo_buf) {
+                        goto unlock;
+                }
+
+                if (!local->dict) {
+                        local->dict = dict_new ();
+                        if (!local->dict) {
+                                local->op_ret = -1;
+                                local->op_errno = ENOMEM;
+                                goto unlock;
+                        }
+                }
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        if (lockinfo_buf != NULL) {
+                lockinfo = dict_new ();
+                if (lockinfo == NULL) {
+                        local->op_ret = -1;
+                        local->op_errno = ENOMEM;
+                } else {
+                        op_ret = dict_unserialize (lockinfo_buf, len,
+                                                   &lockinfo);
+
+                        if (lockinfo && local->dict) {
+                                dict_copy (lockinfo, local->dict);
+                        }
+                }
+        }
+
+        if (xdata && local->xdata_rsp) {
+                dict_copy (xdata, local->xdata_rsp);
+        }
+
+        if (!call_cnt) {
+                newdict = dict_new ();
+                if (!newdict) {
+                        local->op_ret = -1;
+                        local->op_errno = ENOMEM;
+                        goto unwind;
+                }
+
+                len = dict_serialized_length (local->dict);
+                if (len <= 0) {
+                        goto unwind;
+                }
 
-        int               read_child = -1;
+                lockinfo_buf = GF_CALLOC (1, len, gf_common_mt_char);
+                if (!lockinfo_buf) {
+                        local->op_ret = -1;
+                        local->op_errno = ENOMEM;
+                        goto unwind;
+                }
 
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+                op_ret = dict_serialize (local->dict, lockinfo_buf);
+                if (op_ret < 0) {
+                        local->op_ret = -1;
+                        local->op_errno = -op_ret;
+                }
 
+                op_ret = dict_set_dynptr (newdict, GF_XATTR_LOCKINFO_KEY,
+                                          (void *)lockinfo_buf, len);
+                if (op_ret < 0) {
+                        local->op_ret = -1;
+                        local->op_errno = -op_ret;
+                        goto unwind;
+                }
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        unwind:
+                AFR_STACK_UNWIND (fgetxattr, frame, op_ret,
+                                  op_errno, newdict,
+                                  local->xdata_rsp);
+        }
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+        dict_unref (lockinfo);
 
-	children = priv->children;
+        return 0;
+}
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-	frame->local = local;
+int32_t
+afr_fgetxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
+                            xlator_t *this, int32_t op_ret, int32_t op_errno,
+                            dict_t *dict, dict_t *xdata)
+{
+        afr_local_t *local          = NULL;
+        int32_t      callcnt        = 0;
+        int          ret            = 0;
+        char        *xattr          = NULL;
+        char        *xattr_serz     = NULL;
+        char        xattr_cky[1024] = {0,};
+        dict_t      *nxattr         = NULL;
+        long         cky            = 0;
+        int32_t      padding        = 0;
+        int32_t      tlen           = 0;
+
+        if (!frame || !frame->local || !this) {
+                gf_log ("", GF_LOG_ERROR, "possible NULL deref");
+                goto out;
+        }
 
-        if (name) {
-                if (!strncmp (name, AFR_XATTR_PREFIX,
-                              strlen (AFR_XATTR_PREFIX))) {
+        local = frame->local;
+        cky = (long) cookie;
 
-                        op_errno = ENODATA;
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
+
+                if (!dict || (op_ret < 0))
                         goto out;
+
+                if (!local->dict)
+                        local->dict = dict_new ();
+
+                if (local->dict) {
+                        ret = dict_get_str (dict,
+                                            local->cont.getxattr.name,
+                                            &xattr);
+                        if (ret)
+                                goto out;
+
+                        xattr = gf_strdup (xattr);
+
+                        (void)snprintf (xattr_cky, 1024, "%s-%ld",
+                                        local->cont.getxattr.name, cky);
+                        ret = dict_set_dynstr (local->dict,
+                                               xattr_cky, xattr);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Cannot set xattr cookie key");
+                                goto out;
+                        }
+
+                        local->cont.getxattr.xattr_len
+                                += strlen (xattr) + 1;
+                }
+        }
+out:
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                if (!local->cont.getxattr.xattr_len)
+                        goto unwind;
+
+                nxattr = dict_new ();
+                if (!nxattr)
+                        goto unwind;
+
+                /* extra bytes for decorations (brackets and <>'s) */
+                padding += strlen (this->name)
+                        + strlen (AFR_PATHINFO_HEADER) + 4;
+                local->cont.getxattr.xattr_len += (padding + 2);
+
+                xattr_serz = GF_CALLOC (local->cont.getxattr.xattr_len,
+                                        sizeof (char), gf_common_mt_char);
+
+                if (!xattr_serz)
+                        goto unwind;
+
+                /* the xlator info */
+                (void) sprintf (xattr_serz, "(<"AFR_PATHINFO_HEADER"%s> ",
+                                this->name);
+
+                /* actual series of pathinfo */
+                ret = dict_serialize_value_with_delim (local->dict,
+                                                       xattr_serz
+                                                       + strlen (xattr_serz),
+                                                       &tlen, ' ');
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Error serializing"
+                                " dictionary");
+                        goto unwind;
+                }
+
+                /* closing part */
+                *(xattr_serz + padding + tlen) = ')';
+                *(xattr_serz + padding + tlen + 1) = '\0';
+
+                ret = dict_set_dynstr (nxattr, local->cont.getxattr.name,
+                                       xattr_serz);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR, "Cannot set pathinfo"
+                                " key in dict");
+
+        unwind:
+                AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, nxattr,
+                                  xdata);
+
+                if (nxattr)
+                        dict_unref (nxattr);
+        }
+
+        return ret;
+}
+
+int32_t
+afr_getxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
+                           xlator_t *this, int32_t op_ret, int32_t op_errno,
+                           dict_t *dict, dict_t *xdata)
+{
+        afr_local_t *local          = NULL;
+        int32_t      callcnt        = 0;
+        int          ret            = 0;
+        char        *xattr          = NULL;
+        char        *xattr_serz     = NULL;
+        char        xattr_cky[1024] = {0,};
+        dict_t      *nxattr         = NULL;
+        long         cky            = 0;
+        int32_t      padding        = 0;
+        int32_t      tlen           = 0;
+
+        if (!frame || !frame->local || !this) {
+                gf_log ("", GF_LOG_ERROR, "possible NULL deref");
+                goto out;
+        }
+
+        local = frame->local;
+        cky = (long) cookie;
+
+        LOCK (&frame->lock);
+                {
+                        callcnt = --local->call_count;
+
+                        if (!dict || (op_ret < 0))
+                                goto out;
+
+                        if (!local->dict)
+                                local->dict = dict_new ();
+
+                        if (local->dict) {
+                                ret = dict_get_str (dict,
+                                                    local->cont.getxattr.name,
+                                                    &xattr);
+                                if (ret)
+                                        goto out;
+
+                                xattr = gf_strdup (xattr);
+
+                                (void)snprintf (xattr_cky, 1024, "%s-%ld",
+                                                local->cont.getxattr.name, cky);
+                                ret = dict_set_dynstr (local->dict,
+                                                       xattr_cky, xattr);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Cannot set xattr cookie key");
+                                        goto out;
+                                }
+
+                                local->cont.getxattr.xattr_len += strlen (xattr) + 1;
+                        }
+                }
+ out:
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                if (!local->cont.getxattr.xattr_len)
+                        goto unwind;
+
+                nxattr = dict_new ();
+                if (!nxattr)
+                        goto unwind;
+
+                /* extra bytes for decorations (brackets and <>'s) */
+                padding += strlen (this->name) + strlen (AFR_PATHINFO_HEADER) + 4;
+                local->cont.getxattr.xattr_len += (padding + 2);
+
+                xattr_serz = GF_CALLOC (local->cont.getxattr.xattr_len,
+                                        sizeof (char), gf_common_mt_char);
+
+                if (!xattr_serz)
+                        goto unwind;
+
+                /* the xlator info */
+                (void) sprintf (xattr_serz, "(<"AFR_PATHINFO_HEADER"%s> ",
+                                this->name);
+
+                /* actual series of pathinfo */
+                ret = dict_serialize_value_with_delim (local->dict,
+                                                       xattr_serz + strlen (xattr_serz),
+                                                       &tlen, ' ');
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Error serializing"
+                                " dictionary");
+                        goto unwind;
+                }
+
+                /* closing part */
+                *(xattr_serz + padding + tlen) = ')';
+                *(xattr_serz + padding + tlen + 1) = '\0';
+
+                ret = dict_set_dynstr (nxattr, local->cont.getxattr.name,
+                                       xattr_serz);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR, "Cannot set pathinfo"
+                                " key in dict");
+
+        unwind:
+                AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, nxattr,
+                                  xdata);
+
+                if (nxattr)
+                        dict_unref (nxattr);
+        }
+
+        return ret;
+}
+
+static int
+afr_aggregate_stime_xattr (dict_t *this, char *key, data_t *value, void *data)
+{
+        int ret = 0;
+
+        if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0)
+                ret = gf_get_min_stime (THIS, data, key, value);
+
+        return ret;
+}
+
+int32_t
+afr_common_getxattr_stime_cbk (call_frame_t *frame, void *cookie,
+                               xlator_t *this, int32_t op_ret, int32_t op_errno,
+                               dict_t *dict, dict_t *xdata)
+{
+        afr_local_t *local          = NULL;
+        int32_t      callcnt        = 0;
+
+        if (!frame || !frame->local || !this) {
+                gf_log ("", GF_LOG_ERROR, "possible NULL deref");
+                goto out;
+        }
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
+
+                if (!dict || (op_ret < 0)) {
+                        local->op_errno = op_errno;
+                        goto cleanup;
                 }
 
+                if (!local->dict)
+                        local->dict = dict_copy_with_ref (dict, NULL);
+                else
+                        dict_foreach (dict, afr_aggregate_stime_xattr,
+                                      local->dict);
+                local->op_ret = 0;
+        }
+
+cleanup:
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
+                                  local->op_errno, local->dict, xdata);
         }
 
-        read_child = afr_read_child (this, loc->inode);
+out:
+        return 0;
+}
 
-        if ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
 
-                local->cont.getxattr.last_tried = -1;
+static gf_boolean_t
+afr_is_special_xattr (const char *name, fop_getxattr_cbk_t *cbk,
+                      gf_boolean_t is_fgetxattr)
+{
+        gf_boolean_t    is_spl = _gf_true;
+
+        GF_ASSERT (cbk);
+        if (!cbk) {
+                is_spl = _gf_false;
+                goto out;
+        }
+
+        if (!strcmp (name, GF_XATTR_PATHINFO_KEY)) {
+                if (is_fgetxattr) {
+                        *cbk = afr_fgetxattr_pathinfo_cbk;
+                } else {
+                        *cbk = afr_getxattr_pathinfo_cbk;
+                }
+        } else if (!strncmp (name, GF_XATTR_CLRLK_CMD,
+                             strlen (GF_XATTR_CLRLK_CMD))) {
+                if (is_fgetxattr) {
+                        *cbk = afr_fgetxattr_clrlk_cbk;
+                } else {
+                        *cbk = afr_getxattr_clrlk_cbk;
+                }
+        } else if (!strncmp (name, GF_XATTR_LOCKINFO_KEY,
+                             strlen (GF_XATTR_LOCKINFO_KEY))) {
+                if (is_fgetxattr) {
+                        *cbk = afr_fgetxattr_lockinfo_cbk;
+                } else {
+                        *cbk = afr_getxattr_lockinfo_cbk;
+                }
+        } else if (fnmatch (GF_XATTR_STIME_PATTERN, name, FNM_NOESCAPE) == 0) {
+                *cbk = afr_common_getxattr_stime_cbk;
         } else {
-                call_child = afr_first_up_child (priv);
+                is_spl = _gf_false;
+        }
+
+out:
+        return is_spl;
+}
+
+static void
+afr_getxattr_frm_all_children (xlator_t *this, call_frame_t *frame,
+                               const char *name, loc_t *loc,
+                               fop_getxattr_cbk_t cbk)
+{
+        afr_private_t   *priv           = NULL;
+        afr_local_t     *local          = NULL;
+        xlator_t        **children      = NULL;
+        int             i               = 0;
+
+        priv     = this->private;
+        children = priv->children;
+
+        local = frame->local;
+        local->call_count = priv->child_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                STACK_WIND_COOKIE (frame, cbk,
+                                   (void *) (long) i,
+                                   children[i], children[i]->fops->getxattr,
+                                   loc, name, NULL);
+        }
+        return;
+}
+
+int32_t
+afr_getxattr (call_frame_t *frame, xlator_t *this,
+              loc_t *loc, const char *name, dict_t *xdata)
+{
+        afr_private_t           *priv         = NULL;
+        xlator_t                **children    = NULL;
+        int                     call_child    = 0;
+        afr_local_t             *local        = NULL;
+        xlator_list_t           *trav         = NULL;
+        xlator_t                **sub_volumes = NULL;
+        int                     i             = 0;
+        int32_t                 op_errno      = 0;
+        int32_t                 read_child    = -1;
+        int                     ret           = -1;
+        fop_getxattr_cbk_t      cbk           = NULL;
+        int                     afr_xtime_gauge[MCNT_MAX] = {0,};
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        priv     = this->private;
+        VALIDATE_OR_GOTO (priv->children, out);
+
+        children = priv->children;
+
+        AFR_SBRAIN_CHECK_LOC (loc, out);
+
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
+
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
+
+        loc_copy (&local->loc, loc);
+        if (!name)
+                goto no_name;
+
+        local->cont.getxattr.name = gf_strdup (name);
 
-                if (call_child == -1) {
-                        op_errno = ENOTCONN;
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "no child is up");
+        if (!strncmp (name, AFR_XATTR_PREFIX,
+                      strlen (AFR_XATTR_PREFIX))) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "%s: no data present for key %s",
+                        loc->path, name);
+                op_errno = ENODATA;
+                goto out;
+        }
+        if ((strcmp (GF_XATTR_MARKER_KEY, name) == 0)
+            && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+
+                local->marker.call_count = priv->child_count;
+
+                sub_volumes = alloca ( priv->child_count * sizeof (xlator_t *));
+                for (i = 0, trav = this->children; trav ;
+                     trav = trav->next, i++) {
+
+                        *(sub_volumes + i)  = trav->xlator;
+                }
+
+                if (cluster_getmarkerattr (frame, this, loc, name,
+                                           local, afr_getxattr_unwind,
+                                           sub_volumes,
+                                           priv->child_count,
+                                           MARKER_UUID_TYPE,
+                                           marker_uuid_default_gauge,
+                                           priv->vol_uuid)) {
+
+                        gf_log (this->name, GF_LOG_INFO,
+                                "%s: failed to get marker attr (%s)",
+                                loc->path, name);
+                        op_errno = EINVAL;
                         goto out;
                 }
 
-                local->cont.getxattr.last_tried = call_child;
+                return 0;
+        }
+
+        /*
+         * if we are doing getxattr with pathinfo as the key then we
+         * collect information from all childs
+         */
+        if (afr_is_special_xattr (name, &cbk, 0)) {
+                afr_getxattr_frm_all_children (this, frame, name,
+                                               loc, cbk);
+                return 0;
+        }
+
+        if (XATTR_IS_NODE_UUID (name)) {
+                i = 0;
+                STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
+                                   (void *) (long) i,
+                                   children[i],
+                                   children[i]->fops->getxattr,
+                                   loc, name, xdata);
+                return 0;
+        }
+
+        if (*priv->vol_uuid) {
+                if ((match_uuid_local (name, priv->vol_uuid) == 0)
+                    && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+                        local->marker.call_count = priv->child_count;
+
+                        sub_volumes = alloca ( priv->child_count
+                                               * sizeof (xlator_t *));
+                        for (i = 0, trav = this->children; trav ;
+                             trav = trav->next, i++) {
+
+                                *(sub_volumes + i)  = trav->xlator;
+
+                        }
+
+                        /* don't err out on getting ENOTCONN (brick down)
+                         * from a subset of the bricks
+                         */
+                        memcpy (afr_xtime_gauge, marker_xtime_default_gauge,
+                                sizeof (afr_xtime_gauge));
+                        afr_xtime_gauge[MCNT_NOTFOUND] = 0;
+                        afr_xtime_gauge[MCNT_ENOTCONN] = 0;
+                        if (cluster_getmarkerattr (frame, this, loc,
+                                                   name, local,
+                                                   afr_getxattr_unwind,
+                                                   sub_volumes,
+                                                   priv->child_count,
+                                                   MARKER_XTIME_TYPE,
+                                                   afr_xtime_gauge,
+                                                   priv->vol_uuid)) {
+                                gf_log (this->name, GF_LOG_INFO,
+                                        "%s: failed to get marker attr (%s)",
+                                        loc->path, name);
+                                op_errno = EINVAL;
+                                goto out;
+                        }
+
+                        return 0;
+                }
+        }
+
+no_name:
+        local->fresh_children = afr_children_create (priv->child_count);
+        if (!local->fresh_children) {
+                op_errno = ENOMEM;
+                goto out;
         }
 
-	loc_copy (&local->loc, loc);
-	if (name)
-	  local->cont.getxattr.name       = gf_strdup (name);
+        read_child = afr_inode_get_read_ctx (this, loc->inode,
+                                             local->fresh_children);
+        ret = afr_get_call_child (this, local->child_up, read_child,
+                                     local->fresh_children,
+                                     &call_child,
+                                     &local->cont.getxattr.last_index);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto out;
+        }
+
+        STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
+                           (void *) (long) call_child,
+                           children[call_child],
+                           children[call_child]->fops->getxattr,
+                           loc, name, xdata);
+
+        ret = 0;
+out:
+        if (ret < 0)
+                AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
+        return 0;
+}
+
+/* {{{ fgetxattr */
 
-	STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
-			   (void *) (long) call_child,
-			   children[call_child], children[call_child]->fops->getxattr,
-			   loc, name);
 
-	op_ret = 0;
+int32_t
+afr_fgetxattr_cbk (call_frame_t *frame, void *cookie,
+                   xlator_t *this, int32_t op_ret, int32_t op_errno,
+                   dict_t *dict, dict_t *xdata)
+{
+        afr_private_t * priv            = NULL;
+        afr_local_t *   local           = NULL;
+        xlator_t **     children        = NULL;
+        int             unwind          = 1;
+        int32_t         *last_index     = NULL;
+        int32_t         next_call_child = -1;
+        int32_t         read_child      = -1;
+        int32_t         *fresh_children  = NULL;
+
+        priv     = this->private;
+        children = priv->children;
+
+        local = frame->local;
+
+        read_child = (long) cookie;
+
+        if (op_ret == -1) {
+                last_index = &local->cont.getxattr.last_index;
+                fresh_children = local->fresh_children;
+                next_call_child = afr_next_call_child (fresh_children,
+                                                       local->child_up,
+                                                       priv->child_count,
+                                                       last_index, read_child);
+                if (next_call_child < 0)
+                        goto out;
+
+                unwind = 0;
+                STACK_WIND_COOKIE (frame, afr_fgetxattr_cbk,
+                                   (void *) (long) read_child,
+                                   children[next_call_child],
+                                   children[next_call_child]->fops->fgetxattr,
+                                   local->fd,
+                                   local->cont.getxattr.name,
+                                   NULL);
+        }
+
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
+        if (unwind) {
+                if (op_ret >= 0 && dict)
+                        __filter_xattrs (dict);
+
+                AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict,
+                                  xdata);
+        }
+
+        return 0;
+}
+
+int32_t
+afr_fgetxattr_unwind (call_frame_t *frame,
+                      int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+
+{
+        AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+static void
+afr_fgetxattr_frm_all_children (xlator_t *this, call_frame_t *frame,
+                                const char *name, fd_t *fd,
+                                fop_fgetxattr_cbk_t cbk)
+{
+        afr_private_t   *priv           = NULL;
+        afr_local_t     *local          = NULL;
+        xlator_t        **children      = NULL;
+        int             i               = 0;
+
+        priv     = this->private;
+        children = priv->children;
+
+        local = frame->local;
+        local->call_count = priv->child_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                STACK_WIND_COOKIE (frame, cbk,
+                                   (void *) (long) i,
+                                   children[i], children[i]->fops->fgetxattr,
+                                   fd, name, NULL);
+        }
+
+        return;
+}
+
+int32_t
+afr_fgetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, const char *name, dict_t *xdata)
+{
+        afr_private_t        *priv       = NULL;
+        xlator_t            **children   = NULL;
+        int                   call_child = 0;
+        afr_local_t          *local      = NULL;
+        int32_t               op_ret     = -1;
+        int32_t               op_errno   = 0;
+        int32_t               read_child = -1;
+        fop_fgetxattr_cbk_t   cbk        = NULL;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        priv     = this->private;
+        VALIDATE_OR_GOTO (priv->children, out);
+
+        children = priv->children;
+
+        AFR_SBRAIN_CHECK_FD (fd, out);
+
+        AFR_LOCAL_ALLOC_OR_GOTO (local, out);
+        frame->local = local;
+
+        op_ret = afr_local_init (local, priv, &op_errno);
+        if (op_ret < 0) {
+                op_errno = -op_ret;
+                goto out;
+        }
+
+        local->fd = fd_ref (fd);
+        if (name)
+                local->cont.getxattr.name = gf_strdup (name);
+
+        /* pathinfo gets handled only in getxattr(), but we need to handle
+         * lockinfo.
+         * If we are doing fgetxattr with lockinfo as the key then we
+         * collect information from all children.
+         */
+        if (afr_is_special_xattr (name, &cbk, 1)) {
+                afr_fgetxattr_frm_all_children (this, frame, name,
+                                                fd, cbk);
+                return 0;
+        }
+
+
+        local->fresh_children = afr_children_create (priv->child_count);
+        if (!local->fresh_children) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        read_child = afr_inode_get_read_ctx (this, fd->inode,
+                                             local->fresh_children);
+        op_ret = afr_get_call_child (this, local->child_up, read_child,
+                                     local->fresh_children,
+                                     &call_child,
+                                     &local->cont.getxattr.last_index);
+        if (op_ret < 0) {
+                op_errno = -op_ret;
+                op_ret = -1;
+                goto out;
+        }
+
+        STACK_WIND_COOKIE (frame, afr_fgetxattr_cbk,
+                           (void *) (long) call_child,
+                           children[call_child],
+                           children[call_child]->fops->fgetxattr,
+                           fd, name, xdata);
+
+        op_ret = 0;
+out:
+        if (op_ret == -1) {
+                AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, NULL,
+                                  NULL);
+        }
+        return 0;
 }
 
 
@@ -748,140 +1811,130 @@ out:
 
 int32_t
 afr_readv_cbk (call_frame_t *frame, void *cookie,
-	       xlator_t *this, int32_t op_ret, int32_t op_errno,
-	       struct iovec *vector, int32_t count, struct iatt *buf,
-               struct iobref *iobref)
+               xlator_t *this, int32_t op_ret, int32_t op_errno,
+               struct iovec *vector, int32_t count, struct iatt *buf,
+               struct iobref *iobref, dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
-
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try   = -1;
-        int read_child = -1;
+        afr_private_t * priv            = NULL;
+        afr_local_t *   local           = NULL;
+        xlator_t **     children        = NULL;
+        int             unwind          = 1;
+        int32_t         *last_index     = NULL;
+        int32_t         next_call_child = -1;
+        int32_t         *fresh_children  = NULL;
+        int32_t         read_child      = -1;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+        priv     = this->private;
+        VALIDATE_OR_GOTO (priv->children, out);
 
-	children = priv->children;
+        children = priv->children;
 
-	local = frame->local;
+        local = frame->local;
 
         read_child = (long) cookie;
 
-	if (op_ret == -1) {
-	retry:
-		last_tried = local->cont.readv.last_tried;
-
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try = ++local->cont.readv.last_tried;
-
-		if (this_try == read_child) {
-			/*
-			   skip the read child since if we are here
-			   we must have already tried that child
-			*/
-			goto retry;
-		}
-
-		unwind = 0;
-
-		STACK_WIND_COOKIE (frame, afr_readv_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->readv,
-				   local->fd, local->cont.readv.size,
-				   local->cont.readv.offset);
-	}
+        if (op_ret == -1) {
+                last_index = &local->cont.readv.last_index;
+                fresh_children = local->fresh_children;
+                next_call_child = afr_next_call_child (fresh_children,
+                                                       local->child_up,
+                                                       priv->child_count,
+                                                       last_index, read_child);
+                if (next_call_child < 0)
+                        goto out;
 
-out:
-	if (unwind) {
-                if (buf && local)
-                        buf->ia_ino = local->cont.readv.ino;
+                unwind = 0;
 
-		AFR_STACK_UNWIND (readv, frame, op_ret, op_errno,
-                                  vector, count, buf, iobref);
-	}
+                STACK_WIND_COOKIE (frame, afr_readv_cbk,
+                                   (void *) (long) read_child,
+                                   children[next_call_child],
+                                   children[next_call_child]->fops->readv,
+                                   local->fd, local->cont.readv.size,
+                                   local->cont.readv.offset,
+                                   local->cont.readv.flags,
+                                   NULL);
+        }
 
-	return 0;
+out:
+        if (unwind) {
+                AFR_STACK_UNWIND (readv, frame, op_ret, op_errno,
+                                  vector, count, buf, iobref, xdata);
+        }
+
+        return 0;
 }
 
 
 int32_t
 afr_readv (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd, size_t size, off_t offset)
+           fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
 {
-	afr_private_t * priv       = NULL;
-	afr_local_t   * local      = NULL;
-	xlator_t **     children   = NULL;
-
+        afr_private_t * priv       = NULL;
+        afr_local_t   * local      = NULL;
+        xlator_t **     children   = NULL;
+        int             call_child = 0;
+        int32_t         op_errno   = 0;
         int32_t         read_child = -1;
-	int             call_child = 0;
+        int             ret        = -1;
 
-	int32_t         op_ret     = -1;
-	int32_t         op_errno   = 0;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+        VALIDATE_OR_GOTO (fd, out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-	VALIDATE_OR_GOTO (fd, out);
+        priv     = this->private;
+        children = priv->children;
 
-	priv     = this->private;
-	children = priv->children;
+        AFR_SBRAIN_CHECK_FD (fd, out);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	frame->local = local;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-        read_child = afr_read_child (this, fd->inode);
+        local->fresh_children = afr_children_create (priv->child_count);
+        if (!local->fresh_children) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-        if ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
+        read_child = afr_inode_get_read_ctx (this, fd->inode, local->fresh_children);
+        ret = afr_get_call_child (this, local->child_up, read_child,
+                                     local->fresh_children,
+                                     &call_child,
+                                     &local->cont.readv.last_index);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto out;
+        }
 
-		/*
-		   if read fails from the read child, we try
-		   all children starting with the first one
-		*/
-                local->cont.readv.last_tried = -1;
+        local->fd                    = fd_ref (fd);
 
-        } else {
-		call_child = afr_first_up_child (priv);
-		if (call_child == -1) {
-			op_errno = ENOTCONN;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"no child is up");
-			goto out;
-		}
-
-		local->cont.readv.last_tried = call_child;
-	}
-
-	local->fd                    = fd_ref (fd);
-
-        local->cont.readv.ino        = fd->inode->ino;
-	local->cont.readv.size       = size;
-	local->cont.readv.offset     = offset;
-
-	STACK_WIND_COOKIE (frame, afr_readv_cbk,
-			   (void *) (long) call_child,
-			   children[call_child],
-			   children[call_child]->fops->readv,
-			   fd, size, offset);
-
-	op_ret = 0;
+        local->cont.readv.size       = size;
+        local->cont.readv.offset     = offset;
+        local->cont.readv.flags      = flags;
+
+        afr_open_fd_fix (fd, this);
+
+        STACK_WIND_COOKIE (frame, afr_readv_cbk,
+                           (void *) (long) call_child,
+                           children[call_child],
+                           children[call_child]->fops->readv,
+                           fd, size, offset, flags, xdata);
+
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (readv, frame, op_ret, op_errno, NULL, 0, NULL,
-                                  NULL);
-	}
-	return 0;
+        if (ret < 0) {
+                AFR_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL,
+                                  NULL, NULL);
+        }
+        return 0;
 }
 
 /* }}} */
diff --git a/xlators/cluster/afr/src/afr-inode-read.h b/xlators/cluster/afr/src/afr-inode-read.h
index 2b22131db..e4091a793 100644
--- a/xlators/cluster/afr/src/afr-inode-read.h
+++ b/xlators/cluster/afr/src/afr-inode-read.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __INODE_READ_H__
@@ -22,26 +13,30 @@
 
 int32_t
 afr_access (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t mask);
+	    loc_t *loc, int32_t mask, dict_t *xdata);
 
 int32_t
 afr_stat (call_frame_t *frame, xlator_t *this,
-	  loc_t *loc);
+	  loc_t *loc, dict_t *xdata);
 
 int32_t
 afr_fstat (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd);
+	   fd_t *fd, dict_t *xdata);
 
 int32_t
 afr_readlink (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, size_t size);
+	      loc_t *loc, size_t size, dict_t *xdata);
 
 int32_t
 afr_readv (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd, size_t size, off_t offset);
+	   fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata);
 
 int32_t
 afr_getxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, const char *name);
+	      loc_t *loc, const char *name, dict_t *xdata);
+
+int32_t
+afr_fgetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, const char *name, dict_t *xdata);
 
 #endif /* __INODE_READ_H__ */
diff --git a/xlators/cluster/afr/src/afr-inode-write.c b/xlators/cluster/afr/src/afr-inode-write.c
index 7a908acd7..c1ec69a55 100644
--- a/xlators/cluster/afr/src/afr-inode-write.c
+++ b/xlators/cluster/afr/src/afr-inode-write.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -46,270 +37,549 @@
 
 #include "afr.h"
 #include "afr-transaction.h"
+#include "afr-self-heal-common.h"
+
+void
+__inode_write_fop_cbk (call_frame_t *frame, int child_index, int read_child,
+                       xlator_t *this, int32_t *op_ret, int32_t *op_errno,
+                       struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+        afr_local_t     *local = NULL;
+
+        local = frame->local;
+
+        if (afr_fop_failed (*op_ret, *op_errno)) {
+                local->child_errno[child_index] = *op_errno;
+
+                switch (local->op) {
+                case GF_FOP_TRUNCATE:
+                case GF_FOP_FTRUNCATE:
+                        if (*op_errno != EFBIG)
+                                afr_transaction_fop_failed (frame, this,
+                                                            child_index);
+                break;
+                default:
+                        afr_transaction_fop_failed (frame, this, child_index);
+                break;
+                }
+                local->op_errno = *op_errno;
+                goto out;
+        }
+
+        if ((local->success_count == 0) || (read_child == child_index)) {
+                local->op_ret              = *op_ret;
+                if (prebuf)
+                        local->cont.inode_wfop.prebuf  = *prebuf;
+                if (postbuf)
+                        local->cont.inode_wfop.postbuf = *postbuf;
+        }
+
+        local->success_count++;
+out:
+        return;
+}
 
 /* {{{ writev */
 
-int
-afr_writev_unwind (call_frame_t *frame, xlator_t *this)
+void
+afr_writev_copy_outvars (call_frame_t *src_frame, call_frame_t *dst_frame)
 {
-	afr_local_t *   local = NULL;
-	call_frame_t   *main_frame = NULL;
+        afr_local_t *src_local = NULL;
+        afr_local_t *dst_local = NULL;
 
-	local = frame->local;
+        src_local = src_frame->local;
+        dst_local = dst_frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame)
-			main_frame = local->transaction.main_frame;
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        dst_local->op_ret = src_local->op_ret;
+        dst_local->op_errno = src_local->op_errno;
+        dst_local->cont.inode_wfop.prebuf = src_local->cont.inode_wfop.prebuf;
+        dst_local->cont.inode_wfop.postbuf = src_local->cont.inode_wfop.postbuf;
+}
 
-	if (main_frame) {
-                local->cont.writev.prebuf.ia_ino  = local->cont.writev.ino;
-                local->cont.writev.postbuf.ia_ino = local->cont.writev.ino;
+void
+afr_writev_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *   local = NULL;
+        local = frame->local;
 
-		AFR_STACK_UNWIND (writev, main_frame,
-                                  local->op_ret, local->op_errno,
-                                  &local->cont.writev.prebuf,
-                                  &local->cont.writev.postbuf);
-	}
-	return 0;
+        AFR_STACK_UNWIND (writev, frame,
+                          local->op_ret, local->op_errno,
+                          &local->cont.inode_wfop.prebuf,
+                          &local->cont.inode_wfop.postbuf,
+                          NULL);
 }
 
+call_frame_t*
+afr_transaction_detach_fop_frame (call_frame_t *frame)
+{
+        afr_local_t *   local = NULL;
+        call_frame_t   *fop_frame = NULL;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                fop_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
+
+        return fop_frame;
+}
 
 int
-afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                     struct iatt *postbuf)
+afr_transaction_writev_unwind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
+        call_frame_t *fop_frame = NULL;
 
-	int child_index = (long) cookie;
-	int call_count  = -1;
+        fop_frame = afr_transaction_detach_fop_frame (frame);
+
+        if (fop_frame) {
+                afr_writev_copy_outvars (frame, fop_frame);
+                afr_writev_unwind (fop_frame, this);
+        }
+        return 0;
+}
+
+static void
+afr_writev_handle_short_writes (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t   *local = NULL;
+        afr_private_t *priv  = NULL;
+        int           i      = 0;
+
+        local = frame->local;
+        priv = this->private;
+        /*
+         * We already have the best case result of the writev calls staged
+         * as the return value. Any writev that returns some value less
+         * than the best case is now out of sync, so mark the fop as
+         * failed. Note that fops that have returned with errors have
+         * already been marked as failed.
+         */
+        for (i = 0; i < priv->child_count; i++) {
+                if ((!local->replies[i].valid) ||
+                    (local->replies[i].op_ret == -1))
+                        continue;
+
+                if (local->replies[i].op_ret < local->op_ret)
+                        afr_transaction_fop_failed(frame, this, i);
+        }
+}
+
+int
+afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
+{
+        afr_local_t *   local = NULL;
+        afr_private_t  *priv  = NULL;
+        call_frame_t    *fop_frame = NULL;
+        int child_index = (long) cookie;
+        int call_count  = -1;
         int read_child  = 0;
+        int      ret = 0;
+        uint32_t open_fd_count = 0;
+        uint32_t write_is_append = 0;
 
-	local = frame->local;
+        local = frame->local;
+        priv  = this->private;
 
-        read_child = afr_read_child (this, local->fd->inode);
+        read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
 
-	LOCK (&frame->lock);
-	{
+        LOCK (&frame->lock);
+        {
                 if (child_index == read_child) {
                         local->read_child_returned = _gf_true;
                 }
 
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret              = op_ret;
-				local->cont.writev.prebuf  = *prebuf;
-				local->cont.writev.postbuf = *postbuf;
-			}
+                __inode_write_fop_cbk (frame, child_index, read_child, this,
+                                       &op_ret, &op_errno, prebuf, postbuf,
+                                       xdata);
 
-                        if (child_index == read_child) {
-                                local->cont.writev.prebuf  = *prebuf;
-                                local->cont.writev.postbuf = *postbuf;
-                        }
-		}
+		local->replies[child_index].valid = 1;
+		local->replies[child_index].op_ret = op_ret;
+		local->replies[child_index].op_errno = op_errno;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+		/* stage the best case return value for unwind */
+                if ((local->success_count == 0) || (op_ret > local->op_ret)) {
+                        local->op_ret              = op_ret;
+			local->op_errno		   = op_errno;
+		}
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+		if (op_ret != -1) {
+                        if (xdata) {
+                                ret = dict_get_uint32 (xdata,
+                                                       GLUSTERFS_OPEN_FD_COUNT,
+                                                       &open_fd_count);
+                                if ((ret == 0) &&
+                                    (open_fd_count > local->open_fd_count)) {
+                                        local->open_fd_count = open_fd_count;
+                                        local->update_open_fd_count = _gf_true;
+                                }
+
+				write_is_append = 0;
+                                ret = dict_get_uint32 (xdata,
+                                                       GLUSTERFS_WRITE_IS_APPEND,
+                                                       &write_is_append);
+                                if (ret || !write_is_append)
+					local->append_write = _gf_false;
+                        }
 
-		local->transaction.resume (frame, this);
-	}
-	return 0;
+		}
+        }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+
+                if (local->update_open_fd_count)
+                        afr_handle_open_fd_count (frame, this);
+
+                if (!local->stable_write && !local->append_write)
+			/* An appended write removes the necessity to
+			   fsync() the file. This is because self-heal
+			   has the logic to check for larger file when
+			   the xattrs are not reliably pointing at
+			   a stale file.
+			*/
+                        afr_fd_report_unstable_write (this, local->fd);
+
+                afr_writev_handle_short_writes (frame, this);
+                if (afr_any_fops_failed (local, priv)) {
+                        //Don't unwind until post-op is complete
+                        local->transaction.resume (frame, this);
+                } else {
+                /*
+                 * Generally inode-write fops do transaction.unwind then
+                 * transaction.resume, but writev needs to make sure that
+                 * delayed post-op frame is placed in fdctx before unwind
+                 * happens. This prevents the race of flush doing the
+                 * changelog wakeup first in fuse thread and then this
+                 * writev placing its delayed post-op frame in fdctx.
+                 * This helps flush make sure all the delayed post-ops are
+                 * completed.
+                 */
+
+                        fop_frame = afr_transaction_detach_fop_frame (frame);
+                        afr_writev_copy_outvars (frame, fop_frame);
+                        local->transaction.resume (frame, this);
+                        afr_writev_unwind (fop_frame, this);
+                }
+        }
+        return 0;
 }
 
-
 int
 afr_writev_wind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-	
-	int i = 0;
-	int call_count = -1;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int i = 0;
+        int call_count = -1;
+        dict_t *xdata = NULL;
+        GF_UNUSED int     ret = 0;
 
-	local = frame->local;
-	priv = this->private;
+        local = frame->local;
+        priv = this->private;
+
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
+        local->call_count = call_count;
+	local->replies = GF_CALLOC(priv->child_count, sizeof(*local->replies),
+				   gf_afr_mt_reply_t);
+	if (!local->replies) {
+		local->op_ret = -1;
+		local->op_errno = ENOMEM;
+		local->transaction.unwind(frame, this);
+		local->transaction.resume(frame, this);
 		return 0;
 	}
 
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_writev_wind_cbk, 
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->writev,
-					   local->fd, 
-					   local->cont.writev.vector,
-					   local->cont.writev.count, 
-					   local->cont.writev.offset,
-                                           local->cont.writev.iobref);
-		
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        xdata = dict_new ();
+        if (xdata) {
+                ret = dict_set_uint32 (xdata, GLUSTERFS_OPEN_FD_COUNT,
+                                       sizeof (uint32_t));
+		ret = dict_set_uint32 (xdata, GLUSTERFS_WRITE_IS_APPEND,
+				       0);
+		/* Set append_write to be true speculatively. If on any
+		   server it turns not be true, we unset it in the
+		   callback.
+		*/
+		local->append_write = _gf_true;
+        }
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_writev_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->writev,
+                                           local->fd,
+                                           local->cont.writev.vector,
+                                           local->cont.writev.count,
+                                           local->cont.writev.offset,
+                                           local->cont.writev.flags,
+                                           local->cont.writev.iobref,
+                                           xdata);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        if (xdata)
+                dict_unref (xdata);
+
+        return 0;
 }
 
 
 int
 afr_writev_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
+        afr_local_t *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
         iobref_unref (local->cont.writev.iobref);
-	local->cont.writev.iobref = NULL;
+        local->cont.writev.iobref = NULL;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
+        AFR_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_do_writev (call_frame_t *frame, xlator_t *this)
 {
-        call_frame_t * transaction_frame = NULL;
-        afr_local_t *  local             = NULL;
-
-        int op_ret   = -1;
-        int op_errno = 0;
+        call_frame_t    *transaction_frame = NULL;
+        afr_local_t     *local             = NULL;
+        int             op_ret   = -1;
+        int             op_errno = 0;
 
         local = frame->local;
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
                 op_errno = ENOMEM;
-		goto out;
-	}
+                goto out;
+        }
 
-	transaction_frame->local = local;
-        frame->local = NULL;
+        transaction_frame->local = local;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
 
-	local->op = GF_FOP_WRITE;
+        local->op = GF_FOP_WRITE;
 
         local->success_count      = 0;
 
-	local->transaction.fop    = afr_writev_wind;
-	local->transaction.done   = afr_writev_done;
-	local->transaction.unwind = afr_writev_unwind;
-
-	local->transaction.main_frame = frame;
-	if (local->fd->flags & O_APPEND) {
-		local->transaction.start   = 0;
-		local->transaction.len     = 0;
-	} else {
-		local->transaction.start   = local->cont.writev.offset;
-		local->transaction.len     = iov_length (local->cont.writev.vector,
+        local->transaction.fop    = afr_writev_wind;
+        local->transaction.done   = afr_writev_done;
+        local->transaction.unwind = afr_transaction_writev_unwind;
+
+        local->transaction.main_frame = frame;
+        if (local->fd->flags & O_APPEND) {
+               /*
+                * Backend vfs ignores the 'offset' for append mode fd so
+                * locking just the region provided for the writev does not
+                * give consistency gurantee. The actual write may happen at a
+                * completely different range than the one provided by the
+                * offset, len in the fop. So lock the entire file.
+                */
+                local->transaction.start   = 0;
+                local->transaction.len     = 0;
+        } else {
+                local->transaction.start   = local->cont.writev.offset;
+                local->transaction.len     = iov_length (local->cont.writev.vector,
                                                          local->cont.writev.count);
-	}
+        }
 
-	afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        if (op_ret < 0) {
+            op_errno = -op_ret;
+            goto out;
+        }
 
         op_ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (writev, frame, op_ret, op_errno, NULL, NULL);
-	}
+        if (op_ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (writev, frame, op_ret, op_errno, NULL, NULL, NULL);
+        }
 
         return 0;
 }
 
+static void
+afr_trigger_open_fd_self_heal (fd_t *fd, xlator_t *this)
+{
+        call_frame_t    *frame   = NULL;
+        afr_local_t     *local   = NULL;
+        afr_self_heal_t *sh      = NULL;
+        char            *reason  = NULL;
+        int32_t         op_errno = 0;
+        int             ret      = 0;
+
+        if (!fd || !fd->inode || uuid_is_null (fd->inode->gfid)) {
+                gf_log_callingfn (this->name, GF_LOG_ERROR, "Invalid args: "
+                                  "fd: %p, inode: %p", fd,
+                                  fd ? fd->inode : NULL);
+                goto out;
+        }
 
-int
-afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, 
-	    struct iovec *vector, int32_t count, off_t offset,
-            struct iobref *iobref)
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame)
+                goto out;
+
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
+        ret = afr_local_init (local, this->private, &op_errno);
+        if (ret < 0)
+                goto out;
+
+        local->loc.inode = inode_ref (fd->inode);
+        ret = loc_path (&local->loc, NULL);
+        if (ret < 0)
+                goto out;
+
+        sh    = &local->self_heal;
+        sh->do_metadata_self_heal = _gf_true;
+        if (fd->inode->ia_type == IA_IFREG)
+                sh->do_data_self_heal = _gf_true;
+        else if (fd->inode->ia_type == IA_IFDIR)
+                sh->do_entry_self_heal = _gf_true;
+
+        reason = "subvolume came online";
+        afr_launch_self_heal (frame, this, fd->inode, _gf_true,
+                              fd->inode->ia_type, reason, NULL, NULL);
+        return;
+out:
+        AFR_STACK_DESTROY (frame);
+}
+
+void
+afr_open_fd_fix (fd_t *fd, xlator_t *this)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
+        int           ret             = 0;
+        int           i               = 0;
+        afr_fd_ctx_t  *fd_ctx         = NULL;
+        gf_boolean_t  need_self_heal  = _gf_false;
+        int           *need_open      = NULL;
+        size_t        need_open_count = 0;
+        afr_private_t *priv           = NULL;
 
-	int ret = -1;
+        priv  = this->private;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        if (!afr_is_fd_fixable (fd))
+                goto out;
 
-        uint64_t ctx;
-        afr_fd_ctx_t *fd_ctx = NULL;
+        fd_ctx = afr_fd_ctx_get (fd, this);
+        if (!fd_ctx)
+                goto out;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        LOCK (&fd->lock);
+        {
+                if (fd_ctx->up_count < priv->up_count) {
+                        need_self_heal = _gf_true;
+                        fd_ctx->up_count   = priv->up_count;
+                        fd_ctx->down_count = priv->down_count;
+                }
 
-	priv = this->private;
+                need_open = alloca (priv->child_count * sizeof (*need_open));
+                for (i = 0; i < priv->child_count; i++) {
+                        need_open[i] = 0;
+                        if (fd_ctx->opened_on[i] != AFR_FD_NOT_OPENED)
+                                continue;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+                        if (!priv->child_up[i])
+                                continue;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+                        fd_ctx->opened_on[i] = AFR_FD_OPENING;
 
-        frame->local = local;
+                        need_open[i] = 1;
+                        need_open_count++;
+                }
+        }
+        UNLOCK (&fd->lock);
+        if (ret)
+                goto out;
 
-	local->cont.writev.vector     = iov_dup (vector, count);
-	local->cont.writev.count      = count;
-	local->cont.writev.offset     = offset;
-	local->cont.writev.ino        = fd->inode->ino;
-        local->cont.writev.iobref     = iobref_ref (iobref);
+        if (need_self_heal)
+                afr_trigger_open_fd_self_heal (fd, this);
 
-	local->fd                = fd_ref (fd);
+        if (!need_open_count)
+                goto out;
 
-        ret = fd_ctx_get (fd, this, &ctx);
-        if (ret < 0) {
+        afr_fix_open (this, fd, need_open_count, need_open);
+out:
+        return;
+}
+
+int
+afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+            struct iovec *vector, int32_t count, off_t offset,
+            uint32_t flags, struct iobref *iobref, dict_t *xdata)
+{
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
+        int ret = -1;
+        int op_errno = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        priv = this->private;
+
+        if (afr_is_split_brain (this, fd->inode)) {
+                op_errno = EIO;
                 goto out;
         }
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        QUORUM_CHECK(writev,out);
 
-        if (fd_ctx->down_count < priv->down_count) {
-                local->up_down_flush_cbk = afr_do_writev;
-                afr_up_down_flush (frame, this, fd, AFR_CHILD_DOWN_FLUSH);
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-        } else if (fd_ctx->up_count < priv->up_count) {
-                local->up_down_flush_cbk = afr_do_writev;
-                afr_up_down_flush (frame, this, fd, AFR_CHILD_UP_FLUSH);
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-        } else {
-                afr_do_writev (frame, this);
-        }
+        local->cont.writev.vector     = iov_dup (vector, count);
+        local->cont.writev.count      = count;
+        local->cont.writev.offset     = offset;
+        local->cont.writev.flags      = flags;
+        local->cont.writev.iobref     = iobref_ref (iobref);
+
+        local->fd                = fd_ref (fd);
+
+	/* detect here, but set it in writev_wind_cbk *after* the unstable
+	   write is performed
+	*/
+	local->stable_write = !!((fd->flags|flags)&(O_SYNC|O_DSYNC));
+
+        afr_open_fd_fix (fd, this);
 
-	op_ret = 0;
+        afr_do_writev (frame, this);
+
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (writev, frame, op_ret, op_errno, NULL, NULL);
-	}
+        if (ret < 0)
+                AFR_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
@@ -320,212 +590,188 @@ out:
 int
 afr_truncate_unwind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	call_frame_t   *main_frame = NULL;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame)
-			main_frame = local->transaction.main_frame;
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        local = frame->local;
 
-	if (main_frame) {
-                local->cont.truncate.prebuf.ia_ino  = local->cont.truncate.ino;
-                local->cont.truncate.postbuf.ia_ino = local->cont.truncate.ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame)
+                        main_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
+        if (main_frame) {
                 AFR_STACK_UNWIND (truncate, main_frame, local->op_ret,
                                   local->op_errno,
-                                  &local->cont.truncate.prebuf,
-                                  &local->cont.truncate.postbuf);
+                                  &local->cont.inode_wfop.prebuf,
+                                  &local->cont.inode_wfop.postbuf,
+                                  NULL);
         }
 
-	return 0;
+        return 0;
 }
 
 
 int
-afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		       int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                       struct iatt *postbuf)
+afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                       struct iatt *postbuf, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int child_index = (long) cookie;
+        afr_local_t *   local = NULL;
+        int child_index = (long) cookie;
         int read_child  = 0;
-	int call_count  = -1;
-	int need_unwind = 0;
+        int call_count  = -1;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
 
-        read_child = afr_read_child (this, local->loc.inode);
+        read_child = afr_inode_get_read_ctx (this, local->loc.inode, NULL);
 
-	LOCK (&frame->lock);
-	{
+        LOCK (&frame->lock);
+        {
                 if (child_index == read_child) {
                         local->read_child_returned = _gf_true;
                 }
 
-		if (afr_fop_failed (op_ret, op_errno) && op_errno != EFBIG)
-			afr_transaction_fop_failed (frame, this, child_index);
-
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-				local->cont.truncate.prebuf  = *prebuf;
-				local->cont.truncate.postbuf = *postbuf;
-			}
-
-                        if (child_index == read_child) {
-                                local->cont.truncate.prebuf  = *prebuf;
-                                local->cont.truncate.postbuf = *postbuf;
-                        }
-
-			local->success_count++;
-
-			if ((local->success_count >= priv->wait_count)
-                            && local->read_child_returned) {
-				need_unwind = 1;
-			}
-		}
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                if (op_ret != -1) {
+			if (prebuf->ia_size != postbuf->ia_size)
+				local->stable_write = _gf_false;
+                }
+                __inode_write_fop_cbk (frame, child_index, read_child, this,
+                                       &op_ret, &op_errno, prebuf, postbuf,
+                                       xdata);
+        }
+        UNLOCK (&frame->lock);
 
-	if (need_unwind)
-		local->transaction.unwind (frame, this);
+        call_count = afr_frame_return (frame);
 
-	call_count = afr_frame_return (frame);
+        if (call_count == 0) {
+		if (local->stable_write && afr_txn_nothing_failed (frame, this))
+			local->transaction.unwind (frame, this);
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
+                local->transaction.resume (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_truncate_wind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
-	local = frame->local;
-	priv = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+        local = frame->local;
+        priv = this->private;
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-	local->call_count = call_count;
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_truncate_wind_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->truncate,
-					   &local->loc,
-					   local->cont.truncate.offset);
+        local->call_count = call_count;
+	local->stable_write = _gf_true;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_truncate_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->truncate,
+                                           &local->loc,
+                                           local->cont.truncate.offset,
+                                           NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int
 afr_truncate_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
+        afr_local_t *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
+        AFR_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_truncate (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, off_t offset)
+              loc_t *loc, off_t offset, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t   *transaction_frame = NULL;
-
-	int ret = -1;
-
-	int op_ret   = -1;
-	int op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
+        call_frame_t   *transaction_frame = NULL;
+        int ret = -1;
+        int op_errno = 0;
 
-	priv = this->private;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
+        priv = this->private;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        QUORUM_CHECK(truncate,out);
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	transaction_frame->local = local;
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
 
-	local->op_ret = -1;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	local->cont.truncate.offset  = offset;
-	local->cont.truncate.ino     = loc->inode->ino;
+        local->cont.truncate.offset  = offset;
 
-	local->transaction.fop    = afr_truncate_wind;
-	local->transaction.done   = afr_truncate_done;
-	local->transaction.unwind = afr_truncate_unwind;
+        local->transaction.fop    = afr_truncate_wind;
+        local->transaction.done   = afr_truncate_done;
+        local->transaction.unwind = afr_truncate_unwind;
 
-	loc_copy (&local->loc, loc);
+        loc_copy (&local->loc, loc);
 
-	local->transaction.main_frame = frame;
-	local->transaction.start   = 0;
-	local->transaction.len     = offset;
+        local->transaction.main_frame = frame;
+        local->transaction.start   = offset;
+        local->transaction.len     = 0;
 
-	afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (truncate, frame, op_ret, op_errno, NULL, NULL);
-	}
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
+        }
 
-	return 0;
+        return 0;
 }
 
 
@@ -537,145 +783,126 @@ out:
 int
 afr_ftruncate_unwind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	call_frame_t   *main_frame = NULL;
-
-	local = frame->local;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame)
-			main_frame = local->transaction.main_frame;
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        local = frame->local;
 
-	if (main_frame) {
-                local->cont.ftruncate.prebuf.ia_ino  = local->cont.ftruncate.ino;
-                local->cont.ftruncate.postbuf.ia_ino = local->cont.ftruncate.ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame)
+                        main_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-		AFR_STACK_UNWIND (ftruncate, main_frame, local->op_ret,
+        if (main_frame) {
+                AFR_STACK_UNWIND (ftruncate, main_frame, local->op_ret,
                                   local->op_errno,
-                                  &local->cont.ftruncate.prebuf,
-                                  &local->cont.ftruncate.postbuf);
-	}
-	return 0;
+                                  &local->cont.inode_wfop.prebuf,
+                                  &local->cont.inode_wfop.postbuf,
+                                  NULL);
+        }
+        return 0;
 }
 
 
 int
-afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-			int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                        struct iatt *postbuf)
+afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                        struct iatt *postbuf, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int child_index = (long) cookie;
-	int call_count  = -1;
-	int need_unwind = 0;
+        afr_local_t *   local = NULL;
+        int child_index = (long) cookie;
+        int call_count  = -1;
         int read_child  = 0;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
 
-        read_child = afr_read_child (this, local->fd->inode);
+        read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
 
-	LOCK (&frame->lock);
-	{
+        LOCK (&frame->lock);
+        {
                 if (child_index == read_child) {
                         local->read_child_returned = _gf_true;
                 }
 
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-				local->cont.ftruncate.prebuf  = *prebuf;
-				local->cont.ftruncate.postbuf = *postbuf;
-			}
-
-                        if (child_index == read_child) {
-                                local->cont.ftruncate.prebuf  = *prebuf;
-                                local->cont.ftruncate.postbuf = *postbuf;
-                        }
-
-			local->success_count++;
+                if (op_ret != -1) {
+			if (prebuf->ia_size != postbuf->ia_size)
+				local->stable_write = _gf_false;
+                }
+                __inode_write_fop_cbk (frame, child_index, read_child, this,
+                                       &op_ret, &op_errno, prebuf, postbuf,
+                                       xdata);
+        }
+        UNLOCK (&frame->lock);
 
-			if ((local->success_count >= priv->wait_count)
-                            && local->read_child_returned) {
-				need_unwind = 1;
-			}
-		}
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        call_count = afr_frame_return (frame);
 
-	if (need_unwind)
-		local->transaction.unwind (frame, this);
+        if (call_count == 0) {
+		if (local->stable_write && afr_txn_nothing_failed (frame, this))
+			local->transaction.unwind (frame, this);
 
-	call_count = afr_frame_return (frame);
+                local->transaction.resume (frame, this);
+        }
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int
 afr_ftruncate_wind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-	
-	int call_count = -1;
-	int i = 0;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
-	local = frame->local;
-	priv = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+        local = frame->local;
+        priv = this->private;
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-	local->call_count = call_count;
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_ftruncate_wind_cbk,
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->ftruncate,
-					   local->fd, local->cont.ftruncate.offset);
+        local->call_count = call_count;
+	local->stable_write = _gf_true;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_ftruncate_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->ftruncate,
+                                           local->fd,
+                                           local->cont.ftruncate.offset,
+                                           NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int
 afr_ftruncate_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
+        afr_local_t *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
+        AFR_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
 
@@ -684,42 +911,44 @@ afr_do_ftruncate (call_frame_t *frame, xlator_t *this)
 {
         call_frame_t * transaction_frame = NULL;
         afr_local_t *  local             = NULL;
-
         int op_ret   = -1;
         int op_errno = 0;
 
         local = frame->local;
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                goto out;
+        }
 
         transaction_frame->local = local;
         frame->local = NULL;
 
-	local->op = GF_FOP_FTRUNCATE;
+        local->op = GF_FOP_FTRUNCATE;
 
-	local->transaction.fop    = afr_ftruncate_wind;
-	local->transaction.done   = afr_ftruncate_done;
-	local->transaction.unwind = afr_ftruncate_unwind;
+        local->transaction.fop    = afr_ftruncate_wind;
+        local->transaction.done   = afr_ftruncate_done;
+        local->transaction.unwind = afr_ftruncate_unwind;
 
-	local->transaction.main_frame = frame;
+        local->transaction.main_frame = frame;
 
-	local->transaction.start   = 0;
-	local->transaction.len     = local->cont.ftruncate.offset;
+        local->transaction.start   = local->cont.ftruncate.offset;
+        local->transaction.len     = 0;
 
-	afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        if (op_ret < 0) {
+            op_errno = -op_ret;
+            goto out;
+        }
 
         op_ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, NULL, NULL);
-	}
+        if (op_ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, NULL,
+                                  NULL, NULL);
+        }
 
         return 0;
 }
@@ -727,64 +956,50 @@ out:
 
 int
 afr_ftruncate (call_frame_t *frame, xlator_t *this,
-	       fd_t *fd, off_t offset)
+               fd_t *fd, off_t offset, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t   *transaction_frame = NULL;
-
-	int ret = -1;
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
+        call_frame_t   *transaction_frame = NULL;
+        int ret = -1;
+        int op_errno = 0;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-        uint64_t ctx;
-        afr_fd_ctx_t *fd_ctx = NULL;
+        priv = this->private;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        if (afr_is_split_brain (this, fd->inode)) {
+                op_errno = EIO;
+                goto out;
+        }
+        QUORUM_CHECK(ftruncate,out);
 
-	priv = this->private;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        local->cont.ftruncate.offset  = offset;
 
-        frame->local = local;
+        local->fd = fd_ref (fd);
 
-	local->cont.ftruncate.offset  = offset;
-	local->cont.ftruncate.ino     = fd->inode->ino;
+        afr_open_fd_fix (fd, this);
 
-	local->fd = fd_ref (fd);
+        afr_do_ftruncate (frame, this);
 
-        ret = fd_ctx_get (fd, this, &ctx);
+        ret = 0;
+out:
         if (ret < 0) {
-                goto out;
-        }
-
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-        if (fd_ctx->down_count < priv->down_count) {
-                local->up_down_flush_cbk = afr_do_ftruncate;
-                afr_up_down_flush (frame, this, fd, AFR_CHILD_DOWN_FLUSH);
-        } else {
-                afr_do_ftruncate (frame, this);
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
         }
 
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, NULL, NULL);
-	}
-
-	return 0;
+        return 0;
 }
 
 /* }}} */
@@ -794,215 +1009,193 @@ out:
 int
 afr_setattr_unwind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	call_frame_t   *main_frame = NULL;
-
-	local = frame->local;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame)
-			main_frame = local->transaction.main_frame;
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        local = frame->local;
 
-	if (main_frame) {
-                local->cont.setattr.preop_buf.ia_ino  = local->cont.setattr.ino;
-                local->cont.setattr.postop_buf.ia_ino = local->cont.setattr.ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame)
+                        main_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
+        if (main_frame) {
                 AFR_STACK_UNWIND (setattr, main_frame, local->op_ret,
                                   local->op_errno,
-                                  &local->cont.setattr.preop_buf,
-                                  &local->cont.setattr.postop_buf);
+                                  &local->cont.inode_wfop.prebuf,
+                                  &local->cont.inode_wfop.postbuf,
+                                  NULL);
         }
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                       int32_t op_ret, int32_t op_errno,
-                      struct iatt *preop, struct iatt *postop)
+                      struct iatt *preop, struct iatt *postop, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int child_index = (long) cookie;
+        afr_local_t *   local = NULL;
+        afr_private_t * priv  = NULL;
+        int child_index = (long) cookie;
         int read_child  = 0;
-	int call_count  = -1;
-	int need_unwind = 0;
+        int call_count  = -1;
+        int need_unwind = 0;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
+        priv  = this->private;
 
-        read_child = afr_read_child (this, local->loc.inode);
+        read_child = afr_inode_get_read_ctx (this, local->loc.inode, NULL);
 
-	LOCK (&frame->lock);
-	{
+        LOCK (&frame->lock);
+        {
                 if (child_index == read_child) {
                         local->read_child_returned = _gf_true;
                 }
 
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-				local->cont.setattr.preop_buf  = *preop;
-                                local->cont.setattr.postop_buf = *postop;
-			}
-
-                        if (child_index == read_child) {
-                                local->cont.setattr.preop_buf  = *preop;
-                                local->cont.setattr.postop_buf = *postop;
-                        }
+                __inode_write_fop_cbk (frame, child_index, read_child, this,
+                                       &op_ret, &op_errno, preop, postop,
+                                       xdata);
 
-			local->success_count++;
-
-			if ((local->success_count >= priv->wait_count)
-                            && local->read_child_returned) {
-				need_unwind = 1;
-			}
-		}
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                if ((local->success_count >= priv->wait_count)
+                    && local->read_child_returned) {
+                        need_unwind = 1;
+                }
+        }
+        UNLOCK (&frame->lock);
 
-	if (need_unwind)
-		local->transaction.unwind (frame, this);
+        if (need_unwind)
+                local->transaction.unwind (frame, this);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_setattr_wind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv = this->private;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+        local = frame->local;
+        priv = this->private;
 
-	local->call_count = call_count;
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_setattr_wind_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->setattr,
-					   &local->loc,
-					   &local->cont.setattr.in_buf,
-                                           local->cont.setattr.valid);
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
-			if (!--call_count)
-				break;
-		}
-	}
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_setattr_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->setattr,
+                                           &local->loc,
+                                           &local->cont.setattr.in_buf,
+                                           local->cont.setattr.valid,
+                                           NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_setattr_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
+        afr_local_t *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
+        AFR_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_setattr (call_frame_t *frame, xlator_t *this,
-             loc_t *loc, struct iatt *buf, int32_t valid)
+             loc_t *loc, struct iatt *buf, int32_t valid, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t   *transaction_frame = NULL;
-
-	int ret = -1;
-
-	int op_ret   = -1;
-	int op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-
-	priv = this->private;
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
+        call_frame_t   *transaction_frame = NULL;
+        int ret = -1;
+        int op_errno = 0;
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        priv = this->private;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        QUORUM_CHECK(setattr,out);
 
-	transaction_frame->local = local;
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	local->op_ret = -1;
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
 
-	local->cont.setattr.ino     = loc->inode->ino;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
         local->cont.setattr.in_buf = *buf;
         local->cont.setattr.valid  = valid;
 
-	local->transaction.fop    = afr_setattr_wind;
-	local->transaction.done   = afr_setattr_done;
-	local->transaction.unwind = afr_setattr_unwind;
+        local->transaction.fop    = afr_setattr_wind;
+        local->transaction.done   = afr_setattr_done;
+        local->transaction.unwind = afr_setattr_unwind;
 
-	loc_copy (&local->loc, loc);
+        loc_copy (&local->loc, loc);
 
-	local->transaction.main_frame = frame;
-	local->transaction.start   = LLONG_MAX - 1;
-	local->transaction.len     = 0;
+        local->transaction.main_frame = frame;
+        local->transaction.start   = LLONG_MAX - 1;
+        local->transaction.len     = 0;
 
-	afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (setattr, frame, op_ret, op_errno, NULL, NULL);
-	}
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
+        }
 
-	return 0;
+        return 0;
 }
 
 /* {{{ fsetattr */
@@ -1010,217 +1203,199 @@ out:
 int
 afr_fsetattr_unwind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	call_frame_t   *main_frame = NULL;
-
-	local = frame->local;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame)
-			main_frame = local->transaction.main_frame;
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        local = frame->local;
 
-	if (main_frame) {
-                local->cont.fsetattr.preop_buf.ia_ino  =
-                        local->cont.fsetattr.ino;
-                local->cont.fsetattr.postop_buf.ia_ino =
-                        local->cont.fsetattr.ino;
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame)
+                        main_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
+        if (main_frame) {
                 AFR_STACK_UNWIND (fsetattr, main_frame, local->op_ret,
                                   local->op_errno,
-                                  &local->cont.fsetattr.preop_buf,
-                                  &local->cont.fsetattr.postop_buf);
+                                  &local->cont.inode_wfop.prebuf,
+                                  &local->cont.inode_wfop.postbuf,
+                                  NULL);
         }
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_fsetattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno,
-                      struct iatt *preop, struct iatt *postop)
+                       int32_t op_ret, int32_t op_errno,
+                       struct iatt *preop, struct iatt *postop, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int child_index = (long) cookie;
+        afr_local_t *   local = NULL;
+        afr_private_t * priv  = NULL;
+        int child_index = (long) cookie;
         int read_child  = 0;
-	int call_count  = -1;
-	int need_unwind = 0;
+        int call_count  = -1;
+        int need_unwind = 0;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
+        priv  = this->private;
 
-        read_child = afr_read_child (this, local->fd->inode);
+        read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
 
-	LOCK (&frame->lock);
-	{
+        LOCK (&frame->lock);
+        {
                 if (child_index == read_child) {
                         local->read_child_returned = _gf_true;
                 }
 
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
+                __inode_write_fop_cbk (frame, child_index, read_child, this,
+                                       &op_ret, &op_errno, preop, postop,
+                                       xdata);
 
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-				local->cont.fsetattr.preop_buf  = *preop;
-                                local->cont.fsetattr.postop_buf = *postop;
-			}
-
-                        if (child_index == read_child) {
-                                local->cont.fsetattr.preop_buf  = *preop;
-                                local->cont.fsetattr.postop_buf = *postop;
-                        }
-
-			local->success_count++;
-
-			if ((local->success_count >= priv->wait_count)
-                            && local->read_child_returned) {
-				need_unwind = 1;
-			}
-		}
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+                if ((local->success_count >= priv->wait_count)
+                    && local->read_child_returned) {
+                        need_unwind = 1;
+                }
+        }
+        UNLOCK (&frame->lock);
 
-	if (need_unwind)
-		local->transaction.unwind (frame, this);
+        if (need_unwind)
+                local->transaction.unwind (frame, this);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_fsetattr_wind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv = this->private;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+        local = frame->local;
+        priv = this->private;
 
-	local->call_count = call_count;
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_fsetattr_wind_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->fsetattr,
-					   local->fd,
-					   &local->cont.fsetattr.in_buf,
-                                           local->cont.fsetattr.valid);
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
-			if (!--call_count)
-				break;
-		}
-	}
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_fsetattr_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->fsetattr,
+                                           local->fd,
+                                           &local->cont.fsetattr.in_buf,
+                                           local->cont.fsetattr.valid,
+                                           NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_fsetattr_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
+        afr_local_t *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
+        AFR_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
-
 int
 afr_fsetattr (call_frame_t *frame, xlator_t *this,
-              fd_t *fd, struct iatt *buf, int32_t valid)
+              fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t   *transaction_frame = NULL;
-
-	int ret = -1;
-
-	int op_ret   = -1;
-	int op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
+        call_frame_t   *transaction_frame = NULL;
+        int ret = -1;
+        int op_errno = 0;
 
-	priv = this->private;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
+        priv = this->private;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        if (afr_is_split_brain (this, fd->inode)) {
+                op_errno = EIO;
+                goto out;
+        }
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        QUORUM_CHECK(fsetattr,out);
 
-	transaction_frame->local = local;
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	local->op_ret = -1;
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
 
-	local->cont.fsetattr.ino     = fd->inode->ino;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
         local->cont.fsetattr.in_buf = *buf;
         local->cont.fsetattr.valid  = valid;
 
-	local->transaction.fop    = afr_fsetattr_wind;
-	local->transaction.done   = afr_fsetattr_done;
-	local->transaction.unwind = afr_fsetattr_unwind;
+        local->transaction.fop    = afr_fsetattr_wind;
+        local->transaction.done   = afr_fsetattr_done;
+        local->transaction.unwind = afr_fsetattr_unwind;
 
         local->fd                 = fd_ref (fd);
 
-	local->transaction.main_frame = frame;
-	local->transaction.start   = LLONG_MAX - 1;
-	local->transaction.len     = 0;
+        afr_open_fd_fix (fd, this);
 
-	afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        local->transaction.main_frame = frame;
+        local->transaction.start   = LLONG_MAX - 1;
+        local->transaction.len     = 0;
 
-	op_ret = 0;
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
+
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (fsetattr, frame, op_ret, op_errno, NULL, NULL);
-	}
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
+        }
 
-	return 0;
+        return 0;
 }
 
 
@@ -1230,370 +1405,1457 @@ out:
 int
 afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	call_frame_t   *main_frame = NULL;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame)
-			main_frame = local->transaction.main_frame;
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame)
+                        main_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-	if (main_frame) {
-		AFR_STACK_UNWIND (setxattr, main_frame,
-                                  local->op_ret, local->op_errno)
-	}
-	return 0;
+        if (main_frame) {
+                AFR_STACK_UNWIND (setxattr, main_frame,
+                                  local->op_ret, local->op_errno,
+                                  NULL);
+        }
+        return 0;
 }
 
 
 int
-afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		       int32_t op_ret, int32_t op_errno)
+afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
+        afr_local_t   *local      = NULL;
+        afr_private_t *priv       = NULL;
+        int           call_count  = -1;
+        int           need_unwind = 0;
+        int           child_index = (long) cookie;
 
-	int call_count  = -1;
-	int need_unwind = 0;
+        local = frame->local;
+        priv = this->private;
+
+        LOCK (&frame->lock);
+        {
+                __inode_write_fop_cbk (frame, child_index, -1, this,
+                                       &op_ret, &op_errno, NULL, NULL,
+                                       xdata);
+                if (local->success_count == priv->child_count) {
+                        need_unwind = 1;
+                }
+        }
+        UNLOCK (&frame->lock);
 
-	local = frame->local;
-	priv = this->private;
+        if (need_unwind)
+                local->transaction.unwind (frame, this);
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-			}
-			local->success_count++;
-
-			if (local->success_count == priv->child_count) {
-				need_unwind = 1;
-			}
-		}
+        call_count = afr_frame_return (frame);
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+        }
 
-	if (need_unwind)
-		local->transaction.unwind (frame, this);
+        return 0;
+}
 
-	call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
+int
+afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t      *local         = NULL;
+        afr_private_t    *priv          = NULL;
+        int               call_count    = -1;
+        int               i             = 0;
+
+        local = frame->local;
+        priv = this->private;
+
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
+
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_setxattr_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->setxattr,
+                                           &local->loc,
+                                           local->cont.setxattr.dict,
+                                           local->cont.setxattr.flags,
+                                           NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
 }
 
 
 int
-afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
+afr_setxattr_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
+        afr_local_t   *local    = frame->local;
 
-	int call_count = -1;
-	int i = 0;
+        local->transaction.unwind (frame, this);
 
-	local = frame->local;
-	priv = this->private;
+        AFR_STACK_DESTROY (frame);
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+        return 0;
+}
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+int
+afr_setxattr (call_frame_t *frame, xlator_t *this,
+              loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        afr_private_t  *priv              = NULL;
+        afr_local_t    *local             = NULL;
+        call_frame_t   *transaction_frame = NULL;
+        int             ret               = -1;
+        int             op_errno          = EINVAL;
 
-	local->call_count = call_count;
+        VALIDATE_OR_GOTO (this, out);
 
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_setxattr_wind_cbk,
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->setxattr,
-					   &local->loc, 
-					   local->cont.setxattr.dict,
-					   local->cont.setxattr.flags); 
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
+                                   op_errno, out);
 
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
+                                   op_errno, out);
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        priv = this->private;
+
+        QUORUM_CHECK(setxattr,out);
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
+
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
+
+        local->cont.setxattr.dict  = dict_ref (dict);
+        local->cont.setxattr.flags = flags;
+
+        local->transaction.fop    = afr_setxattr_wind;
+        local->transaction.done   = afr_setxattr_done;
+        local->transaction.unwind = afr_setxattr_unwind;
+
+        loc_copy (&local->loc, loc);
+
+        local->transaction.main_frame = frame;
+        local->transaction.start   = LLONG_MAX - 1;
+        local->transaction.len     = 0;
+
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
+
+        ret = 0;
+out:
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
+        }
+
+        return 0;
 }
 
+/* {{{ fsetxattr */
+
 
 int
-afr_setxattr_done (call_frame_t *frame, xlator_t *this)
+afr_fsetxattr_unwind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t * local = frame->local;
+        afr_local_t    *local         = NULL;
+        call_frame_t   *main_frame    = NULL;
 
-	local->transaction.unwind (frame, this);
+        local = frame->local;
 
-	AFR_STACK_DESTROY (frame);
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame)
+                        main_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-	return 0;
+        if (main_frame) {
+                AFR_STACK_UNWIND (fsetxattr, main_frame,
+                                  local->op_ret, local->op_errno,
+                                  NULL);
+        }
+        return 0;
 }
 
+
 int
-afr_setxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, dict_t *dict, int32_t flags)
+afr_fsetxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t   *transaction_frame = NULL;
+        afr_local_t   *local      = NULL;
+        afr_private_t *priv       = NULL;
+        int           call_count  = -1;
+        int           need_unwind = 0;
+        int           child_index = (long) cookie;
 
-	int ret = -1;
+        local = frame->local;
+        priv = this->private;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        LOCK (&frame->lock);
+        {
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+                __inode_write_fop_cbk (frame, child_index, -1, this,
+                                       &op_ret, &op_errno, NULL, NULL,
+                                       xdata);
+                if (local->success_count == priv->child_count) {
+                        need_unwind = 1;
+                }
+        }
+        UNLOCK (&frame->lock);
 
-	priv = this->private;
+        if (need_unwind)
+                local->transaction.unwind (frame, this);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        call_count = afr_frame_return (frame);
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+        }
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
+        return 0;
+}
+
+
+int
+afr_fsetxattr_wind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t        *local       = NULL;
+        afr_private_t      *priv        = NULL;
+        int                 call_count  = -1;
+        int                 i           = 0;
+
+        local = frame->local;
+        priv = this->private;
 
-	transaction_frame->local = local;
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
+
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_fsetxattr_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->fsetxattr,
+                                           local->fd,
+                                           local->cont.fsetxattr.dict,
+                                           local->cont.fsetxattr.flags,
+                                           NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
+}
+
+
+int
+afr_fsetxattr_done (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t   *local   = frame->local;
+
+        local->transaction.unwind (frame, this);
+
+        AFR_STACK_DESTROY (frame);
+
+        return 0;
+}
+
+int
+afr_fsetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        afr_private_t    *priv              = NULL;
+        afr_local_t      *local             = NULL;
+        call_frame_t     *transaction_frame = NULL;
+        int               ret               = -1;
+        int               op_errno          = EINVAL;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
+                                   op_errno, out);
+
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
+                                   op_errno, out);
+
+        priv = this->private;
+
+        if (afr_is_split_brain (this, fd->inode)) {
+                op_errno = EIO;
+                goto out;
+        }
+
+        QUORUM_CHECK(fsetxattr,out);
+
+        AFR_LOCAL_ALLOC_OR_GOTO (local, out);
+
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
+
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                goto out;
+        }
+
+        transaction_frame->local = local;
 
-	local->op_ret = -1;
+        local->op_ret = -1;
 
-	local->cont.setxattr.dict  = dict_ref (dict);
-	local->cont.setxattr.flags = flags;
+        local->cont.fsetxattr.dict  = dict_ref (dict);
+        local->cont.fsetxattr.flags = flags;
 
-	local->transaction.fop    = afr_setxattr_wind;
-	local->transaction.done   = afr_setxattr_done;
-	local->transaction.unwind = afr_setxattr_unwind;
+        local->transaction.fop    = afr_fsetxattr_wind;
+        local->transaction.done   = afr_fsetxattr_done;
+        local->transaction.unwind = afr_fsetxattr_unwind;
 
-	loc_copy (&local->loc, loc);
+        local->fd                 = fd_ref (fd);
 
-	local->transaction.main_frame = frame;
-	local->transaction.start   = LLONG_MAX - 1;
-	local->transaction.len     = 0;
+        local->transaction.main_frame = frame;
+        local->transaction.start  = LLONG_MAX - 1;
+        local->transaction.len    = 0;
 
-	afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (setxattr, frame, op_ret, op_errno);
-	}
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
+        }
 
-	return 0;
+        return 0;
 }
 
 /* }}} */
 
+
 /* {{{ removexattr */
 
 
 int
 afr_removexattr_unwind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	call_frame_t   *main_frame = NULL;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame)
-			main_frame = local->transaction.main_frame;
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame)
+                        main_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
 
-	if (main_frame) {
-		AFR_STACK_UNWIND (removexattr, main_frame,
-                                  local->op_ret, local->op_errno)
-	}
-	return 0;
+        if (main_frame) {
+                AFR_STACK_UNWIND (removexattr, main_frame,
+                                  local->op_ret, local->op_errno,
+                                  NULL);
+        }
+        return 0;
 }
 
 
 int
-afr_removexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-			  int32_t op_ret, int32_t op_errno)
+afr_removexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
+        afr_local_t   *local      = NULL;
+        afr_private_t *priv       = NULL;
+        int           call_count  = -1;
+        int           need_unwind = 0;
+        int           child_index = (long) cookie;
 
-	int call_count  = -1;
-	int need_unwind = 0;
-
-	local = frame->local;
-	priv = this->private;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-			}
-			local->success_count++;
-
-			if (local->success_count == priv->wait_count) {
-				need_unwind = 1;
-			}
-		}
+        local = frame->local;
+        priv = this->private;
+
+        LOCK (&frame->lock);
+        {
+                __inode_write_fop_cbk (frame, child_index, -1, this,
+                                       &op_ret, &op_errno, NULL, NULL,
+                                       xdata);
+                if (local->success_count == priv->wait_count) {
+                        need_unwind = 1;
+                }
+        }
+        UNLOCK (&frame->lock);
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        if (need_unwind)
+                local->transaction.unwind (frame, this);
 
-	if (need_unwind)
-		local->transaction.unwind (frame, this);
+        call_count = afr_frame_return (frame);
 
-	call_count = afr_frame_return (frame);
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+        }
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_removexattr_wind (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
-	int call_count = -1;
-	int i = 0;
+        local = frame->local;
+        priv = this->private;
 
-	local = frame->local;
-	priv = this->private;
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+        local->call_count = call_count;
 
-	local->call_count = call_count;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_removexattr_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->removexattr,
+                                           &local->loc,
+                                           local->cont.removexattr.name,
+                                           NULL);
 
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_removexattr_wind_cbk, 
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->removexattr,
-					   &local->loc, 
-					   local->cont.removexattr.name);
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        return 0;
 }
 
 
 int
 afr_removexattr_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t * local = frame->local;
+        afr_local_t * local = frame->local;
+
+        local->transaction.unwind (frame, this);
 
-	local->transaction.unwind (frame, this);
+        AFR_STACK_DESTROY (frame);
 
-	AFR_STACK_DESTROY (frame);
-	
-	return 0;
+        return 0;
 }
 
 
 int
 afr_removexattr (call_frame_t *frame, xlator_t *this,
-		 loc_t *loc, const char *name)
+                 loc_t *loc, const char *name, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t   *transaction_frame = NULL;
+        afr_private_t   *priv              = NULL;
+        afr_local_t     *local             = NULL;
+        call_frame_t    *transaction_frame = NULL;
+        int              ret               = -1;
+        int              op_errno          = 0;
 
-	int ret = -1;
+        VALIDATE_OR_GOTO (this, out);
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
+                                 name, op_errno, out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-	VALIDATE_OR_GOTO (loc, out);
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
+                                 name, op_errno, out);
 
-	priv = this->private;
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this->private, out);
+        VALIDATE_OR_GOTO (loc, out);
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
+        priv = this->private;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        QUORUM_CHECK(removexattr,out);
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	transaction_frame->local = local;
+        AFR_LOCAL_ALLOC_OR_GOTO (transaction_frame->local, out);
+        local = transaction_frame->local;
 
-	local->op_ret = -1;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-	local->cont.removexattr.name = gf_strdup (name);
+        local->cont.removexattr.name = gf_strdup (name);
 
-	local->transaction.fop    = afr_removexattr_wind;
-	local->transaction.done   = afr_removexattr_done;
-	local->transaction.unwind = afr_removexattr_unwind;
+        local->transaction.fop    = afr_removexattr_wind;
+        local->transaction.done   = afr_removexattr_done;
+        local->transaction.unwind = afr_removexattr_unwind;
 
-	loc_copy (&local->loc, loc);
+        loc_copy (&local->loc, loc);
 
-	local->transaction.main_frame = frame;
-	local->transaction.start   = LLONG_MAX - 1;
-	local->transaction.len     = 0;
+        local->transaction.main_frame = frame;
+        local->transaction.start   = LLONG_MAX - 1;
+        local->transaction.len     = 0;
 
-	afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+        ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (removexattr, frame, op_ret, op_errno);
-	}
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
+        }
+
+        return 0;
+}
+
+/* ffremovexattr */
+int
+afr_fremovexattr_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame)
+                        main_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
+
+        if (main_frame) {
+                AFR_STACK_UNWIND (fremovexattr, main_frame,
+                                  local->op_ret, local->op_errno,
+                                  NULL);
+        }
+        return 0;
+}
+
+
+int
+afr_fremovexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        afr_local_t *   local       = NULL;
+        afr_private_t * priv        = NULL;
+        int             call_count  = -1;
+        int             need_unwind = 0;
+        int             child_index = (long) cookie;
+
+        local = frame->local;
+        priv = this->private;
+
+        LOCK (&frame->lock);
+        {
+                __inode_write_fop_cbk (frame, child_index, -1, this,
+                                       &op_ret, &op_errno, NULL, NULL,
+                                       xdata);
+
+                if (local->success_count == priv->wait_count) {
+                        need_unwind = 1;
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        if (need_unwind)
+                local->transaction.unwind (frame, this);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+        }
+
+        return 0;
+}
+
+
+int32_t
+afr_fremovexattr_wind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
+
+        local = frame->local;
+        priv = this->private;
+
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
+
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_fremovexattr_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->fremovexattr,
+                                           local->fd,
+                                           local->cont.removexattr.name,
+                                           NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
+}
+
+
+int
+afr_fremovexattr_done (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t * local = frame->local;
+
+        local->transaction.unwind (frame, this);
+
+        AFR_STACK_DESTROY (frame);
+
+        return 0;
+}
+
+
+int
+afr_fremovexattr (call_frame_t *frame, xlator_t *this,
+                  fd_t *fd, const char *name, dict_t *xdata)
+{
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
+        call_frame_t   *transaction_frame = NULL;
+        int ret = -1;
+        int op_ret   = -1;
+        int op_errno = 0;
+
+        VALIDATE_OR_GOTO (this, out);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
+                                 name, op_errno, out);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
+                                 name, op_errno, out);
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        priv = this->private;
+        if (afr_is_split_brain (this, fd->inode)) {
+                op_errno = EIO;
+                goto out;
+        }
+
+        QUORUM_CHECK(fremovexattr, out);
+
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                goto out;
+        }
+
+        AFR_LOCAL_ALLOC_OR_GOTO (local, out);
+
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto out;
+        }
+
+        transaction_frame->local = local;
+
+        local->op_ret = -1;
+
+        local->cont.removexattr.name = gf_strdup (name);
+
+        local->transaction.fop    = afr_fremovexattr_wind;
+        local->transaction.done   = afr_fremovexattr_done;
+        local->transaction.unwind = afr_fremovexattr_unwind;
+
+        local->fd = fd_ref (fd);
+
+        local->transaction.main_frame = frame;
+        local->transaction.start   = LLONG_MAX - 1;
+        local->transaction.len     = 0;
+
+        op_ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (op_ret < 0) {
+            op_errno = -op_ret;
+            goto out;
+        }
+
+        op_ret = 0;
+out:
+        if (op_ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, NULL);
+        }
+
+        return 0;
+}
+
+static int
+afr_fallocate_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame)
+                        main_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
+
+        if (main_frame) {
+                AFR_STACK_UNWIND (fallocate, main_frame, local->op_ret,
+                                  local->op_errno,
+                                  &local->cont.inode_wfop.prebuf,
+                                  &local->cont.inode_wfop.postbuf,
+                                  NULL);
+        }
+        return 0;
+}
+
+static int
+afr_fallocate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                        struct iatt *postbuf, dict_t *xdata)
+{
+        afr_local_t *   local = NULL;
+        afr_private_t * priv  = NULL;
+        int child_index = (long) cookie;
+        int call_count  = -1;
+        int need_unwind = 0;
+        int read_child  = 0;
+
+        local = frame->local;
+        priv  = this->private;
+
+        read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
+
+        LOCK (&frame->lock);
+        {
+                if (child_index == read_child) {
+                        local->read_child_returned = _gf_true;
+                }
+
+                __inode_write_fop_cbk (frame, child_index, read_child, this,
+                                       &op_ret, &op_errno, prebuf, postbuf,
+                                       xdata);
+
+                if ((local->success_count >= priv->wait_count)
+                    && local->read_child_returned) {
+                        need_unwind = 1;
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        if (need_unwind)
+                local->transaction.unwind (frame, this);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+        }
+
+        return 0;
+}
+
+static int
+afr_fallocate_wind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
+
+        local = frame->local;
+        priv = this->private;
+
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
+
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_fallocate_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->fallocate,
+                                           local->fd,
+                                           local->cont.fallocate.mode,
+                                           local->cont.fallocate.offset,
+                                           local->cont.fallocate.len,
+                                           NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
+}
+
+static int
+afr_fallocate_done (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+
+        local = frame->local;
+
+        local->transaction.unwind (frame, this);
+
+        AFR_STACK_DESTROY (frame);
+
+        return 0;
+}
+
+static int
+afr_do_fallocate (call_frame_t *frame, xlator_t *this)
+{
+        call_frame_t * transaction_frame = NULL;
+        afr_local_t *  local             = NULL;
+        int op_ret   = -1;
+        int op_errno = 0;
+
+        local = frame->local;
+
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                goto out;
+        }
+
+        transaction_frame->local = local;
+        frame->local = NULL;
+
+        local->op = GF_FOP_FALLOCATE;
+
+        local->transaction.fop    = afr_fallocate_wind;
+        local->transaction.done   = afr_fallocate_done;
+        local->transaction.unwind = afr_fallocate_unwind;
+
+        local->transaction.main_frame = frame;
+
+        local->transaction.start   = local->cont.fallocate.offset;
+        local->transaction.len     = 0;
+
+        /* fallocate can modify the file size */
+        op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        if (op_ret < 0) {
+            op_errno = -op_ret;
+            goto out;
+        }
+
+        op_ret = 0;
+out:
+        if (op_ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (fallocate, frame, op_ret, op_errno, NULL,
+                                  NULL, NULL);
+        }
+
+        return 0;
+}
+
+int
+afr_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+               off_t offset, size_t len, dict_t *xdata)
+{
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
+        call_frame_t   *transaction_frame = NULL;
+        int ret = -1;
+        int op_errno = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        priv = this->private;
+
+        if (afr_is_split_brain (this, fd->inode)) {
+                op_errno = EIO;
+                goto out;
+        }
+        QUORUM_CHECK(fallocate,out);
+
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
+
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
+
+        local->cont.fallocate.mode = mode;
+        local->cont.fallocate.offset  = offset;
+        local->cont.fallocate.len = len;
+
+        local->fd = fd_ref (fd);
+
+        afr_open_fd_fix (fd, this);
+
+        afr_do_fallocate (frame, this);
+
+        ret = 0;
+out:
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+        }
+
+        return 0;
+}
+
+/* }}} */
+
+/* {{{ discard */
+
+static int
+afr_discard_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame)
+                        main_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
+
+        if (main_frame) {
+                AFR_STACK_UNWIND (discard, main_frame, local->op_ret,
+                                  local->op_errno,
+                                  &local->cont.inode_wfop.prebuf,
+                                  &local->cont.inode_wfop.postbuf,
+                                  NULL);
+        }
+        return 0;
+}
+
+static int
+afr_discard_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                      struct iatt *postbuf, dict_t *xdata)
+{
+        afr_local_t *   local = NULL;
+        afr_private_t * priv  = NULL;
+        int child_index = (long) cookie;
+        int call_count  = -1;
+        int need_unwind = 0;
+        int read_child  = 0;
+
+        local = frame->local;
+        priv  = this->private;
+
+        read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
+
+        LOCK (&frame->lock);
+        {
+                if (child_index == read_child) {
+                        local->read_child_returned = _gf_true;
+                }
+
+                __inode_write_fop_cbk (frame, child_index, read_child, this,
+                                       &op_ret, &op_errno, prebuf, postbuf,
+                                       xdata);
+
+                if ((local->success_count >= priv->wait_count)
+                    && local->read_child_returned) {
+                        need_unwind = 1;
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        if (need_unwind)
+                local->transaction.unwind (frame, this);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+        }
+
+        return 0;
+}
+
+static int
+afr_discard_wind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
-	return 0;
+        local = frame->local;
+        priv = this->private;
+
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
+
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_discard_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->discard,
+                                           local->fd,
+                                           local->cont.discard.offset,
+                                           local->cont.discard.len,
+                                           NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
 }
+
+static int
+afr_discard_done (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+
+        local = frame->local;
+
+        local->transaction.unwind (frame, this);
+
+        AFR_STACK_DESTROY (frame);
+
+        return 0;
+}
+
+static int
+afr_do_discard (call_frame_t *frame, xlator_t *this)
+{
+        call_frame_t * transaction_frame = NULL;
+        afr_local_t *  local             = NULL;
+        int op_ret   = -1;
+        int op_errno = 0;
+
+        local = frame->local;
+
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                goto out;
+        }
+
+        transaction_frame->local = local;
+        frame->local = NULL;
+
+        local->op = GF_FOP_DISCARD;
+
+        local->transaction.fop    = afr_discard_wind;
+        local->transaction.done   = afr_discard_done;
+        local->transaction.unwind = afr_discard_unwind;
+
+        local->transaction.main_frame = frame;
+
+        local->transaction.start   = local->cont.discard.offset;
+        local->transaction.len     = 0;
+
+        op_ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        if (op_ret < 0) {
+            op_errno = -op_ret;
+            goto out;
+        }
+
+        op_ret = 0;
+out:
+        if (op_ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (discard, frame, op_ret, op_errno, NULL,
+                                  NULL, NULL);
+        }
+
+        return 0;
+}
+
+int
+afr_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+             size_t len, dict_t *xdata)
+{
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
+        call_frame_t   *transaction_frame = NULL;
+        int ret = -1;
+        int op_errno = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        priv = this->private;
+
+        if (afr_is_split_brain (this, fd->inode)) {
+                op_errno = EIO;
+                goto out;
+        }
+        QUORUM_CHECK(discard, out);
+
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
+
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
+
+        local->cont.discard.offset  = offset;
+        local->cont.discard.len = len;
+
+        local->fd = fd_ref (fd);
+
+        afr_open_fd_fix (fd, this);
+
+        afr_do_discard(frame, this);
+
+        ret = 0;
+out:
+        if (ret < 0) {
+                if (transaction_frame)
+                        AFR_STACK_DESTROY (transaction_frame);
+                AFR_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
+        }
+
+        return 0;
+}
+
+
+/* {{{ zerofill */
+
+static int
+afr_zerofill_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local            = NULL;
+        call_frame_t    *main_frame       = NULL;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                if (local->transaction.main_frame) {
+                        main_frame = local->transaction.main_frame;
+                }
+                local->transaction.main_frame = NULL;
+        }
+        UNLOCK (&frame->lock);
+
+        if (main_frame) {
+                AFR_STACK_UNWIND (zerofill, main_frame, local->op_ret,
+                                  local->op_errno,
+                                  &local->cont.zerofill.prebuf,
+                                  &local->cont.zerofill.postbuf,
+                                  NULL);
+        }
+        return 0;
+}
+
+static int
+afr_zerofill_wind_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
+{
+        afr_local_t       *local             = NULL;
+        afr_private_t     *priv              = NULL;
+        int                child_index       = (long) cookie;
+        int                call_count        = -1;
+        int                need_unwind       = 0;
+        int                read_child        = 0;
+
+        local = frame->local;
+        priv  = this->private;
+
+        read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
+
+        LOCK (&frame->lock);
+        {
+                if (child_index == read_child) {
+                        local->read_child_returned = _gf_true;
+                }
+
+                if (afr_fop_failed (op_ret, op_errno)) {
+                        afr_transaction_fop_failed (frame, this, child_index);
+                }
+
+                if (op_ret != -1) {
+                        if (local->success_count == 0) {
+                                local->op_ret = op_ret;
+                                local->cont.zerofill.prebuf  = *prebuf;
+                                local->cont.zerofill.postbuf = *postbuf;
+                        }
+
+                        if (child_index == read_child) {
+                                local->cont.zerofill.prebuf  = *prebuf;
+                                local->cont.zerofill.postbuf = *postbuf;
+                        }
+
+                        local->success_count++;
+
+                        if ((local->success_count >= priv->wait_count)
+                            && local->read_child_returned) {
+                                need_unwind = 1;
+                        }
+                }
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
+
+        if (need_unwind) {
+                local->transaction.unwind (frame, this);
+        }
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+        }
+
+        return 0;
+}
+
+static int
+afr_zerofill_wind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t    *local         = NULL;
+        afr_private_t  *priv          = NULL;
+        int             call_count    = -1;
+        int             i             = 0;
+
+        local = frame->local;
+        priv = this->private;
+
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
+
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
+
+        local->call_count = call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+                        STACK_WIND_COOKIE (frame, afr_zerofill_wind_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->zerofill,
+                                           local->fd,
+                                           local->cont.zerofill.offset,
+                                           local->cont.zerofill.len,
+                                           NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
+}
+
+static int
+afr_zerofill_done (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+
+        local = frame->local;
+
+        local->transaction.unwind (frame, this);
+
+        AFR_STACK_DESTROY (frame);
+
+        return 0;
+}
+
+static int
+afr_do_zerofill(call_frame_t *frame, xlator_t *this)
+{
+        call_frame_t  *transaction_frame = NULL;
+        afr_local_t   *local             = NULL;
+        int            op_ret            = -1;
+        int            op_errno          = 0;
+
+        local = frame->local;
+
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame) {
+                goto out;
+        }
+
+        transaction_frame->local = local;
+        frame->local = NULL;
+
+        local->op = GF_FOP_ZEROFILL;
+
+        local->transaction.fop    = afr_zerofill_wind;
+        local->transaction.done   = afr_zerofill_done;
+        local->transaction.unwind = afr_zerofill_unwind;
+
+        local->transaction.main_frame = frame;
+
+        local->transaction.start   = local->cont.zerofill.offset;
+        local->transaction.len     = 0;
+
+        op_ret = afr_transaction (transaction_frame, this,
+                                  AFR_DATA_TRANSACTION);
+        if (op_ret < 0) {
+                op_errno = -op_ret;
+                goto out;
+        }
+
+        op_ret = 0;
+out:
+        if (op_ret < 0) {
+                if (transaction_frame) {
+                        AFR_STACK_DESTROY (transaction_frame);
+                }
+                AFR_STACK_UNWIND (zerofill, frame, op_ret, op_errno, NULL,
+                                  NULL, NULL);
+        }
+
+        return 0;
+}
+
+int
+afr_zerofill (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+              size_t len, dict_t *xdata)
+{
+        afr_private_t   *priv               = NULL;
+        afr_local_t     *local              = NULL;
+        call_frame_t    *transaction_frame  = NULL;
+        int              ret                = -1;
+        int              op_errno           = 0;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        priv = this->private;
+
+        if (afr_is_split_brain (this, fd->inode)) {
+                op_errno = EIO;
+                goto out;
+        }
+        QUORUM_CHECK(zerofill, out);
+
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+        local = frame->local;
+
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0) {
+                goto out;
+        }
+        local->cont.zerofill.offset  = offset;
+        local->cont.zerofill.len = len;
+
+        local->fd = fd_ref (fd);
+
+        afr_open_fd_fix (fd, this);
+
+        afr_do_zerofill(frame, this);
+
+        ret = 0;
+out:
+        if (ret < 0) {
+                if (transaction_frame) {
+                        AFR_STACK_DESTROY (transaction_frame);
+                }
+                AFR_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL,
+                                  NULL, NULL);
+        }
+
+        return 0;
+}
+
+/* }}} */
+
+
diff --git a/xlators/cluster/afr/src/afr-inode-write.h b/xlators/cluster/afr/src/afr-inode-write.h
index f0c2fbe00..8e93ca44a 100644
--- a/xlators/cluster/afr/src/afr-inode-write.h
+++ b/xlators/cluster/afr/src/afr-inode-write.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __INODE_WRITE_H__
@@ -22,51 +13,70 @@
 
 int32_t
 afr_chmod (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode);
+	   loc_t *loc, mode_t mode, dict_t *xdata);
 
 int32_t
 afr_chown (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, uid_t uid, gid_t gid);
+	   loc_t *loc, uid_t uid, gid_t gid, dict_t *xdata);
 
 int
 afr_fchown (call_frame_t *frame, xlator_t *this,
-	    fd_t *fd, uid_t uid, gid_t gid);
+	    fd_t *fd, uid_t uid, gid_t gid, dict_t *xdata);
 
 int32_t
 afr_fchmod (call_frame_t *frame, xlator_t *this,
-	    fd_t *fd, mode_t mode);
+	    fd_t *fd, mode_t mode, dict_t *xdata);
 
 int32_t
-afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, 
+afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
 	    struct iovec *vector, int32_t count, off_t offset,
-            struct iobref *iobref);
+            uint32_t flags, struct iobref *iobref, dict_t *xdata);
 
 int32_t
 afr_truncate (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, off_t offset);
+	      loc_t *loc, off_t offset, dict_t *xdata);
 
 int32_t
 afr_ftruncate (call_frame_t *frame, xlator_t *this,
-	       fd_t *fd, off_t offset);
+	       fd_t *fd, off_t offset, dict_t *xdata);
 
 int32_t
 afr_utimens (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, struct timespec tv[2]);
+	     loc_t *loc, struct timespec tv[2], dict_t *xdata);
 
 int
 afr_setattr (call_frame_t *frame, xlator_t *this,
-             loc_t *loc, struct iatt *buf, int32_t valid);
+             loc_t *loc, struct iatt *buf, int32_t valid, dict_t *xdata);
 
 int
 afr_fsetattr (call_frame_t *frame, xlator_t *this,
-              fd_t *fd, struct iatt *buf, int32_t valid);
+              fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata);
 
 int32_t
 afr_setxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, dict_t *dict, int32_t flags);
+              loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata);
+
+int32_t
+afr_fsetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata);
 
 int32_t
 afr_removexattr (call_frame_t *frame, xlator_t *this,
-		 loc_t *loc, const char *name);
+		 loc_t *loc, const char *name, dict_t *xdata);
+
+int32_t
+afr_fremovexattr (call_frame_t *frame, xlator_t *this,
+                  fd_t *fd, const char *name, dict_t *xdata);
 
+int
+afr_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+             size_t len, dict_t *xdata);
+
+int
+afr_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+               off_t offset, size_t len, dict_t *xdata);
+
+int
+afr_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+             size_t len, dict_t *xdata);
 #endif /* __INODE_WRITE_H__ */
diff --git a/xlators/cluster/afr/src/afr-lk-common.c b/xlators/cluster/afr/src/afr-lk-common.c
index 749c6bf9a..060d78f35 100644
--- a/xlators/cluster/afr/src/afr-lk-common.c
+++ b/xlators/cluster/afr/src/afr-lk-common.c
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include "dict.h"
@@ -31,8 +22,69 @@
 #define LOCKED_YES      0x1        /* for DATA, METADATA, ENTRY and higher_path */
 #define LOCKED_LOWER    0x2        /* for lower path */
 
+#define AFR_TRACE_INODELK_IN(frame, this, params ...)           \
+        do {                                                    \
+                afr_private_t *_priv = this->private;           \
+                if (!_priv->inodelk_trace)                      \
+                        break;                                  \
+                afr_trace_inodelk_in (frame, this, params);     \
+        } while (0);
+
+#define AFR_TRACE_INODELK_OUT(frame, this, params ...)          \
+        do {                                                    \
+                afr_private_t *_priv = this->private;           \
+                if (!_priv->inodelk_trace)                      \
+                        break;                                  \
+                afr_trace_inodelk_out (frame, this, params);    \
+        } while (0);
+
+#define AFR_TRACE_ENTRYLK_IN(frame, this, params ...)           \
+        do {                                                    \
+                afr_private_t *_priv = this->private;           \
+                if (!_priv->entrylk_trace)                      \
+                        break;                                  \
+                afr_trace_entrylk_in (frame, this, params);     \
+        } while (0);
+
+#define AFR_TRACE_ENTRYLK_OUT(frame, this, params ...)          \
+        do {                                                    \
+                afr_private_t *_priv = this->private;           \
+                if (!_priv->entrylk_trace)                      \
+                        break;                                  \
+                afr_trace_entrylk_out (frame, this, params);    \
+        } while (0);
+
 int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
+afr_entry_lockee_cmp (const void *l1, const void *l2)
+{
+        const afr_entry_lockee_t       *r1 = l1;
+        const afr_entry_lockee_t       *r2 = l2;
+        int                            ret = 0;
+        uuid_t                         gfid1 = {0};
+        uuid_t                         gfid2 = {0};
+
+        loc_gfid ((loc_t*)&r1->loc, gfid1);
+        loc_gfid ((loc_t*)&r2->loc, gfid2);
+        ret = uuid_compare (gfid1, gfid2);
+        /*Entrylks with NULL basename are the 'smallest'*/
+        if (ret == 0) {
+                if (!r1->basename)
+                        return -1;
+                if (!r2->basename)
+                        return 1;
+                ret = strcmp (r1->basename, r2->basename);
+        }
+
+        if (ret <= 0)
+                return -1;
+        else
+                return 1;
+}
+
+int afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
+
+static int
+afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this);
 
 static uint64_t afr_lock_number = 1;
 
@@ -57,14 +109,13 @@ afr_set_lock_number (call_frame_t *frame, xlator_t *this)
 }
 
 void
-afr_set_lk_owner (call_frame_t *frame, xlator_t *this)
+afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner)
 {
-        if (!frame->root->lk_owner) {
-                gf_log (this->name, GF_LOG_TRACE,
-                        "Setting lk-owner=%llu",
-                        (unsigned long long) (unsigned long)frame->root);
-                frame->root->lk_owner = (uint64_t) (unsigned long)frame->root;
-        }
+        gf_log (this->name, GF_LOG_TRACE,
+                "Setting lk-owner=%llu",
+                (unsigned long long) (unsigned long)lk_owner);
+
+        set_lk_owner_from_ptr (&frame->root->lk_owner, lk_owner);
 }
 
 static int
@@ -90,29 +141,19 @@ is_afr_lock_selfheal (afr_local_t *local)
 }
 
 int32_t
-internal_lock_count (call_frame_t *frame, xlator_t *this,
-                     afr_fd_ctx_t *fd_ctx)
+internal_lock_count (call_frame_t *frame, xlator_t *this)
 {
         afr_local_t   *local = NULL;
         afr_private_t *priv  = NULL;
-
         int32_t call_count = 0;
         int i = 0;
 
         local = frame->local;
         priv  = this->private;
 
-        if (fd_ctx) {
-                GF_ASSERT (local->fd);
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i] && fd_ctx->opened_on[i])
-                                ++call_count;
-                }
-        } else {
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i])
-                                ++call_count;
-                }
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i])
+                        ++call_count;
         }
 
         return call_count;
@@ -120,7 +161,7 @@ internal_lock_count (call_frame_t *frame, xlator_t *this,
 
 static void
 afr_print_inodelk (char *str, int size, int cmd,
-                   struct flock *flock, uint64_t owner)
+                   struct gf_flock *flock, gf_lkowner_t *owner)
 {
         char *cmd_str = NULL;
         char *type_str = NULL;
@@ -168,11 +209,11 @@ afr_print_inodelk (char *str, int size, int cmd,
         }
 
         snprintf (str, size, "lock=INODELK, cmd=%s, type=%s, "
-                  "start=%llu, len=%llu, pid=%llu, lk-owner=%llu",
+                  "start=%llu, len=%llu, pid=%llu, lk-owner=%s",
                   cmd_str, type_str, (unsigned long long) flock->l_start,
                   (unsigned long long) flock->l_len,
                   (unsigned long long) flock->l_pid,
-                  (unsigned long long) owner);
+                  lkowner_utoa (owner));
 
 }
 
@@ -188,11 +229,11 @@ afr_print_lockee (char *str, int size, loc_t *loc, fd_t *fd,
 
 void
 afr_print_entrylk (char *str, int size, const char *basename,
-                   uint64_t owner)
+                   gf_lkowner_t *owner)
 {
-        snprintf (str, size, "Basename=%s, lk-owner=%llu",
+        snprintf (str, size, "Basename=%s, lk-owner=%s",
                   basename ? basename : "<nul>",
-                  (unsigned long long)owner);
+                  lkowner_utoa (owner));
 }
 
 static void
@@ -246,27 +287,20 @@ afr_set_lock_call_type (afr_lock_call_type_t lock_call_type,
 }
 
 static void
-afr_trace_inodelk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
-                       afr_lock_op_type_t lk_op_type, struct flock *flock,
+afr_trace_inodelk_out (call_frame_t *frame, xlator_t *this,
+                       afr_lock_call_type_t lock_call_type,
+                       afr_lock_op_type_t lk_op_type, struct gf_flock *flock,
                        int op_ret, int op_errno, int32_t child_index)
 {
-        xlator_t            *this     = NULL;
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
-        afr_private_t       *priv     = NULL;
 
         char lockee[256];
         char lock_call_type_str[256];
         char verdict[16];
 
-        this     = THIS;
         local    = frame->local;
         int_lock = &local->internal_lock;
-        priv     = this->private;
-
-        if (!priv->inodelk_trace) {
-                return;
-        }
 
         afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
 
@@ -274,45 +308,37 @@ afr_trace_inodelk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
 
         afr_print_verdict (op_ret, op_errno, verdict);
 
-        gf_log (this->name, GF_LOG_NORMAL,
-                "[%s %s] [%s] Lockee={%s} Number={%llu}",
+        gf_log (this->name, GF_LOG_INFO,
+                "[%s %s] [%s] lk-owner=%s Lockee={%s} Number={%llu}",
                 lock_call_type_str,
                 lk_op_type == AFR_LOCK_OP ? "LOCK REPLY" : "UNLOCK REPLY",
-                verdict,
-                lockee,
+                verdict, lkowner_utoa (&frame->root->lk_owner), lockee,
                 (unsigned long long) int_lock->lock_number);
 
 }
 
 static void
-afr_trace_inodelk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
-                      afr_lock_op_type_t lk_op_type, struct flock *flock,
+afr_trace_inodelk_in (call_frame_t *frame, xlator_t *this,
+                      afr_lock_call_type_t lock_call_type,
+                      afr_lock_op_type_t lk_op_type, struct gf_flock *flock,
                       int32_t cmd, int32_t child_index)
 {
-        xlator_t            *this     = NULL;
         afr_local_t         *local    = NULL;
         afr_internal_lock_t *int_lock = NULL;
-        afr_private_t       *priv     = NULL;
 
         char lock[256];
         char lockee[256];
         char lock_call_type_str[256];
 
-        this     = THIS;
         local    = frame->local;
         int_lock = &local->internal_lock;
-        priv     = this->private;
 
-        if (!priv->inodelk_trace) {
-                return;
-        }
-
-        afr_print_inodelk (lock, 256, cmd, flock, frame->root->lk_owner);
+        afr_print_inodelk (lock, 256, cmd, flock, &frame->root->lk_owner);
         afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
 
         afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[%s %s] Lock={%s} Lockee={%s} Number={%llu}",
                 lock_call_type_str,
                 lk_op_type == AFR_LOCK_OP ? "LOCK REQUEST" : "UNLOCK REQUEST",
@@ -322,20 +348,21 @@ afr_trace_inodelk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
 }
 
 static void
-afr_trace_entrylk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
+afr_trace_entrylk_in (call_frame_t *frame, xlator_t *this,
+                      afr_lock_call_type_t lock_call_type,
                       afr_lock_op_type_t lk_op_type, const char *basename,
-                      int32_t child_index)
+                      int32_t cookie)
 {
-        xlator_t            *this     = NULL;
         afr_local_t         *local    = NULL;
         afr_internal_lock_t *int_lock = NULL;
         afr_private_t       *priv     = NULL;
+        int                 child_index = 0;
+        int                 lockee_no = 0;
 
         char lock[256];
         char lockee[256];
         char lock_call_type_str[256];
 
-        this     = THIS;
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
@@ -343,36 +370,41 @@ afr_trace_entrylk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
         if (!priv->entrylk_trace) {
                 return;
         }
+        lockee_no = cookie / priv->child_count;
+        child_index = cookie % priv->child_count;
 
-        afr_print_entrylk (lock, 256, basename, frame->root->lk_owner);
-        afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
+        afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
+        afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+                          child_index);
 
         afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
 
-        gf_log (this->name, GF_LOG_NORMAL,
-                "[%s %s] Lock={%s} Lockee={%s} Number={%llu}",
+        gf_log (this->name, GF_LOG_INFO,
+                "[%s %s] Lock={%s} Lockee={%s} Number={%llu}, Cookie={%d}",
                 lock_call_type_str,
                 lk_op_type == AFR_LOCK_OP ? "LOCK REQUEST" : "UNLOCK REQUEST",
                 lock, lockee,
-                (unsigned long long) int_lock->lock_number);
+                (unsigned long long) int_lock->lock_number,
+                cookie);
 }
 
 static void
-afr_trace_entrylk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
-                       afr_lock_op_type_t lk_op_type, const char *basename, int op_ret,
-                       int op_errno, int32_t child_index)
+afr_trace_entrylk_out (call_frame_t *frame, xlator_t *this,
+                       afr_lock_call_type_t lock_call_type,
+                       afr_lock_op_type_t lk_op_type, const char *basename,
+                       int op_ret, int op_errno, int32_t cookie)
 {
-        xlator_t            *this     = NULL;
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
         afr_private_t       *priv     = NULL;
+        int                 lockee_no = 0;
+        int                 child_index = 0;
 
         char lock[256];
         char lockee[256];
         char lock_call_type_str[256];
         char verdict[16];
 
-        this     = THIS;
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
@@ -380,20 +412,25 @@ afr_trace_entrylk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
         if (!priv->entrylk_trace) {
                 return;
         }
+        lockee_no = cookie / priv->child_count;
+        child_index = cookie % priv->child_count;
 
-        afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
+        afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
+        afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+                          child_index);
 
         afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
 
         afr_print_verdict (op_ret, op_errno, verdict);
 
-        gf_log (this->name, GF_LOG_NORMAL,
-                "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu}",
+        gf_log (this->name, GF_LOG_INFO,
+                "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu} Cookie={%d}",
                 lock_call_type_str,
                 lk_op_type == AFR_LOCK_OP ? "LOCK REPLY" : "UNLOCK REPLY",
                 verdict,
                 lock, lockee,
-                (unsigned long long) int_lock->lock_number);
+                (unsigned long long) int_lock->lock_number,
+                cookie);
 
 }
 
@@ -431,9 +468,8 @@ is_afr_lock_transaction (afr_local_t *local)
         int ret = 0;
 
         switch (local->transaction.type) {
-	case AFR_DATA_TRANSACTION:
-	case AFR_METADATA_TRANSACTION:
-	case AFR_FLUSH_TRANSACTION:
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
                 ret = 1;
                 break;
 
@@ -447,6 +483,47 @@ is_afr_lock_transaction (afr_local_t *local)
         return ret;
 }
 
+int
+afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
+                       loc_t *loc, char *basename, int child_count)
+{
+        int     ret     = -1;
+
+        loc_copy (&lockee->loc, loc);
+        lockee->basename        = (basename)? gf_strdup (basename): NULL;
+        if (basename && !lockee->basename)
+                goto out;
+
+        lockee->locked_count    = 0;
+        lockee->locked_nodes    = GF_CALLOC (child_count,
+                                             sizeof (*lockee->locked_nodes),
+                                             gf_afr_mt_afr_node_character);
+
+        if (!lockee->locked_nodes)
+                goto out;
+
+        ret = 0;
+out:
+        return ret;
+
+}
+
+void
+afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock)
+{
+        int     i   = 0;
+
+        for (i = 0; i < int_lock->lockee_count; i++) {
+                loc_wipe (&int_lock->lockee[i].loc);
+                if (int_lock->lockee[i].basename)
+                        GF_FREE (int_lock->lockee[i].basename);
+                if (int_lock->lockee[i].locked_nodes)
+                       GF_FREE (int_lock->lockee[i].locked_nodes);
+        }
+
+        return;
+}
+
 static int
 initialize_entrylk_variables (call_frame_t *frame, xlator_t *this)
 {
@@ -464,8 +541,13 @@ initialize_entrylk_variables (call_frame_t *frame, xlator_t *this)
         int_lock->lock_op_ret        = -1;
         int_lock->lock_op_errno      = 0;
 
-        for (i = 0; i < priv->child_count; i++) {
-                int_lock->entry_locked_nodes[i] = 0;
+        for (i = 0; i < AFR_LOCKEE_COUNT_MAX; i++) {
+                if (!int_lock->lockee[i].locked_nodes)
+                        break;
+                int_lock->lockee[i].locked_count = 0;
+                memset (int_lock->lockee[i].locked_nodes, 0,
+                        sizeof (*int_lock->lockee[i].locked_nodes) *
+                        priv->child_count);
         }
 
         return 0;
@@ -477,20 +559,23 @@ initialize_inodelk_variables (call_frame_t *frame, xlator_t *this)
         afr_local_t         *local    = NULL;
         afr_internal_lock_t *int_lock = NULL;
         afr_private_t       *priv     = NULL;
-
-        int i = 0;
+        afr_inodelk_t       *inodelk  = NULL;
 
         priv     = this->private;
         local    = frame->local;
         int_lock = &local->internal_lock;
 
-        int_lock->inodelk_lock_count = 0;
-        int_lock->lock_op_ret        = -1;
-        int_lock->lock_op_errno      = 0;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
 
-        for (i = 0; i < priv->child_count; i++) {
-                int_lock->inode_locked_nodes[i] = 0;
-        }
+        inodelk->lock_count    = 0;
+        int_lock->lk_attempted_count = 0;
+        int_lock->lock_op_ret   = -1;
+        int_lock->lock_op_errno = 0;
+
+        memset (inodelk->locked_nodes, 0,
+                sizeof (*inodelk->locked_nodes) * priv->child_count);
+        memset (int_lock->locked_nodes, 0,
+                sizeof (*int_lock->locked_nodes) * priv->child_count);
 
         return 0;
 }
@@ -498,24 +583,36 @@ initialize_inodelk_variables (call_frame_t *frame, xlator_t *this)
 loc_t *
 lower_path (loc_t *l1, const char *b1, loc_t *l2, const char *b2)
 {
-	int ret = 0;
+        int ret = 0;
+
+        ret = uuid_compare (l1->inode->gfid, l2->inode->gfid);
+
+        if (ret == 0)
+                ret = strcmp (b1, b2);
+
+        if (ret <= 0)
+                return l1;
+        else
+                return l2;
+}
 
-	ret = strcmp (l1->path, l2->path);
+int
+afr_lockee_locked_nodes_count (afr_internal_lock_t *int_lock)
+{
+        int call_count  = 0;
+        int i           = 0;
 
-	if (ret == 0)
-		ret = strcmp (b1, b2);
+        for (i = 0; i < int_lock->lockee_count; i++)
+                call_count += int_lock->lockee[i].locked_count;
 
-	if (ret <= 0)
-		return l1;
-	else
-		return l2;
+        return call_count;
 }
 
 int
 afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
 
 {
-        int i;
+        int i = 0;
         int call_count = 0;
 
         for (i = 0; i < child_count; i++) {
@@ -529,44 +626,63 @@ afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
 /* FIXME: What if UNLOCK fails */
 static int32_t
 afr_unlock_common_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno)
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_local_t         *local    = NULL;
+        afr_local_t         *local    = NULL;
         afr_internal_lock_t *int_lock = NULL;
-	int call_count = 0;
+        int call_count = 0;
 
-	local    = frame->local;
+        local    = frame->local;
         int_lock = &local->internal_lock;
 
-	LOCK (&frame->lock);
-	{
-		call_count = --int_lock->lk_call_count;
-	}
-	UNLOCK (&frame->lock);
+        LOCK (&frame->lock);
+        {
+                call_count = --int_lock->lk_call_count;
+        }
+        UNLOCK (&frame->lock);
 
-	if (call_count == 0) {
+        if (call_count == 0) {
                 gf_log (this->name, GF_LOG_TRACE,
                         "All internal locks unlocked");
                 int_lock->lock_cbk (frame, this);
         }
 
-	return 0;
+        return 0;
 }
 
 static int32_t
 afr_unlock_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno)
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        afr_trace_inodelk_out (frame, AFR_INODELK_TRANSACTION,
+        afr_local_t         *local = NULL;
+        afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk = NULL;
+        int32_t             child_index = (long)cookie;
+        afr_private_t       *priv = NULL;
+
+        local = frame->local;
+        int_lock = &local->internal_lock;
+
+        AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_TRANSACTION,
                                AFR_UNLOCK_OP, NULL, op_ret,
-                               op_errno, (long) cookie);
+                               op_errno, child_index);
+
+        priv = this->private;
 
         if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
-                gf_log (this->name, GF_LOG_TRACE,
-                        "Unlock failed for some reason");
+                gf_log (this->name, GF_LOG_INFO, "%s: unlock failed on subvolume %s "
+                        "with lock owner %s", local->loc.path,
+                        priv->children[child_index]->name,
+                        lkowner_utoa (&frame->root->lk_owner));
         }
 
-        afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno);
+
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+        inodelk->locked_nodes[child_index] &= LOCKED_NO;
+        if (local->transaction.eager_lock)
+                local->transaction.eager_lock[child_index] = 0;
+
+        afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, xdata);
 
         return 0;
 
@@ -576,22 +692,30 @@ static int
 afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
         afr_local_t         *local    = NULL;
         afr_private_t       *priv     = NULL;
-
-        struct flock flock;
+        struct gf_flock flock = {0,};
+        struct gf_flock full_flock = {0,};
+        struct gf_flock *flock_use = NULL;
         int call_count = 0;
         int i = 0;
+        int piggyback = 0;
+        afr_fd_ctx_t        *fd_ctx      = NULL;
+
 
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
 
-        flock.l_start = int_lock->lk_flock.l_start;
-        flock.l_len   = int_lock->lk_flock.l_len;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+
+        flock.l_start = inodelk->flock.l_start;
+        flock.l_len   = inodelk->flock.l_len;
         flock.l_type  = F_UNLCK;
 
-        call_count = afr_locked_nodes_count (int_lock->inode_locked_nodes,
+        full_flock.l_type = F_UNLCK;
+        call_count = afr_locked_nodes_count (inodelk->locked_nodes,
                                              priv->child_count);
 
         int_lock->lk_call_count = call_count;
@@ -603,55 +727,107 @@ afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
                 goto out;
         }
 
+        if (local->fd)
+                fd_ctx = afr_fd_ctx_get (local->fd, this);
+
         for (i = 0; i < priv->child_count; i++) {
-                if (int_lock->inode_locked_nodes[i] & LOCKED_YES) {
-                        if (local->fd) {
-                                afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
-                                                      AFR_UNLOCK_OP, &flock, F_SETLK, i);
-
-                                STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
-                                                   (void *) (long)i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->finodelk,
-                                                   this->name, local->fd,
-                                                   F_SETLK, &flock);
+                if ((inodelk->locked_nodes[i] & LOCKED_YES) != LOCKED_YES)
+                        continue;
 
-                                if (!--call_count)
-                                        break;
+                if (local->fd) {
+                        flock_use = &flock;
+                        if (!local->transaction.eager_lock[i]) {
+                                goto wind;
+                        }
 
-                        } else {
-                                afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
-                                                      AFR_UNLOCK_OP, &flock, F_SETLK, i);
+                        piggyback = 0;
 
-                                STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
-                                                   (void *) (long)i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->inodelk,
-                                                   this->name, &local->loc,
-                                                   F_SETLK, &flock);
+                        LOCK (&local->fd->lock);
+                        {
+                                if (fd_ctx->lock_piggyback[i]) {
+                                        fd_ctx->lock_piggyback[i]--;
+                                        piggyback = 1;
+                                } else {
+                                        fd_ctx->lock_acquired[i]--;
+                                }
+                        }
+                        UNLOCK (&local->fd->lock);
 
+                        if (piggyback) {
+                                afr_unlock_inodelk_cbk (frame, (void *) (long) i,
+                                                        this, 1, 0, NULL);
                                 if (!--call_count)
                                         break;
-
+                                continue;
                         }
 
-                }
+                        flock_use = &full_flock;
+                wind:
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_TRANSACTION,
+                                              AFR_UNLOCK_OP, flock_use, F_SETLK,
+                                              i);
 
-        }
+                        STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
+                                           (void *) (long)i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->finodelk,
+                                           int_lock->domain, local->fd,
+                                           F_SETLK, flock_use, NULL);
+
+                        if (!--call_count)
+                                break;
+
+                } else {
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_TRANSACTION,
+                                              AFR_UNLOCK_OP, &flock, F_SETLK, i);
+
+                        STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
+                                           (void *) (long)i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->inodelk,
+                                           int_lock->domain, &local->loc,
+                                           F_SETLK, &flock, NULL);
 
+                        if (!--call_count)
+                                break;
+                }
+        }
 out:
         return 0;
 }
 
 static int32_t
 afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno)
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
-                              AFR_UNLOCK_OP, NULL, op_ret,
-                               op_errno, (long) cookie);
+        afr_local_t         *local = NULL;
+        afr_private_t       *priv  = NULL;
+        afr_internal_lock_t *int_lock = NULL;
+        int32_t             child_index = 0;
+        int                 lockee_no   = 0;
+
+        priv = this->private;
+        lockee_no = (int)((long) cookie) / priv->child_count;
+        child_index = (int) ((long) cookie) % priv->child_count;
+
+        local = frame->local;
+        int_lock = &local->internal_lock;
 
-        afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno);
+        AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
+                               AFR_UNLOCK_OP,
+                               int_lock->lockee[lockee_no].basename, op_ret,
+                               op_errno, (int) ((long)cookie));
+
+        if (op_ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: unlock failed on %d, reason: %s",
+                        local->loc.path, child_index, strerror (op_errno));
+        }
+
+        int_lock->lockee[lockee_no].locked_nodes[child_index] &= LOCKED_NO;
+        afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, NULL);
 
         return 0;
 }
@@ -659,25 +835,22 @@ afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 static int
 afr_unlock_entrylk (call_frame_t *frame, xlator_t *this)
 {
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-        afr_private_t       *priv     = NULL;
-        const char          *basename = NULL;
-        loc_t               *loc      = NULL;
-
-        int call_count = 0;
-        int i = -1;
+        afr_internal_lock_t     *int_lock       = NULL;
+        afr_local_t             *local          = NULL;
+        afr_private_t           *priv           = NULL;
+        int                     call_count      = 0;
+        int                     index           = 0;
+        int                     lockee_no       = 0;
+        int                     copies          = 0;
+        int                     i               = -1;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
+        copies   = priv->child_count;
 
-        basename = int_lock->lk_basename;
-        if (int_lock->lk_loc)
-                loc = int_lock->lk_loc;
+        call_count = afr_lockee_locked_nodes_count (int_lock);
 
-        call_count = afr_locked_nodes_count (int_lock->entry_locked_nodes,
-                                             priv->child_count);
         int_lock->lk_call_count = call_count;
 
         if (!call_count){
@@ -687,18 +860,23 @@ afr_unlock_entrylk (call_frame_t *frame, xlator_t *this)
                 goto out;
         }
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (int_lock->entry_locked_nodes[i] & LOCKED_YES) {
-                        afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
-                                              AFR_UNLOCK_OP, basename, i);
+        for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
+                lockee_no = i / copies;
+                index     = i % copies;
+                if (int_lock->lockee[lockee_no].locked_nodes[index] & LOCKED_YES) {
+                        AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+                                              AFR_UNLOCK_OP,
+                                              int_lock->lockee[lockee_no].basename,
+                                              i);
 
                         STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
                                            (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->entrylk,
-                                           this->name,
-                                           loc, basename,
-                                           ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
+                                           priv->children[index],
+                                           priv->children[index]->fops->entrylk,
+                                           int_lock->domain,
+                                           &int_lock->lockee[lockee_no].loc,
+                                           int_lock->lockee[lockee_no].basename,
+                                           ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL);
 
                         if (!--call_count)
                                 break;
@@ -712,155 +890,85 @@ out:
 
 static int32_t
 afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int32_t op_ret, int32_t op_errno)
+              int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        afr_internal_lock_t *int_lock = NULL;
-	afr_local_t         *local    = NULL;
-	afr_private_t       *priv     = NULL;
-
-	int done = 0;
-	int child_index = (long) cookie;
-
-	local    = frame->local;
-        int_lock = &local->internal_lock;
-	priv     = this->private;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			if (op_errno == ENOSYS) {
-				/* return ENOTSUP */
-				gf_log (this->name, GF_LOG_ERROR,
-					"subvolume does not support locking. "
-					"please load features/posix-locks xlator on server");
-				local->op_ret = op_ret;
-                                int_lock->lock_op_ret = op_ret;
-				done = 1;
-			}
-
-			local->child_up[child_index] = 0;
-			local->op_errno              = op_errno;
-                        int_lock->lock_op_errno      = op_errno;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-        if ((op_ret == -1) &&
-            (op_errno == ENOSYS)) {
-                afr_unlock (frame, this);
-        } else {
-                if (op_ret == 0) {
-                        int_lock->locked_nodes[child_index]
-                                |= LOCKED_YES;
-                        int_lock->lock_count++;
-                }
-                afr_lock_blocking (frame, this, child_index + 1);
-        }
-
-	return 0;
-}
-
-static int32_t
-afr_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                          int32_t op_ret, int32_t op_errno)
-{
-        afr_trace_inodelk_out (frame, AFR_INODELK_TRANSACTION,
-                               AFR_LOCK_OP, NULL, op_ret,
-                               op_errno, (long) cookie);
-
-        afr_lock_cbk (frame, cookie, this, op_ret, op_errno);
-        return 0;
-
-}
-
-static int32_t
-afr_lock_lower_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno)
-{
-        afr_internal_lock_t *int_lock    = NULL;
-        afr_private_t       *priv        = NULL;
-        afr_local_t         *local       = NULL;
-        loc_t               *lower       = NULL;
-        loc_t               *higher      = NULL;
-        const char          *lower_name  = NULL;
-        const char          *higher_name = NULL;
-
-        int child_index = (long) cookie;
+        afr_internal_lock_t     *int_lock       = NULL;
+        afr_local_t             *local          = NULL;
+        afr_private_t           *priv           = NULL;
+        int                     cky             = (long) cookie;
+        int                     child_index     = 0;
+        int                     lockee_no       = 0;
 
         priv     = this->private;
         local    = frame->local;
         int_lock = &local->internal_lock;
 
+        child_index = ((int)cky) % priv->child_count;
+        lockee_no   = ((int)cky) / priv->child_count;
+
         LOCK (&frame->lock);
         {
                 if (op_ret == -1) {
                         if (op_errno == ENOSYS) {
                                 /* return ENOTSUP */
-
                                 gf_log (this->name, GF_LOG_ERROR,
                                         "subvolume does not support locking. "
-                                        "please load features/posix-locks xlator on server");
-
-                                local->op_ret   = op_ret;
+                                        "please load features/locks xlator on server");
+                                local->op_ret = op_ret;
+                                int_lock->lock_op_ret = op_ret;
                         }
 
-                        local->child_up[child_index] = 0;
-                        local->op_errno = op_errno;
+                        local->op_errno              = op_errno;
+                        int_lock->lock_op_errno      = op_errno;
                 }
+
+		int_lock->lk_attempted_count++;
         }
         UNLOCK (&frame->lock);
 
-        if (op_ret != 0) {
+        if ((op_ret == -1) &&
+            (op_errno == ENOSYS)) {
                 afr_unlock (frame, this);
-                goto out;
         } else {
-                int_lock->lower_locked_nodes[child_index] |= LOCKED_LOWER;
-                int_lock->lock_count++;
+                if (op_ret == 0) {
+                        if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
+                            local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
+                                int_lock->lockee[lockee_no].locked_nodes[child_index] |= LOCKED_YES;
+                                int_lock->lockee[lockee_no].locked_count++;
+                                int_lock->entrylk_lock_count++;
+                        } else {
+                                int_lock->locked_nodes[child_index] |= LOCKED_YES;
+                                int_lock->lock_count++;
+                        }
+                }
+                afr_lock_blocking (frame, this, cky + 1);
         }
 
-        /* The lower path has been locked. Now lock the higher path */
-
-        lower = lower_path (&local->transaction.parent_loc,
-                            local->transaction.basename,
-                            &local->transaction.new_parent_loc,
-                            local->transaction.new_basename);
-
-        lower_name = (lower == &local->transaction.parent_loc ?
-                      local->transaction.basename :
-                      local->transaction.new_basename);
-
-        higher = (lower == &local->transaction.parent_loc ?
-                  &local->transaction.new_parent_loc :
-                  &local->transaction.parent_loc);
-
-        higher_name = (higher == &local->transaction.parent_loc ?
-                       local->transaction.basename :
-                       local->transaction.new_basename);
-
-        afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
-                              AFR_LOCK_OP, higher_name, child_index);
-
+        return 0;
+}
 
-        STACK_WIND_COOKIE (frame, afr_lock_cbk,
-                           (void *) (long) child_index,
-                           priv->children[child_index],
-                           priv->children[child_index]->fops->entrylk,
-                           this->name, higher, higher_name,
-                           ENTRYLK_LOCK, ENTRYLK_WRLCK);
+static int32_t
+afr_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_TRANSACTION,
+                               AFR_LOCK_OP, NULL, op_ret,
+                               op_errno, (long) cookie);
 
-out:
+        afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
         return 0;
+
 }
 
 static int32_t
 afr_blocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                          int32_t op_ret, int32_t op_errno)
+                          int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
+        AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
                                AFR_LOCK_OP, NULL, op_ret,
                                op_errno, (long)cookie);
 
-        afr_lock_cbk (frame, cookie, this, op_ret, op_errno);
+        afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
         return 0;
 }
 
@@ -868,6 +976,7 @@ static int
 afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
         afr_local_t         *local    = NULL;
         afr_private_t       *priv     = NULL;
 
@@ -876,21 +985,18 @@ afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
 
         switch (local->transaction.type) {
-	case AFR_DATA_TRANSACTION:
-	case AFR_METADATA_TRANSACTION:
-	case AFR_FLUSH_TRANSACTION:
-                memcpy (int_lock->inode_locked_nodes,
-                        int_lock->locked_nodes,
-                        priv->child_count);
-                int_lock->inodelk_lock_count = int_lock->lock_count;
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
+                inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+                memcpy (inodelk->locked_nodes, int_lock->locked_nodes,
+                        sizeof (*inodelk->locked_nodes) * priv->child_count);
+                inodelk->lock_count = int_lock->lock_count;
                 break;
 
         case AFR_ENTRY_RENAME_TRANSACTION:
         case AFR_ENTRY_TRANSACTION:
-                memcpy (int_lock->entry_locked_nodes,
-                        int_lock->locked_nodes,
-                        priv->child_count);
-                int_lock->entrylk_lock_count = int_lock->lock_count;
+                /*entrylk_count is being used in both non-blocking and blocking
+                 * modes */
                 break;
         }
 
@@ -898,42 +1004,78 @@ afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
 
 }
 
+static inline gf_boolean_t
+afr_is_entrylk (afr_internal_lock_t *int_lock,
+                afr_transaction_type trans_type)
+{
+        gf_boolean_t is_entrylk = _gf_false;
+
+        if ((int_lock->transaction_lk_type == AFR_SELFHEAL_LK) &&
+            int_lock->selfheal_lk_type == AFR_ENTRY_SELF_HEAL_LK) {
+
+                is_entrylk = _gf_true;
+
+        } else if ((int_lock->transaction_lk_type == AFR_TRANSACTION_LK) &&
+                 (trans_type == AFR_ENTRY_TRANSACTION ||
+                  trans_type == AFR_ENTRY_RENAME_TRANSACTION)) {
+
+                is_entrylk = _gf_true;
+
+        } else {
+                is_entrylk = _gf_false;
+        }
+
+        return is_entrylk;
+}
+
+static gf_boolean_t
+_is_lock_wind_needed (afr_local_t *local, int child_index)
+{
+        if (!local->child_up[child_index])
+                return _gf_false;
+
+        return _gf_true;
+}
+
 int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
+afr_lock_blocking (call_frame_t *frame, xlator_t *this, int cookie)
 {
         afr_internal_lock_t *int_lock    = NULL;
+        afr_inodelk_t       *inodelk     = NULL;
         afr_local_t         *local       = NULL;
-	afr_private_t       *priv        = NULL;
-        afr_fd_ctx_t        *fd_ctx      = NULL;
-	loc_t               *lower       = NULL;
-	loc_t               *higher      = NULL;
-	const char          *lower_name  = NULL;
-	const char          *higher_name = NULL;
-
-	struct flock flock;
-        uint64_t ctx;
+        afr_private_t       *priv        = NULL;
+        struct gf_flock flock = {0,};
+        uint64_t ctx = 0;
         int ret = 0;
-
-	local    = frame->local;
-        int_lock = &local->internal_lock;
-	priv     = this->private;
-
-	flock.l_start = int_lock->lk_flock.l_start;
-	flock.l_len   = int_lock->lk_flock.l_len;
-	flock.l_type  = int_lock->lk_flock.l_type;
+        int child_index = 0;
+        int lockee_no   = 0;
+        gf_boolean_t is_entrylk = _gf_false;
+
+        local         = frame->local;
+        int_lock      = &local->internal_lock;
+        priv          = this->private;
+        child_index   = cookie % priv->child_count;
+        lockee_no     = cookie / priv->child_count;
+        is_entrylk    = afr_is_entrylk (int_lock, local->transaction.type);
+
+
+        if (!is_entrylk) {
+                inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+                flock.l_start = inodelk->flock.l_start;
+                flock.l_len   = inodelk->flock.l_len;
+                flock.l_type  = inodelk->flock.l_type;
+        }
 
         if (local->fd) {
                 ret = fd_ctx_get (local->fd, this, &ctx);
 
                 if (ret < 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+                        gf_log (this->name, GF_LOG_INFO,
                                 "unable to get fd ctx for fd=%p",
                                 local->fd);
 
                         local->op_ret           = -1;
                         int_lock->lock_op_ret   = -1;
-                        local->op_errno         = EINVAL;
-                        int_lock->lock_op_errno = EINVAL;
 
                         afr_copy_locked_nodes (frame, this);
 
@@ -941,49 +1083,27 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
 
                         return 0;
                 }
-
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                /* skip over children that or down
-                   or don't have the fd open */
-
-                while ((child_index < priv->child_count)
-                       && (!local->child_up[child_index]
-                           || !fd_ctx->opened_on[child_index]))
-
-                        child_index++;
-        } else {
-                /* skip over children that are down */
-                while ((child_index < priv->child_count)
-                       && !local->child_up[child_index])
-                        child_index++;
         }
 
-	if ((child_index == priv->child_count) &&
-	    int_lock->lock_count == 0) {
-
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unable to lock on even one child");
+        if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
+                if ((is_entrylk && int_lock->entrylk_lock_count == 0) ||
+                    (!is_entrylk && int_lock->lock_count == 0)) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "unable to lock on even one child");
 
-		local->op_ret           = -1;
-                int_lock->lock_op_ret   = -1;
-		local->op_errno         = EAGAIN;
-                int_lock->lock_op_errno = EAGAIN;
-
-                afr_copy_locked_nodes (frame, this);
-
-                afr_unlock(frame, this);
+                        local->op_ret           = -1;
+                        int_lock->lock_op_ret   = -1;
 
-		return 0;
+                        afr_copy_locked_nodes (frame, this);
 
-	}
+                        afr_unlock(frame, this);
 
-	if ((child_index == priv->child_count)
-	    || (int_lock->lock_count ==
-		afr_up_children_count (priv->child_count,
-                                       local->child_up))) {
+                        return 0;
+                }
+        }
 
-		/* we're done locking */
+        if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
+                /* we're done locking */
 
                 gf_log (this->name, GF_LOG_DEBUG,
                         "we're done locking");
@@ -992,108 +1112,85 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
 
                 int_lock->lock_op_ret = 0;
                 int_lock->lock_cbk (frame, this);
-		return 0;
-	}
+                return 0;
+        }
 
-	switch (local->transaction.type) {
-	case AFR_DATA_TRANSACTION:
-	case AFR_METADATA_TRANSACTION:
-	case AFR_FLUSH_TRANSACTION:
+        if (!_is_lock_wind_needed (local, child_index)) {
+                afr_lock_blocking (frame, this, cookie + 1);
+                return 0;
+        }
 
-		if (local->fd) {
-                        afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
+        switch (local->transaction.type) {
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
+
+                if (local->fd) {
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_TRANSACTION,
                                               AFR_LOCK_OP, &flock, F_SETLKW,
                                               child_index);
 
-			STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
-					   (void *) (long) child_index,
-					   priv->children[child_index],
-					   priv->children[child_index]->fops->finodelk,
-					   this->name, local->fd,
-                                           F_SETLKW, &flock);
+                        STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
+                                           (void *) (long) child_index,
+                                           priv->children[child_index],
+                                           priv->children[child_index]->fops->finodelk,
+                                           int_lock->domain, local->fd,
+                                           F_SETLKW, &flock, NULL);
 
-		} else {
-                        afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
+                } else {
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_TRANSACTION,
                                               AFR_LOCK_OP, &flock, F_SETLKW,
                                               child_index);
 
-			STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
-					   (void *) (long) child_index,
-					   priv->children[child_index],
-					   priv->children[child_index]->fops->inodelk,
-					   this->name, &local->loc,
-                                           F_SETLKW, &flock);
-		}
-
-		break;
-
-	case AFR_ENTRY_RENAME_TRANSACTION:
-	{
-		lower = lower_path (&local->transaction.parent_loc,
-				    local->transaction.basename,
-				    &local->transaction.new_parent_loc,
-				    local->transaction.new_basename);
-
-		lower_name = (lower == &local->transaction.parent_loc ?
-			      local->transaction.basename :
-			      local->transaction.new_basename);
-
-		higher = (lower == &local->transaction.parent_loc ?
-			  &local->transaction.new_parent_loc :
-			  &local->transaction.parent_loc);
-
-		higher_name = (higher == &local->transaction.parent_loc ?
-			       local->transaction.basename :
-			       local->transaction.new_basename);
-
-                afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
-                                      AFR_LOCK_OP, lower_name, child_index);
-
-
-		STACK_WIND_COOKIE (frame, afr_lock_lower_cbk,
-				   (void *) (long) child_index,
-				   priv->children[child_index],
-				   priv->children[child_index]->fops->entrylk,
-				   this->name, lower, lower_name,
-				   ENTRYLK_LOCK, ENTRYLK_WRLCK);
-
-		break;
-	}
+                        STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
+                                           (void *) (long) child_index,
+                                           priv->children[child_index],
+                                           priv->children[child_index]->fops->inodelk,
+                                           int_lock->domain, &local->loc,
+                                           F_SETLKW, &flock, NULL);
+                }
 
-	case AFR_ENTRY_TRANSACTION:
-		if (local->fd) {
-                        afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
-                                              AFR_LOCK_OP, local->transaction.basename,
-                                              child_index);
+                break;
 
-			STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
-					   (void *) (long) child_index,
-					   priv->children[child_index],
-					   priv->children[child_index]->fops->fentrylk,
-					   this->name, local->fd,
-					   local->transaction.basename,
-					   ENTRYLK_LOCK, ENTRYLK_WRLCK);
-		} else {
-                        afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
+        case AFR_ENTRY_RENAME_TRANSACTION:
+        case AFR_ENTRY_TRANSACTION:
+                /*Accounting for child_index increments on 'down'
+                 *and 'fd-less' children */
+
+                if (local->fd) {
+                        AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_TRANSACTION,
+                                              AFR_LOCK_OP,
+                                              int_lock->lockee[lockee_no].basename,
+                                              cookie);
+
+                        STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
+                                           (void *) (long) cookie,
+                                           priv->children[child_index],
+                                           priv->children[child_index]->fops->fentrylk,
+                                           int_lock->domain, local->fd,
+                                           int_lock->lockee[lockee_no].basename,
+                                           ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+                } else {
+                        AFR_TRACE_ENTRYLK_IN (frame, this,
+                                              AFR_ENTRYLK_TRANSACTION,
                                               AFR_LOCK_OP, local->transaction.basename,
                                               child_index);
 
-			STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
-					   (void *) (long) child_index,
-					   priv->children[child_index],
-					   priv->children[child_index]->fops->entrylk,
-					   this->name,
-                                           &local->transaction.parent_loc,
-					   local->transaction.basename,
-					   ENTRYLK_LOCK, ENTRYLK_WRLCK);
-		}
-
-		break;
-	}
-
-	return 0;
+                        STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
+                                           (void *) (long) cookie,
+                                           priv->children[child_index],
+                                           priv->children[child_index]->fops->entrylk,
+                                           int_lock->domain,
+                                           &int_lock->lockee[lockee_no].loc,
+                                           int_lock->lockee[lockee_no].basename,
+                                           ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+                }
 
+                break;
+        }
 
+        return 0;
 }
 
 int32_t
@@ -1102,20 +1199,25 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this)
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
         afr_private_t       *priv     = NULL;
+        int                  up_count = 0;
 
         priv     = this->private;
         local    = frame->local;
         int_lock = &local->internal_lock;
 
         switch (local->transaction.type) {
-	case AFR_DATA_TRANSACTION:
-	case AFR_METADATA_TRANSACTION:
-	case AFR_FLUSH_TRANSACTION:
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
                 initialize_inodelk_variables (frame, this);
                 break;
 
         case AFR_ENTRY_RENAME_TRANSACTION:
         case AFR_ENTRY_TRANSACTION:
+                up_count = afr_up_children_count (local->child_up,
+                                                  priv->child_count);
+                int_lock->lk_call_count = int_lock->lk_expected_count
+                                        = (int_lock->lockee_count *
+                                           up_count);
                 initialize_entrylk_variables (frame, this);
                 break;
         }
@@ -1127,60 +1229,68 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this)
 
 static int32_t
 afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                             int32_t op_ret, int32_t op_errno)
+                             int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         afr_internal_lock_t *int_lock = NULL;
-	afr_local_t         *local    = NULL;
-	afr_private_t       *priv     = NULL;
-
+        afr_local_t         *local    = NULL;
         int call_count          = 0;
-	int child_index         = (long) cookie;
+        int child_index         = (long) cookie;
+        int copies              = 0;
+        int index               = 0;
+        int lockee_no           = 0;
+        afr_private_t       *priv = NULL;
+
+        priv = this->private;
 
-	local    = frame->local;
+        copies = priv->child_count;
+        index = child_index % copies;
+        lockee_no = child_index / copies;
+
+        local    = frame->local;
         int_lock = &local->internal_lock;
-	priv     = this->private;
 
-        afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
-                               AFR_LOCK_OP, NULL, op_ret,
+        AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
+                               AFR_LOCK_OP,
+                               int_lock->lockee[lockee_no].basename, op_ret,
                                op_errno, (long) cookie);
 
-        LOCK (&frame->lock);
-        {
-                call_count = --int_lock->lk_call_count;
-        }
-        UNLOCK (&frame->lock);
-
-        if (op_ret < 0 ) {
+	LOCK (&frame->lock);
+	{
+		if (op_ret < 0 ) {
 			if (op_errno == ENOSYS) {
-				/* return ENOTSUP */
+                        /* return ENOTSUP */
 				gf_log (this->name, GF_LOG_ERROR,
 					"subvolume does not support locking. "
-					"please load features/posix-locks xlator on server");
+					"please load features/locks xlator on server");
 				local->op_ret         = op_ret;
-                                int_lock->lock_op_ret = op_ret;
+				int_lock->lock_op_ret = op_ret;
 
-                                local->child_up[child_index] = 0;
-                                int_lock->lock_op_errno      = op_errno;
-                                local->op_errno              = op_errno;
-                        }
-        } else if (op_ret == 0) {
-                        int_lock->entry_locked_nodes[child_index]
-                                |= LOCKED_YES;
-                        int_lock->entrylk_lock_count++;
+				int_lock->lock_op_errno      = op_errno;
+				local->op_errno              = op_errno;
+			}
+		} else if (op_ret == 0) {
+			int_lock->lockee[lockee_no].locked_nodes[index] |= \
+				LOCKED_YES;
+			int_lock->lockee[lockee_no].locked_count++;
+			int_lock->entrylk_lock_count++;
+		}
+
+                call_count = --int_lock->lk_call_count;
         }
+        UNLOCK (&frame->lock);
 
         if (call_count == 0) {
                 gf_log (this->name, GF_LOG_TRACE,
                         "Last locking reply received");
-                /* all locks successfull. Proceed to call FOP */
+                /* all locks successful. Proceed to call FOP */
                 if (int_lock->entrylk_lock_count ==
-                    afr_up_children_count (priv->child_count, local->child_up)) {
+                                int_lock->lk_expected_count) {
                         gf_log (this->name, GF_LOG_TRACE,
                                 "All servers locked. Calling the cbk");
                         int_lock->lock_op_ret = 0;
                         int_lock->lock_cbk (frame, this);
                 }
-                /* Not all locks were successfull. Unlock and try locking
+                /* Not all locks were successful. Unlock and try locking
                    again, this time with serially blocking locks */
                 else {
                         gf_log (this->name, GF_LOG_TRACE,
@@ -1197,33 +1307,27 @@ afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 int
 afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
 {
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-        afr_private_t       *priv     = NULL;
-        afr_fd_ctx_t        *fd_ctx   = NULL;
-        const char          *basename = NULL;
-        loc_t               *loc      = NULL;
-
-        int32_t call_count = 0;
+        afr_internal_lock_t *int_lock   = NULL;
+        afr_local_t         *local      = NULL;
+        afr_private_t       *priv       = NULL;
+        afr_fd_ctx_t        *fd_ctx     = NULL;
+        int                 copies      = 0;
+        int                 index       = 0;
+        int                 lockee_no   = 0;
+        int32_t             call_count  = 0;
         int i = 0;
-        uint64_t  ctx;
-        int ret = 0;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
 
+        copies = priv->child_count;
         initialize_entrylk_variables (frame, this);
 
-        basename = int_lock->lk_basename;
-        if (int_lock->lk_loc)
-                loc = int_lock->lk_loc;
-
         if (local->fd) {
-                ret = fd_ctx_get (local->fd, this, &ctx);
-
-                if (ret < 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+                fd_ctx = afr_fd_ctx_get (local->fd, this);
+                if (!fd_ctx) {
+                        gf_log (this->name, GF_LOG_INFO,
                                 "unable to get fd ctx for fd=%p",
                                 local->fd);
 
@@ -1232,113 +1336,147 @@ afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
                         local->op_errno         = EINVAL;
                         int_lock->lock_op_errno = EINVAL;
 
+			afr_unlock (frame, this);
                         return -1;
                 }
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                call_count = internal_lock_count (frame, this, fd_ctx);
+                call_count = int_lock->lockee_count * internal_lock_count (frame, this);
                 int_lock->lk_call_count = call_count;
+                int_lock->lk_expected_count = call_count;
+
+                if (!call_count) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "fd not open on any subvolumes. aborting.");
+                        afr_unlock (frame, this);
+                        goto out;
+                }
 
                 /* Send non-blocking entrylk calls only on up children
                    and where the fd has been opened */
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i] && fd_ctx->opened_on[i]) {
-                                afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
-                                                      AFR_LOCK_OP, basename, i);
+                for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
+                        index = i%copies;
+                        lockee_no = i/copies;
+                        if (local->child_up[index]) {
+                                AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+                                                      AFR_LOCK_OP,
+                                                      int_lock->lockee[lockee_no].basename,
+                                                      i);
 
                                 STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
                                                    (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->fentrylk,
+                                                   priv->children[index],
+                                                   priv->children[index]->fops->fentrylk,
                                                    this->name, local->fd,
-                                                   basename,
-                                                   ENTRYLK_LOCK_NB, ENTRYLK_WRLCK);
+                                                   int_lock->lockee[lockee_no].basename,
+                                                   ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
+                                                   NULL);
+                                if (!--call_count)
+                                        break;
                         }
                 }
         } else {
-                GF_ASSERT (loc);
-
-                call_count = internal_lock_count (frame, this, NULL);
+                call_count = int_lock->lockee_count * internal_lock_count (frame, this);
                 int_lock->lk_call_count = call_count;
+                int_lock->lk_expected_count = call_count;
 
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i]) {
-                                afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
-                                                      AFR_LOCK_OP, basename, i);
+                for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
+                        index = i%copies;
+                        lockee_no = i/copies;
+                        if (local->child_up[index]) {
+                                AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+                                                      AFR_LOCK_OP,
+                                                      int_lock->lockee[lockee_no].basename,
+                                                      i);
 
                                 STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
                                                    (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->entrylk,
-                                                   this->name, loc, basename,
-                                                   ENTRYLK_LOCK, ENTRYLK_WRLCK);
+                                                   priv->children[index],
+                                                   priv->children[index]->fops->entrylk,
+                                                   this->name, &int_lock->lockee[lockee_no].loc,
+                                                   int_lock->lockee[lockee_no].basename,
+                                                   ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
+                                                   NULL);
 
                                 if (!--call_count)
                                         break;
-
                         }
                 }
         }
-
+out:
         return 0;
 }
 
 int32_t
 afr_nonblocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                             int32_t op_ret, int32_t op_errno)
+                             int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         afr_internal_lock_t *int_lock = NULL;
-	afr_local_t         *local    = NULL;
-	afr_private_t       *priv     = NULL;
-
+        afr_inodelk_t       *inodelk  = NULL;
+        afr_local_t         *local    = NULL;
         int call_count  = 0;
-	int child_index = (long) cookie;
+        int child_index = (long) cookie;
+        afr_fd_ctx_t        *fd_ctx = NULL;
+
 
-	local    = frame->local;
+        local    = frame->local;
         int_lock = &local->internal_lock;
-	priv     = this->private;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
 
-        afr_trace_inodelk_out (frame, AFR_INODELK_NB_TRANSACTION,
+        AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_NB_TRANSACTION,
                                AFR_LOCK_OP, NULL, op_ret,
                                op_errno, (long) cookie);
 
+	if (local->fd)
+		fd_ctx = afr_fd_ctx_get (local->fd, this);
+
         LOCK (&frame->lock);
         {
-                call_count = --int_lock->lk_call_count;
-        }
-        UNLOCK (&frame->lock);
-
-        if (op_ret < 0 ) {
+		if (op_ret < 0) {
 			if (op_errno == ENOSYS) {
 				/* return ENOTSUP */
 				gf_log (this->name, GF_LOG_ERROR,
 					"subvolume does not support locking. "
-					"please load features/posix-locks xlator on server");
+					"please load features/locks xlator on "
+					"server");
 				local->op_ret                = op_ret;
-                                int_lock->lock_op_ret        = op_ret;
-                                local->child_up[child_index] = 0;
-                                int_lock->lock_op_errno      = op_errno;
-                                local->op_errno              = op_errno;
-                        }
-        } else if (op_ret == 0) {
-                        int_lock->inode_locked_nodes[child_index]
-                                |= LOCKED_YES;
-                        int_lock->inodelk_lock_count++;
+				int_lock->lock_op_ret        = op_ret;
+				int_lock->lock_op_errno      = op_errno;
+				local->op_errno              = op_errno;
+			}
+			if (local->transaction.eager_lock)
+				local->transaction.eager_lock[child_index] = 0;
+		} else {
+			inodelk->locked_nodes[child_index] |= LOCKED_YES;
+			inodelk->lock_count++;
+
+			if (local->transaction.eager_lock &&
+			    local->transaction.eager_lock[child_index] &&
+			    local->fd) {
+				/* piggybacked */
+				if (op_ret == 1) {
+					/* piggybacked */
+				} else if (op_ret == 0) {
+					/* lock acquired from server */
+                                        fd_ctx->lock_acquired[child_index]++;
+				}
+			}
+		}
+
+                call_count = --int_lock->lk_call_count;
         }
+        UNLOCK (&frame->lock);
 
         if (call_count == 0) {
                 gf_log (this->name, GF_LOG_TRACE,
                         "Last inode locking reply received");
-                /* all locks successfull. Proceed to call FOP */
-                if (int_lock->inodelk_lock_count ==
-                    afr_up_children_count (priv->child_count, local->child_up)) {
+                /* all locks successful. Proceed to call FOP */
+                if (inodelk->lock_count == int_lock->lk_expected_count) {
                         gf_log (this->name, GF_LOG_TRACE,
                                 "All servers locked. Calling the cbk");
                         int_lock->lock_op_ret = 0;
                         int_lock->lock_cbk (frame, this);
                 }
-                /* Not all locks were successfull. Unlock and try locking
+                /* Not all locks were successful. Unlock and try locking
                    again, this time with serially blocking locks */
                 else {
                         gf_log (this->name, GF_LOG_TRACE,
@@ -1356,31 +1494,36 @@ int
 afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
         afr_local_t         *local    = NULL;
         afr_private_t       *priv     = NULL;
         afr_fd_ctx_t        *fd_ctx   = NULL;
-
-        int32_t  call_count = 0;
-        uint64_t ctx        = 0;
-        int      i          = 0;
-        int      ret        = 0;
-        struct flock flock;
+        int32_t             call_count = 0;
+        int                 i          = 0;
+        int                 ret        = 0;
+        struct              gf_flock flock = {0,};
+        struct              gf_flock full_flock = {0,};
+        struct              gf_flock *flock_use = NULL;
+        int                 piggyback = 0;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
 
-	flock.l_start = int_lock->lk_flock.l_start;
-	flock.l_len   = int_lock->lk_flock.l_len;
-	flock.l_type  = int_lock->lk_flock.l_type;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+
+        flock.l_start = inodelk->flock.l_start;
+        flock.l_len   = inodelk->flock.l_len;
+        flock.l_type  = inodelk->flock.l_type;
+
+        full_flock.l_type = inodelk->flock.l_type;
 
         initialize_inodelk_variables (frame, this);
 
         if (local->fd) {
-                ret = fd_ctx_get (local->fd, this, &ctx);
-
-                if (ret < 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+                fd_ctx = afr_fd_ctx_get (local->fd, this);
+                if (!fd_ctx) {
+                        gf_log (this->name, GF_LOG_INFO,
                                 "unable to get fd ctx for fd=%p",
                                 local->fd);
 
@@ -1389,296 +1532,643 @@ afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
                         local->op_errno         = EINVAL;
                         int_lock->lock_op_errno = EINVAL;
 
+			afr_unlock (frame, this);
                         ret = -1;
                         goto out;
                 }
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                call_count = internal_lock_count (frame, this, fd_ctx);
+                call_count = internal_lock_count (frame, this);
                 int_lock->lk_call_count = call_count;
+                int_lock->lk_expected_count = call_count;
+
+                if (!call_count) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "fd not open on any subvolumes. aborting.");
+                        afr_unlock (frame, this);
+                        goto out;
+                }
 
                 /* Send non-blocking inodelk calls only on up children
                    and where the fd has been opened */
                 for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i] && fd_ctx->opened_on[i]) {
-                                afr_trace_inodelk_in (frame, AFR_INODELK_NB_TRANSACTION,
-                                                      AFR_LOCK_OP, &flock, F_SETLK, i);
+                        if (!local->child_up[i])
+                                continue;
 
-                                STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->finodelk,
-                                                   this->name, local->fd,
-                                                   F_SETLK, &flock);
+                        flock_use = &flock;
+                        if (!local->transaction.eager_lock_on) {
+                                goto wind;
+                        }
+
+                        piggyback = 0;
+                        local->transaction.eager_lock[i] = 1;
+
+			afr_set_delayed_post_op (frame, this);
+
+                        LOCK (&local->fd->lock);
+                        {
+                                if (fd_ctx->lock_acquired[i]) {
+                                        fd_ctx->lock_piggyback[i]++;
+                                        piggyback = 1;
+                                }
+                        }
+                        UNLOCK (&local->fd->lock);
 
+                        if (piggyback) {
+                                /* (op_ret == 1) => indicate piggybacked lock */
+                                afr_nonblocking_inodelk_cbk (frame, (void *) (long) i,
+                                                             this, 1, 0, NULL);
                                 if (!--call_count)
                                         break;
-
+                                continue;
                         }
+                        flock_use = &full_flock;
+                wind:
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_NB_TRANSACTION,
+                                              AFR_LOCK_OP, flock_use, F_SETLK, i);
 
+                        STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->finodelk,
+                                           int_lock->domain, local->fd,
+                                           F_SETLK, flock_use, NULL);
+
+                        if (!--call_count)
+                                break;
                 }
         } else {
-                call_count = internal_lock_count (frame, this, NULL);
+                call_count = internal_lock_count (frame, this);
                 int_lock->lk_call_count = call_count;
+                int_lock->lk_expected_count = call_count;
 
                 for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i]) {
-                                afr_trace_inodelk_in (frame, AFR_INODELK_NB_TRANSACTION,
-                                                      AFR_LOCK_OP, &flock, F_SETLK, i);
-
-                                STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->inodelk,
-                                                   this->name, &local->loc,
-                                                   F_SETLK, &flock);
+                        if (!local->child_up[i])
+                                continue;
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_NB_TRANSACTION,
+                                              AFR_LOCK_OP, &flock, F_SETLK, i);
 
-                                if (!--call_count)
-                                        break;
+                        STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->inodelk,
+                                           int_lock->domain, &local->loc,
+                                           F_SETLK, &flock, NULL);
 
-                        }
+                        if (!--call_count)
+                                break;
                 }
         }
-
 out:
         return ret;
 }
 
-static int
-__is_lower_locked (call_frame_t *frame, xlator_t *this)
+int32_t
+afr_unlock (call_frame_t *frame, xlator_t *this)
 {
-        afr_internal_lock_t *int_lock = NULL;
-        afr_private_t       *priv     = NULL;
-        afr_local_t         *local    = NULL;
+        afr_local_t *local = NULL;
 
-        int count = 0;
-        int i     = 0;
+        local = frame->local;
 
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-        priv     = this->private;
+        if (transaction_lk_op (local)) {
+                if (is_afr_lock_transaction (local))
+                        afr_unlock_inodelk (frame, this);
+                else
+                        afr_unlock_entrylk (frame, this);
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (int_lock->lower_locked_nodes[i] & LOCKED_LOWER)
-                        count++;
+        } else {
+                if (is_afr_lock_selfheal (local))
+                        afr_unlock_inodelk (frame, this);
+                else
+                        afr_unlock_entrylk (frame, this);
         }
 
-        return count;
-
+        return 0;
 }
 
-static int
-__is_higher_locked (call_frame_t *frame, xlator_t *this)
+int
+afr_mark_locked_nodes (xlator_t *this, fd_t *fd,
+                       unsigned char *locked_nodes)
 {
-        afr_internal_lock_t *int_lock = NULL;
-        afr_private_t       *priv     = NULL;
-        afr_local_t         *local    = NULL;
+        afr_private_t *priv  = NULL;
+        afr_fd_ctx_t  *fdctx = NULL;
+        uint64_t       tmp   = 0;
+        int            ret   = 0;
 
-        int count = 0;
-        int i = 0;
+        priv = this->private;
 
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-        priv     = this->private;
+        ret = afr_fd_ctx_set (this, fd);
+        if (ret)
+                goto out;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (int_lock->locked_nodes[i] & LOCKED_YES)
-                        count++;
+        ret = fd_ctx_get (fd, this, &tmp);
+        if (ret) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "failed to get the fd ctx");
+                goto out;
         }
+        fdctx = (afr_fd_ctx_t *) (long) tmp;
 
-        return count;
+        GF_ASSERT (fdctx->locked_on);
 
+        memcpy (fdctx->locked_on, locked_nodes,
+                priv->child_count);
+
+out:
+        return ret;
 }
 
 static int
-afr_unlock_lower_entrylk (call_frame_t *frame, xlator_t *this)
+__is_fd_saved (xlator_t *this, fd_t *fd)
 {
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-        afr_private_t       *priv     = NULL;
-        const char          *basename = NULL;
-        loc_t               *loc      = NULL;
+        afr_locked_fd_t *locked_fd = NULL;
+        afr_private_t   *priv      = NULL;
+        int              found     = 0;
 
-        int call_count = 0;
-        int i = -1;
+        priv = this->private;
 
-        local = frame->local;
-        int_lock = &local->internal_lock;
-        priv  = this->private;
+        list_for_each_entry (locked_fd, &priv->saved_fds, list) {
+                if (locked_fd->fd == fd) {
+                        found = 1;
+                        break;
+                }
+        }
 
-        basename = int_lock->lk_basename;
-        if (int_lock->lk_loc)
-                loc = int_lock->lk_loc;
+        return found;
+}
 
-        call_count = __is_lower_locked (frame, this);
-        int_lock->lk_call_count = call_count;
+static int
+__afr_save_locked_fd (xlator_t *this, fd_t *fd)
+{
+        afr_private_t   *priv      = NULL;
+        afr_locked_fd_t *locked_fd = NULL;
+        int              ret       = 0;
 
-        if (!call_count){
-                gf_log (this->name, GF_LOG_TRACE,
-                        "No internal locks unlocked");
-                int_lock->lock_cbk (frame, this);
+        priv = this->private;
+
+        locked_fd = GF_CALLOC (1, sizeof (*locked_fd),
+                               gf_afr_mt_locked_fd);
+        if (!locked_fd) {
+                ret = -1;
                 goto out;
         }
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (int_lock->lower_locked_nodes[i] & LOCKED_LOWER) {
-                        afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
-                                              AFR_UNLOCK_OP, basename, i);
+        locked_fd->fd = fd;
+        INIT_LIST_HEAD (&locked_fd->list);
 
-                        STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->entrylk,
-                                           this->name,
-                                           loc, basename,
-                                           ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
+        list_add_tail (&locked_fd->list, &priv->saved_fds);
 
-                        if (!--call_count)
-                                break;
+out:
+        return ret;
+}
 
+int
+afr_save_locked_fd (xlator_t *this, fd_t *fd)
+{
+        afr_private_t   *priv      = NULL;
+        int              ret       = 0;
+
+        priv = this->private;
+
+        pthread_mutex_lock (&priv->mutex);
+        {
+                if (__is_fd_saved (this, fd)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "fd=%p already saved", fd);
+                        goto unlock;
+                }
+
+                ret = __afr_save_locked_fd (this, fd);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "fd=%p could not be saved", fd);
+                        goto unlock;
                 }
         }
+unlock:
+        pthread_mutex_unlock (&priv->mutex);
+
+        return ret;
+}
+
+static int
+afr_lock_recovery_cleanup (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local     = NULL;
+        afr_locked_fd_t *locked_fd = NULL;
+
+        local = frame->local;
+
+        locked_fd = local->locked_fd;
+
+        STACK_DESTROY (frame->root);
+        afr_local_cleanup (local, this);
+
+        afr_save_locked_fd (this, locked_fd->fd);
 
-out:
         return 0;
 
 }
 
-
 static int
-afr_post_unlock_higher_cbk (call_frame_t *frame, xlator_t *this)
+afr_get_source_lock_recovery (xlator_t *this, fd_t *fd)
 {
-        afr_local_t *local = NULL;
+        afr_fd_ctx_t  *fdctx        = NULL;
+        afr_private_t *priv         = NULL;
+        uint64_t      tmp           = 0;
+        int           i             = 0;
+        int           source_child  = -1;
+        int           ret           = 0;
+
+        priv = this->private;
+
+        ret = fd_ctx_get (fd, this, &tmp);
+        if (ret)
+                goto out;
+
+        fdctx = (afr_fd_ctx_t *) (long) tmp;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (fdctx->locked_on[i]) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Found lock recovery source=%d", i);
+                        source_child = i;
+                        break;
+                }
+        }
+
+out:
+        return source_child;
+
+}
+
+int32_t
+afr_get_locks_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+                      dict_t *xdata);
+int32_t
+afr_recover_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+                      dict_t *xdata)
+{
+        afr_local_t   *local = NULL;
+        afr_private_t *priv  = NULL;
+        int32_t source_child = 0;
+        struct gf_flock flock   = {0,};
 
         local = frame->local;
+        priv  = this->private;
+
+        if (op_ret) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "lock recovery failed");
+                goto cleanup;
+        }
+
+        source_child = local->source_child;
+
+        memcpy (&flock, lock, sizeof (*lock));
+
+        STACK_WIND_COOKIE (frame, afr_get_locks_fd_cbk,
+                           (void *) (long) source_child,
+                           priv->children[source_child],
+                           priv->children[source_child]->fops->lk,
+                           local->fd, F_GETLK_FD, &flock, NULL);
+
+        return 0;
+
+cleanup:
+        afr_lock_recovery_cleanup (frame, this);
+        return 0;
+}
+
+int
+afr_recover_lock (call_frame_t *frame, xlator_t *this,
+                  struct gf_flock *flock)
+{
+        afr_local_t   *local             = NULL;
+        afr_private_t *priv              = NULL;
+        int32_t      lock_recovery_child = 0;
+
+        priv  = this->private;
+        local = frame->local;
+
+        lock_recovery_child = local->lock_recovery_child;
+
+        frame->root->lk_owner = flock->l_owner;
+
+        STACK_WIND_COOKIE (frame, afr_recover_lock_cbk,
+                           (void *) (long) lock_recovery_child,
+                           priv->children[lock_recovery_child],
+                           priv->children[lock_recovery_child]->fops->lk,
+                           local->fd, F_SETLK, flock, NULL);
 
-        local->transaction.done (frame, this);
         return 0;
 }
 
 static int
-afr_post_unlock_lower_cbk (call_frame_t *frame, xlator_t *this)
+is_afr_lock_eol (struct gf_flock *lock)
 {
-        afr_internal_lock_t *int_lock    = NULL;
-        afr_local_t         *local       = NULL;
-	loc_t               *lower       = NULL;
-	loc_t               *higher      = NULL;
-	const char          *lower_name  = NULL;
-	const char          *higher_name = NULL;
+        int ret = 0;
 
-        local    = frame->local;
-        int_lock = &local->internal_lock;
+        if ((lock->l_type == GF_LK_EOL))
+                ret = 1;
 
-        lower = lower_path (&local->transaction.parent_loc,
-                            local->transaction.basename,
-                            &local->transaction.new_parent_loc,
-                            local->transaction.new_basename);
+        return ret;
+}
+
+int32_t
+afr_get_locks_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+                      dict_t *xdata)
+{
+        if (op_ret) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Failed to get locks on fd");
+                goto cleanup;
+        }
 
-        lower_name = (lower == &local->transaction.parent_loc ?
-                      local->transaction.basename :
-                      local->transaction.new_basename);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Got a lock on fd");
 
-        higher = (lower == &local->transaction.parent_loc ?
-                  &local->transaction.new_parent_loc :
-                  &local->transaction.parent_loc);
+        if (is_afr_lock_eol (lock)) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Reached EOL on locks on fd");
+                goto cleanup;
+        }
 
-        higher_name = (higher == &local->transaction.parent_loc ?
-                       local->transaction.basename :
-                       local->transaction.new_basename);
+        afr_recover_lock (frame, this, lock);
 
-        if (__is_higher_locked (frame, this)) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "unlocking higher");
-                int_lock->lk_basename = higher_name;
-                int_lock->lk_loc      = higher;
-                int_lock->lock_cbk    = afr_post_unlock_higher_cbk;
+        return 0;
 
-                afr_unlock_entrylk (frame, this);
-        } else
-                local->transaction.done (frame, this);
+cleanup:
+        afr_lock_recovery_cleanup (frame, this);
 
         return 0;
 }
 
 static int
-afr_rename_unlock (call_frame_t *frame, xlator_t *this)
+afr_lock_recovery (call_frame_t *frame, xlator_t *this)
 {
-        afr_internal_lock_t *int_lock    = NULL;
-        afr_local_t         *local       = NULL;
-	loc_t               *lower       = NULL;
-	loc_t               *higher      = NULL;
-	const char          *lower_name  = NULL;
-	const char          *higher_name = NULL;
+        afr_local_t   *local        = NULL;
+        afr_private_t *priv         = NULL;
+        fd_t          *fd           = NULL;
+        int            ret          = 0;
+        int32_t        source_child = 0;
+        struct gf_flock   flock        = {0,};
 
-        local    = frame->local;
-        int_lock = &local->internal_lock;
+        priv  = this->private;
+        local = frame->local;
 
-        lower = lower_path (&local->transaction.parent_loc,
-                            local->transaction.basename,
-                            &local->transaction.new_parent_loc,
-                            local->transaction.new_basename);
+        fd = local->fd;
 
-        lower_name = (lower == &local->transaction.parent_loc ?
-                      local->transaction.basename :
-                      local->transaction.new_basename);
+        source_child = afr_get_source_lock_recovery (this, fd);
+        if (source_child < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not recover locks due to lock "
+                        "split brain");
+                ret = -1;
+                goto out;
+        }
 
-        higher = (lower == &local->transaction.parent_loc ?
-                  &local->transaction.new_parent_loc :
-                  &local->transaction.parent_loc);
+        local->source_child = source_child;
 
-        higher_name = (higher == &local->transaction.parent_loc ?
-                       local->transaction.basename :
-                       local->transaction.new_basename);
+        /* the flock can be zero filled as we're querying incrementally
+           the locks held on the fd.
+        */
+        STACK_WIND_COOKIE (frame, afr_get_locks_fd_cbk,
+                           (void *) (long) source_child,
+                           priv->children[source_child],
+                           priv->children[source_child]->fops->lk,
+                           local->fd, F_GETLK_FD, &flock, NULL);
+
+out:
+        return ret;
+}
 
 
-        if (__is_lower_locked (frame, this)) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "unlocking lower");
-                int_lock->lk_basename = lower_name;
-                int_lock->lk_loc      = lower;
-                int_lock->lock_cbk    = afr_post_unlock_lower_cbk;
+static int
+afr_mark_fd_opened (xlator_t *this, fd_t *fd, int32_t child_index)
+{
+        afr_fd_ctx_t *fdctx = NULL;
+        uint64_t      tmp   = 0;
+        int           ret   = 0;
+
+        ret = fd_ctx_get (fd, this, &tmp);
+        if (ret)
+                goto out;
+
+        fdctx = (afr_fd_ctx_t *) (long) tmp;
+
+        fdctx->opened_on[child_index] = AFR_FD_OPENED;
+
+out:
+        return ret;
+}
+
+int32_t
+afr_lock_recovery_preopen_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                               int32_t op_ret, int32_t op_errno, fd_t *fd,
+                               dict_t *xdata)
+{
+        int32_t child_index = (long )cookie;
+        int ret = 0;
+
+        if (op_ret) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Reopen during lock-recovery failed");
+                goto cleanup;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Open succeeded => proceed to recover locks");
+
+        ret = afr_lock_recovery (frame, this);
+        if (ret) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Lock recovery failed");
+                goto cleanup;
+        }
+
+        ret = afr_mark_fd_opened (this, fd, child_index);
+        if (ret) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Marking fd open failed");
+                goto cleanup;
+        }
 
-                afr_unlock_lower_entrylk (frame, this);
-        } else
-                afr_post_unlock_lower_cbk (frame, this);
+        return 0;
 
+cleanup:
+        afr_lock_recovery_cleanup (frame, this);
         return 0;
 }
 
 static int
-afr_rename_transaction (call_frame_t *frame, xlator_t *this)
+afr_lock_recovery_preopen (call_frame_t *frame, xlator_t *this)
 {
-        afr_local_t *local = NULL;
+        afr_private_t *priv  = NULL;
+        afr_local_t   *local = NULL;
+        uint64_t       tmp   = 0;
+        afr_fd_ctx_t  *fdctx = NULL;
+        loc_t          loc   = {0,};
+        int32_t        child_index = 0;
+        int            ret   = 0;
 
+        priv  = this->private;
         local = frame->local;
 
-        return (local->transaction.type ==
-                AFR_ENTRY_RENAME_TRANSACTION);
+        GF_ASSERT (local && local->fd);
+
+        ret = fd_ctx_get (local->fd, this, &tmp);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to get the context of fd",
+                        uuid_utoa (local->fd->inode->gfid));
+        fdctx = (afr_fd_ctx_t *) (long) tmp;
+        /* TODO: instead we should return from the function */
+        GF_ASSERT (fdctx);
+
+        child_index = local->lock_recovery_child;
+
+        inode_path (local->fd->inode, NULL, (char **)&loc.path);
+        loc.name   = strrchr (loc.path, '/');
+        loc.inode  = inode_ref (local->fd->inode);
+        loc.parent = inode_parent (local->fd->inode, 0, NULL);
+
+
+        STACK_WIND_COOKIE (frame, afr_lock_recovery_preopen_cbk,
+                           (void *)(long) child_index,
+                           priv->children[child_index],
+                           priv->children[child_index]->fops->open,
+                           &loc, fdctx->flags, local->fd, NULL);
 
+        return 0;
 }
 
-int32_t
-afr_unlock (call_frame_t *frame, xlator_t *this)
+static int
+is_fd_opened (fd_t *fd, int32_t child_index)
 {
-        afr_local_t *local = NULL;
+        afr_fd_ctx_t *fdctx = NULL;
+        uint64_t      tmp = 0;
+        int           ret = 0;
+
+        ret = fd_ctx_get (fd, THIS, &tmp);
+        if (ret)
+                goto out;
+
+        fdctx = (afr_fd_ctx_t *) (long) tmp;
+
+        if (fdctx->opened_on[child_index] == AFR_FD_OPENED)
+                ret = 1;
+
+out:
+        return ret;
+}
+
+int
+afr_attempt_lock_recovery (xlator_t *this, int32_t child_index)
+{
+        call_frame_t    *frame      = NULL;
+        afr_private_t   *priv       = NULL;
+        afr_local_t     *local      = NULL;
+        afr_locked_fd_t *locked_fd  = NULL;
+        afr_locked_fd_t  *tmp       = NULL;
+        int              ret        = -1;
+        struct list_head locks_list = {0,};
+        int32_t          op_errno   = 0;
+
+
+        priv = this->private;
 
+        if (list_empty (&priv->saved_fds))
+                goto out;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame) {
+                ret = -1;
+                goto out;
+        }
+
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
         local = frame->local;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0) {
+                ret = -1;
+                goto out;
+        }
 
-        if (transaction_lk_op (local)) {
-                if (is_afr_lock_transaction (local))
-                        afr_unlock_inodelk (frame, this);
-                else
-                        if (!afr_rename_transaction (frame, this))
-                                afr_unlock_entrylk (frame, this);
-                        else
-                                afr_rename_unlock (frame, this);
-        } else {
-                if (is_afr_lock_selfheal (local))
-                        afr_unlock_inodelk (frame, this);
-                else
-                        afr_unlock_entrylk (frame, this);
+        frame->local = local;
+
+        INIT_LIST_HEAD (&locks_list);
+
+        pthread_mutex_lock (&priv->mutex);
+        {
+                list_splice_init (&priv->saved_fds, &locks_list);
         }
+        pthread_mutex_unlock (&priv->mutex);
 
-        return 0;
+        list_for_each_entry_safe (locked_fd, tmp,
+                                  &locks_list, list) {
+
+                list_del_init (&locked_fd->list);
+
+                local->fd                  = fd_ref (locked_fd->fd);
+                local->lock_recovery_child = child_index;
+                local->locked_fd           = locked_fd;
+
+                if (!is_fd_opened (locked_fd->fd, child_index)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "attempting open before lock "
+                                "recovery");
+                        afr_lock_recovery_preopen (frame, this);
+                } else {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "attempting lock recovery "
+                                "without a preopen");
+                        afr_lock_recovery (frame, this);
+                }
+        }
+
+out:
+        if ((ret < 0) && frame)
+                AFR_STACK_DESTROY (frame);
+        return ret;
+}
+
+int
+afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src, char *dom,
+                          unsigned int child_count)
+{
+        afr_local_t         *dst_local   = NULL;
+        afr_local_t         *src_local   = NULL;
+        afr_internal_lock_t *dst_lock    = NULL;
+        afr_internal_lock_t *src_lock    = NULL;
+        afr_inodelk_t       *dst_inodelk = NULL;
+        afr_inodelk_t       *src_inodelk = NULL;
+        int                 ret = -1;
+
+        src_local = src->local;
+        src_lock  = &src_local->internal_lock;
+        src_inodelk = afr_get_inodelk (src_lock, dom);
+        dst_local = dst->local;
+        dst_lock  = &dst_local->internal_lock;
+        dst_inodelk = afr_get_inodelk (dst_lock, dom);
+        if (!dst_inodelk || !src_inodelk)
+                goto out;
+        if (src_inodelk->locked_nodes) {
+                memcpy (dst_inodelk->locked_nodes, src_inodelk->locked_nodes,
+                        sizeof (*dst_inodelk->locked_nodes) * child_count);
+                memset (src_inodelk->locked_nodes, 0,
+                        sizeof (*src_inodelk->locked_nodes) * child_count);
+        }
+
+        dst_lock->transaction_lk_type = src_lock->transaction_lk_type;
+        dst_lock->selfheal_lk_type    = src_lock->selfheal_lk_type;
+        dst_inodelk->lock_count = src_inodelk->lock_count;
+        src_inodelk->lock_count = 0;
+        ret = 0;
+out:
+        return ret;
 }
diff --git a/xlators/cluster/afr/src/afr-mem-types.h b/xlators/cluster/afr/src/afr-mem-types.h
index c1a86b827..73594f265 100644
--- a/xlators/cluster/afr/src/afr-mem-types.h
+++ b/xlators/cluster/afr/src/afr-mem-types.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -26,7 +17,6 @@
 enum gf_afr_mem_types_ {
         gf_afr_mt_iovec  = gf_common_mt_end + 1,
         gf_afr_mt_afr_fd_ctx_t,
-        gf_afr_mt_afr_local_t,
         gf_afr_mt_afr_private_t,
         gf_afr_mt_int32_t,
         gf_afr_mt_char,
@@ -41,6 +31,20 @@ enum gf_afr_mem_types_ {
         gf_afr_mt_loc_t,
         gf_afr_mt_entry_name,
         gf_afr_mt_pump_priv,
+        gf_afr_mt_locked_fd,
+        gf_afr_mt_inode_ctx_t,
+        gf_afr_fd_paused_call_t,
+        gf_afr_mt_crawl_data_t,
+        gf_afr_mt_brick_pos_t,
+        gf_afr_mt_shd_bool_t,
+        gf_afr_mt_shd_timer_t,
+        gf_afr_mt_shd_event_t,
+        gf_afr_mt_time_t,
+        gf_afr_mt_pos_data_t,
+        gf_afr_mt_reply_t,
+        gf_afr_mt_stats_t,
+        gf_afr_mt_shd_crawl_event_t,
+        gf_afr_mt_uint64_t,
         gf_afr_mt_end
 };
 #endif
diff --git a/xlators/cluster/afr/src/afr-open.c b/xlators/cluster/afr/src/afr-open.c
index dc830f7d5..643a5d692 100644
--- a/xlators/cluster/afr/src/afr-open.c
+++ b/xlators/cluster/afr/src/afr-open.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <libgen.h>
@@ -55,436 +46,337 @@
 #include "afr-self-heal.h"
 #include "afr-self-heal-common.h"
 
-
 int
-afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-			int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                        struct iatt *postbuf)
+afr_stale_child_up (afr_local_t *local, xlator_t *this)
 {
-	afr_local_t * local = frame->local;
-
-	AFR_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
-			  local->fd);
-	return 0;
-}
-
-
-int
-afr_open_cbk (call_frame_t *frame, void *cookie,
-	      xlator_t *this, int32_t op_ret, int32_t op_errno,
-	      fd_t *fd)
-{
-	afr_local_t *  local = NULL;
-
-        int child_index = (long) cookie;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-        int ret = 0;
-
-	int call_count = -1;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-		}
-
-		if (op_ret >= 0) {
-			local->op_ret = op_ret;
-                        local->success_count++;
-
-                        ret = afr_fd_ctx_set (this, fd);
-
-                        if (ret < 0) {
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "could not set fd ctx for fd=%p",
-                                        fd);
-
-                                local->op_ret   = -1;
-                                local->op_errno = -ret;
-                        }
+        int             i = 0;
+        afr_private_t   *priv = NULL;
+        int             up = -1;
 
-                        ret = fd_ctx_get (fd, this, &ctx);
+        priv = this->private;
 
-                        if (ret < 0) {
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "could not get fd ctx for fd=%p", fd);
-                                local->op_ret   = -1;
-                                local->op_errno = -ret;
-                        }
-
-                        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                        fd_ctx->opened_on[child_index] = 1;
-                        fd_ctx->flags                  = local->cont.open.flags;
-                        fd_ctx->wbflags                = local->cont.open.wbflags;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
+        if (!local->fresh_children)
+                local->fresh_children = afr_children_create (priv->child_count);
+        if (!local->fresh_children)
+                goto out;
 
-	if (call_count == 0) {
-		if ((local->cont.open.flags & O_TRUNC)
-		    && (local->op_ret >= 0)) {
-			STACK_WIND (frame, afr_open_ftruncate_cbk,
-				    this, this->fops->ftruncate,
-				    fd, 0);
-		} else {
-                        AFR_STACK_UNWIND (open, frame, local->op_ret,
-                                          local->op_errno, local->fd);
-		}
-	}
+        afr_inode_get_read_ctx (this, local->fd->inode, local->fresh_children);
+        if (priv->child_count == afr_get_children_count (local->fresh_children,
+                                                         priv->child_count))
+                goto out;
 
-	return 0;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!local->child_up[i])
+                        continue;
+                if (afr_is_child_present (local->fresh_children,
+                                          priv->child_count, i))
+                        continue;
+                up = i;
+                break;
+        }
+out:
+        return up;
 }
 
-
-int
-afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-          fd_t *fd, int32_t wbflags)
+void
+afr_perform_data_self_heal (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t *   local = NULL;
-
-	int     i = 0;
-	int   ret = -1;
-
-	int32_t call_count = 0;	
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-	int32_t wind_flags = flags & (~O_TRUNC);
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-	VALIDATE_OR_GOTO (loc, out);
-
-	priv = this->private;
-
-        if (afr_is_split_brain (this, loc->inode)) {
-		/* self-heal failed */
-		op_errno = EIO;
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
-
-	frame->local = local;
-	call_count   = local->call_count;
-
-        loc_copy (&local->loc, loc);
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        inode_t         *inode = NULL;
+        int             st_child = -1;
+        char            reason[64] = {0};
 
-	local->cont.open.flags   = flags;
-        local->cont.open.wbflags = wbflags;
+        local = frame->local;
+        sh = &local->self_heal;
+        inode = local->fd->inode;
 
-	local->fd = fd_ref (fd);
+        if (!IA_ISREG (inode->ia_type))
+                goto out;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_open_cbk, (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->open,
-					   loc, wind_flags, fd, wbflags);
+        st_child = afr_stale_child_up (local, this);
+        if (st_child < 0)
+                goto out;
 
-			if (!--call_count)
-				break;
-		}
-	}
+        sh->do_data_self_heal          = _gf_true;
+        sh->do_metadata_self_heal      = _gf_true;
+        sh->do_gfid_self_heal          = _gf_true;
+        sh->do_missing_entry_self_heal = _gf_true;
 
-	op_ret = 0;
+        snprintf (reason, sizeof (reason), "stale subvolume %d detected",
+                  st_child);
+        afr_launch_self_heal (frame, this, inode, _gf_true, inode->ia_type,
+                              reason, NULL, NULL);
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (open, frame, op_ret, op_errno, fd);
-	}
-
-	return 0;
+        return;
 }
 
-
 int
-afr_up_down_flush_open_cbk (call_frame_t *frame, void *cookie,
-                            xlator_t *this, int32_t op_ret, int32_t op_errno,
-                            fd_t *fd)
+afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                        struct iatt *postbuf, dict_t *xdata)
 {
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-        afr_private_t       *priv     = NULL;
+        afr_local_t * local = frame->local;
+        afr_private_t *priv = NULL;
 
-        int ret = 0;
+        priv = this->private;
+        if (afr_open_only_data_self_heal (priv->data_self_heal))
+                afr_perform_data_self_heal (frame, this);
+        AFR_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
+                          local->fd, xdata);
+        return 0;
+}
 
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
 
-        int call_count  = 0;
-        int child_index = (long) cookie;
+int
+afr_open_cbk (call_frame_t *frame, void *cookie,
+              xlator_t *this, int32_t op_ret, int32_t op_errno,
+              fd_t *fd, dict_t *xdata)
+{
+        afr_local_t *  local       = NULL;
+        int            ret         = 0;
+        int            call_count  = -1;
+        int            child_index = (long) cookie;
+        afr_private_t *priv        = NULL;
 
-        priv     = this->private;
-        local    = frame->local;
-        int_lock = &local->internal_lock;
+        priv = this->private;
+        local = frame->local;
 
         LOCK (&frame->lock);
         {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                }
+
                 if (op_ret >= 0) {
-                        ret = fd_ctx_get (fd, this, &ctx);
+                        local->op_ret = op_ret;
+                        local->success_count++;
 
-                        if (ret < 0) {
-                                goto out;
+                        ret = afr_child_fd_ctx_set (this, fd, child_index,
+                                                    local->cont.open.flags);
+                        if (ret) {
+                                local->op_ret = -1;
+                                local->op_errno = -ret;
+                                goto unlock;
                         }
-
-                        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                        fd_ctx->opened_on[child_index] = 1;
-
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "fd for %s opened successfully on subvolume %s",
-                                local->loc.path, priv->children[child_index]->name);
                 }
         }
-out:
+unlock:
         UNLOCK (&frame->lock);
 
         call_count = afr_frame_return (frame);
 
         if (call_count == 0) {
-                int_lock->lock_cbk = local->transaction.done;
-                local->transaction.post_post_op (frame, this);
+                if ((local->cont.open.flags & O_TRUNC)
+                    && (local->op_ret >= 0)) {
+                        STACK_WIND (frame, afr_open_ftruncate_cbk,
+                                    this, this->fops->ftruncate,
+                                    fd, 0, NULL);
+                } else {
+                        if (afr_open_only_data_self_heal (priv->data_self_heal))
+                                afr_perform_data_self_heal (frame, this);
+                        AFR_STACK_UNWIND (open, frame, local->op_ret,
+                                          local->op_errno, local->fd, xdata);
+                }
         }
 
         return 0;
 }
 
-
-static int
-__unopened_count (int child_count, unsigned char *opened_on, unsigned char *child_up)
-{
-        int i;
-        int count = 0;
-
-        for (i = 0; i < child_count; i++) {
-                if (!opened_on[i] && child_up[i])
-                        count++;
-        }
-
-        return count;
-}
-
-
 int
-afr_up_down_flush_sh_unwind (call_frame_t *frame, xlator_t *this)
+afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+          fd_t *fd, dict_t *xdata)
 {
-        afr_local_t *local  = NULL;
-        afr_private_t *priv = NULL;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
+        afr_private_t * priv       = NULL;
+        afr_local_t *   local      = NULL;
+        int             i          = 0;
+        int             ret        = -1;
+        int32_t         call_count = 0;
+        int32_t         op_errno   = 0;
+        int32_t         wind_flags = flags & (~O_TRUNC);
+        //We can't let truncation to happen outside transaction.
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (this, out);
+        VALIDATE_OR_GOTO (this->private, out);
+        VALIDATE_OR_GOTO (loc, out);
+
+        priv = this->private;
+
+        if (flags & (O_CREAT|O_TRUNC)) {
+                QUORUM_CHECK(open,out);
+        }
 
-        int abandon = 0;
-        int ret = 0;
-        int i;
-        int call_count = 0;
+        if (afr_is_split_brain (this, loc->inode)) {
+                /* self-heal failed */
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to open as split brain seen, returning EIO");
+                op_errno = EIO;
+                goto out;
+        }
 
-        priv  = this->private;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
         local = frame->local;
 
-        /*
-         * Some subvolumes might have come up on which we never
-         * opened this fd in the first place. Re-open fd's on those
-         * subvolumes now.
-         */
-
-        ret = fd_ctx_get (local->fd, this, &ctx);
-
-        if (ret < 0) {
-                abandon = 1;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
                 goto out;
-        }
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        call_count   = local->call_count;
+        loc_copy (&local->loc, loc);
 
-        call_count = __unopened_count (priv->child_count, fd_ctx->opened_on,
-                                       local->child_up);
+        local->cont.open.flags   = flags;
 
-        if (call_count == 0) {
-                abandon = 1;
-                goto out;
-        }
-
-        local->call_count = call_count;
+        local->fd = fd_ref (fd);
 
         for (i = 0; i < priv->child_count; i++) {
-                if (!fd_ctx->opened_on[i] && local->child_up[i]) {
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "opening fd for %s on subvolume %s",
-                                local->loc.path, priv->children[i]->name);
-
-                        STACK_WIND_COOKIE (frame, afr_up_down_flush_open_cbk,
-                                           (void *)(long) i,
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, afr_open_cbk, (void *) (long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->open,
-                                           &local->loc, fd_ctx->flags, local->fd,
-                                           fd_ctx->wbflags);
+                                           loc, wind_flags, fd, xdata);
 
                         if (!--call_count)
                                 break;
                 }
         }
 
+        ret = 0;
 out:
-        if (abandon)
-                local->transaction.post_post_op (frame, this);
-
-        return 0;
-}
-
-
-int
-afr_up_down_flush_post_post_op (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t     *local  = NULL;
-        afr_self_heal_t *sh = NULL;
-        char            sh_type_str[256] = {0,};
-
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        inode_path (local->fd->inode, NULL, (char **)&local->loc.path);
-        local->loc.name   = strrchr (local->loc.path, '/');
-        local->loc.inode  = inode_ref (local->fd->inode);
-        local->loc.parent = inode_parent (local->fd->inode, 0, NULL);
-
-        /* forcibly trigger missing-entries self-heal */
-
-        local->success_count    = 1;
-        local->enoent_count     = 1;
-
-        sh->data_lock_held      = _gf_true;
-        sh->need_data_self_heal = _gf_true;
-        sh->type                = local->fd->inode->ia_type;
-        sh->background          = _gf_false;
-        sh->unwind              = afr_up_down_flush_sh_unwind;
-
-        afr_self_heal_type_str_get(&local->self_heal,
-                                   sh_type_str,
-                                   sizeof(sh_type_str));
-        gf_log (this->name, GF_LOG_NORMAL, "%s self-heal triggered. "
-                "path: %s, reason: Replicate up down flush, data lock is held",
-                sh_type_str, local->loc.path);
-
-        afr_self_heal (frame, this);
+        if (ret < 0)
+                AFR_STACK_UNWIND (open, frame, -1, op_errno, fd, xdata);
 
         return 0;
 }
 
-
-int
-afr_up_down_flush_wind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *local = NULL;
-
-	local = frame->local;
-
-        local->transaction.resume (frame, this);
-	return 0;
-}
-
-
 int
-afr_up_down_flush_done (call_frame_t *frame, xlator_t *this)
+afr_openfd_fix_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, fd_t *fd,
+                         dict_t *xdata)
 {
-        afr_private_t *priv = NULL;
-	afr_local_t *local  = NULL;
+        afr_local_t   *local      = NULL;
+        afr_private_t *priv       = NULL;
+        afr_fd_ctx_t  *fd_ctx     = NULL;
+        int           call_count  = 0;
+        int           child_index = (long) cookie;
 
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
+        priv     = this->private;
+        local    = frame->local;
 
-        int _ret = -1;
-        int i    = 0;
+        if (op_ret >= 0) {
+                gf_log (this->name, GF_LOG_DEBUG, "fd for %s opened "
+                        "successfully on subvolume %s", local->loc.path,
+                        priv->children[child_index]->name);
+        } else {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to open %s "
+                        "on subvolume %s", local->loc.path,
+                        priv->children[child_index]->name);
+        }
 
-        priv  = this->private;
-	local = frame->local;
+        fd_ctx = afr_fd_ctx_get (local->fd, this);
+        if (!fd_ctx) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to get fd context, %p", local->fd);
+                goto out;
+        }
 
         LOCK (&local->fd->lock);
         {
-                _ret = __fd_ctx_get (local->fd, this, &ctx);
-
-                if (_ret < 0) {
-                        goto out;
-                }
-
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                fd_ctx->down_count = priv->down_count;
-                fd_ctx->up_count   = priv->up_count;
-
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i])
-                                fd_ctx->pre_op_done[i] = 0;
+                if (op_ret >= 0) {
+                        fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+                } else {
+                        fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
                 }
         }
-out:
         UNLOCK (&local->fd->lock);
+out:
+        call_count = afr_frame_return (frame);
+        if (call_count == 0)
+                AFR_STACK_DESTROY (frame);
 
-        afr_local_transaction_cleanup (local, this);
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "The up/down flush is over");
+        return 0;
+}
 
-        fd_unref (local->fd);
-        local->up_down_flush_cbk (frame, this);
+void
+afr_fix_open (xlator_t *this, fd_t *fd, size_t need_open_count, int *need_open)
+{
+        afr_private_t *priv    = NULL;
+        int           i        = 0;
+        call_frame_t  *frame   = NULL;
+        afr_local_t   *local   = NULL;
+        int           ret      = -1;
+        int32_t       op_errno = 0;
+        afr_fd_ctx_t  *fd_ctx  = NULL;
 
-	return 0;
-}
+        priv  = this->private;
 
+        if (!afr_is_fd_fixable (fd) || !need_open || !need_open_count)
+                goto out;
 
-int
-afr_up_down_flush (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                   afr_flush_type type)
-{
-	afr_local_t   * local = NULL;
+        fd_ctx = afr_fd_ctx_get (fd, this);
+        if (!fd_ctx) {
+                ret = -1;
+                goto out;
+        }
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame) {
+                ret = -1;
+                goto out;
+        }
 
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
         local = frame->local;
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
-        local->op = GF_FOP_FLUSH;
+        local->loc.inode = inode_ref (fd->inode);
+        ret = loc_path (&local->loc, NULL);
+        if (ret < 0)
+                goto out;
 
         local->fd = fd_ref (fd);
+        local->call_count = need_open_count;
 
-        local->transaction.fop          = afr_up_down_flush_wind;
-        local->transaction.done         = afr_up_down_flush_done;
+        gf_log (this->name, GF_LOG_DEBUG, "need open count: %zd",
+                need_open_count);
 
-        switch (type) {
-        case AFR_CHILD_UP_FLUSH:
-                local->transaction.post_post_op = afr_up_down_flush_post_post_op;
-                break;
-
-        case AFR_CHILD_DOWN_FLUSH:
-                local->transaction.post_post_op = NULL;
-                break;
-        }
+        for (i = 0; i < priv->child_count; i++) {
+                if (!need_open[i])
+                        continue;
 
-        local->transaction.start  = 0;
-        local->transaction.len    = 0;
+                if (IA_IFDIR == fd->inode->ia_type) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "opening fd for dir %s on subvolume %s",
+                                local->loc.path, priv->children[i]->name);
 
-        gf_log (this->name, GF_LOG_TRACE,
-                "doing up/down flush on fd=%p",
-                fd);
+                        STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
+                                           (void*) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->opendir,
+                                           &local->loc, local->fd,
+                                           NULL);
+                } else {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "opening fd for file %s on subvolume %s",
+                                local->loc.path, priv->children[i]->name);
 
-        afr_transaction (frame, this, AFR_FLUSH_TRANSACTION);
+                        STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
+                                           (void *)(long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->open,
+                                           &local->loc,
+                                           fd_ctx->flags & (~O_TRUNC),
+                                           local->fd, NULL);
+                }
 
+        }
+        op_errno = 0;
+        ret = 0;
 out:
-	return 0;
+        if (op_errno)
+                ret = -1; //For handling ALLOC_OR_GOTO
+        if (ret && frame)
+                AFR_STACK_DESTROY (frame);
 }
diff --git a/xlators/cluster/afr/src/afr-self-heal-algorithm.c b/xlators/cluster/afr/src/afr-self-heal-algorithm.c
index 0f7570dc2..83846f152 100644
--- a/xlators/cluster/afr/src/afr-self-heal-algorithm.c
+++ b/xlators/cluster/afr/src/afr-self-heal-algorithm.c
@@ -1,23 +1,15 @@
 /*
-   Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
+#include <openssl/md5.h>
 #include "glusterfs.h"
 #include "afr.h"
 #include "xlator.h"
@@ -33,7 +25,6 @@
 #include "compat-errno.h"
 #include "compat.h"
 #include "byte-order.h"
-#include "md5.h"
 
 #include "afr-transaction.h"
 #include "afr-self-heal.h"
@@ -44,751 +35,624 @@
   This file contains the various self-heal algorithms
 */
 
+static int
+sh_loop_driver (call_frame_t *sh_frame, xlator_t *this,
+                gf_boolean_t is_first_call, call_frame_t *old_loop_frame);
+static int
+sh_loop_return (call_frame_t *sh_frame, xlator_t *this, call_frame_t *loop_frame,
+                int32_t op_ret, int32_t op_errno);
+static int
+sh_destroy_frame (call_frame_t *frame, xlator_t *this)
+{
+        if (!frame)
+                goto out;
 
-/*
-  The "full" algorithm. Copies the entire file from
-  source to sinks.
-*/
-
+        AFR_STACK_DESTROY (frame);
+out:
+        return 0;
+}
 
 static void
-sh_full_private_cleanup (call_frame_t *frame, xlator_t *this)
+sh_private_cleanup (call_frame_t *frame, xlator_t *this)
 {
-        afr_local_t *               local   = NULL;
-        afr_self_heal_t *           sh      = NULL;
-        afr_sh_algo_full_private_t *sh_priv = NULL;
+        afr_local_t             *local   = NULL;
+        afr_self_heal_t         *sh      = NULL;
+        afr_sh_algo_private_t   *sh_priv = NULL;
 
         local = frame->local;
         sh    = &local->self_heal;
 
         sh_priv = sh->private;
-
-        if (sh_priv)
-                GF_FREE (sh_priv);
+        GF_FREE (sh_priv);
 }
 
-
-static int
-sh_full_loop_driver (call_frame_t *frame, xlator_t *this);
-
 static int
-sh_full_loop_return (call_frame_t *rw_frame, xlator_t *this, off_t offset)
+sh_number_of_writes_needed (unsigned char *write_needed, int child_count)
 {
-        afr_local_t *               rw_local   = NULL;
-        afr_self_heal_t *           rw_sh      = NULL;
-
-        call_frame_t *sh_frame              = NULL;
-	afr_local_t * sh_local              = NULL;
-	afr_self_heal_t *sh                 = NULL;
-        afr_sh_algo_full_private_t *sh_priv = NULL;
-
-        rw_local = rw_frame->local;
-        rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-        sh_priv  = sh->private;
+        int writes = 0;
+        int i      = 0;
 
-        LOCK (&sh_priv->lock);
-        {
-                sh_priv->loops_running--;
+        for (i = 0; i < child_count; i++) {
+                if (write_needed[i])
+                        writes++;
         }
-        UNLOCK (&sh_priv->lock);
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "loop for offset %"PRId64" returned", offset);
 
-        AFR_STACK_DESTROY (rw_frame);
-
-        sh_full_loop_driver (sh_frame, this);
-
-        return 0;
+        return writes;
 }
 
 
 static int
-sh_full_write_cbk (call_frame_t *rw_frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                   struct iatt *postbuf)
+sh_loop_driver_done (call_frame_t *sh_frame, xlator_t *this,
+                     call_frame_t *last_loop_frame)
 {
-	afr_private_t * priv    = NULL;
-	afr_local_t * rw_local  = NULL;
-	afr_self_heal_t *rw_sh  = NULL;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-	int child_index = (long) cookie;
-	int call_count = 0;
-
-	priv = this->private;
+        afr_local_t             *local        = NULL;
+        afr_self_heal_t         *sh           = NULL;
+        afr_sh_algo_private_t   *sh_priv      = NULL;
+        int32_t                 total_blocks = 0;
+        int32_t                 diff_blocks  = 0;
+
+        local        = sh_frame->local;
+        sh           = &local->self_heal;
+        sh_priv      = sh->private;
+        if (sh_priv) {
+                total_blocks = sh_priv->total_blocks;
+                diff_blocks  = sh_priv->diff_blocks;
+        }
 
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
+        sh_private_cleanup (sh_frame, this);
+        if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+                GF_ASSERT (!last_loop_frame);
+                //loop_finish should have happened and the old_loop should be NULL
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "self-heal aborting on %s",
+                        local->loc.path);
 
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
+                local->self_heal.algo_abort_cbk (sh_frame, this);
+        } else {
+                GF_ASSERT (last_loop_frame);
+                if (diff_blocks == total_blocks) {
+                        gf_log (this->name, GF_LOG_DEBUG, "full self-heal "
+                                "completed on %s",local->loc.path);
+                } else {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "diff self-heal on %s: completed. "
+                                "(%d blocks of %d were different (%.2f%%))",
+                                local->loc.path, diff_blocks, total_blocks,
+                                ((diff_blocks * 1.0)/total_blocks) * 100);
+                }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"wrote %d bytes of data from %s to child %d, offset %"PRId64"",
-		op_ret, sh_local->loc.path, child_index,
-                rw_sh->offset - op_ret);
+                sh->old_loop_frame = last_loop_frame;
+                local->self_heal.algo_completion_cbk (sh_frame, this);
+        }
 
-	LOCK (&sh_frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"write to %s failed on subvolume %s (%s)",
-				sh_local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
+        return 0;
+}
 
-			sh->op_failed = 1;
-		}
-	}
-	UNLOCK (&sh_frame->lock);
+int
+sh_loop_finish (call_frame_t *loop_frame, xlator_t *this)
+{
+        afr_local_t             *loop_local = NULL;
+        afr_self_heal_t         *loop_sh = NULL;
 
-	call_count = afr_frame_return (rw_frame);
+        if (!loop_frame)
+                goto out;
 
-	if (call_count == 0) {
-		sh_full_loop_return (rw_frame, this, rw_sh->offset - op_ret);
-	}
+        loop_local = loop_frame->local;
+        if (loop_local) {
+                loop_sh = &loop_local->self_heal;
+        }
 
-	return 0;
+        if (loop_sh && loop_sh->data_lock_held) {
+                afr_sh_data_unlock (loop_frame, this, this->name,
+                                    sh_destroy_frame);
+        } else {
+                sh_destroy_frame (loop_frame, this);
+        }
+out:
+        return 0;
 }
 
-
 static int
-sh_full_read_cbk (call_frame_t *rw_frame, void *cookie,
-                  xlator_t *this, int32_t op_ret, int32_t op_errno,
-                  struct iovec *vector, int32_t count, struct iatt *buf,
-                  struct iobref *iobref)
+sh_loop_lock_success (call_frame_t *loop_frame, xlator_t *this)
 {
-	afr_private_t * priv    = NULL;
-	afr_local_t * rw_local  = NULL;
-	afr_self_heal_t *rw_sh  = NULL;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-	int i = 0;
-	int call_count = 0;
-
-	off_t offset = (long) cookie;
-
-	priv = this->private;
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-
-	call_count = sh->active_sinks;
-
-        rw_local->call_count = call_count;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"read %d bytes of data from %s, offset %"PRId64"",
-		op_ret, sh_local->loc.path, offset);
-
-	if (op_ret <= 0) {
-                sh->op_failed = 1;
-
-                sh_full_loop_return (rw_frame, this, offset);
-		return 0;
-	}
+        afr_local_t                 *loop_local = NULL;
+        afr_self_heal_t             *loop_sh    = NULL;
 
-	rw_sh->offset += op_ret;
+        loop_local = loop_frame->local;
+        loop_sh = &loop_local->self_heal;
 
-	if (sh->file_has_holes) {
-		if (iov_0filled (vector, count) == 0) {
-			/* the iter function depends on the
-			   sh->offset already being updated
-			   above
-			*/
+        sh_loop_finish (loop_sh->old_loop_frame, this);
+        loop_sh->old_loop_frame = NULL;
 
-                        sh_full_loop_return (rw_frame, this, offset);
-			goto out;
-		}
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] || !sh_local->child_up[i])
-			continue;
-
-		/* this is a sink, so write to it */
+        gf_log (this->name, GF_LOG_DEBUG, "Acquired lock for range %"PRIu64
+                " %"PRIu64, loop_sh->offset, loop_sh->block_size);
+        loop_sh->data_lock_held = _gf_true;
+        loop_sh->sh_data_algo_start (loop_frame, this);
+        return 0;
+}
 
-		STACK_WIND_COOKIE (rw_frame, sh_full_write_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->writev,
-				   sh->healing_fd, vector, count, offset,
-                                   iobref);
+static int
+sh_loop_lock_failure (call_frame_t *loop_frame, xlator_t *this)
+{
+        call_frame_t                *sh_frame = NULL;
+        afr_local_t                 *loop_local = NULL;
+        afr_self_heal_t             *loop_sh    = NULL;
+
+        loop_local = loop_frame->local;
+        loop_sh = &loop_local->self_heal;
+        sh_frame = loop_sh->sh_frame;
+
+        gf_log (this->name, GF_LOG_ERROR, "failed lock for range %"PRIu64
+                " %"PRIu64, loop_sh->offset, loop_sh->block_size);
+        sh_loop_finish (loop_sh->old_loop_frame, this);
+        loop_sh->old_loop_frame = NULL;
+        sh_loop_return (sh_frame, this, loop_frame, -1, ENOTCONN);
+        return 0;
+}
 
-		if (!--call_count)
-			break;
-	}
+static int
+sh_loop_frame_create (call_frame_t *sh_frame, xlator_t *this,
+                      call_frame_t *old_loop_frame, call_frame_t **loop_frame)
+{
+        call_frame_t                *new_loop_frame = NULL;
+        afr_local_t                 *local          = NULL;
+        afr_self_heal_t             *sh             = NULL;
+        afr_local_t                 *new_loop_local = NULL;
+        afr_self_heal_t             *new_loop_sh    = NULL;
+        afr_private_t               *priv           = NULL;
+
+        GF_ASSERT (sh_frame);
+        GF_ASSERT (loop_frame);
+
+        *loop_frame = NULL;
+        local   = sh_frame->local;
+        sh      = &local->self_heal;
+        priv    = this->private;
+
+        new_loop_frame = copy_frame (sh_frame);
+        if (!new_loop_frame)
+                goto out;
+        //We want the frame to have same lk_owner as sh_frame
+        //so that locks translator allows conflicting locks
+        new_loop_local = afr_self_heal_local_init (local, this);
+        if (!new_loop_local)
+                goto out;
+        new_loop_frame->local = new_loop_local;
 
+        new_loop_sh = &new_loop_local->self_heal;
+        new_loop_sh->sources = memdup (sh->sources,
+                                       priv->child_count * sizeof (*sh->sources));
+        if (!new_loop_sh->sources)
+                goto out;
+        new_loop_sh->write_needed = GF_CALLOC (priv->child_count,
+                                               sizeof (*new_loop_sh->write_needed),
+                                               gf_afr_mt_char);
+        if (!new_loop_sh->write_needed)
+                goto out;
+        new_loop_sh->checksum = GF_CALLOC (priv->child_count, MD5_DIGEST_LENGTH,
+                                           gf_afr_mt_uint8_t);
+        if (!new_loop_sh->checksum)
+                goto out;
+        new_loop_sh->inode      = inode_ref (sh->inode);
+        new_loop_sh->sh_data_algo_start = sh->sh_data_algo_start;
+        new_loop_sh->source = sh->source;
+        new_loop_sh->active_sinks = sh->active_sinks;
+        new_loop_sh->healing_fd = fd_ref (sh->healing_fd);
+        new_loop_sh->file_has_holes = sh->file_has_holes;
+        new_loop_sh->old_loop_frame = old_loop_frame;
+        new_loop_sh->sh_frame = sh_frame;
+        *loop_frame = new_loop_frame;
+        return 0;
 out:
-	return 0;
+        sh_destroy_frame (new_loop_frame, this);
+        return -ENOMEM;
 }
 
-
 static int
-sh_full_read_write (call_frame_t *frame, xlator_t *this, off_t offset)
+sh_loop_start (call_frame_t *sh_frame, xlator_t *this, off_t offset,
+               call_frame_t *old_loop_frame)
 {
-	afr_private_t * priv    = NULL;
-	afr_local_t * local     = NULL;
-	afr_local_t * rw_local  = NULL;
-	afr_self_heal_t *rw_sh  = NULL;
-	afr_self_heal_t *sh     = NULL;
+        call_frame_t                *new_loop_frame = NULL;
+        afr_local_t                 *local          = NULL;
+        afr_self_heal_t             *sh             = NULL;
+        afr_local_t                 *new_loop_local = NULL;
+        afr_self_heal_t             *new_loop_sh    = NULL;
+        int                         ret             = 0;
 
-        call_frame_t *rw_frame = NULL;
+        GF_ASSERT (sh_frame);
 
-        int32_t op_errno = 0;
+        local   = sh_frame->local;
+        sh      = &local->self_heal;
 
-	priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        rw_frame = copy_frame (frame);
-        if (!rw_frame)
+        ret = sh_loop_frame_create (sh_frame, this, old_loop_frame,
+                                    &new_loop_frame);
+        if (ret)
                 goto out;
-
-        ALLOC_OR_GOTO (rw_local, afr_local_t, out);
-
-        rw_frame->local = rw_local;
-	rw_sh           = &rw_local->self_heal;
-
-        rw_sh->offset       = offset;
-        rw_sh->sh_frame     = frame;
-
-	STACK_WIND_COOKIE (rw_frame, sh_full_read_cbk,
-			   (void *) (long) offset,
-			   priv->children[sh->source],
-			   priv->children[sh->source]->fops->readv,
-			   sh->healing_fd, sh->block_size,
-			   offset);
+        new_loop_local = new_loop_frame->local;
+        new_loop_sh = &new_loop_local->self_heal;
+        new_loop_sh->offset = offset;
+        new_loop_sh->block_size = sh->block_size;
+        afr_sh_data_lock (new_loop_frame, this, offset, new_loop_sh->block_size,
+                          _gf_true, this->name, sh_loop_lock_success, sh_loop_lock_failure);
         return 0;
-
 out:
-        sh->op_failed = 1;
-
-        sh_full_loop_driver (frame, this);
-
-	return 0;
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+        if (old_loop_frame)
+                sh_loop_finish (old_loop_frame, this);
+        sh_loop_return (sh_frame, this, new_loop_frame, -1, ENOMEM);
+        return 0;
 }
 
-
 static int
-sh_full_loop_driver (call_frame_t *frame, xlator_t *this)
+sh_loop_driver (call_frame_t *sh_frame, xlator_t *this,
+                gf_boolean_t is_first_call, call_frame_t *old_loop_frame)
 {
-	afr_private_t * priv = NULL;
-	afr_local_t * local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-        afr_sh_algo_full_private_t *sh_priv = NULL;
-
-        int   loop    = 0;
-        int   recurse = 0;
-
-        off_t offset  = 0;
-
-	priv    = this->private;
-	local   = frame->local;
-	sh      = &local->self_heal;
+        afr_local_t *               local          = NULL;
+        afr_self_heal_t *           sh             = NULL;
+        afr_sh_algo_private_t       *sh_priv        = NULL;
+        gf_boolean_t                is_driver_done = _gf_false;
+        blksize_t                   block_size     = 0;
+        int                         loop           = 0;
+        off_t                       offset         = 0;
+        afr_private_t               *priv          = NULL;
+
+        priv    = this->private;
+        local   = sh_frame->local;
+        sh      = &local->self_heal;
         sh_priv = sh->private;
 
-	if (sh->op_failed) {
-                if (sh_priv->loops_running == 0) {
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "full self-heal aborting on %s",
-                                local->loc.path);
-
-                        sh_full_private_cleanup (frame, this);
-                        local->self_heal.algo_abort_cbk (frame, this);
-                }
-
-		goto out;
-	}
-
-	if (sh_priv->offset >= sh->file_size) {
-                if (sh_priv->loops_running == 0) {
-
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "full self-heal completed on %s",
-                                local->loc.path);
-
-                        sh_full_private_cleanup (frame, this);
-                        local->self_heal.algo_completion_cbk (frame, this);
-                }
-
-		goto out;
-	}
-
-spawn:
-        loop    = 0;
-        recurse = 0;
-
         LOCK (&sh_priv->lock);
         {
-                if ((sh_priv->loops_running < priv->data_self_heal_window_size)
-                    && (sh_priv->offset < sh->file_size)) {
-
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "spawning a loop for offset %"PRId64,
-                                sh_priv->offset);
-
-                        offset           = sh_priv->offset;
-                        sh_priv->offset += sh->block_size;
-
+                if (!is_first_call)
+                        sh_priv->loops_running--;
+                offset = sh_priv->offset;
+                block_size = sh->block_size;
+                while ((!sh->eof_reached) &&
+                       (!is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) &&
+                     (sh_priv->loops_running < priv->data_self_heal_window_size)
+                       && (sh_priv->offset < sh->file_size)) {
+
+                        loop++;
+                        sh_priv->offset += block_size;
                         sh_priv->loops_running++;
 
-                        loop = 1;
-
-                        if (sh_priv->offset < sh->file_size)
-                                recurse = 1;
+                        if (!is_first_call)
+                                break;
+                }
+                if (0 == sh_priv->loops_running) {
+                        is_driver_done = _gf_true;
                 }
         }
         UNLOCK (&sh_priv->lock);
 
-        if (loop) {
-                sh_full_read_write (frame, this, offset);
-                if (recurse)
-                        goto spawn;
+        if (0 == loop) {
+                //loop finish does unlock, but the erasing of the pending
+                //xattrs needs to happen before that so do not finish the loop
+                if (is_driver_done &&
+                    !is_self_heal_failed (sh, AFR_CHECK_SPECIFIC))
+                        goto driver_done;
+                if (old_loop_frame) {
+                        sh_loop_finish (old_loop_frame, this);
+                        old_loop_frame = NULL;
+                }
         }
 
-out:
-	return 0;
-}
-
-
-int
-afr_sh_algo_full (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *               local   = NULL;
-        afr_self_heal_t *           sh      = NULL;
-        afr_sh_algo_full_private_t *sh_priv = NULL;
-
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        sh_priv = GF_CALLOC (1, sizeof (*sh_priv),
-                             gf_afr_mt_afr_private_t);
-
-        LOCK_INIT (&sh_priv->lock);
-
-        sh->private = sh_priv;
-
-        local->call_count = 0;
-
-        sh_full_loop_driver (frame, this);
-        return 0;
-}
-
-
-/*
- * The "diff" algorithm. Copies only those blocks whose checksums
- * don't match with those of source.
- */
-
-
-static void
-sh_diff_private_cleanup (call_frame_t *frame, xlator_t *this)
-{
-        afr_private_t *             priv    = NULL;
-        afr_local_t *               local   = NULL;
-        afr_self_heal_t *           sh      = NULL;
-        afr_sh_algo_diff_private_t *sh_priv = NULL;
-
-        int i;
-
-        priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        sh_priv = sh->private;
-
-        for (i = 0; i < priv->data_self_heal_window_size; i++) {
-                if (sh_priv->loops[i]) {
-                        if (sh_priv->loops[i]->write_needed)
-                                GF_FREE (sh_priv->loops[i]->write_needed);
-
-                        if (sh_priv->loops[i]->checksum)
-                                GF_FREE (sh_priv->loops[i]->checksum);
+        //If we have more loops to form we should finish previous loop after
+        //the next loop lock
+        while (loop--) {
+                if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+                        // op failed in other loop, stop spawning more loops
+                        if (old_loop_frame) {
+                                sh_loop_finish (old_loop_frame, this);
+                                old_loop_frame = NULL;
+                        }
+                        sh_loop_driver (sh_frame, this, _gf_false, NULL);
+                } else {
+                        gf_log (this->name, GF_LOG_TRACE, "spawning a loop "
+                                "for offset %"PRId64, offset);
 
-                        GF_FREE (sh_priv->loops[i]);
+                        sh_loop_start (sh_frame, this, offset, old_loop_frame);
+                        old_loop_frame = NULL;
+                        offset += block_size;
                 }
         }
 
-        if (sh_priv) {
-                if (sh_priv->loops)
-                        GF_FREE (sh_priv->loops);
-
-                GF_FREE (sh_priv);
+driver_done:
+        if (is_driver_done) {
+                sh_loop_driver_done (sh_frame, this, old_loop_frame);
         }
-
-
-}
-
-
-static uint32_t
-__make_cookie (int loop_index, int child_index)
-{
-        uint32_t ret = (loop_index << 16) | child_index;
-        return ret;
-}
-
-
-static int
-__loop_index (uint32_t cookie)
-{
-        return (cookie & 0xFFFF0000) >> 16;
+        return 0;
 }
 
-
 static int
-__child_index (uint32_t cookie)
+sh_loop_return (call_frame_t *sh_frame, xlator_t *this, call_frame_t *loop_frame,
+                int32_t op_ret, int32_t op_errno)
 {
-        return (cookie & 0x0000FFFF);
-}
-
+        afr_local_t *               loop_local = NULL;
+        afr_self_heal_t *           loop_sh    = NULL;
+        afr_local_t *               sh_local = NULL;
+        afr_self_heal_t            *sh       = NULL;
 
-static void
-sh_diff_loop_state_reset (struct sh_diff_loop_state *loop_state, int child_count)
-{
-        loop_state->active = _gf_false;
-//        loop_state->offset = 0;
-
-        memset (loop_state->write_needed,
-                0, sizeof (*loop_state->write_needed) * child_count);
-
-        memset (loop_state->checksum,
-                0, MD5_DIGEST_LEN * child_count);
-}
-
-
-static int
-sh_diff_number_of_writes_needed (unsigned char *write_needed, int child_count)
-{
-        int writes = 0;
-        int i;
+        sh_local = sh_frame->local;
+        sh       = &sh_local->self_heal;
 
-        for (i = 0; i < child_count; i++) {
-                if (write_needed[i])
-                        writes++;
+        if (loop_frame) {
+                loop_local = loop_frame->local;
+                if (loop_local)
+                        loop_sh    = &loop_local->self_heal;
+                if (loop_sh)
+                        gf_log (this->name, GF_LOG_TRACE, "loop for offset "
+                                "%"PRId64" returned", loop_sh->offset);
         }
 
-        return writes;
-}
-
+        if (op_ret == -1) {
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                afr_sh_set_error (sh, op_errno);
+                if (loop_frame) {
+                        sh_loop_finish (loop_frame, this);
+                        loop_frame = NULL;
+                }
+        }
 
-static int
-sh_diff_loop_driver (call_frame_t *frame, xlator_t *this);
+        sh_loop_driver (sh_frame, this, _gf_false, loop_frame);
 
+        return 0;
+}
 
 static int
-sh_diff_loop_return (call_frame_t *rw_frame, xlator_t *this,
-                     struct sh_diff_loop_state *loop_state)
+sh_loop_write_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                   struct iatt *postbuf, dict_t *xdata)
 {
-        afr_private_t *             priv       = NULL;
-        afr_local_t *               rw_local   = NULL;
-        afr_self_heal_t *           rw_sh      = NULL;
-
-        call_frame_t *sh_frame              = NULL;
-	afr_local_t * sh_local              = NULL;
-	afr_self_heal_t *sh                 = NULL;
-        afr_sh_algo_diff_private_t *sh_priv = NULL;
-
-        priv  = this->private;
-
-        rw_local = rw_frame->local;
-        rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
+        afr_private_t *             priv        = NULL;
+        afr_local_t *               loop_local    = NULL;
+        afr_self_heal_t *           loop_sh       = NULL;
+        call_frame_t               *sh_frame    = NULL;
+        afr_local_t *               sh_local    = NULL;
+        afr_self_heal_t            *sh          = NULL;
+        int                         call_count  = 0;
+        int                         child_index = 0;
+
+        priv     = this->private;
+        loop_local = loop_frame->local;
+        loop_sh    = &loop_local->self_heal;
+
+        sh_frame = loop_sh->sh_frame;
         sh_local = sh_frame->local;
         sh       = &sh_local->self_heal;
-        sh_priv  = sh->private;
+
+        child_index =  (long) cookie;
 
         gf_log (this->name, GF_LOG_TRACE,
-                "loop for offset %"PRId64" returned", loop_state->offset);
+                "wrote %d bytes of data from %s to child %d, offset %"PRId64"",
+                op_ret, sh_local->loc.path, child_index, loop_sh->offset);
 
-        LOCK (&sh_priv->lock);
-        {
-                sh_priv->loops_running--;
-                sh_diff_loop_state_reset (loop_state, priv->child_count);
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "write to %s failed on subvolume %s (%s)",
+                        sh_local->loc.path,
+                        priv->children[child_index]->name,
+                        strerror (op_errno));
+
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                afr_sh_set_error (loop_sh, op_errno);
+        } else if (op_ret < loop_local->cont.writev.vector->iov_len) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "incomplete write to %s on subvolume %s "
+                        "(expected %lu, returned %d)", sh_local->loc.path,
+                        priv->children[child_index]->name,
+                        loop_local->cont.writev.vector->iov_len, op_ret);
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
         }
-        UNLOCK (&sh_priv->lock);
 
-        AFR_STACK_DESTROY (rw_frame);
+        call_count = afr_frame_return (loop_frame);
+
+        if (call_count == 0) {
+		iobref_unref(loop_local->cont.writev.iobref);
 
-        sh_diff_loop_driver (sh_frame, this);
+                sh_loop_return (sh_frame, this, loop_frame,
+                                loop_sh->op_ret, loop_sh->op_errno);
+        }
 
         return 0;
 }
 
-
-static int
-sh_diff_write_cbk (call_frame_t *rw_frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *buf,
-                   struct iatt *postbuf)
+static void
+sh_prune_writes_needed (call_frame_t *sh_frame, call_frame_t *loop_frame,
+                        afr_private_t *priv)
 {
-	afr_private_t *   priv     = NULL;
-	afr_local_t *     rw_local = NULL;
-	afr_self_heal_t * rw_sh    = NULL;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
+        afr_local_t     *sh_local     = NULL;
+        afr_self_heal_t *sh           = NULL;
+        afr_local_t     *loop_local   = NULL;
+        afr_self_heal_t *loop_sh      = NULL;
+        int             i             = 0;
 
-        afr_sh_algo_diff_private_t *sh_priv;
-        struct sh_diff_loop_state *loop_state;
+        sh_local   = sh_frame->local;
+        sh         = &sh_local->self_heal;
 
-	int call_count  = 0;
-        int child_index = 0;
-        int loop_index  = 0;
+        if (!strcmp (sh->algo->name, "diff"))
+                return;
 
-	priv     = this->private;
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
+        loop_local = loop_frame->local;
+        loop_sh    = &loop_local->self_heal;
 
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-        sh_priv  = sh->private;
-
-        child_index = __child_index ((uint32_t) (long) cookie);
-        loop_index  = __loop_index ((uint32_t) (long) cookie);
-        loop_state  = sh_priv->loops[loop_index];
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"wrote %d bytes of data from %s to child %d, offset %"PRId64"",
-		op_ret, sh_local->loc.path, child_index,
-                loop_state->offset);
-
-	LOCK (&sh_frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"write to %s failed on subvolume %s (%s)",
-				sh_local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-
-			sh->op_failed = 1;
-		}
-	}
-	UNLOCK (&sh_frame->lock);
-
-	call_count = afr_frame_return (rw_frame);
-
-	if (call_count == 0) {
-		sh_diff_loop_return (rw_frame, this, loop_state);
-	}
-
-	return 0;
+        /* full self-heal guarantees there exists atleast 1 file with size 0
+         * That means for other files we can preserve holes that come after
+         * its size before 'trim'
+         */
+        for (i = 0; i < priv->child_count; i++) {
+                if (loop_sh->write_needed[i] &&
+                    ((loop_sh->offset + 1) > sh->buf[i].ia_size))
+                        loop_sh->write_needed[i] = 0;
+        }
 }
 
-
 static int
-sh_diff_read_cbk (call_frame_t *rw_frame, void *cookie,
+sh_loop_read_cbk (call_frame_t *loop_frame, void *cookie,
                   xlator_t *this, int32_t op_ret, int32_t op_errno,
                   struct iovec *vector, int32_t count, struct iatt *buf,
-                  struct iobref *iobref)
+                  struct iobref *iobref, dict_t *xdata)
 {
-	afr_private_t *   priv     = NULL;
-	afr_local_t *     rw_local = NULL;
-	afr_self_heal_t * rw_sh    = NULL;
-
-        afr_sh_algo_diff_private_t * sh_priv = NULL;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-        int loop_index;
-        struct sh_diff_loop_state *loop_state;
-
-        uint32_t wcookie;
-
-	int i = 0;
-	int call_count = 0;
-
-	priv     = this->private;
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
+        afr_private_t *               priv       = NULL;
+        afr_local_t *                 loop_local   = NULL;
+        afr_self_heal_t *             loop_sh      = NULL;
+        call_frame_t                 *sh_frame   = NULL;
+        int                           i          = 0;
+        int                           call_count = 0;
+        afr_local_t *                 sh_local   = NULL;
+        afr_self_heal_t *             sh      = NULL;
+
+        priv       = this->private;
+        loop_local = loop_frame->local;
+        loop_sh    = &loop_local->self_heal;
+
+        sh_frame = loop_sh->sh_frame;
         sh_local = sh_frame->local;
         sh       = &sh_local->self_heal;
-        sh_priv  = sh->private;
-
-        loop_index = __loop_index ((uint32_t) (long) cookie);
-        loop_state = sh_priv->loops[loop_index];
-
-	call_count = sh_diff_number_of_writes_needed (loop_state->write_needed,
-                                                      priv->child_count);
 
-	rw_local->call_count = call_count;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"read %d bytes of data from %s, offset %"PRId64"",
-		op_ret, sh_local->loc.path, loop_state->offset);
-
-	if ((op_ret <= 0) ||
-            (call_count == 0)) {
-                sh_diff_loop_return (rw_frame, this, loop_state);
+        gf_log (this->name, GF_LOG_TRACE,
+                "read %d bytes of data from %s, offset %"PRId64"",
+                op_ret, loop_local->loc.path, loop_sh->offset);
+
+        if (op_ret <= 0) {
+                if (op_ret < 0) {
+                        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                        gf_log (this->name, GF_LOG_ERROR, "read failed on %d "
+                                "for %s reason :%s", sh->source,
+                                sh_local->loc.path, strerror (errno));
+                } else {
+                        sh->eof_reached = _gf_true;
+                        gf_log (this->name, GF_LOG_DEBUG, "Eof reached for %s",
+                                sh_local->loc.path);
+                }
+                sh_loop_return (sh_frame, this, loop_frame, op_ret, op_errno);
+                goto out;
+        }
 
-		return 0;
-	}
+        if (loop_sh->file_has_holes && iov_0filled (vector, count) == 0)
+                sh_prune_writes_needed (sh_frame, loop_frame, priv);
 
-	if (sh->file_has_holes) {
-		if (iov_0filled (vector, count) == 0) {
+        call_count = sh_number_of_writes_needed (loop_sh->write_needed,
+                                                 priv->child_count);
+        if (call_count == 0) {
+                sh_loop_return (sh_frame, this, loop_frame, 0, 0);
+                goto out;
+        }
 
-                        sh_diff_loop_return (rw_frame, this, loop_state);
-			goto out;
-		}
-	}
+        loop_local->call_count = call_count;
 
-	for (i = 0; i < priv->child_count; i++) {
-                if (loop_state->write_needed[i]) {
-                        wcookie = __make_cookie (loop_index, i);
+	/*
+	 * We only really need the request size at the moment, but the buffer
+	 * is required if we want to issue a retry in the event of a short write.
+	 * Therefore, we duplicate the vector and ref the iobref here...
+	 */
+	loop_local->cont.writev.vector = iov_dup(vector, count);
+	loop_local->cont.writev.iobref = iobref_ref(iobref);
 
-                        STACK_WIND_COOKIE (rw_frame, sh_diff_write_cbk,
-                                           (void *) (long) wcookie,
-                                           priv->children[i],
-                                           priv->children[i]->fops->writev,
-                                           sh->healing_fd, vector, count,
-                                           loop_state->offset, iobref);
+        for (i = 0; i < priv->child_count; i++) {
+                if (!loop_sh->write_needed[i])
+                        continue;
+                STACK_WIND_COOKIE (loop_frame, sh_loop_write_cbk,
+                                   (void *) (long) i,
+                                   priv->children[i],
+                                   priv->children[i]->fops->writev,
+                                   loop_sh->healing_fd, vector, count,
+                                   loop_sh->offset, 0, iobref, NULL);
 
-                        if (!--call_count)
-                                break;
-                }
+                if (!--call_count)
+                        break;
         }
 
 out:
-	return 0;
+        return 0;
 }
 
 
 static int
-sh_diff_read (call_frame_t *rw_frame, xlator_t *this,
-              int loop_index)
+sh_loop_read (call_frame_t *loop_frame, xlator_t *this)
 {
-	afr_private_t *   priv     = NULL;
-	afr_local_t *     rw_local = NULL;
-	afr_self_heal_t * rw_sh    = NULL;
-
-        afr_sh_algo_diff_private_t * sh_priv = NULL;
-        struct sh_diff_loop_state *loop_state;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-        uint32_t cookie;
-
-	priv     = this->private;
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-        sh_priv  = sh->private;
-
-        loop_state = sh_priv->loops[loop_index];
+        afr_private_t           *priv       = NULL;
+        afr_local_t             *loop_local   = NULL;
+        afr_self_heal_t         *loop_sh      = NULL;
 
-        cookie = __make_cookie (loop_index, sh->source);
+        priv     = this->private;
+        loop_local = loop_frame->local;
+        loop_sh    = &loop_local->self_heal;
 
-	STACK_WIND_COOKIE (rw_frame, sh_diff_read_cbk,
-			   (void *) (long) cookie,
-			   priv->children[sh->source],
-			   priv->children[sh->source]->fops->readv,
-			   sh->healing_fd, sh_priv->block_size,
-			   loop_state->offset);
+        STACK_WIND_COOKIE (loop_frame, sh_loop_read_cbk,
+                           (void *) (long) loop_sh->source,
+                           priv->children[loop_sh->source],
+                           priv->children[loop_sh->source]->fops->readv,
+                           loop_sh->healing_fd, loop_sh->block_size,
+                           loop_sh->offset, 0, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 static int
-sh_diff_checksum_cbk (call_frame_t *rw_frame, void *cookie, xlator_t *this,
+sh_diff_checksum_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
                       int32_t op_ret, int32_t op_errno,
-                      uint32_t weak_checksum, uint8_t *strong_checksum)
+                      uint32_t weak_checksum, uint8_t *strong_checksum,
+                      dict_t *xdata)
 {
-	afr_private_t * priv    = NULL;
-	afr_local_t * rw_local  = NULL;
-	afr_self_heal_t *rw_sh  = NULL;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-        afr_sh_algo_diff_private_t * sh_priv = NULL;
+        afr_private_t                 *priv         = NULL;
+        afr_local_t                   *loop_local   = NULL;
+        afr_self_heal_t               *loop_sh      = NULL;
+        call_frame_t                  *sh_frame     = NULL;
+        afr_local_t                   *sh_local     = NULL;
+        afr_self_heal_t               *sh           = NULL;
+        afr_sh_algo_private_t         *sh_priv      = NULL;
+        int                           child_index  = 0;
+        int                           call_count   = 0;
+        int                           i            = 0;
+        int                           write_needed = 0;
 
-        int loop_index = 0;
-        int child_index = 0;
-        struct sh_diff_loop_state *loop_state;
-
-        int call_count   = 0;
-        int i            = 0;
-        int write_needed = 0;
-
-	priv  = this->private;
+        priv  = this->private;
 
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
+        loop_local = loop_frame->local;
+        loop_sh    = &loop_local->self_heal;
 
-        sh_frame = rw_sh->sh_frame;
+        sh_frame = loop_sh->sh_frame;
         sh_local = sh_frame->local;
         sh       = &sh_local->self_heal;
 
         sh_priv = sh->private;
 
-        child_index = __child_index ((uint32_t) (long) cookie);
-        loop_index  = __loop_index ((uint32_t) (long) cookie);
-
-        loop_state  = sh_priv->loops[loop_index];
+        child_index = (long) cookie;
 
         if (op_ret < 0) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "checksum on %s failed on subvolume %s (%s)",
                         sh_local->loc.path, priv->children[child_index]->name,
                         strerror (op_errno));
-
-                sh->op_failed = 1;
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
         } else {
-                memcpy (loop_state->checksum + child_index * MD5_DIGEST_LEN,
-                        strong_checksum,
-                        MD5_DIGEST_LEN);
+                memcpy (loop_sh->checksum + child_index * MD5_DIGEST_LENGTH,
+                        strong_checksum, MD5_DIGEST_LENGTH);
         }
 
-        call_count = afr_frame_return (rw_frame);
+        call_count = afr_frame_return (loop_frame);
 
         if (call_count == 0) {
                 for (i = 0; i < priv->child_count; i++) {
                         if (sh->sources[i] || !sh_local->child_up[i])
                                 continue;
 
-                        if (memcmp (loop_state->checksum + (i * MD5_DIGEST_LEN),
-                                    loop_state->checksum + (sh->source * MD5_DIGEST_LEN),
-                                    MD5_DIGEST_LEN)) {
+                        if (memcmp (loop_sh->checksum + (i * MD5_DIGEST_LENGTH),
+                                    loop_sh->checksum + (sh->source * MD5_DIGEST_LENGTH),
+                                    MD5_DIGEST_LENGTH)) {
                                 /*
-                                   Checksums differ, so this block
-                                   must be written to this sink
+                                  Checksums differ, so this block
+                                  must be written to this sink
                                 */
 
-                                gf_log (this->name, GF_LOG_TRACE,
+                                gf_log (this->name, GF_LOG_DEBUG,
                                         "checksum on subvolume %s at offset %"
                                         PRId64" differs from that on source",
-                                        priv->children[i]->name, loop_state->offset);
+                                        priv->children[i]->name, loop_sh->offset);
 
-                                write_needed = loop_state->write_needed[i] = 1;
+                                write_needed = loop_sh->write_needed[i] = 1;
                         }
                 }
 
@@ -800,270 +664,171 @@ sh_diff_checksum_cbk (call_frame_t *rw_frame, void *cookie, xlator_t *this,
                 }
                 UNLOCK (&sh_priv->lock);
 
-                if (write_needed && !sh->op_failed) {
-                        sh_diff_read (rw_frame, this, loop_index);
+                if (write_needed &&
+                    !is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+                        sh_loop_read (loop_frame, this);
                 } else {
-                        sh->offset += sh_priv->block_size;
-
-                        sh_diff_loop_return (rw_frame, this, loop_state);
+                        sh_loop_return (sh_frame, this, loop_frame,
+                                        op_ret, op_errno);
                 }
         }
 
         return 0;
 }
 
-
-static int
-sh_diff_find_unused_loop (afr_sh_algo_diff_private_t *sh_priv, int max)
-{
-        int i;
-
-        LOCK (&sh_priv->lock);
-        {
-                for (i = 0; i < max; i++) {
-                        if (sh_priv->loops[i]->active == _gf_false) {
-                                sh_priv->loops[i]->active = _gf_true;
-                                break;
-                        }
-                }
-        }
-        UNLOCK (&sh_priv->lock);
-
-        if (i == max) {
-                gf_log ("[sh-diff]", GF_LOG_ERROR,
-                        "no free loops found! This shouldn't happen. Please"
-                        " report this to gluster-devel@nongnu.org");
-        }
-
-        return i;
-}
-
-
 static int
-sh_diff_checksum (call_frame_t *frame, xlator_t *this, off_t offset)
+sh_diff_checksum (call_frame_t *loop_frame, xlator_t *this)
 {
-	afr_private_t *   priv     = NULL;
-	afr_local_t *     local    = NULL;
-	afr_local_t *     rw_local = NULL;
-	afr_self_heal_t * sh       = NULL;
-	afr_self_heal_t * rw_sh    = NULL;
-
-        afr_sh_algo_diff_private_t * sh_priv = NULL;
-
-        call_frame_t *rw_frame = NULL;
-
-        uint32_t cookie;
-        int loop_index = 0;
-        struct sh_diff_loop_state *loop_state = NULL;
-
-        int32_t op_errno = 0;
-
-        int call_count = 0;
-        int i          = 0;
-
-	priv    = this->private;
-	local   = frame->local;
-	sh      = &local->self_heal;
+        afr_private_t           *priv         = NULL;
+        afr_local_t             *loop_local   = NULL;
+        afr_self_heal_t         *loop_sh      = NULL;
+        int                     call_count    = 0;
+        int                     i             = 0;
 
-        sh_priv = sh->private;
-
-        rw_frame = copy_frame (frame);
-        if (!rw_frame)
-                goto out;
-
-        ALLOC_OR_GOTO (rw_local, afr_local_t, out);
-
-        rw_frame->local = rw_local;
-	rw_sh           = &rw_local->self_heal;
-
-        rw_sh->offset       = sh->offset;
-        rw_sh->sh_frame     = frame;
+        priv         = this->private;
+        loop_local   = loop_frame->local;
+        loop_sh      = &loop_local->self_heal;
 
-        call_count = sh->active_sinks + 1;  /* sinks and source */
+        call_count = loop_sh->active_sinks + 1;  /* sinks and source */
 
-        rw_local->call_count = call_count;
+        loop_local->call_count = call_count;
 
-        loop_index = sh_diff_find_unused_loop (sh_priv, priv->data_self_heal_window_size);
-
-        loop_state = sh_priv->loops[loop_index];
-        loop_state->offset       = offset;
-
-        /* we need to send both the loop index and child index,
-           so squeeze them both into a 32-bit number */
-
-        cookie = __make_cookie (loop_index, sh->source);
-
-        STACK_WIND_COOKIE (rw_frame, sh_diff_checksum_cbk,
-                           (void *) (long) cookie,
-                           priv->children[sh->source],
-                           priv->children[sh->source]->fops->rchecksum,
-                           sh->healing_fd,
-                           offset, sh_priv->block_size);
+        STACK_WIND_COOKIE (loop_frame, sh_diff_checksum_cbk,
+                           (void *) (long) loop_sh->source,
+                           priv->children[loop_sh->source],
+                           priv->children[loop_sh->source]->fops->rchecksum,
+                           loop_sh->healing_fd,
+                           loop_sh->offset, loop_sh->block_size, NULL);
 
         for (i = 0; i < priv->child_count; i++) {
-                if (sh->sources[i] || !local->child_up[i])
+                if (loop_sh->sources[i] || !loop_local->child_up[i])
                         continue;
 
-                cookie = __make_cookie (loop_index, i);
-
-                STACK_WIND_COOKIE (rw_frame, sh_diff_checksum_cbk,
-                                   (void *) (long) cookie,
+                STACK_WIND_COOKIE (loop_frame, sh_diff_checksum_cbk,
+                                   (void *) (long) i,
                                    priv->children[i],
                                    priv->children[i]->fops->rchecksum,
-                                   sh->healing_fd,
-                                   offset, sh_priv->block_size);
+                                   loop_sh->healing_fd,
+                                   loop_sh->offset, loop_sh->block_size, NULL);
 
                 if (!--call_count)
                         break;
         }
 
         return 0;
-
-out:
-        sh->op_failed = 1;
-
-        sh_diff_loop_driver (frame, this);
-
-        return 0;
 }
 
-
 static int
-sh_diff_loop_driver (call_frame_t *frame, xlator_t *this)
+sh_full_read_write_to_sinks (call_frame_t *loop_frame, xlator_t *this)
 {
-	afr_private_t *   priv  = NULL;
-	afr_local_t *     local = NULL;
-	afr_self_heal_t * sh    = NULL;
-        afr_sh_algo_diff_private_t *sh_priv = NULL;
-
-        int   loop    = 0;
-        int   recurse = 0;
-
-        off_t offset = 0;
-        char  sh_type_str[256] = {0,};
- 
-	priv    = this->private;
-	local   = frame->local;
-	sh      = &local->self_heal;
-        sh_priv = sh->private;
-
-        afr_self_heal_type_str_get(sh, sh_type_str, sizeof(sh_type_str));
-
-	if (sh->op_failed) {
-                if (sh_priv->loops_running == 0) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "diff %s self-heal aborting on %s",
-                                sh_type_str, local->loc.path);
-
-                        sh_diff_private_cleanup (frame, this);
-                        local->self_heal.algo_abort_cbk (frame, this);
-                }
-
-		goto out;
-	}
-
-	if (sh_priv->offset >= sh->file_size) {
-                if (sh_priv->loops_running == 0) {
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "diff %s self-heal completed on %s",
-                                sh_type_str, local->loc.path);
+        afr_private_t           *priv         = NULL;
+        afr_local_t             *loop_local   = NULL;
+        afr_self_heal_t         *loop_sh      = NULL;
+        int                     i             = 0;
 
+        priv         = this->private;
+        loop_local   = loop_frame->local;
+        loop_sh      = &loop_local->self_heal;
 
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "diff %s self-heal on %s: %d blocks of %d were different (%.2f%%)",
-                                sh_type_str, local->loc.path,
-                                sh_priv->diff_blocks, sh_priv->total_blocks,
-                                ((sh_priv->diff_blocks * 1.0)/sh_priv->total_blocks) * 100);
-
-                        sh_diff_private_cleanup (frame, this);
-                        local->self_heal.algo_completion_cbk (frame, this);
-                }
-
-		goto out;
-	}
-
-spawn:
-        loop    = 0;
-        recurse = 0;
-
-        LOCK (&sh_priv->lock);
-        {
-                if ((sh_priv->loops_running < priv->data_self_heal_window_size)
-                    && (sh_priv->offset < sh->file_size)) {
-
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "spawning a loop for offset %"PRId64,
-                                sh_priv->offset);
-
-                        offset           = sh_priv->offset;
-                        sh_priv->offset += sh_priv->block_size;
-
-                        sh_priv->loops_running++;
-
-                        loop = 1;
-
-                        if (sh_priv->offset < sh->file_size)
-                                recurse = 1;
-                }
-        }
-        UNLOCK (&sh_priv->lock);
-
-        if (loop) {
-                sh_diff_checksum (frame, this, offset);
-                if (recurse)
-                    goto spawn;
+        for (i = 0; i < priv->child_count; i++) {
+                if (loop_sh->sources[i] || !loop_local->child_up[i])
+                        continue;
+                loop_sh->write_needed[i] = 1;
         }
-
-out:
+        sh_loop_read (loop_frame, this);
         return 0;
 }
 
-
-int
-afr_sh_algo_diff (call_frame_t *frame, xlator_t *this)
+afr_sh_algo_private_t*
+afr_sh_priv_init ()
 {
-        afr_private_t *             priv    = NULL;
-        afr_local_t *               local   = NULL;
-        afr_self_heal_t *           sh      = NULL;
-        afr_sh_algo_diff_private_t *sh_priv = NULL;
-
-        int i;
-
-        priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
+        afr_sh_algo_private_t   *sh_priv = NULL;
 
         sh_priv = GF_CALLOC (1, sizeof (*sh_priv),
                              gf_afr_mt_afr_private_t);
-
-        sh_priv->block_size = this->ctx->page_size;
-
-        sh->private = sh_priv;
+        if (!sh_priv)
+                goto out;
 
         LOCK_INIT (&sh_priv->lock);
+out:
+        return sh_priv;
+}
 
-        local->call_count = 0;
+int
+afr_sh_transfer_lock (call_frame_t *dst, call_frame_t *src, char *dom,
+                      unsigned int child_count)
+{
+        afr_local_t             *dst_local   = NULL;
+        afr_self_heal_t         *dst_sh      = NULL;
+        afr_local_t             *src_local   = NULL;
+        afr_self_heal_t         *src_sh      = NULL;
+        int                     ret          = -1;
+
+        dst_local = dst->local;
+        dst_sh = &dst_local->self_heal;
+        src_local = src->local;
+        src_sh = &src_local->self_heal;
+        GF_ASSERT (src_sh->data_lock_held);
+        GF_ASSERT (!dst_sh->data_lock_held);
+        ret = afr_lk_transfer_datalock (dst, src, dom, child_count);
+        if (ret)
+                return ret;
+        src_sh->data_lock_held = _gf_false;
+        dst_sh->data_lock_held = _gf_true;
+        return 0;
+}
 
-        sh_priv->loops = GF_CALLOC (priv->data_self_heal_window_size,
-                                    sizeof (*sh_priv->loops),
-                                    gf_afr_mt_sh_diff_loop_state);
+int
+afr_sh_start_loops (call_frame_t *sh_frame, xlator_t *this,
+                    afr_sh_algo_fn sh_data_algo_start)
+{
+        call_frame_t            *first_loop_frame = NULL;
+        afr_local_t             *local   = NULL;
+        afr_self_heal_t         *sh      = NULL;
+        int                     ret      = 0;
+        afr_private_t           *priv    = NULL;
 
-        for (i = 0; i < priv->data_self_heal_window_size; i++) {
-                sh_priv->loops[i]               = GF_CALLOC (1, sizeof (*sh_priv->loops[i]),
-                                                             gf_afr_mt_sh_diff_loop_state);
+        local = sh_frame->local;
+        sh    = &local->self_heal;
+        priv  = this->private;
 
-                sh_priv->loops[i]->checksum     = GF_CALLOC (priv->child_count,
-                                                             MD5_DIGEST_LEN, gf_afr_mt_uint8_t);
-                sh_priv->loops[i]->write_needed = GF_CALLOC (priv->child_count,
-                                                             sizeof (*sh_priv->loops[i]->write_needed),
-                                                             gf_afr_mt_char);
+        sh->sh_data_algo_start = sh_data_algo_start;
+        local->call_count = 0;
+        ret = sh_loop_frame_create (sh_frame, this, NULL, &first_loop_frame);
+        if (ret)
+                goto out;
+        ret = afr_sh_transfer_lock (first_loop_frame, sh_frame, this->name,
+                                    priv->child_count);
+        if (ret)
+                goto out;
+        sh->private = afr_sh_priv_init ();
+        if (!sh->private) {
+                ret = -1;
+                goto out;
         }
+        sh_loop_driver (sh_frame, this, _gf_true, first_loop_frame);
+        ret = 0;
+out:
+        if (ret) {
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                sh_loop_driver_done (sh_frame, this, NULL);
+        }
+        return 0;
+}
 
-        sh_diff_loop_driver (frame, this);
-
+int
+afr_sh_algo_diff (call_frame_t *sh_frame, xlator_t *this)
+{
+        afr_sh_start_loops (sh_frame, this, sh_diff_checksum);
         return 0;
 }
 
+int
+afr_sh_algo_full (call_frame_t *sh_frame, xlator_t *this)
+{
+        afr_sh_start_loops (sh_frame, this, sh_full_read_write_to_sinks);
+        return 0;
+}
 
 struct afr_sh_algorithm afr_self_heal_algorithms[] = {
         {.name = "full",  .fn = afr_sh_algo_full},
diff --git a/xlators/cluster/afr/src/afr-self-heal-algorithm.h b/xlators/cluster/afr/src/afr-self-heal-algorithm.h
index 0bdae3aa7..6b20789b1 100644
--- a/xlators/cluster/afr/src/afr-self-heal-algorithm.h
+++ b/xlators/cluster/afr/src/afr-self-heal-algorithm.h
@@ -1,26 +1,16 @@
 /*
-   Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __AFR_SELF_HEAL_ALGORITHM_H__
 #define __AFR_SELF_HEAL_ALGORITHM_H__
 
-
 typedef int (*afr_sh_algo_fn) (call_frame_t *frame,
                                xlator_t *this);
 
@@ -30,31 +20,13 @@ struct afr_sh_algorithm {
 };
 
 extern struct afr_sh_algorithm afr_self_heal_algorithms[3];
-
 typedef struct {
         gf_lock_t lock;
         unsigned int loops_running;
         off_t offset;
-} afr_sh_algo_full_private_t;
-
-struct sh_diff_loop_state {
-        off_t   offset;
-        unsigned char *write_needed;
-        uint8_t *checksum;
-        gf_boolean_t active;
-};
-
-typedef struct {
-        size_t block_size;
-
-        gf_lock_t lock;
-        unsigned int loops_running;
-        off_t offset;
 
         int32_t total_blocks;
         int32_t diff_blocks;
-
-        struct sh_diff_loop_state **loops;
-} afr_sh_algo_diff_private_t;
+} afr_sh_algo_private_t;
 
 #endif /* __AFR_SELF_HEAL_ALGORITHM_H__ */
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index 2e677d775..ef92b4205 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include "glusterfs.h"
@@ -27,6 +18,67 @@
 #include "afr-self-heal.h"
 #include "pump.h"
 
+#define ADD_FMT_STRING(msg, off, sh_str, status, print_log)                 \
+        do {                                                                \
+                if (AFR_SELF_HEAL_NOT_ATTEMPTED != status) {                \
+                        off += snprintf (msg + off, sizeof (msg) - off,     \
+                                         " "sh_str" self heal %s,",         \
+                                         get_sh_completion_status (status));\
+                        print_log = 1;                                      \
+                }                                                           \
+        } while (0)
+
+#define ADD_FMT_STRING_SYNC(msg, off, sh_str, status, print_log)            \
+        do {                                                                \
+                if (AFR_SELF_HEAL_SYNC_BEGIN == status ||                   \
+                    AFR_SELF_HEAL_FAILED == status)  {                      \
+                        off += snprintf (msg + off, sizeof (msg) - off,     \
+                                         " "sh_str" self heal %s,",         \
+                                         get_sh_completion_status (status));\
+                        print_log = 1;                                      \
+                }                                                           \
+        } while (0)
+
+
+void
+afr_sh_reset (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
+
+        memset (sh->child_errno, 0,
+                sizeof (*sh->child_errno) * priv->child_count);
+        memset (sh->buf, 0, sizeof (*sh->buf) * priv->child_count);
+        memset (sh->parentbufs, 0,
+                sizeof (*sh->parentbufs) * priv->child_count);
+        memset (sh->success, 0, sizeof (*sh->success) * priv->child_count);
+        memset (sh->locked_nodes, 0,
+                sizeof (*sh->locked_nodes) * priv->child_count);
+        sh->active_sinks = 0;
+
+        afr_reset_xattr (sh->xattr, priv->child_count);
+}
+
+//Intersection[child]=1 if child is part of intersection
+void
+afr_children_intersection_get (int32_t *set1, int32_t *set2,
+                               int *intersection, unsigned int child_count)
+{
+        int                      i = 0;
+
+        memset (intersection, 0, sizeof (*intersection) * child_count);
+        for (i = 0; i < child_count; i++) {
+                intersection[i] = afr_is_child_present (set1, child_count, i)
+                                     && afr_is_child_present (set2, child_count,
+                                                              i);
+        }
+}
+
 /**
  * select_source - select a source and return it
  */
@@ -34,118 +86,215 @@
 int
 afr_sh_select_source (int sources[], int child_count)
 {
-	int i;
-	for (i = 0; i < child_count; i++)
-		if (sources[i])
-			return i;
+        int i = 0;
+        for (i = 0; i < child_count; i++)
+                if (sources[i])
+                        return i;
 
-	return -1;
+        return -1;
 }
 
+void
+afr_sh_mark_source_sinks (call_frame_t *frame, xlator_t *this)
+{
+        int              i = 0;
+        afr_local_t     *local      = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              active_sinks = 0;
 
-/**
- * sink_count - return number of sinks in sources array
- */
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-int
-afr_sh_sink_count (int sources[], int child_count)
-{
-	int i;
-	int sinks = 0;
-	for (i = 0; i < child_count; i++)
-		if (!sources[i])
-			sinks++;
-	return sinks;
+        for (i = 0; i < priv->child_count; i++) {
+                if (sh->sources[i] == 0 && local->child_up[i] == 1) {
+                        active_sinks++;
+                        sh->success[i] = 1;
+                } else if (sh->sources[i] == 1 && local->child_up[i] == 1) {
+                        sh->success[i] = 1;
+                }
+        }
+        sh->active_sinks = active_sinks;
 }
 
 int
 afr_sh_source_count (int sources[], int child_count)
 {
-	int i;
-	int nsource = 0;
+        int i = 0;
+        int nsource = 0;
 
-	for (i = 0; i < child_count; i++)
-		if (sources[i])
-			nsource++;
-	return nsource;
+        for (i = 0; i < child_count; i++)
+                if (sources[i])
+                        nsource++;
+        return nsource;
 }
 
+void
+afr_sh_set_error (afr_self_heal_t *sh, int32_t op_errno)
+{
+        sh->op_ret = -1;
+	sh->op_errno = afr_most_important_error(sh->op_errno, op_errno,
+						_gf_false);
+}
 
-int
-afr_sh_supress_errenous_children (int sources[], int child_errno[],
-				  int child_count)
+void
+afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this)
 {
-	int i = 0;
+        afr_private_t *  priv = this->private;
+        char            *buf  = NULL;
+        char            *ptr  = NULL;
+        int              i    = 0;
+        int              j    = 0;
 
-	for (i = 0; i < child_count; i++) {
-		if (child_errno[i] && sources[i]) {
-			sources[i] = 0;
-		}
-	}
+        /* 10 digits per entry + 1 space + '[' and ']' */
+        buf = GF_MALLOC (priv->child_count * 11 + 8, gf_afr_mt_char);
 
-	return 0;
+        for (i = 0; i < priv->child_count; i++) {
+                ptr = buf;
+                ptr += sprintf (ptr, "[ ");
+                for (j = 0; j < priv->child_count; j++) {
+                        ptr += sprintf (ptr, "%d ", pending_matrix[i][j]);
+                }
+                sprintf (ptr, "]");
+                gf_log (this->name, GF_LOG_DEBUG, "pending_matrix: %s", buf);
+        }
+
+        GF_FREE (buf);
 }
 
+char*
+afr_get_pending_matrix_str (int32_t *pending_matrix[], xlator_t *this)
+{
+        afr_private_t *  priv = this->private;
+        char            *buf  = NULL;
+        char            *ptr  = NULL;
+        int              i    = 0;
+        int              j    = 0;
+        int             child_count = priv->child_count;
+        char            *matrix_begin = "[ [ ";
+        char            *matrix_end = "] ]";
+        char            *seperator = "] [ ";
+        int             pending_entry_strlen = 12; //Including space after entry
+        int             matrix_begin_strlen = 0;
+        int             matrix_end_strlen = 0;
+        int             seperator_strlen = 0;
+        int             string_length = 0;
+        char            *msg = "- Pending matrix:  ";
+
+        /*
+         *  for a list of lists of [ [ a b ] [ c d ] ]
+         * */
+
+        matrix_begin_strlen = strlen (matrix_begin);
+        matrix_end_strlen = strlen (matrix_end);
+        seperator_strlen = strlen (seperator);
+        string_length = matrix_begin_strlen + matrix_end_strlen
+                        + (child_count -1) * seperator_strlen
+                        + (child_count * child_count * pending_entry_strlen);
+
+        buf = GF_CALLOC (1, 1 + strlen (msg) + string_length , gf_afr_mt_char);
+        if (!buf)
+                goto out;
+
+        ptr = buf;
+        ptr += sprintf (ptr, "%s", msg);
+        ptr += sprintf (ptr, "%s", matrix_begin);
+        for (i = 0; i < priv->child_count; i++) {
+                for (j = 0; j < priv->child_count; j++) {
+                        ptr += sprintf (ptr, "%d ", pending_matrix[i][j]);
+                }
+                if (i < priv->child_count -1)
+                        ptr += sprintf (ptr, "%s", seperator);
+        }
+
+        ptr += sprintf (ptr, "%s", matrix_end);
+
+out:
+        return buf;
+}
 
 void
-afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this)
+afr_sh_print_split_brain_log (int32_t *pending_matrix[], xlator_t *this,
+                              const char *loc)
 {
-	afr_private_t * priv = this->private;
+        char *buf      = NULL;
+        char *free_ptr = NULL;
 
-	char *buf = NULL;
-	char *ptr = NULL;
+        buf = afr_get_pending_matrix_str (pending_matrix, this);
+        if (buf)
+                free_ptr = buf;
+        else
+                buf = "";
 
-	int i, j;
 
-        /* 10 digits per entry + 1 space + '[' and ']' */
-	buf = GF_MALLOC (priv->child_count * 11 + 8, gf_afr_mt_char);
+        gf_log (this->name, GF_LOG_ERROR, "Unable to self-heal contents of '%s'"
+                " (possible split-brain). Please delete the file from all but "
+                "the preferred subvolume.%s", loc, buf);
+        GF_FREE (free_ptr);
+        return;
+}
 
-	for (i = 0; i < priv->child_count; i++) {
-		ptr = buf;
-		ptr += sprintf (ptr, "[ ");
-		for (j = 0; j < priv->child_count; j++) {
-			ptr += sprintf (ptr, "%d ", pending_matrix[i][j]);
-		}
-		sprintf (ptr, "]");
-		gf_log (this->name, GF_LOG_TRACE,
-			"pending_matrix: %s", buf);
-	}
 
-	GF_FREE (buf);
-}
+void
+afr_init_pending_matrix (int32_t **pending_matrix, size_t child_count)
+{
+        int             i   = 0;
+        int             j   = 0;
+
+        GF_ASSERT (pending_matrix);
 
+        for (i = 0; i < child_count; i++) {
+                for (j = 0; j < child_count; j++) {
+                        pending_matrix[i][j] = 0;
+                }
+        }
+}
 
 void
-afr_sh_build_pending_matrix (afr_private_t *priv,
-                             int32_t *pending_matrix[], dict_t *xattr[],
-			     int child_count, afr_transaction_type type)
+afr_mark_ignorant_subvols_as_pending (int32_t **pending_matrix,
+                                      unsigned char *ignorant_subvols,
+                                      size_t  child_count)
 {
-	int i, j, k;
+        int            i                = 0;
+        int            j                = 0;
 
-	/* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-	int32_t pending[3];
-	void *pending_raw = NULL;
-	int ret = -1;
+        GF_ASSERT (pending_matrix);
+        GF_ASSERT (ignorant_subvols);
 
-        unsigned char *ignorant_subvols = NULL;
+        for (i = 0; i < child_count; i++) {
+                if (ignorant_subvols[i]) {
+                        for (j = 0; j < child_count; j++) {
+                                if (!ignorant_subvols[j])
+                                        pending_matrix[j][i] += 1;
+                        }
+                }
+        }
+}
 
-        ignorant_subvols = GF_CALLOC (sizeof (*ignorant_subvols), child_count,
-                                      gf_afr_mt_char);
+int
+afr_build_pending_matrix (char **pending_key, int32_t **pending_matrix,
+                          unsigned char *ignorant_subvols,
+                          dict_t *xattr[], afr_transaction_type type,
+                          size_t child_count)
+{
+        /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
+        int32_t        pending[3]       = {0,};
+        void          *pending_raw      = NULL;
+        int            ret              = -1;
+        int            i                = 0;
+        int            j                = 0;
+        int            k                = 0;
 
-	/* start clean */
-	for (i = 0; i < child_count; i++) {
-		for (j = 0; j < child_count; j++) {
-			pending_matrix[i][j] = 0;
-		}
-	}
+        afr_init_pending_matrix (pending_matrix, child_count);
 
-	for (i = 0; i < child_count; i++) {
-		pending_raw = NULL;
+        for (i = 0; i < child_count; i++) {
+                pending_raw = NULL;
 
                 for (j = 0; j < child_count; j++) {
-                        ret = dict_get_ptr (xattr[i], priv->pending_key[j],
+                        ret = dict_get_ptr (xattr[i], pending_key[j],
                                             &pending_raw);
-                        
+
                         if (ret != 0) {
                                 /*
                                  * There is no xattr present. This means this
@@ -153,63 +302,26 @@ afr_sh_build_pending_matrix (afr_private_t *priv,
                                  * subvolume.
                                  */
 
-                                ignorant_subvols[i] = 1;
+                                if (ignorant_subvols)
+                                        ignorant_subvols[i] = 1;
                                 continue;
                         }
 
-			memcpy (pending, pending_raw, sizeof(pending));
+                        memcpy (pending, pending_raw, sizeof(pending));
                         k = afr_index_for_transaction_type (type);
-                        
-                        pending_matrix[i][j] = ntoh32 (pending[k]);
-                }
-	}
 
-        /*
-         * Make all non-ignorant subvols point towards the ignorant
-         * subvolumes.
-         */
-
-        for (i = 0; i < child_count; i++) {
-                if (ignorant_subvols[i]) {
-                        for (j = 0; j < child_count; j++) {
-                                if (!ignorant_subvols[j])
-                                        pending_matrix[j][i] += 1;
-                        }
+                        pending_matrix[i][j] = ntoh32 (pending[k]);
                 }
         }
 
-        GF_FREE (ignorant_subvols);
+        return ret;
 }
 
-
-/**
- * mark_sources: Mark all 'source' nodes and return number of source
- * nodes found
- *
- * A node (a row in the pending matrix) belongs to one of
- * three categories:
- *
- * M is the pending matrix.
- *
- * 'innocent' - M[i] is all zeroes
- * 'fool'     - M[i] has i'th element = 1 (self-reference)
- * 'wise'     - M[i] has i'th element = 0, others are 1 or 0.
- *
- * All 'innocent' nodes are sinks. If all nodes are innocent, no self-heal is
- * needed.
- *
- * A 'wise' node can be a source. If two 'wise' nodes conflict, it is 
- * a split-brain. If one wise node refers to the other but the other doesn't
- * refer back, the referrer is a source.
- *
- * All fools are sinks, unless there are no 'wise' nodes. In that case,
- * one of the fools is made a source.
- */
-
 typedef enum {
+        AFR_NODE_INVALID,
         AFR_NODE_INNOCENT,
         AFR_NODE_FOOL,
-        AFR_NODE_WISE
+        AFR_NODE_WISE,
 } afr_node_type;
 
 typedef struct {
@@ -250,7 +362,7 @@ afr_sh_is_wise (int32_t *array, int i, int child_count)
 
 
 static int
-afr_sh_all_nodes_innocent (afr_node_character *characters, 
+afr_sh_all_nodes_innocent (afr_node_character *characters,
                            int child_count)
 {
         int i   = 0;
@@ -286,10 +398,10 @@ afr_sh_wise_nodes_exist (afr_node_character *characters, int child_count)
 
 /*
  * The 'wisdom' of a wise node is 0 if any other wise node accuses it.
- * It is 1 if no other wise node accuses it. 
+ * It is 1 if no other wise node accuses it.
  * Only wise nodes with wisdom 1 are sources.
  *
- * If no nodes with wisdom 1 exist, a split-brain has occured.
+ * If no nodes with wisdom 1 exist, a split-brain has occurred.
  */
 
 static void
@@ -306,7 +418,7 @@ afr_sh_compute_wisdom (int32_t *pending_matrix[],
                         for (j = 0; j < child_count; j++) {
                                 if ((characters[j].type == AFR_NODE_WISE)
                                     && pending_matrix[j][i]) {
-                                        
+
                                         characters[i].wisdom = 0;
                                 }
                         }
@@ -316,7 +428,7 @@ afr_sh_compute_wisdom (int32_t *pending_matrix[],
 
 
 static int
-afr_sh_wise_nodes_conflict (afr_node_character *characters, 
+afr_sh_wise_nodes_conflict (afr_node_character *characters,
                             int child_count)
 {
         int i   = 0;
@@ -337,13 +449,12 @@ afr_sh_wise_nodes_conflict (afr_node_character *characters,
 
 
 static int
-afr_sh_mark_wisest_as_sources (int sources[], 
-                               afr_node_character *characters, 
+afr_sh_mark_wisest_as_sources (int sources[],
+                               afr_node_character *characters,
                                int child_count)
 {
         int nsources = 0;
-        
-        int i = 0;
+        int i        = 0;
 
         for (i = 0; i < child_count; i++) {
                 if (characters[i].wisdom == 1) {
@@ -355,1165 +466,1778 @@ afr_sh_mark_wisest_as_sources (int sources[],
         return nsources;
 }
 
-
-static int
-afr_sh_mark_if_size_differs (afr_self_heal_t *sh, int child_count)
+static void
+afr_compute_witness_of_fools (int32_t *witnesses, int32_t **pending_matrix,
+                              afr_node_character *characters,
+                              int32_t child_count)
 {
-        int32_t ** pending_matrix;
-        int i, j;
-
-        int size_differs = 0;
+        int i       = 0;
+        int j       = 0;
+        int witness = 0;
 
-        pending_matrix = sh->pending_matrix;
+        GF_ASSERT (witnesses);
+        GF_ASSERT (pending_matrix);
+        GF_ASSERT (characters);
+        GF_ASSERT (child_count > 0);
 
         for (i = 0; i < child_count; i++) {
+                if (characters[i].type != AFR_NODE_FOOL)
+                        continue;
+
+                witness = 0;
                 for (j = 0; j < child_count; j++) {
-                        if (!sh->buf)
-                                break;
+                        if (i == j)
+                                continue;
+                        witness += pending_matrix[i][j];
+                }
+                witnesses[i] = witness;
+        }
+}
 
-                        if (SIZE_DIFFERS (&sh->buf[i], &sh->buf[j])
-                            && (pending_matrix[i][j] == 0)
-                            && (pending_matrix[j][i] == 0)) {
+static int32_t
+afr_find_biggest_witness_among_fools (int32_t *witnesses,
+                                      afr_node_character *characters,
+                                      int32_t child_count)
+{
+        int i               = 0;
+        int biggest_witness = -1;
+        int biggest_witness_idx = -1;
+        int biggest_witness_cnt = -1;
 
-                                pending_matrix[i][j] = 1;
-                                pending_matrix[j][i] = 1;
+        GF_ASSERT (witnesses);
+        GF_ASSERT (characters);
+        GF_ASSERT (child_count > 0);
 
-                                size_differs = 1;
-                        }
-                }
+        for (i = 0; i < child_count; i++) {
+                if (characters[i].type != AFR_NODE_FOOL)
+                        continue;
+
+                if (biggest_witness < witnesses[i]) {
+                        biggest_witness = witnesses[i];
+			biggest_witness_idx = i;
+			biggest_witness_cnt = 1;
+			continue;
+		}
+
+		if (biggest_witness == witnesses[i])
+			biggest_witness_cnt++;
         }
 
-        return size_differs;
+	if (biggest_witness_cnt != 1)
+		return -1;
+
+        return biggest_witness_idx;
 }
 
-        
-static int
-afr_sh_mark_biggest_fool_as_source (afr_self_heal_t *sh,
-                                    afr_node_character *characters, 
-                                    int child_count)
+int
+afr_mark_fool_as_source_by_witness (int32_t *sources, int32_t *witnesses,
+                                    afr_node_character *characters,
+                                    int32_t child_count, int32_t witness)
 {
-        int i = 0;
-        int biggest = 0;
+        int i        = 0;
+        int nsources = 0;
 
-        for (i = 0; i < child_count; i++) {
-                if (characters[i].type == AFR_NODE_FOOL) {
-                        biggest = i;
-                        break;
-                }
-        }
+        GF_ASSERT (sources);
+        GF_ASSERT (witnesses);
+        GF_ASSERT (characters);
+        GF_ASSERT (child_count > 0);
 
         for (i = 0; i < child_count; i++) {
                 if (characters[i].type != AFR_NODE_FOOL)
                         continue;
 
-                if (!sh->buf)
-                        break;
-
-                if (SIZE_GREATER (&sh->buf[i], &sh->buf[biggest])) {
-                        biggest = i;
+                if (witness == witnesses[i]) {
+                        sources[i] = 1;
+                        nsources++;
                 }
         }
+        return nsources;
+}
 
-        sh->sources[biggest] = 1;
 
-        return 1;
+int
+afr_mark_fool_as_source_by_idx (int32_t *sources, int child_count, int idx)
+{
+	if (idx >= 0 && idx < child_count) {
+		sources[idx] = 1;
+		return 1;
+	}
+	return 0;
 }
 
 
 static int
-afr_sh_mark_biggest_as_source (afr_self_heal_t *sh, int child_count)
+afr_find_largest_file_size (struct iatt *bufs, int32_t *success_children,
+			    int child_count)
 {
-        int biggest = 0;
-        int i;
+	int idx = -1;
+	int i = -1;
+	int child = -1;
+	uint64_t max_size = 0;
+        uint64_t min_size = 0;
+        int      num_children = 0;
+
+	for (i = 0; i < child_count; i++) {
+		if (success_children[i] == -1)
+			break;
+
+		child = success_children[i];
+		if (bufs[child].ia_size > max_size) {
+			max_size = bufs[child].ia_size;
+			idx = child;
+		}
+
+                if ((num_children == 0) || (bufs[child].ia_size < min_size)) {
+                        min_size = bufs[child].ia_size;
+                }
+
+                num_children++;
+	}
+
+        /* If sizes are same for all of them, finding sources will have to
+         * happen with pending changelog. So return -1
+         */
+        if ((num_children > 1) && (min_size == max_size))
+                return -1;
+	return idx;
+}
+
+
+static int
+afr_find_newest_file (struct iatt *bufs, int32_t *success_children,
+		      int child_count)
+{
+	int idx = -1;
+	int i = -1;
+	int child = -1;
+	uint64_t max_ctime = 0;
+
+	for (i = 0; i < child_count; i++) {
+		if (success_children[i] == -1)
+			break;
+
+		child = success_children[i];
+		if (bufs[child].ia_ctime > max_ctime) {
+			max_ctime = bufs[child].ia_ctime;
+			idx = child;
+		}
+	}
+
+	return idx;
+}
+
+
+static int
+afr_mark_biggest_of_fools_as_source (int32_t *sources, int32_t **pending_matrix,
+                                     afr_node_character *characters,
+				     int32_t *success_children,
+                                     int child_count, struct iatt *bufs)
+{
+        int32_t       biggest_witness = 0;
+        int           nsources        = 0;
+        int32_t       *witnesses      = NULL;
+
+        GF_ASSERT (child_count > 0);
+
+	biggest_witness = afr_find_largest_file_size (bufs, success_children,
+						      child_count);
+	if (biggest_witness != -1)
+		goto found;
+
+        witnesses = GF_CALLOC (child_count, sizeof (*witnesses),
+                               gf_afr_mt_int32_t);
+        if (NULL == witnesses) {
+                nsources = -1;
+                goto out;
+        }
+
+        afr_compute_witness_of_fools (witnesses, pending_matrix, characters,
+                                      child_count);
+        biggest_witness = afr_find_biggest_witness_among_fools (witnesses,
+                                                                characters,
+                                                                child_count);
+	if (biggest_witness != -1)
+		goto found;
+
+	biggest_witness = afr_find_newest_file (bufs, success_children,
+						child_count);
+
+found:
+	nsources = afr_mark_fool_as_source_by_idx (sources, child_count,
+						   biggest_witness);
+out:
+        GF_FREE (witnesses);
+        return nsources;
+}
+
+int
+afr_mark_child_as_source_by_uid (int32_t *sources, struct iatt *bufs,
+                                 int32_t *success_children,
+                                 unsigned int child_count, uint32_t uid)
+{
+        int     i        = 0;
+        int     nsources = 0;
+        int     child    = 0;
 
         for (i = 0; i < child_count; i++) {
-                if (!sh->buf)
+                if (-1 == success_children[i])
                         break;
 
-                if (SIZE_GREATER (&sh->buf[i], &sh->buf[biggest])) {
-                        biggest = i;
+                child = success_children[i];
+                if (uid == bufs[child].ia_uid) {
+                        sources[child] = 1;
+                        nsources++;
                 }
         }
+        return nsources;
+}
 
-        sh->sources[biggest] = 1;
+int
+afr_get_child_with_lowest_uid (struct iatt *bufs, int32_t *success_children,
+                               unsigned int child_count)
+{
+        int     i        = 0;
+        int     smallest = -1;
+        int     child    = 0;
 
-        return 1;
+        for (i = 0; i < child_count; i++) {
+                if (-1 == success_children[i])
+                        break;
+                child = success_children[i];
+                if ((smallest == -1) ||
+                    (bufs[child].ia_uid < bufs[smallest].ia_uid)) {
+                        smallest = child;
+                }
+        }
+        return smallest;
 }
 
-
 static int
-afr_sh_mark_loweia_uid_as_source (afr_self_heal_t *sh, int child_count)
+afr_sh_mark_lowest_uid_as_source (struct iatt *bufs, int32_t *success_children,
+                                  int child_count, int32_t *sources)
 {
-        uid_t smallest = 0;
-        int i;
+        int   nsources              = 0;
+        int   smallest              = 0;
 
-        for (i = 0; i < child_count; i++) {
-                if (!sh->buf)
+        smallest = afr_get_child_with_lowest_uid (bufs, success_children,
+                                                  child_count);
+        if (smallest < 0) {
+                nsources = -1;
+                goto out;
+        }
+        nsources = afr_mark_child_as_source_by_uid (sources, bufs,
+                                                    success_children, child_count,
+                                                    bufs[smallest].ia_uid);
+out:
+        return nsources;
+}
+
+int
+afr_get_no_xattr_dir_read_child (xlator_t *this, int32_t *success_children,
+                                 struct iatt *bufs)
+{
+        afr_private_t *priv = NULL;
+        int            i = 0;
+        int            child = -1;
+        int            read_child = -1;
+
+        priv = this->private;
+        for (i = 0; i < priv->child_count; i++) {
+                child = success_children[i];
+                if (child < 0)
                         break;
+                if (read_child < 0)
+                        read_child = child;
+                else if (bufs[read_child].ia_size < bufs[child].ia_size)
+                        read_child = child;
+        }
+        return read_child;
+}
+
+int
+afr_sh_mark_zero_size_file_as_sink (struct iatt *bufs, int32_t *success_children,
+                                    int child_count, int32_t *sources)
+{
+        int             nsources = 0;
+        int             i = 0;
+        int             child = 0;
+        gf_boolean_t    sink_exists = _gf_false;
+        gf_boolean_t    source_exists = _gf_false;
+        int             source = -1;
 
-                if (sh->buf[i].ia_uid < sh->buf[smallest].ia_uid) {
-                        smallest = i;
+        for (i = 0; i < child_count; i++) {
+                child = success_children[i];
+                if (child < 0)
+                        break;
+                if (!bufs[child].ia_size) {
+                        sink_exists = _gf_true;
+                        continue;
+                }
+                if (!source_exists) {
+                        source_exists = _gf_true;
+                        source = child;
+                        continue;
+                }
+                if (bufs[source].ia_size != bufs[child].ia_size) {
+                        nsources = -1;
+                        goto out;
                 }
         }
+        if (!source_exists && !sink_exists) {
+                nsources = -1;
+                goto out;
+        }
 
-        sh->sources[smallest] = 1;
+        if (!source_exists || !sink_exists)
+                goto out;
 
-        return 1;
+        for (i = 0; i < child_count; i++) {
+                child = success_children[i];
+                if (child < 0)
+                        break;
+                if (bufs[child].ia_size) {
+                        sources[child] = 1;
+                        nsources++;
+                }
+        }
+out:
+        return nsources;
 }
 
+char *
+afr_get_character_str (afr_node_type type)
+{
+        char *character = NULL;
+
+        switch (type) {
+        case AFR_NODE_INNOCENT:
+                character = "innocent";
+                break;
+        case AFR_NODE_FOOL:
+                character = "fool";
+                break;
+        case AFR_NODE_WISE:
+                character = "wise";
+                break;
+        default:
+                character = "invalid";
+                break;
+        }
+        return character;
+}
+
+afr_node_type
+afr_find_child_character_type (int32_t *pending_row, int32_t child,
+                               unsigned int child_count)
+{
+        afr_node_type type = AFR_NODE_INVALID;
+
+        GF_ASSERT ((child >= 0) && (child < child_count));
+
+        if (afr_sh_is_innocent (pending_row, child_count))
+                type = AFR_NODE_INNOCENT;
+        else if (afr_sh_is_fool (pending_row, child, child_count))
+                type = AFR_NODE_FOOL;
+        else if (afr_sh_is_wise (pending_row, child, child_count))
+                type = AFR_NODE_WISE;
+        return type;
+}
 
 int
-afr_sh_mark_sources (afr_self_heal_t *sh, int child_count,
-                     afr_self_heal_type type)
+afr_build_sources (xlator_t *this, dict_t **xattr, struct iatt *bufs,
+                   int32_t **pending_matrix, int32_t *sources,
+                   int32_t *success_children, afr_transaction_type type,
+                   int32_t *subvol_status, gf_boolean_t ignore_ignorant)
 {
-	int i = 0;
+        afr_private_t           *priv = NULL;
+        afr_self_heal_type      sh_type    = AFR_SELF_HEAL_INVALID;
+        int                     nsources   = -1;
+        unsigned char           *ignorant_subvols = NULL;
+        unsigned int            child_count = 0;
 
-        int32_t ** pending_matrix;
-        int *      sources;
+        priv = this->private;
+        child_count = priv->child_count;
 
-        int size_differs = 0;
+        if (afr_get_children_count (success_children, priv->child_count) == 0)
+                goto out;
 
-        pending_matrix = sh->pending_matrix;
-        sources        = sh->sources;
+        if (!ignore_ignorant) {
+                ignorant_subvols = GF_CALLOC (sizeof (*ignorant_subvols),
+                                              child_count, gf_afr_mt_char);
+                if (NULL == ignorant_subvols)
+                        goto out;
+        }
 
-	int nsources = 0;
+        afr_build_pending_matrix (priv->pending_key, pending_matrix,
+                                  ignorant_subvols, xattr, type,
+                                  priv->child_count);
 
-        /* stores the 'characters' (innocent, fool, wise) of the nodes */
-        afr_node_character *
-                characters = GF_CALLOC (sizeof (afr_node_character), 
-                                        child_count,
-                                        gf_afr_mt_afr_node_character) ;
+        if (!ignore_ignorant)
+                afr_mark_ignorant_subvols_as_pending (pending_matrix,
+                                                      ignorant_subvols,
+                                                      priv->child_count);
+        sh_type = afr_self_heal_type_for_transaction (type);
+        if (AFR_SELF_HEAL_INVALID == sh_type)
+                goto out;
 
-	/* start clean */
-	for (i = 0; i < child_count; i++) {
-		sources[i] = 0;
-	}
-        
-        for (i = 0; i < child_count; i++) {
-                if (afr_sh_is_innocent (pending_matrix[i], child_count)) {
-                        characters[i].type = AFR_NODE_INNOCENT;
+        afr_sh_print_pending_matrix (pending_matrix, this);
 
-                } else if (afr_sh_is_fool (pending_matrix[i], i, child_count)) {
-                        characters[i].type = AFR_NODE_FOOL;
+        nsources = afr_mark_sources (this, sources, pending_matrix, bufs,
+                                     sh_type, success_children, subvol_status);
+out:
+        GF_FREE (ignorant_subvols);
+        return nsources;
+}
 
-                } else if (afr_sh_is_wise (pending_matrix[i], i, child_count)) {
-                        characters[i].type = AFR_NODE_WISE;
+void
+afr_find_character_types (afr_node_character *characters,
+                          int32_t **pending_matrix, int32_t *success_children,
+                          unsigned int child_count)
+{
+        afr_node_type type  = AFR_NODE_INVALID;
+        int           child = 0;
+        int           i     = 0;
 
-                } else {
-                        gf_log ("[module:replicate]", GF_LOG_ERROR,
-                                "Could not determine the state of subvolume %d!"
-                                " (This message should never appear."
-                                " Please file a bug report to "
-                                "<gluster-devel@nongnu.org>.)", i);
-                }
+        for (i = 0; i < child_count; i++) {
+                child = success_children[i];
+                if (child == -1)
+                        break;
+                type = afr_find_child_character_type (pending_matrix[child],
+                                                      child, child_count);
+                characters[child].type = type;
         }
+}
 
-        if (type == AFR_SELF_HEAL_DATA) {
-                size_differs = afr_sh_mark_if_size_differs (sh, child_count);
+void
+afr_mark_success_children_sources (int32_t *sources, int32_t *success_children,
+                                   unsigned int child_count)
+{
+        int i = 0;
+        for (i = 0; i < child_count; i++) {
+                if (success_children[i] == -1)
+                        break;
+                sources[success_children[i]] = 1;
         }
+}
+/**
+ * mark_sources: Mark all 'source' nodes and return number of source
+ * nodes found
+ *
+ * A node (a row in the pending matrix) belongs to one of
+ * three categories:
+ *
+ * M is the pending matrix.
+ *
+ * 'innocent' - M[i] is all zeroes
+ * 'fool'     - M[i] has i'th element = 1 (self-reference)
+ * 'wise'     - M[i] has i'th element = 0, others are 1 or 0.
+ *
+ * All 'innocent' nodes are sinks. If all nodes are innocent, no self-heal is
+ * needed.
+ *
+ * A 'wise' node can be a source. If two 'wise' nodes conflict, it is
+ * a split-brain. If one wise node refers to the other but the other doesn't
+ * refer back, the referrer is a source.
+ *
+ * All fools are sinks, unless there are no 'wise' nodes. In that case,
+ * one of the fools is made a source.
+ */
 
-        if ((type == AFR_SELF_HEAL_METADATA)
-            && afr_sh_all_nodes_innocent (characters, child_count)) {
+int
+afr_mark_sources (xlator_t *this, int32_t *sources, int32_t **pending_matrix,
+                  struct iatt *bufs, afr_self_heal_type type,
+                  int32_t *success_children, int32_t *subvol_status)
+{
+        /* stores the 'characters' (innocent, fool, wise) of the nodes */
+        afr_node_character *characters =  NULL;
+        int                nsources    = -1;
+        unsigned int       child_count = 0;
+        afr_private_t      *priv       = NULL;
 
-                nsources = afr_sh_mark_loweia_uid_as_source (sh, child_count);
+        priv = this->private;
+        child_count = priv->child_count;
+        characters = GF_CALLOC (sizeof (afr_node_character),
+                                child_count, gf_afr_mt_afr_node_character);
+        if (!characters)
                 goto out;
-        }
 
+        this = THIS;
+
+        /* start clean */
+        memset (sources, 0, sizeof (*sources) * child_count);
+        nsources = 0;
+        afr_find_character_types (characters, pending_matrix, success_children,
+                                  child_count);
         if (afr_sh_all_nodes_innocent (characters, child_count)) {
-                if (size_differs) {
-                        nsources = afr_sh_mark_biggest_as_source (sh,
-                                                                  child_count);
+                switch (type) {
+                case AFR_SELF_HEAL_METADATA:
+                        nsources = afr_sh_mark_lowest_uid_as_source (bufs,
+                                                             success_children,
+                                                             child_count,
+                                                             sources);
+                        break;
+                case AFR_SELF_HEAL_DATA:
+                        nsources = afr_sh_mark_zero_size_file_as_sink (bufs,
+                                                             success_children,
+                                                             child_count,
+                                                             sources);
+                        if ((nsources < 0) && subvol_status)
+                                *subvol_status |= SPLIT_BRAIN;
+                        break;
+                default:
+                        break;
                 }
+                goto out;
+        }
 
-        } else if (afr_sh_wise_nodes_exist (characters, child_count)) {
+        if (afr_sh_wise_nodes_exist (characters, child_count)) {
                 afr_sh_compute_wisdom (pending_matrix, characters, child_count);
 
                 if (afr_sh_wise_nodes_conflict (characters, child_count)) {
-                        /* split-brain */
-
+                        if (subvol_status)
+                                *subvol_status |= SPLIT_BRAIN;
                         nsources = -1;
-                        goto out;
-
                 } else {
-                        nsources = afr_sh_mark_wisest_as_sources (sources, 
+                        nsources = afr_sh_mark_wisest_as_sources (sources,
                                                                   characters,
                                                                   child_count);
                 }
         } else {
-                nsources = afr_sh_mark_biggest_fool_as_source (sh, characters,
-                                                               child_count);
+                if (subvol_status)
+                        *subvol_status |= ALL_FOOLS;
+                nsources = afr_mark_biggest_of_fools_as_source (sources,
+                                                                pending_matrix,
+                                                                characters,
+								success_children,
+                                                                child_count, bufs);
         }
 
 out:
+        if (nsources == 0)
+                afr_mark_success_children_sources (sources, success_children,
+                                                   child_count);
         GF_FREE (characters);
 
-	return nsources;
+        gf_log (this->name, GF_LOG_DEBUG, "Number of sources: %d", nsources);
+        return nsources;
 }
 
-
 void
 afr_sh_pending_to_delta (afr_private_t *priv, dict_t **xattr,
-                         int32_t *delta_matrix[], int success[],
+                         int32_t *delta_matrix[], unsigned char success[],
                          int child_count, afr_transaction_type type)
 {
-	int i = 0;
-	int j = 0;
-        int k = 0;
-
-	/* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-	int32_t   pending[3];
-	void    * pending_raw = NULL;
-        int       ret         = 0;
-
-	/* start clean */
-	for (i = 0; i < child_count; i++) {
-		for (j = 0; j < child_count; j++) {
-			delta_matrix[i][j] = 0;
-		}
-	}
-
-	for (i = 0; i < child_count; i++) {
-                pending_raw = NULL;
-
-                for (j = 0; j < child_count; j++) {
-                        ret = dict_get_ptr (xattr[i], priv->pending_key[j],
-                                            &pending_raw);
-                        if (ret < 0)
-                                gf_log ("afr_sh_pending_to_delta",
-                                        GF_LOG_WARNING,
-                                        "Unable to get dict value.");
-
-                        if (!success[j])
-                                continue;
+        int     tgt     = 0;
+        int     src     = 0;
+        int     value   = 0;
 
-                        k = afr_index_for_transaction_type (type);
+        afr_build_pending_matrix (priv->pending_key, delta_matrix, NULL,
+                                  xattr, type, priv->child_count);
 
-                        if (pending_raw) {
-				memcpy (pending, pending_raw, sizeof(pending));
-                                delta_matrix[i][j] = -(ntoh32 (pending[k]));
-                        } else {
-                                delta_matrix[i][j]  = 0;
+        /*
+         * The algorithm here has two parts.  First, for each subvol indexed
+         * as tgt, we try to figure out what count everyone should have for it.
+         * If the self-heal succeeded, that's easy; the value is zero.
+         * Otherwise, the value is the maximum of the succeeding nodes' counts.
+         * Once we know the value, we loop through (possibly for a second time)
+         * setting each count to the difference so that when we're done all
+         * succeeding nodes will have the same count for tgt.
+         */
+        for (tgt = 0; tgt < priv->child_count; ++tgt) {
+                value = 0;
+                if (!success[tgt]) {
+                        /* Find the maximum. */
+                        for (src = 0; src < priv->child_count; ++src) {
+                                if (!success[src]) {
+                                        continue;
+                                }
+                                if (delta_matrix[src][tgt] > value) {
+                                        value = delta_matrix[src][tgt];
+                                }
                         }
-
                 }
-	}
+                /* Force everyone who succeeded to the chosen value. */
+                for (src = 0; src < priv->child_count; ++src) {
+                        if (success[src]) {
+                                delta_matrix[src][tgt] = value
+                                                       - delta_matrix[src][tgt];
+                        }
+                        else {
+                                delta_matrix[src][tgt] = 0;
+                        }
+                }
+        }
 }
 
 
 int
-afr_sh_delta_to_xattr (afr_private_t *priv,
+afr_sh_delta_to_xattr (xlator_t *this,
                        int32_t *delta_matrix[], dict_t *xattr[],
-		       int child_count, afr_transaction_type type)
+                       int child_count, afr_transaction_type type)
 {
-	int i = 0;
-	int j = 0;
-        int k = 0;
+        int              i       = 0;
+        int              j       = 0;
+        int              k       = 0;
+        int              ret     = 0;
+        int32_t         *pending = NULL;
+        int32_t         *local_pending = NULL;
+        afr_private_t   *priv = NULL;
 
-	int ret = 0;
-
-	int32_t *pending = 0;
-
-	for (i = 0; i < child_count; i++) {
-		if (!xattr[i])
-			continue;
+        priv = this->private;
+        for (i = 0; i < child_count; i++) {
+                if (!xattr[i])
+                        continue;
 
-		for (j = 0; j < child_count; j++) {
+                local_pending = NULL;
+                for (j = 0; j < child_count; j++) {
                         pending = GF_CALLOC (sizeof (int32_t), 3,
                                              gf_afr_mt_int32_t);
+
+                        if (!pending) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to allocate pending entry "
+                                        "for %s[%d] on %s",
+                                        priv->pending_key[j], type,
+                                        priv->children[i]->name);
+                                continue;
+                        }
                         /* 3 = data+metadata+entry */
 
                         k = afr_index_for_transaction_type (type);
 
-			pending[k] = hton32 (delta_matrix[i][j]);
+                        pending[k] = hton32 (delta_matrix[i][j]);
 
+                        if (j == i) {
+                                local_pending = pending;
+                                continue;
+                        }
                         ret = dict_set_bin (xattr[i], priv->pending_key[j],
                                             pending,
-                                            3 * sizeof (int32_t));
-                        if (ret < 0)
-                                gf_log ("afr_sh_delta_to_xattr",
-                                        GF_LOG_WARNING,
+                                        AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+                        if (ret < 0) {
+                                gf_log (this->name, GF_LOG_WARNING,
                                         "Unable to set dict value.");
-		}
-	}
-	return 0;
+                                GF_FREE (pending);
+                        }
+                }
+                if (local_pending) {
+                        ret = dict_set_bin (xattr[i], priv->pending_key[i],
+                                            local_pending,
+                                        AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+                        if (ret < 0) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "Unable to set dict value.");
+                                GF_FREE (local_pending);
+                        }
+                }
+        }
+        return 0;
 }
 
 
 int
-afr_sh_has_metadata_pending (dict_t *xattr, int child_count, xlator_t *this)
+afr_sh_missing_entries_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t *priv = NULL;
-	/* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-	int32_t       pending[3];
-	void          *pending_raw = NULL;
-
-	int           ret = -1;
-	int            i  = 0;
-        int            j  = 0;
-
-	priv = this->private;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
 
-        for (i = 0; i < priv->child_count; i++) {
-                ret = dict_get_ptr (xattr, priv->pending_key[i],
-                                    &pending_raw);
+        local = frame->local;
+        sh = &local->self_heal;
 
-                if (ret != 0)
-                        return 0;
+        afr_sh_reset (frame, this);
 
-		memcpy (pending, pending_raw, sizeof(pending));
-                j = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
+        if (local->unhealable) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "split brain found, aborting selfheal of %s",
+                        local->loc.path);
+        }
 
-                if (pending[j])
-                        return 1;
+        if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+                sh->completion_cbk (frame, this);
+        } else {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "proceeding to metadata check on %s",
+                        local->loc.path);
+                afr_self_heal_metadata (frame, this);
         }
 
-	return 0;
+        return 0;
 }
 
 
-int
-afr_sh_has_data_pending (dict_t *xattr, int child_count, xlator_t *this)
+static int
+afr_sh_missing_entries_finish (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t *priv = NULL;
-	/* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-	int32_t       pending[3];
-	void          *pending_raw = NULL;
+        afr_internal_lock_t *int_lock = NULL;
+        afr_local_t         *local    = NULL;
 
-	int           ret = -1;
-	int            i  = 0;
-        int            j  = 0;
+        local = frame->local;
+        int_lock = &local->internal_lock;
 
-	priv = this->private;
+        int_lock->lock_cbk = afr_sh_missing_entries_done;
+        afr_unlock (frame, this);
 
-        for (i = 0; i < priv->child_count; i++) {
-                ret = dict_get_ptr (xattr, priv->pending_key[i],
-                                    &pending_raw);
+        return 0;
+}
 
-                if (ret != 0)
-                        return 0;
-                
-		memcpy (pending, pending_raw, sizeof(pending));
-                j = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+int
+afr_sh_common_create (afr_self_heal_t *sh, unsigned int child_count)
+{
+        int     ret = -ENOMEM;
+        sh->buf = GF_CALLOC (child_count, sizeof (*sh->buf),
+                             gf_afr_mt_iatt);
+        if (!sh->buf)
+                goto out;
+        sh->parentbufs = GF_CALLOC (child_count, sizeof (*sh->parentbufs),
+                                    gf_afr_mt_iatt);
+        if (!sh->parentbufs)
+                goto out;
+        sh->child_errno = GF_CALLOC (child_count, sizeof (*sh->child_errno),
+                                     gf_afr_mt_int);
+        if (!sh->child_errno)
+                goto out;
+        sh->success_children = afr_children_create (child_count);
+        if (!sh->success_children)
+                goto out;
+        sh->fresh_children = afr_children_create (child_count);
+        if (!sh->fresh_children)
+                goto out;
+        sh->xattr = GF_CALLOC (child_count, sizeof (*sh->xattr),
+                               gf_afr_mt_dict_t);
+        if (!sh->xattr)
+                goto out;
+        ret = 0;
+out:
+        return ret;
+}
 
-                if (pending[j])
-                        return 1;
-        }
+void
+afr_sh_common_lookup_resp_handler (call_frame_t *frame, void *cookie,
+                                   xlator_t *this,
+                                   int32_t op_ret, int32_t op_errno,
+                                   inode_t *inode, struct iatt *buf,
+                                   dict_t *xattr, struct iatt *postparent,
+                                   loc_t *loc)
+{
+        int              child_index = 0;
+        afr_local_t     *local = NULL;
+        afr_private_t   *priv = NULL;
+        afr_self_heal_t *sh = NULL;
 
-	return 0;
+        local = frame->local;
+        priv = this->private;
+        sh   = &local->self_heal;
+        child_index = (long) cookie;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0) {
+                        sh->buf[child_index] = *buf;
+                        sh->parentbufs[child_index] = *postparent;
+                        sh->success_children[sh->success_count] = child_index;
+                        sh->success_count++;
+                        sh->xattr[child_index] = dict_ref (xattr);
+                } else {
+                        gf_log (this->name, GF_LOG_DEBUG, "path %s on subvolume"
+                                " %s => -1 (%s)", loc->path,
+                                priv->children[child_index]->name,
+                                strerror (op_errno));
+                        local->self_heal.child_errno[child_index] = op_errno;
+                }
+        }
+        UNLOCK (&frame->lock);
+        return;
 }
 
+gf_boolean_t
+afr_valid_ia_type (ia_type_t ia_type)
+{
+        switch (ia_type) {
+        case IA_IFSOCK:
+        case IA_IFREG:
+        case IA_IFBLK:
+        case IA_IFCHR:
+        case IA_IFIFO:
+        case IA_IFLNK:
+        case IA_IFDIR:
+                return _gf_true;
+        default:
+                return _gf_false;
+        }
+        return _gf_false;
+}
 
 int
-afr_sh_has_entry_pending (dict_t *xattr, int child_count, xlator_t *this)
+afr_impunge_frame_create (call_frame_t *frame, xlator_t *this,
+                          int active_source, call_frame_t **impunge_frame)
 {
-        afr_private_t *priv = NULL;
-	/* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-	int32_t       pending[3];
-	void          *pending_raw = NULL;
-
-	int           ret = -1;
-	int            i  = 0;
-        int            j  = 0;
+        afr_local_t     *local         = NULL;
+        afr_local_t     *impunge_local = NULL;
+        afr_self_heal_t *impunge_sh    = NULL;
+        int32_t         op_errno       = 0;
+        afr_private_t   *priv          = NULL;
+        int             ret            = 0;
+        call_frame_t    *new_frame     = NULL;
+
+        op_errno = ENOMEM;
+        priv = this->private;
+        new_frame = copy_frame (frame);
+        if (!new_frame) {
+                goto out;
+        }
 
-	priv = this->private;
+        AFR_LOCAL_ALLOC_OR_GOTO (impunge_local, out);
 
-        for (i = 0; i < priv->child_count; i++) {
-                ret = dict_get_ptr (xattr, priv->pending_key[i],
-                                    &pending_raw);
+        local = frame->local;
+        new_frame->local = impunge_local;
+        impunge_sh = &impunge_local->self_heal;
+        impunge_sh->sh_frame = frame;
+        impunge_sh->active_source = active_source;
+        impunge_local->child_up  = memdup (local->child_up,
+                                           sizeof (*local->child_up) *
+                                           priv->child_count);
+        if (!impunge_local->child_up)
+                goto out;
 
-                if (ret != 0)
-                        return 0;
-                
-		memcpy (pending, pending_raw, sizeof(pending));
-                j = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
+        impunge_local->pending = afr_matrix_create (priv->child_count,
+                                                    AFR_NUM_CHANGE_LOGS);
+        if (!impunge_local->pending)
+                goto out;
 
-                if (pending[j])
-                        return 1;
+        ret = afr_sh_common_create (impunge_sh, priv->child_count);
+        if (ret) {
+                op_errno = -ret;
+                goto out;
         }
-
-	return 0;
+        op_errno = 0;
+        *impunge_frame = new_frame;
+out:
+        if (op_errno && new_frame)
+                AFR_STACK_DESTROY (new_frame);
+        return -op_errno;
 }
 
+void
+afr_sh_missing_entry_call_impunge_recreate (call_frame_t *frame, xlator_t *this,
+                                            struct iatt *buf,
+                                            struct iatt *postparent,
+                                            afr_impunge_done_cbk_t impunge_done)
+{
+        call_frame_t    *impunge_frame = NULL;
+        afr_local_t     *local = NULL;
+        afr_local_t     *impunge_local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_self_heal_t *impunge_sh = NULL;
+        int             ret = 0;
+        unsigned int    enoent_count = 0;
+        afr_private_t   *priv = NULL;
+        int             i = 0;
+        int32_t         op_errno = 0;
 
-/**
- * is_matrix_zero - return true if pending matrix is all zeroes
- */
+        local = frame->local;
+        sh    = &local->self_heal;
+        priv  = this->private;
+
+        enoent_count = afr_errno_count (NULL, sh->child_errno,
+                                        priv->child_count, ENOENT);
+        if (!enoent_count) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no missing files - %s. proceeding to metadata check",
+                        local->loc.path);
+                goto out;
+        }
+        sh->impunge_done = impunge_done;
+        ret = afr_impunge_frame_create (frame, this, sh->source, &impunge_frame);
+        if (ret)
+                goto out;
+        impunge_local = impunge_frame->local;
+        impunge_sh    = &impunge_local->self_heal;
+        loc_copy (&impunge_local->loc, &local->loc);
+        ret = afr_build_parent_loc (&impunge_sh->parent_loc,
+                                    &impunge_local->loc, &op_errno);
+        if (ret) {
+                ret = -op_errno;
+                goto out;
+        }
+        impunge_local->call_count = enoent_count;
+        impunge_sh->entrybuf = sh->buf[sh->source];
+        impunge_sh->parentbuf = sh->parentbufs[sh->source];
+        for (i = 0; i < priv->child_count; i++) {
+                if (!impunge_local->child_up[i]) {
+                        impunge_sh->child_errno[i] = ENOTCONN;
+                        continue;
+                }
+                if (sh->child_errno[i] != ENOENT) {
+                        impunge_sh->child_errno[i] = EEXIST;
+                        continue;
+                }
+        }
+        for (i = 0; i < priv->child_count; i++) {
+                if (sh->child_errno[i] != ENOENT)
+                        continue;
+                afr_sh_entry_impunge_create (impunge_frame, this, i);
+                enoent_count--;
+        }
+        GF_ASSERT (!enoent_count);
+        return;
+out:
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "impunge of %s failed, "
+                        "reason: %s", local->loc.path, strerror (-ret));
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+        }
+        afr_sh_missing_entries_finish (frame, this);
+}
 
 int
-afr_sh_is_matrix_zero (int32_t *pending_matrix[], int child_count)
+afr_sh_create_entry_cbk (call_frame_t *frame, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno)
 {
-	int i, j;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
 
-	for (i = 0; i < child_count; i++) 
-		for (j = 0; j < child_count; j++) 
-			if (pending_matrix[i][j]) 
-				return 0;
-	return 1;
+        local = frame->local;
+        sh = &local->self_heal;
+        if (op_ret < 0)
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+        afr_sh_missing_entries_finish (frame, this);
+        return 0;
 }
 
-
-int
-afr_sh_missing_entries_done (call_frame_t *frame, xlator_t *this)
+static int
+sh_missing_entries_create (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        int              type = 0;
+        struct iatt     *buf = NULL;
+        struct iatt     *postparent = NULL;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
 
-//	memset (sh->child_errno, 0, sizeof (int) * priv->child_count);
-	memset (sh->buf, 0, sizeof (struct iatt) * priv->child_count);
+        buf = &sh->buf[sh->source];
+        postparent = &sh->parentbufs[sh->source];
 
-        for (i = 0; i < priv->child_count; i++) {
-                sh->locked_nodes[i] = 0;
+        type = buf->ia_type;
+        if (!afr_valid_ia_type (type)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: unknown file type: 0%o", local->loc.path, type);
+                afr_set_local_for_unhealable (local);
+                afr_sh_missing_entries_finish (frame, this);
+                goto out;
         }
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i])
-			dict_unref (sh->xattr[i]);
-		sh->xattr[i] = NULL;
-	}
-
-	if (local->govinda_gOvinda) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"aborting selfheal of %s",
-			local->loc.path);
-		sh->completion_cbk (frame, this);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"proceeding to metadata check on %s",
-			local->loc.path);
-		afr_self_heal_metadata (frame, this);
-	}
-
-	return 0;
+        afr_sh_missing_entry_call_impunge_recreate (frame, this,
+                                                    buf, postparent,
+                                                    afr_sh_create_entry_cbk);
+out:
+        return 0;
 }
 
-
-int
-sh_missing_entries_unlck_cbk (call_frame_t *frame, void *cookie,
-			      xlator_t *this,
-			      int32_t op_ret, int32_t op_errno)
+void
+afr_sh_missing_entries_lookup_done (call_frame_t *frame, xlator_t *this,
+                                    int32_t op_ret, int32_t op_errno)
 {
-	afr_local_t     *local = NULL;
-	int              call_count = 0;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        ia_type_t       ia_type = IA_INVAL;
+        int32_t         nsources = 0;
+        loc_t           *loc = NULL;
+        int32_t         subvol_status = 0;
+        afr_transaction_type txn_type = AFR_DATA_TRANSACTION;
+        gf_boolean_t    split_brain = _gf_false;
+        int             read_child = -1;
 
-	local = frame->local;
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
+        loc = &local->loc;
 
-	LOCK (&frame->lock);
-	{
-	}
-	UNLOCK (&frame->lock);
+        if (op_ret < 0) {
+                if (op_errno == EIO) {
+                        afr_set_local_for_unhealable (local);
+                }
+                // EIO can happen if finding the fresh parent dir failed
+                goto out;
+        }
 
-	call_count = afr_frame_return (frame);
+        //now No chance for the ia_type to conflict
+        ia_type = sh->buf[sh->success_children[0]].ia_type;
+        txn_type = afr_transaction_type_get (ia_type);
+        nsources = afr_build_sources (this, sh->xattr, sh->buf,
+                                      sh->pending_matrix, sh->sources,
+                                      sh->success_children, txn_type,
+                                      &subvol_status, _gf_false);
+        if (nsources < 0) {
+                gf_log (this->name, GF_LOG_INFO, "No sources for dir of %s,"
+                        " in missing entry self-heal, continuing with the rest"
+                        " of the self-heals", local->loc.path);
+                if (subvol_status & SPLIT_BRAIN) {
+                        split_brain = _gf_true;
+                        switch (txn_type) {
+                        case AFR_DATA_TRANSACTION:
+                                nsources = 1;
+                                sh->sources[sh->success_children[0]] = 1;
+                                break;
+                        case AFR_ENTRY_TRANSACTION:
+                                read_child = afr_get_no_xattr_dir_read_child
+                                                          (this,
+                                                           sh->success_children,
+                                                           sh->buf);
+                                sh->sources[read_child] = 1;
+                                nsources = 1;
+                                break;
+                        default:
+                                op_errno = EIO;
+                                goto out;
+                        }
+                } else {
+                        op_errno = EIO;
+                        goto out;
+                }
+        }
 
-	if (call_count == 0) {
-		afr_sh_missing_entries_done (frame, this);
-	}
+        afr_get_fresh_children (sh->success_children, sh->sources,
+                                sh->fresh_children, priv->child_count);
+        sh->source = sh->fresh_children[0];
+        if (sh->source == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "No active sources found.");
+                op_errno = EIO;
+                goto out;
+        }
 
-	return 0;
+        if (sh->gfid_sh_success_cbk)
+                sh->gfid_sh_success_cbk (frame, this);
+        sh->type = sh->buf[sh->source].ia_type;
+        if (uuid_is_null (loc->inode->gfid))
+                uuid_copy (loc->gfid, sh->buf[sh->source].ia_gfid);
+        if (split_brain) {
+                afr_sh_missing_entries_finish (frame, this);
+        } else {
+                sh_missing_entries_create (frame, this);
+        }
+        return;
+out:
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+        afr_sh_set_error (sh, op_errno);
+        afr_sh_missing_entries_finish (frame, this);
+        return;
 }
 
-
 static int
-sh_missing_entries_finish (call_frame_t *frame, xlator_t *this)
+afr_sh_common_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int32_t op_ret, int32_t op_errno, inode_t *inode,
+                          struct iatt *buf, dict_t *xattr,
+                          struct iatt *postparent)
 {
-	afr_private_t      *priv = NULL;
-	afr_local_t        *local = NULL;
-	int                 i = 0;
-	int                 call_count = 0;
-	afr_self_heal_t    *sh = NULL;
+        int                     call_count = 0;
+        afr_local_t             *local = NULL;
+        afr_self_heal_t         *sh    = NULL;
+        afr_private_t           *priv  = NULL;
 
+        local = frame->local;
+        sh    = &local->self_heal;
+        priv  = this->private;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        afr_sh_common_lookup_resp_handler (frame, cookie, this, op_ret,
+                                           op_errno, inode, buf, xattr,
+                                           postparent, &sh->lookup_loc);
+        call_count = afr_frame_return (frame);
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (sh->locked_nodes[i])
-                        call_count++;
+        if (call_count)
+                goto out;
+        op_ret = -1;
+        if (!sh->success_count) {
+                op_errno = afr_resultant_errno_get (NULL, sh->child_errno,
+                                                    priv->child_count);
+                gf_log (this->name, GF_LOG_ERROR, "Failed to lookup %s, "
+                        "reason %s", sh->lookup_loc.path,
+                        strerror (op_errno));
+                goto done;
         }
 
-        if (call_count == 0) {
-                afr_sh_missing_entries_done (frame, this);
-                return 0;
+        if ((sh->lookup_flags & AFR_LOOKUP_FAIL_CONFLICTS) &&
+            (afr_conflicting_iattrs (sh->buf, sh->success_children,
+                                     priv->child_count,
+                                     sh->lookup_loc.path, this->name))) {
+                op_errno = EIO;
+                gf_log (this->name, GF_LOG_ERROR, "Conflicting entries "
+                        "for %s", sh->lookup_loc.path);
+                goto done;
         }
 
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->locked_nodes[i]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"unlocking %"PRId64"/%s on subvolume %s",
-				sh->parent_loc.inode->ino, local->loc.name,
-				priv->children[i]->name);
-
-			STACK_WIND (frame, sh_missing_entries_unlck_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->entrylk,
-                                    this->name,
-				    &sh->parent_loc, local->loc.name,
-				    ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
+        if ((sh->lookup_flags & AFR_LOOKUP_FAIL_MISSING_GFIDS) &&
+            (afr_gfid_missing_count (this->name, sh->success_children,
+                                     sh->buf, priv->child_count,
+                                     sh->lookup_loc.path))) {
+                op_errno = ENODATA;
+                gf_log (this->name, GF_LOG_ERROR, "Missing Gfids "
+                        "for %s", sh->lookup_loc.path);
+                goto done;
+        }
+        op_ret = 0;
 
-			if (!--call_count)
-				break;
-		}
-	}
-	return 0;
+done:
+        sh->lookup_done (frame, this, op_ret, op_errno);
+out:
+        return 0;
 }
 
-
-static int
-sh_destroy_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int32_t op_ret, int op_errno,
-                struct iatt *preop, struct iatt *postop)
+int
+afr_sh_remove_entry_cbk (call_frame_t *frame, xlator_t *this, int child,
+                         int32_t op_ret, int32_t op_errno)
 {
-        afr_local_t *local = NULL;
-
-        loc_t *parent_loc = cookie;
-
-        int call_count = 0;
+        int             call_count = 0;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
 
         local = frame->local;
+        sh = &local->self_heal;
 
-        if (op_ret == -1) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "setattr on %s failed: %s",
-                        local->loc.path, strerror (op_errno));
-        }
-
-        if (parent_loc) {
-                loc_wipe (parent_loc);
-                GF_FREE (parent_loc);
+        GF_ASSERT (sh->post_remove_call);
+        if ((op_ret == -1) && (op_errno != ENOENT)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "purge entry %s failed, on child %d reason, %s",
+                        local->loc.path, child, strerror (op_errno));
+                LOCK (&frame->lock);
+                {
+                        afr_sh_set_error (sh, EIO);
+                        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                }
+                UNLOCK (&frame->lock);
         }
-
         call_count = afr_frame_return (frame);
-        
-        if (call_count == 0) {
-                STACK_DESTROY (frame->root);
-        }
-        
-	return 0;
+        if (call_count == 0)
+                sh->post_remove_call (frame, this);
+        return 0;
 }
 
+void
+afr_sh_call_entry_expunge_remove (call_frame_t *frame, xlator_t *this,
+                                  int child_index, struct iatt *buf,
+                                  struct iatt *parentbuf,
+                                  afr_expunge_done_cbk_t expunge_done)
+{
+        call_frame_t    *expunge_frame = NULL;
+        afr_local_t     *local = NULL;
+        afr_local_t     *expunge_local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_self_heal_t *expunge_sh = NULL;
+        int32_t         op_errno = 0;
+        int             ret = 0;
 
-static int
-sh_missing_entries_newentry_cbk (call_frame_t *frame, void *cookie,
-				 xlator_t *this,
-				 int32_t op_ret, int32_t op_errno,
-				 inode_t *inode, struct iatt *buf,
-                                 struct iatt *preparent,
-                                 struct iatt *postparent)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	call_frame_t    *setattr_frame = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
-
-        loc_t *parent_loc = NULL;
-
-	struct iatt     stbuf;
-        int32_t valid;
-
-	local = frame->local;
-	sh    = &local->self_heal;
-	priv  = this->private;
-
-	child_index = (long) cookie;
-
-	stbuf.ia_atime = sh->buf[sh->source].ia_atime;
-	stbuf.ia_atime_nsec = sh->buf[sh->source].ia_atime_nsec;
-	stbuf.ia_mtime = sh->buf[sh->source].ia_mtime;
-	stbuf.ia_mtime_nsec = sh->buf[sh->source].ia_mtime_nsec;
-
-        stbuf.ia_uid = sh->buf[sh->source].ia_uid;
-        stbuf.ia_gid = sh->buf[sh->source].ia_gid;
-        
-        valid = GF_SET_ATTR_UID   | GF_SET_ATTR_GID |
-                GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-        
-	if (op_ret == 0) {
-		setattr_frame = copy_frame (frame);
-                
-                setattr_frame->local = GF_CALLOC (1, sizeof (afr_local_t),
-                                                  gf_afr_mt_afr_local_t);
-
-                ((afr_local_t *)setattr_frame->local)->call_count = 2;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"setattr (%s) on subvolume %s",
-			local->loc.path, priv->children[child_index]->name);
-
-		STACK_WIND_COOKIE (setattr_frame, sh_destroy_cbk,
-                                   (void *) (long) 0,
-                                   priv->children[child_index],
-                                   priv->children[child_index]->fops->setattr,
-                                   &local->loc, &stbuf, valid);
-
-                valid      = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-                parent_loc = GF_CALLOC (1, sizeof (*parent_loc), 
-                                        gf_afr_mt_loc_t);
-                afr_build_parent_loc (parent_loc, &local->loc);
-
-                STACK_WIND_COOKIE (setattr_frame, sh_destroy_cbk,
-                                   (void *) (long) parent_loc,
-                                   priv->children[child_index],
-                                   priv->children[child_index]->fops->setattr,
-                                   parent_loc, &sh->parentbuf, valid);
-	}
-
-	call_count = afr_frame_return (frame);
+        expunge_frame = copy_frame (frame);
+        if (!expunge_frame) {
+                goto out;
+        }
 
-	if (call_count == 0) {
-		sh_missing_entries_finish (frame, this);
-	}
+        AFR_LOCAL_ALLOC_OR_GOTO (expunge_local, out);
 
-	return 0;
+        local = frame->local;
+        sh = &local->self_heal;
+        expunge_frame->local = expunge_local;
+        expunge_sh = &expunge_local->self_heal;
+        expunge_sh->sh_frame = frame;
+        loc_copy (&expunge_local->loc, &local->loc);
+        ret = afr_build_parent_loc (&expunge_sh->parent_loc,
+                                    &expunge_local->loc, &op_errno);
+        if (ret) {
+                ret = -op_errno;
+                goto out;
+        }
+        sh->expunge_done = expunge_done;
+        afr_sh_entry_expunge_remove (expunge_frame, this, child_index, buf,
+                                     parentbuf);
+        return;
+out:
+        gf_log (this->name, GF_LOG_ERROR, "Expunge of %s failed, reason: %s",
+                local->loc.path, strerror (op_errno));
+        expunge_done (frame, this, child_index, -1, op_errno);
 }
 
-
-static int
-sh_missing_entries_mknod (call_frame_t *frame, xlator_t *this)
+void
+afr_sh_remove_stale_lookup_info (afr_self_heal_t *sh, int32_t *success_children,
+                                 int32_t *fresh_children,
+                                 unsigned int child_count)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
-        int              ret = 0;
-	int              enoent_count = 0;
-	int              call_count = 0;
-	mode_t           st_mode = 0;
-	dev_t            ia_dev = 0;
-        dict_t          *dict = NULL;
+        int     i = 0;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	for (i = 0; i < priv->child_count; i++)
-		if (sh->child_errno[i] == ENOENT)
-			enoent_count++;
+        for (i = 0; i < child_count; i++) {
+                if (afr_is_child_present (success_children, child_count, i) &&
+                    !afr_is_child_present (fresh_children, child_count, i)) {
+                        sh->child_errno[i] = ENOENT;
+                        GF_ASSERT (sh->xattr[i]);
+                        dict_unref (sh->xattr[i]);
+                        sh->xattr[i] = NULL;
+                }
+        }
+}
 
-	call_count = enoent_count;
-	local->call_count = call_count;
+int
+afr_sh_purge_stale_entries_done (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t         *local    = NULL;
+        afr_self_heal_t     *sh       = NULL;
+        afr_private_t       *priv     = NULL;
 
-	st_mode = st_mode_from_ia (sh->buf[sh->source].ia_prot,
-                                   sh->buf[sh->source].ia_type);
-	ia_dev  = sh->buf[sh->source].ia_dev;
+        local    = frame->local;
+        sh       = &local->self_heal;
+        priv     = this->private;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"mknod %s mode 0%o on %d subvolumes",
-		local->loc.path, st_mode, enoent_count);
+        if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+                afr_sh_missing_entries_finish (frame, this);
+        } else {
+                if (afr_gfid_missing_count (this->name, sh->fresh_children,
+                                            sh->buf, priv->child_count,
+                                            local->loc.path)) {
+                        afr_sh_common_lookup (frame, this, &local->loc,
+                                              afr_sh_missing_entries_lookup_done,
+                                              sh->sh_gfid_req,
+                                              AFR_LOOKUP_FAIL_CONFLICTS|
+                                              AFR_LOOKUP_FAIL_MISSING_GFIDS,
+                                              NULL);
+                } else {
+                        //No need to set gfid so goto missing entries lookup done
+                        //Behave as if you have done the lookup
+                        afr_sh_remove_stale_lookup_info (sh,
+                                                         sh->success_children,
+                                                         sh->fresh_children,
+                                                         priv->child_count);
+                        afr_children_copy (sh->success_children,
+                                           sh->fresh_children,
+                                           priv->child_count);
+                        afr_sh_missing_entries_lookup_done (frame, this, 0, 0);
+                }
+        }
+        return 0;
+}
 
-        dict = dict_new ();
-        if (!dict)
-                gf_log (this->name, GF_LOG_ERROR, "out of memory");
+gf_boolean_t
+afr_sh_purge_entry_condition (afr_local_t *local, afr_private_t *priv,
+                              int child)
+{
+        afr_self_heal_t *sh = NULL;
 
-        ret = afr_set_dict_gfid (dict, sh->buf[sh->source].ia_gfid);
-        if (ret)
-                gf_log (this->name, GF_LOG_DEBUG, "gfid set failed");
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->child_errno[i] == ENOENT) {
-			STACK_WIND_COOKIE (frame,
-					   sh_missing_entries_newentry_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->mknod,
-					   &local->loc, st_mode, ia_dev, dict);
-			if (!--call_count)
-				break;
-		}
-	}
+        sh = &local->self_heal;
 
-        if (dict)
-                dict_unref (dict);
+        if (local->child_up[child] &&
+            (!afr_is_child_present (sh->fresh_parent_dirs, priv->child_count,
+                                    child))
+            && (sh->child_errno[child] != ENOENT))
+                return _gf_true;
 
-	return 0;
+        return _gf_false;
 }
 
-
-static int
-sh_missing_entries_mkdir (call_frame_t *frame, xlator_t *this)
+gf_boolean_t
+afr_sh_purge_stale_entry_condition (afr_local_t *local, afr_private_t *priv,
+                                    int child)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-        dict_t          *dict = NULL;
-	int              i = 0;
-        int              ret = 0;
-	int              enoent_count = 0;
-	int              call_count = 0;
-	mode_t           st_mode = 0;
+        afr_self_heal_t *sh = NULL;
 
+        sh = &local->self_heal;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        if (local->child_up[child] &&
+            (!afr_is_child_present (sh->fresh_children, priv->child_count,
+                                    child))
+             && (sh->child_errno[child] != ENOENT))
+                return _gf_true;
 
-	for (i = 0; i < priv->child_count; i++)
-		if (sh->child_errno[i] == ENOENT)
-			enoent_count++;
+        return _gf_false;
+}
 
-	call_count = enoent_count;
-	local->call_count = call_count;
+void
+afr_sh_purge_entry_common (call_frame_t *frame, xlator_t *this,
+                           gf_boolean_t purge_condition (afr_local_t *local,
+                                                         afr_private_t *priv,
+                                                         int child))
+{
+        afr_local_t     *local = NULL;
+        afr_private_t   *priv = NULL;
+        afr_self_heal_t *sh = NULL;
+        int             i = 0;
+        int             call_count = 0;
 
-	st_mode = st_mode_from_ia (sh->buf[sh->source].ia_prot,
-                                   sh->buf[sh->source].ia_type);
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-        dict = dict_new ();
-        if (!dict) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                sh_missing_entries_finish (frame, this);
-                return 0;
+        for (i = 0; i < priv->child_count; i++) {
+                if (purge_condition (local, priv, i))
+                        call_count++;
         }
 
-        ret = afr_set_dict_gfid (dict, sh->buf[sh->source].ia_gfid);
-        if (ret)
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "inode gfid set failed");
-
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"mkdir %s mode 0%o on %d subvolumes",
-		local->loc.path, st_mode, enoent_count);
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->child_errno[i] == ENOENT) {
-                        if (!strcmp (local->loc.path, "/")) {
-                                /* We shouldn't try to create "/" */
+        if (call_count == 0) {
+                sh->post_remove_call (frame, this);
+                goto out;
+        }
 
-                                sh_missing_entries_finish (frame, this);
+        local->call_count = call_count;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!purge_condition (local, priv, i))
+                        continue;
+                gf_log (this->name, GF_LOG_INFO, "purging the stale entry %s "
+                        "on %s", local->loc.path, priv->children[i]->name);
+                afr_sh_call_entry_expunge_remove (frame, this,
+                                                  (long) i, &sh->buf[i],
+                                                  &sh->parentbufs[i],
+                                                  afr_sh_remove_entry_cbk);
+        }
+out:
+        return;
+}
 
-                                return 0;
-                        } else {
-                                STACK_WIND_COOKIE (frame,
-                                                   sh_missing_entries_newentry_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->mkdir,
-                                                   &local->loc, st_mode, dict);
-                                if (!--call_count)
-                                        break;
-                        }
-		}
-	}
+void
+afr_sh_purge_entry (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
 
-        if (dict)
-                dict_unref (dict);
+        local = frame->local;
+        sh = &local->self_heal;
+        sh->post_remove_call = afr_sh_missing_entries_finish;
 
-	return 0;
+        afr_sh_purge_entry_common (frame, this, afr_sh_purge_entry_condition);
 }
 
+void
+afr_sh_purge_stale_entry (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int             i = 0;
 
-static int
-sh_missing_entries_symlink (call_frame_t *frame, xlator_t *this,
-			    const char *link, struct iatt *buf)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-        dict_t          *dict = NULL;
-	int              i = 0;
-        int              ret = 0;
-	int              enoent_count = 0;
-	int              call_count = 0;
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
+        sh->post_remove_call = afr_sh_purge_stale_entries_done;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        for (i = 0; i < priv->child_count; i++) {
+                if (afr_is_child_present (sh->fresh_children,
+                                          priv->child_count, i))
+                        continue;
 
-	for (i = 0; i < priv->child_count; i++)
-		if (sh->child_errno[i] == ENOENT)
-			enoent_count++;
+                if ((!local->child_up[i]) || sh->child_errno[i] != 0)
+                        continue;
 
-	call_count = enoent_count;
-	local->call_count = call_count;
+                GF_ASSERT (!uuid_is_null (sh->entrybuf.ia_gfid) ||
+                           uuid_is_null (sh->buf[i].ia_gfid));
 
-        dict = dict_new ();
-        if (!dict) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                sh_missing_entries_finish (frame, this);
-                return 0;
-        }
+                if ((sh->entrybuf.ia_type != sh->buf[i].ia_type) ||
+                    (uuid_compare (sh->buf[i].ia_gfid,
+                                   sh->entrybuf.ia_gfid)))
+                        continue;
 
-        ret = afr_set_dict_gfid (dict, buf->ia_gfid);
-        if (ret)
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "dict gfid set failed");
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"symlink %s -> %s on %d subvolumes",
-		local->loc.path, link, enoent_count);
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->child_errno[i] == ENOENT) {
-			STACK_WIND_COOKIE (frame,
-					   sh_missing_entries_newentry_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->symlink,
-					   link, &local->loc, dict);
-			if (!--call_count)
-				break;
-		}
-	}
+                afr_children_add_child (sh->fresh_children, i,
+                                        priv->child_count);
 
-	return 0;
+        }
+        afr_sh_purge_entry_common (frame, this,
+                                   afr_sh_purge_stale_entry_condition);
 }
 
-
-static int
-sh_missing_entries_readlink_cbk (call_frame_t *frame, void *cookie,
-				 xlator_t *this,
-				 int32_t op_ret, int32_t op_errno,
-				 const char *link, struct iatt *sbuf)
+void
+afr_sh_save_child_iatts_from_policy (int32_t *children, struct iatt *bufs,
+                                     struct iatt *save,
+                                     unsigned int child_count)
 {
-	if (op_ret > 0)
-		sh_missing_entries_symlink (frame, this, link, sbuf);
-	else
-		sh_missing_entries_finish (frame, this);
+        int             i = 0;
+        int             child = 0;
+        gf_boolean_t    saved = _gf_false;
 
-	return 0;
+        GF_ASSERT (save);
+        //if iatt buf with gfid exists sets it
+        for (i = 0; i < child_count; i++) {
+                child = children[i];
+                if (child == -1)
+                        break;
+                *save = bufs[child];
+                saved = _gf_true;
+                if (!uuid_is_null (save->ia_gfid))
+                        break;
+        }
+        GF_ASSERT (saved);
 }
 
-
-static int
-sh_missing_entries_readlink (call_frame_t *frame, xlator_t *this)
+void
+afr_get_children_of_fresh_parent_dirs (afr_self_heal_t *sh,
+                                       unsigned int child_count)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	STACK_WIND (frame, sh_missing_entries_readlink_cbk,
-		    priv->children[sh->source],
-		    priv->children[sh->source]->fops->readlink,
-		    &local->loc, 4096);
-
-	return 0;
+        afr_children_intersection_get (sh->success_children,
+                                       sh->fresh_parent_dirs,
+                                       sh->sources, child_count);
+        afr_get_fresh_children (sh->success_children, sh->sources,
+                                sh->fresh_children, child_count);
+        memset (sh->sources, 0, sizeof (*sh->sources) * child_count);
 }
 
-
-static int
-sh_missing_entries_create (call_frame_t *frame, xlator_t *this)
+void
+afr_sh_children_lookup_done (call_frame_t *frame, xlator_t *this,
+                             int32_t op_ret, int32_t op_errno)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	int              type = 0;
-	int              i = 0;
-	afr_private_t   *priv = NULL;
-	int              enoent_count = 0;
-	int              govinda_gOvinda = 0;
+        afr_local_t      *local = NULL;
+        afr_self_heal_t  *sh = NULL;
+        afr_private_t    *priv = NULL;
+        int32_t          fresh_child_enoents = 0;
+        int32_t          fresh_parent_count = 0;
 
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        if (op_ret < 0)
+                goto fail;
+        afr_get_children_of_fresh_parent_dirs (sh, priv->child_count);
+        fresh_parent_count = afr_get_children_count (sh->fresh_parent_dirs,
+                                                     priv->child_count);
+        //we need the enoent count of the subvols present in fresh_parent_dirs
+        fresh_child_enoents = afr_errno_count (sh->fresh_parent_dirs,
+                                               sh->child_errno,
+                                               priv->child_count, ENOENT);
+        if (fresh_child_enoents == fresh_parent_count) {
+                afr_sh_set_error (sh, ENOENT);
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                afr_sh_purge_entry (frame, this);
+        } else if (!afr_conflicting_iattrs (sh->buf, sh->fresh_children,
+                                            priv->child_count, local->loc.path,
+                                            this->name)) {
+                afr_sh_save_child_iatts_from_policy (sh->fresh_children,
+                                                     sh->buf, &sh->entrybuf,
+                                                     priv->child_count);
+                afr_update_gfid_from_iatts (sh->sh_gfid_req, sh->buf,
+                                            sh->fresh_children,
+                                            priv->child_count);
+                afr_sh_purge_stale_entry (frame, this);
+        } else {
+                op_errno = EIO;
+                afr_set_local_for_unhealable (local);
+                goto fail;
+        }
 
-	for (i = 0; i < priv->child_count; i++) {
-                if (!local->child_up[i])
-                        continue;
+        return;
 
-		if (sh->child_errno[i]) {
-			if (sh->child_errno[i] == ENOENT)
-				enoent_count++;
-		} else {
-			if (type) {
-				if (type != sh->buf[i].ia_type) {
-                                        gf_log (this->name, GF_LOG_TRACE,
-                                                "file %s is govinda!",
-                                                local->loc.path);
-
-					govinda_gOvinda = 1;
-                                }
-			} else {
-				sh->source = i;
-				type = sh->buf[i].ia_type;
-			}
-		}
-	}
+fail:
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+        afr_sh_set_error (sh, op_errno);
+        afr_sh_missing_entries_finish (frame, this);
+        return;
+}
 
-	if (govinda_gOvinda) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"conflicting filetypes exist for path %s. returning.",
-			local->loc.path);
+static void
+afr_sh_find_fresh_parents (call_frame_t *frame, xlator_t *this,
+                           int32_t op_ret, int32_t op_errno)
+{
+        afr_self_heal_t *sh  = NULL;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *local = NULL;
+        int             enoent_count = 0;
+        int             nsources = 0;
+        int             source  = -1;
+        int32_t         subvol_status = 0;
 
-		local->govinda_gOvinda = 1;
-		sh_missing_entries_finish (frame, this);
-		return 0;
-	}
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-	if (!type) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"no source found for %s. all nodes down?. returning.",
-			local->loc.path);
-		/* subvolumes down and/or file does not exist */
-		sh_missing_entries_finish (frame, this);
-		return 0;
-	}
+        if (op_ret < 0)
+                goto out;
+        enoent_count = afr_errno_count (NULL, sh->child_errno,
+                                        priv->child_count, ENOENT);
+        if (enoent_count > 0) {
+                gf_log (this->name, GF_LOG_INFO, "Parent dir missing for %s,"
+                        " in missing entry self-heal, aborting missing-entry "
+                        "self-heal",
+                        local->loc.path);
+                afr_sh_missing_entries_finish (frame, this);
+                return;
+        }
 
-	if (enoent_count == 0) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"no missing files - %s. proceeding to metadata check",
-			local->loc.path);
-		/* proceed to next step - metadata self-heal */
-		sh_missing_entries_finish (frame, this);
-		return 0;
-	}
+        nsources = afr_build_sources (this, sh->xattr, sh->buf,
+                                      sh->pending_matrix, sh->sources,
+                                      sh->success_children,
+                                      AFR_ENTRY_TRANSACTION, &subvol_status,
+                                      _gf_true);
+        if ((subvol_status & ALL_FOOLS) ||
+            (subvol_status & SPLIT_BRAIN)) {
+                gf_log (this->name, GF_LOG_INFO, "%s: Performing conservative "
+                        "merge", sh->parent_loc.path);
+                afr_mark_success_children_sources (sh->sources,
+                                                   sh->success_children,
+                                                   priv->child_count);
+        } else if (nsources < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "No sources for dir "
+                        "of %s, in missing entry self-heal, aborting "
+                        "self-heal", local->loc.path);
+                op_errno = EIO;
+                goto out;
+        }
 
-	switch (type) {
-	case IA_IFSOCK:
-	case IA_IFREG:
-	case IA_IFBLK:
-	case IA_IFCHR:
-	case IA_IFIFO:
-		sh_missing_entries_mknod (frame, this);
-		break;
-	case IA_IFLNK:
-		sh_missing_entries_readlink (frame, this);
-		break;
-	case IA_IFDIR:
-		sh_missing_entries_mkdir (frame, this);
-		break;
-	default:
-		gf_log (this->name, GF_LOG_ERROR,
-			"unknown file type: 0%o", type);
-		local->govinda_gOvinda = 1;
-		sh_missing_entries_finish (frame, this);
-	}
+        source = afr_sh_select_source (sh->sources, priv->child_count);
+        if (source == -1) {
+                GF_ASSERT (0);
+                gf_log (this->name, GF_LOG_DEBUG, "No active sources found.");
+                op_errno = EIO;
+                goto out;
+        }
+        afr_get_fresh_children (sh->success_children, sh->sources,
+                                sh->fresh_parent_dirs, priv->child_count);
+        afr_sh_common_lookup (frame, this, &local->loc,
+                              afr_sh_children_lookup_done, NULL, 0,
+                              NULL);
+        return;
 
-	return 0;
+out:
+        afr_sh_set_error (sh, op_errno);
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+	afr_sh_missing_entries_finish (frame, this);
+        return;
 }
 
+void
+afr_sh_common_reset (afr_self_heal_t *sh, unsigned int child_count)
+{
+        int             i = 0;
 
-static int
-sh_missing_entries_lookup_cbk (call_frame_t *frame, void *cookie,
-			       xlator_t *this,
-			       int32_t op_ret, int32_t op_errno,
-                               inode_t *inode, struct iatt *buf, dict_t *xattr,
-                               struct iatt *postparent)
-{
-	int              child_index = 0;
-	afr_local_t     *local = NULL;
-	int              call_count = 0;
-	afr_private_t   *priv = NULL;
-
-
-	local = frame->local;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"path %s on subvolume %s is of mode 0%o",
-				local->loc.path,
-				priv->children[child_index]->name,
-				buf->ia_type);
-
-			local->self_heal.buf[child_index] = *buf;
-                        local->self_heal.parentbuf        = *postparent;
-		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"path %s on subvolume %s => -1 (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-
-			local->self_heal.child_errno[child_index] = op_errno;
-		}
-
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		sh_missing_entries_create (frame, this);
-	}
-
-	return 0;
+        for (i = 0; i < child_count; i++) {
+                memset (&sh->buf[i], 0, sizeof (sh->buf[i]));
+                memset (&sh->parentbufs[i], 0, sizeof (sh->parentbufs[i]));
+                sh->child_errno[i] = 0;
+        }
+        memset (&sh->parentbuf, 0, sizeof (sh->parentbuf));
+        sh->success_count = 0;
+        afr_reset_children (sh->success_children, child_count);
+        afr_reset_children (sh->fresh_children, child_count);
+        afr_reset_xattr (sh->xattr, child_count);
+        loc_wipe (&sh->lookup_loc);
 }
 
-
-static int
-sh_missing_entries_lookup (call_frame_t *frame, xlator_t *this)
+/* afr self-heal state will be lost if this call is made
+ * please check the afr_sh_common_reset that is called in this function
+ */
+int
+afr_sh_common_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                      afr_lookup_done_cbk_t lookup_done , uuid_t gfid,
+                      int32_t flags, dict_t *xdata)
 {
-	afr_local_t    *local = NULL;
-	int             i = 0;
-	int             call_count = 0;
-	afr_private_t  *priv = NULL;
-	dict_t         *xattr_req = NULL;
-	int             ret = -1;
+        afr_local_t    *local = NULL;
+        int             i = 0;
+        int             call_count = 0;
+        afr_private_t  *priv = NULL;
+        dict_t         *xattr_req = NULL;
+        afr_self_heal_t *sh = NULL;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
+        priv  = this->private;
+        sh    = &local->self_heal;
 
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
+        call_count = afr_up_children_count (local->child_up, priv->child_count);
 
-	local->call_count = call_count;
+        local->call_count = call_count;
 
-	xattr_req = dict_new();
+        xattr_req = dict_new();
 
-	if (xattr_req) {
-                for (i = 0; i < priv->child_count; i++) {
-                        ret = dict_set_uint64 (xattr_req,
-                                               priv->pending_key[i],
-                                               3 * sizeof(int32_t));
-                        if (ret < 0)
-                                gf_log (this->name, GF_LOG_WARNING,
-                                        "Unable to set dict value.");
+        if (xattr_req) {
+                afr_xattr_req_prepare (this, xattr_req, loc->path);
+                if (gfid) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "looking up %s with gfid: %s",
+                                loc->path, uuid_utoa (gfid));
+                        GF_ASSERT (!uuid_is_null (gfid));
+                        afr_set_dict_gfid (xattr_req, gfid);
                 }
         }
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"looking up %s on subvolume %s",
-				local->loc.path, priv->children[i]->name);
-
-			STACK_WIND_COOKIE (frame,
-					   sh_missing_entries_lookup_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->lookup,
-					   &local->loc, xattr_req);
-
-			if (!--call_count)
-				break;
-		}
-	}
+        afr_sh_common_reset (sh, priv->child_count);
+        sh->lookup_done = lookup_done;
+        loc_copy (&sh->lookup_loc, loc);
+        sh->lookup_flags = flags;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "looking up %s on subvolume %s",
+                                loc->path, priv->children[i]->name);
+
+                        STACK_WIND_COOKIE (frame,
+                                           afr_sh_common_lookup_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->lookup,
+                                           loc, xattr_req);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-	if (xattr_req)
-		dict_unref (xattr_req);
+        if (xattr_req)
+                dict_unref (xattr_req);
 
-	return 0;
+        return 0;
 }
 
 
 
 int
-afr_sh_post_nonblocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
+afr_sh_post_nb_entrylk_missing_entry_sh_cbk (call_frame_t *frame,
+                                             xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
+        afr_self_heal_t     *sh       = NULL;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
+        sh       = &local->self_heal;
 
         if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_INFO,
                         "Non blocking entrylks failed.");
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
                 afr_sh_missing_entries_done (frame, this);
         } else {
 
                 gf_log (this->name, GF_LOG_DEBUG,
                         "Non blocking entrylks done. Proceeding to FOP");
-                sh_missing_entries_lookup (frame, this);
+                afr_sh_common_lookup (frame, this, &sh->parent_loc,
+                                      afr_sh_find_fresh_parents,
+                                      NULL, AFR_LOOKUP_FAIL_CONFLICTS,
+                                      NULL);
         }
 
         return 0;
 }
 
-static int
-afr_sh_entrylk (call_frame_t *frame, xlator_t *this)
+int
+afr_sh_entrylk (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                char *base_name, afr_lock_cbk_t lock_cbk)
 {
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
-        afr_self_heal_t     *sh       = NULL;
+        afr_private_t       *priv     = NULL;
 
+        priv     = this->private;
         local    = frame->local;
         int_lock = &local->internal_lock;
-        sh       = &local->self_heal;
 
         int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
         int_lock->selfheal_lk_type    = AFR_ENTRY_SELF_HEAL_LK;
 
         afr_set_lock_number (frame, this);
 
-        int_lock->lk_basename = local->loc.name;
-        int_lock->lk_loc      = &sh->parent_loc;
-        int_lock->lock_cbk    = afr_sh_post_nonblocking_entrylk_cbk;
+        int_lock->lk_basename = base_name;
+        int_lock->lk_loc      = loc;
+        int_lock->lock_cbk    = lock_cbk;
+        int_lock->domain      = this->name;
 
+        int_lock->lockee_count = 0;
+        afr_init_entry_lockee (&int_lock->lockee[0], local, loc,
+                               base_name, priv->child_count);
+        int_lock->lockee_count++;
         afr_nonblocking_entrylk (frame, this);
 
         return 0;
 }
 
 static int
-afr_self_heal_missing_entries (call_frame_t *frame, xlator_t *this)
+afr_self_heal_parent_entrylk (call_frame_t *frame, xlator_t *this,
+                              afr_lock_cbk_t lock_cbk)
 {
+        afr_local_t         *local    = NULL;
+        afr_self_heal_t     *sh       = NULL;
         afr_internal_lock_t *int_lock = NULL;
-	afr_local_t         *local    = NULL;
-	afr_self_heal_t     *sh       = NULL;
-	afr_private_t       *priv     = NULL;
+        int                 ret       = -1;
+        int32_t             op_errno  = 0;
 
-	local    = frame->local;
-        int_lock = &local->internal_lock;
-	sh       = &local->self_heal;
-	priv     = this->private;
+        local    = frame->local;
+        sh       = &local->self_heal;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"attempting to recreate missing entries for path=%s",
-		local->loc.path);
+        gf_log (this->name, GF_LOG_TRACE,
+                "attempting to recreate missing entries for path=%s",
+                local->loc.path);
 
-	afr_build_parent_loc (&sh->parent_loc, &local->loc);
+        ret = afr_build_parent_loc (&sh->parent_loc, &local->loc, &op_errno);
+        if (ret)
+                goto out;
 
-        afr_sh_entrylk (frame, this);
-	return 0;
+        afr_sh_entrylk (frame, this, &sh->parent_loc, NULL,
+                        lock_cbk);
+        return 0;
+out:
+        int_lock = &local->internal_lock;
+        int_lock->lock_op_ret = -1;
+        lock_cbk (frame, this);
+        return 0;
 }
 
-afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
+static int
+afr_self_heal_missing_entries (call_frame_t *frame, xlator_t *this)
 {
-        afr_private_t *priv = NULL;
-        afr_local_t   *lc     = NULL;
+        afr_local_t     *local = NULL;
         afr_self_heal_t *sh = NULL;
-        afr_self_heal_t *shc = NULL;
 
+        local = frame->local;
+        sh = &local->self_heal;
+
+        sh->sh_type_in_action  = AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY;
+
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
+
+        afr_self_heal_parent_entrylk (frame, this,
+                                      afr_sh_post_nb_entrylk_missing_entry_sh_cbk);
+        return 0;
+}
+
+afr_local_t*
+afr_self_heal_local_init (afr_local_t *l, xlator_t *this)
+{
+        afr_private_t   *priv  = NULL;
+        afr_local_t     *lc    = NULL;
+        afr_self_heal_t *sh    = NULL;
+        afr_self_heal_t *shc   = NULL;
+        int             ret    = 0;
 
         priv = this->private;
 
         sh = &l->self_heal;
 
-        lc = GF_CALLOC (1, sizeof (afr_local_t),
-                        gf_afr_mt_afr_local_t);
+        lc = mem_get0 (this->local_pool);
+        if (!lc)
+                goto out;
 
         shc = &lc->self_heal;
 
         shc->unwind = sh->unwind;
-        shc->need_data_self_heal = sh->need_data_self_heal;
-        shc->need_metadata_self_heal = sh->need_metadata_self_heal;
-        shc->need_entry_self_heal = sh->need_entry_self_heal;
+        shc->gfid_sh_success_cbk = sh->gfid_sh_success_cbk;
+        shc->do_missing_entry_self_heal = sh->do_missing_entry_self_heal;
+        shc->do_gfid_self_heal = sh->do_gfid_self_heal;
+        shc->do_data_self_heal = sh->do_data_self_heal;
+        shc->do_metadata_self_heal = sh->do_metadata_self_heal;
+        shc->do_entry_self_heal = sh->do_entry_self_heal;
+        shc->force_confirm_spb = sh->force_confirm_spb;
         shc->forced_merge = sh->forced_merge;
-        shc->healing_fd_opened = sh->healing_fd_opened;
-        shc->data_lock_held = sh->data_lock_held;
-        if (sh->healing_fd && !sh->healing_fd_opened)
-                shc->healing_fd = fd_ref (sh->healing_fd);
-        else
-                shc->healing_fd = sh->healing_fd;
         shc->background = sh->background;
         shc->type = sh->type;
+        shc->data_sh_info = "";
+        shc->metadata_sh_info =  "";
+
+        uuid_copy (shc->sh_gfid_req, sh->sh_gfid_req);
+        if (l->loc.path) {
+                ret = loc_copy (&lc->loc, &l->loc);
+                if (ret < 0)
+                        goto out;
+        }
 
-        if (l->loc.path)
-                loc_copy (&lc->loc, &l->loc);
+        lc->child_up  = memdup (l->child_up,
+                                sizeof (*lc->child_up) * priv->child_count);
+        if (!lc->child_up) {
+                ret = -1;
+                goto out;
+        }
 
-        lc->child_up  = memdup (l->child_up, priv->child_count);
         if (l->xattr_req)
                 lc->xattr_req = dict_ref (l->xattr_req);
 
@@ -1521,69 +2245,67 @@ afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
                 lc->cont.lookup.inode = inode_ref (l->cont.lookup.inode);
         if (l->cont.lookup.xattr)
                 lc->cont.lookup.xattr = dict_ref (l->cont.lookup.xattr);
-        if (l->internal_lock.inode_locked_nodes)
-                lc->internal_lock.inode_locked_nodes =
-                        memdup (l->internal_lock.inode_locked_nodes,
-                                priv->child_count);
-        else
-                lc->internal_lock.inode_locked_nodes =
-                        GF_CALLOC (sizeof (*l->internal_lock.inode_locked_nodes),
-                                   priv->child_count,
-                                   gf_afr_mt_char);
-        if (l->internal_lock.entry_locked_nodes)
-                lc->internal_lock.entry_locked_nodes =
-                        memdup (l->internal_lock.entry_locked_nodes,
-                                priv->child_count);
-        else
-                lc->internal_lock.entry_locked_nodes =
-                        GF_CALLOC (sizeof (*l->internal_lock.entry_locked_nodes),
-                                   priv->child_count,
-                                   gf_afr_mt_char);
-        if (l->internal_lock.locked_nodes)
-                lc->internal_lock.locked_nodes =
-                        memdup (l->internal_lock.locked_nodes,
-                                priv->child_count);
-        else
-                lc->internal_lock.locked_nodes =
-                        GF_CALLOC (sizeof (*l->internal_lock.locked_nodes),
-                                   priv->child_count,
-                                   gf_afr_mt_char);
 
-        lc->internal_lock.inodelk_lock_count =
-                l->internal_lock.inodelk_lock_count;
-        lc->internal_lock.entrylk_lock_count =
-                l->internal_lock.entrylk_lock_count;
+        lc->internal_lock.locked_nodes =
+                             GF_CALLOC (sizeof (*l->internal_lock.locked_nodes),
+                                        priv->child_count, gf_afr_mt_char);
+        if (!lc->internal_lock.locked_nodes) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = afr_inodelk_init (&lc->internal_lock.inodelk[0],
+                                this->name, priv->child_count);
+        if (ret)
+                goto out;
 
+out:
+        if (ret) {
+                afr_local_cleanup (lc, this);
+                lc = NULL;
+        }
         return lc;
 }
 
-
 int
 afr_self_heal_completion_cbk (call_frame_t *bgsh_frame, xlator_t *this)
 {
         afr_private_t *   priv  = NULL;
-	afr_local_t *     local = NULL;
+        afr_local_t *     local = NULL;
         afr_self_heal_t * sh    = NULL;
+        afr_local_t *     orig_frame_local = NULL;
+        afr_self_heal_t * orig_frame_sh = NULL;
         char              sh_type_str[256] = {0,};
+        gf_loglevel_t     loglevel = 0;
 
         priv  = this->private;
-	local = bgsh_frame->local;
+        local = bgsh_frame->local;
         sh    = &local->self_heal;
 
-	if (local->govinda_gOvinda) {
-                afr_set_split_brain (this, local->cont.lookup.inode, _gf_true);
-	} else {
-                afr_set_split_brain (this, local->cont.lookup.inode, _gf_false);
-	}
+        if (local->unhealable) {
+                afr_set_split_brain (this, sh->inode, SPB, SPB);
+        }
 
-        afr_self_heal_type_str_get(sh, sh_type_str,
-                                   sizeof(sh_type_str));
-        gf_log (this->name, GF_LOG_NORMAL,
-                "background %s self-heal completed on %s", sh_type_str,
-                local->loc.path);
+        afr_self_heal_type_str_get (sh, sh_type_str,
+                                    sizeof(sh_type_str));
+        if (is_self_heal_failed (sh, AFR_CHECK_ALL) && !priv->shd.iamshd) {
+                loglevel = GF_LOG_ERROR;
+        } else if (!is_self_heal_failed (sh, AFR_CHECK_ALL)) {
+                loglevel = GF_LOG_INFO;
+        } else {
+                loglevel = GF_LOG_DEBUG;
+        }
 
-        if (!sh->unwound) {
-                sh->unwind (sh->orig_frame, this);
+        afr_log_self_heal_completion_status (local, loglevel);
+
+        FRAME_SU_UNDO (bgsh_frame, afr_local_t);
+
+        if (!sh->unwound && sh->unwind) {
+                orig_frame_local = sh->orig_frame->local;
+                orig_frame_sh = &orig_frame_local->self_heal;
+                orig_frame_sh->actual_sh_started = _gf_true;
+                sh->unwind (sh->orig_frame, this, sh->op_ret, sh->op_errno,
+                            is_self_heal_failed (sh, AFR_CHECK_ALL));
         }
 
         if (sh->background) {
@@ -1596,119 +2318,495 @@ afr_self_heal_completion_cbk (call_frame_t *bgsh_frame, xlator_t *this)
 
         AFR_STACK_DESTROY (bgsh_frame);
 
-	return 0;
+        return 0;
 }
 
 int
-afr_self_heal (call_frame_t *frame, xlator_t *this)
+afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int32_t          op_errno = 0;
+        int              ret = 0;
+        afr_self_heal_t *orig_sh = NULL;
+        call_frame_t    *sh_frame = NULL;
+        afr_local_t     *sh_local = NULL;
+        loc_t           *loc   = NULL;
 
-        call_frame_t *sh_frame = NULL;
-        afr_local_t  *sh_local = NULL;
+        local = frame->local;
+        orig_sh = &local->self_heal;
+        priv  = this->private;
+
+        GF_ASSERT (local->loc.path);
+
+        gf_log (this->name, GF_LOG_TRACE,
+                "performing self heal on %s (metadata=%d data=%d entry=%d)",
+                local->loc.path,
+                local->self_heal.do_metadata_self_heal,
+                local->self_heal.do_data_self_heal,
+                local->self_heal.do_entry_self_heal);
+
+        op_errno        = ENOMEM;
+        sh_frame        = copy_frame (frame);
+        if (!sh_frame)
+                goto out;
+        afr_set_lk_owner (sh_frame, this, sh_frame->root);
+        afr_set_low_priority (sh_frame);
+
+        sh_local        = afr_self_heal_local_init (local, this);
+        if (!sh_local)
+                goto out;
+        sh_frame->local = sh_local;
+        sh              = &sh_local->self_heal;
+
+        sh->inode       = inode_ref (inode);
+        sh->orig_frame  = frame;
 
+        sh->completion_cbk = afr_self_heal_completion_cbk;
 
-	local = frame->local;
-	priv  = this->private;
+        sh->success = GF_CALLOC (priv->child_count, sizeof (*sh->success),
+                                 gf_afr_mt_char);
+        if (!sh->success)
+                goto out;
+        sh->sources = GF_CALLOC (sizeof (*sh->sources), priv->child_count,
+                                 gf_afr_mt_int);
+        if (!sh->sources)
+                goto out;
+        sh->locked_nodes = GF_CALLOC (sizeof (*sh->locked_nodes),
+                                      priv->child_count,
+                                      gf_afr_mt_int);
+        if (!sh->locked_nodes)
+                goto out;
 
-        afr_set_lk_owner (frame, this);
+        sh->pending_matrix = afr_matrix_create (priv->child_count,
+                                                priv->child_count);
+        if (!sh->pending_matrix)
+                goto out;
+
+        sh->delta_matrix = afr_matrix_create (priv->child_count,
+                                              priv->child_count);
+        if (!sh->delta_matrix)
+                goto out;
+
+        sh->fresh_parent_dirs = afr_children_create (priv->child_count);
+        if (!sh->fresh_parent_dirs)
+                goto out;
+        ret = afr_sh_common_create (sh, priv->child_count);
+        if (ret) {
+                op_errno = -ret;
+                goto out;
+        }
 
         if (local->self_heal.background) {
                 LOCK (&priv->lock);
                 {
                         if (priv->background_self_heals_started
-                            > priv->background_self_heal_count) {
+                            < priv->background_self_heal_count) {
+                                priv->background_self_heals_started++;
 
-                                local->self_heal.background = _gf_false;
 
                         } else {
-                                priv->background_self_heals_started++;
+                                local->self_heal.background = _gf_false;
+                                sh->background = _gf_false;
                         }
                 }
                 UNLOCK (&priv->lock);
         }
 
-        gf_log (this->name, GF_LOG_TRACE,
-                "performing self heal on %s (metadata=%d data=%d entry=%d)",
-                local->loc.path,
-                local->self_heal.need_metadata_self_heal,
-                local->self_heal.need_data_self_heal,
-                local->self_heal.need_entry_self_heal);
+        if (!local->loc.parent) {
+                sh->do_missing_entry_self_heal = _gf_false;
+                sh->do_gfid_self_heal = _gf_false;
+        }
 
-        sh_frame        = copy_frame (frame);
-        sh_local        = afr_local_copy (local, this);
-        sh_frame->local = sh_local;
-        sh              = &sh_local->self_heal;
+        sh->sh_type_in_action = AFR_SELF_HEAL_INVALID;
 
-        sh->orig_frame  = frame;
+        FRAME_SU_DO (sh_frame, afr_local_t);
+        if (sh->do_missing_entry_self_heal || sh->do_gfid_self_heal) {
+                afr_self_heal_missing_entries (sh_frame, this);
+        } else {
+                loc = &sh_local->loc;
+                if (uuid_is_null (loc->inode->gfid) && uuid_is_null (loc->gfid)) {
+                        if (!uuid_is_null (inode->gfid))
+                                GF_ASSERT (!uuid_compare (inode->gfid,
+                                           sh->sh_gfid_req));
+                        uuid_copy (loc->gfid, sh->sh_gfid_req);
+                }
+                gf_log (this->name, GF_LOG_TRACE,
+                        "proceeding to metadata check on %s",
+                        local->loc.path);
 
-        sh->completion_cbk = afr_self_heal_completion_cbk;
+                afr_sh_missing_entries_done (sh_frame, this);
+        }
+        op_errno = 0;
 
+out:
+        if (op_errno) {
+                orig_sh->unwind (frame, this, -1, op_errno, 1);
+                if (sh_frame)
+                        AFR_STACK_DESTROY (sh_frame);
+        }
+        return 0;
+}
 
-	sh->buf = GF_CALLOC (priv->child_count, sizeof (struct iatt),
-                             gf_afr_mt_iatt);
-	sh->child_errno = GF_CALLOC (priv->child_count, sizeof (int),
-                                     gf_afr_mt_int);
-	sh->success = GF_CALLOC (priv->child_count, sizeof (int),
-                                gf_afr_mt_int);
-	sh->xattr = GF_CALLOC (priv->child_count, sizeof (dict_t *),
-                                gf_afr_mt_dict_t);
-	sh->sources = GF_CALLOC (sizeof (*sh->sources), priv->child_count,
-                                gf_afr_mt_int);
-	sh->locked_nodes = GF_CALLOC (sizeof (*sh->locked_nodes), 
-                                      priv->child_count,
-                                      gf_afr_mt_int);
+void
+afr_self_heal_type_str_get (afr_self_heal_t *self_heal_p, char *str,
+                            size_t size)
+{
+        GF_ASSERT (str && (size > strlen (" missing-entry gfid "
+                                          "meta-data data entry")));
 
-	sh->pending_matrix = GF_CALLOC (sizeof (int32_t *), priv->child_count,
-                                        gf_afr_mt_int32_t);
+        if (self_heal_p->do_metadata_self_heal) {
+                snprintf (str, size, " meta-data");
+        }
 
-	for (i = 0; i < priv->child_count; i++) {
-		sh->pending_matrix[i] = GF_CALLOC (sizeof (int32_t),
-					  	   priv->child_count,
-                                                   gf_afr_mt_int32_t);
-	}
+        if (self_heal_p->do_data_self_heal) {
+                snprintf (str + strlen(str), size - strlen(str), " data");
+        }
 
-	sh->delta_matrix = GF_CALLOC (sizeof (int32_t *), priv->child_count,
-                                      gf_afr_mt_int32_t);
-	for (i = 0; i < priv->child_count; i++) {
-		sh->delta_matrix[i] = GF_CALLOC (sizeof (int32_t),
-					         priv->child_count,
-                                                 gf_afr_mt_int32_t);
-	}
+        if (self_heal_p->do_entry_self_heal) {
+                snprintf (str + strlen(str), size - strlen(str), " entry");
+        }
 
-	if (local->success_count && local->enoent_count) {
-		afr_self_heal_missing_entries (sh_frame, this);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"proceeding to metadata check on %s",
-			local->loc.path);
+        if (self_heal_p->do_missing_entry_self_heal) {
+                snprintf (str + strlen(str), size - strlen(str),
+                         " missing-entry");
+        }
 
-		afr_sh_missing_entries_done (sh_frame, this);
-	}
+        if (self_heal_p->do_gfid_self_heal) {
+                snprintf (str + strlen(str), size - strlen(str), " gfid");
+        }
+}
 
-	return 0;
+afr_self_heal_type
+afr_self_heal_type_for_transaction (afr_transaction_type type)
+{
+        afr_self_heal_type sh_type = AFR_SELF_HEAL_INVALID;
+
+        switch (type) {
+        case AFR_DATA_TRANSACTION:
+                sh_type = AFR_SELF_HEAL_DATA;
+                break;
+        case AFR_METADATA_TRANSACTION:
+                sh_type = AFR_SELF_HEAL_METADATA;
+                break;
+        case AFR_ENTRY_TRANSACTION:
+                sh_type = AFR_SELF_HEAL_ENTRY;
+                break;
+        case AFR_ENTRY_RENAME_TRANSACTION:
+                GF_ASSERT (0);
+                break;
+        }
+        return sh_type;
+}
+
+int
+afr_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
+{
+        int   ret = -1;
+        uuid_t pargfid = {0};
+
+        if (!child)
+                goto out;
+
+        if (!uuid_is_null (parent->inode->gfid))
+                uuid_copy (pargfid, parent->inode->gfid);
+        else if (!uuid_is_null (parent->gfid))
+                uuid_copy (pargfid, parent->gfid);
+
+        if (uuid_is_null (pargfid))
+                goto out;
+
+        if (strcmp (parent->path, "/") == 0)
+                ret = gf_asprintf ((char **)&child->path, "/%s", name);
+        else
+                ret = gf_asprintf ((char **)&child->path, "%s/%s", parent->path,
+                                   name);
+
+        if (-1 == ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "asprintf failed while setting child path");
+        }
+
+        child->name = strrchr (child->path, '/');
+        if (child->name)
+                child->name++;
+
+        child->parent = inode_ref (parent->inode);
+        child->inode = inode_new (parent->inode->table);
+        uuid_copy (child->pargfid, pargfid);
+
+        if (!child->inode) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if ((ret == -1) && child)
+                loc_wipe (child);
+
+        return ret;
+}
+
+int
+afr_sh_erase_pending (call_frame_t *frame, xlator_t *this,
+                      afr_transaction_type type, afr_fxattrop_cbk_t cbk,
+                      int (*finish)(call_frame_t *frame, xlator_t *this))
+{
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              call_count = 0;
+        int              i = 0;
+        dict_t          **erase_xattr = NULL;
+        int             ret = -1;
+
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
+
+        afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix,
+                                 sh->success, priv->child_count, type);
+
+        erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
+                                 gf_afr_mt_dict_t);
+        if (!erase_xattr)
+                goto out;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (sh->xattr[i]) {
+                        call_count++;
+                        erase_xattr[i] = dict_new ();
+                        if (!erase_xattr[i])
+                                goto out;
+                }
+        }
+
+        afr_sh_delta_to_xattr (this, sh->delta_matrix, erase_xattr,
+                               priv->child_count, type);
+
+        gf_log (this->name, GF_LOG_DEBUG, "Delta matrix for: %s",
+                lkowner_utoa (&frame->root->lk_owner));
+        afr_sh_print_pending_matrix (sh->delta_matrix, this);
+        local->call_count = call_count;
+        if (call_count == 0) {
+                ret = 0;
+                finish (frame, this);
+                goto out;
+        }
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (!erase_xattr[i])
+                        continue;
+
+                if (sh->healing_fd) {//true for ENTRY, reg file DATA transaction
+                        STACK_WIND_COOKIE (frame, cbk, (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->fxattrop,
+                                           sh->healing_fd,
+                                           GF_XATTROP_ADD_ARRAY, erase_xattr[i],
+                                           NULL);
+                } else {
+                        STACK_WIND_COOKIE (frame, cbk, (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->xattrop,
+                                           &local->loc,
+                                           GF_XATTROP_ADD_ARRAY, erase_xattr[i],
+                                           NULL);
+                }
+        }
+
+        ret = 0;
+out:
+        if (erase_xattr) {
+                for (i = 0; i < priv->child_count; i++) {
+                        if (erase_xattr[i]) {
+                                dict_unref (erase_xattr[i]);
+                        }
+                }
+        }
+
+        GF_FREE (erase_xattr);
+
+        if (ret < 0) {
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                finish (frame, this);
+        }
+
+        return 0;
 }
 
 void
-afr_self_heal_type_str_get (afr_self_heal_t *self_heal_p, char *str,
-                            size_t size)
+afr_set_self_heal_status(afr_self_heal_t *sh, afr_self_heal_status status)
+{
+        xlator_t                *this = NULL;
+        afr_sh_status_for_all_type *sh_status = &(sh->afr_all_sh_status);
+        afr_self_heal_type  sh_type_in_action = sh->sh_type_in_action;
+        this = THIS;
+
+        if (!sh) {
+                gf_log_callingfn (this->name, GF_LOG_ERROR, "Null self heal"
+                                  "Structure");
+                goto out;
+        }
+
+        switch (sh_type_in_action) {
+                case AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY:
+                       sh_status->gfid_or_missing_entry_self_heal = status;
+                        break;
+                case AFR_SELF_HEAL_METADATA:
+                        sh_status->metadata_self_heal = status;
+                        break;
+                case AFR_SELF_HEAL_DATA:
+                        sh_status->data_self_heal = status;
+                        break;
+                case AFR_SELF_HEAL_ENTRY:
+                        sh_status->entry_self_heal = status;
+                        break;
+                case AFR_SELF_HEAL_INVALID:
+                        gf_log_callingfn (this->name, GF_LOG_ERROR, "Invalid"
+                                          "self heal type in action");
+                        break;
+        }
+out:
+        return;
+}
+
+void
+afr_set_local_for_unhealable (afr_local_t *local)
+{
+        afr_self_heal_t  *sh = NULL;
+
+        sh = &local->self_heal;
+
+        local->unhealable = 1;
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+}
+
+int
+is_self_heal_failed (afr_self_heal_t *sh, afr_sh_fail_check_type type)
 {
-        assert(str && (size > 0));
+        afr_sh_status_for_all_type      sh_status = sh->afr_all_sh_status;
+        afr_self_heal_type   sh_type_in_action =  AFR_SELF_HEAL_INVALID;
+        afr_self_heal_status    status = AFR_SELF_HEAL_FAILED;
+        xlator_t                *this = NULL;
+        int                     sh_failed = 0;
+
+        this = THIS;
+
+        if (!sh) {
+                gf_log_callingfn (this->name, GF_LOG_ERROR, "Null self heal "
+                                  "structure");
+                sh_failed = 1;
+                goto out;
+        }
+
+        if (type == AFR_CHECK_ALL) {
+                if ((sh_status.gfid_or_missing_entry_self_heal == AFR_SELF_HEAL_FAILED)
+                    || (sh_status.metadata_self_heal == AFR_SELF_HEAL_FAILED)
+                    || (sh_status.data_self_heal == AFR_SELF_HEAL_FAILED)
+                    || (sh_status.entry_self_heal == AFR_SELF_HEAL_FAILED))
+                sh_failed = 1;
+        } else if (type == AFR_CHECK_SPECIFIC) {
+                sh_type_in_action = sh->sh_type_in_action;
+                switch (sh_type_in_action) {
+                        case AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY:
+                             status = sh_status.gfid_or_missing_entry_self_heal;
+                                break;
+                        case AFR_SELF_HEAL_METADATA:
+                                status = sh_status.metadata_self_heal;
+                                break;
+                        case AFR_SELF_HEAL_ENTRY:
+                                status = sh_status.entry_self_heal;
+                                break;
+                        case AFR_SELF_HEAL_DATA:
+                                status = sh_status.data_self_heal;
+                                break;
+                        case AFR_SELF_HEAL_INVALID:
+                                status = AFR_SELF_HEAL_NOT_ATTEMPTED;
+                                break;
+                }
+                if (status == AFR_SELF_HEAL_FAILED)
+                        sh_failed = 1;
 
-        if (self_heal_p->need_metadata_self_heal) {
-                snprintf(str, size, " meta-data");
         }
 
-        if (self_heal_p->need_data_self_heal) {
-                snprintf(str + strlen(str), size - strlen(str),
-                         " data");
+out:
+        return sh_failed;
+}
+
+char *
+get_sh_completion_status (afr_self_heal_status status)
+{
+
+        char *not_attempted       = " is not attempted";
+        char *failed              = " failed";
+        char *started             = " is started";
+        char *sync_begin          = " is successfully completed";
+        char *result              = " has unknown status";
+
+        switch (status)
+        {
+                case AFR_SELF_HEAL_NOT_ATTEMPTED:
+                        result = not_attempted;
+                        break;
+                case AFR_SELF_HEAL_FAILED:
+                        result = failed;
+                        break;
+                case AFR_SELF_HEAL_STARTED:
+                        result = started;
+                        break;
+                case AFR_SELF_HEAL_SYNC_BEGIN:
+                        result = sync_begin;
+                        break;
         }
 
-        if (self_heal_p->need_entry_self_heal) {
-                snprintf(str + strlen(str), size - strlen(str),
-                         " entry");
+        return result;
+
+}
+
+void
+afr_log_self_heal_completion_status (afr_local_t *local, gf_loglevel_t loglvl)
+{
+
+        char sh_log[4096]              = {0};
+        afr_self_heal_t *sh            = &local->self_heal;
+        afr_sh_status_for_all_type   all_status = sh->afr_all_sh_status;
+        xlator_t      *this            = NULL;
+        size_t        off              = 0;
+        int           data_sh          = 0;
+        int           metadata_sh      = 0;
+        int           print_log        = 0;
+
+        this = THIS;
+
+        ADD_FMT_STRING (sh_log, off, "gfid or missing entry",
+                        all_status.gfid_or_missing_entry_self_heal, print_log);
+        ADD_FMT_STRING_SYNC (sh_log, off, "metadata",
+                             all_status.metadata_self_heal, print_log);
+        if (sh->background) {
+                ADD_FMT_STRING_SYNC (sh_log, off, "backgroung data",
+                                all_status.data_self_heal, print_log);
+        } else {
+                ADD_FMT_STRING_SYNC (sh_log, off, "foreground data",
+                                all_status.data_self_heal, print_log);
         }
+        ADD_FMT_STRING_SYNC (sh_log, off, "entry", all_status.entry_self_heal,
+                             print_log);
+
+        if (AFR_SELF_HEAL_SYNC_BEGIN == all_status.data_self_heal &&
+	    strcmp (sh->data_sh_info, "") && sh->data_sh_info )
+                data_sh = 1;
+        if (AFR_SELF_HEAL_SYNC_BEGIN == all_status.metadata_self_heal &&
+	    strcmp (sh->metadata_sh_info, "") && sh->metadata_sh_info)
+                metadata_sh = 1;
+
+        if (!print_log)
+                return;
+
+        gf_log (this->name, loglvl, "%s %s %s on %s", sh_log,
+                ((data_sh == 1) ? sh->data_sh_info : ""),
+                ((metadata_sh == 1) ? sh->metadata_sh_info : ""),
+                local->loc.path);
 }
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.h b/xlators/cluster/afr/src/afr-self-heal-common.h
index 31f02f3f9..473264776 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.h
+++ b/xlators/cluster/afr/src/afr-self-heal-common.h
@@ -1,73 +1,144 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __AFR_SELF_HEAL_COMMON_H__
 #define __AFR_SELF_HEAL_COMMON_H__
 
 #define FILE_HAS_HOLES(buf) (((buf)->ia_size) > ((buf)->ia_blocks * 512))
+#define AFR_SH_MIN_PARTICIPANTS 2
 
 typedef enum {
-        AFR_SELF_HEAL_ENTRY,
-        AFR_SELF_HEAL_METADATA,
-        AFR_SELF_HEAL_DATA,
-} afr_self_heal_type;
+        AFR_LOOKUP_FAIL_CONFLICTS = 1,
+        AFR_LOOKUP_FAIL_MISSING_GFIDS = 2,
+} afr_lookup_flags_t;
 
 int
 afr_sh_select_source (int sources[], int child_count);
 
 int
-afr_sh_sink_count (int sources[], int child_count);
-
-int
 afr_sh_source_count (int sources[], int child_count);
 
-int
-afr_sh_supress_errenous_children (int sources[], int child_errno[],
-				  int child_count);
-
 void
 afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this);
 
 void
-afr_sh_build_pending_matrix (afr_private_t *priv,
-                             int32_t *pending_matrix[], dict_t *xattr[],
-			     int child_count, afr_transaction_type type);
+afr_sh_print_split_brain_log (int32_t *pending_matrix[], xlator_t *this,
+                              const char *loc);
+
+int
+afr_build_pending_matrix (char **pending_key, int32_t **pending_matrix,
+                          unsigned char *ignorant_subvols,
+                          dict_t *xattr[], afr_transaction_type type,
+                          size_t child_count);
 
 void
 afr_sh_pending_to_delta (afr_private_t *priv, dict_t **xattr,
-                         int32_t *delta_matrix[], int success[],
+                         int32_t *delta_matrix[], unsigned char success[],
                          int child_count, afr_transaction_type type);
 
 int
-afr_sh_mark_sources (afr_self_heal_t *sh, int child_count,
-                     afr_self_heal_type type);
+afr_mark_sources (xlator_t *this, int32_t *sources, int32_t **pending_matrix,
+                  struct iatt *bufs, afr_self_heal_type type,
+                  int32_t *success_children, int32_t *subvol_status);
 
 int
-afr_sh_delta_to_xattr (afr_private_t *priv,
+afr_sh_delta_to_xattr (xlator_t *this,
                        int32_t *delta_matrix[], dict_t *xattr[],
 		       int child_count, afr_transaction_type type);
 
-int
-afr_sh_is_matrix_zero (int32_t *pending_matrix[], int child_count);
-
 void
 afr_self_heal_type_str_get (afr_self_heal_t *self_heal_p, char *str,
                             size_t size);
 
+afr_self_heal_type
+afr_self_heal_type_for_transaction (afr_transaction_type type);
+
+int
+afr_build_sources (xlator_t *this, dict_t **xattr, struct iatt *bufs,
+                   int32_t **pending_matrix, int32_t *sources,
+                   int32_t *success_children, afr_transaction_type type,
+                   int32_t *subvol_status, gf_boolean_t ignore_ignorant);
+void
+afr_sh_common_reset (afr_self_heal_t *sh, unsigned int child_count);
+
+void
+afr_sh_common_lookup_resp_handler (call_frame_t *frame, void *cookie,
+                                   xlator_t *this,
+                                   int32_t op_ret, int32_t op_errno,
+                                   inode_t *inode, struct iatt *buf,
+                                   dict_t *xattr, struct iatt *postparent,
+                                   loc_t *loc);
+
+int
+afr_sh_common_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                      afr_lookup_done_cbk_t lookup_cbk, uuid_t uuid,
+                      int32_t flags, dict_t *xdata);
+int
+afr_sh_entry_expunge_remove (call_frame_t *expunge_frame, xlator_t *this,
+                             int active_src, struct iatt *buf,
+                             struct iatt *parentbuf);
+int
+afr_sh_entrylk (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                char *base_name, afr_lock_cbk_t lock_cbk);
+int
+afr_sh_entry_impunge_create (call_frame_t *impunge_frame, xlator_t *this,
+                             int child_index);
+int
+afr_sh_data_unlock (call_frame_t *frame, xlator_t *this, char *dom,
+                    afr_lock_cbk_t lock_cbk);
+afr_local_t *
+afr_self_heal_local_init (afr_local_t *l, xlator_t *this);
+int
+afr_sh_data_lock (call_frame_t *frame, xlator_t *this,
+                  off_t start, off_t len, gf_boolean_t block, char *dom,
+                  afr_lock_cbk_t success_handler,
+                  afr_lock_cbk_t failure_handler);
+void
+afr_sh_set_error (afr_self_heal_t *sh, int32_t op_errno);
+void
+afr_sh_mark_source_sinks (call_frame_t *frame, xlator_t *this);
+typedef int
+(*afr_fxattrop_cbk_t) (call_frame_t *frame, void *cookie,
+                       xlator_t *this, int32_t op_ret, int32_t op_errno,
+                       dict_t *xattr, dict_t *xdata);
+int
+afr_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name);
+int
+afr_impunge_frame_create (call_frame_t *frame, xlator_t *this,
+                          int active_source, call_frame_t **impunge_frame);
+void
+afr_sh_reset (call_frame_t *frame, xlator_t *this);
+
+void
+afr_children_intersection_get (int32_t *set1, int32_t *set2,
+                               int *intersection, unsigned int child_count);
+int
+afr_get_no_xattr_dir_read_child (xlator_t *this, int32_t *success_children,
+                                 struct iatt *bufs);
+int
+afr_sh_erase_pending (call_frame_t *frame, xlator_t *this,
+                      afr_transaction_type type, afr_fxattrop_cbk_t cbk,
+                      int (*finish)(call_frame_t *frame, xlator_t *this));
+
+void
+afr_set_local_for_unhealable (afr_local_t *local);
+
+int
+is_self_heal_failed (afr_self_heal_t *sh, afr_sh_fail_check_type type);
+
+void
+afr_set_self_heal_status (afr_self_heal_t *sh, afr_self_heal_status status);
+
+void
+afr_log_self_heal_completion_status (afr_local_t *local, gf_loglevel_t  logl);
+
+char*
+afr_get_pending_matrix_str (int32_t *pending_matrix[], xlator_t *this);
 #endif /* __AFR_SELF_HEAL_COMMON_H__ */
diff --git a/xlators/cluster/afr/src/afr-self-heal-data.c b/xlators/cluster/afr/src/afr-self-heal-data.c
index 0398d21b1..9de26ee56 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <libgen.h>
@@ -49,421 +40,476 @@
 #include "afr-self-heal-common.h"
 #include "afr-self-heal-algorithm.h"
 
-
 int
-afr_sh_data_done (call_frame_t *frame, xlator_t *this)
+afr_sh_data_fail (call_frame_t *frame, xlator_t *this);
+
+static inline gf_boolean_t
+afr_sh_data_proceed (unsigned int success_count)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
+        return (success_count >= AFR_SH_MIN_PARTICIPANTS);
+}
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+extern int
+sh_loop_finish (call_frame_t *loop_frame, xlator_t *this);
 
-	/* 
-	   TODO: cleanup sh->* 
-	 */
+int
+afr_post_sh_big_lock_success (call_frame_t *frame, xlator_t *this);
 
-        if (sh->healing_fd && !sh->healing_fd_opened) {
-                /* unref only if we created the fd ourselves */
+int
+afr_post_sh_big_lock_failure (call_frame_t *frame, xlator_t *this);
 
-		fd_unref (sh->healing_fd);
-		sh->healing_fd = NULL;
-        }
+int
+afr_sh_data_finish (call_frame_t *frame, xlator_t *this);
 
-/*         for (i = 0; i < priv->child_count; i++) */
-/*                 sh->locked_nodes[i] = 0; */
+int
+afr_sh_data_done (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"self heal of %s completed",
-		local->loc.path);
+        local = frame->local;
+        sh = &local->self_heal;
 
-	sh->completion_cbk (frame, this);
+        sh->completion_cbk (frame, this);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_data_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno)
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_local_t     *local = NULL;
-	afr_private_t   *priv  = NULL;
-	int              call_count = 0;
-
-        int child_index = (long) cookie;
+        afr_local_t   *local       = NULL;
+        afr_private_t *priv        = NULL;
+        int            call_count  = 0;
+        int            child_index = (long) cookie;
 
-	local = frame->local;
-	priv = this->private;
+        local = frame->local;
+        priv = this->private;
 
-	LOCK (&frame->lock);
-	{
+        LOCK (&frame->lock);
+        {
                 if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "flush or setattr failed on %s on subvolume %s: %s",
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "flush failed on %s on subvolume %s: %s",
                                 local->loc.path, priv->children[child_index]->name,
                                 strerror (op_errno));
                 }
-	}
-	UNLOCK (&frame->lock);
+        }
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		afr_sh_data_done (frame, this);
-	}
+        if (call_count == 0) {
+                afr_sh_data_done (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
-
 int
-afr_sh_data_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno, struct iatt *statpre, struct iatt *statpost)
+afr_sh_data_close (call_frame_t *frame, xlator_t *this)
 {
-        afr_sh_data_flush_cbk (frame, cookie, this, op_ret, op_errno);
+        afr_local_t     *local      = NULL;
+        afr_private_t   *priv       = NULL;
+        afr_self_heal_t *sh         = NULL;
+        int              i          = 0;
+        int              call_count = 0;
+
+        local = frame->local;
+        sh    = &local->self_heal;
+        priv  = this->private;
+
+        if (!sh->healing_fd) {
+                //This happens when file is non-reg
+                afr_sh_data_done (frame, this);
+                return 0;
+        }
+        call_count        = afr_set_elem_count_get (sh->success,
+                                                    priv->child_count);
+        local->call_count = call_count;
+
+        if (call_count == 0) {
+                afr_sh_data_done (frame, this);
+                return 0;
+        }
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (!sh->success[i])
+                        continue;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "closing fd of %s on %s",
+                        local->loc.path, priv->children[i]->name);
+
+                STACK_WIND_COOKIE (frame, afr_sh_data_flush_cbk,
+                                   (void *) (long) i,
+                                   priv->children[i],
+                                   priv->children[i]->fops->flush,
+                                   sh->healing_fd, NULL);
+
+                if (!--call_count)
+                        break;
+        }
 
         return 0;
 }
 
-
 int
-afr_sh_data_close (call_frame_t *frame, xlator_t *this)
+afr_sh_dom_unlock (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t     *local = NULL;
-	afr_private_t   *priv  = NULL;
-	afr_self_heal_t *sh    = NULL;
-
-	int  i            = 0;
-	int  call_count   = 0;
-        int  source       = 0;
-        int32_t valid     = 0;
-
-        struct iatt stbuf = {0,};
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh    = NULL;
+        afr_private_t   *priv  = NULL;
 
         local = frame->local;
         sh    = &local->self_heal;
         priv  = this->private;
 
-        source       = sh->source;
+        if (sh->sh_dom_lock_held)
+                afr_sh_data_unlock (frame, this, priv->sh_domain,
+                                    afr_sh_data_close);
+        else
+                afr_sh_data_close (frame, this);
+        return 0;
+}
 
-        valid |= (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME);
+int
+afr_sh_data_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                         struct iatt *statpost, dict_t *xdata)
+{
 
-	stbuf.ia_atime = sh->buf[source].ia_atime;
-	stbuf.ia_atime_nsec = sh->buf[source].ia_atime_nsec;
-	stbuf.ia_mtime = sh->buf[source].ia_mtime;
-	stbuf.ia_mtime_nsec = sh->buf[source].ia_mtime_nsec;
+        afr_local_t   *local       = NULL;
+        afr_private_t *priv        = NULL;
+        int            call_count  = 0;
+        int            child_index = (long) cookie;
 
-        if (sh->healing_fd_opened) {
-                /* not our job to close the fd */
+        local = frame->local;
+        priv = this->private;
 
-                afr_sh_data_done (frame, this);
-                return 0;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "setattr failed on %s on subvolume %s: %s",
+                                local->loc.path, priv->children[child_index]->name,
+                                strerror (op_errno));
+                }
         }
+        UNLOCK (&frame->lock);
 
-	if (!sh->healing_fd) {
-		afr_sh_data_done (frame, this);
-		return 0;
-	}
-
-	call_count        = (sh->active_sinks + 1) * 2;
-	local->call_count = call_count;
+        call_count = afr_frame_return (frame);
 
-	/* closed source */
-	gf_log (this->name, GF_LOG_TRACE,
-		"closing fd of %s on %s",
-		local->loc.path, priv->children[sh->source]->name);
-
-	STACK_WIND_COOKIE (frame, afr_sh_data_flush_cbk,
-			   (void *) (long) sh->source,
-			   priv->children[sh->source],
-			   priv->children[sh->source]->fops->flush,
-			   sh->healing_fd);
-	call_count--;
+        if (call_count == 0) {
+                afr_sh_data_finish (frame, this);
+        }
 
-        STACK_WIND_COOKIE (frame, afr_sh_data_setattr_cbk,
-                           (void *) (long) sh->source,
-                           priv->children[sh->source],
-                           priv->children[sh->source]->fops->setattr,
-                           &local->loc, &stbuf, valid);
+        return 0;
+}
 
-        call_count--;
+int
+afr_sh_data_setattr (call_frame_t *frame, xlator_t *this, struct iatt* stbuf)
+{
+        afr_local_t     *local      = NULL;
+        afr_private_t   *priv       = NULL;
+        afr_self_heal_t *sh         = NULL;
+        int              i          = 0;
+        int              call_count = 0;
+        int32_t          valid      = 0;
 
-        if (call_count == 0)
-                return 0;
+        local = frame->local;
+        sh    = &local->self_heal;
+        priv  = this->private;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] || !local->child_up[i])
-			continue;
+        valid = (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME);
 
-		gf_log (this->name, GF_LOG_TRACE,
-			"closing fd of %s on %s",
-			local->loc.path, priv->children[i]->name);
+        call_count        = afr_set_elem_count_get (sh->success,
+                                                    priv->child_count);
+        local->call_count = call_count;
 
-		STACK_WIND_COOKIE (frame, afr_sh_data_flush_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->flush,
-				   sh->healing_fd);
+        if (call_count == 0) {
+                GF_ASSERT (0);
+                afr_sh_data_finish (frame, this);
+                return 0;
+        }
 
-                call_count--;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!sh->success[i])
+                        continue;
 
                 STACK_WIND_COOKIE (frame, afr_sh_data_setattr_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->setattr,
-				   &local->loc, &stbuf, valid);
+                                   (void *) (long) i,
+                                   priv->children[i],
+                                   priv->children[i]->fops->setattr,
+                                   &local->loc, stbuf, valid, NULL);
 
-		if (!--call_count)
-			break;
-	}
+                if (!--call_count)
+                        break;
+        }
 
-	return 0;
+        return 0;
 }
 
-
 int
-afr_sh_data_unlck_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno)
+afr_sh_data_setattr_fstat_cbk (call_frame_t *frame, void *cookie,
+                               xlator_t *this, int32_t op_ret, int32_t op_errno,
+                               struct iatt *buf, dict_t *xdata)
 {
-	afr_local_t * local = NULL;
-	int           call_count = 0;
-	int           child_index = (long) cookie;
-
-	
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"locking inode of %s on child %d failed: %s",
-				local->loc.path, child_index,
-				strerror (op_errno));
-		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"inode of %s on child %d locked",
-				local->loc.path, child_index);
-		}
-	}
-	UNLOCK (&frame->lock);
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        int child_index = (long) cookie;
 
-	call_count = afr_frame_return (frame);
+        local = frame->local;
+        sh = &local->self_heal;
 
-	if (call_count == 0) {
-		afr_sh_data_close (frame, this);
-	}
+        GF_ASSERT (sh->source == child_index);
+        if (op_ret != -1) {
+                sh->buf[child_index] = *buf;
+                afr_sh_data_setattr (frame, this, buf);
+        } else {
+                gf_log (this->name, GF_LOG_ERROR, "%s: Failed to set "
+                        "time-stamps after self-heal", local->loc.path);
+                afr_sh_data_fail (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
+/*
+ * If there are any writes after the self-heal is triggered then the
+ * stbuf stored in local->self_heal.buf[] will be invalid so we do one more
+ * stat on the source and then set the [am]times
+ */
+int
+afr_sh_set_timestamps (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local      = NULL;
+        afr_private_t   *priv       = NULL;
+        afr_self_heal_t *sh         = NULL;
+
+        local = frame->local;
+        sh    = &local->self_heal;
+        priv  = this->private;
+
+        STACK_WIND_COOKIE (frame, afr_sh_data_setattr_fstat_cbk,
+                           (void *) (long) sh->source,
+                           priv->children[sh->source],
+                           priv->children[sh->source]->fops->fstat,
+                           sh->healing_fd, NULL);
+        return 0;
+}
 
+//Fun fact, lock_cbk is being used for both lock & unlock
 int
-afr_sh_data_unlock (call_frame_t *frame, xlator_t *this)
+afr_sh_data_unlock (call_frame_t *frame, xlator_t *this, char *dom,
+                    afr_lock_cbk_t lock_cbk)
 {
         afr_local_t         *local    = NULL;
         afr_internal_lock_t *int_lock = NULL;
-	afr_self_heal_t     *sh       = NULL;
+        afr_self_heal_t     *sh       = NULL;
+        afr_private_t       *priv     = NULL;
+        int                 ret       = 0;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
-	sh       = &local->self_heal;
-
-        GF_ASSERT (!sh->data_lock_held);
+        sh       = &local->self_heal;
+        priv     = this->private;
 
-        int_lock->lock_cbk = afr_sh_data_close;
+        if (strcmp (dom, this->name) == 0) {
+                sh->data_lock_held = _gf_false;
+        } else if (strcmp (dom, priv->sh_domain) == 0) {
+                sh->sh_dom_lock_held = _gf_false;
+        } else {
+                ret = -1;
+                goto out;
+        }
+        int_lock->lock_cbk = lock_cbk;
+        int_lock->domain = dom;
         afr_unlock (frame, this);
 
-	return 0;
+out:
+        if (ret) {
+                int_lock->lock_op_ret = -1;
+                int_lock->lock_cbk (frame, this);
+        }
+        return 0;
 }
 
-
 int
 afr_sh_data_finish (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t   *local = NULL;
-	afr_self_heal_t *sh = NULL;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh    = NULL;
 
-	local = frame->local;
-	sh = &local->self_heal;
+        local = frame->local;
+        sh = &local->self_heal;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"finishing data selfheal of %s", local->loc.path);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "finishing data selfheal of %s", local->loc.path);
 
-        if (!sh->data_lock_held)
-                afr_sh_data_unlock (frame, this);
+        if (sh->data_lock_held)
+                afr_sh_data_unlock (frame, this, this->name, afr_sh_dom_unlock);
         else
-                afr_sh_data_close (frame, this);
+                afr_sh_dom_unlock (frame, this);
 
-	return 0;
+        return 0;
 }
 
-
 int
-afr_sh_data_erase_pending_cbk (call_frame_t *frame, void *cookie,
-			       xlator_t *this, int32_t op_ret,
-			       int32_t op_errno, dict_t *xattr)
+afr_sh_data_fail (call_frame_t *frame, xlator_t *this)
 {
-	int             call_count = 0;
+        afr_local_t   *local = NULL;
+        afr_self_heal_t *sh = NULL;
 
-	call_count = afr_frame_return (frame);
+        local = frame->local;
+        sh = &local->self_heal;
 
-	if (call_count == 0)
-		afr_sh_data_finish (frame, this);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "finishing failed data selfheal of %s", local->loc.path);
 
-	return 0;
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+        afr_sh_data_finish (frame, this);
+        return 0;
 }
 
-
 int
-afr_sh_data_erase_pending (call_frame_t *frame, xlator_t *this)
+afr_sh_data_erase_pending_cbk (call_frame_t *frame, void *cookie,
+                               xlator_t *this, int32_t op_ret,
+                               int32_t op_errno, dict_t *xattr, dict_t *xdata)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              i = 0;
-	dict_t          **erase_xattr = NULL;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, sh->success,
-                                 priv->child_count, AFR_DATA_TRANSACTION);
-
-	erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
-                              gf_afr_mt_dict_t);
+        int             call_count = 0;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int32_t         child_index = (long) cookie;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i]) {
-			call_count++;
-
-			erase_xattr[i] = get_new_dict();
-			dict_ref (erase_xattr[i]);
-		}
-	}
-
-	afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
-			       priv->child_count, AFR_DATA_TRANSACTION);
-
-	local->call_count = call_count;
-	for (i = 0; i < priv->child_count; i++) {
-		if (!erase_xattr[i])
-			continue;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"erasing pending flags from %s on %s",
-			local->loc.path, priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_data_erase_pending_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->fxattrop,
-				   sh->healing_fd,
-				   GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
-		if (!--call_count)
-			break;
-	}
+        priv  = this->private;
+        local = frame->local;
+        sh    = &local->self_heal;
+        if (op_ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Erasing of pending change "
+                        "log failed on %s for subvol %s, reason: %s",
+                        local->loc.path, priv->children[child_index]->name,
+                        strerror (op_errno));
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+        }
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (erase_xattr[i]) {
-			dict_unref (erase_xattr[i]);
-		}
-	}
-	GF_FREE (erase_xattr);
+        call_count = afr_frame_return (frame);
 
-	return 0;
+        if (call_count == 0) {
+                if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+                        if (sh->old_loop_frame)
+                                sh_loop_finish (sh->old_loop_frame, this);
+                        sh->old_loop_frame = NULL;
+                        afr_sh_data_fail (frame, this);
+                        goto out;
+                }
+                if (!IA_ISREG (sh->type)) {
+                        afr_sh_data_finish (frame, this);
+                        goto out;
+                }
+                GF_ASSERT (sh->old_loop_frame);
+                afr_sh_data_lock (frame, this, 0, 0, _gf_true, this->name,
+                                  afr_post_sh_big_lock_success,
+                                  afr_post_sh_big_lock_failure);
+        }
+out:
+        return 0;
 }
 
-
 int
-afr_sh_data_trim_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                      struct iatt *postbuf)
+afr_sh_data_erase_pending (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t * priv = NULL;
-	afr_local_t * local  = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
-
-	priv = this->private;
-	local = frame->local;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1)
-			gf_log (this->name, GF_LOG_DEBUG,
-				"ftruncate of %s on subvolume %s failed (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-		else
-			gf_log (this->name, GF_LOG_TRACE,
-				"ftruncate of %s on subvolume %s completed",
-				local->loc.path,
-				priv->children[child_index]->name);
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		afr_sh_data_erase_pending (frame, this);
-	}
-
-	return 0;
+        afr_sh_erase_pending (frame, this, AFR_DATA_TRANSACTION,
+                              afr_sh_data_erase_pending_cbk,
+                              afr_sh_data_finish);
+        return 0;
 }
 
-
 int
-afr_sh_data_trim_sinks (call_frame_t *frame, xlator_t *this)
+afr_sh_data_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int op_ret, int op_errno, struct iatt *pre,
+                       struct iatt *post, dict_t *xdata)
 {
-	afr_private_t * priv = NULL;
-	afr_local_t * local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int             *sources = NULL;
-	int              call_count = 0;
-	int              i = 0;
+        afr_local_t     *local      = NULL;
+        afr_private_t   *priv       = NULL;
+        afr_self_heal_t *sh         = NULL;
+        int             call_count  = 0;
+        int             child_index = (long) cookie;
 
+        local = frame->local;
+        priv = this->private;
+        sh   = &local->self_heal;
+
+        if (op_ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "%s: Failed to fsync on "
+                        "%s - %s", local->loc.path,
+                        priv->children[child_index]->name, strerror (op_errno));
+                LOCK (&frame->lock);
+                {
+                        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                }
+                UNLOCK (&frame->lock);
+                if (sh->old_loop_frame)
+                        sh_loop_finish (sh->old_loop_frame, this);
+                sh->old_loop_frame = NULL;
+        }
 
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
+        call_count = afr_frame_return (frame);
+        if (call_count == 0) {
+                if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC))
+                        afr_sh_data_fail (frame, this);
+                else
+                        afr_sh_data_erase_pending (frame, this);
+        }
+        return 0;
+}
 
-	sources = sh->sources;
-	call_count = sh->active_sinks;
+/*
+ * Before erasing xattrs, make sure the data is written to disk
+ */
+int
+afr_sh_data_fsync (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local     = NULL;
+        afr_private_t   *priv      = NULL;
+        afr_self_heal_t *sh        = NULL;
+        int             i          = 0;
+        int             call_count = 0;
 
-	local->call_count = call_count;
+        local = frame->local;
+        priv = this->private;
+        sh   = &local->self_heal;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (sources[i] || !local->child_up[i])
-			continue;
+        call_count        = sh->active_sinks;
+        if (call_count == 0) {
+                afr_sh_data_erase_pending (frame, this);
+                return 0;
+        }
 
-		STACK_WIND_COOKIE (frame, afr_sh_data_trim_cbk,
-				   (void *) (long) i,
-				   priv->children[i], 
-				   priv->children[i]->fops->ftruncate,
-				   sh->healing_fd, sh->file_size); 
+        local->call_count = call_count;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!sh->success[i] || sh->sources[i])
+                        continue;
 
-		if (!--call_count)
-			break;
-	}
+                STACK_WIND_COOKIE (frame, afr_sh_data_fsync_cbk,
+                                   (void *) (long) i, priv->children[i],
+                                   priv->children[i]->fops->fsync,
+                                   sh->healing_fd, 1, NULL);
+        }
 
-	return 0;
+        return 0;
 }
 
-
 static struct afr_sh_algorithm *
 sh_algo_from_name (xlator_t *this, char *name)
 {
         int i = 0;
 
+        if (name == NULL)
+                goto out;
+
         while (afr_self_heal_algorithms[i].name) {
                 if (!strcmp (name, afr_self_heal_algorithms[i].name)) {
                         return &afr_self_heal_algorithms[i];
@@ -472,17 +518,22 @@ sh_algo_from_name (xlator_t *this, char *name)
                 i++;
         }
 
+out:
         return NULL;
 }
 
 
 static int
-sh_zero_byte_files_exist (afr_self_heal_t *sh, int child_count)
+sh_zero_byte_files_exist (afr_local_t *local, int child_count)
 {
-        int i;
-        int ret = 0;
+        int             i = 0;
+        int             ret = 0;
+        afr_self_heal_t *sh = NULL;
 
+        sh = &local->self_heal;
         for (i = 0; i < child_count; i++) {
+                if (!local->child_up[i] || sh->child_errno[i])
+                        continue;
                 if (sh->buf[i].ia_size == 0) {
                         ret = 1;
                         break;
@@ -509,9 +560,9 @@ afr_sh_data_pick_algo (call_frame_t *frame, xlator_t *this)
         if (algo == NULL) {
                 /* option not set, so fall back on heuristics */
 
-                if ((local->enoent_count != 0)
-                    || sh_zero_byte_files_exist (sh, priv->child_count)
-                    || (sh->file_size <= (priv->data_self_heal_window_size * this->ctx->page_size))) {
+                if (sh_zero_byte_files_exist (local, priv->child_count)
+                    || (sh->file_size <= (priv->data_self_heal_window_size *
+                                          this->ctx->page_size))) {
 
                         /*
                          * If the file does not exist on one of the subvolumes,
@@ -540,398 +591,860 @@ afr_sh_data_pick_algo (call_frame_t *frame, xlator_t *this)
 int
 afr_sh_data_sync_prepare (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              active_sinks = 0;
-	int              source = 0;
-	int              i = 0;
-
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
         struct afr_sh_algorithm *sh_algo = NULL;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
 
-	source = sh->source;
+        sh->algo_completion_cbk = afr_sh_data_fsync;
+        sh->algo_abort_cbk      = afr_sh_data_fail;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] == 0 && local->child_up[i] == 1) {
-			active_sinks++;
-			sh->success[i] = 1;
-		}
-	}
-	sh->success[source] = 1;
-
-	if (active_sinks == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"no active sinks for performing self-heal on file %s",
-			local->loc.path);
-		afr_sh_data_finish (frame, this);
-		return 0;
-	}
-	sh->active_sinks = active_sinks;
+        sh_algo = afr_sh_data_pick_algo (frame, this);
 
-	gf_log (this->name, GF_LOG_DEBUG,
-		"self-healing file %s from subvolume %s to %d other",
-		local->loc.path, priv->children[source]->name, active_sinks);
+        sh->algo = sh_algo;
+        sh_algo->fn (frame, this);
 
-        sh->algo_completion_cbk = afr_sh_data_trim_sinks;
-        sh->algo_abort_cbk      = afr_sh_data_finish;
+        return 0;
+}
 
-        sh_algo = afr_sh_data_pick_algo (frame, this);
+int
+afr_sh_data_trim_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                      struct iatt *postbuf, dict_t *xdata)
+{
+        int              call_count = 0;
+        int              child_index = 0;
+        afr_private_t    *priv = NULL;
+        afr_local_t      *local  = NULL;
+        afr_self_heal_t  *sh = NULL;
 
-        sh_algo->fn (frame, this);
+        priv  = this->private;
+        local = frame->local;
+        sh    = &local->self_heal;
+
+        child_index = (long) cookie;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "ftruncate of %s on subvolume %s failed (%s)",
+                                local->loc.path,
+                                priv->children[child_index]->name,
+                                strerror (op_errno));
+                        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                } else {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "ftruncate of %s on subvolume %s completed",
+                                local->loc.path,
+                                priv->children[child_index]->name);
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+                if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC))
+                        afr_sh_data_fail (frame, this);
+                else
+                        afr_sh_data_sync_prepare (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
-afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
+afr_sh_data_trim_sinks (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t     *local      = NULL;
-        afr_local_t *    orig_local = NULL;
+        afr_private_t * priv = NULL;
+        afr_local_t * local  = NULL;
+        afr_self_heal_t *sh  = NULL;
+        int             *sources = NULL;
+        int              call_count = 0;
+        int              i = 0;
 
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              nsources = 0;
-	int              source = 0;
-	int              i = 0;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
 
-	afr_sh_build_pending_matrix (priv, sh->pending_matrix, sh->xattr, 
-				     priv->child_count, AFR_DATA_TRANSACTION);
+        sources = sh->sources;
+        call_count = sh->active_sinks;
 
-	afr_sh_print_pending_matrix (sh->pending_matrix, this);
+        local->call_count = call_count;
 
-	nsources = afr_sh_mark_sources (sh, priv->child_count,
-                                        AFR_SELF_HEAL_DATA);
+        for (i = 0; i < priv->child_count; i++) {
+                if (sources[i] || !local->child_up[i])
+                        continue;
 
-	afr_sh_supress_errenous_children (sh->sources, sh->child_errno,
-					  priv->child_count);
+                STACK_WIND_COOKIE (frame, afr_sh_data_trim_cbk,
+                                   (void *) (long) i,
+                                   priv->children[i],
+                                   priv->children[i]->fops->ftruncate,
+                                   sh->healing_fd, sh->file_size,
+                                   NULL);
 
-        if (nsources == 0) {
-                gf_log (this->name, GF_LOG_TRACE,
-                        "No self-heal needed for %s",
-                        local->loc.path);
+                if (!--call_count)
+                        break;
+        }
 
-                afr_sh_data_finish (frame, this);
-                return 0;
+        return 0;
+}
+
+int
+afr_sh_inode_set_read_ctx (afr_self_heal_t *sh, xlator_t *this)
+{
+        afr_private_t   *priv = NULL;
+        int             ret = 0;
+        int             i = 0;
+
+        priv = this->private;
+        sh->source = afr_sh_select_source (sh->sources, priv->child_count);
+        if (sh->source < 0) {
+                ret = -1;
+                goto out;
         }
 
-	if ((nsources == -1)
-	    && (priv->favorite_child != -1)
-	    && (sh->child_errno[priv->favorite_child] == 0)) {
+        /* detect changes not visible through pending flags -- JIC */
+        for (i = 0; i < priv->child_count; i++) {
+                if (i == sh->source || sh->child_errno[i])
+                        continue;
+
+                if (SIZE_DIFFERS (&sh->buf[i], &sh->buf[sh->source]))
+                        sh->sources[i] = 0;
+        }
 
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Picking favorite child %s as authentic source to resolve conflicting data of %s",
-			priv->children[priv->favorite_child]->name,
-			local->loc.path);
+        afr_reset_children (sh->fresh_children, priv->child_count);
+        afr_get_fresh_children (sh->success_children, sh->sources,
+                                sh->fresh_children, priv->child_count);
+        afr_inode_set_read_ctx (this, sh->inode, sh->source,
+                                sh->fresh_children);
+out:
+        return ret;
+}
 
-		sh->sources[priv->favorite_child] = 1;
+char*
+afr_get_sizes_str (afr_local_t *local, struct iatt *bufs, xlator_t *this)
+{
+        afr_private_t *priv = NULL;
+        int           i     = 0;
+        char          num[1024] = {0};
+        size_t        len = 0;
+        char          *sizes_str = NULL;
+        size_t        off = 0;
+        char          *fmt_str = "%llu bytes on %s, ";
+        char          *child_down =  " %s,";
+        char          *child_unknown = " %s,";
+        int           down_child_present = 0;
+        int           down_count = 0;
+        int           unknown_count = 0;
+        int           unknown_child_present = 0;
+        char          *down_subvol_1 = " down subvolume is ";
+        char          *unknown_subvol_1 = " unknown subvolume is ";
+        char          *down_subvol_2 = " down subvolumes are ";
+        char          *unknown_subvol_2 = " unknown subvolumes are ";
+
+        priv = this->private;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i] == 1) {
+                        len += snprintf (num, sizeof (num), fmt_str,
+                                         (unsigned long long) bufs[i].ia_size,
+                                         priv->children[i]->name);
+                } else if (local->child_up[i] == 0) {
+                        len += snprintf (num, sizeof (num), child_down,
+                                         priv->children[i]->name);
+                        if (!down_child_present)
+                                down_child_present = 1;
+                        down_count ++;
+                } else if (local->child_up[i] == -1) {
+                        len += snprintf (num, sizeof (num), child_unknown,
+                                         priv->children[i]->name);
+                        if (!unknown_child_present)
+                                unknown_child_present = 1;
+                        unknown_count++;
+                }
 
-		nsources = afr_sh_source_count (sh->sources,
-						priv->child_count);
-	}
+        }
 
-	if (nsources == -1) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Unable to self-heal contents of '%s' (possible split-brain). "
-                        "Please delete the file from all but the preferred "
-                        "subvolume.", local->loc.path);
+        if (down_child_present) {
+                if (down_count > 1)
+                        len += snprintf (num, sizeof (num), "%s",
+                                         down_subvol_2);
+                else
+                        len += snprintf (num, sizeof (num), "%s",
+                                        down_subvol_1);
+        }
+        if (unknown_child_present) {
+                if (unknown_count > 1)
+                        len += snprintf (num, sizeof (num), "%s",
+                                         unknown_subvol_2);
+                else
+                        len += snprintf (num, sizeof (num), "%s",
+                                         unknown_subvol_1);
+        }
 
-		local->govinda_gOvinda = 1;
+        len++;//for '\0'
 
-		afr_sh_data_finish (frame, this);
-		return 0;
-	}
+        sizes_str = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
 
-	source = afr_sh_select_source (sh->sources, priv->child_count);
-        
-        if (source == -1) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "No active sources found.");
+        if (!sizes_str)
+                return NULL;
 
-                afr_sh_data_finish (frame, this);
-                return 0;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i] == 1) {
+                        off += snprintf (sizes_str + off, len - off, fmt_str,
+                                         (unsigned long long) bufs[i].ia_size,
+                                         priv->children[i]->name);
+                }
         }
 
-	sh->source     = source;
-	sh->block_size = 65536;
-	sh->file_size  = sh->buf[source].ia_size;
+        if (down_child_present) {
+                if (down_count > 1) {
+                        off += snprintf (sizes_str + off, len - off, "%s",
+                                         down_subvol_2);
+                } else {
+                        off += snprintf (sizes_str + off, len - off, "%s",
+                                         down_subvol_1);
+                }
+        }
 
-	if (FILE_HAS_HOLES (&sh->buf[source]))
-		sh->file_has_holes = 1;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i] == 0) {
+                        off += snprintf (sizes_str + off, len - off, child_down,
+                                         priv->children[i]->name);
+                }
+        }
 
-        orig_local = sh->orig_frame->local;
-        orig_local->cont.lookup.buf.ia_size = sh->buf[source].ia_size;
+        if (unknown_child_present) {
+                if (unknown_count > 1) {
+                        off += snprintf (sizes_str + off, len - off, "%s",
+                                        unknown_subvol_2);
+                } else {
+                        off += snprintf (sizes_str + off, len - off, "%s",
+                                         unknown_subvol_1);
+                }
+        }
 
-	/* detect changes not visible through pending flags -- JIC */
-	for (i = 0; i < priv->child_count; i++) {
-		if (i == source || sh->child_errno[i])
-			continue;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i] == -1) {
+                        off += snprintf (sizes_str + off, len - off,
+                                         child_unknown,
+                                         priv->children[i]->name);
 
-		if (SIZE_DIFFERS (&sh->buf[i], &sh->buf[source]))
-			sh->sources[i] = 0;
-	}
+                }
+        }
 
-        afr_set_read_child (this, local->loc.inode, sh->source);
+        return sizes_str;
+}
 
-	/*
-	   quick-read might have read the file, so send xattr from
-	   the source subvolume (http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=815)
-	*/
+char*
+afr_get_sinks_str (xlator_t *this, afr_local_t *local, afr_self_heal_t *sh)
+{
+        afr_private_t   *priv = NULL;
+        int             i = 0;
+        char            num[1024] = {0};
+        size_t          len = 0;
+        char            *sinks_str = NULL;
+        char            *temp_str = " to sinks ";
+        char            *str_format = " %s,";
+        char            off = 0;
+
+        priv = this->private;
+
+        len += snprintf (num, sizeof (num), "%s", temp_str);
+        for (i = 0; i < priv->child_count; i++) {
+                if ((sh->sources[i] == 0) && (local->child_up[i] == 1)) {
+                        len += snprintf (num, sizeof (num), str_format,
+                                         priv->children[i]->name);
+                }
+        }
 
-	dict_unref (orig_local->cont.lookup.xattr);
-        if (orig_local->cont.lookup.xattrs)
-                orig_local->cont.lookup.xattr = dict_ref (orig_local->cont.lookup.xattrs[sh->source]);
+        len ++;
 
-        if (sh->background) {
-                sh->unwind (sh->orig_frame, this);
-                sh->unwound = _gf_true;
+        sinks_str = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
+
+        if (!sinks_str)
+                return NULL;
+
+        off += snprintf (sinks_str + off, len - off, "%s", temp_str);
+
+        for (i = 0; i < priv->child_count; i++) {
+                if ((sh->sources[i] == 0) && (local->child_up[i] == 1)) {
+                        off += snprintf (sinks_str + off, len - off,
+                                         str_format,
+                                         priv->children[i]->name);
+                }
         }
 
-	afr_sh_data_sync_prepare (frame, this);
+        return sinks_str;
 
-	return 0;
 }
 
 
+void
+afr_set_data_sh_info_str (afr_local_t *local, afr_self_heal_t *sh, xlator_t *this)
+{
+        char            *pending_matrix_str = NULL;
+        char            *sizes_str = NULL;
+        char            *sinks_str = NULL;
+        afr_private_t   *priv = NULL;
+
+        priv = this->private;
+
+        pending_matrix_str = afr_get_pending_matrix_str (sh->pending_matrix,
+                                                         this);
+        if (!pending_matrix_str)
+                pending_matrix_str = "";
+
+        sizes_str = afr_get_sizes_str (local, sh->buf, this);
+        if (!sizes_str)
+                sizes_str = "";
+
+        sinks_str = afr_get_sinks_str (this, local, sh);
+        if (!sinks_str)
+                sinks_str = "";
+
+        gf_asprintf (&sh->data_sh_info, " data self heal from %s %s with "
+                     "%s data %s", priv->children[sh->source]->name, sinks_str,
+                     sizes_str, pending_matrix_str);
+
+        if (pending_matrix_str && strcmp (pending_matrix_str, ""))
+                GF_FREE (pending_matrix_str);
+
+        if (sizes_str && strcmp (sizes_str, ""))
+                GF_FREE (sizes_str);
+}
+
+void
+afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
+{
+        int              source = 0;
+        afr_local_t     *local      = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
+
+        source     = sh->source;
+        sh->block_size = this->ctx->page_size;
+        sh->file_size  = sh->buf[source].ia_size;
+
+        if (FILE_HAS_HOLES (&sh->buf[source]))
+                sh->file_has_holes = 1;
+
+        if (sh->background && sh->unwind && !sh->unwound) {
+                sh->unwind (sh->orig_frame, this, sh->op_ret, sh->op_errno,
+                            is_self_heal_failed (sh, AFR_CHECK_SPECIFIC));
+                sh->unwound = _gf_true;
+        }
+
+        afr_sh_mark_source_sinks (frame, this);
+        if (sh->active_sinks == 0) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no active sinks for performing self-heal on file %s",
+                        local->loc.path);
+                afr_sh_data_finish (frame, this);
+                return;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "self-healing file %s from subvolume %s to %d other",
+                local->loc.path, priv->children[sh->source]->name,
+                sh->active_sinks);
+
+        sh->actual_sh_started = _gf_true;
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_SYNC_BEGIN);
+        afr_sh_data_trim_sinks (frame, this);
+}
+
 int
-afr_self_heal_get_source (xlator_t *this, afr_local_t *local, dict_t **xattr)
+afr_sh_data_fxattrop_fstat_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_self_heal_t *sh   = NULL;
-	afr_private_t   *priv = NULL;
+        afr_local_t     *local      = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              nsources = 0;
+        int              ret = 0;
+        int             *old_sources = NULL;
+        int             tstamp_source = 0;
+        int             i = 0;
 
-	int              source = 0;
-	int              i = 0;
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Pending matrix for: %s",
+                lkowner_utoa (&frame->root->lk_owner));
+        if (sh->sync_done) {
+                //store sources before sync so that mtime can be set using the
+                //iatt buf from one of them.
+                old_sources = alloca (priv->child_count*sizeof (*old_sources));
+                memcpy (old_sources, sh->sources,
+                        priv->child_count * sizeof (*old_sources));
+        }
 
-	sh   = &local->self_heal;
-	priv = this->private;
+        nsources = afr_build_sources (this, sh->xattr, sh->buf, sh->pending_matrix,
+                                      sh->sources, sh->success_children,
+                                      AFR_DATA_TRANSACTION, NULL, _gf_true);
+        if ((nsources == -1)
+            && (priv->favorite_child != -1)
+            && (sh->child_errno[priv->favorite_child] == 0)) {
 
-	sh->pending_matrix = GF_CALLOC (sizeof (int32_t *), priv->child_count,
-                                        gf_afr_mt_int32_t);
-	for (i = 0; i < priv->child_count; i++) {
-		sh->pending_matrix[i] = GF_CALLOC (sizeof (int32_t),
-						priv->child_count,
-                                                gf_afr_mt_int32_t);
-	}
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Picking favorite child %s as authentic source to "
+                        "resolve conflicting data of %s",
+                        priv->children[priv->favorite_child]->name,
+                        local->loc.path);
+
+                sh->sources[priv->favorite_child] = 1;
+
+                nsources = afr_sh_source_count (sh->sources,
+                                                priv->child_count);
+        }
+
+        if (nsources == -1) {
+                afr_sh_print_split_brain_log (sh->pending_matrix, this,
+                                              local->loc.path);
+                afr_set_split_brain (this, sh->inode, DONT_KNOW, SPB);
 
-	sh->sources = GF_CALLOC (priv->child_count, sizeof (*sh->sources),
-                                 gf_afr_mt_int32_t);
+                afr_sh_data_fail (frame, this);
+                return 0;
+        }
+
+        afr_set_split_brain (this, sh->inode, DONT_KNOW, NO_SPB);
+
+        ret = afr_sh_inode_set_read_ctx (sh, this);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "No active sources found.");
 
-	afr_sh_build_pending_matrix (priv, sh->pending_matrix, xattr,
-				     priv->child_count, AFR_DATA_TRANSACTION);
+                afr_sh_data_fail (frame, this);
+                return 0;
+        }
 
-	(void)afr_sh_mark_sources (sh, priv->child_count, AFR_SELF_HEAL_DATA);
+        if (sh->sync_done) {
+                /* Perform setattr from one of the old_sources if possible
+                 * Because only they have the correct mtime, the new sources
+                 * (i.e. old sinks) have mtime from last writev in sync.
+                 */
+                tstamp_source = sh->source;
+                for (i = 0; i < priv->child_count; i++) {
+                        if (old_sources[i] && sh->sources[i])
+                                tstamp_source = i;
+                }
+                afr_sh_data_setattr (frame, this, &sh->buf[tstamp_source]);
+        } else {
+                afr_set_data_sh_info_str (local, sh, this);
+                if (nsources == 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "No self-heal needed for %s",
+                                local->loc.path);
 
-	source = afr_sh_select_source (sh->sources, priv->child_count);
+                        afr_sh_data_finish (frame, this);
+                        return 0;
+                }
 
-	return source;
+                if (sh->do_data_self_heal &&
+                    afr_data_self_heal_enabled (priv->data_self_heal))
+                        afr_sh_data_fix (frame, this);
+                else
+                        afr_sh_data_finish (frame, this);
+        }
+        return 0;
 }
 
+int
+afr_lookup_select_read_child_by_txn_type (xlator_t *this, afr_local_t *local,
+                                          dict_t **xattr,
+                                          afr_transaction_type txn_type,
+                                          uuid_t gfid)
+{
+        afr_private_t            *priv      = NULL;
+        int                      read_child = -1;
+        int32_t                  **pending_matrix = NULL;
+        int32_t                  *sources         = NULL;
+        int32_t                  *success_children   = NULL;
+        struct iatt              *bufs            = NULL;
+        int32_t                  nsources         = 0;
+        int32_t                  prev_read_child  = -1;
+        int32_t                  config_read_child = -1;
+        int32_t                  subvol_status = 0;
+
+        priv = this->private;
+        bufs = local->cont.lookup.bufs;
+        success_children = local->cont.lookup.success_children;
+
+        pending_matrix = local->cont.lookup.pending_matrix;
+        sources = local->cont.lookup.sources;
+        memset (sources, 0, sizeof (*sources) * priv->child_count);
+
+        nsources = afr_build_sources (this, xattr, bufs, pending_matrix,
+                                      sources, success_children, txn_type,
+                                      &subvol_status, _gf_false);
+        if (subvol_status & SPLIT_BRAIN) {
+                gf_log (this->name, GF_LOG_DEBUG, "%s: Possible split-brain",
+                        local->loc.path);
+                switch (txn_type) {
+                case AFR_DATA_TRANSACTION:
+                        local->cont.lookup.possible_spb = _gf_true;
+                        nsources = 1;
+                        sources[success_children[0]] = 1;
+                        break;
+                case AFR_ENTRY_TRANSACTION:
+                        read_child = afr_get_no_xattr_dir_read_child (this,
+                                                             success_children,
+                                                             bufs);
+                        sources[read_child] = 1;
+                        nsources = 1;
+                        break;
+                default:
+                        break;
+                }
+        }
+        if (nsources < 0)
+                goto out;
+
+        prev_read_child = local->read_child_index;
+        config_read_child = priv->read_child;
+        read_child = afr_select_read_child_from_policy (success_children,
+                                                        priv->child_count,
+                                                        prev_read_child,
+                                                        config_read_child,
+                                                        sources,
+                                                        priv->hash_mode, gfid);
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "returning read_child: %d",
+                read_child);
+        return read_child;
+}
 
 int
 afr_sh_data_fstat_cbk (call_frame_t *frame, void *cookie,
                        xlator_t *this, int32_t op_ret, int32_t op_errno,
-                       struct iatt *buf)
+                       struct iatt *buf, dict_t *xdata)
 {
-	afr_private_t   *priv  = NULL;
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
+        afr_private_t   *priv  = NULL;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        int call_count  = -1;
+        int child_index = (long) cookie;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret != -1) {
+        LOCK (&frame->lock);
+        {
+                if (op_ret != -1) {
                         gf_log (this->name, GF_LOG_TRACE,
                                 "fstat of %s on %s succeeded",
                                 local->loc.path,
                                 priv->children[child_index]->name);
 
-			sh->buf[child_index] = *buf;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		afr_sh_data_fix (frame, this);
-	}
-
-	return 0;
+                        sh->buf[child_index] = *buf;
+                        sh->success_children[sh->success_count] = child_index;
+                        sh->success_count++;
+                } else {
+                        gf_log (this->name, GF_LOG_ERROR, "%s: fstat failed "
+                                "on %s, reason %s", local->loc.path,
+                                priv->children[child_index]->name,
+                                strerror (op_errno));
+                        sh->child_errno[child_index] = op_errno;
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+                /* Previous versions of glusterfs might have set
+                 * the pending data xattrs which need to be erased
+                 */
+                if (!afr_sh_data_proceed (sh->success_count)) {
+                        gf_log (this->name, GF_LOG_ERROR, "inspecting metadata "
+                                "succeeded on < %d children, aborting "
+                                "self-heal for %s", AFR_SH_MIN_PARTICIPANTS,
+                                local->loc.path);
+                        afr_sh_data_fail (frame, this);
+                        goto out;
+                }
+                afr_sh_data_fxattrop_fstat_done (frame, this);
+        }
+out:
+        return 0;
 }
 
 
 int
 afr_sh_data_fstat (call_frame_t *frame, xlator_t *this)
 {
-	afr_self_heal_t *sh    = NULL; 
-	afr_local_t     *local = NULL;
-	afr_private_t   *priv  = NULL;
+        afr_self_heal_t *sh    = NULL;
+        afr_local_t     *local = NULL;
+        afr_private_t   *priv  = NULL;
+        int             call_count = 0;
+        int             i = 0;
+        int             child = 0;
+        int32_t         *fstat_children = NULL;
 
-	int call_count = 0;
-	int i = 0;
-
-	priv  = this->private;
-	local = frame->local;
-	sh    = &local->self_heal;
-
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_sh_data_fstat_cbk,
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->fstat,
-					   sh->healing_fd);
-
-			if (!--call_count)
-				break;
-		}
-	}
+        priv  = this->private;
+        local = frame->local;
+        sh    = &local->self_heal;
 
-	return 0;
+        fstat_children = memdup (sh->success_children,
+                                 sizeof (*fstat_children) * priv->child_count);
+        if (!fstat_children) {
+                afr_sh_data_fail (frame, this);
+                goto out;
+        }
+        call_count = sh->success_count;
+        local->call_count = call_count;
+
+        memset (sh->buf, 0, sizeof (*sh->buf) * priv->child_count);
+        afr_reset_children (sh->success_children, priv->child_count);
+        sh->success_count = 0;
+        for (i = 0; i < priv->child_count; i++) {
+                child = fstat_children[i];
+                if (child == -1)
+                        break;
+                STACK_WIND_COOKIE (frame, afr_sh_data_fstat_cbk,
+                                   (void *) (long) child,
+                                   priv->children[child],
+                                   priv->children[child]->fops->fstat,
+                                   sh->healing_fd, NULL);
+                --call_count;
+        }
+        GF_ASSERT (!call_count);
+out:
+        GF_FREE (fstat_children);
+        return 0;
 }
 
-
-int
-afr_sh_data_fxattrop_cbk (call_frame_t *frame, void *cookie,
-                          xlator_t *this, int32_t op_ret, int32_t op_errno,
-                          dict_t *xattr)
+void
+afr_sh_common_fxattrop_resp_handler (call_frame_t *frame, void *cookie,
+                                     xlator_t *this, int32_t op_ret,
+                                     int32_t op_errno, dict_t *xattr)
 {
-	afr_private_t   *priv  = NULL;
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
+        afr_private_t   *priv  = NULL;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        int child_index = (long) cookie;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret != -1) {
+        LOCK (&frame->lock);
+        {
+                if (op_ret != -1) {
                         gf_log (this->name, GF_LOG_TRACE,
                                 "fxattrop of %s on %s succeeded",
                                 local->loc.path,
                                 priv->children[child_index]->name);
 
-			sh->xattr[child_index] = dict_ref (xattr);
-		}
-	}
-	UNLOCK (&frame->lock);
+                        sh->xattr[child_index] = dict_ref (xattr);
+                        sh->success_children[sh->success_count] = child_index;
+                        sh->success_count++;
+                } else {
+                        gf_log (this->name, GF_LOG_ERROR, "fxattrop of %s "
+                                "failed on %s, reason %s", local->loc.path,
+                                priv->children[child_index]->name,
+                                strerror (op_errno));
+                        sh->child_errno[child_index] = op_errno;
+                }
+        }
+        UNLOCK (&frame->lock);
+}
 
-	call_count = afr_frame_return (frame);
+int
+afr_sh_data_fxattrop_cbk (call_frame_t *frame, void *cookie,
+                          xlator_t *this, int32_t op_ret, int32_t op_errno,
+                          dict_t *xattr, dict_t *xdata)
+{
+        int             call_count  = -1;
+        afr_local_t     *local  = NULL;
+        afr_self_heal_t *sh     = NULL;
 
-	if (call_count == 0) {
-		afr_sh_data_fstat (frame, this);
-	}
+        local = frame->local;
+        sh    = &local->self_heal;
 
-	return 0;
+        afr_sh_common_fxattrop_resp_handler (frame, cookie, this, op_ret,
+                                             op_errno, xattr);
+
+        call_count = afr_frame_return (frame);
+        if (call_count == 0) {
+                if (!afr_sh_data_proceed (sh->success_count)) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s, inspecting "
+                                "change log succeeded on < %d children",
+                                local->loc.path, AFR_SH_MIN_PARTICIPANTS);
+                        afr_sh_data_fail (frame, this);
+                        goto out;
+                }
+                afr_sh_data_fstat (frame, this);
+        }
+out:
+        return 0;
 }
 
 
 int
 afr_sh_data_fxattrop (call_frame_t *frame, xlator_t *this)
 {
-	afr_self_heal_t *sh    = NULL;
-	afr_local_t     *local = NULL;
-	afr_private_t   *priv  = NULL;
-	dict_t          *xattr_req = NULL;
+        afr_self_heal_t *sh    = NULL;
+        afr_local_t     *local = NULL;
+        afr_private_t   *priv  = NULL;
+        dict_t          **xattr_req;
+        int32_t         *zero_pending = NULL;
+        int call_count = 0;
+        int i = 0;
+        int ret = 0;
+	int j;
 
-        int32_t zero_pending[3] = {0, 0, 0};
+        priv  = this->private;
+        local = frame->local;
+        sh    = &local->self_heal;
 
-	int call_count = 0;
-	int i = 0;
-	int ret = 0;
+        call_count = afr_up_children_count (local->child_up,
+                                            priv->child_count);
 
-	priv  = this->private;
-	local = frame->local;
-	sh    = &local->self_heal;
+        local->call_count = call_count;
 
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
+	xattr_req = GF_CALLOC(priv->child_count, sizeof(struct dict_t *),
+			      gf_afr_mt_dict_t);
+	if (!xattr_req)
+		goto out;
 
-	local->call_count = call_count;
+	for (i = 0; i < priv->child_count; i++) {
+		xattr_req[i] = dict_new();
+		if (!xattr_req[i]) {
+			ret = -1;
+			goto out;
+		}
+	}
 
-	xattr_req = dict_new();
-	if (xattr_req) {
-                for (i = 0; i < priv->child_count; i++) {
-                        ret = dict_set_static_bin (xattr_req, priv->pending_key[i],
-                                                   zero_pending, 3 * sizeof(int32_t));
-                        if (ret < 0)
+	for (i = 0; i < priv->child_count; i++) {
+		for (j = 0; j < priv->child_count; j++) {
+			zero_pending = GF_CALLOC (3, sizeof (*zero_pending),
+						  gf_afr_mt_int32_t);
+			if (!zero_pending) {
+				ret = -1;
+				goto out;
+			}
+			ret = dict_set_dynptr (xattr_req[i], priv->pending_key[j],
+					       zero_pending,
+					       3 * sizeof (*zero_pending));
+			if (ret < 0) {
 				gf_log (this->name, GF_LOG_WARNING,
 					"Unable to set dict value");
+				goto out;
+			} else {
+				zero_pending = NULL;
+			}
+		}
+	}
+
+        afr_reset_xattr (sh->xattr, priv->child_count);
+        afr_reset_children (sh->success_children, priv->child_count);
+        memset (sh->child_errno, 0,
+                sizeof (*sh->child_errno) * priv->child_count);
+        sh->success_count = 0;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, afr_sh_data_fxattrop_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->fxattrop,
+                                           sh->healing_fd, GF_XATTROP_ADD_ARRAY,
+                                           xattr_req[i], NULL);
+
+                        if (!--call_count)
+                                break;
                 }
         }
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_sh_data_fxattrop_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->fxattrop,
-					   sh->healing_fd, GF_XATTROP_ADD_ARRAY,
-                                           xattr_req);
-
-			if (!--call_count)
-				break;
-		}
+out:
+	if (xattr_req) {
+		for (i = 0; i < priv->child_count; i++)
+			if (xattr_req[i])
+				dict_unref(xattr_req[i]);
+		GF_FREE(xattr_req);
 	}
 
-	if (xattr_req)
-		dict_unref (xattr_req);
+        if (ret) {
+                GF_FREE (zero_pending);
+                afr_sh_data_fail (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
+int
+afr_sh_data_big_lock_success (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t   *local = NULL;
+        afr_self_heal_t *sh = NULL;
+
+        local = frame->local;
+        sh = &local->self_heal;
+
+        sh->data_lock_held = _gf_true;
+        afr_sh_data_fxattrop (frame, this);
+        return 0;
+}
 
 int
-afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this);
+afr_sh_dom_lock_success (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t   *local = NULL;
+        afr_self_heal_t *sh = NULL;
+
+        local = frame->local;
+        sh = &local->self_heal;
+
+        sh->sh_dom_lock_held = _gf_true;
+        afr_sh_data_lock (frame, this, 0, 0, _gf_true, this->name,
+                          afr_sh_data_big_lock_success,
+                          afr_sh_data_fail);
+        return 0;
+}
 
 int
-afr_sh_data_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
+afr_sh_data_post_blocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
+        afr_self_heal_t     *sh       = NULL;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
+        sh       = &local->self_heal;
 
         if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking inodelks failed.");
-                afr_sh_data_done (frame, this);
+                gf_log (this->name, GF_LOG_ERROR, "Blocking data inodelks "
+                        "failed for %s. by %s",
+                        local->loc.path, lkowner_utoa (&frame->root->lk_owner));
+
+                sh->data_lock_failure_handler (frame, this);
         } else {
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking inodelks done. Proceeding to FOP");
-                afr_sh_data_fxattrop (frame, this);
+                gf_log (this->name, GF_LOG_DEBUG, "Blocking data inodelks "
+                        "done for %s by %s. Proceding to self-heal",
+                        local->loc.path, lkowner_utoa (&frame->root->lk_owner));
+
+                sh->data_lock_success_handler (frame, this);
         }
 
         return 0;
 }
 
 int
-afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this)
+afr_sh_data_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
@@ -941,173 +1454,301 @@ afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
         sh       = &local->self_heal;
 
+        if (int_lock->lock_op_ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG, "Non Blocking data inodelks "
+                        "failed for %s. by %s",
+                        local->loc.path, lkowner_utoa (&frame->root->lk_owner));
+
+		if (!sh->data_lock_block) {
+			sh->data_lock_failure_handler(frame, this);
+		} else {
+			int_lock->lock_cbk =
+				afr_sh_data_post_blocking_inodelk_cbk;
+			afr_blocking_lock (frame, this);
+		}
+        } else {
+
+                gf_log (this->name, GF_LOG_DEBUG, "Non Blocking data inodelks "
+                        "done for %s by %s. Proceeding to self-heal",
+                        local->loc.path, lkowner_utoa (&frame->root->lk_owner));
+                sh->data_lock_success_handler (frame, this);
+        }
+
+        return 0;
+}
+
+int
+afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this, char *dom,
+                      off_t start, off_t len)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
+        afr_local_t         *local    = NULL;
+
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+
         int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
         int_lock->selfheal_lk_type    = AFR_DATA_SELF_HEAL_LK;
 
         afr_set_lock_number (frame, this);
 
-        int_lock->lk_flock.l_start = 0;
-        int_lock->lk_flock.l_len   = 0;
-        int_lock->lk_flock.l_type  = F_WRLCK;
         int_lock->lock_cbk         = afr_sh_data_post_nonblocking_inodelk_cbk;
 
-        afr_nonblocking_inodelk (frame, this);
+        int_lock->domain = dom;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+        inodelk->flock.l_start = start;
+        inodelk->flock.l_len   = len;
+        inodelk->flock.l_type  = F_WRLCK;
 
+        afr_nonblocking_inodelk (frame, this);
 
-	return 0;
+        return 0;
 }
 
+int
+afr_post_sh_big_lock_success (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local     = NULL;
+        afr_self_heal_t *sh        = NULL;
+
+        local = frame->local;
+        sh = &local->self_heal;
+
+        GF_ASSERT (sh->old_loop_frame);
+        sh_loop_finish (sh->old_loop_frame, this);
+        sh->old_loop_frame = NULL;
+        sh->data_lock_held = _gf_true;
+        sh->sync_done = _gf_true;
+        afr_sh_data_fxattrop (frame, this);
+        return 0;
+}
 
 int
-afr_sh_data_lock (call_frame_t *frame, xlator_t *this)
+afr_post_sh_big_lock_failure (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	afr_self_heal_t * sh  = NULL;
+        afr_local_t     *local     = NULL;
+        afr_self_heal_t *sh        = NULL;
 
+        local = frame->local;
+        sh = &local->self_heal;
 
-	local = frame->local;
-	sh    = &local->self_heal;
-	priv  = this->private;
+        GF_ASSERT (sh->old_loop_frame);
+        sh_loop_finish (sh->old_loop_frame, this);
+        sh->old_loop_frame = NULL;
+        afr_sh_set_timestamps (frame, this);
+        return 0;
+}
 
-        if (sh->data_lock_held) {
-                /* caller has held the lock already,
-                   so skip locking */
 
-                afr_sh_data_fxattrop (frame, this);
-                return 0;
-        }
+int
+afr_sh_data_lock (call_frame_t *frame, xlator_t *this,
+                  off_t start, off_t len, gf_boolean_t block,
+                  char *dom, afr_lock_cbk_t success_handler,
+                  afr_lock_cbk_t failure_handler)
+{
+        afr_local_t *   local = NULL;
+        afr_self_heal_t * sh  = NULL;
 
-        return afr_sh_data_lock_rec (frame, this);
-}
+        local = frame->local;
+        sh    = &local->self_heal;
 
+        sh->data_lock_success_handler = success_handler;
+        sh->data_lock_failure_handler = failure_handler;
+	sh->data_lock_block = block;
+        return afr_sh_data_lock_rec (frame, this, dom, start, len);
+}
 
 int
 afr_sh_data_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		      int32_t op_ret, int32_t op_errno, fd_t *fd)
+                      int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	/* TODO: some of the open's might fail.
-	   In that case, modify cleanup fn to send flush on those 
-	   fd's which are already open */
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"open of %s failed on child %s (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-			sh->op_failed = 1;
-		}
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              call_count = 0;
+        int              child_index = 0;
 
-                gf_log (this->name, GF_LOG_TRACE,
-                        "open of %s succeeded on child %s",
-                        local->loc.path,
-                        priv->children[child_index]->name);
-	}
-	UNLOCK (&frame->lock);
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-	call_count = afr_frame_return (frame);
+        child_index = (long) cookie;
 
-	if (call_count == 0) {
-		if (sh->op_failed) {
-			afr_sh_data_finish (frame, this);
-			return 0;
-		}
+        /* TODO: some of the open's might fail.
+           In that case, modify cleanup fn to send flush on those
+           fd's which are already open */
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "open of %s failed on child %s (%s)",
+                                local->loc.path,
+                                priv->children[child_index]->name,
+                                strerror (op_errno));
+                        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                } else {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "open of %s succeeded on child %s",
+                                local->loc.path,
+                                priv->children[child_index]->name);
+                }
+        }
+        UNLOCK (&frame->lock);
 
-		gf_log (this->name, GF_LOG_TRACE,
-			"fd for %s opened, commencing sync",
-			local->loc.path);
+        call_count = afr_frame_return (frame);
 
-                afr_sh_data_lock (frame, this);
-	}
+        if (call_count == 0) {
+                if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+                        afr_sh_data_fail (frame, this);
+                        return 0;
+                }
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "fd for %s opened, commencing sync",
+                        local->loc.path);
+
+                afr_sh_data_lock (frame, this, 0, 0, _gf_true, priv->sh_domain,
+                                  afr_sh_dom_lock_success, afr_sh_data_fail);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_data_open (call_frame_t *frame, xlator_t *this)
 {
-	int i = 0;
-	int call_count = 0;
+        int i = 0;
+        int call_count = 0;
+        fd_t *fd = NULL;
+        afr_local_t *   local = NULL;
+        afr_private_t * priv  = NULL;
+        afr_self_heal_t *sh = NULL;
+
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-	fd_t *fd = NULL;
+        call_count = afr_up_children_count (local->child_up, priv->child_count);
+        local->call_count = call_count;
 
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	afr_self_heal_t *sh = NULL;
+        fd = fd_create (local->loc.inode, frame->root->pid);
+        sh->healing_fd = fd;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        /* open sinks */
+        for (i = 0; i < priv->child_count; i++) {
+                if(!local->child_up[i])
+                        continue;
 
-        if (sh->healing_fd_opened) {
-                /* caller has opened the fd for us already, so skip open */
+                STACK_WIND_COOKIE (frame, afr_sh_data_open_cbk,
+                                   (void *) (long) i,
+                                   priv->children[i],
+                                   priv->children[i]->fops->open,
+                                   &local->loc,
+                                   O_RDWR|O_LARGEFILE, fd, NULL);
 
-                afr_sh_data_lock (frame, this);
-                return 0;
+                if (!--call_count)
+                        break;
         }
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-	local->call_count = call_count;
+        return 0;
+}
 
-	fd = fd_create (local->loc.inode, frame->root->pid);
-	sh->healing_fd = fd;
+void
+afr_sh_non_reg_fix (call_frame_t *frame, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno)
+{
+        afr_private_t   *priv = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_local_t     *local = NULL;
+        int             i = 0;
+
+        if (op_ret < 0) {
+                afr_sh_data_fail (frame, this);
+                return;
+        }
 
-	/* open sinks */
-	for (i = 0; i < priv->child_count; i++) {
-		if(!local->child_up[i])
-			continue;
-
-		STACK_WIND_COOKIE (frame, afr_sh_data_open_cbk,
-				   (void *) (long) i,
-				   priv->children[i], 
-				   priv->children[i]->fops->open,
-				   &local->loc, 
-				   O_RDWR|O_LARGEFILE, fd, 0); 
-
-		if (!--call_count)
-			break;
-	}
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-	return 0;
+        for (i = 0; i < priv->child_count ; i++) {
+                if (1 == local->child_up[i])
+                        sh->success[i] = 1;
+        }
+
+        afr_sh_erase_pending (frame, this, AFR_DATA_TRANSACTION,
+                              afr_sh_data_erase_pending_cbk,
+                              afr_sh_data_finish);
+}
+
+int
+afr_sh_non_reg_lock_success (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t   *local = NULL;
+        afr_self_heal_t *sh = NULL;
+
+        local = frame->local;
+        sh = &local->self_heal;
+        sh->data_lock_held = _gf_true;
+        afr_sh_common_lookup (frame, this, &local->loc,
+                              afr_sh_non_reg_fix, NULL,
+                              AFR_LOOKUP_FAIL_CONFLICTS |
+                              AFR_LOOKUP_FAIL_MISSING_GFIDS,
+                              NULL);
+        return 0;
 }
 
+gf_boolean_t
+afr_can_start_data_self_heal (afr_self_heal_t *sh, afr_private_t *priv)
+{
+        if (sh->force_confirm_spb)
+                return _gf_true;
+        if (sh->do_data_self_heal &&
+            afr_data_self_heal_enabled (priv->data_self_heal))
+                return _gf_true;
+        return _gf_false;
+}
 
 int
 afr_self_heal_data (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t   *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t *priv = this->private;
-
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh    = NULL;
+        afr_private_t   *priv  = this->private;
+        int             ret    = -1;
 
-	local = frame->local;
-	sh = &local->self_heal;
+        local = frame->local;
+        sh = &local->self_heal;
+
+        sh->sh_type_in_action = AFR_SELF_HEAL_DATA;
+
+        if (afr_can_start_data_self_heal (sh, priv)) {
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
+                ret = afr_inodelk_init (&local->internal_lock.inodelk[1],
+                                        priv->sh_domain, priv->child_count);
+                if (ret < 0) {
+                        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                        afr_sh_data_done (frame, this);
+                        return 0;
+                }
 
-	if (sh->need_data_self_heal && priv->data_self_heal) {
-		afr_sh_data_open (frame, this);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"not doing data self heal on %s",
-			local->loc.path);
-		afr_sh_data_done (frame, this);
-	}
+                if (IA_ISREG (sh->type)) {
+                        afr_sh_data_open (frame, this);
+                } else {
+                        afr_sh_data_lock (frame, this, 0, 0, _gf_true,
+                                          this->name,
+                                          afr_sh_non_reg_lock_success,
+                                          afr_sh_data_fail);
+                }
+        } else {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "not doing data self heal on %s",
+                        local->loc.path);
+                afr_sh_data_done (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
-
diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index 446b3d868..53491a1d7 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <libgen.h>
@@ -49,39 +40,30 @@
 #include "afr-self-heal.h"
 #include "afr-self-heal-common.h"
 
-int
-afr_sh_post_nonblocking_entrylk_cbk (call_frame_t *frame, xlator_t *this);
+#define AFR_INIT_SH_FRAME_VALS(_frame, _local, _sh, _sh_frame, _sh_local, _sh_sh)\
+        do {\
+                _local = _frame->local;\
+                _sh = &_local->self_heal;\
+                _sh_frame = _sh->sh_frame;\
+                _sh_local = _sh_frame->local;\
+                _sh_sh    = &_sh_local->self_heal;\
+        } while (0);
 
 int
+afr_sh_entry_impunge_create_file (call_frame_t *impunge_frame, xlator_t *this,
+                                  int child_index);
+int
 afr_sh_entry_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	/* 
-	   TODO: cleanup sh->* 
-	*/
-
-        if (sh->healing_fd)
-                fd_unref (sh->healing_fd);
-        sh->healing_fd = NULL;
-
-/*         for (i = 0; i < priv->child_count; i++) { */
-/*                 sh->locked_nodes[i] = 0; */
-/*         } */
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"self heal of %s completed",
-		local->loc.path);
+        local = frame->local;
+        sh = &local->self_heal;
 
-	sh->completion_cbk (frame, this);
+        sh->completion_cbk (frame, this);
 
-	return 0;
+        return 0;
 }
 
 
@@ -104,323 +86,268 @@ afr_sh_entry_unlock (call_frame_t *frame, xlator_t *this)
 int
 afr_sh_entry_finish (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t   *local = NULL;
+        afr_local_t   *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"finishing entry selfheal of %s", local->loc.path);
+        gf_log (this->name, GF_LOG_TRACE,
+                "finishing entry selfheal of %s", local->loc.path);
 
-	afr_sh_entry_unlock (frame, this);
+        afr_sh_entry_unlock (frame, this);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_entry_erase_pending_cbk (call_frame_t *frame, void *cookie,
-				xlator_t *this, int32_t op_ret,
-				int32_t op_errno, dict_t *xattr)
+                                xlator_t *this, int32_t op_ret,
+                                int32_t op_errno, dict_t *xattr, dict_t *xdata)
 {
-	int             call_count = 0;
+        long                 i          = 0;
+        int                  call_count = 0;
+        afr_local_t         *local      = NULL;
+        afr_self_heal_t     *sh         = NULL;
+        afr_local_t         *orig_local = NULL;
+        call_frame_t        *orig_frame = NULL;
+        afr_private_t       *priv       = NULL;
+        int32_t             read_child  = -1;
 
-	call_count = afr_frame_return (frame);
+        local = frame->local;
+        priv  = this->private;
+        sh = &local->self_heal;
+        i = (long)cookie;
 
-	if (call_count == 0)
-		afr_sh_entry_finish (frame, this);
 
-	return 0;
+        afr_children_add_child (sh->fresh_children, i, priv->child_count);
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "%s: failed to erase pending xattrs on %s (%s)",
+                        local->loc.path, priv->children[i]->name,
+                        strerror (op_errno));
+        }
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+                if (sh->source == -1) {
+                        //this happens if the forced merge option is set
+                        read_child = sh->fresh_children[0];
+                } else {
+                        read_child = sh->source;
+                }
+                afr_inode_set_read_ctx (this, sh->inode, read_child,
+                                        sh->fresh_children);
+                orig_frame = sh->orig_frame;
+                orig_local = orig_frame->local;
+
+                if (sh->source != -1) {
+                        orig_local->cont.lookup.buf.ia_nlink = sh->buf[sh->source].ia_nlink;
+                }
+
+                afr_sh_entry_finish (frame, this);
+        }
+
+        return 0;
 }
 
 
 int
 afr_sh_entry_erase_pending (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              i = 0;
-	dict_t          **erase_xattr = NULL;
-        int              need_unwind = 0;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
 
+        local = frame->local;
+        sh = &local->self_heal;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        if (sh->entries_skipped) {
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                goto out;
+        }
+        afr_sh_erase_pending (frame, this, AFR_ENTRY_TRANSACTION,
+                              afr_sh_entry_erase_pending_cbk,
+                              afr_sh_entry_finish);
+        return 0;
+out:
+        afr_sh_entry_finish (frame, this);
+        return 0;
+}
 
-	afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, sh->success,
-                                 priv->child_count, AFR_ENTRY_TRANSACTION);
 
-	erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
-                                 gf_afr_mt_dict_t);
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i]) {
-			call_count++;
+static int
+next_active_source (call_frame_t *frame, xlator_t *this,
+                    int current_active_source)
+{
+        afr_private_t   *priv = NULL;
+        afr_local_t     *local  = NULL;
+        afr_self_heal_t *sh  = NULL;
+        int              source = -1;
+        int              next_active_source = -1;
+        int              i = 0;
 
-			erase_xattr[i] = get_new_dict();
-			dict_ref (erase_xattr[i]);
-		}
-	}
+        priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
 
-        if (call_count == 0)
-                need_unwind = 1;
-
-	afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
-			       priv->child_count, AFR_ENTRY_TRANSACTION);
-
-	local->call_count = call_count;
-	for (i = 0; i < priv->child_count; i++) {
-		if (!erase_xattr[i])
-			continue;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"erasing pending flags from %s on %s",
-			local->loc.path, priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_entry_erase_pending_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->xattrop,
-				   &local->loc,
-				   GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
-		if (!--call_count)
-			break;
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (erase_xattr[i]) {
-			dict_unref (erase_xattr[i]);
-		}
-	}
-	GF_FREE (erase_xattr);
-
-        if (need_unwind)
-                afr_sh_entry_finish (frame, this);
+        source = sh->source;
 
-	return 0;
-}
+        if (source != -1) {
+                if (current_active_source != source)
+                        next_active_source = source;
+                goto out;
+        }
 
+        /*
+          the next active sink becomes the source for the
+          'conservative decision' of merging all entries
+        */
 
+        for (i = 0; i < priv->child_count; i++) {
+                if ((sh->sources[i] == 0)
+                    && (local->child_up[i] == 1)
+                    && (i > current_active_source)) {
 
-static int
-next_active_source (call_frame_t *frame, xlator_t *this,
-		    int current_active_source)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              source = -1;
-	int              next_active_source = -1;
-	int              i = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	source = sh->source;
-
-	if (source != -1) {
-		if (current_active_source != source)
-			next_active_source = source;
-		goto out;
-	}
-
-	/*
-	  the next active sink becomes the source for the
-	  'conservative decision' of merging all entries
-	*/
-
-	for (i = 0; i < priv->child_count; i++) {
-		if ((sh->sources[i] == 0)
-		    && (local->child_up[i] == 1)
-		    && (i > current_active_source)) {
-
-			next_active_source = i;
-			break;
-		}
-	}
+                        next_active_source = i;
+                        break;
+                }
+        }
 out:
-	return next_active_source;
+        return next_active_source;
 }
 
 
 
 static int
 next_active_sink (call_frame_t *frame, xlator_t *this,
-		  int current_active_sink)
+                  int current_active_sink)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              next_active_sink = -1;
-	int              i = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	/*
-	  the next active sink becomes the source for the
-	  'conservative decision' of merging all entries
-	*/
-
-	for (i = 0; i < priv->child_count; i++) {
-		if ((sh->sources[i] == 0)
-		    && (local->child_up[i] == 1)
-		    && (i > current_active_sink)) {
-
-			next_active_sink = i;
-			break;
-		}
-	}
-
-	return next_active_sink;
-}
-
+        afr_private_t   *priv = NULL;
+        afr_local_t     *local  = NULL;
+        afr_self_heal_t *sh  = NULL;
+        int              next_active_sink = -1;
+        int              i = 0;
 
-int
-build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
-{
-	int   ret = -1;
+        priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
 
-	if (!child) {
-		goto out;
-	}
+        /*
+          the next active sink becomes the source for the
+          'conservative decision' of merging all entries
+        */
 
-	if (strcmp (parent->path, "/") == 0)
-		ret = gf_asprintf ((char **)&child->path, "/%s", name);
-	else
-		ret = gf_asprintf ((char **)&child->path, "%s/%s", parent->path, 
-                                name);
+        for (i = 0; i < priv->child_count; i++) {
+                if ((sh->sources[i] == 0)
+                    && (local->child_up[i] == 1)
+                    && (i > current_active_sink)) {
 
-        if (-1 == ret) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "asprintf failed while setting child path");
+                        next_active_sink = i;
+                        break;
+                }
         }
 
-	if (!child->path) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	child->name = strrchr (child->path, '/');
-	if (child->name)
-		child->name++;
-
-	child->parent = inode_ref (parent->inode);
-	child->inode = inode_new (parent->inode->table);
-
-	if (!child->inode) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ret = 0;
-out:
-	if (ret == -1)
-		loc_wipe (child);
-
-	return ret;
+        return next_active_sink;
 }
 
+int
+afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this);
+
+int
+afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this);
 
 int
 afr_sh_entry_expunge_all (call_frame_t *frame, xlator_t *this);
 
 int
 afr_sh_entry_expunge_subvol (call_frame_t *frame, xlator_t *this,
-			     int active_src);
+                             int active_src);
 
 int
 afr_sh_entry_expunge_entry_done (call_frame_t *frame, xlator_t *this,
-				 int active_src)
+                                 int active_src, int32_t op_ret,
+                                 int32_t op_errno)
 {
-	int              call_count = 0;
+        int              call_count = 0;
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0)
-		afr_sh_entry_expunge_subvol (frame, this, active_src);
+        if (call_count == 0)
+                afr_sh_entry_expunge_subvol (frame, this, active_src);
 
-	return 0;
+        return 0;
 }
 
 int
 afr_sh_entry_expunge_parent_setattr_cbk (call_frame_t *expunge_frame,
                                          void *cookie, xlator_t *this,
                                          int32_t op_ret, int32_t op_errno,
-                                         struct iatt *preop, struct iatt *postop)
+                                         struct iatt *preop, struct iatt *postop,
+                                         dict_t *xdata)
 {
-	afr_private_t   *priv          = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh    = NULL;
+        afr_private_t   *priv          = NULL;
+        afr_local_t     *expunge_local = NULL;
+        afr_self_heal_t *expunge_sh    = NULL;
         call_frame_t    *frame         = NULL;
-
-        int active_src = (long) cookie;
+        int              active_src    = (long) cookie;
+        afr_self_heal_t *sh            = NULL;
+        afr_local_t     *local         = NULL;
 
         priv          = this->private;
         expunge_local = expunge_frame->local;
-	expunge_sh    = &expunge_local->self_heal;
-	frame         = expunge_sh->sh_frame;
+        expunge_sh    = &expunge_local->self_heal;
+        frame         = expunge_sh->sh_frame;
+        local         = frame->local;
+        sh            = &local->self_heal;
 
         if (op_ret != 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_ERROR,
                         "setattr on parent directory of %s on subvolume %s failed: %s",
                         expunge_local->loc.path,
                         priv->children[active_src]->name, strerror (op_errno));
         }
 
-	AFR_STACK_DESTROY (expunge_frame);
-	afr_sh_entry_expunge_entry_done (frame, this, active_src);
+        AFR_STACK_DESTROY (expunge_frame);
+        sh->expunge_done (frame, this, active_src, op_ret, op_errno);
 
         return 0;
 }
 
 
 int
-afr_sh_entry_expunge_rename_cbk (call_frame_t *expunge_frame, void *cookie,
-				 xlator_t *this,
-				 int32_t op_ret, int32_t op_errno,
-                                 struct iatt *buf,
-                                 struct iatt *preoldparent,
-                                 struct iatt *postoldparent,
-                                 struct iatt *prenewparent,
-                                 struct iatt *postnewparent)
+afr_sh_entry_expunge_remove_cbk (call_frame_t *expunge_frame, void *cookie,
+                                 xlator_t *this,
+                                 int32_t op_ret, int32_t op_errno,
+                                 struct iatt *preparent,
+                                 struct iatt *postparent, dict_t *xdata)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	int              active_src = 0;
-	call_frame_t    *frame = NULL;
-
-        int32_t valid = 0;
-
-	priv = this->private;
-	expunge_local = expunge_frame->local;
-	expunge_sh = &expunge_local->self_heal;
-	frame = expunge_sh->sh_frame;
-
-	active_src = (long) cookie;
-
-	if (op_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"removed %s on %s",
-			expunge_local->loc.path,
-			priv->children[active_src]->name);
-	} else {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"removing %s on %s failed (%s)",
-			expunge_local->loc.path,
-			priv->children[active_src]->name,
-			strerror (op_errno));
-	}
+        afr_private_t   *priv = NULL;
+        afr_local_t     *expunge_local = NULL;
+        afr_self_heal_t *expunge_sh = NULL;
+        int              active_src = 0;
+        int32_t          valid = 0;
+
+        priv = this->private;
+        expunge_local = expunge_frame->local;
+        expunge_sh = &expunge_local->self_heal;
+
+        active_src = (long) cookie;
+
+        if (op_ret == 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "removed %s on %s",
+                        expunge_local->loc.path,
+                        priv->children[active_src]->name);
+        } else {
+                gf_log (this->name, GF_LOG_INFO,
+                        "removing %s on %s failed (%s)",
+                        expunge_local->loc.path,
+                        priv->children[active_src]->name,
+                        strerror (op_errno));
+        }
 
         valid = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-        afr_build_parent_loc (&expunge_sh->parent_loc, &expunge_local->loc);
 
         STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_parent_setattr_cbk,
                            (void *) (long) active_src,
@@ -428,915 +355,972 @@ afr_sh_entry_expunge_rename_cbk (call_frame_t *expunge_frame, void *cookie,
                            priv->children[active_src]->fops->setattr,
                            &expunge_sh->parent_loc,
                            &expunge_sh->parentbuf,
-                           valid);
+                           valid, NULL);
 
         return 0;
 }
 
 
-static void
-init_trash_loc (loc_t *trash_loc, inode_table_t *table)
-{
-        trash_loc->path   = gf_strdup ("/" GF_REPLICATE_TRASH_DIR);
-        trash_loc->name   = GF_REPLICATE_TRASH_DIR;
-        trash_loc->parent = table->root;
-        trash_loc->inode  = inode_new (table);
-}
-
-
-char *
-make_trash_path (const char *path)
+int
+afr_sh_entry_expunge_unlink (call_frame_t *expunge_frame, xlator_t *this,
+                             int active_src)
 {
-        char *c  = NULL;
-        char *tp = NULL;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *expunge_local = NULL;
 
-        tp = GF_CALLOC (strlen ("/" GF_REPLICATE_TRASH_DIR) + strlen (path) + 1, sizeof (char),
-                        gf_afr_mt_char);
+        priv          = this->private;
+        expunge_local = expunge_frame->local;
 
-        strcpy (tp, GF_REPLICATE_TRASH_DIR);
-        strcat (tp, path);
+        gf_log (this->name, GF_LOG_TRACE,
+                "expunging file %s on %s",
+                expunge_local->loc.path, priv->children[active_src]->name);
 
-        c = strchr (tp, '/') + 1;
-        while (*c++)
-                if (*c == '/')
-                        *c = '-';
+        STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_remove_cbk,
+                           (void *) (long) active_src,
+                           priv->children[active_src],
+                           priv->children[active_src]->fops->unlink,
+                           &expunge_local->loc, 0, NULL);
 
-        return tp;
+        return 0;
 }
 
 
+
 int
-afr_sh_entry_expunge_rename (call_frame_t *expunge_frame, xlator_t *this,
-                             int active_src, inode_t *trash_inode)
+afr_sh_entry_expunge_rmdir (call_frame_t *expunge_frame, xlator_t *this,
+                            int active_src)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-
-        loc_t rename_loc;
-
-	priv          = this->private;
-	expunge_local = expunge_frame->local;
-
-        rename_loc.inode = inode_ref (expunge_local->loc.inode);
-        rename_loc.path  = make_trash_path (expunge_local->loc.path);
-        rename_loc.name  = strrchr (rename_loc.path, '/') + 1;
-        rename_loc.parent = trash_inode;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *expunge_local = NULL;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"moving file/directory %s on %s to %s",
-		expunge_local->loc.path, priv->children[active_src]->name,
-                rename_loc.path);
+        priv          = this->private;
+        expunge_local = expunge_frame->local;
 
-	STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_rename_cbk,
-			   (void *) (long) active_src,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->rename,
-			   &expunge_local->loc, &rename_loc);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "expunging directory %s on %s",
+                expunge_local->loc.path, priv->children[active_src]->name);
 
-        loc_wipe (&rename_loc);
+        STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_remove_cbk,
+                           (void *) (long) active_src,
+                           priv->children[active_src],
+                           priv->children[active_src]->fops->rmdir,
+                           &expunge_local->loc, 1, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
-afr_sh_entry_expunge_mkdir_cbk (call_frame_t *expunge_frame, void *cookie, xlator_t *this,
-                                int32_t op_ret, int32_t op_errno, inode_t *inode,
-                                struct iatt *buf, struct iatt *preparent,
-                                struct iatt *postparent)
+afr_sh_entry_expunge_remove (call_frame_t *expunge_frame, xlator_t *this,
+                             int active_src, struct iatt *buf,
+                             struct iatt *parentbuf)
 {
-        afr_private_t *priv            = NULL;
+        afr_private_t   *priv = NULL;
         afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh    = NULL;
-        call_frame_t    *frame         = NULL;
-
-        int active_src = (long) cookie;
-
-        inode_t *trash_inode = NULL;
-
-        priv          = this->private;
+        afr_self_heal_t *expunge_sh = NULL;
+        call_frame_t    *frame = NULL;
+        int              type = 0;
+        afr_self_heal_t *sh            = NULL;
+        afr_local_t     *local         = NULL;
+        loc_t           *loc           = NULL;
+
+        priv = this->private;
         expunge_local = expunge_frame->local;
-	expunge_sh    = &expunge_local->self_heal;
-	frame         = expunge_sh->sh_frame;
-
-        if (op_ret != 0) {
+        expunge_sh = &expunge_local->self_heal;
+        frame = expunge_sh->sh_frame;
+        local         = frame->local;
+        sh            = &local->self_heal;
+        loc           = &expunge_local->loc;
+
+        type = buf->ia_type;
+        if (loc->parent && uuid_is_null (loc->parent->gfid))
+                uuid_copy (loc->pargfid, parentbuf->ia_gfid);
+
+        switch (type) {
+        case IA_IFSOCK:
+        case IA_IFREG:
+        case IA_IFBLK:
+        case IA_IFCHR:
+        case IA_IFIFO:
+        case IA_IFLNK:
+                afr_sh_entry_expunge_unlink (expunge_frame, this, active_src);
+                break;
+        case IA_IFDIR:
+                afr_sh_entry_expunge_rmdir (expunge_frame, this, active_src);
+                break;
+        default:
                 gf_log (this->name, GF_LOG_ERROR,
-                        "mkdir of /" GF_REPLICATE_TRASH_DIR " failed on %s",
-                        priv->children[active_src]->name);
-
+                        "%s has unknown file type on %s: 0%o",
+                        expunge_local->loc.path,
+                        priv->children[active_src]->name, type);
                 goto out;
+                break;
         }
 
-        /* mkdir successful */
-
-        trash_inode = inode_link (inode, expunge_local->loc.inode->table->root,
-                                  GF_REPLICATE_TRASH_DIR, buf);
-
-        afr_sh_entry_expunge_rename (expunge_frame, this, active_src,
-                                     trash_inode);
         return 0;
 out:
         AFR_STACK_DESTROY (expunge_frame);
-        afr_sh_entry_expunge_entry_done (frame, this, active_src);
+        sh->expunge_done (frame, this, active_src, -1, EINVAL);
+
         return 0;
 }
 
 
 int
-afr_sh_entry_expunge_lookup_trash_cbk (call_frame_t *expunge_frame, void *cookie,
-                                       xlator_t *this,
-                                       int32_t op_ret, int32_t op_errno,
-                                       inode_t *inode, struct iatt *buf,
-                                       dict_t *xattr, struct iatt *postparent)
+afr_sh_entry_expunge_lookup_cbk (call_frame_t *expunge_frame, void *cookie,
+                                 xlator_t *this,
+                                 int32_t op_ret, int32_t op_errno,
+                                 inode_t *inode, struct iatt *buf, dict_t *x,
+                                 struct iatt *postparent)
 {
-        afr_private_t *priv            = NULL;
+        afr_private_t   *priv = NULL;
         afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh    = NULL;
-        call_frame_t    *frame         = NULL;
-        dict_t          *dict          = NULL;
-
-        int active_src = (long) cookie;
-        int ret = 0;
-
-        inode_t *trash_inode;
-        loc_t    trash_loc;
+        afr_self_heal_t *expunge_sh = NULL;
+        call_frame_t    *frame = NULL;
+        int              active_src = 0;
+        afr_self_heal_t *sh            = NULL;
+        afr_local_t     *local         = NULL;
 
-        priv          = this->private;
+        priv = this->private;
         expunge_local = expunge_frame->local;
-	expunge_sh    = &expunge_local->self_heal;
-	frame         = expunge_sh->sh_frame;
-
-        if ((op_ret != 0) && (op_errno == ENOENT)) {
-                init_trash_loc (&trash_loc, expunge_local->loc.inode->table);
-
-                dict = dict_new ();
-                if (!dict) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        goto out;
-                }
-
-                ret = afr_set_dict_gfid (dict, buf->ia_gfid);
-                if (ret)
-                        gf_log (this->name, GF_LOG_DEBUG, "gfid set failed");
-
-
-                gf_log (this->name, GF_LOG_TRACE,
-                        "creating directory " GF_REPLICATE_TRASH_DIR " on subvolume %s",
-                        priv->children[active_src]->name);
-
-                STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_mkdir_cbk,
-                                   (void *) (long) active_src,
-                                   priv->children[active_src],
-                                   priv->children[active_src]->fops->mkdir,
-                                   &trash_loc, 0777, dict);
-
-                loc_wipe (&trash_loc);
-                if (dict)
-                        dict_unref (dict);
-                return 0;
-        }
+        expunge_sh = &expunge_local->self_heal;
+        frame = expunge_sh->sh_frame;
+        active_src = (long) cookie;
+        local         = frame->local;
+        sh            = &local->self_heal;
 
-        if (op_ret != 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "lookup of /" GF_REPLICATE_TRASH_DIR " failed on %s",
-                        priv->children[active_src]->name);
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "lookup of %s on %s failed (%s)",
+                        expunge_local->loc.path,
+                        priv->children[active_src]->name,
+                        strerror (op_errno));
                 goto out;
         }
 
-        /* lookup successful */
+        afr_sh_entry_expunge_remove (expunge_frame, this, active_src, buf,
+                                     postparent);
 
-        trash_inode = inode_link (inode, expunge_local->loc.inode->table->root,
-                                  GF_REPLICATE_TRASH_DIR, buf);
-
-        afr_sh_entry_expunge_rename (expunge_frame, this, active_src,
-                                     trash_inode);
         return 0;
 out:
         AFR_STACK_DESTROY (expunge_frame);
-        afr_sh_entry_expunge_entry_done (frame, this, active_src);
+        sh->expunge_done (frame, this, active_src, op_ret, op_errno);
+
         return 0;
 }
 
 
 int
-afr_sh_entry_expunge_lookup_trash (call_frame_t *expunge_frame, xlator_t *this,
-                                   int active_src)
+afr_sh_entry_expunge_purge (call_frame_t *expunge_frame, xlator_t *this,
+                            int active_src)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *expunge_local = NULL;
 
-        inode_t *root  = NULL;
-        inode_t *trash = NULL;
-        loc_t trash_loc;
+        priv = this->private;
+        expunge_local = expunge_frame->local;
 
-	priv          = this->private;
-	expunge_local = expunge_frame->local;
+        gf_log (this->name, GF_LOG_TRACE,
+                "looking up %s on %s",
+                expunge_local->loc.path, priv->children[active_src]->name);
 
-        root = expunge_local->loc.inode->table->root;
+        STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_lookup_cbk,
+                           (void *) (long) active_src,
+                           priv->children[active_src],
+                           priv->children[active_src]->fops->lookup,
+                           &expunge_local->loc, NULL);
 
-        trash = inode_grep (root->table, root, GF_REPLICATE_TRASH_DIR);
+        return 0;
+}
 
-        if (trash) {
-                /* inode is in cache, so no need to mkdir */
+int
+afr_sh_entry_expunge_entry_cbk (call_frame_t *expunge_frame, void *cookie,
+                                xlator_t *this,
+                                int32_t op_ret, int32_t op_errno,
+                                inode_t *inode, struct iatt *buf, dict_t *x,
+                                struct iatt *postparent)
+{
+        afr_private_t   *priv = NULL;
+        afr_local_t     *expunge_local = NULL;
+        afr_self_heal_t *expunge_sh = NULL;
+        int              source = 0;
+        call_frame_t    *frame = NULL;
+        int              active_src = 0;
+        int              need_expunge = 0;
+        afr_self_heal_t *sh            = NULL;
+        afr_local_t     *local         = NULL;
+
+        priv = this->private;
+        expunge_local = expunge_frame->local;
+        expunge_sh = &expunge_local->self_heal;
+        frame = expunge_sh->sh_frame;
+        active_src = expunge_sh->active_source;
+        source = (long) cookie;
+        local         = frame->local;
+        sh            = &local->self_heal;
+
+        if (op_ret == -1 && op_errno == ENOENT)
+                need_expunge = 1;
+        else if (op_ret == -1)
+                goto out;
 
-                afr_sh_entry_expunge_rename (expunge_frame, this, active_src,
-                                             trash);
-                return 0;
+        if (!uuid_is_null (expunge_sh->entrybuf.ia_gfid) &&
+            !uuid_is_null (buf->ia_gfid) &&
+            (uuid_compare (expunge_sh->entrybuf.ia_gfid, buf->ia_gfid) != 0)) {
+                char uuidbuf1[64];
+                char uuidbuf2[64];
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "entry %s found on %s with mismatching gfid (%s/%s)",
+                        expunge_local->loc.path,
+                        priv->children[source]->name,
+                        uuid_utoa_r (expunge_sh->entrybuf.ia_gfid, uuidbuf1),
+                        uuid_utoa_r (buf->ia_gfid, uuidbuf2));
+                need_expunge = 1;
         }
 
-        /* Not in cache, so look it up */
+        if (need_expunge) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Entry %s is missing on %s and deleting from "
+                        "replica's other bricks",
+                        expunge_local->loc.path,
+                        priv->children[source]->name);
+
+                if (postparent)
+                        expunge_sh->parentbuf = *postparent;
 
-        init_trash_loc (&trash_loc, expunge_local->loc.inode->table);
+                afr_sh_entry_expunge_purge (expunge_frame, this, active_src);
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"looking up /" GF_REPLICATE_TRASH_DIR " on %s",
-                priv->children[active_src]->name);
+                return 0;
+        }
 
-	STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_lookup_trash_cbk,
-			   (void *) (long) active_src,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->lookup,
-			   &trash_loc, NULL);
+out:
+        if (op_ret == 0) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "%s exists under %s",
+                        expunge_local->loc.path,
+                        priv->children[source]->name);
+        } else {
+                gf_log (this->name, GF_LOG_INFO,
+                        "looking up %s under %s failed (%s)",
+                        expunge_local->loc.path,
+                        priv->children[source]->name,
+                        strerror (op_errno));
+        }
 
-        loc_wipe (&trash_loc);
+        AFR_STACK_DESTROY (expunge_frame);
+        sh->expunge_done (frame, this, active_src, op_ret, op_errno);
 
-	return 0;
+        return 0;
 }
 
-
-int
-afr_sh_entry_expunge_remove (call_frame_t *expunge_frame, xlator_t *this,
-			     int active_src, struct iatt *buf)
+static gf_boolean_t
+can_skip_entry_self_heal (char *name, loc_t *parent_loc)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	int              source = 0;
-	call_frame_t    *frame = NULL;
-	int              type = 0;
-
-	priv = this->private;
-	expunge_local = expunge_frame->local;
-	expunge_sh = &expunge_local->self_heal;
-	frame = expunge_sh->sh_frame;
-	source = expunge_sh->source;
-
-	type = buf->ia_type;
-
-	switch (type) {
-	case IA_IFSOCK:
-	case IA_IFREG:
-	case IA_IFBLK:
-	case IA_IFCHR:
-	case IA_IFIFO:
-	case IA_IFLNK:
-	case IA_IFDIR:
-		afr_sh_entry_expunge_lookup_trash (expunge_frame, this, active_src);
-		break;
-	default:
-		gf_log (this->name, GF_LOG_ERROR,
-			"%s has unknown file type on %s: 0%o",
-			expunge_local->loc.path,
-			priv->children[source]->name, type);
-		goto out;
-		break;
-	}
-
-	return 0;
-out:
-	AFR_STACK_DESTROY (expunge_frame);
-	afr_sh_entry_expunge_entry_done (frame, this, active_src);
-
-	return 0;
+        if (strcmp (name, ".") == 0) {
+                return _gf_true;
+        } else if (strcmp (name, "..") == 0) {
+                return _gf_true;
+        } else if (loc_is_root (parent_loc) &&
+                   (strcmp (name, GF_REPLICATE_TRASH_DIR) == 0)) {
+                return _gf_true;
+        }
+        return _gf_false;
 }
 
-
 int
-afr_sh_entry_expunge_lookup_cbk (call_frame_t *expunge_frame, void *cookie,
-				xlator_t *this,
-				int32_t op_ret,	int32_t op_errno,
-                                inode_t *inode, struct iatt *buf, dict_t *x,
-                                struct iatt *postparent)
+afr_sh_entry_expunge_entry (call_frame_t *frame, xlator_t *this,
+                            gf_dirent_t *entry)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
-
-	priv = this->private;
-	expunge_local = expunge_frame->local;
-	expunge_sh = &expunge_local->self_heal;
-	frame = expunge_sh->sh_frame;
-	active_src = (long) cookie;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"lookup of %s on %s failed (%s)",
-			expunge_local->loc.path,
-			priv->children[active_src]->name,
-			strerror (op_errno));
-		goto out;
-	}
-
-	afr_sh_entry_expunge_remove (expunge_frame, this, active_src, buf);
-
-	return 0;
-out:
-	AFR_STACK_DESTROY (expunge_frame);
-	afr_sh_entry_expunge_entry_done (frame, this, active_src);
+        afr_private_t   *priv = NULL;
+        afr_local_t     *local  = NULL;
+        afr_self_heal_t *sh  = NULL;
+        int              ret = -1;
+        call_frame_t    *expunge_frame = NULL;
+        afr_local_t     *expunge_local = NULL;
+        afr_self_heal_t *expunge_sh = NULL;
+        int              active_src = 0;
+        int              source = 0;
+        int              op_errno = 0;
+        char            *name = NULL;
+        int             op_ret = -1;
+
+        priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
 
-	return 0;
-}
+        active_src = sh->active_source;
+        source = sh->source;
+        sh->expunge_done = afr_sh_entry_expunge_entry_done;
 
+        name = entry->d_name;
+        if (can_skip_entry_self_heal (name, &local->loc)) {
+                op_ret = 0;
+                goto out;
+        }
 
-int
-afr_sh_entry_expunge_purge (call_frame_t *expunge_frame, xlator_t *this,
-			    int active_src)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-
-	priv = this->private;
-	expunge_local = expunge_frame->local;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"looking up %s on %s",
-		expunge_local->loc.path, priv->children[active_src]->name);
-	
-	STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_lookup_cbk,
-			   (void *) (long) active_src,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->lookup,
-			   &expunge_local->loc, 0);
-
-	return 0;
-}
+        gf_log (this->name, GF_LOG_TRACE,
+                "inspecting existence of %s under %s",
+                name, local->loc.path);
 
+        expunge_frame = copy_frame (frame);
+        if (!expunge_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-int
-afr_sh_entry_expunge_entry_cbk (call_frame_t *expunge_frame, void *cookie,
-				xlator_t *this,
-				int32_t op_ret,	int32_t op_errno,
-                                inode_t *inode, struct iatt *buf, dict_t *x,
-                                struct iatt *postparent)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	int              source = 0;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
-
-
-	priv = this->private;
-	expunge_local = expunge_frame->local;
-	expunge_sh = &expunge_local->self_heal;
-	frame = expunge_sh->sh_frame;
-	active_src = expunge_sh->active_source;
-	source = (long) cookie;
-
-	if (op_ret == -1 && op_errno == ENOENT && postparent) {
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"missing entry %s on %s",
-			expunge_local->loc.path,
-			priv->children[source]->name);
-
-                expunge_sh->parentbuf = *postparent;
-
-		afr_sh_entry_expunge_purge (expunge_frame, this, active_src);
-
-		return 0;
-	}
-
-	if (op_ret == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"%s exists under %s",
-			expunge_local->loc.path,
-			priv->children[source]->name);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"looking up %s under %s failed (%s)",
-			expunge_local->loc.path,
-			priv->children[source]->name,
-			strerror (op_errno));
-	}
-
-	AFR_STACK_DESTROY (expunge_frame);
-	afr_sh_entry_expunge_entry_done (frame, this, active_src);
-
-	return 0;
-}
+        AFR_LOCAL_ALLOC_OR_GOTO (expunge_local, out);
 
+        expunge_frame->local = expunge_local;
+        expunge_sh = &expunge_local->self_heal;
+        expunge_sh->sh_frame = frame;
+        expunge_sh->active_source = active_src;
+        expunge_sh->entrybuf = entry->d_stat;
+        loc_copy (&expunge_sh->parent_loc, &local->loc);
 
-int
-afr_sh_entry_expunge_entry (call_frame_t *frame, xlator_t *this,
-			    char *name)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              ret = -1;
-	call_frame_t    *expunge_frame = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	int              active_src = 0;
-	int              source = 0;
-	int              op_errno = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	active_src = sh->active_source;
-	source = sh->source;
-
-	if ((strcmp (name, ".") == 0)
-	    || (strcmp (name, "..") == 0)
-            || ((strcmp (local->loc.path, "/") == 0)
-                && (strcmp (name, GF_REPLICATE_TRASH_DIR) == 0))) {
+        ret = afr_build_child_loc (this, &expunge_local->loc, &local->loc,
+                                   name);
+        if (ret != 0) {
+                op_errno = EINVAL;
+                goto out;
+        }
 
-                gf_log (this->name, GF_LOG_TRACE,
-			"skipping inspection of %s under %s",
-			name, local->loc.path);
-		goto out;
-	}
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"inspecting existance of %s under %s",
-		name, local->loc.path);
-
-	expunge_frame = copy_frame (frame);
-	if (!expunge_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (expunge_local, afr_local_t, out);
-
-	expunge_frame->local = expunge_local;
-	expunge_sh = &expunge_local->self_heal;
-	expunge_sh->sh_frame = frame;
-	expunge_sh->active_source = active_src;
-
-	ret = build_child_loc (this, &expunge_local->loc, &local->loc, name);
-	if (ret != 0) {
-		goto out;
-	}
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"looking up %s on %s", expunge_local->loc.path,
-		priv->children[source]->name);
-
-	STACK_WIND_COOKIE (expunge_frame,
-			   afr_sh_entry_expunge_entry_cbk,
-			   (void *) (long) source,
-			   priv->children[source],
-			   priv->children[source]->fops->lookup,
-			   &expunge_local->loc, 0);
-
-	ret = 0;
+        gf_log (this->name, GF_LOG_TRACE,
+                "looking up %s on %s", expunge_local->loc.path,
+                priv->children[source]->name);
+
+        STACK_WIND_COOKIE (expunge_frame,
+                           afr_sh_entry_expunge_entry_cbk,
+                           (void *) (long) source,
+                           priv->children[source],
+                           priv->children[source]->fops->lookup,
+                           &expunge_local->loc, NULL);
+
+        ret = 0;
 out:
-	if (ret == -1)
-		afr_sh_entry_expunge_entry_done (frame, this, active_src);
+        if (ret == -1)
+                sh->expunge_done (frame, this, active_src, op_ret, op_errno);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_entry_expunge_readdir_cbk (call_frame_t *frame, void *cookie,
-				  xlator_t *this,
-				  int32_t op_ret, int32_t op_errno,
-				  gf_dirent_t *entries)
+                                  xlator_t *this,
+                                  int32_t op_ret, int32_t op_errno,
+                                  gf_dirent_t *entries, dict_t *xdata)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	gf_dirent_t     *entry = NULL;
-	off_t            last_offset = 0;
-	int              active_src = 0;
-	int              entry_count = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	active_src = sh->active_source;
-
-	if (op_ret <= 0) {
-		if (op_ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"readdir of %s on subvolume %s failed (%s)",
-				local->loc.path,
-				priv->children[active_src]->name,
-				strerror (op_errno));
-		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"readdir of %s on subvolume %s complete",
-				local->loc.path,
-				priv->children[active_src]->name);
-		}
-
-		afr_sh_entry_expunge_all (frame, this);
-		return 0;
-	}
-
-	list_for_each_entry (entry, &entries->list, list) {
-		last_offset = entry->d_off;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *local  = NULL;
+        afr_self_heal_t *sh  = NULL;
+        gf_dirent_t     *entry = NULL;
+        off_t            last_offset = 0;
+        int              active_src = 0;
+        int              entry_count = 0;
+
+        priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
+
+        active_src = sh->active_source;
+
+        if (op_ret <= 0) {
+                if (op_ret < 0) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "readdir of %s on subvolume %s failed (%s)",
+                                local->loc.path,
+                                priv->children[active_src]->name,
+                                strerror (op_errno));
+                } else {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "readdir of %s on subvolume %s complete",
+                                local->loc.path,
+                                priv->children[active_src]->name);
+                }
+
+                afr_sh_entry_expunge_all (frame, this);
+                return 0;
+        }
+
+        list_for_each_entry (entry, &entries->list, list) {
+                last_offset = entry->d_off;
                 entry_count++;
-	}
+        }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"readdir'ed %d entries from %s",
-		entry_count, priv->children[active_src]->name);
+        gf_log (this->name, GF_LOG_TRACE,
+                "readdir'ed %d entries from %s",
+                entry_count, priv->children[active_src]->name);
 
-	sh->offset = last_offset;
-	local->call_count = entry_count;
+        sh->offset = last_offset;
+        local->call_count = entry_count;
 
-	list_for_each_entry (entry, &entries->list, list) {
-                afr_sh_entry_expunge_entry (frame, this, entry->d_name);
-	}
+        list_for_each_entry (entry, &entries->list, list) {
+                afr_sh_entry_expunge_entry (frame, this, entry);
+        }
 
-	return 0;
+        return 0;
 }
 
 int
 afr_sh_entry_expunge_subvol (call_frame_t *frame, xlator_t *this,
-			     int active_src)
+                             int active_src)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *local  = NULL;
+        afr_self_heal_t *sh  = NULL;
 
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
+        priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
 
-	STACK_WIND (frame, afr_sh_entry_expunge_readdir_cbk,
-		    priv->children[active_src],
-		    priv->children[active_src]->fops->readdir,
-		    sh->healing_fd, sh->block_size, sh->offset);
+        STACK_WIND (frame, afr_sh_entry_expunge_readdir_cbk,
+                    priv->children[active_src],
+                    priv->children[active_src]->fops->readdirp,
+                    sh->healing_fd, sh->block_size, sh->offset, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_entry_expunge_all (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              active_src = -1;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *local  = NULL;
+        afr_self_heal_t *sh  = NULL;
+        int              active_src = -1;
 
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
+        priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
 
-	sh->offset = 0;
+        sh->offset = 0;
 
-	if (sh->source == -1) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"no active sources for %s to expunge entries",
-			local->loc.path);
-		goto out;
-	}
+        if (sh->source == -1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no active sources for %s to expunge entries",
+                        local->loc.path);
+                goto out;
+        }
 
-	active_src = next_active_sink (frame, this, sh->active_source);
-	sh->active_source = active_src;
+        active_src = next_active_sink (frame, this, sh->active_source);
+        sh->active_source = active_src;
 
-	if (sh->op_failed) {
-		goto out;
-	}
+        if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+                goto out;
+        }
 
-	if (active_src == -1) {
-		/* completed creating missing files on all subvolumes */
-		goto out;
-	}
+        if (active_src == -1) {
+                /* completed creating missing files on all subvolumes */
+                goto out;
+        }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"expunging entries of %s on %s to other sinks",
-		local->loc.path, priv->children[active_src]->name);
+        gf_log (this->name, GF_LOG_TRACE,
+                "expunging entries of %s on %s to other sinks",
+                local->loc.path, priv->children[active_src]->name);
 
-	afr_sh_entry_expunge_subvol (frame, this, active_src);
+        afr_sh_entry_expunge_subvol (frame, this, active_src);
 
-	return 0;
+        return 0;
 out:
-	afr_sh_entry_erase_pending (frame, this);
-	return 0;
+        afr_sh_entry_impunge_all (frame, this);
+        return 0;
 
 }
 
 
 int
-afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this);
-
-int
-afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this,
-			     int active_src);
-
-int
 afr_sh_entry_impunge_entry_done (call_frame_t *frame, xlator_t *this,
-				 int active_src)
+                                 int32_t op_ret, int32_t op_errno)
 {
-	int              call_count = 0;
-
-	call_count = afr_frame_return (frame);
+        int              call_count = 0;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
 
-	if (call_count == 0)
-		afr_sh_entry_impunge_subvol (frame, this, active_src);
+        local = frame->local;
+        sh = &local->self_heal;
+        if (op_ret < 0)
+                sh->entries_skipped = _gf_true;
+        call_count = afr_frame_return (frame);
+        if (call_count == 0)
+                afr_sh_entry_impunge_subvol (frame, this);
 
-	return 0;
+        return 0;
 }
 
-
-int
-afr_sh_entry_impunge_setattr_cbk (call_frame_t *impunge_frame, void *cookie,
-				  xlator_t *this,
-                                  int32_t op_ret, int32_t op_errno,
-                                  struct iatt *preop, struct iatt *postop)
+void
+afr_sh_entry_call_impunge_done (call_frame_t *impunge_frame, xlator_t *this,
+                                int32_t op_ret, int32_t op_errno)
 {
-	int              call_count = 0;
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
-	int              child_index = 0;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	frame = impunge_sh->sh_frame;
-        local = frame->local;
-        sh    = &local->self_heal;
-        active_src = sh->active_source;
-	child_index = (long) cookie;
-
-	if (op_ret == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"setattr done for %s on %s",
-			impunge_local->loc.path,
-			priv->children[child_index]->name);
-	} else {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"setattr (%s) on %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[child_index]->name,
-			strerror (op_errno));
-	}
-
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-	}
-
-	return 0;
-}
+        afr_local_t     *impunge_local = NULL;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_self_heal_t *impunge_sh = NULL;
+        call_frame_t    *frame = NULL;
 
+        AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+                                frame, local, sh);
+
+        AFR_STACK_DESTROY (impunge_frame);
+        sh->impunge_done (frame, this, op_ret, op_errno);
+}
 
 int
-afr_sh_entry_impunge_xattrop_cbk (call_frame_t *impunge_frame, void *cookie,
+afr_sh_entry_impunge_setattr_cbk (call_frame_t *impunge_frame, void *cookie,
                                   xlator_t *this,
                                   int32_t op_ret, int32_t op_errno,
-                                  dict_t *xattr)
+                                  struct iatt *preop, struct iatt *postop,
+                                  dict_t *xdata)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              child_index = 0;
-
-        struct iatt stbuf;
-        int32_t     valid = 0;
-
-	priv          = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh    = &impunge_local->self_heal;
-
-	child_index = (long) cookie;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"setting ownership of %s on %s to %d/%d",
-		impunge_local->loc.path,
-		priv->children[child_index]->name,
-		impunge_local->cont.lookup.buf.ia_uid,
-		impunge_local->cont.lookup.buf.ia_gid);
-
-	stbuf.ia_atime = impunge_local->cont.lookup.buf.ia_atime;
-	stbuf.ia_atime_nsec = impunge_local->cont.lookup.buf.ia_atime_nsec;
-	stbuf.ia_mtime = impunge_local->cont.lookup.buf.ia_mtime;
-	stbuf.ia_mtime_nsec = impunge_local->cont.lookup.buf.ia_mtime_nsec;
-
-        stbuf.ia_uid = impunge_local->cont.lookup.buf.ia_uid;
-        stbuf.ia_gid = impunge_local->cont.lookup.buf.ia_gid;
-
-        valid = GF_SET_ATTR_UID   | GF_SET_ATTR_GID |
-                GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_setattr_cbk,
-			   (void *) (long) child_index,
-			   priv->children[child_index],
-			   priv->children[child_index]->fops->setattr,
-			   &impunge_local->loc,
-                           &stbuf, valid);
+        int              call_count = 0;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *impunge_local = NULL;
+        int              child_index = 0;
+
+        priv = this->private;
+        impunge_local = impunge_frame->local;
+        child_index = (long) cookie;
+
+        if (op_ret == 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "setattr done for %s on %s",
+                        impunge_local->loc.path,
+                        priv->children[child_index]->name);
+        } else {
+                gf_log (this->name, GF_LOG_INFO,
+                        "setattr (%s) on %s failed (%s)",
+                        impunge_local->loc.path,
+                        priv->children[child_index]->name,
+                        strerror (op_errno));
+        }
+
+        call_count = afr_frame_return (impunge_frame);
+        if (call_count == 0) {
+                afr_sh_entry_call_impunge_done (impunge_frame, this,
+                                                0, op_errno);
+        }
+
         return 0;
 }
 
-
 int
 afr_sh_entry_impunge_parent_setattr_cbk (call_frame_t *setattr_frame,
                                          void *cookie, xlator_t *this,
                                          int32_t op_ret, int32_t op_errno,
-                                         struct iatt *preop, struct iatt *postop)
+                                         struct iatt *preop, struct iatt *postop,
+                                         dict_t *xdata)
 {
-        loc_t *parent_loc = cookie;
+        int             call_count = 0;
+        afr_local_t     *setattr_local = NULL;
 
+        setattr_local = setattr_frame->local;
         if (op_ret != 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "setattr on parent directory failed: %s",
-                        strerror (op_errno));
+                gf_log (this->name, GF_LOG_INFO,
+                        "setattr on parent directory (%s) failed: %s",
+                        setattr_local->loc.path, strerror (op_errno));
         }
 
-        loc_wipe (parent_loc);
+        call_count = afr_frame_return (setattr_frame);
+        if (call_count == 0)
+                AFR_STACK_DESTROY (setattr_frame);
+        return 0;
+}
+
+int
+afr_sh_entry_impunge_setattr (call_frame_t *impunge_frame, xlator_t *this)
+{
+        afr_private_t   *priv = NULL;
+        afr_local_t     *impunge_local = NULL;
+        afr_local_t     *setattr_local = NULL;
+        afr_self_heal_t *impunge_sh = NULL;
+        call_frame_t    *setattr_frame = NULL;
+        int32_t          valid = 0;
+        int32_t          op_errno = 0;
+        int              child_index = 0;
+        int              call_count = 0;
+        int              i = 0;
+
+        priv          = this->private;
+        impunge_local = impunge_frame->local;
+        impunge_sh    = &impunge_local->self_heal;
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "setting ownership of %s on %s to %d/%d",
+                impunge_local->loc.path,
+                priv->children[child_index]->name,
+                impunge_sh->entrybuf.ia_uid,
+                impunge_sh->entrybuf.ia_gid);
+
+        setattr_frame = copy_frame (impunge_frame);
+        if (!setattr_frame) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+        AFR_LOCAL_ALLOC_OR_GOTO (setattr_frame->local, out);
+        setattr_local = setattr_frame->local;
+        call_count = afr_errno_count (NULL, impunge_sh->child_errno,
+                                      priv->child_count, 0);
+        loc_copy (&setattr_local->loc, &impunge_sh->parent_loc);
+        impunge_local->call_count = call_count;
+        setattr_local->call_count = call_count;
+        for (i = 0; i < priv->child_count; i++) {
+                if (impunge_sh->child_errno[i])
+                        continue;
+                valid         = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
+                STACK_WIND_COOKIE (setattr_frame,
+                                   afr_sh_entry_impunge_parent_setattr_cbk,
+                                   (void *) (long) i, priv->children[i],
+                                   priv->children[i]->fops->setattr,
+                                   &setattr_local->loc,
+                                   &impunge_sh->parentbuf, valid, NULL);
+
+                valid = GF_SET_ATTR_UID   | GF_SET_ATTR_GID |
+                        GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
+                STACK_WIND_COOKIE (impunge_frame,
+                                   afr_sh_entry_impunge_setattr_cbk,
+                                   (void *) (long) i, priv->children[i],
+                                   priv->children[i]->fops->setattr,
+                                   &impunge_local->loc,
+                                   &impunge_sh->entrybuf, valid, NULL);
+                call_count--;
+        }
+        GF_ASSERT (!call_count);
+        return 0;
+out:
+        if (setattr_frame)
+                AFR_STACK_DESTROY (setattr_frame);
+        afr_sh_entry_call_impunge_done (impunge_frame, this, 0, op_errno);
+        return 0;
+}
+
+int
+afr_sh_entry_impunge_xattrop_cbk (call_frame_t *impunge_frame, void *cookie,
+                                  xlator_t *this,
+                                  int32_t op_ret, int32_t op_errno,
+                                  dict_t *xattr, dict_t *xdata)
+{
+        afr_private_t   *priv = NULL;
+        afr_local_t     *impunge_local = NULL;
+        int              child_index = 0;
+
+        priv          = this->private;
+        impunge_local = impunge_frame->local;
 
-        GF_FREE (parent_loc);
+        child_index = (long) cookie;
 
-        AFR_STACK_DESTROY (setattr_frame);
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "%s: failed to perform xattrop on %s (%s)",
+                        impunge_local->loc.path,
+                        priv->children[child_index]->name,
+                        strerror (op_errno));
+                goto out;
+        }
+
+        afr_sh_entry_impunge_setattr (impunge_frame, this);
+        return 0;
+out:
+        afr_sh_entry_call_impunge_done (impunge_frame, this,
+                                        -1, op_errno);
         return 0;
 }
 
+int
+afr_sh_entry_impunge_perform_xattrop (call_frame_t *impunge_frame,
+                                      xlator_t *this)
+{
+        int              active_src       = 0;
+        dict_t          *xattr            = NULL;
+        afr_private_t   *priv             = NULL;
+        afr_local_t     *impunge_local    = NULL;
+        afr_self_heal_t *impunge_sh       = NULL;
+        int32_t         op_errno          = 0;
+
+        priv = this->private;
+        impunge_local = impunge_frame->local;
+        impunge_sh = &impunge_local->self_heal;
+        active_src = impunge_sh->active_source;
+
+        afr_prepare_new_entry_pending_matrix (impunge_local->pending,
+                                              afr_is_errno_unset,
+                                              impunge_sh->child_errno,
+                                              &impunge_sh->entrybuf,
+                                              priv->child_count);
+        xattr = dict_new ();
+        if (!xattr) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        afr_set_pending_dict (priv, xattr, impunge_local->pending, active_src,
+                              LOCAL_LAST);
+
+        STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_xattrop_cbk,
+                           (void *) (long) active_src,
+                           priv->children[active_src],
+                           priv->children[active_src]->fops->xattrop,
+                           &impunge_local->loc, GF_XATTROP_ADD_ARRAY, xattr, NULL);
+
+        if (xattr)
+                dict_unref (xattr);
+        return 0;
+out:
+        afr_sh_entry_call_impunge_done (impunge_frame, this,
+                                        -1, op_errno);
+        return 0;
+}
 
 int
 afr_sh_entry_impunge_newfile_cbk (call_frame_t *impunge_frame, void *cookie,
-				  xlator_t *this,
-				  int32_t op_ret, int32_t op_errno,
+                                  xlator_t *this,
+                                  int32_t op_ret, int32_t op_errno,
                                   inode_t *inode, struct iatt *stbuf,
                                   struct iatt *preparent,
-                                  struct iatt *postparent)
+                                  struct iatt *postparent, dict_t *xdata)
 {
-	int              call_count = 0;
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
-	int              child_index = 0;
-        int              pending_array[3] = {0, };
-        dict_t          *xattr = NULL;
-        int              ret = 0;
-        int              idx = 0;
-        afr_local_t     *local = NULL;
-        afr_self_heal_t *sh = NULL;
-
-        call_frame_t *setattr_frame = NULL;
-        int32_t valid = 0;
-        loc_t *parent_loc = NULL;
-        struct iatt parentbuf;
+        int              call_count       = 0;
+        afr_private_t   *priv             = NULL;
+        afr_local_t     *impunge_local    = NULL;
+        afr_self_heal_t *impunge_sh       = NULL;
+        int              child_index      = 0;
 
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	frame = impunge_sh->sh_frame;
-        local = frame->local;
-        sh    = &local->self_heal;
-        active_src = sh->active_source;
+        priv = this->private;
+        impunge_local = impunge_frame->local;
+        impunge_sh = &impunge_local->self_heal;
 
-	child_index = (long) cookie;
+        child_index = (long) cookie;
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"creation of %s on %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[child_index]->name,
-			strerror (op_errno));
-		goto out;
-	}
+        if (op_ret == -1) {
+                impunge_sh->child_errno[child_index] = op_errno;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "creation of %s on %s failed (%s)",
+                        impunge_local->loc.path,
+                        priv->children[child_index]->name,
+                        strerror (op_errno));
+        } else {
+                impunge_sh->child_errno[child_index] = 0;
+        }
 
-	inode->ia_type = stbuf->ia_type;
+        call_count = afr_frame_return (impunge_frame);
+        if (call_count == 0) {
+                if (!afr_errno_count (NULL, impunge_sh->child_errno,
+                                      priv->child_count, 0)) {
+                        // new_file creation failed every where
+                        afr_sh_entry_call_impunge_done (impunge_frame, this,
+                                                        -1, op_errno);
+                        goto out;
+                }
+                afr_sh_entry_impunge_perform_xattrop (impunge_frame, this);
+        }
+out:
+        return 0;
+}
 
-        xattr = get_new_dict ();
-        dict_ref (xattr);
+int
+afr_sh_entry_impunge_hardlink_cbk (call_frame_t *impunge_frame, void *cookie,
+                                   xlator_t *this, int32_t op_ret,
+                                   int32_t op_errno, inode_t *inode,
+                                   struct iatt *buf, struct iatt *preparent,
+                                   struct iatt *postparent, dict_t *xdata)
+{
+        int              call_count        = 0;
+        afr_local_t     *impunge_local = NULL;
+        afr_self_heal_t *impunge_sh  = NULL;
+
+        impunge_local = impunge_frame->local;
+        impunge_sh = &impunge_local->self_heal;
+
+        if (IA_IFLNK == impunge_sh->entrybuf.ia_type) {
+                //For symlinks impunge is attempted un-conditionally
+                //So the file can already exist.
+                if ((op_ret < 0) && (op_errno == EEXIST))
+                        op_ret = 0;
+        }
 
-        idx = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
-        pending_array[idx] = hton32 (1);
-        if (IA_ISDIR (stbuf->ia_type))
-                idx = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
-        else
-                idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
-        pending_array[idx] = hton32 (1);
+        call_count = afr_frame_return (impunge_frame);
+        if (call_count == 0)
+                afr_sh_entry_call_impunge_done (impunge_frame, this,
+                                                op_ret, op_errno);
 
-        ret = dict_set_static_bin (xattr, priv->pending_key[child_index],
-                                   pending_array, sizeof (pending_array));
-        if (ret < 0)
-                gf_log (this->name, GF_LOG_WARNING,
-                        "Unable to set dict value.");
+        return 0;
+}
 
-        valid         = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-        parentbuf     = impunge_sh->parentbuf;
-        setattr_frame = copy_frame (impunge_frame);
+int
+afr_sh_entry_impunge_hardlink (call_frame_t *impunge_frame, xlator_t *this,
+                               int child_index)
+{
+        afr_private_t   *priv          = NULL;
+        afr_local_t     *impunge_local = NULL;
+        afr_self_heal_t *impunge_sh  = NULL;
+        loc_t           *loc           = NULL;
+        struct iatt     *buf           = NULL;
+        loc_t            oldloc        = {0};
 
-        parent_loc = GF_CALLOC (1, sizeof (*parent_loc),
-                                gf_afr_mt_loc_t);
-        afr_build_parent_loc (parent_loc, &impunge_local->loc);
+        priv = this->private;
+        impunge_local = impunge_frame->local;
+        impunge_sh = &impunge_local->self_heal;
+        loc = &impunge_local->loc;
+        buf = &impunge_sh->entrybuf;
 
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_xattrop_cbk,
-			   (void *) (long) child_index,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->xattrop,
-			   &impunge_local->loc, GF_XATTROP_ADD_ARRAY, xattr);
+        oldloc.inode = inode_ref (loc->inode);
+        uuid_copy (oldloc.gfid, buf->ia_gfid);
+        gf_log (this->name, GF_LOG_DEBUG, "linking missing file %s on %s",
+                loc->path, priv->children[child_index]->name);
 
-        STACK_WIND_COOKIE (setattr_frame, afr_sh_entry_impunge_parent_setattr_cbk,
-                           (void *) (long) parent_loc,
+        STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_hardlink_cbk,
+                           (void *) (long) child_index,
                            priv->children[child_index],
-                           priv->children[child_index]->fops->setattr,
-                           parent_loc, &parentbuf, valid);
+                           priv->children[child_index]->fops->link,
+                           &oldloc, loc, NULL);
+        loc_wipe (&oldloc);
 
-        dict_unref (xattr);
+        return 0;
+}
+
+int
+afr_sh_nameless_lookup_cbk (call_frame_t *impunge_frame, void *cookie,
+                            xlator_t *this,
+                            int32_t op_ret, int32_t op_errno, inode_t *inode,
+                            struct iatt *buf, dict_t *xattr,
+                            struct iatt *postparent)
+{
+        if (op_ret < 0) {
+                 afr_sh_entry_impunge_create_file (impunge_frame, this,
+                                                   (long)cookie);
+        } else {
+                afr_sh_entry_impunge_hardlink (impunge_frame, this,
+                                               (long)cookie);
+        }
+        return 0;
+}
 
-	return 0;
+int
+afr_sh_entry_impunge_check_hardlink (call_frame_t *impunge_frame,
+                                     xlator_t *this,
+                                     int child_index, struct iatt *stbuf)
+{
+        afr_private_t   *priv          = NULL;
+        call_frame_t    *frame             = NULL;
+        afr_local_t     *impunge_local     = NULL;
+        afr_local_t     *local             = NULL;
+        afr_self_heal_t *impunge_sh        = NULL;
+        afr_self_heal_t *sh                = NULL;
+        loc_t           *loc           = NULL;
+        dict_t          *xattr_req     = NULL;
+        loc_t            oldloc        = {0};
+        int              ret           = -1;
+
+        priv = this->private;
+        AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+                                frame, local, sh);
+        loc = &impunge_local->loc;
+
+        xattr_req = dict_new ();
+        if (!xattr_req)
+                goto out;
+        oldloc.inode = inode_ref (loc->inode);
+        uuid_copy (oldloc.gfid, stbuf->ia_gfid);
 
+        STACK_WIND_COOKIE (impunge_frame, afr_sh_nameless_lookup_cbk,
+                           (void *) (long) child_index,
+                           priv->children[child_index],
+                           priv->children[child_index]->fops->lookup,
+                           &oldloc, xattr_req);
+        ret = 0;
 out:
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-	}
-
-	return 0;
+        if (xattr_req)
+                dict_unref (xattr_req);
+        loc_wipe (&oldloc);
+        if (ret)
+                sh->impunge_done (frame, this, -1, ENOMEM);
+        return 0;
 }
 
-
 int
 afr_sh_entry_impunge_mknod (call_frame_t *impunge_frame, xlator_t *this,
-			    int child_index, struct iatt *stbuf)
+                            int child_index, struct iatt *stbuf)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-        dict_t          *dict = NULL;
+        afr_private_t *priv          = NULL;
+        afr_local_t   *impunge_local = NULL;
+        dict_t        *dict          = NULL;
+        int            ret           = 0;
 
-        int ret = 0;
+        priv = this->private;
+        impunge_local = impunge_frame->local;
 
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-
-	gf_log (this->name, GF_LOG_DEBUG,
-		"creating missing file %s on %s",
-		impunge_local->loc.path,
-		priv->children[child_index]->name);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "creating missing file %s on %s",
+                impunge_local->loc.path,
+                priv->children[child_index]->name);
 
         dict = dict_new ();
         if (!dict)
                 gf_log (this->name, GF_LOG_ERROR, "Out of memory");
 
+        GF_ASSERT (!uuid_is_null (stbuf->ia_gfid));
         ret = afr_set_dict_gfid (dict, stbuf->ia_gfid);
         if (ret)
-                gf_log (this->name, GF_LOG_DEBUG, "gfid set failed");
+                gf_log (this->name, GF_LOG_INFO, "%s: gfid set failed",
+                        impunge_local->loc.path);
+
+        /*
+         * Reason for adding GLUSTERFS_INTERNAL_FOP_KEY :
+         *
+         * Problem:
+         * While a brick is down in a replica pair, lets say the user creates
+         * one file(file-A) and a hard link to that file(h-file-A). After the
+         * brick comes back up, entry self-heal is attempted on parent dir of
+         * these two files. As part of readdir in self-heal it reads both the
+         * entries file-A and h-file-A for both of them it does name less lookup
+         * to check if there are any hardlinks already present in the
+         * destination brick. It finds that there are no hard links already
+         * present for files file-A, h-file-A. Self-heal does mknods for both
+         * file-A and h-file-A. This leads to file-A and h-file-A not being
+         * hardlinks anymore.
+         *
+         * Fix: (More like shrinking of race-window, the race itself is still
+         * present in posix-mknod).
+         * If mknod comes with the presence of GLUSTERFS_INTERNAL_FOP_KEY then
+         * posix_mknod checks if there are already any gfid-links and does
+         * link() instead of mknod. There still can be a race where two
+         * posix_mknods same gfid see that
+         * gfid-link file is not present and proceeds with mknods and result in
+         * two different files with same gfid.
+         */
+        ret = dict_set_str (dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
+        if (ret)
+                gf_log (this->name, GF_LOG_INFO, "%s: %s set failed",
+                        impunge_local->loc.path, GLUSTERFS_INTERNAL_FOP_KEY);
 
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
-			   (void *) (long) child_index,
-			   priv->children[child_index],
-			   priv->children[child_index]->fops->mknod,
-			   &impunge_local->loc,
-			   st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type),
-                           stbuf->ia_rdev, dict);
+        STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
+                           (void *) (long) child_index,
+                           priv->children[child_index],
+                           priv->children[child_index]->fops->mknod,
+                           &impunge_local->loc,
+                           st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type),
+                           makedev (ia_major (stbuf->ia_rdev),
+                                    ia_minor (stbuf->ia_rdev)), 0, dict);
 
         if (dict)
                 dict_unref (dict);
 
-	return 0;
+        return 0;
 }
 
 
 
 int
 afr_sh_entry_impunge_mkdir (call_frame_t *impunge_frame, xlator_t *this,
-			    int child_index, struct iatt *stbuf)
+                            int child_index, struct iatt *stbuf)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *impunge_local = NULL;
         dict_t          *dict = NULL;
 
         int ret = 0;
 
-	priv = this->private;
-	impunge_local = impunge_frame->local;
+        priv = this->private;
+        impunge_local = impunge_frame->local;
 
         dict = dict_new ();
         if (!dict) {
@@ -1345,73 +1329,76 @@ afr_sh_entry_impunge_mkdir (call_frame_t *impunge_frame, xlator_t *this,
                 return 0;
         }
 
+        GF_ASSERT (!uuid_is_null (stbuf->ia_gfid));
         ret = afr_set_dict_gfid (dict, stbuf->ia_gfid);
         if (ret)
-                gf_log (this->name, GF_LOG_DEBUG, "gfid set failed");
-
-	gf_log (this->name, GF_LOG_DEBUG,
-		"creating missing directory %s on %s",
-		impunge_local->loc.path,
-		priv->children[child_index]->name);
-
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
-			   (void *) (long) child_index,
-			   priv->children[child_index],
-			   priv->children[child_index]->fops->mkdir,
-			   &impunge_local->loc,
+                gf_log (this->name, GF_LOG_INFO, "%s: gfid set failed",
+                        impunge_local->loc.path);
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "creating missing directory %s on %s",
+                impunge_local->loc.path,
+                priv->children[child_index]->name);
+
+        STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
+                           (void *) (long) child_index,
+                           priv->children[child_index],
+                           priv->children[child_index]->fops->mkdir,
+                           &impunge_local->loc,
                            st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type),
-                           dict);
+                           0, dict);
 
         if (dict)
                 dict_unref (dict);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_entry_impunge_symlink (call_frame_t *impunge_frame, xlator_t *this,
-			      int child_index, const char *linkname)
+                              int child_index, const char *linkname)
 {
-	afr_private_t   *priv          = NULL;
-	afr_local_t     *impunge_local = NULL;
+        afr_private_t   *priv          = NULL;
+        afr_local_t     *impunge_local = NULL;
         dict_t          *dict          = NULL;
         struct iatt     *buf           = NULL;
+        int              ret           = 0;
 
-        int ret = 0;
+        priv = this->private;
+        impunge_local = impunge_frame->local;
 
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-
-        buf = &impunge_local->cont.symlink.buf;
+        buf = &impunge_local->cont.dir_fop.buf;
 
         dict = dict_new ();
         if (!dict) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                afr_sh_entry_impunge_entry_done (impunge_frame, this, 0);
+                afr_sh_entry_call_impunge_done (impunge_frame, this,
+                                                -1, ENOMEM);
+                goto out;
         }
 
+        GF_ASSERT (!uuid_is_null (buf->ia_gfid));
         ret = afr_set_dict_gfid (dict, buf->ia_gfid);
         if (ret)
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "dict set gfid failed");
+                gf_log (this->name, GF_LOG_INFO,
+                        "%s: dict set gfid failed",
+                        impunge_local->loc.path);
 
-	gf_log (this->name, GF_LOG_DEBUG,
-		"creating missing symlink %s -> %s on %s",
-		impunge_local->loc.path, linkname,
-		priv->children[child_index]->name);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "creating missing symlink %s -> %s on %s",
+                impunge_local->loc.path, linkname,
+                priv->children[child_index]->name);
 
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
-			   (void *) (long) child_index,
-			   priv->children[child_index],
-			   priv->children[child_index]->fops->symlink,
-			   linkname, &impunge_local->loc, dict);
+        STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
+                           (void *) (long) child_index,
+                           priv->children[child_index],
+                           priv->children[child_index]->fops->symlink,
+                           linkname, &impunge_local->loc, 0, dict);
 
         if (dict)
                 dict_unref (dict);
-
-	return 0;
+out:
+        return 0;
 }
 
 
@@ -1420,50 +1407,45 @@ afr_sh_entry_impunge_symlink_unlink_cbk (call_frame_t *impunge_frame,
                                          void *cookie, xlator_t *this,
                                          int32_t op_ret, int32_t op_errno,
                                          struct iatt *preparent,
-                                         struct iatt *postparent)
+                                         struct iatt *postparent, dict_t *xdata)
 {
         afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              child_index = -1;
-	call_frame_t    *frame = NULL;
-	int              call_count = -1;
-	int              active_src = -1;
-
-	priv          = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh    = &impunge_local->self_heal;
-	frame         = impunge_sh->sh_frame;
-	active_src    = impunge_sh->active_source;
-
-	child_index = (long) cookie;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unlink of %s on %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[child_index]->name,
-			strerror (op_errno));
-		goto out;
-	}
+        afr_local_t     *impunge_local = NULL;
+        afr_self_heal_t *impunge_sh = NULL;
+        int              child_index = -1;
+        int              call_count = -1;
+
+        priv          = this->private;
+        impunge_local = impunge_frame->local;
+        impunge_sh    = &impunge_local->self_heal;
+
+        child_index = (long) cookie;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "unlink of %s on %s failed (%s)",
+                        impunge_local->loc.path,
+                        priv->children[child_index]->name,
+                        strerror (op_errno));
+                goto out;
+        }
 
         afr_sh_entry_impunge_symlink (impunge_frame, this, child_index,
                                       impunge_sh->linkname);
 
         return 0;
 out:
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-	}
-
-	return 0;
+        LOCK (&impunge_frame->lock);
+        {
+                call_count = --impunge_local->call_count;
+        }
+        UNLOCK (&impunge_frame->lock);
+
+        if (call_count == 0)
+                afr_sh_entry_call_impunge_done (impunge_frame, this,
+                                                op_ret, op_errno);
+
+        return 0;
 }
 
 
@@ -1471,22 +1453,22 @@ int
 afr_sh_entry_impunge_symlink_unlink (call_frame_t *impunge_frame, xlator_t *this,
                                      int child_index)
 {
-	afr_private_t   *priv          = NULL;
-	afr_local_t     *impunge_local = NULL;
+        afr_private_t   *priv          = NULL;
+        afr_local_t     *impunge_local = NULL;
 
-	priv          = this->private;
-	impunge_local = impunge_frame->local;
+        priv          = this->private;
+        impunge_local = impunge_frame->local;
 
-	gf_log (this->name, GF_LOG_DEBUG,
-		"unlinking symlink %s with wrong target on %s",
-		impunge_local->loc.path,
-		priv->children[child_index]->name);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "unlinking symlink %s with wrong target on %s",
+                impunge_local->loc.path,
+                priv->children[child_index]->name);
 
         STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_symlink_unlink_cbk,
                            (void *) (long) child_index,
                            priv->children[child_index],
                            priv->children[child_index]->fops->unlink,
-                           &impunge_local->loc);
+                           &impunge_local->loc, 0, NULL);
 
         return 0;
 }
@@ -1496,32 +1478,30 @@ int
 afr_sh_entry_impunge_readlink_sink_cbk (call_frame_t *impunge_frame, void *cookie,
                                         xlator_t *this,
                                         int32_t op_ret, int32_t op_errno,
-                                        const char *linkname, struct iatt *sbuf)
+                                        const char *linkname, struct iatt *sbuf, dict_t *xdata)
 {
         afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              child_index = -1;
-	call_frame_t    *frame = NULL;
-	int              call_count = -1;
-	int              active_src = -1;
-
-	priv          = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh    = &impunge_local->self_heal;
-	frame         = impunge_sh->sh_frame;
-	active_src    = impunge_sh->active_source;
-
-	child_index = (long) cookie;
-
-	if ((op_ret == -1) && (op_errno != ENOENT)) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"readlink of %s on %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[active_src]->name,
-			strerror (op_errno));
-		goto out;
-	}
+        afr_local_t     *impunge_local = NULL;
+        afr_self_heal_t *impunge_sh = NULL;
+        int              child_index = -1;
+        int              call_count = -1;
+        int              active_src = -1;
+
+        priv          = this->private;
+        impunge_local = impunge_frame->local;
+        impunge_sh    = &impunge_local->self_heal;
+        active_src    = impunge_sh->active_source;
+
+        child_index = (long) cookie;
+
+        if ((op_ret == -1) && (op_errno != ENOENT)) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "readlink of %s on %s failed (%s)",
+                        impunge_local->loc.path,
+                        priv->children[active_src]->name,
+                        strerror (op_errno));
+                goto out;
+        }
 
         /* symlink doesn't exist on the sink */
 
@@ -1548,18 +1528,17 @@ afr_sh_entry_impunge_readlink_sink_cbk (call_frame_t *impunge_frame, void *cooki
         }
 
 out:
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-	}
-
-	return 0;
+        LOCK (&impunge_frame->lock);
+        {
+                call_count = --impunge_local->call_count;
+        }
+        UNLOCK (&impunge_frame->lock);
+
+        if (call_count == 0)
+                afr_sh_entry_call_impunge_done (impunge_frame, this,
+                                                op_ret, op_errno);
+
+        return 0;
 }
 
 
@@ -1567,821 +1546,770 @@ int
 afr_sh_entry_impunge_readlink_sink (call_frame_t *impunge_frame, xlator_t *this,
                                     int child_index)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *impunge_local = NULL;
 
-	priv = this->private;
-	impunge_local = impunge_frame->local;
+        priv = this->private;
+        impunge_local = impunge_frame->local;
 
-	gf_log (this->name, GF_LOG_DEBUG,
-		"checking symlink target of %s on %s",
-		impunge_local->loc.path, priv->children[child_index]->name);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "checking symlink target of %s on %s",
+                impunge_local->loc.path, priv->children[child_index]->name);
 
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_sink_cbk,
-			   (void *) (long) child_index,
-			   priv->children[child_index],
-			   priv->children[child_index]->fops->readlink,
-			   &impunge_local->loc, 4096);
+        STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_sink_cbk,
+                           (void *) (long) child_index,
+                           priv->children[child_index],
+                           priv->children[child_index]->fops->readlink,
+                           &impunge_local->loc, 4096, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_entry_impunge_readlink_cbk (call_frame_t *impunge_frame, void *cookie,
-				   xlator_t *this,
-				   int32_t op_ret, int32_t op_errno,
-				   const char *linkname, struct iatt *sbuf)
+                                   xlator_t *this,
+                                   int32_t op_ret, int32_t op_errno,
+                                   const char *linkname, struct iatt *sbuf, dict_t *xdata)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              child_index = -1;
-	call_frame_t    *frame = NULL;
-	int              call_count = -1;
-	int              active_src = -1;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	frame = impunge_sh->sh_frame;
-	active_src = impunge_sh->active_source;
-
-	child_index = (long) cookie;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"readlink of %s on %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[active_src]->name,
-			strerror (op_errno));
-		goto out;
-	}
+        afr_private_t   *priv = NULL;
+        afr_local_t     *impunge_local = NULL;
+        afr_self_heal_t *impunge_sh = NULL;
+        int              child_index = -1;
+        int              call_count = -1;
+        int              active_src = -1;
+
+        priv = this->private;
+        impunge_local = impunge_frame->local;
+        impunge_sh = &impunge_local->self_heal;
+        active_src = impunge_sh->active_source;
+
+        child_index = (long) cookie;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "readlink of %s on %s failed (%s)",
+                        impunge_local->loc.path,
+                        priv->children[active_src]->name,
+                        strerror (op_errno));
+                goto out;
+        }
 
         impunge_sh->linkname = gf_strdup (linkname);
-	afr_sh_entry_impunge_readlink_sink (impunge_frame, this, child_index);
+        afr_sh_entry_impunge_readlink_sink (impunge_frame, this, child_index);
 
-	return 0;
+        return 0;
 
 out:
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-	}
-
-	return 0;
+        LOCK (&impunge_frame->lock);
+        {
+                call_count = --impunge_local->call_count;
+        }
+        UNLOCK (&impunge_frame->lock);
+
+        if (call_count == 0)
+                afr_sh_entry_call_impunge_done (impunge_frame, this,
+                                                op_ret, op_errno);
+
+        return 0;
 }
 
 
 int
 afr_sh_entry_impunge_readlink (call_frame_t *impunge_frame, xlator_t *this,
-			       int child_index, struct iatt *stbuf)
+                               int child_index, struct iatt *stbuf)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              active_src = -1;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	active_src = impunge_sh->active_source;
-        impunge_local->cont.symlink.buf = *stbuf;
-
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_cbk,
-			   (void *) (long) child_index,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->readlink,
-			   &impunge_local->loc, 4096);
-
-	return 0;
-}
+        afr_private_t   *priv = NULL;
+        afr_local_t     *impunge_local = NULL;
+        afr_self_heal_t *impunge_sh = NULL;
+        int              active_src = -1;
+
+        priv = this->private;
+        impunge_local = impunge_frame->local;
+        impunge_sh = &impunge_local->self_heal;
+        active_src = impunge_sh->active_source;
+        impunge_local->cont.dir_fop.buf = *stbuf;
 
+        STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_cbk,
+                           (void *) (long) child_index,
+                           priv->children[active_src],
+                           priv->children[active_src]->fops->readlink,
+                           &impunge_local->loc, 4096, NULL);
+
+        return 0;
+}
 
 int
-afr_sh_entry_impunge_recreate_lookup_cbk (call_frame_t *impunge_frame,
-					  void *cookie, xlator_t *this,
-					  int32_t op_ret, int32_t op_errno,
-					  inode_t *inode, struct iatt *buf,
-					  dict_t *xattr,struct iatt *postparent)
+afr_sh_entry_impunge_create (call_frame_t *impunge_frame, xlator_t *this,
+                             int child_index)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              active_src = 0;
-	int              type = 0;
-	int              child_index = 0;
-	call_frame_t    *frame = NULL;
-	int              call_count = 0;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	frame = impunge_sh->sh_frame;
-
-	child_index = (long) cookie;
-
-	active_src = impunge_sh->active_source;
-
-	if (op_ret != 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"looking up %s on %s (for %s) failed (%s)",
-			impunge_local->loc.path,
-			priv->children[active_src]->name,
-			priv->children[child_index]->name,
-			strerror (op_errno));
-		goto out;
-	}
-
-        impunge_sh->parentbuf = *postparent;
-
-	impunge_local->cont.lookup.buf = *buf;
-	type = buf->ia_type;
-
-	switch (type) {
-	case IA_IFSOCK:
-	case IA_IFREG:
-	case IA_IFBLK:
-	case IA_IFCHR:
-	case IA_IFIFO:
-		afr_sh_entry_impunge_mknod (impunge_frame, this,
-					    child_index, buf);
-		break;
-	case IA_IFLNK:
-		afr_sh_entry_impunge_readlink (impunge_frame, this,
-					       child_index, buf);
-		break;
-	case IA_IFDIR:
-		afr_sh_entry_impunge_mkdir (impunge_frame, this,
-					    child_index, buf);
-		break;
-	default:
-		gf_log (this->name, GF_LOG_ERROR,
-			"%s has unknown file type on %s: 0%o",
-			impunge_local->loc.path,
-			priv->children[active_src]->name, type);
-		goto out;
-		break;
-	}
-
-	return 0;
+        call_frame_t    *frame             = NULL;
+        afr_local_t     *impunge_local     = NULL;
+        afr_local_t     *local             = NULL;
+        afr_self_heal_t *impunge_sh        = NULL;
+        afr_self_heal_t *sh                = NULL;
+        afr_private_t   *priv = NULL;
+        ia_type_t       type = IA_INVAL;
+        int             active_src = 0;
+        struct iatt     *buf = NULL;
+
+        AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+                                frame, local, sh);
+        active_src = impunge_sh->active_source;
+        afr_update_loc_gfids (&impunge_local->loc, &impunge_sh->entrybuf,
+                              &impunge_sh->parentbuf);
+
+        buf = &impunge_sh->entrybuf;
+        type = buf->ia_type;
+
+        switch (type) {
+        case IA_IFSOCK:
+        case IA_IFREG:
+        case IA_IFBLK:
+        case IA_IFCHR:
+        case IA_IFIFO:
+        case IA_IFLNK:
+                afr_sh_entry_impunge_check_hardlink (impunge_frame, this,
+                                                     child_index, buf);
+                break;
+        case IA_IFDIR:
+                afr_sh_entry_impunge_mkdir (impunge_frame, this,
+                                            child_index, buf);
+                break;
+        default:
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s has unknown file type on %s: 0%o",
+                        impunge_local->loc.path,
+                        priv->children[active_src]->name, type);
+                sh->impunge_done (frame, this, -1, EINVAL);
+                break;
+        }
 
-out:
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-	}
-
-	return 0;
+        return 0;
 }
 
-
 int
-afr_sh_entry_impunge_recreate (call_frame_t *impunge_frame, xlator_t *this,
-			       int child_index)
+afr_sh_entry_impunge_create_file (call_frame_t *impunge_frame, xlator_t *this,
+                                  int child_index)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              active_src = 0;
+        call_frame_t    *frame             = NULL;
+        afr_local_t     *impunge_local     = NULL;
+        afr_local_t     *local             = NULL;
+        afr_self_heal_t *impunge_sh        = NULL;
+        afr_self_heal_t *sh                = NULL;
+        afr_private_t   *priv = NULL;
+        ia_type_t       type = IA_INVAL;
+        int             active_src = 0;
+        struct iatt     *buf = NULL;
+
+        AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+                                frame, local, sh);
+        active_src = impunge_sh->active_source;
+        buf = &impunge_sh->entrybuf;
+        type = buf->ia_type;
+
+        switch (type) {
+        case IA_IFSOCK:
+        case IA_IFREG:
+        case IA_IFBLK:
+        case IA_IFCHR:
+        case IA_IFIFO:
+                afr_sh_entry_impunge_mknod (impunge_frame, this,
+                                            child_index, buf);
+                break;
+        case IA_IFLNK:
+                afr_sh_entry_impunge_readlink (impunge_frame, this,
+                                               child_index, buf);
+                break;
+        default:
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s has unknown file type on %s: 0%o",
+                        impunge_local->loc.path,
+                        priv->children[active_src]->name, type);
+                sh->impunge_done (frame, this, -1, EINVAL);
+                break;
+        }
 
+        return 0;
+}
 
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
+gf_boolean_t
+afr_sh_need_recreate (afr_self_heal_t *impunge_sh, unsigned int child,
+                      unsigned int child_count)
+{
+        gf_boolean_t    recreate = _gf_false;
 
-	active_src = impunge_sh->active_source;
+        GF_ASSERT (impunge_sh->child_errno);
 
-	STACK_WIND_COOKIE (impunge_frame,
-			   afr_sh_entry_impunge_recreate_lookup_cbk,
-			   (void *) (long) child_index,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->lookup,
-			   &impunge_local->loc, 0);
+        if (child == impunge_sh->active_source)
+                goto out;
 
-	return 0;
+        if (IA_IFLNK == impunge_sh->entrybuf.ia_type) {
+                recreate = _gf_true;
+                goto out;
+        }
+
+        if (impunge_sh->child_errno[child] == ENOENT)
+                recreate = _gf_true;
+out:
+        return recreate;
 }
 
+unsigned int
+afr_sh_recreate_count (afr_self_heal_t *impunge_sh, int *sources,
+                       unsigned int child_count)
+{
+        int             count = 0;
+        int             i = 0;
+
+        for (i = 0; i < child_count; i++) {
+                if (afr_sh_need_recreate (impunge_sh, i, child_count))
+                        count++;
+        }
+
+        return count;
+}
 
 int
-afr_sh_entry_impunge_entry_cbk (call_frame_t *impunge_frame, void *cookie,
-				xlator_t *this,
-				int32_t op_ret,	int32_t op_errno,
-                                inode_t *inode, struct iatt *buf, dict_t *x,
-                                struct iatt *postparent)
+afr_sh_entry_call_impunge_recreate (call_frame_t *impunge_frame,
+                                    xlator_t *this)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	frame = impunge_sh->sh_frame;
-	child_index = (long) cookie;
-	active_src = impunge_sh->active_source;
-
-	if ((op_ret == -1 && op_errno == ENOENT)
-            || (IA_ISLNK (impunge_sh->impunging_entry_mode))) {
-
-                /*
-                 * A symlink's target might have changed, so
-                 * always go down the recreate path for them.
-                 */
+        afr_private_t   *priv = NULL;
+        afr_local_t     *impunge_local = NULL;
+        afr_self_heal_t *impunge_sh = NULL;
+        call_frame_t    *frame = NULL;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        unsigned int     recreate_count = 0;
+        int              i = 0;
+        int              active_src = 0;
 
-		/* decrease call_count in recreate-callback */
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"missing entry %s on %s",
-			impunge_local->loc.path,
-			priv->children[child_index]->name);
-
-		afr_sh_entry_impunge_recreate (impunge_frame, this,
-					       child_index);
-		return 0;
-	}
-
-	if (op_ret == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"%s exists under %s",
-			impunge_local->loc.path,
-			priv->children[child_index]->name);
-
-                impunge_sh->parentbuf = *postparent;
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"looking up %s under %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[child_index]->name,
-			strerror (op_errno));
-	}
-
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-	}
-
-	return 0;
+        priv          = this->private;
+        AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+                                frame, local, sh);
+        active_src = impunge_sh->active_source;
+        impunge_sh->entrybuf = impunge_sh->buf[active_src];
+        impunge_sh->parentbuf = impunge_sh->parentbufs[active_src];
+        recreate_count = afr_sh_recreate_count (impunge_sh, sh->sources,
+                                                priv->child_count);
+        if (!recreate_count) {
+                afr_sh_entry_call_impunge_done (impunge_frame, this, 0, 0);
+                goto out;
+        }
+        impunge_local->call_count = recreate_count;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!impunge_local->child_up[i]) {
+                        impunge_sh->child_errno[i] = ENOTCONN;
+                        continue;
+                }
+                if (!afr_sh_need_recreate (impunge_sh, i, priv->child_count)) {
+                        impunge_sh->child_errno[i] = EEXIST;
+                        continue;
+                }
+        }
+        for (i = 0; i < priv->child_count; i++) {
+                if (!afr_sh_need_recreate (impunge_sh, i, priv->child_count))
+                        continue;
+                (void)afr_sh_entry_impunge_create (impunge_frame, this, i);
+                recreate_count--;
+        }
+        GF_ASSERT (!recreate_count);
+out:
+        return 0;
 }
 
+void
+afr_sh_entry_common_lookup_done (call_frame_t *impunge_frame, xlator_t *this,
+                                 int32_t op_ret, int32_t op_errno)
+{
+        afr_private_t   *priv = NULL;
+        afr_local_t     *impunge_local = NULL;
+        afr_self_heal_t *impunge_sh = NULL;
+        call_frame_t    *frame = NULL;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        unsigned int     gfid_miss_count = 0;
+        unsigned int     children_up_count = 0;
+        uuid_t           gfid = {0};
+        int              active_src = 0;
+
+        priv          = this->private;
+        AFR_INIT_SH_FRAME_VALS (impunge_frame, impunge_local, impunge_sh,
+                                frame, local, sh);
+        active_src    = impunge_sh->active_source;
+
+        if (op_ret < 0)
+                goto done;
+        if (impunge_sh->child_errno[active_src]) {
+                op_ret = -1;
+                op_errno = impunge_sh->child_errno[active_src];
+                goto done;
+        }
+
+        gfid_miss_count = afr_gfid_missing_count (this->name,
+                                                  impunge_sh->success_children,
+                                                  impunge_sh->buf, priv->child_count,
+                                                  impunge_local->loc.path);
+        children_up_count = afr_up_children_count (impunge_local->child_up,
+                                                   priv->child_count);
+        if ((gfid_miss_count == children_up_count) &&
+            (children_up_count < priv->child_count)) {
+                op_ret = -1;
+                op_errno = ENODATA;
+                gf_log (this->name, GF_LOG_ERROR, "Not all children are up, "
+                        "gfid should not be assigned in this state for %s",
+                        impunge_local->loc.path);
+                goto done;
+        }
+
+        if (gfid_miss_count) {
+                afr_update_gfid_from_iatts (gfid, impunge_sh->buf,
+                                            impunge_sh->success_children,
+                                            priv->child_count);
+                if (uuid_is_null (gfid)) {
+                        sh->entries_skipped = _gf_true;
+                        gf_log (this->name, GF_LOG_INFO, "%s: Skipping entry "
+                                "self-heal because of gfid absence",
+                                impunge_local->loc.path);
+                        goto done;
+                }
+                afr_sh_common_lookup (impunge_frame, this, &impunge_local->loc,
+                                      afr_sh_entry_common_lookup_done, gfid,
+                                      AFR_LOOKUP_FAIL_CONFLICTS |
+                                      AFR_LOOKUP_FAIL_MISSING_GFIDS,
+                                      NULL);
+        } else {
+                afr_sh_entry_call_impunge_recreate (impunge_frame, this);
+        }
+        return;
+done:
+        afr_sh_entry_call_impunge_done (impunge_frame, this,
+                                        op_ret, op_errno);
+        return;
+}
 
 int
 afr_sh_entry_impunge_entry (call_frame_t *frame, xlator_t *this,
-			    gf_dirent_t *entry)
+                            gf_dirent_t *entry)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              ret = -1;
-	call_frame_t    *impunge_frame = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              active_src = 0;
-	int              i = 0;
-	int              call_count = 0;
-	int              op_errno = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	active_src = sh->active_source;
-
-	if ((strcmp (entry->d_name, ".") == 0)
-	    || (strcmp (entry->d_name, "..") == 0)
-            || ((strcmp (local->loc.path, "/") == 0)
-                && (strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR) == 0))) {
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"skipping inspection of %s under %s",
-			entry->d_name, local->loc.path);
-		goto out;
-	}
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"inspecting existance of %s under %s",
-		entry->d_name, local->loc.path);
-
-	impunge_frame = copy_frame (frame);
-	if (!impunge_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (impunge_local, afr_local_t, out);
-
-	impunge_frame->local = impunge_local;
-	impunge_sh = &impunge_local->self_heal;
-	impunge_sh->sh_frame = frame;
-	impunge_sh->active_source = active_src;
-
-        impunge_sh->impunging_entry_mode =
-                st_mode_from_ia (entry->d_stat.ia_prot, entry->d_stat.ia_type);
-
-	ret = build_child_loc (this, &impunge_local->loc, &local->loc, entry->d_name);
-	if (ret != 0) {
-		goto out;
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (i == active_src)
-			continue;
-		if (local->child_up[i] == 0)
-			continue;
-		if (sh->sources[i] == 1)
-			continue;
-		call_count++;
-	}
-
-	impunge_local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (i == active_src)
-			continue;
-		if (local->child_up[i] == 0)
-			continue;
-		if (sh->sources[i] == 1)
-			continue;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"looking up %s on %s", impunge_local->loc.path,
-			priv->children[i]->name);
-
-		STACK_WIND_COOKIE (impunge_frame,
-				   afr_sh_entry_impunge_entry_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->lookup,
-				   &impunge_local->loc, 0);
-
-		if (!--call_count)
-			break;
-	}
-
-	ret = 0;
+        afr_local_t     *local  = NULL;
+        afr_self_heal_t *sh  = NULL;
+        afr_self_heal_t *impunge_sh  = NULL;
+        int              ret = -1;
+        call_frame_t    *impunge_frame = NULL;
+        afr_local_t     *impunge_local = NULL;
+        int              active_src = 0;
+        int              op_errno = 0;
+        int              op_ret = -1;
+
+        local = frame->local;
+        sh = &local->self_heal;
+
+        active_src = sh->active_source;
+        sh->impunge_done = afr_sh_entry_impunge_entry_done;
+
+        if (can_skip_entry_self_heal (entry->d_name, &local->loc)) {
+                op_ret = 0;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_TRACE,
+                "inspecting existence of %s under %s",
+                entry->d_name, local->loc.path);
+
+        ret = afr_impunge_frame_create (frame, this, active_src,
+                                        &impunge_frame);
+        if (ret) {
+                op_errno = -ret;
+                goto out;
+        }
+
+        impunge_local = impunge_frame->local;
+        impunge_sh = &impunge_local->self_heal;
+        ret = afr_build_child_loc (this, &impunge_local->loc, &local->loc,
+                                   entry->d_name);
+        loc_copy (&impunge_sh->parent_loc, &local->loc);
+        if (ret != 0) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        afr_sh_common_lookup (impunge_frame, this, &impunge_local->loc,
+                              afr_sh_entry_common_lookup_done, NULL,
+                              AFR_LOOKUP_FAIL_CONFLICTS, NULL);
+
+        op_ret = 0;
 out:
-	if (ret == -1)
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
+        if (ret) {
+                if (impunge_frame)
+                        AFR_STACK_DESTROY (impunge_frame);
+                sh->impunge_done (frame, this, op_ret, op_errno);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_entry_impunge_readdir_cbk (call_frame_t *frame, void *cookie,
-				  xlator_t *this,
-				  int32_t op_ret, int32_t op_errno,
-				  gf_dirent_t *entries)
+                                  xlator_t *this,
+                                  int32_t op_ret, int32_t op_errno,
+                                  gf_dirent_t *entries, dict_t *xdata)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	gf_dirent_t     *entry = NULL;
-	off_t            last_offset = 0;
-	int              active_src = 0;
-	int              entry_count = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	active_src = sh->active_source;
-
-	if (op_ret <= 0) {
-		if (op_ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"readdir of %s on subvolume %s failed (%s)",
-				local->loc.path,
-				priv->children[active_src]->name,
-				strerror (op_errno));
-		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"readdir of %s on subvolume %s complete",
-				local->loc.path,
-				priv->children[active_src]->name);
-		}
-
-		afr_sh_entry_impunge_all (frame, this);
-		return 0;
-	}
-
-	list_for_each_entry (entry, &entries->list, list) {
-		last_offset = entry->d_off;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *local  = NULL;
+        afr_self_heal_t *sh  = NULL;
+        gf_dirent_t     *entry = NULL;
+        off_t            last_offset = 0;
+        int              active_src = 0;
+        int              entry_count = 0;
+
+        priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
+
+        active_src = sh->active_source;
+
+        if (op_ret <= 0) {
+                if (op_ret < 0) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "readdir of %s on subvolume %s failed (%s)",
+                                local->loc.path,
+                                priv->children[active_src]->name,
+                                strerror (op_errno));
+                        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                } else {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "readdir of %s on subvolume %s complete",
+                                local->loc.path,
+                                priv->children[active_src]->name);
+                }
+
+                afr_sh_entry_impunge_all (frame, this);
+                return 0;
+        }
+
+        list_for_each_entry (entry, &entries->list, list) {
+                last_offset = entry->d_off;
                 entry_count++;
-	}
+        }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"readdir'ed %d entries from %s",
-		entry_count, priv->children[active_src]->name);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "readdir'ed %d entries from %s",
+                entry_count, priv->children[active_src]->name);
 
-	sh->offset = last_offset;
-	local->call_count = entry_count;
+        sh->offset = last_offset;
+        local->call_count = entry_count;
 
-	list_for_each_entry (entry, &entries->list, list) {
+        list_for_each_entry (entry, &entries->list, list) {
                 afr_sh_entry_impunge_entry (frame, this, entry);
-	}
+        }
 
-	return 0;
+        return 0;
 }
-				  
+
 
 int
-afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this,
-			     int active_src)
+afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *local  = NULL;
+        afr_self_heal_t *sh  = NULL;
+        int32_t         active_src = 0;
 
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
+        priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
+        active_src = sh->active_source;
+        gf_log (this->name, GF_LOG_DEBUG, "%s: readdir from offset %zd",
+                local->loc.path, sh->offset);
 
-	STACK_WIND (frame, afr_sh_entry_impunge_readdir_cbk,
-		    priv->children[active_src],
-		    priv->children[active_src]->fops->readdirp,
-		    sh->healing_fd, sh->block_size, sh->offset);
+        STACK_WIND (frame, afr_sh_entry_impunge_readdir_cbk,
+                    priv->children[active_src],
+                    priv->children[active_src]->fops->readdirp,
+                    sh->healing_fd, sh->block_size, sh->offset, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              active_src = -1;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *local  = NULL;
+        afr_self_heal_t *sh  = NULL;
+        int              active_src = -1;
 
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
+        priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
 
-	sh->offset = 0;
+        sh->offset = 0;
 
-	active_src = next_active_source (frame, this, sh->active_source);
-	sh->active_source = active_src;
+        active_src = next_active_source (frame, this, sh->active_source);
+        sh->active_source = active_src;
 
-	if (sh->op_failed) {
-		afr_sh_entry_finish (frame, this);
-		return 0;
-	}
+        if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+                afr_sh_entry_finish (frame, this);
+                return 0;
+        }
 
-	if (active_src == -1) {
-		/* completed creating missing files on all subvolumes */
-		afr_sh_entry_expunge_all (frame, this);
-		return 0;
-	}
+        if (active_src == -1) {
+                /* completed creating missing files on all subvolumes */
+                afr_sh_entry_erase_pending (frame, this);
+                return 0;
+        }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"impunging entries of %s on %s to other sinks",
-		local->loc.path, priv->children[active_src]->name);
+        gf_log (this->name, GF_LOG_TRACE,
+                "impunging entries of %s on %s to other sinks",
+                local->loc.path, priv->children[active_src]->name);
 
-	afr_sh_entry_impunge_subvol (frame, this, active_src);
+        afr_sh_entry_impunge_subvol (frame, this);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_entry_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			  int32_t op_ret, int32_t op_errno, fd_t *fd)
+                          int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	/* TODO: some of the open's might fail.
-	   In that case, modify cleanup fn to send flush on those 
-	   fd's which are already open */
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"opendir of %s failed on child %s (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-			sh->op_failed = 1;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		if (sh->op_failed) {
-			afr_sh_entry_finish (frame, this);
-			return 0;
-		}
-		gf_log (this->name, GF_LOG_TRACE,
-			"fd for %s opened, commencing sync",
-			local->loc.path);
-
-		sh->active_source = -1;
-		afr_sh_entry_impunge_all (frame, this);
-	}
-
-	return 0;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              call_count = 0;
+        int              child_index = 0;
+
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
+
+        child_index = (long) cookie;
+
+        /* TODO: some of the open's might fail.
+           In that case, modify cleanup fn to send flush on those
+           fd's which are already open */
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "opendir of %s failed on child %s (%s)",
+                                local->loc.path,
+                                priv->children[child_index]->name,
+                                strerror (op_errno));
+                        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+                if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
+                        afr_sh_entry_finish (frame, this);
+                        return 0;
+                }
+                gf_log (this->name, GF_LOG_TRACE,
+                        "fd for %s opened, commencing sync",
+                        local->loc.path);
+
+                sh->active_source = -1;
+                afr_sh_entry_expunge_all (frame, this);
+        }
+
+        return 0;
 }
 
 
 int
 afr_sh_entry_open (call_frame_t *frame, xlator_t *this)
 {
-	int i = 0;				
-	int call_count = 0;		     
+        int i = 0;
+        int call_count = 0;
 
-	int source = -1;
-	int *sources = NULL;
+        int source = -1;
+        int *sources = NULL;
 
-	fd_t *fd = NULL;
+        fd_t *fd = NULL;
 
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	afr_self_heal_t *sh = NULL;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        afr_local_t *   local = NULL;
+        afr_private_t * priv  = NULL;
+        afr_self_heal_t *sh = NULL;
 
-	source  = local->self_heal.source;
-	sources = local->self_heal.sources;
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-	sh->block_size = 65536; //131072
-	sh->offset = 0;
+        source  = local->self_heal.source;
+        sources = local->self_heal.sources;
 
-	call_count = sh->active_sinks;
-	if (source != -1)
-		call_count++;
+        sh->block_size = priv->sh_readdir_size;
+        sh->offset = 0;
 
-	local->call_count = call_count;
+        call_count = sh->active_sinks;
+        if (source != -1)
+                call_count++;
 
-	fd = fd_create (local->loc.inode, frame->root->pid);
-	sh->healing_fd = fd;
+        local->call_count = call_count;
 
-	if (source != -1) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"opening directory %s on subvolume %s (source)",
-			local->loc.path, priv->children[source]->name);
+        fd = fd_create (local->loc.inode, frame->root->pid);
+        sh->healing_fd = fd;
 
-		/* open source */
-		STACK_WIND_COOKIE (frame, afr_sh_entry_opendir_cbk,
-				   (void *) (long) source,
-				   priv->children[source],
-				   priv->children[source]->fops->opendir,
-				   &local->loc, fd);
-		call_count--;
-	}
+        if (source != -1) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "opening directory %s on subvolume %s (source)",
+                        local->loc.path, priv->children[source]->name);
+
+                /* open source */
+                STACK_WIND_COOKIE (frame, afr_sh_entry_opendir_cbk,
+                                   (void *) (long) source,
+                                   priv->children[source],
+                                   priv->children[source]->fops->opendir,
+                                   &local->loc, fd, NULL);
+                call_count--;
+        }
 
-	/* open sinks */
-	for (i = 0; i < priv->child_count; i++) {
-		if (sources[i] || !local->child_up[i])
-			continue;
+        /* open sinks */
+        for (i = 0; i < priv->child_count; i++) {
+                if (sources[i] || !local->child_up[i])
+                        continue;
 
-		gf_log (this->name, GF_LOG_TRACE,
-			"opening directory %s on subvolume %s (sink)",
-			local->loc.path, priv->children[i]->name);
+                gf_log (this->name, GF_LOG_TRACE,
+                        "opening directory %s on subvolume %s (sink)",
+                        local->loc.path, priv->children[i]->name);
 
-		STACK_WIND_COOKIE (frame, afr_sh_entry_opendir_cbk,
-				   (void *) (long) i,
-				   priv->children[i], 
-				   priv->children[i]->fops->opendir,
-				   &local->loc, fd);
+                STACK_WIND_COOKIE (frame, afr_sh_entry_opendir_cbk,
+                                   (void *) (long) i,
+                                   priv->children[i],
+                                   priv->children[i]->fops->opendir,
+                                   &local->loc, fd, NULL);
 
-		if (!--call_count)
-			break;
-	}
+                if (!--call_count)
+                        break;
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_entry_sync_prepare (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              active_sinks = 0;
-	int              source = 0;
-	int              i = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	source = sh->source;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] == 0 && local->child_up[i] == 1) {
-			active_sinks++;
-			sh->success[i] = 1;
-		}
-	}
-	if (source != -1)
-		sh->success[source] = 1;
-
-	if (active_sinks == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"no active sinks for self-heal on dir %s",
-			local->loc.path);
-		afr_sh_entry_finish (frame, this);
-		return 0;
-	}
-	if (source == -1 && active_sinks < 2) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"cannot sync with 0 sources and 1 sink on dir %s",
-			local->loc.path);
-		afr_sh_entry_finish (frame, this);
-		return 0;
-	}
-	sh->active_sinks = active_sinks;
-
-	if (source != -1)
-		gf_log (this->name, GF_LOG_DEBUG,
-			"self-healing directory %s from subvolume %s to "
-                        "%d other",
-			local->loc.path, priv->children[source]->name,
-			active_sinks);
-	else
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no active sources for %s found. "
-			"merging all entries as a conservative decision",
-			local->loc.path);
-
-	afr_sh_entry_open (frame, this);
-
-	return 0;
-}
-
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              source = 0;
 
-int
-afr_sh_entry_fix (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              source = 0;
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-        int nsources = 0;
+        source = sh->source;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        afr_sh_mark_source_sinks (frame, this);
+        if (source != -1)
+                sh->success[source] = 1;
 
-        if (sh->forced_merge) {
-                sh->source = -1;
-                goto heal;
+        if (sh->active_sinks == 0) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "no active sinks for self-heal on dir %s",
+                        local->loc.path);
+                afr_sh_entry_finish (frame, this);
+                return 0;
         }
-
-	afr_sh_build_pending_matrix (priv, sh->pending_matrix, sh->xattr, 
-				     priv->child_count, AFR_ENTRY_TRANSACTION);
-
-	afr_sh_print_pending_matrix (sh->pending_matrix, this);
-
-	nsources = afr_sh_mark_sources (sh, priv->child_count,
-                                        AFR_SELF_HEAL_ENTRY);
-
-        if (nsources == 0) {
+        if (source == -1 && sh->active_sinks < 2) {
                 gf_log (this->name, GF_LOG_TRACE,
-                        "No self-heal needed for %s",
+                        "cannot sync with 0 sources and 1 sink on dir %s",
                         local->loc.path);
-
                 afr_sh_entry_finish (frame, this);
                 return 0;
         }
 
-	afr_sh_supress_errenous_children (sh->sources, sh->child_errno,
-					  priv->child_count);
-
-	source = afr_sh_select_source (sh->sources, priv->child_count);
-
-        sh->source = source;
+        if (source != -1)
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "self-healing directory %s from subvolume %s to "
+                        "%d other",
+                        local->loc.path, priv->children[source]->name,
+                        sh->active_sinks);
+        else
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no active sources for %s found. "
+                        "merging all entries as a conservative decision",
+                        local->loc.path);
 
-heal:
-	afr_sh_entry_sync_prepare (frame, this);
+        sh->actual_sh_started = _gf_true;
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_SYNC_BEGIN);
+        afr_sh_entry_open (frame, this);
 
-	return 0;
+        return 0;
 }
 
 
-
-int
-afr_sh_entry_lookup_cbk (call_frame_t *frame, void *cookie,
-			 xlator_t *this, int32_t op_ret, int32_t op_errno,
-                         inode_t *inode, struct iatt *buf, dict_t *xattr,
-                         struct iatt *postparent)
+void
+afr_sh_entry_fix (call_frame_t *frame, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
-
-	local = frame->local;
-	sh = &local->self_heal;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret != -1) {
-			sh->xattr[child_index] = dict_ref (xattr);
-			sh->buf[child_index] = *buf;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              source = 0;
+        int              nsources = 0;
+        int32_t          subvol_status = 0;
 
-	if (call_count == 0) {
-		afr_sh_entry_fix (frame, this);
-	}
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-	return 0;
-}
+        if (op_ret < 0) {
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                afr_sh_set_error (sh, op_errno);
+                afr_sh_entry_finish (frame, this);
+                goto out;
+        }
 
+        if (sh->forced_merge) {
+                sh->source = -1;
+                goto heal;
+        }
 
+        nsources = afr_build_sources (this, sh->xattr, sh->buf,
+                                      sh->pending_matrix, sh->sources,
+                                      sh->success_children,
+                                      AFR_ENTRY_TRANSACTION, &subvol_status,
+                                      _gf_true);
+        if ((subvol_status & ALL_FOOLS) ||
+            (subvol_status & SPLIT_BRAIN)) {
+                gf_log (this->name, GF_LOG_INFO, "%s: Performing conservative "
+                        "merge", local->loc.path);
+                source = -1;
+                memset (sh->sources, 0,
+                        sizeof (*sh->sources) * priv->child_count);
+        } else if (nsources == 0) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "No self-heal needed for %s",
+                        local->loc.path);
 
-int
-afr_sh_entry_lookup (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t    *  local = NULL;
-	afr_private_t  *  priv  = NULL;
-	dict_t         *xattr_req = NULL;
-	int ret = 0;
-	int call_count = 0;
-	int i = 0;
-
-	priv  = this->private;
-	local = frame->local;
-
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
-
-	local->call_count = call_count;
-
-	xattr_req = dict_new();
-	if (xattr_req) {
-                for (i = 0; i < priv->child_count; i++) {
-                        ret = dict_set_uint64 (xattr_req,
-                                               priv->pending_key[i],
-                                               3 * sizeof(int32_t));
-                        if (ret < 0)
-                                gf_log (this->name, GF_LOG_WARNING,
-                                        "Unable to set dict value.");
-                }
+                afr_sh_entry_finish (frame, this);
+                return;
+        } else {
+                source = afr_sh_select_source (sh->sources, priv->child_count);
         }
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame,
-					   afr_sh_entry_lookup_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->lookup,
-					   &local->loc, xattr_req);
-			if (!--call_count)
-				break;
-		}
-	}
+        sh->source = source;
 
-	if (xattr_req)
-		dict_unref (xattr_req);
+        afr_reset_children (sh->fresh_children, priv->child_count);
+        afr_get_fresh_children (sh->success_children, sh->sources,
+                                sh->fresh_children, priv->child_count);
+        if (sh->source >= 0)
+                afr_inode_set_read_ctx (this, sh->inode, sh->source,
+                                        sh->fresh_children);
 
-	return 0;
+heal:
+        afr_sh_entry_sync_prepare (frame, this);
+out:
+        return;
 }
 
 int
@@ -2389,68 +2317,54 @@ afr_sh_post_nonblocking_entry_cbk (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
+        afr_self_heal_t     *sh       = NULL;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
+        sh       = &local->self_heal;
 
         if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking entrylks failed.");
+                gf_log (this->name, GF_LOG_ERROR, "Non Blocking entrylks "
+                        "failed for %s.", local->loc.path);
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
                 afr_sh_entry_done (frame, this);
         } else {
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking entrylks done. Proceeding to FOP");
-                afr_sh_entry_lookup(frame, this);
+                gf_log (this->name, GF_LOG_DEBUG, "Non Blocking entrylks done "
+                        "for %s. Proceeding to FOP", local->loc.path);
+                afr_sh_common_lookup (frame, this, &local->loc,
+                                      afr_sh_entry_fix, NULL,
+                                      AFR_LOOKUP_FAIL_CONFLICTS |
+                                      AFR_LOOKUP_FAIL_MISSING_GFIDS,
+                                      NULL);
         }
 
         return 0;
 }
 
 int
-afr_sh_entry_lock (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
-        int_lock->selfheal_lk_type    = AFR_ENTRY_SELF_HEAL_LK;
-
-        afr_set_lock_number (frame, this);
-
-        int_lock->lk_basename = NULL;
-        int_lock->lk_loc      = &local->loc;
-        int_lock->lock_cbk    = afr_sh_post_nonblocking_entry_cbk;
-
-        afr_nonblocking_entrylk (frame, this);
-
-
-	return 0;
-}
-
-
-int
 afr_self_heal_entry (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t   *local = NULL;
-	afr_private_t   *priv = NULL;
+        afr_local_t     *local = NULL;
+        afr_private_t   *priv = NULL;
+        afr_self_heal_t *sh = NULL;
 
+        priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
 
-	priv = this->private;
-	local = frame->local;
+        sh->sh_type_in_action = AFR_SELF_HEAL_ENTRY;
 
-	if (local->self_heal.need_entry_self_heal && priv->entry_self_heal) {
-		afr_sh_entry_lock (frame, this);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"proceeding to completion on %s",
-			local->loc.path);
-		afr_sh_entry_done (frame, this);
-	}
+        if (local->self_heal.do_entry_self_heal && priv->entry_self_heal) {
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
+                afr_sh_entrylk (frame, this, &local->loc, NULL,
+                                afr_sh_post_nonblocking_entry_cbk);
+        } else {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "proceeding to completion on %s",
+                        local->loc.path);
+                afr_sh_entry_done (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
-
diff --git a/xlators/cluster/afr/src/afr-self-heal-metadata.c b/xlators/cluster/afr/src/afr-self-heal-metadata.c
index bd70608a2..fd5da6cfd 100644
--- a/xlators/cluster/afr/src/afr-self-heal-metadata.c
+++ b/xlators/cluster/afr/src/afr-self-heal-metadata.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <libgen.h>
@@ -52,73 +43,26 @@
 int
 afr_sh_metadata_done (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-//	memset (sh->child_errno, 0, sizeof (int) * priv->child_count);
-	memset (sh->buf, 0, sizeof (struct iatt) * priv->child_count);
-	memset (sh->success, 0, sizeof (int) * priv->child_count);
-
-/*         for (i = 0; i < priv->child_count; i++) { */
-/*                 sh->locked_nodes[i] = 1; */
-/*         } */
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i])
-			dict_unref (sh->xattr[i]);
-		sh->xattr[i] = NULL;
-	}
-
-	if (local->govinda_gOvinda) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"aborting selfheal of %s",
-			local->loc.path);
-		sh->completion_cbk (frame, this);
-	} else {
-		if (IA_ISREG (sh->type)) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"proceeding to data check on %s",
-				local->loc.path);
-			afr_self_heal_data (frame, this);
-			return 0;
-		}
-
-		if (IA_ISDIR (sh->type)) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"proceeding to entry check on %s",
-				local->loc.path);
-			afr_self_heal_entry (frame, this);
-			return 0;
-		}
-		gf_log (this->name, GF_LOG_DEBUG,
-			"completed self heal of %s",
-			local->loc.path);
-
-		sh->completion_cbk (frame, this);
-	}
-
-	return 0;
-}
-
-
-int
-afr_sh_metadata_unlck_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			   int32_t op_ret, int32_t op_errno)
-{
-	int               call_count = 0;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
 
-	call_count = afr_frame_return (frame);
+        local = frame->local;
+        sh = &local->self_heal;
 
-	if (call_count == 0)
-		afr_sh_metadata_done (frame, this);
+        afr_sh_reset (frame, this);
+        if (IA_ISDIR (sh->type)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "proceeding to entry check on %s",
+                        local->loc.path);
+                afr_self_heal_entry (frame, this);
+        } else {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "proceeding to data check on %s",
+                        local->loc.path);
+                afr_self_heal_data (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 int
@@ -141,530 +85,604 @@ afr_sh_metadata_finish (call_frame_t *frame, xlator_t *this)
 {
         afr_sh_inode_unlock (frame, this);
 
-	return 0;
+        return 0;
 }
 
-
 int
-afr_sh_metadata_erase_pending_cbk (call_frame_t *frame, void *cookie,
-				   xlator_t *this, int32_t op_ret,
-				   int32_t op_errno, dict_t *xattr)
+afr_sh_metadata_fail (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t     *local = NULL;
-	int             call_count = 0;
+        afr_local_t         *local    = NULL;
+        afr_self_heal_t     *sh       = NULL;
 
-	local = frame->local;
+        local    = frame->local;
+        sh       = &local->self_heal;
 
-	LOCK (&frame->lock);
-	{
-	}
-	UNLOCK (&frame->lock);
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+        afr_sh_metadata_finish (frame, this);
+        return 0;
+}
 
-	call_count = afr_frame_return (frame);
+int
+afr_sh_metadata_erase_pending_cbk (call_frame_t *frame, void *cookie,
+                                   xlator_t *this, int32_t op_ret,
+                                   int32_t op_errno, dict_t *xattr, dict_t *xdata)
+{
+        afr_local_t     *local     = NULL;
+        int             call_count = 0;
+        long            i          = 0;
+        afr_self_heal_t *sh        = NULL;
+        afr_private_t   *priv      = NULL;
+
+        local = frame->local;
+        priv  = this->private;
+        sh = &local->self_heal;
+        i = (long)cookie;
+
+        if ((!IA_ISREG (sh->buf[sh->source].ia_type)) &&
+            (!IA_ISDIR (sh->buf[sh->source].ia_type))) {
+                afr_children_add_child (sh->fresh_children, i,
+                                        priv->child_count);
+        }
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0)
-		afr_sh_metadata_finish (frame, this);
+        if (call_count == 0) {
+                if ((!IA_ISREG (sh->buf[sh->source].ia_type)) &&
+                    (!IA_ISDIR (sh->buf[sh->source].ia_type))) {
+                        afr_inode_set_read_ctx (this, sh->inode, sh->source,
+                                                sh->fresh_children);
+                }
+                afr_sh_metadata_finish (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
-
 int
 afr_sh_metadata_erase_pending (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              i = 0;
-	dict_t          **erase_xattr = NULL;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, 
-                                 sh->success, priv->child_count,
-                                 AFR_METADATA_TRANSACTION);
-
-	erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
-                                 gf_afr_mt_dict_t);
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i]) {
-			call_count++;
-
-			erase_xattr[i] = get_new_dict();
-			dict_ref (erase_xattr[i]);
-		}
-	}
-
-	afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
-			       priv->child_count, AFR_METADATA_TRANSACTION);
-
-	local->call_count = call_count;
-
-	if (call_count == 0) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"metadata of %s not healed on any subvolume",
-			local->loc.path);
-
-		afr_sh_metadata_finish (frame, this);
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (!erase_xattr[i])
-			continue;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"erasing pending flags from %s on %s",
-			local->loc.path, priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_metadata_erase_pending_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->xattrop,
-				   &local->loc,
-				   GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
-		if (!--call_count)
-			break;
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (erase_xattr[i]) {
-			dict_unref (erase_xattr[i]);
-		}
-	}
-	GF_FREE (erase_xattr);
-
-	return 0;
+         afr_sh_erase_pending (frame, this, AFR_METADATA_TRANSACTION,
+                               afr_sh_metadata_erase_pending_cbk,
+                               afr_sh_metadata_finish);
+         return 0;
 }
 
 
 int
 afr_sh_metadata_sync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			  int32_t op_ret, int32_t op_errno)
+                          int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              call_count = 0;
+        int              child_index = 0;
 
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-	child_index = (long) cookie;
+        child_index = (long) cookie;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"setting attributes failed for %s on %s (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "setting attributes failed for %s on %s (%s)",
+                                local->loc.path,
+                                priv->children[child_index]->name,
+                                strerror (op_errno));
 
-			sh->success[child_index] = 0;
-		}
-	}
-	UNLOCK (&frame->lock);
+                        sh->success[child_index] = 0;
+                }
+        }
+        UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
+        call_count = afr_frame_return (frame);
 
-	if (call_count == 0)
-		afr_sh_metadata_erase_pending (frame, this);
+        if (call_count == 0) {
+                if (local->xattr_req) {
+                        dict_unref (local->xattr_req);
+                        local->xattr_req = NULL;
+                }
+                afr_sh_metadata_erase_pending (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_metadata_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                              int32_t op_ret, int32_t op_errno,
-                             struct iatt *preop, struct iatt *postop)
+                             struct iatt *preop, struct iatt *postop, dict_t *xdata)
 {
-	afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno);
+        afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno, xdata);
 
-	return 0;
+        return 0;
 }
 
 
 int
 afr_sh_metadata_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			   int32_t op_ret, int32_t op_errno)
+                           int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno);
+        afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno, xdata);
 
-	return 0;
+        return 0;
 }
 
+int
+afr_sh_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno,
+                        dict_t *xdata)
+{
+        int            i     = 0;
+        afr_private_t *priv  = NULL;
+        afr_local_t   *local = NULL;
+
+        priv = this->private;
+        local = frame->local;
+
+        if (op_ret < 0) {
+                afr_sh_metadata_sync_cbk (frame, cookie,
+                                          this, -1, op_errno, xdata);
+                goto out;
+        }
+
+        i = (long) cookie;
+
+        STACK_WIND_COOKIE (frame, afr_sh_metadata_xattr_cbk,
+                           (void *) (long) i,
+                           priv->children[i],
+                           priv->children[i]->fops->setxattr,
+                           &local->loc, local->xattr_req, 0, NULL);
+
+ out:
+        return 0;
+}
+
+inline void
+afr_prune_special_keys (dict_t *xattr_dict)
+{
+        dict_del (xattr_dict, GF_SELINUX_XATTR_KEY);
+}
+
+inline void
+afr_prune_pending_keys (dict_t *xattr_dict, afr_private_t *priv)
+{
+        int i = 0;
+
+        for (; i < priv->child_count; i++) {
+                dict_del (xattr_dict, priv->pending_key[i]);
+        }
+}
+
+int
+afr_sh_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xattr,
+                     dict_t *xdata)
+{
+        int            i     = 0;
+        afr_private_t *priv  = NULL;
+        afr_local_t   *local = NULL;
+
+        priv = this->private;
+        local = frame->local;
+
+        if (op_ret < 0) {
+                afr_sh_metadata_sync_cbk (frame, cookie,
+                                          this, -1, op_errno, xdata);
+                goto out;
+        }
+
+        afr_prune_pending_keys (xattr, priv);
+
+        afr_prune_special_keys (xattr);
+
+        i = (long) cookie;
+
+        /* send removexattr in bulk via xdata */
+        STACK_WIND_COOKIE (frame, afr_sh_removexattr_cbk,
+                           cookie,
+                           priv->children[i],
+                           priv->children[i]->fops->removexattr,
+                           &local->loc, "", xattr);
+
+ out:
+        return 0;
+}
 
 int
 afr_sh_metadata_sync (call_frame_t *frame, xlator_t *this, dict_t *xattr)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              source = 0;
-	int              active_sinks = 0;
-	int              call_count = 0;
-	int              i = 0;
-
-        struct iatt      stbuf;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              source = 0;
+        int              active_sinks = 0;
+        int              call_count = 0;
+        int              i = 0;
+
+        struct iatt      stbuf = {0,};
         int32_t          valid = 0;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-	source = sh->source;
-	active_sinks = sh->active_sinks;
+        source = sh->source;
+        active_sinks = sh->active_sinks;
 
-	/*
-	 * 2 calls per sink - setattr, setxattr
-	 */
-	if (xattr)
-		call_count = active_sinks * 2;
-	else
-		call_count = active_sinks;
+        /*
+         * 2 calls per sink - setattr, setxattr
+         */
+        if (xattr) {
+                call_count = active_sinks * 2;
+                local->xattr_req = dict_ref (xattr);
+        } else
+                call_count = active_sinks;
 
-	local->call_count = call_count;
+        local->call_count = call_count;
 
-	stbuf.ia_atime = sh->buf[source].ia_atime;
-	stbuf.ia_atime_nsec = sh->buf[source].ia_atime_nsec;
-	stbuf.ia_mtime = sh->buf[source].ia_mtime;
-	stbuf.ia_mtime_nsec = sh->buf[source].ia_mtime_nsec;
+        stbuf.ia_atime = sh->buf[source].ia_atime;
+        stbuf.ia_atime_nsec = sh->buf[source].ia_atime_nsec;
+        stbuf.ia_mtime = sh->buf[source].ia_mtime;
+        stbuf.ia_mtime_nsec = sh->buf[source].ia_mtime_nsec;
 
         stbuf.ia_uid = sh->buf[source].ia_uid;
         stbuf.ia_gid = sh->buf[source].ia_gid;
 
         stbuf.ia_type = sh->buf[source].ia_type;
         stbuf.ia_prot = sh->buf[source].ia_prot;
-        
-        valid = GF_SET_ATTR_MODE  | 
+
+        valid = GF_SET_ATTR_MODE  |
                 GF_SET_ATTR_UID   | GF_SET_ATTR_GID |
                 GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (call_count == 0) {
-			break;
-		}
-		if (sh->sources[i] || !local->child_up[i])
-			continue;
-
-		gf_log (this->name, GF_LOG_DEBUG,
-			"self-healing metadata of %s from %s to %s",
-			local->loc.path, priv->children[source]->name,
-			priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_metadata_setattr_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->setattr,
-				   &local->loc, &stbuf, valid);
-
-		call_count--;
-
-		if (!xattr)
-			continue;
-
-		STACK_WIND_COOKIE (frame, afr_sh_metadata_xattr_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->setxattr,
-				   &local->loc, xattr, 0);
-		call_count--;
-	}
-
-	return 0;
+        for (i = 0; i < priv->child_count; i++) {
+                if (call_count == 0) {
+                        break;
+                }
+                if (sh->sources[i] || !local->child_up[i])
+                        continue;
+
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "self-healing metadata of %s from %s to %s",
+                        local->loc.path, priv->children[source]->name,
+                        priv->children[i]->name);
+
+                STACK_WIND_COOKIE (frame, afr_sh_metadata_setattr_cbk,
+                                   (void *) (long) i,
+                                   priv->children[i],
+                                   priv->children[i]->fops->setattr,
+                                   &local->loc, &stbuf, valid, NULL);
+
+                call_count--;
+
+                if (!xattr)
+                        continue;
+
+                STACK_WIND_COOKIE (frame, afr_sh_getxattr_cbk,
+                                   (void *) (long) i,
+                                   priv->children[i],
+                                   priv->children[i]->fops->getxattr,
+                                   &local->loc, NULL, NULL);
+                call_count--;
+        }
+
+        return 0;
 }
 
 
 int
-afr_sh_metadata_getxattr_cbk (call_frame_t *frame, void *cookie,
-			      xlator_t *this,
-			      int32_t op_ret, int32_t op_errno, dict_t *xattr)
+afr_sh_metadata_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                              int32_t op_ret, int32_t op_errno, dict_t *xattr,
+                              dict_t *xdata)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              source = 0;
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              source = 0;
+
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
 
-        int i;
+        source = sh->source;
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "getxattr of %s failed on subvolume %s (%s). proceeding without xattr",
+                        local->loc.path, priv->children[source]->name,
+                        strerror (op_errno));
+
+                afr_sh_metadata_sync (frame, this, NULL);
+        } else {
+                afr_prune_pending_keys (xattr, priv);
+                afr_sh_metadata_sync (frame, this, xattr);
+        }
 
-	source = sh->source;
+        return 0;
+}
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"getxattr of %s failed on subvolume %s (%s). proceeding without xattr",
-			local->loc.path, priv->children[source]->name,
-			strerror (op_errno));
+static void
+afr_set_metadata_sh_info_str (afr_local_t *local, afr_self_heal_t *sh,
+                              xlator_t *this)
+{
+        afr_private_t    *priv = NULL;
+        int              i = 0;
+        char             num[1024] = {0};
+        size_t           len = 0;
+        char             *string = NULL;
+        size_t           off = 0;
+        char             *source_child =  " from source %s to";
+        char             *format = " %s, ";
+        char             *string_msg = " metadata self heal";
+        char             *pending_matrix_str = NULL;
+        int              down_child_present = 0;
+        int              unknown_child_present = 0;
+        char             *down_subvol_1 = " down subvolume is ";
+        char             *unknown_subvol_1 = " unknown subvolume is";
+        char             *down_subvol_2 = " down subvolumes are ";
+        char             *unknown_subvol_2 = " unknown subvolumes are ";
+        int              down_count = 0;
+        int              unknown_count = 0;
+
+        priv = this->private;
+
+        pending_matrix_str = afr_get_pending_matrix_str (sh->pending_matrix,
+                                                         this);
+
+        if (!pending_matrix_str)
+                pending_matrix_str = "";
+
+        len += snprintf (num, sizeof (num), "%s", string_msg);
+
+        for (i = 0; i < priv->child_count; i++) {
+                if ((sh->source == i) && (local->child_up[i] == 1)) {
+                        len += snprintf (num, sizeof (num), source_child,
+                                         priv->children[i]->name);
+                } else if ((local->child_up[i] == 1) && (sh->sources[i] == 0)) {
+                        len += snprintf (num, sizeof (num), format,
+                                         priv->children[i]->name);
+                } else if (local->child_up[i] == 0) {
+                        len += snprintf (num, sizeof (num), format,
+                                         priv->children[i]->name);
+                        if (!down_child_present)
+                                down_child_present = 1;
+                        down_count++;
+                } else if (local->child_up[i] == -1) {
+                        len += snprintf (num, sizeof (num), format,
+                                         priv->children[i]->name);
+                        if (!unknown_child_present)
+                                unknown_child_present = 1;
+                        unknown_count++;
+                }
+        }
 
-		afr_sh_metadata_sync (frame, this, NULL);
-	} else {
-                for (i = 0; i < priv->child_count; i++) {
-                        dict_del (xattr, priv->pending_key[i]);
+        if (down_child_present) {
+                if (down_count > 1) {
+                        len += snprintf (num, sizeof (num), "%s",
+                                         down_subvol_2);
+                } else {
+                        len += snprintf (num, sizeof (num), "%s",
+                                         down_subvol_1);
                 }
+        }
+        if (unknown_child_present) {
+                if (unknown_count > 1) {
+                        len += snprintf (num, sizeof (num), "%s",
+                                         unknown_subvol_2);
+                } else {
+                        len += snprintf (num, sizeof (num), "%s",
+                                         unknown_subvol_1);
+                }
+        }
 
-		afr_sh_metadata_sync (frame, this, xattr);
-	}
+        len ++;
 
-	return 0;
-}
+        string = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
+        if (!string)
+                return;
 
+        off += snprintf (string + off, len - off, "%s", string_msg);
+        for (i=0; i < priv->child_count; i++) {
+                if ((sh->source == i) && (local->child_up[i] == 1))
+                        off += snprintf (string + off, len - off, source_child,
+                                         priv->children[i]->name);
+        }
 
-int
-afr_sh_metadata_sync_prepare (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              active_sinks = 0;
-	int              source = 0;
-	int              i = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	source = sh->source;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] == 0 && local->child_up[i] == 1) {
-			active_sinks++;
-			sh->success[i] = 1;
-		}
-	}
-	sh->success[source] = 1;
-
-	if (active_sinks == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no active sinks for performing self-heal on file %s",
-			local->loc.path);
-		afr_sh_metadata_finish (frame, this);
-		return 0;
-	}
-	sh->active_sinks = active_sinks;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"syncing metadata of %s from subvolume %s to %d active sinks",
-		local->loc.path, priv->children[source]->name, active_sinks);
-
-	STACK_WIND (frame, afr_sh_metadata_getxattr_cbk,
-		    priv->children[source],
-		    priv->children[source]->fops->getxattr,
-		    &local->loc, NULL);
-
-	return 0;
-}
+        for (i = 0; i < priv->child_count; i++) {
+                if ((local->child_up[i] == 1)&& (sh->sources[i] == 0))
+                        off += snprintf (string + off, len - off, format,
+                                         priv->children[i]->name);
+        }
 
+        if (down_child_present) {
+                if (down_count > 1) {
+                        off += snprintf (string + off, len - off, "%s",
+                                         down_subvol_2);
+                } else {
+                        off += snprintf (string + off, len - off, "%s",
+                                         down_subvol_1);
+                }
+        }
 
-int
-afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              nsources = 0;
-	int              source = 0;
-	int              i = 0;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i] == 0)
+                        off += snprintf (string + off, len - off, format,
+                                         priv->children[i]->name);
+        }
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+        if (unknown_child_present) {
+                if (unknown_count > 1) {
+                        off += snprintf (string + off, len - off, "%s",
+                                 unknown_subvol_2);
+                } else {
+                        off += snprintf (string + off, len - off, "%s",
+                                         unknown_subvol_1);
+                }
+        }
 
-	afr_sh_build_pending_matrix (priv, sh->pending_matrix, sh->xattr, 
-				     priv->child_count, 
-                                     AFR_METADATA_TRANSACTION);
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i] == -1)
+                        off += snprintf (string + off, len - off, format,
+                                         priv->children[i]->name);
+        }
 
-	afr_sh_print_pending_matrix (sh->pending_matrix, this);
+        gf_asprintf (&sh->metadata_sh_info, "%s metadata %s,", string,
+                     pending_matrix_str);
 
-	nsources = afr_sh_mark_sources (sh, priv->child_count,
-                                        AFR_SELF_HEAL_METADATA);
+        if (pending_matrix_str && strcmp (pending_matrix_str, ""))
+                GF_FREE (pending_matrix_str);
 
-	afr_sh_supress_errenous_children (sh->sources, sh->child_errno,
-					  priv->child_count);
+        if (string && strcmp (string, ""))
+                GF_FREE (string);
+}
 
-        if (nsources == 0) {
-                gf_log (this->name, GF_LOG_TRACE,
-                        "No self-heal needed for %s",
-                        local->loc.path);
+int
+afr_sh_metadata_sync_prepare (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              source = 0;
+
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
+
+        source = sh->source;
 
+        afr_sh_mark_source_sinks (frame, this);
+        if (sh->active_sinks == 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no active sinks for performing self-heal on file %s",
+                        local->loc.path);
                 afr_sh_metadata_finish (frame, this);
                 return 0;
         }
 
-	if ((nsources == -1)
-	    && (priv->favorite_child != -1)
-	    && (sh->child_errno[priv->favorite_child] == 0)) {
+        gf_log (this->name, GF_LOG_TRACE,
+                "syncing metadata of %s from subvolume %s to %d active sinks",
+                local->loc.path, priv->children[source]->name,
+                sh->active_sinks);
 
-		gf_log (this->name, GF_LOG_WARNING,
-			"Picking favorite child %s as authentic source to resolve conflicting metadata of %s",
-			priv->children[priv->favorite_child]->name,
-			local->loc.path);
+        sh->actual_sh_started = _gf_true;
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_SYNC_BEGIN);
+        afr_set_metadata_sh_info_str (local, sh, this);
+        STACK_WIND (frame, afr_sh_metadata_getxattr_cbk,
+                    priv->children[source],
+                    priv->children[source]->fops->getxattr,
+                    &local->loc, NULL, NULL);
 
-		sh->sources[priv->favorite_child] = 1;
+        return 0;
+}
 
-		nsources = afr_sh_source_count (sh->sources,
-						priv->child_count);
-	}
 
-	if (nsources == -1) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Unable to self-heal permissions/ownership of '%s' "
-                        "(possible split-brain). Please fix the file on "
-                        "all backend volumes", local->loc.path);
+void
+afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno)
+{
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = NULL;
+        afr_private_t   *priv = NULL;
+        int              nsources = 0;
+        int              source = 0;
+        int              i = 0;
+
+        local = frame->local;
+        sh = &local->self_heal;
+        priv = this->private;
+
+        if (op_ret < 0) {
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+                afr_sh_set_error (sh, op_errno);
+                afr_sh_metadata_finish (frame, this);
+                goto out;
+        }
+        nsources = afr_build_sources (this, sh->xattr, sh->buf,
+                                      sh->pending_matrix, sh->sources,
+                                      sh->success_children,
+                                      AFR_METADATA_TRANSACTION, NULL, _gf_false);
+        if ((nsources == -1)
+            && (priv->favorite_child != -1)
+            && (sh->child_errno[priv->favorite_child] == 0)) {
+
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Picking favorite child %s as authentic source to resolve conflicting metadata of %s",
+                        priv->children[priv->favorite_child]->name,
+                        local->loc.path);
 
-		local->govinda_gOvinda = 1;
+                sh->sources[priv->favorite_child] = 1;
 
-		afr_sh_metadata_finish (frame, this);
-		return 0;
-	}
+                nsources = afr_sh_source_count (sh->sources,
+                                                priv->child_count);
+        }
 
-	source = afr_sh_select_source (sh->sources, priv->child_count);
+        if (nsources == -1) {
+                afr_sh_print_split_brain_log (sh->pending_matrix, this,
+                                              local->loc.path);
+                afr_set_split_brain (this, sh->inode, SPB, DONT_KNOW);
+                afr_sh_metadata_fail (frame, this);
+                goto out;
+        }
 
-        if (source == -1) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "No active sources found.");
+        afr_set_split_brain (this, sh->inode, NO_SPB, DONT_KNOW);
+        if (nsources == 0) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "No self-heal needed for %s",
+                        local->loc.path);
 
                 afr_sh_metadata_finish (frame, this);
-                return 0;
+                goto out;
         }
 
-	sh->source = source;
-
-	/* detect changes not visible through pending flags -- JIC */
-	for (i = 0; i < priv->child_count; i++) {
-		if (i == source || sh->child_errno[i])
-			continue;
+        source = afr_sh_select_source (sh->sources, priv->child_count);
 
-		if (PERMISSION_DIFFERS (&sh->buf[i], &sh->buf[source]))
-			sh->sources[i] = 0;
+        if (source == -1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "No active sources found.");
 
-		if (OWNERSHIP_DIFFERS (&sh->buf[i], &sh->buf[source]))
-			sh->sources[i] = 0;
-	}
+                afr_sh_metadata_finish (frame, this);
+                goto out;
+        }
 
-	afr_sh_metadata_sync_prepare (frame, this);
+        sh->source = source;
 
-	return 0;
-}
+        /* detect changes not visible through pending flags -- JIC */
+        for (i = 0; i < priv->child_count; i++) {
+                if (i == source || sh->child_errno[i])
+                        continue;
 
+                if (PERMISSION_DIFFERS (&sh->buf[i], &sh->buf[source]))
+                        sh->sources[i] = 0;
 
-int
-afr_sh_metadata_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			    int32_t op_ret, int32_t op_errno,
-                            inode_t *inode, struct iatt *buf, dict_t *xattr,
-                            struct iatt *postparent)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"path %s on subvolume %s is of mode 0%o",
-				local->loc.path,
-				priv->children[child_index]->name,
-				buf->ia_type);
-
-			sh->buf[child_index] = *buf;
-			if (xattr)
-				sh->xattr[child_index] = dict_ref (xattr);
-		} else {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"path %s on subvolume %s => -1 (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-
-			sh->child_errno[child_index] = op_errno;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_metadata_fix (frame, this);
-
-	return 0;
-}
-
+                if (OWNERSHIP_DIFFERS (&sh->buf[i], &sh->buf[source]))
+                        sh->sources[i] = 0;
+        }
 
-int
-afr_sh_metadata_lookup (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
-	int              call_count = 0;
-	dict_t          *xattr_req = NULL;
-	int              ret = 0;
-
-	local = frame->local;
-	priv = this->private;
-
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
-	local->call_count = call_count;
-
-	xattr_req = dict_new();
-
-	if (xattr_req) {
-                for (i = 0; i < priv->child_count; i++) {
-                        ret = dict_set_uint64 (xattr_req,
-                                               priv->pending_key[i],
-                                               3 * sizeof(int32_t));
-                        if (ret < 0)
-                                gf_log (this->name, GF_LOG_WARNING,
-                                        "Unable to set dict value.");
-                }
+        if ((!IA_ISREG (sh->buf[source].ia_type)) &&
+            (!IA_ISDIR (sh->buf[source].ia_type))) {
+                afr_reset_children (sh->fresh_children, priv->child_count);
+                afr_get_fresh_children (sh->success_children, sh->sources,
+                                        sh->fresh_children, priv->child_count);
+                afr_inode_set_read_ctx (this, sh->inode, sh->source,
+                                        sh->fresh_children);
         }
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"looking up %s on %s",
-				local->loc.path, priv->children[i]->name);
-
-			STACK_WIND_COOKIE (frame, afr_sh_metadata_lookup_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->lookup,
-					   &local->loc, xattr_req);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	if (xattr_req)
-		dict_unref (xattr_req);
-
-	return 0;
+        if (sh->do_metadata_self_heal && priv->metadata_self_heal)
+                afr_sh_metadata_sync_prepare (frame, this);
+        else
+                afr_sh_metadata_finish (frame, this);
+out:
+        return;
 }
 
 int
-afr_sh_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
+afr_sh_metadata_post_nonblocking_inodelk_cbk (call_frame_t *frame,
+                                              xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
@@ -673,14 +691,21 @@ afr_sh_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
 
         if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking inodelks failed.");
+                gf_log (this->name, GF_LOG_DEBUG, "Non Blocking metadata "
+                        "inodelks failed for %s.", local->loc.path);
+                gf_log (this->name, GF_LOG_DEBUG, "Metadata self-heal "
+                        "failed for %s.", local->loc.path);
                 afr_sh_metadata_done (frame, this);
         } else {
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking inodelks done. Proceeding to FOP");
-                afr_sh_metadata_lookup (frame, this);
+                gf_log (this->name, GF_LOG_DEBUG, "Non Blocking metadata "
+                        "inodelks done for %s. Proceeding to FOP",
+                        local->loc.path);
+                afr_sh_common_lookup (frame, this, &local->loc,
+                                      afr_sh_metadata_fix, NULL,
+                                      AFR_LOOKUP_FAIL_CONFLICTS |
+                                      AFR_LOOKUP_FAIL_MISSING_GFIDS,
+                                      NULL);
         }
 
         return 0;
@@ -690,42 +715,56 @@ int
 afr_sh_metadata_lock (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
         afr_local_t         *local    = NULL;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
 
+        int_lock->domain           = this->name;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
         int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
         int_lock->selfheal_lk_type    = AFR_METADATA_SELF_HEAL_LK;
 
         afr_set_lock_number (frame, this);
 
-        int_lock->lk_flock.l_start = 0;
-        int_lock->lk_flock.l_len   = 0;
-        int_lock->lk_flock.l_type  = F_WRLCK;
-        int_lock->lock_cbk         = afr_sh_post_nonblocking_inodelk_cbk;
+        inodelk->flock.l_start = LLONG_MAX - 1;
+        inodelk->flock.l_len   = 0;
+        inodelk->flock.l_type  = F_WRLCK;
+        int_lock->lock_cbk         = afr_sh_metadata_post_nonblocking_inodelk_cbk;
 
         afr_nonblocking_inodelk (frame, this);
 
-	return 0;
+        return 0;
 }
 
+gf_boolean_t
+afr_can_start_metadata_self_heal (afr_self_heal_t *sh, afr_private_t *priv)
+{
+        if (sh->force_confirm_spb)
+                return _gf_true;
+        if (sh->do_metadata_self_heal && priv->metadata_self_heal)
+                return _gf_true;
+        return _gf_false;
+}
 
 int
 afr_self_heal_metadata (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t   *local = NULL;
-	afr_private_t *priv = this->private;
+        afr_local_t   *local = NULL;
+        afr_private_t *priv = this->private;
+        afr_self_heal_t *sh = &local->self_heal;
 
+        local = frame->local;
+        sh = &local->self_heal;
+        sh->sh_type_in_action = AFR_SELF_HEAL_METADATA;
 
-	local = frame->local;
-
-	if (local->self_heal.need_metadata_self_heal && priv->metadata_self_heal) {
-		afr_sh_metadata_lock (frame, this);
-	} else {
-		afr_sh_metadata_done (frame, this);
-	}
+        if (afr_can_start_metadata_self_heal (sh, priv)) {
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_STARTED);
+                afr_sh_metadata_lock (frame, this);
+        } else {
+                afr_sh_metadata_done (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
-
diff --git a/xlators/cluster/afr/src/afr-self-heal.h b/xlators/cluster/afr/src/afr-self-heal.h
index c43473a43..7c9bc8111 100644
--- a/xlators/cluster/afr/src/afr-self-heal.h
+++ b/xlators/cluster/afr/src/afr-self-heal.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __AFR_SELF_HEAL_H__
@@ -30,13 +21,6 @@
 #define SIZE_GREATER(buf1,buf2) ((buf1)->ia_size > (buf2)->ia_size)
 
 int
-afr_sh_has_metadata_pending (dict_t *xattr, int child_count, xlator_t *this);
-int
-afr_sh_has_entry_pending (dict_t *xattr, int child_count, xlator_t *this);
-int
-afr_sh_has_data_pending (dict_t *xattr, int child_count, xlator_t *this);
-
-int
 afr_self_heal_entry (call_frame_t *frame, xlator_t *this);
 
 int
@@ -49,6 +33,11 @@ int
 afr_self_heal_get_source (xlator_t *this, afr_local_t *local, dict_t **xattr);
 
 int
-afr_self_heal (call_frame_t *frame, xlator_t *this);
+afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode);
 
+int
+afr_lookup_select_read_child_by_txn_type (xlator_t *this, afr_local_t *local,
+                                          dict_t **xattr,
+                                          afr_transaction_type txn_type,
+                                          uuid_t gfid);
 #endif /* __AFR_SELF_HEAL_H__ */
diff --git a/xlators/cluster/afr/src/afr-self-heald.c b/xlators/cluster/afr/src/afr-self-heald.c
new file mode 100644
index 000000000..1b48a1bca
--- /dev/null
+++ b/xlators/cluster/afr/src/afr-self-heald.c
@@ -0,0 +1,1787 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+#include "afr.h"
+#include "syncop.h"
+#include "afr-self-heald.h"
+#include "afr-self-heal-common.h"
+#include "protocol-common.h"
+#include "event-history.h"
+
+typedef enum {
+        STOP_CRAWL_ON_SINGLE_SUBVOL = 1
+} afr_crawl_flags_t;
+
+typedef enum {
+        HEAL = 1,
+        INFO,
+        STATISTICS_TO_BE_HEALED,
+} shd_crawl_op;
+
+typedef struct shd_dump {
+        dict_t   *dict;
+        xlator_t *this;
+        int      child;
+} shd_dump_t;
+
+typedef struct shd_event_ {
+        int     child;
+        char    *path;
+} shd_event_t;
+
+typedef struct shd_pos_ {
+        int     child;
+        xlator_t *this;
+        afr_child_pos_t pos;
+} shd_pos_t;
+
+typedef int
+(*afr_crawl_done_cbk_t)  (int ret, call_frame_t *sync_frame, void *crawl_data);
+
+void
+afr_start_crawl (xlator_t *this, int idx, afr_crawl_type_t crawl,
+                 process_entry_cbk_t process_entry, void *op_data,
+                 gf_boolean_t exclusive, int crawl_flags,
+                 afr_crawl_done_cbk_t crawl_done);
+
+static int
+_crawl_directory (fd_t *fd, loc_t *loc, afr_crawl_data_t *crawl_data);
+
+/* For calling straight through (e.g. already in a synctask). */
+int
+afr_find_child_position (xlator_t *this, int child, afr_child_pos_t *pos);
+
+/* For deferring through a new synctask. */
+int
+afr_syncop_find_child_position (void *data);
+
+static int
+_loc_assign_gfid_path (loc_t *loc)
+{
+        int  ret = -1;
+        char gfid_path[64] = {0};
+
+        if (loc->inode && !uuid_is_null (loc->inode->gfid)) {
+                ret = inode_path (loc->inode, NULL, (char**)&loc->path);
+        } else if (!uuid_is_null (loc->gfid)) {
+                snprintf (gfid_path, sizeof (gfid_path), "<gfid:%s>",
+                          uuid_utoa (loc->gfid));
+                loc->path = gf_strdup (gfid_path);
+                if (loc->path)
+                        ret = 0;
+        }
+        return ret;
+}
+
+void
+_destroy_crawl_event_data (void *data)
+{
+        shd_crawl_event_t        *crawl_event = NULL;
+
+        if (!data)
+                goto out;
+
+        crawl_event = (shd_crawl_event_t *)data;
+        GF_FREE (crawl_event->start_time_str);
+        GF_FREE (crawl_event->end_time_str);
+
+out:
+        return;
+}
+
+void
+_destroy_shd_event_data (void *data)
+{
+        shd_event_t             *event = NULL;
+        if (!data)
+                goto out;
+        event = (shd_event_t*)data;
+        GF_FREE (event->path);
+out:
+        return;
+}
+void
+shd_cleanup_event (void *event)
+{
+        shd_event_t *shd_event = event;
+
+        if (!shd_event)
+                goto out;
+        GF_FREE (shd_event->path);
+        GF_FREE (shd_event);
+out:
+        return;
+}
+
+int
+afr_get_local_child (afr_self_heald_t *shd, unsigned int child_count)
+{
+        int i = 0;
+        int ret = -1;
+        for (i = 0; i < child_count; i++) {
+                if (shd->pos[i] == AFR_POS_LOCAL) {
+                        ret = i;
+                        break;
+                }
+        }
+        return ret;
+}
+
+static int
+_build_index_loc (xlator_t *this, loc_t *loc, char *name, loc_t *parent)
+{
+        int             ret = 0;
+
+        uuid_copy (loc->pargfid, parent->inode->gfid);
+        loc->path = "";
+        loc->name = name;
+        loc->parent = inode_ref (parent->inode);
+        if (!loc->parent) {
+                loc->path = NULL;
+                loc_wipe (loc);
+                ret = -1;
+        }
+        return ret;
+}
+
+int
+_add_crawl_stats_to_dict (xlator_t *this, dict_t *output, int child,
+                          shd_crawl_event_t *shd_event, struct timeval *tv)
+{
+        int             ret = 0;
+        uint64_t        count = 0;
+        char            key[256] = {0};
+        int             xl_id = 0;
+        uint64_t        healed_count = 0;
+        uint64_t        split_brain_count = 0;
+        uint64_t        heal_failed_count = 0;
+        char            *start_time_str = NULL;
+        char            *end_time_str = NULL;
+        char            *crawl_type = NULL;
+        int             progress = -1;
+
+        healed_count = shd_event->healed_count;
+        split_brain_count = shd_event->split_brain_count;
+        heal_failed_count = shd_event->heal_failed_count;
+        start_time_str = shd_event->start_time_str;
+        end_time_str = shd_event->end_time_str;
+        crawl_type = shd_event->crawl_type;
+
+        if (!start_time_str) {
+                ret = -1;
+                goto out;
+        }
+
+
+        ret = dict_get_int32 (output, this->name, &xl_id);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "xl does not have id");
+                goto out;
+        }
+
+        snprintf (key, sizeof (key), "statistics-%d-%d-count", xl_id, child);
+        ret = dict_get_uint64 (output, key, &count);
+
+        snprintf (key, sizeof (key), "statistics_healed_cnt-%d-%d-%"PRIu64,
+                  xl_id, child, count);
+        ret = dict_set_uint64(output, key, healed_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+                        "healed_count to outout");
+                goto out;
+         }
+        snprintf (key, sizeof (key), "statistics_sb_cnt-%d-%d-%"PRIu64,
+                  xl_id, child, count);
+        ret = dict_set_uint64 (output, key, split_brain_count);
+         if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+                        "split_brain_count to outout");
+                goto out;
+        }
+        snprintf (key, sizeof (key), "statistics_crawl_type-%d-%d-%"PRIu64,
+                  xl_id, child, count);
+        ret = dict_set_dynstr (output, key, gf_strdup (crawl_type));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+                        "crawl_type to output");
+                goto out;
+        }
+        snprintf (key, sizeof (key), "statistics_heal_failed_cnt-%d-%d-%"PRIu64,
+                  xl_id, child, count);
+        ret = dict_set_uint64 (output, key, heal_failed_count);
+         if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+                        "healed_failed_count to outout");
+                goto out;
+        }
+        snprintf (key, sizeof (key), "statistics_strt_time-%d-%d-%"PRIu64,
+                  xl_id, child, count);
+        ret = dict_set_dynstr (output, key, gf_strdup(start_time_str));
+
+         if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+                        "crawl_start_time to outout");
+                goto out;
+        }
+
+        snprintf (key, sizeof (key), "statistics_end_time-%d-%d-%"PRIu64,
+                  xl_id, child, count);
+
+        if (!end_time_str)
+                end_time_str = "Could not determine the end time";
+        ret = dict_set_dynstr (output, key, gf_strdup(end_time_str));
+         if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+                        "crawl_end_time to outout");
+                goto out;
+        }
+        snprintf (key, sizeof (key), "statistics_inprogress-%d-%d-%"PRIu64,
+                  xl_id, child, count);
+
+        if (shd_event->crawl_inprogress == _gf_true)
+                progress = 1;
+        else
+                progress = 0;
+
+        ret = dict_set_int32 (output, key, progress);
+         if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not add statistics_"
+                        "inprogress to outout");
+                goto out;
+        }
+
+         snprintf (key, sizeof (key), "statistics-%d-%d-count",xl_id, child);
+         ret = dict_set_uint64 (output, key, count + 1);
+         if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not increment the "
+                        "counter.");
+                goto out;
+         }
+out:
+        return ret;
+}
+
+int
+_add_path_to_dict (xlator_t *this, dict_t *output, int child, char *path,
+                   struct timeval *tv, gf_boolean_t dyn)
+{
+        //subkey not used for now
+        int             ret = -1;
+        uint64_t        count = 0;
+        char            key[256] = {0};
+        int             xl_id = 0;
+
+        ret = dict_get_int32 (output, this->name, &xl_id);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "xl does not have id");
+                goto out;
+        }
+
+        snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
+        ret = dict_get_uint64 (output, key, &count);
+
+        snprintf (key, sizeof (key), "%d-%d-%"PRIu64, xl_id, child, count);
+        if (dyn)
+                ret = dict_set_dynstr (output, key, path);
+        else
+                ret = dict_set_str (output, key, path);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s: Could not add to output",
+                        path);
+                goto out;
+        }
+
+        if (!tv)
+                goto inc_count;
+        snprintf (key, sizeof (key), "%d-%d-%"PRIu64"-time", xl_id,
+                  child, count);
+        ret = dict_set_uint32 (output, key, tv->tv_sec);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s: Could not set time",
+                        path);
+                goto out;
+        }
+
+inc_count:
+        snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
+        ret = dict_set_uint64 (output, key, count + 1);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not increment count");
+                goto out;
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+_get_path_from_gfid_loc (xlator_t *this, xlator_t *readdir_xl, loc_t *child,
+                         char **fpath, gf_boolean_t *missing)
+{
+        dict_t          *xattr = NULL;
+        char            *path = NULL;
+        int             ret = -1;
+
+        ret = syncop_getxattr (readdir_xl, child, &xattr, GFID_TO_PATH_KEY);
+        if (ret < 0) {
+                if ((errno == ENOENT) && missing)
+                        *missing = _gf_true;
+                goto out;
+        }
+        ret = dict_get_str (xattr, GFID_TO_PATH_KEY, &path);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get path for "
+                        "gfid %s", uuid_utoa (child->gfid));
+                goto out;
+        }
+        path = gf_strdup (path);
+        if (!path) {
+                ret = -1;
+                goto out;
+        }
+        ret = 0;
+out:
+        if (!ret)
+                *fpath = path;
+        if (xattr)
+                dict_unref (xattr);
+        return ret;
+}
+
+int
+_add_event_to_dict (circular_buffer_t *cb, void *data)
+{
+        int               ret = 0;
+        shd_dump_t        *dump_data = NULL;
+        shd_event_t       *shd_event = NULL;
+
+        dump_data = data;
+        shd_event = cb->data;
+        if (shd_event->child != dump_data->child)
+                goto out;
+        ret = _add_path_to_dict (dump_data->this, dump_data->dict,
+                                 dump_data->child, shd_event->path, &cb->tv,
+                                 _gf_false);
+out:
+        return ret;
+}
+
+int
+_add_crawl_event_statistics_to_dict (circular_buffer_t *cb, void *data)
+{
+        int               ret = 0;
+        shd_dump_t        *dump_data = NULL;
+        shd_crawl_event_t *shd_event = NULL;
+
+        dump_data = data;
+        shd_event = cb->data;
+        ret = _add_crawl_stats_to_dict (dump_data->this, dump_data->dict,
+                                        dump_data->child, shd_event, &cb->tv);
+        return ret;
+}
+
+int
+_add_eh_to_dict (xlator_t *this, eh_t *eh, dict_t *dict, int child)
+{
+        shd_dump_t dump_data = {0};
+
+        dump_data.this = this;
+        dump_data.dict = dict;
+        dump_data.child = child;
+        eh_dump (eh, &dump_data, _add_event_to_dict);
+        return 0;
+}
+
+
+int
+_add_statistics_to_dict (xlator_t *this, dict_t *dict, int child)
+{
+        shd_dump_t              dump_data = {0};
+        afr_private_t           *priv  = NULL;
+        afr_self_heald_t        *shd = NULL;
+
+        priv = this->private;
+        shd = &priv->shd;
+
+        dump_data.this = this;
+        dump_data.dict = dict;
+        dump_data.child = child;
+        eh_dump (shd->statistics[child], &dump_data,
+                 _add_crawl_event_statistics_to_dict);
+        return 0;
+
+}
+
+void
+_remove_stale_index (xlator_t *this, xlator_t *readdir_xl,
+                     loc_t *parent, char *fname)
+{
+        int              ret = 0;
+        loc_t            index_loc = {0};
+
+        ret = _build_index_loc (this, &index_loc, fname, parent);
+        if (ret)
+                goto out;
+        gf_log (this->name, GF_LOG_DEBUG, "Removing stale index "
+                "for %s on %s", index_loc.name, readdir_xl->name);
+        ret = syncop_unlink (readdir_xl, &index_loc);
+        if(ret && (errno != ENOENT)) {
+                gf_log(this->name, GF_LOG_ERROR, "%s: Failed to remove index "
+                       "on %s - %s",index_loc.name, readdir_xl->name,
+                       strerror (errno));
+        }
+        index_loc.path = NULL;
+        loc_wipe (&index_loc);
+out:
+        return;
+}
+
+int
+_count_hard_links_under_base_indices_dir (xlator_t *this,
+                                           afr_crawl_data_t *crawl_data,
+                                           gf_dirent_t *entry, loc_t *childloc,
+                                           loc_t *parentloc, struct iatt *iattr)
+{
+        xlator_t                *readdir_xl = crawl_data->readdir_xl;
+        struct iatt             parent = {0};
+        int                     ret = 0;
+        dict_t                  *output = NULL;
+        int                     xl_id =  0;
+        char                    key[256] = {0};
+        int                     child  = -1;
+        uint64_t                     hardlinks = 0;
+
+        output = crawl_data->op_data;
+        child = crawl_data->child;
+
+        ret = syncop_lookup (readdir_xl, childloc, NULL, iattr, NULL, &parent);
+        if (ret)
+                goto out;
+
+        ret = dict_get_int32 (output, this->name, &xl_id);
+        if (ret)
+                goto out;
+
+        snprintf (key, sizeof (key), "%d-%d-hardlinks", xl_id, child);
+        ret =  dict_get_uint64 (output, key, &hardlinks);
+
+        /*Removing the count of base_entry under indices/base_indicies and
+         * entry under indices/xattrop */
+        hardlinks = hardlinks + iattr->ia_nlink - 2;
+        ret = dict_set_uint64 (output, key, hardlinks);
+        if (ret)
+                goto out;
+
+out:
+        return ret;
+}
+
+int
+_add_summary_to_dict (xlator_t *this, afr_crawl_data_t *crawl_data,
+                      gf_dirent_t *entry,
+                      loc_t *childloc, loc_t *parentloc, struct iatt *iattr)
+{
+        dict_t          *output = NULL;
+        xlator_t        *readdir_xl = NULL;
+        int             ret = -1;
+        char            *path = NULL;
+        gf_boolean_t    missing = _gf_false;
+        char            gfid_str[64] = {0};
+
+        if (uuid_is_null (childloc->gfid))
+                goto out;
+
+        output = crawl_data->op_data;
+        readdir_xl = crawl_data->readdir_xl;
+
+        ret = _get_path_from_gfid_loc (this, readdir_xl, childloc, &path,
+                                       &missing);
+        if (ret == 0) {
+                ret = _add_path_to_dict (this, output, crawl_data->child, path,
+                                         NULL, _gf_true);
+        } else if (missing) {
+                _remove_stale_index (this, readdir_xl, parentloc,
+                                     uuid_utoa_r (childloc->gfid, gfid_str));
+        }
+
+out:
+        if (ret && path)
+                GF_FREE (path);
+        return ret;
+}
+
+void
+_crawl_post_sh_action (xlator_t *this, loc_t *parent, loc_t *child,
+                       int32_t op_ret, int32_t op_errno, dict_t *xattr_rsp,
+                       afr_crawl_data_t *crawl_data)
+{
+        int                ret = 0;
+        afr_private_t      *priv = NULL;
+        afr_self_heald_t   *shd = NULL;
+        eh_t               *eh = NULL;
+        char               *path = NULL;
+        char               gfid_str[64] = {0};
+        shd_event_t        *event = NULL;
+        int32_t            sh_failed = 0;
+        gf_boolean_t       split_brain = 0;
+        int32_t            actual_sh_done = 0;
+        shd_crawl_event_t  **shd_crawl_event = NULL;
+
+        priv = this->private;
+        shd  = &priv->shd;
+        if (crawl_data->crawl == INDEX) {
+                if ((op_ret < 0) && (op_errno == ENOENT)) {
+                        _remove_stale_index (this, crawl_data->readdir_xl,
+                                             parent, uuid_utoa_r (child->gfid,
+                                                                  gfid_str));
+                        goto out;
+                }
+                ret = _get_path_from_gfid_loc (this, crawl_data->readdir_xl,
+                                               child, &path, NULL);
+                if (ret)
+                        goto out;
+        } else {
+                path = gf_strdup (child->path);
+                if (!path) {
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        if (xattr_rsp) {
+                ret = dict_get_int32 (xattr_rsp, "sh-failed", &sh_failed);
+                ret = dict_get_int32 (xattr_rsp, "actual-sh-done", &actual_sh_done);
+        }
+
+        shd_crawl_event = (shd_crawl_event_t**)(shd->crawl_events);
+
+        split_brain = afr_is_split_brain (this, child->inode);
+        if ((op_ret < 0 && op_errno == EIO) || split_brain) {
+                eh = shd->split_brain;
+                shd_crawl_event[crawl_data->child]->split_brain_count += 1;
+        } else if ((op_ret < 0) || sh_failed) {
+                eh = shd->heal_failed;
+                shd_crawl_event[crawl_data->child]->heal_failed_count += 1;
+        } else if (actual_sh_done == 1) {
+                eh = shd->healed;
+                shd_crawl_event[crawl_data->child]->healed_count += 1;
+        }
+        ret = -1;
+
+        if (eh != NULL) {
+                event = GF_CALLOC (1, sizeof (*event), gf_afr_mt_shd_event_t);
+                if (!event)
+                        goto out;
+                event->child = crawl_data->child;
+                event->path = path;
+
+                ret = eh_save_history (eh, event);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s:Failed to save "
+                                "to event history, (%d, %s)", path, op_ret,
+                                strerror (op_errno));
+
+                        goto out;
+                }
+        } else {
+                gf_log (this->name, GF_LOG_DEBUG, "%s:Self heal already done ",
+                        path);
+
+        }
+        ret = 0;
+out:
+        if (ret && path)
+                GF_FREE (path);
+        return;
+}
+
+int
+_link_inode_update_loc (xlator_t *this, loc_t *loc, struct iatt *iattr)
+{
+        inode_t       *link_inode = NULL;
+        int           ret = -1;
+
+        link_inode = inode_link (loc->inode, NULL, NULL, iattr);
+        if (link_inode == NULL) {
+                gf_log (this->name, GF_LOG_ERROR, "inode link failed "
+                        "on the inode (%s)", uuid_utoa (iattr->ia_gfid));
+                goto out;
+        }
+        inode_unref (loc->inode);
+        loc->inode = link_inode;
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+_self_heal_entry (xlator_t *this, afr_crawl_data_t *crawl_data, gf_dirent_t *entry,
+                  loc_t *child, loc_t *parent, struct iatt *iattr)
+{
+        struct iatt      parentbuf = {0};
+        int              ret = 0;
+        dict_t           *xattr_rsp = NULL;
+        dict_t           *xattr_req = NULL;
+
+        xattr_req = dict_new ();
+        if (!xattr_req) {
+                errno = ENOMEM;
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_int32 (xattr_req, "allow-sh-for-running-transaction", 1);
+
+        gf_log (this->name, GF_LOG_DEBUG, "lookup %s", child->path);
+
+        ret = syncop_lookup (this, child, xattr_req,
+                             iattr, &xattr_rsp, &parentbuf);
+        _crawl_post_sh_action (this, parent, child, ret, errno, xattr_rsp,
+                               crawl_data);
+        if (xattr_rsp)
+                dict_unref (xattr_rsp);
+        if (ret == 0)
+                ret = _link_inode_update_loc (this, child, iattr);
+
+out:
+        if (xattr_req)
+                dict_unref(xattr_req);
+        return ret;
+}
+
+static int
+afr_crawl_done  (int ret, call_frame_t *sync_frame, void *data)
+{
+        GF_FREE (data);
+        STACK_DESTROY (sync_frame->root);
+        return 0;
+}
+
+void
+_do_self_heal_on_subvol (xlator_t *this, int child, afr_crawl_type_t crawl)
+{
+        afr_start_crawl (this, child, crawl, _self_heal_entry,
+                         NULL, _gf_true, STOP_CRAWL_ON_SINGLE_SUBVOL,
+                         afr_crawl_done);
+}
+
+gf_boolean_t
+_crawl_proceed (xlator_t *this, int child, int crawl_flags, char **reason)
+{
+        afr_private_t           *priv = NULL;
+        afr_self_heald_t        *shd = NULL;
+        gf_boolean_t            proceed = _gf_false;
+        char                    *msg = NULL;
+
+        priv = this->private;
+        shd  = &priv->shd;
+        if (!shd->enabled) {
+                msg = "Self-heal daemon is not enabled";
+                gf_log (this->name, GF_LOG_DEBUG, "%s", msg);
+                goto out;
+        }
+        if (!priv->child_up[child]) {
+                gf_log (this->name, GF_LOG_DEBUG, "Stopping crawl for %s , "
+                        "subvol went down", priv->children[child]->name);
+                msg = "Brick is Not connected";
+                goto out;
+        }
+
+        if (crawl_flags & STOP_CRAWL_ON_SINGLE_SUBVOL) {
+                if (afr_up_children_count (priv->child_up,
+                                           priv->child_count) < 2) {
+                        gf_log (this->name, GF_LOG_DEBUG, "Stopping crawl as "
+                                "< 2 children are up");
+                        msg = "< 2 bricks in replica are running";
+                        goto out;
+                }
+        }
+        proceed = _gf_true;
+out:
+        if (reason)
+                *reason = msg;
+        return proceed;
+}
+
+int
+_do_crawl_op_on_local_subvols (xlator_t *this, afr_crawl_type_t crawl,
+                               shd_crawl_op op, dict_t *output)
+{
+        afr_private_t       *priv = NULL;
+        char                *status = NULL;
+        char                *subkey = NULL;
+        char                key[256] = {0};
+        shd_pos_t           pos_data = {0};
+        int                 op_ret = -1;
+        int                 xl_id = -1;
+        int                 i = 0;
+        int                 ret = 0;
+        int                 crawl_flags = 0;
+
+        priv = this->private;
+        if (op == HEAL)
+                crawl_flags |= STOP_CRAWL_ON_SINGLE_SUBVOL;
+
+        if (output) {
+                ret = dict_get_int32 (output, this->name, &xl_id);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Invalid input, "
+                                "translator-id is not available");
+                        goto out;
+                }
+        }
+        pos_data.this = this;
+        subkey = "status";
+        for (i = 0; i < priv->child_count; i++) {
+                if (_crawl_proceed (this, i, crawl_flags, &status)) {
+                        pos_data.child = i;
+                        /*
+                         * We're already in a synctask in this case, so we
+                         * don't need to defer through a second (and in fact
+                         * that can cause deadlock).  Just call straight
+                         * through instead.
+                         */
+                        ret = afr_find_child_position(pos_data.this,
+                                                      pos_data.child,
+                                                      &pos_data.pos);
+                        if (ret) {
+                                status = "Not able to find brick location";
+                        } else if (pos_data.pos == AFR_POS_REMOTE) {
+                                status = "brick is remote";
+                        } else {
+                                op_ret = 0;
+                                if (op == HEAL) {
+                                        status = "Started self-heal";
+                                        _do_self_heal_on_subvol (this, i,
+                                                                 crawl);
+                                } else if (output && (op == INFO)) {
+                                        status = "";
+                                        afr_start_crawl (this, i, INDEX,
+                                                         _add_summary_to_dict,
+                                                         output, _gf_false, 0,
+                                                         NULL);
+                                } else if (output &&
+                                           (op == STATISTICS_TO_BE_HEALED)) {
+                                            status = "";
+                                            afr_start_crawl (this, i,
+                                                             INDEX_TO_BE_HEALED,
+                                       _count_hard_links_under_base_indices_dir,
+                                                             output, _gf_false,
+                                                             0, NULL);
+                                }
+                        }
+                        if (output) {
+                                snprintf (key, sizeof (key), "%d-%d-%s", xl_id,
+                                          i, subkey);
+                                ret = dict_set_str (output, key, status);
+                        }
+                        if (!op_ret && (crawl == FULL))
+                                break;
+                }
+                if (output) {
+                        snprintf (key, sizeof (key), "%d-%d-%s", xl_id, i,
+                                  subkey);
+                        ret = dict_set_str (output, key, status);
+                }
+        }
+out:
+        return op_ret;
+}
+
+int
+_do_self_heal_on_local_subvols (xlator_t *this, afr_crawl_type_t crawl,
+                                dict_t *output)
+{
+        return _do_crawl_op_on_local_subvols (this, crawl, HEAL, output);
+}
+
+int
+_get_index_summary_on_local_subvols (xlator_t *this, dict_t *output)
+{
+        return _do_crawl_op_on_local_subvols (this, INDEX, INFO, output);
+}
+
+void
+afr_fill_completed_crawl_statistics_to_dict (xlator_t *this, dict_t *dict)
+{
+        afr_private_t           *priv  = NULL;
+        afr_self_heald_t        *shd = NULL;
+        int                     i = 0;
+        priv = this->private;
+        shd= &priv->shd;
+        for (i = 0; i < priv->child_count; i++) {
+                if (shd->pos[i] != AFR_POS_LOCAL)
+                        continue;
+                _add_statistics_to_dict (this, dict, i);
+        }
+
+        return ;
+}
+
+static void
+reset_crawl_event (shd_crawl_event_t *crawl_event)
+{
+    crawl_event->healed_count = 0;
+    crawl_event->split_brain_count = 0;
+    crawl_event->heal_failed_count = 0;
+    GF_FREE (crawl_event->start_time_str);
+    crawl_event->start_time_str = NULL;
+    crawl_event->end_time_str = NULL;
+    crawl_event->crawl_type = NULL;
+    crawl_event->crawl_inprogress = _gf_false;
+    return;
+}
+
+static void
+afr_copy_crawl_event_struct (shd_crawl_event_t *src, shd_crawl_event_t *dst)
+{
+        dst->healed_count = src->healed_count;
+        dst->split_brain_count = src->split_brain_count;
+        dst->heal_failed_count = src->heal_failed_count;
+        dst->start_time_str = gf_strdup (src->start_time_str);
+        dst->end_time_str = "Crawl is already in progress";
+        dst->crawl_type = src->crawl_type;
+        dst->crawl_inprogress = _gf_true;
+        return;
+}
+
+static int
+afr_fill_crawl_statistics_of_running_crawl(xlator_t *this, dict_t *dict)
+{
+        shd_crawl_event_t       *evnt = NULL;
+        int                     ret = 0;
+        afr_private_t           *priv = NULL;
+        afr_self_heald_t        *shd = NULL;
+        int                     i = 0;
+        priv = this->private;
+        shd = &priv->shd;
+
+        evnt = GF_CALLOC (1, sizeof (shd_crawl_event_t),
+                          gf_afr_mt_shd_crawl_event_t);
+        if (!evnt) {
+                ret = -1;
+                goto out;
+        }
+        LOCK (&priv->lock);
+        {
+                for (i = 0; i < priv->child_count; i++) {
+                        if (shd->pos[i] != AFR_POS_LOCAL)
+                                continue;
+
+                        reset_crawl_event (evnt);
+
+                        if (!shd->crawl_events[i]) {
+                                continue;
+                        }
+
+                        afr_copy_crawl_event_struct (shd->crawl_events[i],
+                                                     evnt);
+                        _add_crawl_stats_to_dict (this, dict, i, evnt, NULL);
+
+                }
+        }
+        UNLOCK (&priv->lock);
+        reset_crawl_event (evnt);
+        GF_FREE (evnt);
+
+out:
+        return ret;
+}
+
+static int
+_add_local_subvols_crawl_statistics_to_dict (xlator_t *this, dict_t *dict)
+{
+        int ret = 0;
+        afr_fill_completed_crawl_statistics_to_dict (this, dict);
+        ret = afr_fill_crawl_statistics_of_running_crawl (this, dict);
+        return ret;
+}
+int
+_add_local_subvols_eh_to_dict (xlator_t *this, eh_t *eh, dict_t *dict)
+{
+        afr_private_t           *priv = NULL;
+        afr_self_heald_t        *shd = NULL;
+        int                     i = 0;
+
+        priv = this->private;
+        shd = &priv->shd;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (shd->pos[i] != AFR_POS_LOCAL)
+                        continue;
+                _add_eh_to_dict (this, eh, dict, i);
+        }
+        return 0;
+}
+
+int
+afr_xl_op (xlator_t *this, dict_t *input, dict_t *output)
+{
+        gf_xl_afr_op_t   op = GF_AFR_OP_INVALID;
+        int              ret = 0;
+        afr_private_t    *priv = NULL;
+        afr_self_heald_t *shd = NULL;
+        int              xl_id = 0;
+
+        priv = this->private;
+        shd = &priv->shd;
+
+        ret = dict_get_int32 (input, "xl-op", (int32_t*)&op);
+        if (ret)
+                goto out;
+        ret = dict_get_int32 (input, this->name, &xl_id);
+        if (ret)
+                goto out;
+        ret = dict_set_int32 (output, this->name, xl_id);
+        if (ret)
+                goto out;
+        switch (op) {
+        case GF_AFR_OP_HEAL_INDEX:
+                ret = _do_self_heal_on_local_subvols (this, INDEX, output);
+                break;
+        case GF_AFR_OP_HEAL_FULL:
+                ret = _do_self_heal_on_local_subvols (this, FULL, output);
+                break;
+        case GF_AFR_OP_INDEX_SUMMARY:
+                (void)_get_index_summary_on_local_subvols (this, output);
+                ret = 0;
+                break;
+        case GF_AFR_OP_HEALED_FILES:
+                ret = _add_local_subvols_eh_to_dict (this, shd->healed, output);
+                break;
+        case GF_AFR_OP_HEAL_FAILED_FILES:
+                ret = _add_local_subvols_eh_to_dict (this, shd->heal_failed,
+                                                   output);
+                break;
+        case GF_AFR_OP_SPLIT_BRAIN_FILES:
+                ret = _add_local_subvols_eh_to_dict (this, shd->split_brain,
+                                                   output);
+                break;
+        case GF_AFR_OP_STATISTICS:
+                ret = _add_local_subvols_crawl_statistics_to_dict (this, output);
+                break;
+        case GF_AFR_OP_STATISTICS_HEAL_COUNT:
+        case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA:
+                ret = _do_crawl_op_on_local_subvols (this, INDEX_TO_BE_HEALED,
+                                                     STATISTICS_TO_BE_HEALED,
+                                                     output);
+                break;
+        default:
+                gf_log (this->name, GF_LOG_ERROR, "Unknown set op %d", op);
+                break;
+        }
+out:
+        dict_del (output, this->name);
+        return ret;
+}
+
+void
+afr_poll_self_heal (void *data)
+{
+        afr_private_t    *priv = NULL;
+        afr_self_heald_t *shd = NULL;
+        struct timespec  timeout = {0};
+        xlator_t         *this = NULL;
+        long             child = (long)data;
+        gf_timer_t       *old_timer = NULL;
+        gf_timer_t       *new_timer = NULL;
+        shd_pos_t        pos_data = {0};
+        int              ret = 0;
+
+        this = THIS;
+        priv = this->private;
+        shd = &priv->shd;
+
+        if (shd->pos[child] == AFR_POS_UNKNOWN) {
+                pos_data.this = this;
+                pos_data.child = child;
+                ret = synctask_new (this->ctx->env,
+                                    afr_syncop_find_child_position,
+                                    NULL, NULL, &pos_data);
+                if (!ret)
+                        shd->pos[child] = pos_data.pos;
+        }
+        if (shd->enabled && (shd->pos[child] == AFR_POS_LOCAL))
+                _do_self_heal_on_subvol (this, child, INDEX);
+        timeout.tv_sec = shd->timeout;
+        timeout.tv_nsec = 0;
+        //notify and previous timer should be synchronized.
+        LOCK (&priv->lock);
+        {
+                old_timer = shd->timer[child];
+                if (shd->pos[child] == AFR_POS_REMOTE)
+                        goto unlock;
+                shd->timer[child] = gf_timer_call_after (this->ctx, timeout,
+                                                         afr_poll_self_heal,
+                                                         data);
+                new_timer = shd->timer[child];
+        }
+unlock:
+        UNLOCK (&priv->lock);
+
+        if (old_timer)
+                gf_timer_call_cancel (this->ctx, old_timer);
+        if (!new_timer && (shd->pos[child] != AFR_POS_REMOTE)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Could not create self-heal polling timer for %s",
+                        priv->children[child]->name);
+        }
+        return;
+}
+
+static int
+afr_handle_child_up  (int ret, call_frame_t *sync_frame, void *data)
+{
+        afr_self_heald_t *shd = NULL;
+        shd_pos_t        *pos_data = data;
+        afr_private_t    *priv = NULL;
+
+        if (ret)
+                goto out;
+
+        priv = pos_data->this->private;
+        shd = &priv->shd;
+        shd->pos[pos_data->child] = pos_data->pos;
+        if (pos_data->pos != AFR_POS_REMOTE)
+                afr_poll_self_heal ((void*)(long)pos_data->child);
+        _do_self_heal_on_local_subvols (THIS, INDEX, NULL);
+out:
+        GF_FREE (data);
+        return 0;
+}
+
+void
+afr_proactive_self_heal (void *data)
+{
+        xlator_t         *this = NULL;
+        long             child = (long)data;
+        shd_pos_t        *pos_data = NULL;
+        int              ret = 0;
+
+        this = THIS;
+
+        //Position of brick could have changed and it could be local now.
+        //Compute the position again
+        pos_data = GF_CALLOC (1, sizeof (*pos_data), gf_afr_mt_pos_data_t);
+        if (!pos_data)
+                goto out;
+        pos_data->this = this;
+        pos_data->child = child;
+        ret = synctask_new (this->ctx->env, afr_syncop_find_child_position,
+                            afr_handle_child_up, NULL, pos_data);
+        if (ret)
+                goto out;
+out:
+        return;
+}
+
+static int
+get_pathinfo_host (char *pathinfo, char *hostname, size_t size)
+{
+        char    *start = NULL;
+        char    *end = NULL;
+        int     ret  = -1;
+        int     i    = 0;
+
+        if (!pathinfo)
+                goto out;
+
+        start = strchr (pathinfo, ':');
+        if (!start)
+                goto out;
+        end = strrchr (pathinfo, ':');
+        if (start == end)
+                goto out;
+
+        memset (hostname, 0, size);
+        i = 0;
+        while (++start != end)
+                hostname[i++] = *start;
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+afr_local_pathinfo (char *pathinfo, gf_boolean_t *local)
+{
+        int             ret   = 0;
+        char            pathinfohost[1024] = {0};
+        char            localhost[1024] = {0};
+        xlator_t        *this = THIS;
+
+        *local = _gf_false;
+        ret = get_pathinfo_host (pathinfo, pathinfohost, sizeof (pathinfohost));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Invalid pathinfo: %s",
+                        pathinfo);
+                goto out;
+        }
+
+        ret = gethostname (localhost, sizeof (localhost));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "gethostname() failed, "
+                        "reason: %s", strerror (errno));
+                goto out;
+        }
+
+        if (!strcmp (localhost, pathinfohost))
+                *local = _gf_true;
+out:
+        return ret;
+}
+
+int
+afr_crawl_build_start_loc (xlator_t *this, afr_crawl_data_t *crawl_data,
+                           loc_t *dirloc)
+{
+        afr_private_t *priv = NULL;
+        dict_t        *xattr = NULL;
+        void          *index_gfid = NULL;
+        void          *base_indices_holder_vgfid = NULL;
+        loc_t         rootloc = {0};
+        struct iatt   iattr = {0};
+        struct iatt   parent = {0};
+        int           ret = 0;
+        xlator_t      *readdir_xl = crawl_data->readdir_xl;
+
+        priv = this->private;
+        if (crawl_data->crawl == FULL) {
+                afr_build_root_loc (this, dirloc);
+        } else if (crawl_data->crawl == INDEX) {
+                afr_build_root_loc (this, &rootloc);
+                ret = syncop_getxattr (readdir_xl, &rootloc, &xattr,
+                                       GF_XATTROP_INDEX_GFID);
+                if (ret < 0)
+                        goto out;
+                ret = dict_get_ptr (xattr, GF_XATTROP_INDEX_GFID, &index_gfid);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to get index "
+                                "dir gfid on %s", readdir_xl->name);
+                        goto out;
+                }
+                if (!index_gfid) {
+                        gf_log (this->name, GF_LOG_ERROR, "index gfid empty "
+                                "on %s", readdir_xl->name);
+                        ret = -1;
+                        goto out;
+                }
+                uuid_copy (dirloc->gfid, index_gfid);
+                dirloc->path = "";
+                dirloc->inode = inode_new (priv->root_inode->table);
+                ret = syncop_lookup (readdir_xl, dirloc, NULL,
+                                     &iattr, NULL, &parent);
+                if (ret < 0) {
+                        if (errno != ENOENT) {
+                                gf_log (this->name, GF_LOG_ERROR, "lookup "
+                                        "failed on index dir on %s - (%s)",
+                                        readdir_xl->name, strerror (errno));
+                        }
+                        goto out;
+                }
+                ret = _link_inode_update_loc (this, dirloc, &iattr);
+                if (ret)
+                        goto out;
+        } else if (crawl_data->crawl == INDEX_TO_BE_HEALED) {
+                afr_build_root_loc (this, &rootloc);
+                ret = syncop_getxattr (readdir_xl, &rootloc, &xattr,
+                                       GF_BASE_INDICES_HOLDER_GFID);
+                if (ret < 0)
+                        goto out;
+                ret = dict_get_ptr (xattr, GF_BASE_INDICES_HOLDER_GFID,
+                                    &base_indices_holder_vgfid);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "index gfid empty "
+                                "on %s", readdir_xl->name);
+                        ret = -1;
+                        goto out;
+                }
+                if (!base_indices_holder_vgfid) {
+                        gf_log (this->name, GF_LOG_ERROR, "Base indices holder"
+                                "virtual gfid is null on %s", readdir_xl->name);
+                        ret = -1;
+                        goto out;
+                }
+                uuid_copy (dirloc->gfid,  base_indices_holder_vgfid);
+                dirloc->path = "";
+                dirloc->inode = inode_new (priv->root_inode->table);
+                ret = syncop_lookup (readdir_xl, dirloc, NULL, &iattr, NULL,
+                                     &parent);
+                if (ret < 0) {
+                        if (errno != ENOENT) {
+                                gf_log (this->name, GF_LOG_ERROR, "lookup "
+                                        "failed for base_indices_holder dir"
+                                        " on %s - (%s)", readdir_xl->name,
+                                        strerror (errno));
+
+                        } else {
+                                gf_log (this->name, GF_LOG_ERROR, "base_indices"
+                                        "_holder is not yet created.");
+                        }
+                        goto out;
+                }
+                ret = _link_inode_update_loc (this, dirloc, &iattr);
+                if (ret)
+                        goto out;
+        }
+        ret = 0;
+out:
+        if (xattr)
+                dict_unref (xattr);
+        loc_wipe (&rootloc);
+        return ret;
+}
+
+int
+afr_crawl_opendir (xlator_t *this, afr_crawl_data_t *crawl_data, fd_t **dirfd,
+                   loc_t *dirloc)
+{
+        fd_t          *fd   = NULL;
+        int           ret = 0;
+
+        if (crawl_data->crawl == FULL) {
+                fd = fd_create (dirloc->inode, crawl_data->pid);
+                if (!fd) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to create fd for %s", dirloc->path);
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = syncop_opendir (crawl_data->readdir_xl, dirloc, fd);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "opendir failed on %s", dirloc->path);
+                        goto out;
+                }
+        } else {
+                fd = fd_anonymous (dirloc->inode);
+        }
+        ret = 0;
+out:
+        if (!ret)
+                *dirfd = fd;
+        return ret;
+}
+
+xlator_t*
+afr_crawl_readdir_xl_get (xlator_t *this, afr_crawl_data_t *crawl_data)
+{
+        afr_private_t *priv = this->private;
+
+        if (crawl_data->crawl == FULL) {
+                return this;
+        } else {
+                return priv->children[crawl_data->child];
+        }
+        return NULL;
+}
+
+int
+afr_crawl_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent,
+                           gf_dirent_t *entry, afr_crawl_data_t *crawl_data)
+{
+        int           ret = -1;
+        afr_private_t *priv = NULL;
+
+        priv = this->private;
+        if (crawl_data->crawl == FULL) {
+                ret = afr_build_child_loc (this, child, parent, entry->d_name);
+        } else if (crawl_data->crawl == INDEX_TO_BE_HEALED) {
+                ret = _build_index_loc (this, child, entry->d_name, parent);
+                if (ret)
+                        goto out;
+                child->inode = inode_new (priv->root_inode->table);
+                if (!child->inode) {
+                        ret = -1;
+                        goto out;
+                }
+                child->path = NULL;
+        } else {
+                child->inode = inode_new (priv->root_inode->table);
+                if (!child->inode)
+                        goto out;
+                uuid_parse (entry->d_name, child->gfid);
+                ret = _loc_assign_gfid_path (child);
+        }
+out:
+        return ret;
+}
+
+static int
+_process_entries (xlator_t *this, loc_t *parentloc, gf_dirent_t *entries,
+                  off_t *offset, afr_crawl_data_t *crawl_data)
+{
+        gf_dirent_t      *entry = NULL;
+        gf_dirent_t      *tmp = NULL;
+        int              ret = 0;
+        loc_t            entry_loc = {0};
+        fd_t             *fd = NULL;
+        struct iatt      iattr = {0};
+
+        list_for_each_entry_safe (entry, tmp, &entries->list, list) {
+                if (!_crawl_proceed (this, crawl_data->child,
+                                     crawl_data->crawl_flags, NULL)) {
+                        ret = -1;
+                        goto out;
+                }
+                *offset = entry->d_off;
+                if (IS_ENTRY_CWD (entry->d_name) ||
+                    IS_ENTRY_PARENT (entry->d_name))
+                        continue;
+                if ((crawl_data->crawl == FULL) &&
+                     uuid_is_null (entry->d_stat.ia_gfid)) {
+                        gf_log (this->name, GF_LOG_WARNING, "%s/%s: No "
+                                "gfid present skipping",
+                                parentloc->path, entry->d_name);
+                        continue;
+                }
+
+                loc_wipe (&entry_loc);
+                ret = afr_crawl_build_child_loc (this, &entry_loc, parentloc,
+                                                 entry, crawl_data);
+                if (ret)
+                        goto out;
+
+                ret = crawl_data->process_entry (this, crawl_data, entry,
+                                                 &entry_loc, parentloc, &iattr);
+
+                if (crawl_data->crawl == INDEX_TO_BE_HEALED && ret) {
+                       goto out;
+                } else if (ret) {
+                        continue;
+                }
+
+                if ((crawl_data->crawl == INDEX) ||
+                    (crawl_data->crawl == INDEX_TO_BE_HEALED))
+                        continue;
+
+                if (!IA_ISDIR (iattr.ia_type))
+                        continue;
+                fd = NULL;
+                ret = afr_crawl_opendir (this, crawl_data, &fd, &entry_loc);
+                if (ret)
+                        continue;
+                ret = _crawl_directory (fd, &entry_loc, crawl_data);
+                if (fd)
+                        fd_unref (fd);
+        }
+        ret = 0;
+out:
+        if ((crawl_data->crawl == INDEX_TO_BE_HEALED)  && ret) {
+                gf_log (this->name, GF_LOG_ERROR,"Failed to get the hardlink "
+                        "count");
+        }
+        loc_wipe (&entry_loc);
+        return ret;
+}
+
+static int
+_crawl_directory (fd_t *fd, loc_t *loc, afr_crawl_data_t *crawl_data)
+{
+        xlator_t        *this = NULL;
+        off_t           offset   = 0;
+        gf_dirent_t     entries;
+        int             ret = 0;
+        gf_boolean_t    free_entries = _gf_false;
+        xlator_t        *readdir_xl = crawl_data->readdir_xl;
+
+        INIT_LIST_HEAD (&entries.list);
+        this = THIS;
+
+        GF_ASSERT (loc->inode);
+
+        if (crawl_data->crawl == FULL)
+                gf_log (this->name, GF_LOG_DEBUG, "crawling %s", loc->path);
+        else
+                gf_log (this->name, GF_LOG_DEBUG, "crawling INDEX %s",
+                        uuid_utoa (loc->gfid));
+
+        while (1) {
+                if (crawl_data->crawl == FULL)
+                        ret = syncop_readdirp (readdir_xl, fd, 131072, offset,
+                                               NULL, &entries);
+                else
+                        ret = syncop_readdir (readdir_xl, fd, 131072, offset,
+                                              &entries);
+                if (ret <= 0)
+                        break;
+                ret = 0;
+                free_entries = _gf_true;
+
+                if (!_crawl_proceed (this, crawl_data->child,
+                                     crawl_data->crawl_flags, NULL)) {
+                        ret = -1;
+                        goto out;
+                }
+                if (list_empty (&entries.list))
+                        goto out;
+
+                ret = _process_entries (this, loc, &entries, &offset,
+                                        crawl_data);
+                if ((ret < 0) && (crawl_data->crawl == INDEX_TO_BE_HEALED)) {
+                        goto out;
+                }
+                gf_dirent_free (&entries);
+                free_entries = _gf_false;
+        }
+        ret = 0;
+out:
+        if (free_entries)
+                gf_dirent_free (&entries);
+        return ret;
+}
+
+static char*
+position_str_get (afr_child_pos_t pos)
+{
+        switch (pos) {
+        case AFR_POS_UNKNOWN:
+                return "unknown";
+        case AFR_POS_LOCAL:
+                return "local";
+        case AFR_POS_REMOTE:
+                return "remote";
+        }
+        return NULL;
+}
+
+int
+afr_find_child_position (xlator_t *this, int child, afr_child_pos_t *pos)
+{
+        afr_private_t    *priv = NULL;
+        afr_self_heald_t *shd  = NULL;
+        dict_t           *xattr_rsp = NULL;
+        loc_t            loc = {0};
+        int              ret = 0;
+        char             *node_uuid = NULL;
+
+        priv = this->private;
+        shd  = &priv->shd;
+
+        afr_build_root_loc (this, &loc);
+
+        ret = syncop_getxattr (priv->children[child], &loc, &xattr_rsp,
+                               GF_XATTR_NODE_UUID_KEY);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "getxattr failed on %s - "
+                        "(%s)", priv->children[child]->name, strerror (errno));
+                goto out;
+        }
+
+        ret = dict_get_str (xattr_rsp, GF_XATTR_NODE_UUID_KEY, &node_uuid);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "node-uuid key not found on "
+                        "child %s", priv->children[child]->name);
+                goto out;
+        }
+
+        if (!strcmp (node_uuid, shd->node_uuid))
+                *pos = AFR_POS_LOCAL;
+        else
+                *pos = AFR_POS_REMOTE;
+
+        gf_log (this->name, GF_LOG_DEBUG, "child %s is %s",
+                priv->children[child]->name, position_str_get (*pos));
+out:
+        if (ret)
+                *pos = AFR_POS_UNKNOWN;
+        loc_wipe (&loc);
+        return ret;
+}
+
+int
+afr_syncop_find_child_position (void *data)
+{
+        shd_pos_t *pos_data = data;
+        int       ret = 0;
+
+        ret = afr_find_child_position (pos_data->this, pos_data->child,
+                                       &pos_data->pos);
+        return ret;
+}
+
+static int
+afr_dir_crawl (void *data)
+{
+        xlator_t            *this = NULL;
+        int                 ret = -1;
+        xlator_t            *readdir_xl = NULL;
+        fd_t                *fd = NULL;
+        loc_t               dirloc = {0};
+        afr_crawl_data_t    *crawl_data = data;
+
+        this = THIS;
+
+        if (!_crawl_proceed (this, crawl_data->child, crawl_data->crawl_flags,
+                             NULL))
+                goto out;
+
+        readdir_xl = afr_crawl_readdir_xl_get (this, crawl_data);
+        if (!readdir_xl)
+                goto out;
+        crawl_data->readdir_xl = readdir_xl;
+
+        ret = afr_crawl_build_start_loc (this, crawl_data, &dirloc);
+        if (ret)
+                goto out;
+
+        ret = afr_crawl_opendir (this, crawl_data, &fd, &dirloc);
+        if (ret) {
+                if (crawl_data->crawl == INDEX_TO_BE_HEALED) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to open base_"
+                                "indices_holder");
+                }
+                goto out;
+        }
+
+        ret = _crawl_directory (fd, &dirloc, crawl_data);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "Crawl failed on %s",
+                        readdir_xl->name);
+        else
+                gf_log (this->name, GF_LOG_DEBUG, "Crawl completed "
+                        "on %s", readdir_xl->name);
+        if (crawl_data->crawl == INDEX)
+                dirloc.path = NULL;
+out:
+        if (fd)
+                fd_unref (fd);
+        if ((crawl_data->crawl == INDEX) ||
+            (crawl_data->crawl == INDEX_TO_BE_HEALED ))
+                dirloc.path = NULL;
+        loc_wipe (&dirloc);
+        return ret;
+}
+
+char *
+get_crawl_type_in_string (afr_crawl_type_t crawl)
+{
+        char    *index = "INDEX";
+        char    *full  = "FULL";
+        char    *crawl_type = NULL;
+
+        if (crawl == INDEX){
+                crawl_type = index;
+        } else if (crawl == FULL) {
+                crawl_type = full;
+        }
+
+        return  crawl_type;
+}
+
+static int
+afr_allocate_crawl_event (xlator_t *this, int child, afr_crawl_type_t crawl)
+{
+        afr_private_t           *priv = NULL;
+        afr_self_heald_t        *shd = NULL;
+        int                     ret = 0;
+        shd_crawl_event_t       *crawl_event = NULL;
+        time_t                  get_time = 0;
+
+        priv = this->private;
+        shd = &priv->shd;
+
+        crawl_event = GF_CALLOC (sizeof (shd_crawl_event_t), 1,
+                                 gf_afr_mt_shd_crawl_event_t);
+        if (!crawl_event) {
+                ret = -1;
+                goto out;
+        }
+
+        get_time =  time(NULL);
+        if (get_time == ((time_t)-1)) {
+                 ret = -1;
+                goto out;
+        }
+
+        crawl_event->start_time_str = gf_strdup (ctime(&get_time));
+
+        crawl_event->crawl_type = get_crawl_type_in_string (crawl);
+        if (!crawl_event->crawl_type) {
+                ret = -1;
+                goto out;
+        }
+        LOCK (&priv->lock);
+        {
+                shd->crawl_events[child] = crawl_event;
+        }
+        UNLOCK (&priv->lock);
+        ret = 0;
+out:
+        return ret;
+
+}
+
+static int
+afr_put_crawl_event_in_eh (xlator_t *this, int child)
+{
+        afr_private_t           *priv = NULL;
+        afr_self_heald_t        *shd = NULL;
+        int                     ret = 0;
+        time_t                  get_time = 0;
+        shd_crawl_event_t       **crawl_event = NULL;
+
+        priv = this->private;
+        shd = &priv->shd;
+
+        get_time = time(NULL);
+        if (get_time == ((time_t)-1)) {
+                ret = -1;
+                goto out;
+        }
+        crawl_event = (shd_crawl_event_t**)shd->crawl_events;
+        LOCK (&priv->lock);
+        {
+                crawl_event[child]->end_time_str = gf_strdup (ctime(&get_time));
+                ret = eh_save_history (shd->statistics[child],
+                                       crawl_event[child]);
+                crawl_event[child] = NULL;
+        }
+        UNLOCK (&priv->lock);
+out:
+        return ret;
+}
+
+static int
+afr_dir_exclusive_crawl (void *data)
+{
+        afr_private_t    *priv = NULL;
+        afr_self_heald_t *shd = NULL;
+        gf_boolean_t     crawl = _gf_false;
+        int              ret = 0;
+        int              child = -1;
+        xlator_t         *this = NULL;
+        afr_crawl_data_t *crawl_data = data;
+
+        this = THIS;
+        priv = this->private;
+        shd = &priv->shd;
+        child = crawl_data->child;
+
+        LOCK (&priv->lock);
+        {
+                if (shd->inprogress[child]) {
+                        if (shd->pending[child] != FULL)
+                                shd->pending[child] = crawl_data->crawl;
+                } else {
+                        shd->inprogress[child] = _gf_true;
+                        crawl = _gf_true;
+                }
+        }
+        UNLOCK (&priv->lock);
+
+        if (!crawl) {
+                gf_log (this->name, GF_LOG_INFO, "Another crawl is in progress "
+                        "for %s", priv->children[child]->name);
+                goto out;
+        }
+
+        do {
+                ret = afr_allocate_crawl_event (this, child, crawl_data->crawl);
+                if (ret)
+                        goto out;
+                afr_dir_crawl (data);
+
+                ret = afr_put_crawl_event_in_eh (this, child);
+                if (ret < 0)
+                        goto out;
+
+                LOCK (&priv->lock);
+                {
+                        if (shd->pending[child] != NONE) {
+                                crawl_data->crawl = shd->pending[child];
+                                shd->pending[child] = NONE;
+                        } else {
+                                shd->inprogress[child] = _gf_false;
+                                crawl = _gf_false;
+                        }
+                }
+                UNLOCK (&priv->lock);
+        } while (crawl);
+out:
+        return ret;
+}
+
+void
+afr_start_crawl (xlator_t *this, int idx, afr_crawl_type_t crawl,
+                 process_entry_cbk_t process_entry, void *op_data,
+                 gf_boolean_t exclusive, int crawl_flags,
+                 afr_crawl_done_cbk_t crawl_done)
+{
+        afr_private_t              *priv = NULL;
+        call_frame_t               *frame = NULL;
+        afr_crawl_data_t           *crawl_data = NULL;
+        int                        ret = 0;
+        int (*crawler) (void*) = NULL;
+
+        priv = this->private;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame)
+                goto out;
+
+        afr_set_lk_owner (frame, this, frame->root);
+        afr_set_low_priority (frame);
+        crawl_data = GF_CALLOC (1, sizeof (*crawl_data),
+                                gf_afr_mt_crawl_data_t);
+        if (!crawl_data)
+                goto out;
+        crawl_data->process_entry = process_entry;
+        crawl_data->child = idx;
+        crawl_data->pid = frame->root->pid;
+        crawl_data->crawl = crawl;
+        crawl_data->op_data = op_data;
+        crawl_data->crawl_flags = crawl_flags;
+        gf_log (this->name, GF_LOG_DEBUG, "starting crawl %d for %s",
+                crawl_data->crawl, priv->children[idx]->name);
+
+        if (exclusive)
+                crawler = afr_dir_exclusive_crawl;
+        else
+                crawler = afr_dir_crawl;
+        ret = synctask_new (this->ctx->env, crawler,
+                            crawl_done, frame, crawl_data);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "afr crawl failed for child"
+                        " %d with ret %d", idx, ret);
+out:
+        return;
+}
+
+void
+afr_build_root_loc (xlator_t *this, loc_t *loc)
+{
+        afr_private_t   *priv = NULL;
+
+        priv = this->private;
+        loc->path = gf_strdup ("/");
+        loc->name = "";
+        loc->inode = inode_ref (priv->root_inode);
+        uuid_copy (loc->gfid, loc->inode->gfid);
+}
+
+int
+afr_set_root_gfid (dict_t *dict)
+{
+        uuid_t gfid;
+        int ret = 0;
+
+        memset (gfid, 0, 16);
+        gfid[15] = 1;
+
+        ret = afr_set_dict_gfid (dict, gfid);
+
+        return ret;
+}
diff --git a/xlators/cluster/afr/src/afr-self-heald.h b/xlators/cluster/afr/src/afr-self-heald.h
new file mode 100644
index 000000000..e0c083754
--- /dev/null
+++ b/xlators/cluster/afr/src/afr-self-heald.h
@@ -0,0 +1,65 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef __AFR_SELF_HEALD_H__
+#define __AFR_SELF_HEALD_H__
+#include "xlator.h"
+
+#define IS_ROOT_PATH(path) (!strcmp (path, "/"))
+#define IS_ENTRY_CWD(entry) (!strcmp (entry, "."))
+#define IS_ENTRY_PARENT(entry) (!strcmp (entry, ".."))
+#define AFR_ALL_CHILDREN -1
+
+typedef struct afr_crawl_data_ {
+        int                 child;
+        pid_t               pid;
+        afr_crawl_type_t    crawl;
+        xlator_t            *readdir_xl;
+        void                *op_data;
+        int                 crawl_flags;
+        int (*process_entry) (xlator_t *this, struct afr_crawl_data_ *crawl_data,
+                              gf_dirent_t *entry, loc_t *child, loc_t *parent,
+                              struct iatt *iattr);
+} afr_crawl_data_t;
+
+typedef struct crawl_event_stats_ {
+        uint64_t healed_count;
+        uint64_t split_brain_count;
+        uint64_t heal_failed_count;
+        char     *start_time_str;
+        char     *end_time_str;
+        char     *crawl_type;
+        gf_boolean_t crawl_inprogress;
+} shd_crawl_event_t;
+
+void _destroy_crawl_event_data (void *data);
+void _destroy_shd_event_data (void *data);
+
+typedef int (*process_entry_cbk_t) (xlator_t *this, afr_crawl_data_t *crawl_data,
+                              gf_dirent_t *entry, loc_t *child, loc_t *parent,
+                              struct iatt *iattr);
+
+void afr_build_root_loc (xlator_t *this, loc_t *loc);
+
+int afr_set_root_gfid (dict_t *dict);
+
+void
+afr_proactive_self_heal (void *data);
+
+int
+afr_xl_op (xlator_t *this, dict_t *input, dict_t *output);
+
+/*
+ * In addition to its self-heal use, this is used to find a local default
+ * read_child.
+ */
+int
+afr_local_pathinfo (char *pathinfo, gf_boolean_t *local);
+#endif /* __AFR_SELF_HEALD_H__ */
diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c
index a382c12a8..20306e469 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -1,25 +1,17 @@
 /*
-  Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include "dict.h"
 #include "byte-order.h"
 #include "common-utils.h"
+#include "timer.h"
 
 #include "afr.h"
 #include "afr-transaction.h"
@@ -32,39 +24,85 @@
                                       of RENAME */
 #define LOCKED_LOWER    0x2        /* for lower_path of RENAME */
 
+afr_fd_ctx_t *
+__afr_fd_ctx_get (fd_t *fd, xlator_t *this)
+{
+        uint64_t       ctx = 0;
+        int            ret = 0;
+        afr_fd_ctx_t  *fd_ctx = NULL;
+        int            i = 0;
+        afr_private_t *priv = NULL;
+
+        priv = this->private;
+
+        ret = __fd_ctx_get (fd, this, &ctx);
+
+        if (ret < 0 && fd_is_anonymous (fd)) {
+                ret = __afr_fd_ctx_set (this, fd);
+                if (ret < 0)
+                        goto out;
+
+                ret = __fd_ctx_get (fd, this, &ctx);
+                if (ret < 0)
+                        goto out;
+
+                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+                for (i = 0; i < priv->child_count; i++)
+                        fd_ctx->opened_on[i] = AFR_FD_OPENED;
+        }
+
+        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+out:
+        return fd_ctx;
+}
+
+
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this)
+{
+        afr_fd_ctx_t  *fd_ctx = NULL;
+
+        LOCK(&fd->lock);
+        {
+                fd_ctx = __afr_fd_ctx_get (fd, this);
+        }
+        UNLOCK(&fd->lock);
+
+        return fd_ctx;
+}
+
 
 static void
-afr_pid_save (call_frame_t *frame)
+afr_save_lk_owner (call_frame_t *frame)
 {
         afr_local_t * local = NULL;
 
         local = frame->local;
 
-        local->saved_pid = frame->root->pid;
+        local->saved_lk_owner = frame->root->lk_owner;
 }
 
 
 static void
-afr_pid_restore (call_frame_t *frame)
+afr_restore_lk_owner (call_frame_t *frame)
 {
         afr_local_t * local = NULL;
 
         local = frame->local;
 
-        frame->root->pid = local->saved_pid;
+        frame->root->lk_owner = local->saved_lk_owner;
 }
 
-
 static void
 __mark_all_pending (int32_t *pending[], int child_count,
                     afr_transaction_type type)
-{	
-	int i;
-        int j;
+{
+        int i = 0;
+        int j = 0;
 
         for (i = 0; i < child_count; i++) {
                 j = afr_index_for_transaction_type (type);
-		pending[i][j] = hton32 (1);
+                pending[i][j] = hton32 (1);
         }
 }
 
@@ -73,801 +111,1000 @@ static void
 __mark_child_dead (int32_t *pending[], int child_count, int child,
                    afr_transaction_type type)
 {
-        int j;
+        int j = 0;
 
         j = afr_index_for_transaction_type (type);
-        
-	pending[child][j] = 0;
+
+        pending[child][j] = 0;
 }
 
 
 static void
-__mark_fop_failed_on_fd (fd_t *fd, xlator_t *this,
-                         int child_index)
+__mark_pre_op_done_on_fd (call_frame_t *frame, xlator_t *this, int child_index)
 {
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
+        afr_local_t   *local = NULL;
+        afr_fd_ctx_t  *fd_ctx = NULL;
 
-        int ret = 0;
+        local = frame->local;
 
-        ret = fd_ctx_get (fd, this, &ctx);
+        if (!local->fd)
+                return;
 
-        if (ret < 0)
-                goto out;
+        fd_ctx = afr_fd_ctx_get (local->fd, this);
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        if (!fd_ctx)
+                goto out;
 
-        fd_ctx->child_failed[child_index] = 1;
+        LOCK (&local->fd->lock);
+        {
+                if (local->transaction.type == AFR_DATA_TRANSACTION)
+                        fd_ctx->pre_op_done[child_index]++;
+        }
+        UNLOCK (&local->fd->lock);
 out:
         return;
 }
 
-
 static void
-__mark_failed_children (int32_t *pending[], int child_count, 
-                        xlator_t *this, fd_t *fd, afr_transaction_type type)
+__mark_non_participant_children (int32_t *pending[], int child_count,
+                                 unsigned char *participants,
+                                 afr_transaction_type type)
 {
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
-
-        int ret = 0;
-        int i   = 0;
-        int j   = 0;
-
-        ret = fd_ctx_get (fd, this, &ctx);
-
-        if (ret < 0)
-                goto out;
-
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        int i = 0;
+        int j = 0;
 
+        j = afr_index_for_transaction_type (type);
         for (i = 0; i < child_count; i++) {
-                j = afr_index_for_transaction_type (type);
-
-                if (fd_ctx->child_failed[i])
+                if (!participants[i])
                         pending[i][j] = 0;
         }
-        
-out:
-        return;
 }
 
 
-static void
-__mark_pre_op_done_on_fd (call_frame_t *frame, xlator_t *this, int child_index)
+void
+__mark_all_success (int32_t *pending[], int child_count,
+                    afr_transaction_type type)
 {
-        afr_local_t *local = NULL;
-
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
-        int ret = 0;
-
-        local = frame->local;
-
-        ret = fd_ctx_get (local->fd, this, &ctx);
-
-        if (ret < 0)
-                goto out;
-
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        int i;
+        int j;
 
-        if ((local->op == GF_FOP_WRITE)
-            || (local->op == GF_FOP_FTRUNCATE)) {
-                fd_ctx->pre_op_done[child_index] = 1;
+        for (i = 0; i < child_count; i++) {
+                j = afr_index_for_transaction_type (type);
+                pending[i][j] = hton32 (-1);
         }
-
-out:
-        return;
 }
 
-
-static void
-__mark_down_children (int32_t *pending[], int child_count, 
-                      unsigned char *child_up, afr_transaction_type type)
+void
+_set_all_child_errno (int *child_errno, unsigned int child_count)
 {
-	int i;
-	int j;
+        int     i = 0;
 
-	for (i = 0; i < child_count; i++) {
-                j = afr_index_for_transaction_type (type);
-
-		if (!child_up[i])
-			pending[i][j] = 0;
-        }
+        for (i = 0; i < child_count; i++)
+                if (child_errno[i] == 0)
+                        child_errno[i] = ENOTCONN;
 }
 
-
-static void
-__mark_all_success (int32_t *pending[], int child_count,
-                    afr_transaction_type type)
+void
+afr_transaction_perform_fop (call_frame_t *frame, xlator_t *this)
 {
-	int i;
-        int j;
+        afr_local_t     *local = NULL;
+        afr_private_t   *priv = NULL;
+        fd_t            *fd   = NULL;
 
-	for (i = 0; i < child_count; i++) {
-                j = afr_index_for_transaction_type (type);
-		pending[i][j] = hton32 (-1);
-        }
+        local = frame->local;
+        priv  = this->private;
+        fd    = local->fd;
+
+        __mark_all_success (local->pending, priv->child_count,
+                            local->transaction.type);
+
+        _set_all_child_errno (local->child_errno, priv->child_count);
+
+        /*  Perform fops with the lk-owner from top xlator.
+         *  Eg: lk-owner of posix-lk and flush should be same,
+         *  flush cant clear the  posix-lks without that lk-owner.
+         */
+        afr_save_lk_owner (frame);
+        frame->root->lk_owner =
+                local->transaction.main_frame->root->lk_owner;
+
+
+        /* The wake up needs to happen independent of
+           what type of fop arrives here. If it was
+           a write, then it has already inherited the
+           lock and changelog. If it was not a write,
+           then the presumption of the optimization (of
+           optimizing for successive write operations)
+           fails.
+        */
+        if (fd)
+                afr_delayed_changelog_wake_up (this, fd);
+        local->transaction.fop (frame, this);
 }
 
 
 static int
-__is_first_write_on_fd (xlator_t *this, fd_t *fd)
+__changelog_enabled (afr_private_t *priv, afr_transaction_type type)
 {
-        int op_ret     = 0;
-        int _ret       = -1;
-        int i          = 0;
+        int ret = 0;
 
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
+        switch (type) {
+        case AFR_DATA_TRANSACTION:
+                if (priv->data_change_log)
+                        ret = 1;
 
-        afr_private_t *priv = NULL;
-
-        priv = this->private;
+                break;
 
-        LOCK (&fd->lock);
-        {
-                _ret = __fd_ctx_get (fd, this, &ctx);
-                
-                if (_ret < 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "could not get fd ctx on fd=%p",
-                                fd);
-                        goto out;
-                }
+        case AFR_METADATA_TRANSACTION:
+                if (priv->metadata_change_log)
+                        ret = 1;
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+                break;
 
-                op_ret = 1;
-                for (i = 0; i < priv->child_count; i++) {
-                        if (fd_ctx->pre_op_done[i] == 0)
-                                continue;
+        case AFR_ENTRY_TRANSACTION:
+        case AFR_ENTRY_RENAME_TRANSACTION:
+                if (priv->entry_change_log)
+                        ret = 1;
 
-                        op_ret = 0;
-                }
+                break;
         }
-out:
-        UNLOCK (&fd->lock);
 
-        return op_ret;
+        return ret;
 }
 
 
 static int
-__if_fd_pre_op_done (xlator_t *this, fd_t *fd, int child_index)
+__fop_changelog_needed (call_frame_t *frame, xlator_t *this)
 {
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
         int op_ret = 0;
-        int _ret   = -1;
+        afr_transaction_type type = -1;
 
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
+        priv  = this->private;
+        local = frame->local;
+        type  = local->transaction.type;
 
-        LOCK (&fd->lock);
-        {
-                _ret = __fd_ctx_get (fd, this, &ctx);
+        if (__changelog_enabled (priv, type)) {
+                switch (local->op) {
 
-                if (_ret < 0) {
-                        goto out;
-                }
+                case GF_FOP_WRITE:
+                case GF_FOP_FTRUNCATE:
+                        op_ret = 1;
+                        break;
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+                case GF_FOP_FLUSH:
+                        op_ret = 0;
+                        break;
 
-                if (fd_ctx->pre_op_done[child_index]) {
+                default:
                         op_ret = 1;
                 }
-                fd_ctx->pre_op_done[child_index] = 0;
         }
-out:
-        UNLOCK (&fd->lock);
 
         return op_ret;
 }
 
-
-static int
-afr_pre_op_done_count (xlator_t *this, fd_t *fd, unsigned char *child_up)
+int
+afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending,
+                      int child, afr_xattrop_type_t op)
 {
         int i = 0;
-        int count = 0;
+        int ret = 0;
+
+        if (op == LOCAL_FIRST) {
+                ret = dict_set_static_bin (xattr, priv->pending_key[child],
+                                           pending[child],
+                                   AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+                if (ret)
+                        goto out;
+        }
+        for (i = 0; i < priv->child_count; i++) {
+                if (i == child)
+                        continue;
+                ret = dict_set_static_bin (xattr, priv->pending_key[i],
+                                           pending[i],
+                                   AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+                /* 3 = data+metadata+entry */
 
-        int _ret = 0;
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
+                if (ret < 0)
+                        goto out;
+        }
+        if (op == LOCAL_LAST) {
+                ret = dict_set_static_bin (xattr, priv->pending_key[child],
+                                           pending[child],
+                                   AFR_NUM_CHANGE_LOGS * sizeof (int32_t));
+                if (ret)
+                        goto out;
+        }
+out:
+        return ret;
+}
 
-        afr_private_t *priv = NULL;
+int
+afr_lock_server_count (afr_private_t *priv, afr_transaction_type type)
+{
+        int ret = 0;
 
-        priv = this->private;
+        switch (type) {
+        case AFR_DATA_TRANSACTION:
+                ret = priv->child_count;
+                break;
+
+        case AFR_METADATA_TRANSACTION:
+                ret = priv->child_count;
+                break;
+
+        case AFR_ENTRY_TRANSACTION:
+        case AFR_ENTRY_RENAME_TRANSACTION:
+                ret = priv->child_count;
+                break;
+        }
+
+        return ret;
+}
+
+/* {{{ pending */
+
+int32_t
+afr_changelog_post_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int32_t op_ret, int32_t op_errno, dict_t *xattr,
+                           dict_t *xdata)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        afr_private_t       *priv     = NULL;
+        afr_local_t         *local    = NULL;
+        int                  call_count = -1;
+
+        priv     = this->private;
+        local    = frame->local;
+        int_lock = &local->internal_lock;
 
-        LOCK (&fd->lock);
+        LOCK (&frame->lock);
         {
-                _ret = __fd_ctx_get (fd, this, &ctx);
+                call_count = --local->call_count;
+        }
+        UNLOCK (&frame->lock);
 
-                if (_ret < 0) {
-                        goto out;
+        if (call_count == 0) {
+                if (local->transaction.resume_stub) {
+			call_resume (local->transaction.resume_stub);
+                        local->transaction.resume_stub = NULL;
                 }
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                for (i = 0; i < priv->child_count; i++) {
-                        if (fd_ctx->pre_op_done[i] && child_up[i]) {
-                                count++;
-                        }
+                if (afr_lock_server_count (priv, local->transaction.type) == 0) {
+                        local->transaction.done (frame, this);
+                } else {
+                        int_lock->lock_cbk = local->transaction.done;
+                        afr_unlock (frame, this);
                 }
         }
-out:
-        UNLOCK (&fd->lock);
 
-        return count;
+        return 0;
 }
 
 
-static int
-__changelog_enabled (afr_private_t *priv, afr_transaction_type type)
+void
+afr_transaction_rm_stale_children (call_frame_t *frame, xlator_t *this,
+                                   inode_t *inode, afr_transaction_type type)
 {
-	int ret = 0;
-
-	switch (type) {
-	case AFR_DATA_TRANSACTION:
-		if (priv->data_change_log)
-			ret = 1;
-		
-		break;
-
-	case AFR_METADATA_TRANSACTION:
-		if (priv->metadata_change_log)
-			ret = 1;
-
-		break;
-
-	case AFR_ENTRY_TRANSACTION:
-	case AFR_ENTRY_RENAME_TRANSACTION:
-		if (priv->entry_change_log)
-			ret = 1;
-
-		break;
-		
-	case AFR_FLUSH_TRANSACTION:
-		ret = 1;
-	}
+        int             i = -1;
+        int             count = 0;
+        int             read_child = -1;
+        afr_private_t   *priv = NULL;
+        afr_local_t     *local = NULL;
+        int             **pending = NULL;
+        int             idx = 0;
+        int32_t         *stale_children = NULL;
+        int32_t         *fresh_children = NULL;
+        gf_boolean_t    rm_stale_children = _gf_false;
+
+        idx = afr_index_for_transaction_type (type);
 
-	return ret;
-}
+        priv = this->private;
+        local = frame->local;
+        pending = local->pending;
 
+        if (local->op_ret < 0)
+                goto out;
+        fresh_children = local->fresh_children;
+        read_child = afr_inode_get_read_ctx (this, inode, fresh_children);
+        if (read_child < 0) {
+                gf_log (this->name, GF_LOG_DEBUG, "Possible split-brain "
+                        "for %s", uuid_utoa (inode->gfid));
+                goto out;
+        }
 
-static int
-__changelog_needed_pre_op (call_frame_t *frame, xlator_t *this)
-{
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	fd_t *          fd    = NULL;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!afr_is_child_present (fresh_children,
+                                           priv->child_count, i))
+                        continue;
+                if (pending[i][idx])
+                        continue;
+                /* child is down or op failed on it */
+                if (!stale_children)
+                        stale_children = afr_children_create (priv->child_count);
+                if (!stale_children)
+                        goto out;
 
-	int op_ret   = 0;
+                rm_stale_children = _gf_true;
+                stale_children[count++] = i;
+                gf_log (this->name, GF_LOG_DEBUG, "Removing stale child "
+                        "%d for %s", i, uuid_utoa (inode->gfid));
+        }
 
-	priv  = this->private;
-	local = frame->local;
-	
-	if (__changelog_enabled (priv, local->transaction.type)) {
-		switch (local->op) {
-
-		case GF_FOP_WRITE:
-		case GF_FOP_FTRUNCATE:
-			/* 
-			   if it's a data transaction, we write the changelog
-			   only on the first write on an fd 
-			*/
-			
-			fd = local->fd;
-			if (!fd || __is_first_write_on_fd (this, fd))
-				op_ret = 1;
-
-			break;
-
-		case GF_FOP_FLUSH:
-			/* only do post-op on flush() */
-
-			op_ret = 0;
-			break;
-
-		default:
-			op_ret = 1;
-		}
-	}
+        if (!rm_stale_children)
+                goto out;
 
-	return op_ret;
+        afr_inode_rm_stale_children (this, inode, stale_children);
+out:
+        GF_FREE (stale_children);
+        return;
 }
 
+afr_inodelk_t*
+afr_get_inodelk (afr_internal_lock_t *int_lock, char *dom)
+{
+        afr_inodelk_t *inodelk = NULL;
+        int           i = 0;
 
-static int
-__changelog_needed_post_op (call_frame_t *frame, xlator_t *this)
+        for (i = 0; int_lock->inodelk[i].domain; i++) {
+                inodelk = &int_lock->inodelk[i];
+                if (strcmp (dom, inodelk->domain) == 0)
+                        return inodelk;
+        }
+        return NULL;
+}
+
+unsigned char*
+afr_locked_nodes_get (afr_transaction_type type, afr_internal_lock_t *int_lock)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
+        unsigned char *locked_nodes = NULL;
+        afr_inodelk_t *inodelk = NULL;
+        switch (type) {
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
+                inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+                locked_nodes = inodelk->locked_nodes;
+        break;
+
+        case AFR_ENTRY_TRANSACTION:
+        case AFR_ENTRY_RENAME_TRANSACTION:
+                /*Because same set of subvols participate in all lockee
+                 * entities*/
+                locked_nodes = int_lock->lockee[0].locked_nodes;
+        break;
+        }
+        return locked_nodes;
+}
 
-	int op_ret = 0;
-	afr_transaction_type type = -1;
+int
+afr_changelog_pre_op_call_count (afr_transaction_type type,
+                                 afr_internal_lock_t *int_lock,
+                                 unsigned int child_count)
+{
+        int           call_count = 0;
+        unsigned char *locked_nodes = NULL;
 
-	priv  = this->private;
-	local = frame->local;
-	type  = local->transaction.type;
+        locked_nodes = afr_locked_nodes_get (type, int_lock);
+        GF_ASSERT (locked_nodes);
 
-	if (__changelog_enabled (priv, type)) {
-                switch (local->op) {
+        call_count = afr_locked_children_count (locked_nodes, child_count);
+        if (type == AFR_ENTRY_RENAME_TRANSACTION)
+                call_count *= 2;
 
-                case GF_FOP_WRITE:
-                case GF_FOP_FTRUNCATE:
-                        op_ret = 0;
-                        break;
+        return call_count;
+}
 
-                case GF_FOP_FLUSH:
-                        op_ret = 1;
-                        break;
+int
+afr_changelog_post_op_call_count (afr_transaction_type type,
+                                  unsigned char *pre_op,
+                                  unsigned int child_count)
+{
+        int           call_count = 0;
 
-                default:
-                        op_ret = 1;
-                }
-        }
+        call_count = afr_pre_op_done_children_count (pre_op, child_count);
+        if (type == AFR_ENTRY_RENAME_TRANSACTION)
+                call_count *= 2;
 
-	return op_ret;
+        return call_count;
 }
 
-
-static int
-afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending)
+void
+afr_compute_txn_changelog (afr_local_t *local, afr_private_t *priv)
 {
-        int i;
-        int ret = 0;
-
+        int             i = 0;
+        int             index = 0;
+        int32_t         postop = 0;
+        int32_t         preop = 1;
+        int32_t         **txn_changelog = NULL;
+
+        txn_changelog = local->transaction.txn_changelog;
+        index = afr_index_for_transaction_type (local->transaction.type);
         for (i = 0; i < priv->child_count; i++) {
-                ret = dict_set_static_bin (xattr, priv->pending_key[i],
-                                           pending[i], 3 * sizeof (int32_t));
-                /* 3 = data+metadata+entry */
-                
-                if (ret < 0)
-                        goto out;
+                postop = ntoh32 (local->pending[i][index]);
+                txn_changelog[i][index] = hton32 (postop + preop);
         }
-
-out:
-        return ret;
 }
 
+afr_xattrop_type_t
+afr_get_postop_xattrop_type (int32_t **pending, int optimized, int child,
+                             afr_transaction_type type)
+{
+        int                     index = 0;
+        afr_xattrop_type_t      op = LOCAL_LAST;
 
-int
-afr_lock_server_count (afr_private_t *priv, afr_transaction_type type)
+        index = afr_index_for_transaction_type (type);
+        if (optimized && !pending[child][index])
+                op = LOCAL_FIRST;
+        return op;
+}
+
+void
+afr_set_postop_dict (afr_local_t *local, xlator_t *this, dict_t *xattr,
+                     int optimized, int child)
 {
-	int ret = 0;
-
-	switch (type) {
-	case AFR_FLUSH_TRANSACTION:
-	case AFR_DATA_TRANSACTION:
-		ret = priv->child_count;
-		break;
-
-	case AFR_METADATA_TRANSACTION:
-		ret = priv->child_count;
-		break;
-
-	case AFR_ENTRY_TRANSACTION:
-	case AFR_ENTRY_RENAME_TRANSACTION:
-		ret = priv->child_count;
-		break;
-	}
+        int32_t                 **txn_changelog = NULL;
+        int32_t                 **changelog = NULL;
+        afr_private_t           *priv = NULL;
+        int                     ret = 0;
+        afr_xattrop_type_t      op = LOCAL_LAST;
 
-	return ret;
+        priv = this->private;
+        txn_changelog = local->transaction.txn_changelog;
+        op = afr_get_postop_xattrop_type (local->pending, optimized, child,
+                                          local->transaction.type);
+        if (optimized)
+                changelog = txn_changelog;
+        else
+                changelog = local->pending;
+        ret = afr_set_pending_dict (priv, xattr, changelog, child, op);
+        if (ret < 0)
+                gf_log (this->name, GF_LOG_INFO,
+                        "failed to set pending entry");
 }
 
-/* {{{ pending */
 
-int32_t
-afr_changelog_post_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			   int32_t op_ret, int32_t op_errno, dict_t *xattr)
+gf_boolean_t
+afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this)
 {
-        afr_internal_lock_t *int_lock = NULL;
-	afr_private_t       *priv     = NULL;
-	afr_local_t         *local    = NULL;
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        int index = -1;
+        int i = 0;
 
-	int call_count = -1;
+        local = frame->local;
+        priv = this->private;
 
-        int (*post_post_op) (call_frame_t *, xlator_t *);
+        index = afr_index_for_transaction_type (local->transaction.type);
 
-	priv     = this->private;
-	local    = frame->local;
-        int_lock = &local->internal_lock;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->pending[i][index] == 0)
+                        return _gf_false;
+        }
 
-	LOCK (&frame->lock);
-	{
-		call_count = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
+        return _gf_true;
+}
 
-	if (call_count == 0) {
-                if (local->transaction.post_post_op) {
-                        post_post_op = local->transaction.post_post_op;
+static void
+afr_dir_fop_handle_all_fop_failures (call_frame_t *frame)
+{
+        xlator_t        *this = NULL;
+        afr_local_t     *local = NULL;
+        afr_private_t   *priv = NULL;
 
-                        if (afr_lock_server_count (priv, local->transaction.type) == 0) {
-                                local->transaction.post_post_op = local->transaction.done;
-                        } else {
-                                int_lock->lock_cbk = local->transaction.done;
-                                local->transaction.post_post_op = afr_unlock;
-                        }
+        this = frame->this;
+        local = frame->local;
+        priv = this->private;
 
-                        post_post_op (frame, this);
-                } else {
-                        if (afr_lock_server_count (priv, local->transaction.type) == 0) {
-                                local->transaction.done (frame, this);
-                        } else {
-                                int_lock->lock_cbk = local->transaction.done;
-                                afr_unlock (frame, this);
-                        }
-                }
-	}
+        if ((local->transaction.type != AFR_ENTRY_TRANSACTION) &&
+            (local->transaction.type != AFR_ENTRY_RENAME_TRANSACTION))
+                return;
+
+        if (local->op_ret >= 0)
+                goto out;
 
-	return 0;	
+        __mark_all_success (local->pending, priv->child_count,
+                            local->transaction.type);
+out:
+        return;
 }
 
+static void
+afr_data_handle_quota_errors (call_frame_t *frame, xlator_t *this)
+{
+        int     i = 0;
+        afr_private_t *priv = NULL;
+        afr_local_t   *local = NULL;
+        gf_boolean_t  all_quota_failures = _gf_false;
 
-int 
-afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
+        local = frame->local;
+        priv  = this->private;
+        if (local->transaction.type != AFR_DATA_TRANSACTION)
+                return;
+        /*
+         * Idea is to not leave the file in FOOL-FOOL scenario in case on
+         * all the bricks data transaction failed with EDQUOT to avoid
+         * increasing un-necessary load of self-heals in the system.
+         */
+        all_quota_failures = _gf_true;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i] &&
+                    (local->child_errno[i] != EDQUOT)) {
+                        all_quota_failures = _gf_false;
+                        break;
+                }
+        }
+        if (all_quota_failures)
+                __mark_all_success (local->pending, priv->child_count,
+                                    local->transaction.type);
+}
+
+int
+afr_changelog_post_op_now (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t * priv = this->private;
+        afr_private_t * priv = this->private;
         afr_internal_lock_t *int_lock = NULL;
-	int ret        = 0;
-	int i          = 0;				
-	int call_count = 0;
-	
-	afr_local_t *  local = NULL;	
-	dict_t        **xattr = NULL;
-
-	local    = frame->local;
+        int i          = 0;
+        int call_count = 0;
+
+        afr_local_t *  local = NULL;
+        afr_fd_ctx_t  *fdctx = NULL;
+        dict_t        **xattr = NULL;
+        int            piggyback = 0;
+        int            nothing_failed = 1;
+
+        local    = frame->local;
         int_lock = &local->internal_lock;
 
-	__mark_down_children (local->pending, priv->child_count, 
-                              local->child_up, local->transaction.type);
-        
-        if (local->op == GF_FOP_FLUSH) {
-                __mark_failed_children (local->pending, priv->child_count,
-                                        this, local->fd,
-                                        local->transaction.type);
-        }
+        __mark_non_participant_children (local->pending, priv->child_count,
+                                         local->transaction.pre_op,
+                                         local->transaction.type);
+
+        afr_data_handle_quota_errors (frame, this);
+        afr_dir_fop_handle_all_fop_failures (frame);
+
+        if (local->fd)
+                afr_transaction_rm_stale_children (frame, this,
+                                                   local->fd->inode,
+                                                   local->transaction.type);
 
         xattr = alloca (priv->child_count * sizeof (*xattr));
         memset (xattr, 0, (priv->child_count * sizeof (*xattr)));
-	for (i = 0; i < priv->child_count; i++) {
-                xattr[i] = get_new_dict ();
-                dict_ref (xattr[i]);
+        for (i = 0; i < priv->child_count; i++) {
+                xattr[i] = dict_new ();
         }
 
-        if (local->op == GF_FOP_FLUSH) {
-                call_count = afr_pre_op_done_count (this, local->fd, local->child_up);
-        } else {
-                call_count = afr_up_children_count (priv->child_count, local->child_up); 
+        call_count = afr_changelog_post_op_call_count (local->transaction.type,
+                                                       local->transaction.pre_op,
+                                                       priv->child_count);
+        local->call_count = call_count;
 
-                if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
-                        call_count *= 2;
-                }
+        if (local->fd)
+                fdctx = afr_fd_ctx_get (local->fd, this);
+
+        if (call_count == 0) {
+                /* no child is up */
+                int_lock->lock_cbk = local->transaction.done;
+                afr_unlock (frame, this);
+                goto out;
         }
 
-	local->call_count = call_count;		
+        nothing_failed = afr_txn_nothing_failed (frame, this);
 
-	if (call_count == 0) {
-		/* no child is up */
-                for (i = 0; i < priv->child_count; i++) {
-                        dict_unref (xattr[i]);
+        afr_compute_txn_changelog (local , priv);
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (!local->transaction.pre_op[i])
+                        continue;
+
+                if (local->transaction.type != AFR_DATA_TRANSACTION)
+                        afr_set_postop_dict (local, this, xattr[i],
+                                             local->optimistic_change_log, i);
+                switch (local->transaction.type) {
+                case AFR_DATA_TRANSACTION:
+                {
+                        if (!fdctx) {
+                                afr_set_postop_dict (local, this, xattr[i],
+                                                     0, i);
+                                STACK_WIND (frame, afr_changelog_post_op_cbk,
+                                            priv->children[i],
+                                            priv->children[i]->fops->xattrop,
+                                            &local->loc,
+                                            GF_XATTROP_ADD_ARRAY, xattr[i],
+                                            NULL);
+                                break;
+                        }
+
+                        /* local->transaction.postop_piggybacked[] was
+                           precomputed in is_piggyback_postop() when called from
+                           afr_changelog_post_op_safe()
+                        */
+
+                        piggyback = 0;
+                        if (local->transaction.postop_piggybacked[i])
+                                piggyback = 1;
+
+                        afr_set_postop_dict (local, this, xattr[i],
+                                             piggyback, i);
+
+                        if (nothing_failed && piggyback) {
+                                afr_changelog_post_op_cbk (frame, (void *)(long)i,
+                                                           this, 1, 0, xattr[i], NULL);
+                        } else {
+                                STACK_WIND_COOKIE (frame,
+                                                   afr_changelog_post_op_cbk,
+                                                   (void *) (long) i,
+                                                   priv->children[i],
+                                                   priv->children[i]->fops->fxattrop,
+                                                   local->fd,
+                                                   GF_XATTROP_ADD_ARRAY, xattr[i],
+                                                   NULL);
+                        }
                 }
+                break;
+                case AFR_METADATA_TRANSACTION:
+                {
+                        if (nothing_failed && local->optimistic_change_log) {
+                                afr_changelog_post_op_cbk (frame, (void *)(long)i,
+                                                           this, 1, 0, xattr[i],
+                                                           NULL);
+                                break;
+                        }
 
-                int_lock->lock_cbk = local->transaction.done;
-		afr_unlock (frame, this);
-		return 0;
-	}
+                        if (local->fd)
+                                STACK_WIND (frame, afr_changelog_post_op_cbk,
+                                            priv->children[i],
+                                            priv->children[i]->fops->fxattrop,
+                                            local->fd,
+                                            GF_XATTROP_ADD_ARRAY, xattr[i],
+                                            NULL);
+                        else
+                                STACK_WIND (frame, afr_changelog_post_op_cbk,
+                                            priv->children[i],
+                                            priv->children[i]->fops->xattrop,
+                                            &local->loc,
+                                            GF_XATTROP_ADD_ARRAY, xattr[i],
+                                            NULL);
+                }
+                break;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-                        ret = afr_set_pending_dict (priv, xattr[i], 
-                                                    local->pending);
-
-			if (ret < 0)
-				gf_log (this->name, GF_LOG_DEBUG,
-					"failed to set pending entry");
-
-
-			switch (local->transaction.type) {
-			case AFR_DATA_TRANSACTION:
-			case AFR_METADATA_TRANSACTION:
-			{
-				if (local->fd)
-					STACK_WIND (frame, afr_changelog_post_op_cbk,
-						    priv->children[i],
-						    priv->children[i]->fops->fxattrop,
-						    local->fd,
-						    GF_XATTROP_ADD_ARRAY, xattr[i]);
-				else
-					STACK_WIND (frame, afr_changelog_post_op_cbk,
-						    priv->children[i],
-						    priv->children[i]->fops->xattrop,
-						    &local->loc,
-						    GF_XATTROP_ADD_ARRAY, xattr[i]);
-                                call_count--;
-			}
-			break;
-
-			case AFR_FLUSH_TRANSACTION:
-			{
-				if (__if_fd_pre_op_done (this, local->fd, i)) {
-					STACK_WIND (frame, afr_changelog_post_op_cbk,
-						    priv->children[i], 
-						    priv->children[i]->fops->fxattrop,
-						    local->fd,
-						    GF_XATTROP_ADD_ARRAY, xattr[i]);
-                                        call_count--;
-                                }
-			}
-			break;
-
-			case AFR_ENTRY_RENAME_TRANSACTION:
-			{
-				STACK_WIND_COOKIE (frame, afr_changelog_post_op_cbk,
-						   (void *) (long) i,
-						   priv->children[i],
-						   priv->children[i]->fops->xattrop,
-						   &local->transaction.new_parent_loc,
-						   GF_XATTROP_ADD_ARRAY, xattr[i]);
-				
-				call_count--;
-			}
-
-			/* 
-			   set it again because previous stack_wind
-			   might have already returned (think of case
-			   where subvolume is posix) and would have
-			   used the dict as placeholder for return
-			   value
-			*/
-                        
-			ret = afr_set_pending_dict (priv, xattr[i], 
-                                                    local->pending);
-
-			if (ret < 0)
-				gf_log (this->name, GF_LOG_DEBUG,
-					"failed to set pending entry");
-
-			/* fall through */
-
-			case AFR_ENTRY_TRANSACTION:
-			{
-				if (local->fd)
-					STACK_WIND (frame, afr_changelog_post_op_cbk,
-						    priv->children[i], 
-						    priv->children[i]->fops->fxattrop,
-						    local->fd, 
-						    GF_XATTROP_ADD_ARRAY, xattr[i]);
-				else 
-					STACK_WIND (frame, afr_changelog_post_op_cbk,
-						    priv->children[i], 
-						    priv->children[i]->fops->xattrop,
-						    &local->transaction.parent_loc, 
-						    GF_XATTROP_ADD_ARRAY, xattr[i]);
-                                call_count--;
-			}
-			break;
-			}
-
-			if (!call_count)
-				break;
-		}
-	}
+                case AFR_ENTRY_RENAME_TRANSACTION:
+                {
+                        if (nothing_failed && local->optimistic_change_log) {
+                                afr_changelog_post_op_cbk (frame, (void *)(long)i,
+                                                           this, 1, 0, xattr[i],
+                                                           NULL);
+                        } else {
+                                STACK_WIND_COOKIE (frame, afr_changelog_post_op_cbk,
+                                                   (void *) (long) i,
+                                                   priv->children[i],
+                                                   priv->children[i]->fops->xattrop,
+                                                   &local->transaction.new_parent_loc,
+                                                   GF_XATTROP_ADD_ARRAY, xattr[i],
+                                                   NULL);
+                        }
+                        call_count--;
+                }
+
+                /*
+                  set it again because previous stack_wind
+                  might have already returned (think of case
+                  where subvolume is posix) and would have
+                  used the dict as placeholder for return
+                  value
+                */
+
+                afr_set_postop_dict (local, this, xattr[i],
+                                     local->optimistic_change_log, i);
+
+                /* fall through */
+
+                case AFR_ENTRY_TRANSACTION:
+                {
+                        if (nothing_failed && local->optimistic_change_log) {
+                                afr_changelog_post_op_cbk (frame, (void *)(long)i,
+                                                           this, 1, 0, xattr[i],
+                                                           NULL);
+                                break;
+                        }
+
+                        if (local->fd)
+                                STACK_WIND (frame, afr_changelog_post_op_cbk,
+                                            priv->children[i],
+                                            priv->children[i]->fops->fxattrop,
+                                            local->fd,
+                                            GF_XATTROP_ADD_ARRAY, xattr[i],
+                                            NULL);
+                        else
+                                STACK_WIND (frame, afr_changelog_post_op_cbk,
+                                            priv->children[i],
+                                            priv->children[i]->fops->xattrop,
+                                            &local->transaction.parent_loc,
+                                            GF_XATTROP_ADD_ARRAY, xattr[i],
+                                            NULL);
+                }
+                break;
+                }
+
+                if (!--call_count)
+                        break;
+        }
 
+out:
         for (i = 0; i < priv->child_count; i++) {
                 dict_unref (xattr[i]);
         }
 
-	return 0;
+        return 0;
 }
 
 
 int32_t
 afr_changelog_pre_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			      int32_t op_ret, int32_t op_errno, dict_t *xattr)
+                          int32_t op_ret, int32_t op_errno, dict_t *xattr,
+                          dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = this->private;
-	loc_t       *   loc   = NULL;
+        afr_local_t *   local = NULL;
+        afr_private_t * priv  = this->private;
+        int call_count  = -1;
+        int child_index = (long) cookie;
 
-	int call_count  = -1;
-	int child_index = (long) cookie;
-
-	local = frame->local;
-	loc   = &local->loc;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-                if (op_ret == 0) {
+        LOCK (&frame->lock);
+        {
+                switch (op_ret) {
+                case 0:
                         __mark_pre_op_done_on_fd (frame, this, child_index);
+                        //fallthrough we need to mark the pre_op
+                case 1:
+                        local->transaction.pre_op[child_index] = 1;
+                        /* special op_ret for piggyback */
+                        break;
+                case -1:
+                        if (op_errno == ENOTSUP) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "xattrop not supported by %s",
+                                        priv->children[child_index]->name);
+                                local->op_ret = -1;
+
+                        } else if (!child_went_down (op_ret, op_errno)) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "xattrop failed on child %s: %s",
+                                        priv->children[child_index]->name,
+                                        strerror (op_errno));
+                        }
+                        local->op_errno = op_errno;
+                        break;
                 }
 
-		if (op_ret == -1) {
-			local->child_up[child_index] = 0;
-			
-			if (op_errno == ENOTSUP) {
-				gf_log (this->name, GF_LOG_ERROR,
-					"xattrop not supported by %s",
-					priv->children[child_index]->name);
-				local->op_ret = -1;
-                                
-			} else if (!child_went_down (op_ret, op_errno)) {
-				gf_log (this->name, GF_LOG_ERROR,
-					"xattrop failed on child %s: %s",
-					priv->children[child_index]->name, 
-					strerror (op_errno));
-			}
-			local->op_errno = op_errno;
-		}
-
-		call_count = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-
-	if (call_count == 0) {
-		if ((local->op_ret == -1) && 
-		    (local->op_errno == ENOTSUP)) {
-			local->transaction.resume (frame, this);
-		} else {
-                        __mark_all_success (local->pending, priv->child_count,
-                                            local->transaction.type);
-
-                        afr_pid_restore (frame);
+                call_count = --local->call_count;
+        }
+        UNLOCK (&frame->lock);
 
-			local->transaction.fop (frame, this);
-		}
-	}
+        if (call_count == 0) {
+                if ((local->op_ret == -1) &&
+                    (local->op_errno == ENOTSUP)) {
+                        local->transaction.resume (frame, this);
+                } else {
+                        afr_transaction_perform_fop (frame, this);
+                }
+        }
 
-	return 0;	
+        return 0;
 }
 
-
-int 
+int
 afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t * priv = this->private;
-	int i = 0;				
-	int ret = 0;
-	int call_count = 0;		     
-	dict_t **xattr = NULL;
+        afr_private_t * priv = this->private;
+        int i = 0;
+        int ret = 0;
+        int call_count = 0;
+        dict_t **xattr = NULL;
+        afr_fd_ctx_t *fdctx = NULL;
+        afr_local_t *local = NULL;
+        int          piggyback = 0;
+        afr_internal_lock_t *int_lock = NULL;
+        unsigned char       *locked_nodes = NULL;
 
-	afr_local_t *local = NULL;
+        local = frame->local;
+        int_lock = &local->internal_lock;
 
-	local = frame->local;
-        
         xattr = alloca (priv->child_count * sizeof (*xattr));
         memset (xattr, 0, (priv->child_count * sizeof (*xattr)));
 
-	for (i = 0; i < priv->child_count; i++) {
-                xattr[i] = get_new_dict ();
-                dict_ref (xattr[i]);
+        for (i = 0; i < priv->child_count; i++) {
+                xattr[i] = dict_new ();
+        }
+
+        call_count = afr_changelog_pre_op_call_count (local->transaction.type,
+                                                      int_lock,
+                                                      priv->child_count);
+        if (call_count == 0) {
+                local->internal_lock.lock_cbk =
+                        local->transaction.done;
+                afr_unlock (frame, this);
+                goto out;
         }
 
-	call_count = afr_up_children_count (priv->child_count, 
-					    local->child_up); 
+        local->call_count = call_count;
 
-	if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
-		call_count *= 2;
-	}
+        __mark_all_pending (local->pending, priv->child_count,
+                            local->transaction.type);
 
-	if (call_count == 0) {
-		/* no child is up */
-                for (i = 0; i < priv->child_count; i++) {
-                        dict_unref (xattr[i]);
+        if (local->fd)
+                fdctx = afr_fd_ctx_get (local->fd, this);
+
+        locked_nodes = afr_locked_nodes_get (local->transaction.type, int_lock);
+        for (i = 0; i < priv->child_count; i++) {
+                if (!locked_nodes[i])
+                        continue;
+                ret = afr_set_pending_dict (priv, xattr[i], local->pending,
+                                            i, LOCAL_FIRST);
+
+                if (ret < 0)
+                        gf_log (this->name, GF_LOG_INFO,
+                                "failed to set pending entry");
+
+
+                switch (local->transaction.type) {
+                case AFR_DATA_TRANSACTION:
+                {
+                        if (!fdctx) {
+                                STACK_WIND_COOKIE (frame,
+                                                   afr_changelog_pre_op_cbk,
+                                                   (void *) (long) i,
+                                                   priv->children[i],
+                                                   priv->children[i]->fops->xattrop,
+                                                   &(local->loc),
+                                                   GF_XATTROP_ADD_ARRAY, xattr[i],
+                                                   NULL);
+                                break;
+                        }
+
+                        LOCK (&local->fd->lock);
+                        {
+                                piggyback = 0;
+                                if (fdctx->pre_op_done[i]) {
+                                        fdctx->pre_op_piggyback[i]++;
+                                        piggyback = 1;
+                                        fdctx->hit++;
+                                } else {
+                                        fdctx->miss++;
+                                }
+                        }
+                        UNLOCK (&local->fd->lock);
+
+                        afr_set_delayed_post_op (frame, this);
+
+                        if (piggyback)
+                                afr_changelog_pre_op_cbk (frame, (void *)(long)i,
+                                                          this, 1, 0, xattr[i],
+                                                          NULL);
+                        else
+                                STACK_WIND_COOKIE (frame,
+                                                   afr_changelog_pre_op_cbk,
+                                                   (void *) (long) i,
+                                                   priv->children[i],
+                                                   priv->children[i]->fops->fxattrop,
+                                                   local->fd,
+                                                   GF_XATTROP_ADD_ARRAY, xattr[i],
+                                                   NULL);
                 }
+                break;
+                case AFR_METADATA_TRANSACTION:
+                {
+                        if (local->optimistic_change_log) {
+                                afr_changelog_pre_op_cbk (frame, (void *)(long)i,
+                                                          this, 1, 0, xattr[i],
+                                                          NULL);
+                                break;
+                        }
 
-                local->internal_lock.lock_cbk =
-                        local->transaction.done;
-		afr_unlock (frame, this);
-		return 0;
-	}
+                        if (local->fd)
+                                STACK_WIND_COOKIE (frame,
+                                                   afr_changelog_pre_op_cbk,
+                                                   (void *) (long) i,
+                                                   priv->children[i],
+                                                   priv->children[i]->fops->fxattrop,
+                                                   local->fd,
+                                                   GF_XATTROP_ADD_ARRAY, xattr[i],
+                                                   NULL);
+                        else
+                                STACK_WIND_COOKIE (frame,
+                                                   afr_changelog_pre_op_cbk,
+                                                   (void *) (long) i,
+                                                   priv->children[i],
+                                                   priv->children[i]->fops->xattrop,
+                                                   &(local->loc),
+                                                   GF_XATTROP_ADD_ARRAY, xattr[i],
+                                                   NULL);
+                }
+                break;
 
-	local->call_count = call_count;		
+                case AFR_ENTRY_RENAME_TRANSACTION:
+                {
+                        if (local->optimistic_change_log) {
+                                afr_changelog_pre_op_cbk (frame, (void *)(long)i,
+                                                          this, 1, 0, xattr[i],
+                                                          NULL);
+                        } else {
+                                STACK_WIND_COOKIE (frame,
+                                                   afr_changelog_pre_op_cbk,
+                                                   (void *) (long) i,
+                                                   priv->children[i],
+                                                   priv->children[i]->fops->xattrop,
+                                                   &local->transaction.new_parent_loc,
+                                                   GF_XATTROP_ADD_ARRAY, xattr[i],
+                                                   NULL);
+                        }
 
-	__mark_all_pending (local->pending, priv->child_count,
-                            local->transaction.type);
+                        call_count--;
+                }
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			ret = afr_set_pending_dict (priv, xattr[i], 
-                                                    local->pending);
-
-			if (ret < 0)
-				gf_log (this->name, GF_LOG_DEBUG,
-					"failed to set pending entry");
-
-
-			switch (local->transaction.type) {
-			case AFR_DATA_TRANSACTION:
-			case AFR_METADATA_TRANSACTION:
-			case AFR_FLUSH_TRANSACTION:
-			{
-				if (local->fd)
-					STACK_WIND_COOKIE (frame, 
-							   afr_changelog_pre_op_cbk,
-							   (void *) (long) i,
-							   priv->children[i], 
-							   priv->children[i]->fops->fxattrop,
-							   local->fd,
-							   GF_XATTROP_ADD_ARRAY, xattr[i]);
-				else
-					STACK_WIND_COOKIE (frame, 
-							   afr_changelog_pre_op_cbk,
-							   (void *) (long) i,
-							   priv->children[i], 
-							   priv->children[i]->fops->xattrop,
-							   &(local->loc), 
-							   GF_XATTROP_ADD_ARRAY, xattr[i]);
-			}
-			break;
-				
-			case AFR_ENTRY_RENAME_TRANSACTION: 
-			{
-				STACK_WIND_COOKIE (frame, 
-						   afr_changelog_pre_op_cbk,
-						   (void *) (long) i,
-						   priv->children[i], 
-						   priv->children[i]->fops->xattrop,
-						   &local->transaction.new_parent_loc, 
-						   GF_XATTROP_ADD_ARRAY, xattr[i]);
-
-				call_count--;
-			}
-
-
-			/* 
-			   set it again because previous stack_wind
-			   might have already returned (think of case
-			   where subvolume is posix) and would have
-			   used the dict as placeholder for return
-			   value
-			*/
-
-			ret = afr_set_pending_dict (priv, xattr[i], 
-                                                    local->pending);
-
-			if (ret < 0)
-				gf_log (this->name, GF_LOG_DEBUG,
-					"failed to set pending entry");
-
-			/* fall through */
-				
-			case AFR_ENTRY_TRANSACTION:
-			{
-				if (local->fd)
-					STACK_WIND_COOKIE (frame, 
-							   afr_changelog_pre_op_cbk,
-							   (void *) (long) i,
-							   priv->children[i], 
-							   priv->children[i]->fops->fxattrop,
-							   local->fd, 
-							   GF_XATTROP_ADD_ARRAY, xattr[i]);
-				else
-					STACK_WIND_COOKIE (frame, 
-							   afr_changelog_pre_op_cbk,
-							   (void *) (long) i,
-							   priv->children[i], 
-							   priv->children[i]->fops->xattrop,
-							   &local->transaction.parent_loc, 
-							   GF_XATTROP_ADD_ARRAY, xattr[i]);
-			}
-
-			break;
-			}
-
-			if (!--call_count)
-				break;
-		}
-	}
-        
+
+                /*
+                  set it again because previous stack_wind
+                  might have already returned (think of case
+                  where subvolume is posix) and would have
+                  used the dict as placeholder for return
+                  value
+                */
+
+                ret = afr_set_pending_dict (priv, xattr[i], local->pending,
+                                            i, LOCAL_FIRST);
+
+                if (ret < 0)
+                        gf_log (this->name, GF_LOG_INFO,
+                                "failed to set pending entry");
+
+                /* fall through */
+
+                case AFR_ENTRY_TRANSACTION:
+                {
+                        if (local->optimistic_change_log) {
+                                afr_changelog_pre_op_cbk (frame, (void *)(long)i,
+                                                          this, 1, 0, xattr[i],
+                                                          NULL);
+                                break;
+                        }
+
+                        if (local->fd)
+                                STACK_WIND_COOKIE (frame,
+                                                   afr_changelog_pre_op_cbk,
+                                                   (void *) (long) i,
+                                                   priv->children[i],
+                                                   priv->children[i]->fops->fxattrop,
+                                                   local->fd,
+                                                   GF_XATTROP_ADD_ARRAY, xattr[i],
+                                                   NULL);
+                        else
+                                STACK_WIND_COOKIE (frame,
+                                                   afr_changelog_pre_op_cbk,
+                                                   (void *) (long) i,
+                                                   priv->children[i],
+                                                   priv->children[i]->fops->xattrop,
+                                                   &local->transaction.parent_loc,
+                                                   GF_XATTROP_ADD_ARRAY, xattr[i],
+                                                   NULL);
+                }
+                break;
+                }
+
+                if (!--call_count)
+                        break;
+        }
+out:
         for (i = 0; i < priv->child_count; i++) {
                 dict_unref (xattr[i]);
         }
-        
-	return 0;
+
+        return 0;
 }
 
+
 int
 afr_post_blocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
 {
@@ -878,9 +1115,9 @@ afr_post_blocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
 
         if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_INFO,
                         "Blocking inodelks failed.");
-		local->transaction.done (frame, this);
+                local->transaction.done (frame, this);
         } else {
 
                 gf_log (this->name, GF_LOG_DEBUG,
@@ -891,6 +1128,7 @@ afr_post_blocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
         return 0;
 }
 
+
 int
 afr_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
 {
@@ -916,6 +1154,7 @@ afr_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
         return 0;
 }
 
+
 int
 afr_post_blocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
 {
@@ -926,9 +1165,9 @@ afr_post_blocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
 
         if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_INFO,
                         "Blocking entrylks failed.");
-		local->transaction.done (frame, this);
+                local->transaction.done (frame, this);
         } else {
 
                 gf_log (this->name, GF_LOG_DEBUG,
@@ -939,6 +1178,7 @@ afr_post_blocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
         return 0;
 }
 
+
 int
 afr_post_nonblocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
 {
@@ -964,6 +1204,7 @@ afr_post_nonblocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
         return 0;
 }
 
+
 int
 afr_post_blocking_rename_cbk (call_frame_t *frame, xlator_t *this)
 {
@@ -974,9 +1215,9 @@ afr_post_blocking_rename_cbk (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
 
         if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_INFO,
                         "Blocking entrylks failed.");
-		local->transaction.done (frame, this);
+                local->transaction.done (frame, this);
         } else {
 
                 gf_log (this->name, GF_LOG_DEBUG,
@@ -986,7 +1227,9 @@ afr_post_blocking_rename_cbk (call_frame_t *frame, xlator_t *this)
         return 0;
 }
 
-int afr_post_lower_unlock_cbk (call_frame_t *frame, xlator_t *this)
+
+int
+afr_post_lower_unlock_cbk (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
@@ -1002,16 +1245,19 @@ int afr_post_lower_unlock_cbk (call_frame_t *frame, xlator_t *this)
         return 0;
 }
 
+
 int
 afr_set_transaction_flock (afr_local_t *local)
 {
         afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
 
         int_lock = &local->internal_lock;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
 
-        int_lock->lk_flock.l_len   = local->transaction.len;
-        int_lock->lk_flock.l_start = local->transaction.start;
-        int_lock->lk_flock.l_type  = F_WRLCK;
+        inodelk->flock.l_len   = local->transaction.len;
+        inodelk->flock.l_start = local->transaction.start;
+        inodelk->flock.l_type  = F_WRLCK;
 
         return 0;
 }
@@ -1026,25 +1272,25 @@ afr_lock_rec (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
 
         int_lock->transaction_lk_type = AFR_TRANSACTION_LK;
+        int_lock->domain = this->name;
 
-	switch (local->transaction.type) {
-	case AFR_DATA_TRANSACTION:
-	case AFR_METADATA_TRANSACTION:
-	case AFR_FLUSH_TRANSACTION:
+        switch (local->transaction.type) {
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
                 afr_set_transaction_flock (local);
 
                 int_lock->lock_cbk = afr_post_nonblocking_inodelk_cbk;
 
                 afr_nonblocking_inodelk (frame, this);
-		break;
+                break;
 
         case AFR_ENTRY_RENAME_TRANSACTION:
 
-                int_lock->lock_cbk = afr_post_blocking_rename_cbk;
-                afr_blocking_lock (frame, this);
+                int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
+                afr_nonblocking_entrylk (frame, this);
                 break;
 
-	case AFR_ENTRY_TRANSACTION:
+        case AFR_ENTRY_TRANSACTION:
                 int_lock->lk_basename = local->transaction.basename;
                 if (&local->transaction.parent_loc)
                         int_lock->lk_loc = &local->transaction.parent_loc;
@@ -1060,18 +1306,12 @@ afr_lock_rec (call_frame_t *frame, xlator_t *this)
 }
 
 
-int32_t
+int
 afr_lock (call_frame_t *frame, xlator_t *this)
 {
-        afr_pid_save (frame);
-
-        frame->root->pid = (long) frame->root;
-
-        afr_set_lk_owner (frame, this);
-
         afr_set_lock_number (frame, this);
 
-	return afr_lock_rec (frame, this);
+        return afr_lock_rec (frame, this);
 }
 
 
@@ -1080,49 +1320,498 @@ afr_lock (call_frame_t *frame, xlator_t *this)
 int
 afr_internal_lock_finish (call_frame_t *frame, xlator_t *this)
 {
-        afr_local_t   *local = NULL;
-        afr_private_t *priv  = NULL;
+        if (__fop_changelog_needed (frame, this)) {
+                afr_changelog_pre_op (frame, this);
+        } else {
+                afr_transaction_perform_fop (frame, this);
+        }
+
+        return 0;
+}
+
+
+void
+afr_set_delayed_post_op (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t    *local = NULL;
+        afr_private_t  *priv = NULL;
+
+        /* call this function from any of the related optimizations
+           which benefit from delaying post op are enabled, namely:
+
+           - changelog piggybacking
+           - eager locking
+        */
+
+        priv = this->private;
+        if (!priv)
+                return;
+
+        if (!priv->post_op_delay_secs)
+                return;
+
+        local = frame->local;
+        if (!local->transaction.eager_lock_on)
+                return;
+
+        if (!local)
+                return;
+
+        if (!local->fd)
+                return;
+
+        if (local->op == GF_FOP_WRITE)
+                local->delayed_post_op = _gf_true;
+}
+
+gf_boolean_t
+afr_are_multiple_fds_opened (inode_t *inode, xlator_t *this)
+{
+        afr_inode_ctx_t *ictx = NULL;
+
+        if (!inode) {
+                /* If false is returned, it may keep on taking eager-lock
+                 * which may lead to starvation, so return true to avoid that.
+                 */
+                gf_log_callingfn (this->name, GF_LOG_ERROR, "Invalid inode");
+                return _gf_true;
+        }
+        /* Lets say mount1 has eager-lock(full-lock) and after the eager-lock
+         * is taken mount2 opened the same file, it won't be able to
+         * perform any data operations until mount1 releases eager-lock.
+         * To avoid such scenario do not enable eager-lock for this transaction
+         * if open-fd-count is > 1
+         */
+
+        ictx = afr_inode_ctx_get (inode, this);
+        if (!ictx)
+                return _gf_true;
+
+        if (ictx->open_fd_count > 1)
+                return _gf_true;
+
+        return _gf_false;
+}
+
+gf_boolean_t
+afr_any_fops_failed (afr_local_t *local, afr_private_t *priv)
+{
+        if (local->success_count != priv->child_count)
+                return _gf_true;
+        return _gf_false;
+}
+
+gf_boolean_t
+is_afr_delayed_changelog_post_op_needed (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t      *local = NULL;
+        gf_boolean_t      res = _gf_false;
+        afr_private_t    *priv  = NULL;
 
         priv  = this->private;
+
         local = frame->local;
+        if (!local)
+                goto out;
 
-        if (__changelog_needed_pre_op (frame, this)) {
-                afr_changelog_pre_op (frame, this);
-        } else {
-                __mark_all_success (local->pending, priv->child_count,
-                                            local->transaction.type);
+        if (!local->delayed_post_op)
+                goto out;
+
+        //Mark pending changelog ASAP
+        if (afr_any_fops_failed (local, priv))
+                goto out;
+
+        if (local->fd && afr_are_multiple_fds_opened (local->fd->inode, this))
+                goto out;
+
+        res = _gf_true;
+out:
+        return res;
+}
+
+
+void
+afr_delayed_changelog_post_op (xlator_t *this, call_frame_t *frame, fd_t *fd,
+                               call_stub_t *stub);
+
+void
+afr_delayed_changelog_wake_up_cbk (void *data)
+{
+        fd_t           *fd = NULL;
+
+        fd = data;
+
+        afr_delayed_changelog_wake_up (THIS, fd);
+}
+
+
+/*
+  Check if the frame is destined to get optimized away
+  with changelog piggybacking
+*/
+static gf_boolean_t
+is_piggyback_post_op (call_frame_t *frame, fd_t *fd)
+{
+        afr_fd_ctx_t *fdctx = NULL;
+        afr_local_t *local = NULL;
+        gf_boolean_t piggyback = _gf_true;
+        afr_private_t *priv = NULL;
+        int i = 0;
+
+        priv = frame->this->private;
+        local = frame->local;
+        fdctx = afr_fd_ctx_get (fd, frame->this);
+
+        LOCK(&fd->lock);
+        {
+                piggyback = _gf_true;
+
+                for (i = 0; i < priv->child_count; i++) {
+                        if (!local->transaction.pre_op[i])
+                                continue;
+                        if (fdctx->pre_op_piggyback[i]) {
+                                fdctx->pre_op_piggyback[i]--;
+                                local->transaction.postop_piggybacked[i] = 1;
+                        } else {
+                                /* For at least _one_ subvolume we cannot
+                                   piggyback on the changelog, and have to
+                                   perform a hard POST-OP and therefore fsync
+                                   if necesssary
+                                */
+                                piggyback = _gf_false;
+                                GF_ASSERT (fdctx->pre_op_done[i]);
+                                fdctx->pre_op_done[i]--;
+                        }
+                }
+        }
+        UNLOCK(&fd->lock);
+
+        if (!afr_txn_nothing_failed (frame, frame->this)) {
+                /* something failed in this transaction,
+                   we will be performing a hard post-op
+                */
+                return _gf_false;
+        }
+
+        return piggyback;
+}
+
+
+/* SET operation */
+int
+afr_fd_report_unstable_write (xlator_t *this, fd_t *fd)
+{
+        afr_fd_ctx_t *fdctx = NULL;
 
-                afr_pid_restore (frame);
+        fdctx = afr_fd_ctx_get (fd, this);
 
-                local->transaction.fop (frame, this);
+        LOCK(&fd->lock);
+        {
+                fdctx->witnessed_unstable_write = _gf_true;
         }
+        UNLOCK(&fd->lock);
 
         return 0;
 }
 
-int32_t
+/* TEST and CLEAR operation */
+gf_boolean_t
+afr_fd_has_witnessed_unstable_write (xlator_t *this, fd_t *fd)
+{
+        afr_fd_ctx_t *fdctx = NULL;
+        gf_boolean_t witness = _gf_false;
+
+	fdctx = afr_fd_ctx_get (fd, this);
+        if (!fdctx)
+                return _gf_true;
+
+        LOCK(&fd->lock);
+        {
+                if (fdctx->witnessed_unstable_write) {
+                        witness = _gf_true;
+                        fdctx->witnessed_unstable_write = _gf_false;
+                }
+        }
+        UNLOCK (&fd->lock);
+
+        return witness;
+}
+
+
+int
+afr_changelog_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int op_ret, int op_errno, struct iatt *pre,
+                         struct iatt *post, dict_t *xdata)
+{
+        afr_private_t *priv = NULL;
+        int child_index = (long) cookie;
+        int call_count = -1;
+        afr_local_t *local = NULL;
+
+        priv = this->private;
+        local = frame->local;
+
+        if (afr_fop_failed (op_ret, op_errno)) {
+                /* Failure of fsync() is as good as failure of previous
+                   write(). So treat it like one.
+                */
+                gf_log (this->name, GF_LOG_WARNING,
+                        "fsync(%s) failed on subvolume %s. Transaction was %s",
+                        uuid_utoa (local->fd->inode->gfid),
+                        priv->children[child_index]->name,
+                        gf_fop_list[local->op]);
+
+                afr_transaction_fop_failed (frame, this, child_index);
+        }
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0)
+                afr_changelog_post_op_now (frame, this);
+
+        return 0;
+}
+
+
+int
+afr_changelog_fsync (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+        int i = 0;
+        int call_count = 0;
+        afr_private_t *priv = NULL;
+ 	dict_t *xdata = NULL;
+ 	GF_UNUSED int ret = -1;
+
+        local = frame->local;
+        priv = this->private;
+
+        call_count = afr_pre_op_done_children_count (local->transaction.pre_op,
+                                                     priv->child_count);
+
+        if (!call_count) {
+                /* will go straight to unlock */
+                afr_changelog_post_op_now (frame, this);
+                return 0;
+        }
+
+        local->call_count = call_count;
+
+	xdata = dict_new();
+	if (xdata)
+		ret = dict_set_int32 (xdata, "batch-fsync", 1);
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (!local->transaction.pre_op[i])
+                        continue;
+
+                STACK_WIND_COOKIE (frame, afr_changelog_fsync_cbk,
+                                (void *) (long) i, priv->children[i],
+                                priv->children[i]->fops->fsync, local->fd,
+                                1, xdata);
+                if (!--call_count)
+                        break;
+        }
+
+	if (xdata)
+		dict_unref (xdata);
+
+        return 0;
+}
+
+
+        int
+afr_changelog_post_op_safe (call_frame_t *frame, xlator_t *this)
+{
+	afr_local_t    *local = NULL;
+        afr_private_t  *priv = NULL;
+
+	local = frame->local;
+        priv = this->private;
+
+        if (!local->fd || local->transaction.type != AFR_DATA_TRANSACTION) {
+                afr_changelog_post_op_now (frame, this);
+                return 0;
+        }
+
+        if (is_piggyback_post_op (frame, local->fd)) {
+                /* just detected that this post-op is about to
+                   be optimized away as a new write() has
+                   already piggybacked on this frame's changelog.
+                   */
+                afr_changelog_post_op_now (frame, this);
+                return 0;
+        }
+
+        /* Calling afr_changelog_post_op_now() now will result in
+           issuing ->[f]xattrop().
+
+           Performing a hard POST-OP (->[f]xattrop() FOP) is a more
+           responsible operation that what it might appear on the surface.
+
+           The changelog of a file (in the xattr of the file on the server)
+           stores information (pending count) about the state of the file
+           on the OTHER server. This changelog is blindly trusted, and must
+           therefore be updated in such a way it remains trustworthy. This
+           implies that decrementing the pending count (essentially "clearing
+           the dirty flag") must be done STRICTLY after we are sure that the
+           operation on the other server has reached stable storage.
+
+           While the backend filesystem on that server will eventually flush
+           it to stable storage, we (being in userspace) have no mechanism
+           to get notified when the write became "stable".
+
+           This means we need take matter into our own hands and issue an
+           fsync() EVEN IF THE APPLICATION WAS PERFORMING UNSTABLE WRITES,
+           and get an acknowledgement for it. And we need to wait for the
+           fsync() acknowledgement before initiating the hard POST-OP.
+
+           However if the FD itself was opened in O_SYNC or O_DSYNC then
+           we are already guaranteed that the writes were made stable as
+           part of the FOP itself. The same holds true for NFS stable
+           writes which happen on an anonymous FD with O_DSYNC or O_SYNC
+           flag set in the writev() @flags param. For all other write types,
+           mark a flag in the fdctx whenever an unstable write is witnessed.
+           */
+
+        if (!afr_fd_has_witnessed_unstable_write (this, local->fd)) {
+                afr_changelog_post_op_now (frame, this);
+                return 0;
+        }
+
+        /* Check whether users want durability and perform fsync/post-op
+         * accordingly.
+         */
+        if (priv->ensure_durability) {
+                /* Time to fsync() */
+                afr_changelog_fsync (frame, this);
+        } else {
+                afr_changelog_post_op_now (frame, this);
+        }
+
+        return 0;
+}
+
+
+void
+afr_delayed_changelog_post_op (xlator_t *this, call_frame_t *frame, fd_t *fd,
+                               call_stub_t *stub)
+{
+	afr_fd_ctx_t      *fd_ctx = NULL;
+	call_frame_t      *prev_frame = NULL;
+	struct timespec    delta = {0, };
+	afr_private_t     *priv = NULL;
+	afr_local_t       *local = NULL;
+
+	priv = this->private;
+
+	fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!fd_ctx)
+                goto out;
+
+	delta.tv_sec = priv->post_op_delay_secs;
+	delta.tv_nsec = 0;
+
+	pthread_mutex_lock (&fd_ctx->delay_lock);
+	{
+		prev_frame = fd_ctx->delay_frame;
+		fd_ctx->delay_frame = NULL;
+		if (fd_ctx->delay_timer)
+			gf_timer_call_cancel (this->ctx, fd_ctx->delay_timer);
+		fd_ctx->delay_timer = NULL;
+		if (!frame)
+			goto unlock;
+		fd_ctx->delay_timer = gf_timer_call_after (this->ctx, delta,
+							   afr_delayed_changelog_wake_up_cbk,
+							   fd);
+		fd_ctx->delay_frame = frame;
+	}
+unlock:
+	pthread_mutex_unlock (&fd_ctx->delay_lock);
+
+out:
+	if (prev_frame) {
+		local = prev_frame->local;
+		local->transaction.resume_stub = stub;
+		afr_changelog_post_op_safe (prev_frame, this);
+	} else if (stub) {
+		call_resume (stub);
+	}
+}
+
+
+void
+afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t  *local = NULL;
+
+        local = frame->local;
+
+        if (is_afr_delayed_changelog_post_op_needed (frame, this))
+                afr_delayed_changelog_post_op (this, frame, local->fd, NULL);
+        else
+                afr_changelog_post_op_safe (frame, this);
+}
+
+
+
+/* Wake up the sleeping/delayed post-op, and also register
+   a stub to have it resumed after this transaction
+   completely finishes.
+
+   The @stub gets saved in @local and gets resumed in
+   afr_local_cleanup()
+   */
+void
+afr_delayed_changelog_wake_resume (xlator_t *this, fd_t *fd, call_stub_t *stub)
+{
+        afr_delayed_changelog_post_op (this, NULL, fd, stub);
+}
+
+
+void
+afr_delayed_changelog_wake_up (xlator_t *this, fd_t *fd)
+{
+        afr_delayed_changelog_post_op (this, NULL, fd, NULL);
+}
+
+
+        int
 afr_transaction_resume (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
-	afr_local_t         *local    = NULL;
-	afr_private_t       *priv     = NULL;
+        afr_local_t         *local    = NULL;
+        afr_private_t       *priv     = NULL;
 
-	local    = frame->local;
+        local    = frame->local;
         int_lock = &local->internal_lock;
-	priv     = this->private;
-
-	if (__changelog_needed_post_op (frame, this)) {
-		afr_changelog_post_op (frame, this);
-	} else {
-		if (afr_lock_server_count (priv, local->transaction.type) == 0) {
-			local->transaction.done (frame, this);
-		} else {
+        priv     = this->private;
+
+        if (local->transaction.eager_lock_on) {
+                /* We don't need to retain "local" in the
+                   fd list anymore, writes to all subvols
+                   are finished by now */
+                LOCK (&local->fd->lock);
+                {
+                        list_del_init (&local->transaction.eager_locked);
+                }
+                UNLOCK (&local->fd->lock);
+        }
+
+        afr_restore_lk_owner (frame);
+
+        if (__fop_changelog_needed (frame, this)) {
+                afr_changelog_post_op (frame, this);
+        } else {
+                if (afr_lock_server_count (priv, local->transaction.type) == 0) {
+                        local->transaction.done (frame, this);
+                } else {
                         int_lock->lock_cbk = local->transaction.done;
-			afr_unlock (frame, this);
-		}
-	}
+                        afr_unlock (frame, this);
+                }
+        }
 
-	return 0;
+        return 0;
 }
 
 
@@ -1131,54 +1820,144 @@ afr_transaction_resume (call_frame_t *frame, xlator_t *this)
  */
 
 void
-afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this, int child_index)
+afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
+                            int child_index)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
+        afr_local_t *   local = NULL;
+        afr_private_t * priv  = NULL;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
+        priv  = this->private;
 
-        switch (local->op) {
-        case GF_FOP_WRITE:
-                __mark_fop_failed_on_fd (local->fd, this, child_index);
-                break;
-        default:
-                __mark_child_dead (local->pending, priv->child_count,
-                                   child_index, local->transaction.type);
-                break;
+        __mark_child_dead (local->pending, priv->child_count,
+                        child_index, local->transaction.type);
+}
+
+
+
+        static gf_boolean_t
+afr_locals_overlap (afr_local_t *local1, afr_local_t *local2)
+{
+        uint64_t start1 = local1->transaction.start;
+        uint64_t start2 = local2->transaction.start;
+        uint64_t end1 = 0;
+        uint64_t end2 = 0;
+
+        if (local1->transaction.len)
+                end1 = start1 + local1->transaction.len - 1;
+        else
+                end1 = ULLONG_MAX;
+
+        if (local2->transaction.len)
+                end2 = start2 + local2->transaction.len - 1;
+        else
+                end2 = ULLONG_MAX;
+
+        return ((end1 >= start2) && (end2 >= start1));
+}
+
+void
+afr_transaction_eager_lock_init (afr_local_t *local, xlator_t *this)
+{
+        afr_private_t *priv = NULL;
+        afr_fd_ctx_t  *fdctx = NULL;
+        afr_local_t   *each = NULL;
+
+        priv = this->private;
+
+        if (!local->fd)
+                return;
+
+        if (local->transaction.type != AFR_DATA_TRANSACTION)
+                return;
+
+        if (!priv->eager_lock)
+                return;
+
+        fdctx = afr_fd_ctx_get (local->fd, this);
+        if (!fdctx)
+                return;
+
+        if (afr_are_multiple_fds_opened (local->fd->inode, this))
+                return;
+        /*
+         * Once full file lock is acquired in eager-lock phase, overlapping
+         * writes do not compete for inode-locks, instead are transferred to the
+         * next writes. Because of this overlapping writes are not ordered.
+         * This can cause inconsistencies in replication.
+         * Example:
+         * Two overlapping writes w1, w2 are sent in parallel on same fd
+         * in two threads t1, t2.
+         * Both threads can execute afr_writev_wind in the following manner.
+         * t1 winds w1 on brick-0
+         * t2 winds w2 on brick-0
+         * t2 winds w2 on brick-1
+         * t1 winds w1 on brick-1
+         *
+         * This check makes sure the locks are not transferred for
+         * overlapping writes.
+         */
+        LOCK (&local->fd->lock);
+        {
+                list_for_each_entry (each, &fdctx->eager_locked,
+                                     transaction.eager_locked) {
+                        if (afr_locals_overlap (each, local)) {
+                                local->transaction.eager_lock_on = _gf_false;
+                                goto unlock;
+                        }
+                }
+
+                local->transaction.eager_lock_on = _gf_true;
+                list_add_tail (&local->transaction.eager_locked,
+                               &fdctx->eager_locked);
         }
+unlock:
+        UNLOCK (&local->fd->lock);
 }
 
 
-int32_t
+int
 afr_transaction (call_frame_t *frame, xlator_t *this, afr_transaction_type type)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
+        afr_local_t *   local = NULL;
+        afr_private_t * priv  = NULL;
+        fd_t            *fd   = NULL;
+        int             ret   = -1;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
+        priv  = this->private;
 
-	afr_transaction_local_init (local, priv);
+        local->transaction.resume = afr_transaction_resume;
+        local->transaction.type   = type;
 
-	local->transaction.resume = afr_transaction_resume;
-	local->transaction.type   = type;
+        ret = afr_transaction_local_init (local, this);
+        if (ret < 0)
+            goto out;
 
-	if (afr_lock_server_count (priv, local->transaction.type) == 0) {
-		if (__changelog_needed_pre_op (frame, this)) {
-			afr_changelog_pre_op (frame, this);
-		} else {
-                        __mark_all_success (local->pending, priv->child_count,
-                                            local->transaction.type);
+        afr_transaction_eager_lock_init (local, this);
 
-                        afr_pid_restore (frame);
+        if (local->fd && local->transaction.eager_lock_on)
+                afr_set_lk_owner (frame, this, local->fd);
+        else
+                afr_set_lk_owner (frame, this, frame->root);
 
-			local->transaction.fop (frame, this);
-		}
-	} else {
-		afr_lock (frame, this);
-	}
+        if (!local->transaction.eager_lock_on && local->loc.inode) {
+                fd = fd_lookup (local->loc.inode, frame->root->pid);
+                if (fd == NULL)
+                        fd = fd_lookup_anonymous (local->loc.inode);
+
+                if (fd) {
+                        afr_delayed_changelog_wake_up (this, fd);
+                        fd_unref (fd);
+                }
+        }
 
-	return 0;
+        if (afr_lock_server_count (priv, local->transaction.type) == 0) {
+                afr_internal_lock_finish (frame, this);
+        } else {
+                afr_lock (frame, this);
+        }
+        ret = 0;
+out:
+        return ret;
 }
diff --git a/xlators/cluster/afr/src/afr-transaction.h b/xlators/cluster/afr/src/afr-transaction.h
index 0d3d4443e..fa626fd0d 100644
--- a/xlators/cluster/afr/src/afr-transaction.h
+++ b/xlators/cluster/afr/src/afr-transaction.h
@@ -1,25 +1,21 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __TRANSACTION_H__
 #define __TRANSACTION_H__
 
+typedef enum {
+        LOCAL_FIRST = 1,
+        LOCAL_LAST = 2
+} afr_xattrop_type_t;
+
 void
 afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
 			    int child_index);
@@ -27,7 +23,29 @@ afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
 int
 afr_lock_server_count (afr_private_t *priv, afr_transaction_type type);
 
+afr_inodelk_t*
+afr_get_inodelk (afr_internal_lock_t *int_lock, char *dom);
+
 int32_t
 afr_transaction (call_frame_t *frame, xlator_t *this, afr_transaction_type type);
 
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this);
+int
+afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending,
+                      int child, afr_xattrop_type_t op);
+void
+afr_set_delayed_post_op (call_frame_t *frame, xlator_t *this);
+
+void
+afr_delayed_changelog_wake_up (xlator_t *this, fd_t *fd);
+
+void
+__mark_all_success (int32_t *pending[], int child_count,
+                    afr_transaction_type type);
+gf_boolean_t
+afr_any_fops_failed (afr_local_t *local, afr_private_t *priv);
+
+gf_boolean_t
+afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this);
 #endif /* __TRANSACTION_H__ */
diff --git a/xlators/cluster/afr/src/afr.c b/xlators/cluster/afr/src/afr.c
index a5906550d..c724eb2ae 100644
--- a/xlators/cluster/afr/src/afr.c
+++ b/xlators/cluster/afr/src/afr.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <libgen.h>
@@ -30,13 +21,25 @@
 #endif
 #include "afr-common.c"
 
+#define SHD_INODE_LRU_LIMIT          2048
+#define AFR_EH_HEALED_LIMIT          1024
+#define AFR_EH_HEAL_FAIL_LIMIT       1024
+#define AFR_EH_SPLIT_BRAIN_LIMIT     1024
+
+struct volume_options options[];
+
 int32_t
 notify (xlator_t *this, int32_t event,
-	void *data, ...)
+        void *data, ...)
 {
         int ret = -1;
+        va_list         ap;
+        void *data2 = NULL;
 
-        ret = afr_notify (this, event, data);
+        va_start (ap, data);
+        data2 = va_arg (ap, dict_t*);
+        va_end (ap);
+        ret = afr_notify (this, event, data, data2);
 
         return ret;
 }
@@ -53,7 +56,7 @@ mem_acct_init (xlator_t *this)
 
         if (ret != 0) {
                 gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
-                                "failed");
+                       "failed");
                 return ret;
         }
 
@@ -62,644 +65,476 @@ mem_acct_init (xlator_t *this)
 
 
 int
+xlator_subvolume_index (xlator_t *this, xlator_t *subvol)
+{
+        int index = -1;
+        int i = 0;
+        xlator_list_t *list = NULL;
+
+        list = this->children;
+
+        while (list) {
+                if (subvol == list->xlator ||
+                    strcmp (subvol->name, list->xlator->name) == 0) {
+                        index = i;
+                        break;
+                }
+                list = list->next;
+                i++;
+        }
+
+        return index;
+}
+
+void
+fix_quorum_options (xlator_t *this, afr_private_t *priv, char *qtype)
+{
+        if (priv->quorum_count && strcmp(qtype,"fixed")) {
+                gf_log(this->name,GF_LOG_WARNING,
+                       "quorum-type %s overriding quorum-count %u",
+                       qtype, priv->quorum_count);
+        }
+        if (!strcmp(qtype,"none")) {
+                priv->quorum_count = 0;
+        }
+        else if (!strcmp(qtype,"auto")) {
+                priv->quorum_count = AFR_QUORUM_AUTO;
+        }
+}
+
+int
 reconfigure (xlator_t *this, dict_t *options)
 {
+        afr_private_t *priv        = NULL;
+        xlator_t      *read_subvol = NULL;
+        int            read_subvol_index = -1;
+        int            ret         = -1;
+        int            index       = -1;
+        char          *qtype       = NULL;
+
+        priv = this->private;
 
-	gf_boolean_t metadata_self_heal;   /* on/off */
-	gf_boolean_t entry_self_heal;
-	gf_boolean_t data_self_heal;
-	gf_boolean_t data_change_log;       /* on/off */
-	gf_boolean_t metadata_change_log;   /* on/off */
-	gf_boolean_t entry_change_log;      /* on/off */
-	gf_boolean_t strict_readdir;
-
-	afr_private_t * priv        = NULL;
-	xlator_list_t * trav        = NULL;
-	
-	char * read_subvol     = NULL;
-	char * self_heal       = NULL;
-	char * change_log      = NULL;
-	char * str_readdir     = NULL;
-
-        int32_t background_count  = 0;
-        int32_t window_size       = 0;
-
-	int    read_ret      = -1;
-	int    dict_ret      = -1;
-	int    flag	     = 1;
-	int    ret           = 0;
-	int    temp_ret	     = -1;
-	
-	priv = this->private;
-	
-	dict_ret = dict_get_int32 (options, "background-self-heal-count",
-				   &background_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring background self-heal count to %d",
-			background_count);
-
-		priv->background_self_heal_count = background_count;
-	}
-
-	dict_ret = dict_get_str (options, "metadata-self-heal",
-				 &self_heal);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (self_heal, &metadata_self_heal);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Reconfiguration Invalid 'option metadata"
-				"-self-heal %s'. Defaulting to old value.", 
-				self_heal);
-			ret = -1;
-			goto out;
-		} 
-	
-		priv->metadata_self_heal = metadata_self_heal;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option metadata"
-			"-self-heal %s'.", 
-			self_heal);
-	}
-
-	dict_ret = dict_get_str (options, "data-self-heal",
-				 &self_heal);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (self_heal, &data_self_heal);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Reconfiguration Invalid 'option data"
-				"-self-heal %s'. Defaulting to old value.", 
-				self_heal);
-			ret = -1;
-			goto out;
-		} 
-	
-		priv->data_self_heal = data_self_heal;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option data"
-			"-self-heal %s'.", self_heal);
-	}
-
-	dict_ret = dict_get_str (options, "entry-self-heal",
-				 &self_heal);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (self_heal, &entry_self_heal);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Reconfiguration Invalid 'option data"
-				"-self-heal %s'. Defaulting to old value.", 
-				self_heal);
-			ret = -1;
-			goto out;
-		} 
-	
-		priv->entry_self_heal = entry_self_heal;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option entry"
-			"-self-heal %s'.", self_heal);
-	}
-
-
-	dict_ret = dict_get_str (options, "strict-readdir",
-				 &str_readdir);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (str_readdir, &strict_readdir);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option strict-readdir %s'. "
-				"Defaulting to old value.",
-				str_readdir);
-			ret = -1;
-			goto out;
-		}
-		
-		priv->strict_readdir = strict_readdir;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option strict"
-			"-readdir %s'.", str_readdir);
-	}
-
-	dict_ret = dict_get_int32 (options, "data-self-heal-window-size",
-				   &window_size);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring, Setting data self-heal window size to %d",
-			window_size);
-
-		priv->data_self_heal_window_size = window_size;
-	}
-
-	dict_ret = dict_get_str (options, "data-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (change_log, &data_change_log);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Reconfiguration Invalid 'option data-"
-				"change-log %s'. Defaulting to old value.", 
-				change_log);
-			ret = -1;
-			goto out;
-		}
- 
-		priv->data_change_log = data_change_log;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option data-"
-			"change-log %s'.", change_log);
-	}
-
-	dict_ret = dict_get_str (options, "metadata-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (change_log,
-					 &metadata_change_log);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option metadata-change-log %s'. "
-				"Defaulting to metadata-change-log as 'off'.",
-				change_log);
-			ret = -1;
-			goto out;
-		} 
-		
-		priv->metadata_change_log = metadata_change_log;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option metadata-"
-			"change-log %s'.", change_log);
-	}
-
-	dict_ret = dict_get_str (options, "entry-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (change_log, &entry_change_log);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option entry-change-log %s'. "
-				"Defaulting to entry-change-log as 'on'.", 
-				change_log);
-			ret = -1;
-			goto out;
-		} 
-
-		priv->entry_change_log = entry_change_log;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option entry-"
-			"change-log %s'.", change_log);
-	}
-
-	read_ret = dict_get_str (options, "read-subvolume", &read_subvol);
-
-	if (read_ret == -1) 
-		goto next;// No need to traverse, hence set the next option
-	
-	trav = this->children;
-	flag = 0;
-	while (trav) {
-		if (!read_ret && !strcmp (read_subvol, trav->xlator->name)) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"Subvolume '%s' specified as read child.",
-				trav->xlator->name);
-
-			flag = 1;
-			ret = -1; 
-			goto out;
-		}
-
-		
-		trav = trav->next;
-	}
-
-	if (flag == 0 ) {
-
-		gf_log (this->name, GF_LOG_ERROR,
-			"Invalid 'option read-subvolume %s', no such subvolume"
-			, read_subvol);
-		ret = -1;
-		goto out;
-	}
-
-next:
+        GF_OPTION_RECONF ("background-self-heal-count",
+                          priv->background_self_heal_count, options, uint32,
+                          out);
+
+        GF_OPTION_RECONF ("metadata-self-heal",
+                          priv->metadata_self_heal, options, bool, out);
+
+        GF_OPTION_RECONF ("data-self-heal", priv->data_self_heal, options, str,
+                          out);
+
+        GF_OPTION_RECONF ("entry-self-heal", priv->entry_self_heal, options,
+                          bool, out);
+
+        GF_OPTION_RECONF ("strict-readdir", priv->strict_readdir, options, bool,
+                          out);
+
+        GF_OPTION_RECONF ("data-self-heal-window-size",
+                          priv->data_self_heal_window_size, options,
+                          uint32, out);
+
+        GF_OPTION_RECONF ("data-change-log", priv->data_change_log, options,
+                          bool, out);
+
+        GF_OPTION_RECONF ("metadata-change-log",
+                          priv->metadata_change_log, options, bool, out);
+
+        GF_OPTION_RECONF ("entry-change-log", priv->entry_change_log, options,
+                          bool, out);
+
+        GF_OPTION_RECONF ("data-self-heal-algorithm",
+                          priv->data_self_heal_algorithm, options, str, out);
+
+        GF_OPTION_RECONF ("self-heal-daemon", priv->shd.enabled, options, bool, out);
+
+        GF_OPTION_RECONF ("read-subvolume", read_subvol, options, xlator, out);
+
+        GF_OPTION_RECONF ("read-hash-mode", priv->hash_mode,
+                          options, uint32, out);
+
+        if (read_subvol) {
+                index = xlator_subvolume_index (this, read_subvol);
+                if (index == -1) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s not a subvolume",
+                                read_subvol->name);
+                        goto out;
+                }
+                priv->read_child = index;
+        }
+
+        GF_OPTION_RECONF ("read-subvolume-index",read_subvol_index, options,int32,out);
+
+        if (read_subvol_index >-1) {
+                index=read_subvol_index;
+                if (index >= priv->child_count) {
+                        gf_log (this->name, GF_LOG_ERROR, "%d not a subvolume-index",
+                                index);
+                        goto out;
+                }
+                priv->read_child = index;
+        }
+
+        GF_OPTION_RECONF ("eager-lock", priv->eager_lock, options, bool, out);
+        GF_OPTION_RECONF ("quorum-type", qtype, options, str, out);
+        GF_OPTION_RECONF ("quorum-count", priv->quorum_count, options,
+                          uint32, out);
+        fix_quorum_options(this,priv,qtype);
+        GF_OPTION_RECONF ("heal-timeout", priv->shd.timeout, options,
+                          int32, out);
+
+	GF_OPTION_RECONF ("post-op-delay-secs", priv->post_op_delay_secs, options,
+			  uint32, out);
+
+        GF_OPTION_RECONF (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size,
+                          options, size, out);
+        /* Reset this so we re-discover in case the topology changed.  */
+        GF_OPTION_RECONF ("readdir-failover", priv->readdir_failover, options,
+                          bool, out);
+        GF_OPTION_RECONF ("ensure-durability", priv->ensure_durability, options,
+                          bool, out);
+        priv->did_discovery = _gf_false;
+
+        ret = 0;
 out:
-	return ret;
+        return ret;
 
 }
 
 
 static const char *favorite_child_warning_str = "You have specified subvolume '%s' "
-	"as the 'favorite child'. This means that if a discrepancy in the content "
-	"or attributes (ownership, permission, etc.) of a file is detected among "
-	"the subvolumes, the file on '%s' will be considered the definitive "
-	"version and its contents will OVERWRITE the contents of the file on other "
-	"subvolumes. All versions of the file except that on '%s' "
-	"WILL BE LOST.";
-
-static const char *no_lock_servers_warning_str = "You have set lock-server-count = 0. "
-	"This means correctness is NO LONGER GUARANTEED in all cases. If two or more "
-	"applications write to the same region of a file, there is a possibility that "
-	"its copies will be INCONSISTENT. Set it to a value greater than 0 unless you "
-	"are ABSOLUTELY SURE of what you are doing and WILL NOT HOLD GlusterFS "
-	"RESPONSIBLE for inconsistent data. If you are in doubt, set it to a value "
-	"greater than 0.";
-
-int32_t 
+        "as the 'favorite child'. This means that if a discrepancy in the content "
+        "or attributes (ownership, permission, etc.) of a file is detected among "
+        "the subvolumes, the file on '%s' will be considered the definitive "
+        "version and its contents will OVERWRITE the contents of the file on other "
+        "subvolumes. All versions of the file except that on '%s' "
+        "WILL BE LOST.";
+
+
+int32_t
 init (xlator_t *this)
 {
-	afr_private_t * priv        = NULL;
-	int             child_count = 0;
-	xlator_list_t * trav        = NULL;
-	int             i           = 0;
-	int             ret         = -1;
-	int             op_errno    = 0;
-
-	char * read_subvol     = NULL;
-	char * fav_child       = NULL;
-	char * self_heal       = NULL;
-        char * algo            = NULL;
-	char * change_log      = NULL;
-	char * strict_readdir  = NULL;
-        char * inodelk_trace   = NULL;
-        char * entrylk_trace   = NULL;
-
-        int32_t background_count  = 0;
-	int32_t lock_server_count = 1;
-        int32_t window_size       = 0;
-
-	int    fav_ret       = -1;
-	int    read_ret      = -1;
-	int    dict_ret      = -1;
-
-
-	if (!this->children) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"replicate translator needs more than one "
+        afr_private_t *priv        = NULL;
+        int            child_count = 0;
+        xlator_list_t *trav        = NULL;
+        int            i           = 0;
+        int            ret         = -1;
+        GF_UNUSED int  op_errno    = 0;
+        xlator_t      *read_subvol = NULL;
+        int            read_subvol_index = -1;
+        xlator_t      *fav_child   = NULL;
+        char          *qtype       = NULL;
+
+        if (!this->children) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "replicate translator needs more than one "
                         "subvolume defined.");
-		return -1;
-	}
-  
-	if (!this->parents) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"Volume is dangling.");
-	}
+                return -1;
+        }
 
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Volume is dangling.");
+        }
 
-	ALLOC_OR_GOTO (this->private, afr_private_t, out);
+	this->private = GF_CALLOC (1, sizeof (afr_private_t),
+                                   gf_afr_mt_afr_private_t);
+        if (!this->private)
+                goto out;
 
-	priv = this->private;
+        priv = this->private;
+        LOCK_INIT (&priv->lock);
+        LOCK_INIT (&priv->read_child_lock);
+        //lock recovery is not done in afr
+        pthread_mutex_init (&priv->mutex, NULL);
+        INIT_LIST_HEAD (&priv->saved_fds);
 
-	read_ret = dict_get_str (this->options, "read-subvolume", &read_subvol);
-	priv->read_child = -1;
+        child_count = xlator_subvolume_count (this);
 
-	fav_ret = dict_get_str (this->options, "favorite-child", &fav_child);
-	priv->favorite_child = -1;
+        priv->child_count = child_count;
 
-        priv->background_self_heal_count = 16;
+        priv->read_child = -1;
 
-	dict_ret = dict_get_int32 (this->options, "background-self-heal-count",
-				   &background_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting background self-heal count to %d",
-			background_count);
+        GF_OPTION_INIT ("read-subvolume", read_subvol, xlator, out);
+        if (read_subvol) {
+                priv->read_child = xlator_subvolume_index (this, read_subvol);
+                if (priv->read_child == -1) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s not a subvolume",
+                                read_subvol->name);
+                        goto out;
+                }
+        }
+        GF_OPTION_INIT ("read-subvolume-index",read_subvol_index,int32,out);
+        if (read_subvol_index > -1) {
+                if (read_subvol_index >= priv->child_count) {
+                        gf_log (this->name, GF_LOG_ERROR, "%d not a subvolume-index",
+                                read_subvol_index);
+                        goto out;
+                }
+                priv->read_child = read_subvol_index;
+        }
+        GF_OPTION_INIT ("choose-local", priv->choose_local, bool, out);
 
-		priv->background_self_heal_count = background_count;
-	}
+        GF_OPTION_INIT ("read-hash-mode", priv->hash_mode, uint32, out);
 
-	/* Default values */
+        priv->favorite_child = -1;
+        GF_OPTION_INIT ("favorite-child", fav_child, xlator, out);
+        if (fav_child) {
+                priv->favorite_child = xlator_subvolume_index (this, fav_child);
+                if (priv->favorite_child == -1) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s not a subvolume",
+                                fav_child->name);
+                        goto out;
+                }
+                gf_log (this->name, GF_LOG_WARNING,
+                        favorite_child_warning_str, fav_child->name,
+                        fav_child->name, fav_child->name);
+        }
 
-	priv->data_self_heal     = 1;
-	priv->metadata_self_heal = 1;
-	priv->entry_self_heal    = 1;
 
-	dict_ret = dict_get_str (this->options, "data-self-heal", &self_heal);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (self_heal, &priv->data_self_heal);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option data-self-heal %s'. "
-				"Defaulting to data-self-heal as 'on'",
-				self_heal);
-			priv->data_self_heal = 1;
-		}
-	}
+        GF_OPTION_INIT ("background-self-heal-count",
+                        priv->background_self_heal_count, uint32, out);
 
-        priv->data_self_heal_algorithm = "";
+        GF_OPTION_INIT ("data-self-heal", priv->data_self_heal, str, out);
 
-        dict_ret = dict_get_str (this->options, "data-self-heal-algorithm",
-                                 &algo);
-        if (dict_ret == 0) {
-                priv->data_self_heal_algorithm = gf_strdup (algo);
-        }
+        GF_OPTION_INIT ("data-self-heal-algorithm",
+                        priv->data_self_heal_algorithm, str, out);
 
+        GF_OPTION_INIT ("data-self-heal-window-size",
+                        priv->data_self_heal_window_size, uint32, out);
 
-        priv->data_self_heal_window_size = 16;
-
-	dict_ret = dict_get_int32 (this->options, "data-self-heal-window-size",
-				   &window_size);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting data self-heal window size to %d",
-			window_size);
-
-		priv->data_self_heal_window_size = window_size;
-	}
-
-	dict_ret = dict_get_str (this->options, "metadata-self-heal",
-				 &self_heal);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (self_heal, &priv->metadata_self_heal);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option metadata-self-heal %s'. "
-				"Defaulting to metadata-self-heal as 'on'.", 
-				self_heal);
-			priv->metadata_self_heal = 1;
-		} 
-	}
-
-	dict_ret = dict_get_str (this->options, "entry-self-heal", &self_heal);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (self_heal, &priv->entry_self_heal);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option entry-self-heal %s'. "
-				"Defaulting to entry-self-heal as 'on'.", 
-				self_heal);
-			priv->entry_self_heal = 1;
-		} 
-	}
-
-	/* Change log options */
-
-	priv->data_change_log     = 1;
-	priv->metadata_change_log = 1;
-	priv->entry_change_log    = 1;
-
-	dict_ret = dict_get_str (this->options, "data-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (change_log, &priv->data_change_log);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option data-change-log %s'. "
-				"Defaulting to data-change-log as 'on'.", 
-				change_log);
-			priv->data_change_log = 1;
-		} 
-	}
-
-	dict_ret = dict_get_str (this->options, "metadata-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (change_log,
-					 &priv->metadata_change_log);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option metadata-change-log %s'. "
-				"Defaulting to metadata-change-log as 'off'.",
-				change_log);
-			priv->metadata_change_log = 0;
-		} 
-	}
-
-	dict_ret = dict_get_str (this->options, "entry-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (change_log, &priv->entry_change_log);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option entry-change-log %s'. "
-				"Defaulting to entry-change-log as 'on'.", 
-				change_log);
-			priv->entry_change_log = 1;
-		} 
-	}
-
-	/* Locking options */
-
-        priv->inodelk_trace = 0;
-        priv->entrylk_trace = 0;
-
-	dict_ret = dict_get_str (this->options, "inodelk-trace",
-				 &inodelk_trace);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (inodelk_trace, &priv->inodelk_trace);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option inodelk-trace %s' ",
-                                inodelk_trace);
-
-			priv->inodelk_trace = 0;
-		}
-	}
-
-
-	dict_ret = dict_get_str (this->options, "entrylk-trace",
-				 &entrylk_trace);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (entrylk_trace, &priv->entrylk_trace);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option entrylk-trace %s' ",
-                                inodelk_trace);
-
-			priv->entrylk_trace = 0;
-		}
-	}
-
-
-	priv->data_lock_server_count = 1;
-	priv->metadata_lock_server_count = 0;
-	priv->entry_lock_server_count = 1;
-
-	dict_ret = dict_get_int32 (this->options, "data-lock-server-count", 
-				   &lock_server_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting data lock server count to %d.",
-			lock_server_count);
-
-		if (lock_server_count == 0) 
-			gf_log (this->name, GF_LOG_WARNING, "%s",
-                                no_lock_servers_warning_str);
-
-		priv->data_lock_server_count = lock_server_count;
-	}
-
-
-	dict_ret = dict_get_int32 (this->options,
-				   "metadata-lock-server-count", 
-				   &lock_server_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting metadata lock server count to %d.",
-			lock_server_count);
-		priv->metadata_lock_server_count = lock_server_count;
-	}
-
-
-	dict_ret = dict_get_int32 (this->options, "entry-lock-server-count", 
-				   &lock_server_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting entry lock server count to %d.",
-			lock_server_count);
-
-		priv->entry_lock_server_count = lock_server_count;
-	}
-
-	priv->strict_readdir = _gf_false;
-
-	dict_ret = dict_get_str (this->options, "strict-readdir",
-				 &strict_readdir);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (strict_readdir, &priv->strict_readdir);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option strict-readdir %s'. "
-				"Defaulting to strict-readdir as 'off'.",
-				strict_readdir);
-		}
-	}
-
-	trav = this->children;
-	while (trav) {
-		if (!read_ret && !strcmp (read_subvol, trav->xlator->name)) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"Subvolume '%s' specified as read child.",
-				trav->xlator->name);
-
-			priv->read_child = child_count;
-		}
-
-		if (fav_ret == 0 && !strcmp (fav_child, trav->xlator->name)) {
-			gf_log (this->name, GF_LOG_WARNING,
-				favorite_child_warning_str, trav->xlator->name,
-				trav->xlator->name, trav->xlator->name);
-			priv->favorite_child = child_count;
-		}
-
-		child_count++;
-		trav = trav->next;
-	}
-
-	priv->wait_count = 1;
-
-	priv->child_count = child_count;
-
-	LOCK_INIT (&priv->lock);
-        LOCK_INIT (&priv->read_child_lock);
+        GF_OPTION_INIT ("metadata-self-heal", priv->metadata_self_heal, bool,
+                        out);
+
+        GF_OPTION_INIT ("entry-self-heal", priv->entry_self_heal, bool, out);
+
+        GF_OPTION_INIT ("self-heal-daemon", priv->shd.enabled, bool, out);
+
+        GF_OPTION_INIT ("iam-self-heal-daemon", priv->shd.iamshd, bool, out);
+
+        GF_OPTION_INIT ("data-change-log", priv->data_change_log, bool, out);
+
+        GF_OPTION_INIT ("metadata-change-log", priv->metadata_change_log, bool,
+                        out);
 
-	priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
+        GF_OPTION_INIT ("entry-change-log", priv->entry_change_log, bool, out);
+
+        GF_OPTION_INIT ("optimistic-change-log", priv->optimistic_change_log,
+                        bool, out);
+
+        GF_OPTION_INIT ("inodelk-trace", priv->inodelk_trace, bool, out);
+
+        GF_OPTION_INIT ("entrylk-trace", priv->entrylk_trace, bool, out);
+
+        GF_OPTION_INIT ("strict-readdir", priv->strict_readdir, bool, out);
+
+        GF_OPTION_INIT ("eager-lock", priv->eager_lock, bool, out);
+        GF_OPTION_INIT ("quorum-type", qtype, str, out);
+        GF_OPTION_INIT ("quorum-count", priv->quorum_count, uint32, out);
+        GF_OPTION_INIT (AFR_SH_READDIR_SIZE_KEY, priv->sh_readdir_size, size,
+                        out);
+        fix_quorum_options(this,priv,qtype);
+
+	GF_OPTION_INIT ("post-op-delay-secs", priv->post_op_delay_secs, uint32, out);
+        GF_OPTION_INIT ("readdir-failover", priv->readdir_failover, bool, out);
+        GF_OPTION_INIT ("ensure-durability", priv->ensure_durability, bool,
+                        out);
+
+        priv->wait_count = 1;
+
+        priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
                                     gf_afr_mt_char);
-	if (!priv->child_up) {
-		gf_log (this->name, GF_LOG_ERROR,	
-			"Out of memory.");		
-		ret = -ENOMEM;
-		goto out;
-	}
-
-	priv->children = GF_CALLOC (sizeof (xlator_t *), child_count,
+        if (!priv->child_up) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
+        for (i = 0; i < child_count; i++)
+                priv->child_up[i] = -1; /* start with unknown state.
+                                           this initialization needed
+                                           for afr_notify() to work
+                                           reliably
+                                        */
+
+        priv->children = GF_CALLOC (sizeof (xlator_t *), child_count,
                                     gf_afr_mt_xlator_t);
-	if (!priv->children) {
-		gf_log (this->name, GF_LOG_ERROR,	
-			"Out of memory.");		
-		ret = -ENOMEM;
-		goto out;
-	}
-
-        priv->pending_key = GF_CALLOC (sizeof (*priv->pending_key), 
-                                        child_count,
-                                        gf_afr_mt_char);
+        if (!priv->children) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
+        priv->pending_key = GF_CALLOC (sizeof (*priv->pending_key),
+                                       child_count,
+                                       gf_afr_mt_char);
         if (!priv->pending_key) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
                 ret = -ENOMEM;
                 goto out;
         }
 
-	trav = this->children;
-	i = 0;
-	while (i < child_count) {
-		priv->children[i] = trav->xlator;
+        trav = this->children;
+        i = 0;
+        while (i < child_count) {
+                priv->children[i] = trav->xlator;
 
-                ret = gf_asprintf (&priv->pending_key[i], "%s.%s", 
+                ret = gf_asprintf (&priv->pending_key[i], "%s.%s",
                                    AFR_XATTR_PREFIX,
                                    trav->xlator->name);
                 if (-1 == ret) {
-                        gf_log (this->name, GF_LOG_ERROR, 
-                                "asprintf failed to set pending key");
                         ret = -ENOMEM;
                         goto out;
                 }
 
-		trav = trav->next;
-		i++;
-	}
+                trav = trav->next;
+                i++;
+        }
+
+        ret = gf_asprintf (&priv->sh_domain, AFR_SH_DATA_DOMAIN_FMT,
+                           this->name);
+        if (-1 == ret) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
+        priv->last_event = GF_CALLOC (child_count, sizeof (*priv->last_event),
+                                      gf_afr_mt_int32_t);
+        if (!priv->last_event) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
+        /* keep more local here as we may need them for self-heal etc */
+        this->local_pool = mem_pool_new (afr_local_t, 512);
+        if (!this->local_pool) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto out;
+        }
 
-        LOCK_INIT (&priv->root_inode_lk);
         priv->first_lookup = 1;
         priv->root_inode = NULL;
 
-	ret = 0;
+        if (!priv->shd.iamshd) {
+                ret = 0;
+                goto out;
+        }
+
+        ret = -ENOMEM;
+        priv->shd.pos = GF_CALLOC (sizeof (*priv->shd.pos), child_count,
+                                   gf_afr_mt_brick_pos_t);
+        if (!priv->shd.pos)
+                goto out;
+
+        priv->shd.pending = GF_CALLOC (sizeof (*priv->shd.pending), child_count,
+                                       gf_afr_mt_int32_t);
+        if (!priv->shd.pending)
+                goto out;
+
+        priv->shd.inprogress = GF_CALLOC (sizeof (*priv->shd.inprogress),
+                                          child_count, gf_afr_mt_shd_bool_t);
+        if (!priv->shd.inprogress)
+                goto out;
+        priv->shd.timer = GF_CALLOC (sizeof (*priv->shd.timer), child_count,
+                                     gf_afr_mt_shd_timer_t);
+        if (!priv->shd.timer)
+                goto out;
+
+        priv->shd.healed = eh_new (AFR_EH_HEALED_LIMIT, _gf_false,
+                                   _destroy_shd_event_data);
+        if (!priv->shd.healed)
+                goto out;
+
+        priv->shd.heal_failed = eh_new (AFR_EH_HEAL_FAIL_LIMIT, _gf_false,
+                                        _destroy_shd_event_data);
+        if (!priv->shd.heal_failed)
+                goto out;
+
+        priv->shd.split_brain = eh_new (AFR_EH_SPLIT_BRAIN_LIMIT, _gf_false,
+                                        _destroy_shd_event_data);
+        if (!priv->shd.split_brain)
+                goto out;
+
+        this->itable = inode_table_new (SHD_INODE_LRU_LIMIT, this);
+        if (!this->itable)
+                goto out;
+        priv->root_inode = inode_ref (this->itable->root);
+        GF_OPTION_INIT ("node-uuid", priv->shd.node_uuid, str, out);
+        GF_OPTION_INIT ("heal-timeout", priv->shd.timeout, int32, out);
+        ret = afr_initialise_statistics (this);
+        if (ret)
+                goto out;
+        ret = 0;
 out:
-	return ret;
+        return ret;
 }
 
 
 int
 fini (xlator_t *this)
 {
-	return 0;
+        afr_private_t *priv = NULL;
+
+        priv = this->private;
+        this->private = NULL;
+        afr_priv_destroy (priv);
+        if (this->itable);//I dont see any destroy func
+
+        return 0;
 }
 
 
 struct xlator_fops fops = {
-	.lookup      = afr_lookup,
-	.open        = afr_open,
-	.lk          = afr_lk,
-	.flush       = afr_flush,
-	.statfs      = afr_statfs,
-	.fsync       = afr_fsync,
-	.fsyncdir    = afr_fsyncdir,
-	.xattrop     = afr_xattrop,
-	.fxattrop    = afr_fxattrop,
-	.inodelk     = afr_inodelk,
-	.finodelk    = afr_finodelk,
-	.entrylk     = afr_entrylk,
-	.fentrylk    = afr_fentrylk,
-
-	/* inode read */
-	.access      = afr_access,
-	.stat        = afr_stat,
-	.fstat       = afr_fstat,
-	.readlink    = afr_readlink,
-	.getxattr    = afr_getxattr,
-	.readv       = afr_readv,
-
-	/* inode write */
-	.writev      = afr_writev,
-	.truncate    = afr_truncate,
-	.ftruncate   = afr_ftruncate,
-	.setxattr    = afr_setxattr,
+        .lookup      = afr_lookup,
+        .open        = afr_open,
+        .lk          = afr_lk,
+        .flush       = afr_flush,
+        .statfs      = afr_statfs,
+        .fsync       = afr_fsync,
+        .fsyncdir    = afr_fsyncdir,
+        .xattrop     = afr_xattrop,
+        .fxattrop    = afr_fxattrop,
+        .inodelk     = afr_inodelk,
+        .finodelk    = afr_finodelk,
+        .entrylk     = afr_entrylk,
+        .fentrylk    = afr_fentrylk,
+	.fallocate   = afr_fallocate,
+	.discard     = afr_discard,
+        .zerofill    = afr_zerofill,
+
+        /* inode read */
+        .access      = afr_access,
+        .stat        = afr_stat,
+        .fstat       = afr_fstat,
+        .readlink    = afr_readlink,
+        .getxattr    = afr_getxattr,
+        .fgetxattr   = afr_fgetxattr,
+        .readv       = afr_readv,
+
+        /* inode write */
+        .writev      = afr_writev,
+        .truncate    = afr_truncate,
+        .ftruncate   = afr_ftruncate,
+        .setxattr    = afr_setxattr,
+        .fsetxattr   = afr_fsetxattr,
         .setattr     = afr_setattr,
-	.fsetattr    = afr_fsetattr,
-	.removexattr = afr_removexattr,
-
-	/* dir read */
-	.opendir     = afr_opendir,
-	.readdir     = afr_readdir,
-	.readdirp    = afr_readdirp,
-
-	/* dir write */
-	.create      = afr_create,
-	.mknod       = afr_mknod,
-	.mkdir       = afr_mkdir,
-	.unlink      = afr_unlink,
-	.rmdir       = afr_rmdir,
-	.link        = afr_link,
-	.symlink     = afr_symlink,
-	.rename      = afr_rename,
+        .fsetattr    = afr_fsetattr,
+        .removexattr = afr_removexattr,
+        .fremovexattr = afr_fremovexattr,
+
+        /* dir read */
+        .opendir     = afr_opendir,
+        .readdir     = afr_readdir,
+        .readdirp    = afr_readdirp,
+
+        /* dir write */
+        .create      = afr_create,
+        .mknod       = afr_mknod,
+        .mkdir       = afr_mkdir,
+        .unlink      = afr_unlink,
+        .rmdir       = afr_rmdir,
+        .link        = afr_link,
+        .symlink     = afr_symlink,
+        .rename      = afr_rename,
 };
 
 
@@ -709,62 +544,250 @@ struct xlator_dumpops dumpops = {
 
 
 struct xlator_cbks cbks = {
-	.release     = afr_release,
-	.releasedir  = afr_releasedir,
+        .release     = afr_release,
+        .releasedir  = afr_releasedir,
+        .forget      = afr_forget,
 };
 
 
 struct volume_options options[] = {
-	{ .key  = {"read-subvolume" }, 
-	  .type = GF_OPTION_TYPE_XLATOR
-	},
-	{ .key  = {"favorite-child"}, 
-	  .type = GF_OPTION_TYPE_XLATOR
-	},
+        { .key  = {"read-subvolume" },
+          .type = GF_OPTION_TYPE_XLATOR,
+          .description = "inode-read fops happen only on one of the bricks in "
+                         "replicate. Afr will prefer the one specified using "
+                         "this option if it is not stale. Option value must be "
+                         "one of the xlator names of the children. "
+                         "Ex: <volname>-client-0 till "
+                         "<volname>-client-<number-of-bricks - 1>"
+        },
+        { .key  = {"read-subvolume-index" },
+          .type = GF_OPTION_TYPE_INT,
+          .default_value = "-1",
+          .description = "inode-read fops happen only on one of the bricks in "
+                         "replicate. AFR will prefer the one specified using "
+                         "this option if it is not stale. allowed options"
+                         " include -1 till replica-count - 1"
+        },
+        { .key = {"read-hash-mode" },
+          .type = GF_OPTION_TYPE_INT,
+          .min = 0,
+          .max = 2,
+          .default_value = "0",
+          .description = "inode-read fops happen only on one of the bricks in "
+                         "replicate. AFR will prefer the one computed using "
+                         "the method specified using this option"
+                         "0 = first responder, "
+                         "1 = hash by GFID of file (all clients use "
+                                                    "same subvolume), "
+                         "2 = hash by GFID of file and client PID",
+        },
+        { .key  = {"choose-local" },
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "true",
+          .description = "Choose a local subvolume(i.e. Brick) to read from if "
+                         "read-subvolume is not explicitly set.",
+        },
+        { .key  = {"favorite-child"},
+          .type = GF_OPTION_TYPE_XLATOR,
+          .description = "If a split-brain happens choose subvol/brick set by "
+                         "this option as source."
+        },
         { .key  = {"background-self-heal-count"},
           .type = GF_OPTION_TYPE_INT,
-          .min  = 0
+          .min  = 0,
+          .default_value = "16",
+          .validate = GF_OPT_VALIDATE_MIN,
+          .description = "This specifies the number of self-heals that can be "
+                         " performed in background without blocking the fop"
+        },
+        { .key  = {"data-self-heal"},
+          .type = GF_OPTION_TYPE_STR,
+          .value = {"1", "on", "yes", "true", "enable",
+                    "0", "off", "no", "false", "disable",
+                    "open"},
+          .default_value = "on",
+          .description   = "Using this option we can enable/disable data "
+                           "self-heal on the file. \"open\" means data "
+                           "self-heal action will only be triggered by file "
+                           "open operations."
         },
-	{ .key  = {"data-self-heal"},  
-	  .type = GF_OPTION_TYPE_BOOL 
-	},
         { .key  = {"data-self-heal-algorithm"},
-          .type = GF_OPTION_TYPE_STR
+          .type = GF_OPTION_TYPE_STR,
+          .description   = "Select between \"full\", \"diff\". The "
+                           "\"full\" algorithm copies the entire file from "
+                           "source to sink. The \"diff\" algorithm copies to "
+                           "sink only those blocks whose checksums don't match "
+                           "with those of source. If no option is configured "
+                           "the option is chosen dynamically as follows: "
+                           "If the file does not exist on one of the sinks "
+                           "or empty file exists or if the source file size is "
+                           "about the same as page size the entire file will "
+                           "be read and written i.e \"full\" algo, "
+                           "otherwise \"diff\" algo is chosen.",
+          .value = { "diff", "full"}
         },
         { .key  = {"data-self-heal-window-size"},
           .type = GF_OPTION_TYPE_INT,
           .min  = 1,
-          .max  = 1024
+          .max  = 1024,
+          .default_value = "1",
+          .description = "Maximum number blocks per file for which self-heal "
+                         "process would be applied simultaneously."
+        },
+        { .key  = {"metadata-self-heal"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "on",
+          .description = "Using this option we can enable/disable metadata "
+                         "i.e. Permissions, ownerships, xattrs self-heal on "
+                         "the file/directory."
+        },
+        { .key  = {"entry-self-heal"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "on",
+          .description = "Using this option we can enable/disable entry "
+                         "self-heal on the directory."
+        },
+        { .key  = {"data-change-log"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "on",
+          .description = "Data fops like write/truncate will not perform "
+                         "pre/post fop changelog operations in afr transaction "
+                         "if this option is disabled"
+        },
+        { .key  = {"metadata-change-log"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "on",
+          .description = "Metadata fops like setattr/setxattr will not perform "
+                         "pre/post fop changelog operations in afr transaction "
+                         "if this option is disabled"
+        },
+        { .key  = {"entry-change-log"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "on",
+          .description = "Entry fops like create/unlink will not perform "
+                         "pre/post fop changelog operations in afr transaction "
+                         "if this option is disabled"
+        },
+        { .key  = {"optimistic-change-log"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "on",
+          .description = "Entry/Metadata fops will not perform "
+                         "pre fop changelog operations in afr transaction "
+                         "if this option is enabled."
+        },
+        { .key  = {"strict-readdir"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+        },
+        { .key = {"inodelk-trace"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "Enabling this option logs inode lock/unlocks"
+        },
+        { .key = {"entrylk-trace"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "Enabling this option logs entry lock/unlocks"
+        },
+        { .key = {"eager-lock"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "on",
+          .description = "Lock phase of a transaction has two sub-phases. "
+                         "First is an attempt to acquire locks in parallel by "
+                         "broadcasting non-blocking lock requests. If lock "
+                         "aquistion fails on any server, then the held locks "
+                         "are unlocked and revert to a blocking locked mode "
+                         "sequentially on one server after another.  If this "
+                         "option is enabled the initial broadcasting lock "
+                         "request attempt to acquire lock on the entire file. "
+                         "If this fails, we revert back to the sequential "
+                         "\"regional\" blocking lock as before. In the case "
+                         "where such an \"eager\" lock is granted in the "
+                         "non-blocking phase, it gives rise to an opportunity "
+                         "for optimization. i.e, if the next write transaction "
+                         "on the same FD arrives before the unlock phase of "
+                         "the first transaction, it \"takes over\" the full "
+                         "file lock. Similarly if yet another data transaction "
+                         "arrives before the unlock phase of the \"optimized\" "
+                         "transaction, that in turn \"takes over\" the lock as "
+                         "well. The actual unlock now happens at the end of "
+                         "the last \"optimzed\" transaction."
+
+        },
+        { .key = {"self-heal-daemon"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "This option applies to only self-heal-daemon. "
+                         "Index directory crawl and automatic healing of files"
+                         "will not be performed if this option is turned off."
+        },
+        { .key = {"iam-self-heal-daemon"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "This option differentiates if the replicate "
+                         "translator is running as part of self-heal-daemon "
+                         "or not."
+        },
+        { .key = {"quorum-type"},
+          .type = GF_OPTION_TYPE_STR,
+          .value = { "none", "auto", "fixed"},
+          .default_value = "none",
+          .description = "If value is \"fixed\" only allow writes if "
+                         "quorum-count bricks are present.  If value is "
+                         "\"auto\" only allow writes if more than half of "
+                         "bricks, or exactly half including the first, are "
+                         "present.",
+        },
+        { .key = {"quorum-count"},
+          .type = GF_OPTION_TYPE_INT,
+          .min = 1,
+          .max = INT_MAX,
+          .default_value = 0,
+          .description = "If quorum-type is \"fixed\" only allow writes if "
+                         "this many bricks or present.  Other quorum types "
+                         "will OVERWRITE this value.",
+        },
+        { .key  = {"node-uuid"},
+          .type = GF_OPTION_TYPE_STR,
+          .description = "Local glusterd uuid string, used in starting "
+                         "self-heal-daemon so that it can crawl only on "
+                         "local index directories.",
+        },
+        { .key  = {"heal-timeout"},
+          .type = GF_OPTION_TYPE_INT,
+          .min  = 60,
+          .max  = INT_MAX,
+          .default_value = "600",
+          .description = "time interval for checking the need to self-heal "
+                         "in self-heal-daemon"
+        },
+        { .key  = {"post-op-delay-secs"},
+          .type = GF_OPTION_TYPE_INT,
+          .min  = 0,
+          .max  = INT_MAX,
+          .default_value = "1",
+          .description = "Time interval induced artificially before "
+	                 "post-operation phase of the transaction to "
+                         "enhance overlap of adjacent write operations.",
+        },
+        { .key = {AFR_SH_READDIR_SIZE_KEY},
+          .type = GF_OPTION_TYPE_SIZET,
+          .description = "readdirp size for performing entry self-heal",
+          .min = 1024,
+          .max = 131072,
+          .default_value = "1KB",
+        },
+        { .key = {"readdir-failover"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .description = "readdir(p) will not failover if this option is off",
+          .default_value = "on",
+        },
+        { .key = {"ensure-durability"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .description = "Afr performs fsyncs for transactions if this "
+                         "option is on to make sure the changelogs/data is "
+                         "written to the disk",
+          .default_value = "on",
         },
-	{ .key  = {"metadata-self-heal"},  
-	  .type = GF_OPTION_TYPE_BOOL
-	},
-	{ .key  = {"entry-self-heal"},  
-	  .type = GF_OPTION_TYPE_BOOL 
-	},
-	{ .key  = {"data-change-log"},  
-	  .type = GF_OPTION_TYPE_BOOL 
-	},
-	{ .key  = {"metadata-change-log"},  
-	  .type = GF_OPTION_TYPE_BOOL
-	},
-	{ .key  = {"entry-change-log"},  
-	  .type = GF_OPTION_TYPE_BOOL
-	},
-	{ .key  = {"data-lock-server-count"},  
-	  .type = GF_OPTION_TYPE_INT, 
-	  .min  = 0
-	},
-	{ .key  = {"metadata-lock-server-count"},  
-	  .type = GF_OPTION_TYPE_INT, 
-	  .min  = 0
-	},
-	{ .key  = {"entry-lock-server-count"},  
-	  .type = GF_OPTION_TYPE_INT,
-	  .min  = 0
-	},
-	{ .key  = {"strict-readdir"},
-	  .type = GF_OPTION_TYPE_BOOL,
-	},
-	{ .key  = {NULL} },
+        { .key  = {NULL} },
 };
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 8e0121609..21064db58 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -29,152 +20,316 @@
 #include "call-stub.h"
 #include "compat-errno.h"
 #include "afr-mem-types.h"
+#include "afr-self-heal-algorithm.h"
+
+#include "libxlator.h"
+#include "timer.h"
 
 #define AFR_XATTR_PREFIX "trusted.afr"
+#define AFR_PATHINFO_HEADER "REPLICATE:"
+#define AFR_SH_READDIR_SIZE_KEY "self-heal-readdir-size"
+#define AFR_SH_DATA_DOMAIN_FMT "%s:self-heal"
+
+#define AFR_LOCKEE_COUNT_MAX    3
+#define AFR_DOM_COUNT_MAX    3
 
 struct _pump_private;
 
+typedef int (*afr_expunge_done_cbk_t) (call_frame_t *frame, xlator_t *this,
+                                       int child, int32_t op_error,
+                                       int32_t op_errno);
+
+typedef int (*afr_impunge_done_cbk_t) (call_frame_t *frame, xlator_t *this,
+                                       int32_t op_error, int32_t op_errno);
+typedef int (*afr_post_remove_call_t) (call_frame_t *frame, xlator_t *this);
+
+typedef int (*afr_lock_cbk_t) (call_frame_t *frame, xlator_t *this);
+typedef void (*afr_lookup_done_cbk_t) (call_frame_t *frame, xlator_t *this,
+                                      int32_t op_ret, int32_t op_errno);
+
+typedef enum {
+        AFR_POS_UNKNOWN,
+        AFR_POS_LOCAL,
+        AFR_POS_REMOTE
+} afr_child_pos_t;
+
+typedef enum {
+        SPLIT_BRAIN = 1,
+        ALL_FOOLS = 2
+} afr_subvol_status_t;
+
+typedef enum {
+        AFR_INODE_SET_READ_CTX = 1,
+        AFR_INODE_RM_STALE_CHILDREN,
+        AFR_INODE_SET_OPENDIR_DONE,
+        AFR_INODE_GET_READ_CTX,
+        AFR_INODE_GET_OPENDIR_DONE,
+} afr_inode_op_t;
+
+typedef struct afr_inode_params_ {
+        afr_inode_op_t op;
+        union {
+                gf_boolean_t value;
+                struct {
+                        int32_t read_child;
+                        int32_t *children;
+                } read_ctx;
+        } u;
+} afr_inode_params_t;
+
+typedef enum afr_spb_state {
+        DONT_KNOW,
+        SPB,
+        NO_SPB
+} afr_spb_state_t;
+
+typedef struct afr_inode_ctx_ {
+        uint64_t masks;
+        int32_t  *fresh_children;//increasing order of latency
+        afr_spb_state_t mdata_spb;
+        afr_spb_state_t data_spb;
+        uint32_t        open_fd_count;
+} afr_inode_ctx_t;
+
+typedef enum {
+        NONE,
+        INDEX,
+        INDEX_TO_BE_HEALED,
+        FULL,
+} afr_crawl_type_t;
+
+typedef struct afr_self_heald_ {
+        gf_boolean_t            enabled;
+        gf_boolean_t            iamshd;
+        afr_crawl_type_t        *pending;
+        gf_boolean_t            *inprogress;
+        afr_child_pos_t         *pos;
+        gf_timer_t              **timer;
+        eh_t                    *healed;
+        eh_t                    *heal_failed;
+        eh_t                    *split_brain;
+        eh_t                    **statistics;
+        void                    **crawl_events;
+        char                    *node_uuid;
+        int                     timeout;
+} afr_self_heald_t;
+
 typedef struct _afr_private {
-	gf_lock_t lock;               /* to guard access to child_count, etc */
-	unsigned int child_count;     /* total number of children   */
+        gf_lock_t lock;               /* to guard access to child_count, etc */
+        unsigned int child_count;     /* total number of children   */
 
         unsigned int read_child_rr;   /* round-robin index of the read_child */
         gf_lock_t read_child_lock;    /* lock to protect above */
-        
-	xlator_t **children;
 
-        gf_lock_t root_inode_lk;
+        xlator_t **children;
+
         int first_lookup;
         inode_t *root_inode;
 
-	unsigned char *child_up;
+        unsigned char *child_up;
 
         char **pending_key;
 
-	gf_boolean_t data_self_heal;              /* on/off */
+        char         *data_self_heal;              /* on/off/open */
         char *       data_self_heal_algorithm;    /* name of algorithm */
         unsigned int data_self_heal_window_size;  /* max number of pipelined
                                                      read/writes */
 
         unsigned int background_self_heal_count;
         unsigned int background_self_heals_started;
-	gf_boolean_t metadata_self_heal;   /* on/off */
-	gf_boolean_t entry_self_heal;      /* on/off */
-
-	gf_boolean_t data_change_log;       /* on/off */
-	gf_boolean_t metadata_change_log;   /* on/off */
-	gf_boolean_t entry_change_log;      /* on/off */
+        gf_boolean_t metadata_self_heal;   /* on/off */
+        gf_boolean_t entry_self_heal;      /* on/off */
 
-	int read_child;               /* read-subvolume */
-	unsigned int favorite_child;  /* subvolume to be preferred in resolving
-					 split-brain cases */
+        gf_boolean_t data_change_log;       /* on/off */
+        gf_boolean_t metadata_change_log;   /* on/off */
+        gf_boolean_t entry_change_log;      /* on/off */
 
-	unsigned int data_lock_server_count;
-	unsigned int metadata_lock_server_count;
-	unsigned int entry_lock_server_count;
+        int read_child;               /* read-subvolume */
+        unsigned int hash_mode;       /* for when read_child is not set */
+        int favorite_child;  /* subvolume to be preferred in resolving
+                                         split-brain cases */
 
-	gf_boolean_t inodelk_trace;
-	gf_boolean_t entrylk_trace;
+        gf_boolean_t inodelk_trace;
+        gf_boolean_t entrylk_trace;
 
-	gf_boolean_t strict_readdir;
+        gf_boolean_t strict_readdir;
 
-	unsigned int wait_count;      /* # of servers to wait for success */
+        unsigned int wait_count;      /* # of servers to wait for success */
 
         uint64_t up_count;      /* number of CHILD_UPs we have seen */
         uint64_t down_count;    /* number of CHILD_DOWNs we have seen */
 
-	struct _pump_private *pump_private; /* Set if we are loaded as pump */
+        struct _pump_private *pump_private; /* Set if we are loaded as pump */
         int                   use_afr_in_pump;
+
+        pthread_mutex_t  mutex;
+        struct list_head saved_fds;   /* list of fds on which locks have succeeded */
+        gf_boolean_t      optimistic_change_log;
+        gf_boolean_t      eager_lock;
+	uint32_t          post_op_delay_secs;
+        unsigned int      quorum_count;
+
+        char                   vol_uuid[UUID_SIZE + 1];
+        int32_t                *last_event;
+        afr_self_heald_t       shd;
+        gf_boolean_t           choose_local;
+        gf_boolean_t           did_discovery;
+        gf_boolean_t           readdir_failover;
+        uint64_t               sh_readdir_size;
+        gf_boolean_t           ensure_durability;
+        char                   *sh_domain;
 } afr_private_t;
 
+typedef enum {
+        AFR_SELF_HEAL_NOT_ATTEMPTED,
+        AFR_SELF_HEAL_STARTED,
+        AFR_SELF_HEAL_FAILED,
+        AFR_SELF_HEAL_SYNC_BEGIN,
+} afr_self_heal_status;
+
 typedef struct {
+        afr_self_heal_status gfid_or_missing_entry_self_heal;
+        afr_self_heal_status metadata_self_heal;
+        afr_self_heal_status data_self_heal;
+        afr_self_heal_status entry_self_heal;
+} afr_sh_status_for_all_type;
+
+typedef enum {
+        AFR_SELF_HEAL_ENTRY,
+        AFR_SELF_HEAL_METADATA,
+        AFR_SELF_HEAL_DATA,
+        AFR_SELF_HEAL_GFID_OR_MISSING_ENTRY,
+        AFR_SELF_HEAL_INVALID = -1,
+} afr_self_heal_type;
+
+typedef enum {
+        AFR_CHECK_ALL,
+        AFR_CHECK_SPECIFIC,
+} afr_sh_fail_check_type;
+
+struct afr_self_heal_ {
         /* External interface: These are variables (some optional) that
            are set by whoever has triggered self-heal */
 
-        gf_boolean_t need_data_self_heal;
-        gf_boolean_t need_metadata_self_heal;
-        gf_boolean_t need_entry_self_heal;
+        gf_boolean_t do_data_self_heal;
+        gf_boolean_t do_metadata_self_heal;
+        gf_boolean_t do_entry_self_heal;
+        gf_boolean_t do_gfid_self_heal;
+        gf_boolean_t do_missing_entry_self_heal;
+        gf_boolean_t force_confirm_spb; /* Check for split-brains even when
+                                           self-heal is turned off */
 
         gf_boolean_t forced_merge;        /* Is this a self-heal triggered to
                                              forcibly merge the directories? */
 
-        gf_boolean_t healing_fd_opened;   /* true if caller has already
-                                             opened fd */
-
-        gf_boolean_t data_lock_held;      /* true if caller has already
-                                             acquired 0-0 lock */
-
-	fd_t *healing_fd;                 /* set if callers has opened fd */
-
         gf_boolean_t background;          /* do self-heal in background
                                              if possible */
-
         ia_type_t type;                   /* st_mode of the entry we're doing
                                              self-heal on */
+        inode_t   *inode;                 /* inode on which the self-heal is
+                                             performed on */
+        uuid_t  sh_gfid_req;                 /* gfid self-heal needs to be done
+                                             with this gfid if it is not null */
 
         /* Function to call to unwind. If self-heal is being done in the
            background, this function will be called as soon as possible. */
 
-        int (*unwind) (call_frame_t *frame, xlator_t *this);
+        int (*unwind) (call_frame_t *frame, xlator_t *this, int32_t op_ret,
+                       int32_t op_errno, int32_t sh_failed);
 
         /* End of external interface members */
 
 
-	/* array of stat's, one for each child */
-	struct iatt *buf;
+        /* array of stat's, one for each child */
+        struct iatt *buf;
+        struct iatt *parentbufs;
         struct iatt parentbuf;
-
-	/* array of xattr's, one for each child */
-	dict_t **xattr;
-
-	/* array of errno's, one for each child */
-	int *child_errno;
-
-	int32_t **pending_matrix;
-	int32_t **delta_matrix;
-
-	int *sources;
-	int source;
-	int active_source;
-	int active_sinks;
-	int *success;
-	unsigned char *locked_nodes;
+        struct iatt entrybuf;
+
+        afr_expunge_done_cbk_t expunge_done;
+        afr_impunge_done_cbk_t impunge_done;
+
+        /* array of xattr's, one for each child */
+        dict_t **xattr;
+
+        /* array containing if the lookups succeeded in the order of response
+         */
+        int32_t *success_children;
+        int     success_count;
+        /* array containing the fresh children found in the self-heal process */
+        int32_t *fresh_children;
+        /* array containing the fresh children found in the parent lookup */
+        int32_t *fresh_parent_dirs;
+        /* array of errno's, one for each child */
+        int *child_errno;
+        /*loc used for lookup*/
+        loc_t lookup_loc;
+        int32_t lookup_flags;
+        afr_lookup_done_cbk_t lookup_done;
+
+        int32_t **pending_matrix;
+        int32_t **delta_matrix;
+
+        int32_t op_ret;
+        int32_t op_errno;
+
+        int *sources;
+        int source;
+        int active_source;
+        int active_sinks;
+        unsigned char *success;
+        unsigned char *locked_nodes;
         int lock_count;
 
-        mode_t impunging_entry_mode;
         const char *linkname;
-
-	int   op_failed;
-
-	int   file_has_holes;
-	blksize_t block_size;
-	off_t file_size;
-	off_t offset;
-
-	loc_t parent_loc;
-
+        gf_boolean_t entries_skipped;
+
+        gf_boolean_t actual_sh_started;
+        gf_boolean_t sync_done;
+        gf_boolean_t data_lock_held;
+        gf_boolean_t sh_dom_lock_held;
+        gf_boolean_t eof_reached;
+        fd_t  *healing_fd;
+        int   file_has_holes;
+        blksize_t block_size;
+        off_t file_size;
+        off_t offset;
+        unsigned char *write_needed;
+        uint8_t *checksum;
+        afr_post_remove_call_t post_remove_call;
+
+        char    *data_sh_info;
+        char    *metadata_sh_info;
+
+        loc_t parent_loc;
         call_frame_t *orig_frame;
+        call_frame_t *old_loop_frame;
         gf_boolean_t unwound;
 
-        /* private data for the particular self-heal algorithm */
-        void *private;
+        afr_sh_algo_private_t *private;
+        afr_sh_status_for_all_type  afr_all_sh_status;
+        afr_self_heal_type       sh_type_in_action;
 
-        int (*flush_self_heal_cbk) (call_frame_t *frame, xlator_t *this);
-
-	int (*completion_cbk) (call_frame_t *frame, xlator_t *this);
+        struct afr_sh_algorithm  *algo;
+        afr_lock_cbk_t data_lock_success_handler;
+        afr_lock_cbk_t data_lock_failure_handler;
+	gf_boolean_t data_lock_block;
+        int (*completion_cbk) (call_frame_t *frame, xlator_t *this);
+        int (*sh_data_algo_start) (call_frame_t *frame, xlator_t *this);
         int (*algo_completion_cbk) (call_frame_t *frame, xlator_t *this);
         int (*algo_abort_cbk) (call_frame_t *frame, xlator_t *this);
+        void (*gfid_sh_success_cbk) (call_frame_t *sh_frame, xlator_t *this);
 
-	call_frame_t *sh_frame;
-} afr_self_heal_t;
+        call_frame_t *sh_frame;
+};
 
+typedef struct afr_self_heal_ afr_self_heal_t;
 
 typedef enum {
-	AFR_DATA_TRANSACTION,          /* truncate, write, ... */
-	AFR_METADATA_TRANSACTION,      /* chmod, chown, ... */
-	AFR_ENTRY_TRANSACTION,         /* create, rmdir, ... */
-	AFR_ENTRY_RENAME_TRANSACTION,  /* rename */
-	AFR_FLUSH_TRANSACTION,         /* flush */
+        AFR_DATA_TRANSACTION,          /* truncate, write, ... */
+        AFR_METADATA_TRANSACTION,      /* chmod, chown, ... */
+        AFR_ENTRY_TRANSACTION,         /* create, rmdir, ... */
+        AFR_ENTRY_RENAME_TRANSACTION,  /* rename */
 } afr_transaction_type;
 
 typedef enum {
@@ -215,9 +370,8 @@ static inline int
 afr_index_for_transaction_type (afr_transaction_type type)
 {
         switch (type) {
-                
+
         case AFR_DATA_TRANSACTION:
-        case AFR_FLUSH_TRANSACTION:
                 return 0;
 
         case AFR_METADATA_TRANSACTION:
@@ -231,16 +385,31 @@ afr_index_for_transaction_type (afr_transaction_type type)
         return -1;  /* make gcc happy */
 }
 
+typedef struct {
+        loc_t                   loc;
+        char                    *basename;
+        unsigned char           *locked_nodes;
+        int                     locked_count;
 
-typedef enum {
-        AFR_CHILD_UP_FLUSH,
-        AFR_CHILD_DOWN_FLUSH,
-} afr_flush_type;
+} afr_entry_lockee_t;
+
+int
+afr_entry_lockee_cmp (const void *l1, const void *l2);
+
+typedef struct {
+        char    *domain; /* Domain on which inodelk is taken */
+        struct gf_flock flock;
+        unsigned char *locked_nodes;
+        int32_t lock_count;
+} afr_inodelk_t;
 
 typedef struct {
         loc_t *lk_loc;
-        struct flock lk_flock;
 
+        int                     lockee_count;
+        afr_entry_lockee_t      lockee[AFR_LOCKEE_COUNT_MAX];
+
+        afr_inodelk_t       inodelk[AFR_DOM_COUNT_MAX];
         const char *lk_basename;
         const char *lower_basename;
         const char *higher_basename;
@@ -249,402 +418,421 @@ typedef struct {
 
         unsigned char *locked_nodes;
         unsigned char *lower_locked_nodes;
-        unsigned char *inode_locked_nodes;
-        unsigned char *entry_locked_nodes;
 
         selfheal_lk_type_t selfheal_lk_type;
         transaction_lk_type_t transaction_lk_type;
 
         int32_t lock_count;
-        int32_t inodelk_lock_count;
         int32_t entrylk_lock_count;
 
         uint64_t lock_number;
         int32_t lk_call_count;
+        int32_t lk_expected_count;
+        int32_t lk_attempted_count;
 
         int32_t lock_op_ret;
         int32_t lock_op_errno;
+        afr_lock_cbk_t lock_cbk;
+        char *domain; /* Domain on which inode/entry lock/unlock in progress.*/
+} afr_internal_lock_t;
 
-        int (*lock_cbk) (call_frame_t *, xlator_t *);
+typedef struct _afr_locked_fd {
+        fd_t  *fd;
+        struct list_head list;
+} afr_locked_fd_t;
 
-} afr_internal_lock_t;
+struct afr_reply {
+	int	valid;
+	int32_t	op_ret;
+	int32_t	op_errno;
+};
 
 typedef struct _afr_local {
-	unsigned int call_count;
-	unsigned int success_count;
-	unsigned int enoent_count;
+        int     uid;
+        int     gid;
+        unsigned int call_count;
+        unsigned int success_count;
+        unsigned int enoent_count;
+        uint32_t     open_fd_count;
+        gf_boolean_t update_open_fd_count;
 
 
-	unsigned int govinda_gOvinda;
+        unsigned int unhealable;
 
-	unsigned int read_child_index;
+        unsigned int read_child_index;
         unsigned char read_child_returned;
         unsigned int first_up_child;
 
-        pid_t saved_pid;
+	gf_lkowner_t  saved_lk_owner;
+
+        int32_t op_ret;
+        int32_t op_errno;
 
-	int32_t op_ret;
-	int32_t op_errno;
+        int32_t **pending;
 
-	int32_t **pending;
+        loc_t loc;
+        loc_t newloc;
 
-	loc_t loc;
-	loc_t newloc;
+        fd_t *fd;
 
-	fd_t *fd;
+        glusterfs_fop_t fop;
 
-	glusterfs_fop_t fop;
+        unsigned char *child_up;
+        int32_t       *fresh_children; //in the order of response
 
-	unsigned char *child_up; 
+        int32_t *child_errno;
 
-	int32_t *child_errno;
-	
-	dict_t  *xattr_req;
-	int      open_fd_count;
+        dict_t  *xattr_req;
 
-	int32_t  inodelk_count;
-	int32_t  entrylk_count;
+        int32_t  inodelk_count;
+        int32_t  entrylk_count;
 
         afr_internal_lock_t internal_lock;
 
+        afr_locked_fd_t *locked_fd;
+        int32_t          source_child;
+        int32_t          lock_recovery_child;
+
         dict_t  *dict;
+        int      optimistic_change_log;
+	gf_boolean_t      delayed_post_op;
 
-        int (*up_down_flush_cbk) (call_frame_t *, xlator_t *);
 
-	/* 
-	   This struct contains the arguments for the "continuation"
-	   (scheme-like) of fops
+	/* Is the current writev() going to perform a stable write?
+	   i.e, is fd->flags or @flags writev param have O_SYNC or
+	   O_DSYNC?
 	*/
+        gf_boolean_t      stable_write;
 
-	int   op;
-	struct {
-		struct {
-			unsigned char buf_set;
-			struct statvfs buf;
-		} statfs;
+        /* This write appended to the file. Nnot necessarily O_APPEND,
+           just means the offset of write was at the end of file.
+        */
+        gf_boolean_t      append_write;
 
-		struct {
-			inode_t *inode;
-			struct iatt buf;
-                        struct iatt read_child_buf;
+        int allow_sh_for_running_transaction;
+
+
+        /* This struct contains the arguments for the "continuation"
+           (scheme-like) of fops
+        */
+
+        int   op;
+        struct {
+                struct {
+                        unsigned char buf_set;
+                        struct statvfs buf;
+                } statfs;
+
+                struct {
+                        uint32_t parent_entrylk;
+                        uuid_t  gfid_req;
+                        inode_t *inode;
+                        struct iatt buf;
                         struct iatt postparent;
-                        ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			dict_t *xattr;
-			dict_t **xattrs;
-                        gf_boolean_t is_revalidate;
-		} lookup;
+                        dict_t **xattrs;
+                        dict_t *xattr;
+                        struct iatt *postparents;
+                        struct iatt *bufs;
+                        int32_t read_child;
+                        int32_t *sources;
+                        int32_t *success_children;
+                        int32_t **pending_matrix;
+                        gf_boolean_t fresh_lookup;
+                        gf_boolean_t possible_spb;
+                } lookup;
 
-		struct {
-			int32_t flags;
-                        int32_t wbflags;
-		} open;
+                struct {
+                        int32_t flags;
+                } open;
 
-		struct {
-			int32_t cmd;
-                        struct flock user_flock;
-                        struct flock ret_flock;
-			unsigned char *locked_nodes;
-		} lk;
+                struct {
+                        int32_t cmd;
+                        struct gf_flock user_flock;
+                        struct gf_flock ret_flock;
+                        unsigned char *locked_nodes;
+                } lk;
 
-		/* inode read */
+                /* inode read */
 
-		struct {
-			int32_t mask;
-			int last_tried;  /* index of the child we tried previously */
-		} access;
+                struct {
+                        int32_t mask;
+                        int last_index;  /* index of the child we tried previously */
+                } access;
 
-		struct {
-			int last_tried;
-			ino_t ino;
-		} stat;
+                struct {
+                        int last_index;
+                } stat;
 
-		struct {
-			int last_tried;
-			ino_t ino;
-		} fstat;
+                struct {
+                        int last_index;
+                } fstat;
 
-		struct {
-			size_t size;
-			int last_tried;
-                        ino_t ino;
-		} readlink;
+                struct {
+                        size_t size;
+                        int last_index;
+                } readlink;
 
-		struct {
-			char *name;
-			int last_tried;
-		} getxattr;
+                struct {
+                        char *name;
+                        int last_index;
+                        long xattr_len;
+                } getxattr;
 
-		struct {
-                        ino_t ino;
-			size_t size;
-			off_t offset;
-			int last_tried;
-		} readv;
+                struct {
+                        size_t size;
+                        off_t offset;
+                        int last_index;
+                        uint32_t flags;
+                } readv;
 
-		/* dir read */
+                /* dir read */
 
-		struct {
-			int success_count;
-			int32_t op_ret;
-			int32_t op_errno;
+                struct {
+                        int success_count;
+                        int32_t op_ret;
+                        int32_t op_errno;
 
                         uint32_t *checksum;
-		} opendir;
-
-		struct {
-			int32_t op_ret;
-			int32_t op_errno;
-			size_t size;
-			off_t offset;
+                } opendir;
 
+                struct {
+                        int32_t op_ret;
+                        int32_t op_errno;
+                        size_t size;
+                        off_t offset;
+                        dict_t *dict;
                         gf_boolean_t failed;
-			int last_tried;
-		} readdir;
-
-		struct {
-			int32_t op_ret;
-			int32_t op_errno;
+                        int last_index;
+                } readdir;
+                /* inode write */
 
-			size_t size;
-			off_t offset;
-			int32_t flag;
+                struct {
+                        struct iatt prebuf;
+                        struct iatt postbuf;
+                } inode_wfop; //common structure for all inode-write-fops
 
-			int last_tried;
-		} getdents;
+                struct {
+                        int32_t op_ret;
 
-		/* inode write */
+                        struct iovec *vector;
+                        struct iobref *iobref;
+                        int32_t count;
+                        off_t offset;
+                        uint32_t flags;
+                } writev;
 
-		struct {
-			ino_t ino;
-			struct iatt prebuf;
-			struct iatt postbuf;
+                struct {
+                        off_t offset;
+                } truncate;
 
-			int32_t op_ret;
+                struct {
+                        off_t offset;
+                } ftruncate;
 
-			struct iovec *vector;
-			struct iobref *iobref;
-			int32_t count;
-			off_t offset;
-		} writev;
+                struct {
+                        struct iatt in_buf;
+                        int32_t valid;
+                } setattr;
 
                 struct {
-                        ino_t ino;
-                        struct iatt prebuf;
-                        struct iatt postbuf;
-                } fsync;
+                        struct iatt in_buf;
+                        int32_t valid;
+                } fsetattr;
 
-		struct {
-			ino_t ino;
-			off_t offset;
-			struct iatt prebuf;
-                        struct iatt postbuf;
-		} truncate;
+                struct {
+                        dict_t *dict;
+                        int32_t flags;
+                } setxattr;
 
-		struct {
-			ino_t ino;
-			off_t offset;
-			struct iatt prebuf;
-                        struct iatt postbuf;
-		} ftruncate;
+                struct {
+                        dict_t *dict;
+                        int32_t flags;
+                } fsetxattr;
 
-		struct {
-			ino_t ino;
-			struct iatt in_buf;
-                        int32_t valid;
-                        struct iatt preop_buf;
-                        struct iatt postop_buf;
-		} setattr;
+                struct {
+                        char *name;
+                } removexattr;
 
-		struct {
-			ino_t ino;
-			struct iatt in_buf;
-                        int32_t valid;
-                        struct iatt preop_buf;
-                        struct iatt postop_buf;
-		} fsetattr;
+                struct {
+                        dict_t *xattr;
+                } xattrop;
 
-		struct {
-			dict_t *dict;
-			int32_t flags;
-		} setxattr;
+                struct {
+                        dict_t *xattr;
+                } fxattrop;
 
-		struct {
-			char *name;
-		} removexattr;
+                /* dir write */
 
-		/* dir write */
-		
-		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			fd_t *fd;
-                        dict_t *params;
-			int32_t flags;
-			mode_t mode;
-			inode_t *inode;
-			struct iatt buf;
+                struct {
+                        inode_t *inode;
+                        struct iatt buf;
                         struct iatt preparent;
                         struct iatt postparent;
-                        struct iatt read_child_buf;
-		} create;
+                        struct iatt prenewparent;
+                        struct iatt postnewparent;
+                } dir_fop; //common structure for all dir fops
 
-		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			dev_t dev;
-			mode_t mode;
+                struct {
+                        fd_t *fd;
                         dict_t *params;
-			inode_t *inode;
-			struct iatt buf;
-                        struct iatt preparent;
-                        struct iatt postparent;
-                        struct iatt read_child_buf;
-		} mknod;
+                        int32_t flags;
+                        mode_t mode;
+                } create;
 
-		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			int32_t mode;
+                struct {
+                        dev_t dev;
+                        mode_t mode;
                         dict_t *params;
-			inode_t *inode;
-			struct iatt buf;
-                        struct iatt read_child_buf;
-                        struct iatt preparent;
-                        struct iatt postparent;
-		} mkdir;
+                } mknod;
 
-		struct {
-                        ino_t parent_ino;
-			int32_t op_ret;
-			int32_t op_errno;
-                        struct iatt preparent;
-                        struct iatt postparent;
-		} unlink;
+                struct {
+                        int32_t mode;
+                        dict_t *params;
+                } mkdir;
 
-		struct {
-                        ino_t parent_ino;
-			int32_t op_ret;
-			int32_t op_errno;
-                        struct iatt preparent;
-                        struct iatt postparent;
-		} rmdir;
+                struct {
+                        int flags;
+                } rmdir;
 
-		struct {
-                        ino_t oldparent_ino;
-                        ino_t newparent_ino;
-			ino_t ino;
-			struct iatt buf;
-                        struct iatt read_child_buf;
-                        struct iatt preoldparent;
-                        struct iatt prenewparent;
-                        struct iatt postoldparent;
-                        struct iatt postnewparent;
-		} rename;
+                struct {
+                        dict_t *params;
+                        char *linkpath;
+                } symlink;
 
 		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			inode_t *inode;
-			struct iatt buf;
-                        struct iatt read_child_buf;
-                        struct iatt preparent;
-                        struct iatt postparent;
-		} link;
+			int32_t mode;
+			off_t offset;
+			size_t len;
+		} fallocate;
 
 		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			inode_t *inode;
-                        dict_t *params;
-			struct iatt buf;
-                        struct iatt read_child_buf;
-			char *linkpath;
-                        struct iatt preparent;
-                        struct iatt postparent;
-		} symlink;
+			off_t offset;
+			size_t len;
+		} discard;
 
-		struct {
-			int32_t flags;
-			dir_entry_t *entries;
-			int32_t count;
-		} setdents;
-	} cont;
-	
-	struct {
-		off_t start, len;
+                struct {
+                        off_t offset;
+                        size_t len;
+                        struct iatt prebuf;
+                        struct iatt postbuf;
+                } zerofill;
 
-		char *basename;
-		char *new_basename;
 
-		loc_t parent_loc;
-		loc_t new_parent_loc;
+        } cont;
 
-		afr_transaction_type type;
+        struct {
+                off_t start, len;
 
-		int success_count;
-		int erase_pending;
-		int failure_count;
+                gf_boolean_t    eager_lock_on;
+                int *eager_lock;
 
-		int last_tried;
-		int32_t *child_errno;
+                char *basename;
+                char *new_basename;
 
-		call_frame_t *main_frame;
+                loc_t parent_loc;
+                loc_t new_parent_loc;
 
-		int (*fop) (call_frame_t *frame, xlator_t *this);
+                afr_transaction_type type;
 
-		int (*done) (call_frame_t *frame, xlator_t *this);
+		/* pre-compute the post piggyback status before
+		   entering POST-OP phase
+		*/
+		int              *postop_piggybacked;
 
-		int (*resume) (call_frame_t *frame, xlator_t *this);
+		/* stub to resume on destruction
+		   of the transaction frame */
+		call_stub_t      *resume_stub;
 
-		int (*unwind) (call_frame_t *frame, xlator_t *this);
+		struct list_head  eager_locked;
+
+                int32_t         **txn_changelog;//changelog after pre+post ops
+                unsigned char   *pre_op;
+
+                call_frame_t *main_frame;
+
+                int (*fop) (call_frame_t *frame, xlator_t *this);
+
+                int (*done) (call_frame_t *frame, xlator_t *this);
+
+                int (*resume) (call_frame_t *frame, xlator_t *this);
+
+                int (*unwind) (call_frame_t *frame, xlator_t *this);
 
                 /* post-op hook */
-                int (*post_post_op) (call_frame_t *frame, xlator_t *this);
-	} transaction;
+        } transaction;
+
+        afr_self_heal_t self_heal;
+
+        struct marker_str     marker;
 
-	afr_self_heal_t self_heal;
+        /* extra data for fops */
+        dict_t         *xdata_req;
+        dict_t         *xdata_rsp;
+
+        mode_t          umask;
+        int             xflag;
+        gf_boolean_t    do_discovery;
+	struct afr_reply *replies;
 } afr_local_t;
 
+typedef enum {
+        AFR_FD_NOT_OPENED,
+        AFR_FD_OPENED,
+        AFR_FD_OPENING
+} afr_fd_open_status_t;
 
 typedef struct {
-        unsigned char *pre_op_done;
-        unsigned char *opened_on;     /* which subvolumes the fd is open on */
-        unsigned char *child_failed;
+        unsigned int *pre_op_done;
+        afr_fd_open_status_t *opened_on; /* which subvolumes the fd is open on */
+        unsigned int *pre_op_piggyback;
+
+        unsigned int *lock_piggyback;
+        unsigned int *lock_acquired;
+
         int flags;
-        int32_t wbflags;
         uint64_t up_count;   /* number of CHILD_UPs this fd has seen */
         uint64_t down_count; /* number of CHILD_DOWNs this fd has seen */
 
         int32_t last_tried;
+
+        int  hit, miss;
         gf_boolean_t failed_over;
         struct list_head entries; /* needed for readdir failover */
+
+        unsigned char *locked_on; /* which subvolumes locks have been successful */
+
+	/* used for delayed-post-op optimization */
+	pthread_mutex_t    delay_lock;
+	gf_timer_t        *delay_timer;
+	call_frame_t      *delay_frame;
+        int               call_child;
+
+	/* set if any write on this fd was a non stable write
+	   (i.e, without O_SYNC or O_DSYNC)
+	*/
+	gf_boolean_t      witnessed_unstable_write;
+
+	/* list of frames currently in progress */
+	struct list_head  eager_locked;
 } afr_fd_ctx_t;
 
 
 /* try alloc and if it fails, goto label */
-#define ALLOC_OR_GOTO(var, type, label) do {                     \
-		var = GF_CALLOC (sizeof (type), 1,               \
-                                 gf_afr_mt_##type);              \
-		if (!var) {                                      \
-			gf_log (this->name, GF_LOG_ERROR,        \
-				"out of memory :(");             \
-			op_errno = ENOMEM;                       \
-			goto label;                              \
-		}                                                \
-	} while (0);
+#define AFR_LOCAL_ALLOC_OR_GOTO(var, label) do {                    \
+                var = mem_get0 (THIS->local_pool);                  \
+                if (!var) {                                         \
+                        gf_log (this->name, GF_LOG_ERROR,           \
+                                "out of memory :(");                \
+                        op_errno = ENOMEM;                          \
+                        goto label;                                 \
+                }                                                   \
+        } while (0);
 
 
 /* did a call fail due to a child failing? */
-#define child_went_down(op_ret, op_errno) (((op_ret) < 0) &&	      \
-					   ((op_errno == ENOTCONN) || \
-					    (op_errno == EBADFD)))
+#define child_went_down(op_ret, op_errno) (((op_ret) < 0) &&            \
+                                           ((op_errno == ENOTCONN) ||   \
+                                            (op_errno == EBADFD)))
 
 #define afr_fop_failed(op_ret, op_errno) ((op_ret) == -1)
 
@@ -658,11 +846,27 @@ int
 pump_command_reply (call_frame_t *frame, xlator_t *this);
 
 int32_t
-afr_notify (xlator_t *this, int32_t event,
-            void *data, ...);
+afr_notify (xlator_t *this, int32_t event, void *data, void *data2);
+
+int
+afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
+                       loc_t *loc, char *basename, int child_count);
 
 void
-afr_set_lk_owner (call_frame_t *frame, xlator_t *this);
+afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock);
+
+int
+afr_attempt_lock_recovery (xlator_t *this, int32_t child_index);
+
+int
+afr_save_locked_fd (xlator_t *this, fd_t *fd);
+
+int
+afr_mark_locked_nodes (xlator_t *this, fd_t *fd,
+                       unsigned char *locked_nodes);
+
+void
+afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner);
 
 int
 afr_set_lock_number (call_frame_t *frame, xlator_t *this);
@@ -686,32 +890,46 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this);
 int
 afr_internal_lock_finish (call_frame_t *frame, xlator_t *this);
 
+int
+afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src, char *dom,
+                          unsigned int child_count);
 
 int pump_start (call_frame_t *frame, xlator_t *this);
 
 int
-afr_fd_ctx_set (xlator_t *this, fd_t *fd);
+__afr_fd_ctx_set (xlator_t *this, fd_t *fd);
 
-uint64_t
-afr_read_child (xlator_t *this, inode_t *inode);
+int
+afr_fd_ctx_set (xlator_t *this, fd_t *fd);
 
-void
-afr_set_read_child (xlator_t *this, inode_t *inode, int32_t read_child);
+int32_t
+afr_inode_get_read_ctx (xlator_t *this, inode_t *inode, int32_t *fresh_children);
 
 void
-afr_build_parent_loc (loc_t *parent, loc_t *child);
+afr_inode_set_read_ctx (xlator_t *this, inode_t *inode, int32_t read_child,
+                        int32_t *fresh_children);
 
 int
-afr_up_children_count (int child_count, unsigned char *child_up);
+afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno);
 
-int
-afr_locked_nodes_count (unsigned char *locked_nodes, int child_count);
+unsigned int
+afr_up_children_count (unsigned char *child_up, unsigned int child_count);
 
-ino64_t
-afr_itransform (ino64_t ino, int child_count, int child_index);
+unsigned int
+afr_locked_children_count (unsigned char *children, unsigned int child_count);
+
+unsigned int
+afr_pre_op_done_children_count (unsigned char *pre_op,
+                                unsigned int child_count);
+
+gf_boolean_t
+afr_is_fresh_lookup (loc_t *loc, xlator_t *this);
+
+void
+afr_update_loc_gfids (loc_t *loc, struct iatt *buf, struct iatt *postparent);
 
 int
-afr_deitransform (ino64_t ino, int child_count);
+afr_locked_nodes_count (unsigned char *locked_nodes, int child_count);
 
 void
 afr_local_cleanup (afr_local_t *local, xlator_t *this);
@@ -719,23 +937,21 @@ afr_local_cleanup (afr_local_t *local, xlator_t *this);
 int
 afr_frame_return (call_frame_t *frame);
 
-uint64_t
+gf_boolean_t
 afr_is_split_brain (xlator_t *this, inode_t *inode);
 
 void
-afr_set_split_brain (xlator_t *this, inode_t *inode, gf_boolean_t set);
+afr_set_split_brain (xlator_t *this, inode_t *inode, afr_spb_state_t mdata_spb,
+                     afr_spb_state_t data_spb);
 
 int
 afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-          fd_t *fd, int32_t wbflags);
-
-int
-afr_up_down_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, afr_flush_type type);
+          fd_t *fd, dict_t *xdata);
 
 void
 afr_set_opendir_done (xlator_t *this, inode_t *inode);
 
-uint64_t
+gf_boolean_t
 afr_is_opendir_done (xlator_t *this, inode_t *inode);
 
 void
@@ -744,160 +960,253 @@ afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this);
 int
 afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd);
 
-#define AFR_STACK_UNWIND(fop, frame, params ...)        \
-	do {						\
-		afr_local_t *__local = NULL;		\
-		xlator_t    *__this = NULL;		\
-                if (frame) {                            \
-                        __local = frame->local;		\
-                        __this = frame->this;           \
-                        frame->local = NULL;            \
+int
+afr_launch_openfd_self_heal (call_frame_t *frame, xlator_t *this, fd_t *fd);
+
+#define AFR_STACK_UNWIND(fop, frame, params ...)                \
+        do {                                                    \
+                afr_local_t *__local = NULL;                    \
+                xlator_t    *__this = NULL;                     \
+                if (frame) {                                    \
+                        __local = frame->local;                 \
+                        __this = frame->this;                   \
+                        frame->local = NULL;                    \
+                }                                               \
+                STACK_UNWIND_STRICT (fop, frame, params);       \
+                if (__local) {                                  \
+                        afr_local_cleanup (__local, __this);    \
+                        mem_put (__local);                      \
+                }                                               \
+        } while (0)
+
+#define AFR_STACK_DESTROY(frame)                                \
+        do {                                                    \
+                afr_local_t *__local = NULL;                    \
+                xlator_t    *__this = NULL;                     \
+                __local = frame->local;                         \
+                __this = frame->this;                           \
+                frame->local = NULL;                            \
+                STACK_DESTROY (frame->root);                    \
+                if (__local) {                                  \
+                        afr_local_cleanup (__local, __this);    \
+                        mem_put (__local);                      \
                 }                                               \
-		STACK_UNWIND_STRICT (fop, frame, params);       \
-		afr_local_cleanup (__local, __this);            \
-		GF_FREE (__local);				\
-        } while (0);
-
-#define AFR_STACK_DESTROY(frame)			\
-	do {						\
-		afr_local_t *__local = NULL;		\
-		xlator_t    *__this = NULL;		\
-		__local = frame->local;			\
-		__this = frame->this;			\
-		frame->local = NULL;                    \
-		STACK_DESTROY (frame->root);		\
-		afr_local_cleanup (__local, __this);	\
-		GF_FREE (__local);			\
         } while (0);
 
+#define AFR_NUM_CHANGE_LOGS            3 /*data + metadata + entry*/
 /* allocate and return a string that is the basename of argument */
-static inline char * 
-AFR_BASENAME (const char *str)						
+static inline char *
+AFR_BASENAME (const char *str)
 {
-	char *__tmp_str = NULL;				
-	char *__basename_str = NULL;			
-	__tmp_str = gf_strdup (str);			
-	__basename_str = gf_strdup (basename (__tmp_str));	
-	GF_FREE (__tmp_str);
-	return __basename_str;
+        char *__tmp_str = NULL;
+        char *__basename_str = NULL;
+        __tmp_str = gf_strdup (str);
+        __basename_str = gf_strdup (basename (__tmp_str));
+        GF_FREE (__tmp_str);
+        return __basename_str;
 }
 
-/* initialize local_t */
-static inline int
-AFR_LOCAL_INIT (afr_local_t *local, afr_private_t *priv)
-{
-	local->child_up = GF_CALLOC (sizeof (*local->child_up),
-			       	     priv->child_count,
-                                     gf_afr_mt_char);
-	if (!local->child_up) {
-		return -ENOMEM;
-	}
+int
+afr_transaction_local_init (afr_local_t *local, xlator_t *this);
 
-	memcpy (local->child_up, priv->child_up, 
-		sizeof (*local->child_up) * priv->child_count);
+int32_t
+afr_marker_getxattr (call_frame_t *frame, xlator_t *this,
+                     loc_t *loc, const char *name,afr_local_t *local, afr_private_t *priv );
 
+int32_t *
+afr_children_create (int32_t child_count);
 
-	local->call_count = afr_up_children_count (priv->child_count, local->child_up);
-	if (local->call_count == 0)
-		return -ENOTCONN;
+int
+afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno);
 
-	local->transaction.erase_pending = 1;
+int
+afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count,
+                        transaction_lk_type_t lk_type);
 
-	local->op_ret = -1;
-	local->op_errno = EUCLEAN;
+int
+afr_first_up_child (unsigned char *child_up, size_t child_count);
 
-        local->internal_lock.lock_op_ret   = -1;
-        local->internal_lock.lock_op_errno = EUCLEAN;
+int
+afr_select_read_child_from_policy (int32_t *fresh_children, int32_t child_count,
+                                   int32_t prev_read_child,
+                                   int32_t config_read_child, int32_t *sources,
+                                   unsigned int hmode, uuid_t gfid);
 
+void
+afr_set_read_ctx_from_policy (xlator_t *this, inode_t *inode,
+                              int32_t *fresh_children, int32_t prev_read_child,
+                              int32_t config_read_child, uuid_t gfid);
 
-	return 0;
-}
+int32_t
+afr_get_call_child (xlator_t *this, unsigned char *child_up, int32_t read_child,
+                    int32_t *fresh_children,
+                    int32_t *call_child, int32_t *last_index);
+
+int32_t
+afr_next_call_child (int32_t *fresh_children, unsigned char *child_up,
+                     size_t child_count, int32_t *last_index,
+                     int32_t read_child);
+void
+afr_get_fresh_children (int32_t *success_children, int32_t *sources,
+                        int32_t *children, unsigned int child_count);
+void
+afr_children_add_child (int32_t *children, int32_t child,
+                              int32_t child_count);
+void
+afr_children_rm_child (int32_t *children, int32_t child,
+                             int32_t child_count);
+void
+afr_reset_children (int32_t *children, int32_t child_count);
+int32_t
+afr_most_important_error(int32_t old_errno, int32_t new_errno,
+			 gf_boolean_t eio);
+int
+afr_errno_count (int32_t *children, int *child_errno,
+                 unsigned int child_count, int32_t op_errno);
+int
+afr_get_children_count (int32_t *children, unsigned int child_count);
+gf_boolean_t
+afr_is_child_present (int32_t *success_children, int32_t child_count,
+                      int32_t child);
+void
+afr_update_gfid_from_iatts (uuid_t uuid, struct iatt *bufs,
+                            int32_t *success_children,
+                            unsigned int child_count);
+void
+afr_reset_xattr (dict_t **xattr, unsigned int child_count);
+gf_boolean_t
+afr_conflicting_iattrs (struct iatt *bufs, int32_t *success_children,
+                        unsigned int child_count, const char *path,
+                        const char *xlator_name);
+unsigned int
+afr_gfid_missing_count (const char *xlator_name, int32_t *children,
+                        struct iatt *bufs, unsigned int child_count,
+                        const char *path);
+void
+afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req, const char *path);
+void
+afr_children_copy (int32_t *dst, int32_t *src, unsigned int child_count);
+afr_transaction_type
+afr_transaction_type_get (ia_type_t ia_type);
+int32_t
+afr_resultant_errno_get (int32_t *children,
+                         int *child_errno, unsigned int child_count);
+void
+afr_inode_rm_stale_children (xlator_t *this, inode_t *inode,
+                             int32_t *stale_children);
+void
+afr_launch_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode,
+                      gf_boolean_t background, ia_type_t ia_type, char *reason,
+                      void (*gfid_sh_success_cbk) (call_frame_t *sh_frame,
+                                                   xlator_t *this),
+                      int (*unwind) (call_frame_t *frame, xlator_t *this,
+                                     int32_t op_ret, int32_t op_errno,
+                                     int32_t sh_failed));
+void
+afr_fix_open (xlator_t *this, fd_t *fd, size_t need_open_count, int *need_open);
 
+void
+afr_open_fd_fix (fd_t *fd, xlator_t *this);
+int
+afr_set_elem_count_get (unsigned char *elems, int child_count);
 
-/**
- * first_up_child - return the index of the first child that is up
- */
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this);
 
-static inline int
-afr_first_up_child (afr_private_t *priv)
-{
-	xlator_t ** children = NULL;
-	int         ret      = -1;
-	int         i        = 0;
-
-	LOCK (&priv->lock);
-	{
-		children = priv->children;
-		for (i = 0; i < priv->child_count; i++) {
-			if (priv->child_up[i]) {
-				ret = i;
-				break;
-			}
-		}
-	}
-	UNLOCK (&priv->lock);
-
-	return ret;
-}
+gf_boolean_t
+afr_open_only_data_self_heal (char *data_self_heal);
 
+gf_boolean_t
+afr_data_self_heal_enabled (char *data_self_heal);
 
-static inline int
-afr_transaction_local_init (afr_local_t *local, afr_private_t *priv)
-{
-        int i;
-
-        local->first_up_child = afr_first_up_child (priv);
-
-	local->child_errno = GF_CALLOC (sizeof (*local->child_errno),
-				        priv->child_count,
-                                        gf_afr_mt_int32_t);
-	if (!local->child_errno) {
-		return -ENOMEM;
-	}
-
-	local->pending = GF_CALLOC (sizeof (*local->pending),
-                                    priv->child_count,
-                                    gf_afr_mt_int32_t);
-
-	if (!local->pending) {
-		return -ENOMEM;
-	}
-
-        for (i = 0; i < priv->child_count; i++) {
-                local->pending[i] = GF_CALLOC (sizeof (*local->pending[i]),
-                                               3, /* data + metadata + entry */
-                                               gf_afr_mt_int32_t);
-                if (!local->pending[i])
-                        return -ENOMEM;
-        }
+void
+afr_set_low_priority (call_frame_t *frame);
+int
+afr_child_fd_ctx_set (xlator_t *this, fd_t *fd, int32_t child,
+                      int flags);
 
-	local->internal_lock.inode_locked_nodes =
-                GF_CALLOC (sizeof (*local->internal_lock.inode_locked_nodes),
-                           priv->child_count,
-                           gf_afr_mt_char);
+gf_boolean_t
+afr_have_quorum (char *logname, afr_private_t *priv);
 
-	local->internal_lock.entry_locked_nodes =
-                GF_CALLOC (sizeof (*local->internal_lock.entry_locked_nodes),
-                           priv->child_count,
-                           gf_afr_mt_char);
+void
+afr_matrix_cleanup (int32_t **pending, unsigned int m);
 
-	local->internal_lock.locked_nodes =
-                GF_CALLOC (sizeof (*local->internal_lock.locked_nodes),
-                           priv->child_count,
-                           gf_afr_mt_char);
+int32_t**
+afr_matrix_create (unsigned int m, unsigned int n);
 
-	local->internal_lock.lower_locked_nodes
-                = GF_CALLOC (sizeof (*local->internal_lock.lower_locked_nodes),
-                             priv->child_count,
-                             gf_afr_mt_char);
+gf_boolean_t
+afr_is_errno_set (int *child_errno, int child);
 
-	local->transaction.child_errno = GF_CALLOC (sizeof (*local->transaction.child_errno),
-					            priv->child_count,
-                                                    gf_afr_mt_int32_t);
+gf_boolean_t
+afr_is_errno_unset (int *child_errno, int child);
 
-        local->internal_lock.transaction_lk_type = AFR_TRANSACTION_LK;
+gf_boolean_t
+afr_is_fd_fixable (fd_t *fd);
 
-	return 0;
-}
+void
+afr_prepare_new_entry_pending_matrix (int32_t **pending,
+                                      gf_boolean_t (*is_pending) (int *, int),
+                                      int *ctx, struct iatt *buf,
+                                      unsigned int child_count);
+void
+afr_xattr_array_destroy (dict_t **xattr, unsigned int child_count);
+/*
+ * Special value indicating we should use the "auto" quorum method instead of
+ * a fixed value (including zero to turn off quorum enforcement).
+ */
+#define AFR_QUORUM_AUTO INT_MAX
+
+/*
+ * Having this as a macro will make debugging a bit weirder, but does reduce
+ * the probability of functions handling this check inconsistently.
+ */
+#define QUORUM_CHECK(_func,_label) do {                                  \
+        if (priv->quorum_count && !afr_have_quorum(this->name,priv)) { \
+                gf_log(this->name,GF_LOG_WARNING,                        \
+                       "failing "#_func" due to lack of quorum");        \
+                op_errno = EROFS;                                        \
+                goto _label;                                             \
+        }                                                                \
+} while (0);
+
+
+#define AFR_SBRAIN_MSG "Failed on %s as split-brain is seen. Returning EIO."
+
+#define AFR_SBRAIN_CHECK_FD(fd, label) do {                              \
+        if (fd->inode && afr_is_split_brain (this, fd->inode)) {        \
+                op_errno = EIO;                                         \
+                gf_log (this->name, GF_LOG_WARNING,                     \
+                        AFR_SBRAIN_MSG ,uuid_utoa (fd->inode->gfid));   \
+                goto label;                                             \
+        }                                                               \
+} while (0)
+
+#define AFR_SBRAIN_CHECK_LOC(loc, label) do {                           \
+        if (loc->inode && afr_is_split_brain (this, loc->inode)) {      \
+                op_errno = EIO;                                         \
+                loc_path (loc, NULL);                                   \
+                gf_log (this->name, GF_LOG_WARNING,                     \
+                        AFR_SBRAIN_MSG , loc->path);                    \
+                goto label;                                             \
+        }                                                               \
+} while (0)
+
+int
+afr_fd_report_unstable_write (xlator_t *this, fd_t *fd);
+
+gf_boolean_t
+afr_fd_has_witnessed_unstable_write (xlator_t *this, fd_t *fd);
+
+void
+afr_delayed_changelog_wake_resume (xlator_t *this, fd_t *fd, call_stub_t *stub);
+
+int
+afr_inodelk_init (afr_inodelk_t *lk, char *dom, size_t child_count);
+
+void
+afr_handle_open_fd_count (call_frame_t *frame, xlator_t *this);
+
+afr_inode_ctx_t*
+afr_inode_ctx_get (inode_t *inode, xlator_t *this);
 
 #endif /* __AFR_H__ */
diff --git a/xlators/cluster/afr/src/pump.c b/xlators/cluster/afr/src/pump.c
index 306e0e722..a7f72fb30 100644
--- a/xlators/cluster/afr/src/pump.c
+++ b/xlators/cluster/afr/src/pump.c
@@ -1,25 +1,17 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <unistd.h>
 #include <sys/time.h>
 #include <stdlib.h>
+#include <fnmatch.h>
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
@@ -28,6 +20,15 @@
 
 #include "afr-common.c"
 #include "defaults.c"
+#include "glusterfs.h"
+
+static uint64_t pump_pid = 0;
+static inline void
+pump_fill_loc_info (loc_t *loc, struct iatt *iatt, struct iatt *parent)
+{
+        afr_update_loc_gfids (loc, iatt, parent);
+        uuid_copy (loc->inode->gfid, iatt->ia_gfid);
+}
 
 static int
 pump_mark_start_pending (xlator_t *this)
@@ -135,92 +136,24 @@ pump_set_resume_path (xlator_t *this, const char *path)
         priv = this->private;
         pump_priv = priv->pump_private;
 
-        assert (pump_priv);
+        GF_ASSERT (pump_priv);
 
         LOCK (&pump_priv->resume_path_lock);
         {
-                pump_priv->resume_path = strdup (path);
-                if (!pump_priv->resume_path)
-                        ret = -1;
+                strncpy (pump_priv->resume_path, path, strlen (path) + 1);
         }
         UNLOCK (&pump_priv->resume_path_lock);
 
         return ret;
 }
 
-static void
-build_child_loc (loc_t *parent, loc_t *child, char *path, char *name)
-{
-        child->path = path;
-        child->name = name;
-
-        child->parent = inode_ref (parent->inode);
-        child->inode = inode_new (parent->inode->table);
-}
-
-static char *
-build_file_path (loc_t *loc, gf_dirent_t *entry)
-{
-        xlator_t *this = NULL;
-        char *file_path = NULL;
-        int pathlen = 0;
-        int total_size = 0;
-
-        this = THIS;
-
-        pathlen = STRLEN_0 (loc->path);
-
-        if (IS_ROOT_PATH (loc->path)) {
-                total_size = pathlen + entry->d_len;
-                file_path = GF_CALLOC (1, total_size, gf_afr_mt_char);
-                if (!file_path) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        return NULL;
-                }
-
-                gf_log (this->name, GF_LOG_TRACE,
-                        "constructing file path of size=%d"
-                        "pathlen=%d, d_len=%d",
-                        total_size, pathlen,
-                        entry->d_len);
-
-                snprintf(file_path, total_size, "%s%s", loc->path, entry->d_name);
-
-        } else {
-                total_size = pathlen + entry->d_len + 1; /* for the extra '/' in the path */
-                file_path = GF_CALLOC (1, total_size + 1, gf_afr_mt_char);
-                if (!file_path) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        return NULL;
-                }
-
-                gf_log (this->name, GF_LOG_TRACE,
-                        "constructing file path of size=%d"
-                        "pathlen=%d, d_len=%d",
-                        total_size, pathlen,
-                        entry->d_len);
-
-                snprintf(file_path, total_size, "%s/%s", loc->path, entry->d_name);
-        }
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "path=%s and d_name=%s", loc->path, entry->d_name);
-        gf_log (this->name, GF_LOG_TRACE,
-                "constructed file_path=%s of size=%d", file_path, total_size);
-
-        return file_path;
-}
-
 static int
 pump_save_path (xlator_t *this, const char *path)
 {
         afr_private_t *priv = NULL;
-        pump_private_t *pump_priv = NULL;
         pump_state_t state;
         dict_t *dict = NULL;
-        loc_t  loc;
+        loc_t  loc = {0};
         int dict_ret = 0;
         int ret = -1;
 
@@ -229,29 +162,30 @@ pump_save_path (xlator_t *this, const char *path)
                 return 0;
 
         priv = this->private;
-        pump_priv = priv->pump_private;
 
-        assert (priv->root_inode);
+        GF_ASSERT (priv->root_inode);
 
-        build_root_loc (priv->root_inode, &loc);
+        afr_build_root_loc (this, &loc);
 
         dict = dict_new ();
         dict_ret = dict_set_str (dict, PUMP_PATH, (char *)path);
+        if (dict_ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set the key %s", path, PUMP_PATH);
 
         ret = syncop_setxattr (PUMP_SOURCE_CHILD (this), &loc, dict, 0);
 
         if (ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_INFO,
                         "setxattr failed - could not save path=%s", path);
         } else {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "setxattr succeeded - saved path=%s", path);
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Saving path for status info");
         }
 
         dict_unref (dict);
 
+        loc_wipe (&loc);
         return 0;
 }
 
@@ -314,15 +248,9 @@ pump_get_resume_path (xlator_t *this)
 static int
 pump_update_resume_state (xlator_t *this, const char *path)
 {
-        afr_private_t *priv = NULL;
-        pump_private_t *pump_priv = NULL;
-
         pump_state_t state;
         const char *resume_path = NULL;
 
-        priv = this->private;
-        pump_priv = priv->pump_private;
-
         state = pump_get_state ();
 
         if (state == PUMP_STATE_RESUME) {
@@ -350,16 +278,10 @@ pump_update_resume_state (xlator_t *this, const char *path)
 static gf_boolean_t
 is_pump_traversal_allowed (xlator_t *this, const char *path)
 {
-        afr_private_t *priv = NULL;
-        pump_private_t *pump_priv = NULL;
-
         pump_state_t state;
         const char *resume_path = NULL;
         gf_boolean_t ret = _gf_true;
 
-        priv = this->private;
-        pump_priv = priv->pump_private;
-
         state = pump_get_state ();
 
         if (state == PUMP_STATE_RESUME) {
@@ -402,38 +324,33 @@ pump_save_file_stats (xlator_t *this, const char *path)
 static int
 gf_pump_traverse_directory (loc_t *loc)
 {
-        xlator_t *this = NULL;
-        afr_private_t *priv = NULL;
-        fd_t     *fd   = NULL;
-
-        off_t       offset   = 0;
-        loc_t       entry_loc;
-        gf_dirent_t *entry = NULL;
-        gf_dirent_t *tmp = NULL;
-        gf_dirent_t entries;
-
-	struct iatt iatt, parent;
-	dict_t *xattr_rsp;
-
-        int source = 0;
-
-        char *file_path = NULL;
-        int ret = 0;
+        xlator_t        *this              = NULL;
+        fd_t            *fd                = NULL;
+        off_t           offset             = 0;
+        loc_t           entry_loc          = {0};
+        gf_dirent_t     *entry             = NULL;
+        gf_dirent_t     *tmp               = NULL;
+        gf_dirent_t     entries;
+	struct iatt     iatt               = {0};
+        struct iatt     parent             = {0};
+	dict_t          *xattr_rsp         = NULL;
+        int             ret                = 0;
+        gf_boolean_t    is_directory_empty = _gf_true;
+        gf_boolean_t    free_entries       = _gf_false;
 
         INIT_LIST_HEAD (&entries.list);
         this = THIS;
-        priv = this->private;
 
-        assert (loc->inode);
+        GF_ASSERT (loc->inode);
 
-	fd = fd_create (loc->inode, PUMP_PID);
+	fd = fd_create (loc->inode, pump_pid);
         if (!fd) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "Failed to create fd for %s", loc->path);
                 goto out;
         }
 
-        ret = syncop_opendir (priv->children[source], loc, fd);
+        ret = syncop_opendir (this, loc, fd);
         if (ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "opendir failed on %s", loc->path);
@@ -444,7 +361,8 @@ gf_pump_traverse_directory (loc_t *loc)
                 "pump opendir on %s returned=%d",
                 loc->path, ret);
 
-        while (syncop_readdirp (priv->children[source], fd, 131072, offset, &entries)) {
+        while (syncop_readdirp (this, fd, 131072, offset, NULL, &entries)) {
+                free_entries = _gf_true;
 
                 if (list_empty (&entries.list)) {
                         gf_log (this->name, GF_LOG_TRACE,
@@ -456,106 +374,94 @@ gf_pump_traverse_directory (loc_t *loc)
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "found readdir entry=%s", entry->d_name);
 
-                        file_path = build_file_path (loc, entry);
-                        if (!file_path) {
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "file path construction failed");
-                                goto out;
-                        }
-
-                        build_child_loc (loc, &entry_loc, file_path, entry->d_name);
-
-                        ret = syncop_lookup (this, &entry_loc, NULL,
-                                             &iatt, &xattr_rsp, &parent);
-
-                        entry_loc.ino = iatt.ia_ino;
-                        entry_loc.inode->ino = iatt.ia_ino;
-                        memcpy (entry_loc.inode->gfid, iatt.ia_gfid, 16);
-
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "lookup %s => %"PRId64,
-                                entry_loc.path,
-                                iatt.ia_ino);
-
-                        ret = syncop_lookup (this, &entry_loc, NULL,
-                                             &iatt, &xattr_rsp, &parent);
-
-
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "second lookup ret=%d: %s => %"PRId64,
-                                ret,
-                                entry_loc.path,
-                                iatt.ia_ino);
-
-                        pump_update_resume_state (this, entry_loc.path);
-
-                        if (!IS_ENTRY_CWD(entry->d_name) &&
-                            !IS_ENTRY_PARENT (entry->d_name)) {
-                                pump_save_path (this, entry_loc.path);
-                                pump_save_file_stats (this, entry_loc.path);
+                        offset = entry->d_off;
+                        if (uuid_is_null (entry->d_stat.ia_gfid)) {
+                                gf_log (this->name, GF_LOG_WARNING, "%s/%s: No "
+                                        "gfid present skipping",
+                                        loc->path, entry->d_name);
+                                continue;
                         }
-
-                        ret = pump_check_and_update_status (this);
-                        if (ret < 0) {
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "Pump beginning to exit out");
+                        loc_wipe (&entry_loc);
+                        ret = afr_build_child_loc (this, &entry_loc, loc,
+                                                   entry->d_name);
+                        if (ret)
                                 goto out;
-                        }
 
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "type of file=%d, IFDIR=%d",
-                                iatt.ia_type, IA_IFDIR);
-
-                        if (IA_ISDIR (iatt.ia_type) && !IS_ENTRY_CWD(entry->d_name) &&
+                        if (!IS_ENTRY_CWD (entry->d_name) &&
                             !IS_ENTRY_PARENT (entry->d_name)) {
-                                if (is_pump_traversal_allowed (this, entry_loc.path)) {
-                                        gf_log (this->name, GF_LOG_TRACE,
-                                                "entering dir=%s",
-                                                entry->d_name);
-                                        gf_pump_traverse_directory (&entry_loc);
-                                }
-                        }
 
-                        offset = entry->d_off;
-                        loc_wipe (&entry_loc);
+                                    is_directory_empty = _gf_false;
+                                    gf_log (this->name, GF_LOG_DEBUG,
+                                            "lookup %s => %"PRId64,
+                                            entry_loc.path,
+                                            iatt.ia_ino);
+
+                                    ret = syncop_lookup (this, &entry_loc, NULL,
+                                                         &iatt, &xattr_rsp, &parent);
+
+                                    if (ret) {
+                                            gf_log (this->name, GF_LOG_ERROR,
+                                                    "%s: lookup failed",
+                                                    entry_loc.path);
+                                            continue;
+                                    }
+                                    pump_fill_loc_info (&entry_loc, &iatt,
+                                                       &parent);
+
+                                    pump_update_resume_state (this, entry_loc.path);
+
+                                    pump_save_path (this, entry_loc.path);
+                                    pump_save_file_stats (this, entry_loc.path);
+
+                                    ret = pump_check_and_update_status (this);
+                                    if (ret < 0) {
+                                            gf_log (this->name, GF_LOG_DEBUG,
+                                                    "Pump beginning to exit out");
+                                            goto out;
+                                    }
+
+                                    if (IA_ISDIR (iatt.ia_type)) {
+                                            if (is_pump_traversal_allowed (this, entry_loc.path)) {
+                                                    gf_log (this->name, GF_LOG_TRACE,
+                                                            "entering dir=%s",
+                                                            entry->d_name);
+                                                    gf_pump_traverse_directory (&entry_loc);
+                                            }
+                                    }
+                        }
                 }
 
                 gf_dirent_free (&entries);
+                free_entries = _gf_false;
                 gf_log (this->name, GF_LOG_TRACE,
                         "offset incremented to %d",
                         (int32_t ) offset);
 
         }
 
-out:
-        return 0;
-
-}
+        ret = syncop_close (fd);
+        if (ret < 0)
+                gf_log (this->name, GF_LOG_DEBUG, "closing the fd failed");
 
-void
-build_root_loc (inode_t *inode, loc_t *loc)
-{
-        loc->path = "/";
-        loc->name = "";
-        loc->inode = inode;
-        loc->ino = 1;
-        loc->inode->ino = 1;
-        memset (loc->inode->gfid, 0, 16);
-        loc->inode->gfid[15] = 1;
+        if (is_directory_empty && IS_ROOT_PATH (loc->path)) {
+               pump_change_state (this, PUMP_STATE_RUNNING);
+               gf_log (this->name, GF_LOG_INFO, "Empty source brick. "
+                                "Nothing to be done.");
+        }
 
+out:
+        if (entry_loc.path)
+                loc_wipe (&entry_loc);
+        if (free_entries)
+                gf_dirent_free (&entries);
+        return 0;
 }
 
 static int
 pump_update_resume_path (xlator_t *this)
 {
-        afr_private_t *priv = NULL;
-        pump_private_t *pump_priv = NULL;
-
         const char *resume_path = NULL;
 
-        priv = this->private;
-        pump_priv = priv->pump_private;
-
         resume_path = pump_get_resume_path (this);
 
         if (resume_path) {
@@ -574,6 +480,39 @@ pump_update_resume_path (xlator_t *this)
         return 0;
 }
 
+static int32_t
+pump_xattr_cleaner (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        afr_private_t  *priv      = NULL;
+        loc_t           loc       = {0};
+        int             i         = 0;
+        int             ret       = 0;
+        int             source    = 0;
+        int             sink      = 1;
+
+        priv      = this->private;
+
+        afr_build_root_loc (this, &loc);
+
+        ret = syncop_removexattr (priv->children[source], &loc,
+                                          PUMP_PATH);
+
+        ret = syncop_removexattr (priv->children[sink], &loc,
+                                  PUMP_SINK_COMPLETE);
+
+        for (i = 0; i < priv->child_count; i++) {
+                ret = syncop_removexattr (priv->children[i], &loc,
+                                          PUMP_SOURCE_COMPLETE);
+                if (ret)
+                        gf_log (this->name, GF_LOG_DEBUG, "removexattr "
+                                "failed with %s", strerror (errno));
+        }
+
+        loc_wipe (&loc);
+        return pump_command_reply (frame, this);
+}
+
 static int
 pump_complete_migration (xlator_t *this)
 {
@@ -581,16 +520,16 @@ pump_complete_migration (xlator_t *this)
         pump_private_t *pump_priv = NULL;
         dict_t *dict = NULL;
         pump_state_t state;
-        loc_t  loc;
+        loc_t  loc = {0};
         int dict_ret = 0;
         int ret = -1;
 
         priv = this->private;
         pump_priv = priv->pump_private;
 
-        assert (priv->root_inode);
+        GF_ASSERT (priv->root_inode);
 
-        build_root_loc (priv->root_inode, &loc);
+        afr_build_root_loc (this, &loc);
 
         dict = dict_new ();
 
@@ -602,6 +541,10 @@ pump_complete_migration (xlator_t *this)
                 pump_priv->pump_finished = _gf_true;
 
                 dict_ret = dict_set_str (dict, PUMP_SOURCE_COMPLETE, "jargon");
+                if (dict_ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set the key %s",
+                                loc.path, PUMP_SOURCE_COMPLETE);
 
                 ret = syncop_setxattr (PUMP_SOURCE_CHILD (this), &loc, dict, 0);
                 if (ret < 0) {
@@ -609,44 +552,97 @@ pump_complete_migration (xlator_t *this)
                                 "setxattr failed - while  notifying source complete");
                 }
                 dict_ret = dict_set_str (dict, PUMP_SINK_COMPLETE, "jargon");
+                if (dict_ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set the key %s",
+                                loc.path, PUMP_SINK_COMPLETE);
 
                 ret = syncop_setxattr (PUMP_SINK_CHILD (this), &loc, dict, 0);
                 if (ret < 0) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "setxattr failed - while notifying sink complete");
                 }
+
+                pump_save_path (this, "/");
+
+        } else if (state == PUMP_STATE_ABORT) {
+                gf_log (this->name, GF_LOG_DEBUG, "Starting cleanup "
+                        "of pump internal xattrs");
+                call_resume (pump_priv->cleaner);
         }
 
+        loc_wipe (&loc);
         return 0;
 }
 
 static int
+pump_lookup_sink (loc_t *loc)
+{
+        xlator_t *this = NULL;
+	struct iatt iatt, parent;
+	dict_t *xattr_rsp;
+        dict_t *xattr_req = NULL;
+        int ret = 0;
+
+        this = THIS;
+
+        xattr_req = dict_new ();
+
+        ret = afr_set_root_gfid (xattr_req);
+        if (ret)
+                goto out;
+
+        ret = syncop_lookup (PUMP_SINK_CHILD (this), loc,
+                             xattr_req, &iatt, &xattr_rsp, &parent);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Lookup on sink child failed");
+                goto out;
+        }
+
+out:
+        if (xattr_req)
+                dict_unref (xattr_req);
+
+        return ret;
+}
+
+static int
 pump_task (void *data)
 {
 	xlator_t *this = NULL;
         afr_private_t *priv = NULL;
 
 
-        loc_t loc;
+        loc_t loc = {0};
 	struct iatt iatt, parent;
-	dict_t *xattr_rsp;
+	dict_t *xattr_rsp = NULL;
+        dict_t *xattr_req = NULL;
 
         int ret = -1;
 
         this = THIS;
         priv = this->private;
 
-        assert (priv->root_inode);
+        GF_ASSERT (priv->root_inode);
 
-        build_root_loc (priv->root_inode, &loc);
+        afr_build_root_loc (this, &loc);
+        xattr_req = dict_new ();
+        if (!xattr_req) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Out of memory");
+                ret = -1;
+                goto out;
+        }
 
-        ret = syncop_lookup (this, &loc, NULL,
+        afr_set_root_gfid (xattr_req);
+        ret = syncop_lookup (this, &loc, xattr_req,
                              &iatt, &xattr_rsp, &parent);
 
         gf_log (this->name, GF_LOG_TRACE,
-                "lookup: ino=%"PRId64", path=%s",
-                loc.ino,
-                loc.path);
+                "lookup: path=%s gfid=%s",
+                loc.path, uuid_utoa (loc.inode->gfid));
 
         ret = pump_check_and_update_status (this);
         if (ret < 0) {
@@ -655,30 +651,37 @@ pump_task (void *data)
 
         pump_update_resume_path (this);
 
+        afr_set_root_gfid (xattr_req);
+        ret = pump_lookup_sink (&loc);
+        if (ret) {
+                pump_update_resume_path (this);
+                goto out;
+        }
+
         gf_pump_traverse_directory (&loc);
 
         pump_complete_migration (this);
 out:
+        if (xattr_req)
+                dict_unref (xattr_req);
+
+        loc_wipe (&loc);
 	return 0;
 }
 
 
 static int
-pump_task_completion (int ret, void *data)
+pump_task_completion (int ret, call_frame_t *sync_frame, void *data)
 {
         xlator_t *this = NULL;
-        call_frame_t *frame = NULL;
         afr_private_t *priv = NULL;
-        pump_private_t *pump_priv = NULL;
 
         this = THIS;
 
-        frame = (call_frame_t *) data;
-
         priv = this->private;
-        pump_priv = priv->pump_private;
 
         inode_unref (priv->root_inode);
+        STACK_DESTROY (sync_frame->root);
 
         gf_log (this->name, GF_LOG_DEBUG,
                 "Pump xlator exiting");
@@ -696,20 +699,22 @@ pump_start (call_frame_t *pump_frame, xlator_t *this)
 	priv = this->private;
         pump_priv = priv->pump_private;
 
-        if (!pump_frame->root->lk_owner)
-                pump_frame->root->lk_owner = PUMP_LK_OWNER;
+        afr_set_lk_owner (pump_frame, this, pump_frame->root);
+	pump_pid = (uint64_t) (unsigned long)pump_frame->root;
 
-	ret = synctask_new (pump_priv->env, pump_task, pump_task_completion,
-                            pump_frame);
+	ret = synctask_new (pump_priv->env, pump_task,
+                            pump_task_completion,
+                            pump_frame, NULL);
         if (ret == -1) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_ERROR,
                         "starting pump failed");
                 pump_change_state (this, PUMP_STATE_ABORT);
                 goto out;
         }
 
-        gf_log (this->name, GF_LOG_TRACE,
-                "setting pump as started");
+        gf_log (this->name, GF_LOG_DEBUG,
+                "setting pump as started lk_owner: %s %"PRIu64,
+                lkowner_utoa (&pump_frame->root->lk_owner), pump_pid);
 
         priv->use_afr_in_pump = 1;
 out:
@@ -743,9 +748,10 @@ pump_cmd_start_setxattr_cbk (call_frame_t *frame,
                              void *cookie,
                              xlator_t *this,
                              int32_t op_ret,
-                             int32_t op_errno)
+                             int32_t op_errno, dict_t *xdata)
 
 {
+        call_frame_t *prev = NULL;
         afr_local_t *local = NULL;
         int ret = 0;
 
@@ -765,6 +771,11 @@ pump_cmd_start_setxattr_cbk (call_frame_t *frame,
 
         pump_mark_start_pending (this);
 
+        /* send the PARENT_UP as pump is ready now */
+        prev = cookie;
+        if (prev && prev->this)
+                prev->this->notify (prev->this, GF_EVENT_PARENT_UP, this);
+
 out:
         local->op_ret = ret;
         pump_command_reply (frame, this);
@@ -778,8 +789,9 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
         afr_local_t   *local     = NULL;
         afr_private_t *priv      = NULL;
         dict_t        *dict      = NULL;
-        char          *dst_brick = NULL;
-        loc_t loc;
+        data_t        *data      = NULL;
+        char          *clnt_cmd  = NULL;
+        loc_t loc = {0};
 
         int ret = 0;
 
@@ -788,10 +800,11 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
 
         GF_ASSERT (priv->root_inode);
 
-        build_root_loc (priv->root_inode, &loc);
+        afr_build_root_loc (this, &loc);
 
-        ret = dict_get_str (local->dict, PUMP_CMD_START, &dst_brick);
-        if (ret < 0) {
+        data = data_ref (dict_get (local->dict, RB_PUMP_CMD_START));
+        if (!data) {
+                ret = -1;
                 gf_log (this->name, GF_LOG_ERROR,
                         "Could not get destination brick value");
                 goto out;
@@ -799,17 +812,22 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
 
         dict = dict_new ();
         if (!dict) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
                 ret = -1;
                 goto out;
         }
 
-        GF_ASSERT (dst_brick);
-        gf_log (this->name, GF_LOG_DEBUG,
-                "Got destination brick as %s", dst_brick);
+        clnt_cmd = GF_CALLOC (1, data->len+1, gf_common_mt_char);
+        if (!clnt_cmd) {
+                ret = -1;
+                goto out;
+        }
+
+        memcpy (clnt_cmd, data->data, data->len);
+        clnt_cmd[data->len] = '\0';
+        gf_log (this->name, GF_LOG_DEBUG, "Got destination brick %s\n",
+                        clnt_cmd);
 
-        ret = dict_set_str (dict, CLIENT_CMD_CONNECT, dst_brick);
+        ret = dict_set_dynstr (dict, CLIENT_CMD_CONNECT, clnt_cmd);
         if (ret < 0) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "Could not inititiate destination brick "
@@ -823,12 +841,21 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
 		    PUMP_SINK_CHILD(this)->fops->setxattr,
 		    &loc,
 		    dict,
-		    0);
+		    0, NULL);
 
         ret = 0;
 
-        dict_unref (dict);
 out:
+        if (dict)
+                dict_unref (dict);
+
+        if (data)
+                data_unref (data);
+
+        if (ret && clnt_cmd)
+                GF_FREE (clnt_cmd);
+
+        loc_wipe (&loc);
         return ret;
 }
 
@@ -848,7 +875,7 @@ pump_cmd_start_getxattr_cbk (call_frame_t *frame,
                              xlator_t *this,
                              int32_t op_ret,
                              int32_t op_errno,
-                             dict_t *dict)
+                             dict_t *dict, dict_t *xdata)
 {
         afr_local_t *local = NULL;
         char *path = NULL;
@@ -892,6 +919,8 @@ pump_cmd_start_getxattr_cbk (call_frame_t *frame,
         else {
                 /* We're re-starting pump from a previous
                    pause */
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "about to start synctask");
                 ret = pump_start_synctask (this);
                 need_unwind = 1;
         }
@@ -913,6 +942,7 @@ pump_execute_status (call_frame_t *frame, xlator_t *this)
         uint64_t number_files = 0;
 
         char filename[PATH_MAX];
+        char summary[PATH_MAX+256];
         char *dict_str = NULL;
 
         int32_t op_ret = 0;
@@ -941,28 +971,35 @@ pump_execute_status (call_frame_t *frame, xlator_t *this)
         }
 
         if (pump_priv->pump_finished) {
-        snprintf (dict_str, PATH_MAX + 256, "Number of files migrated = %"PRIu64"        Migration complete ",
-                  number_files);
+                snprintf (summary, PATH_MAX+256,
+                          "no_of_files=%"PRIu64, number_files);
         } else {
-        snprintf (dict_str, PATH_MAX + 256, "Number of files migrated = %"PRIu64"       Current file= %s ",
-                  number_files, filename);
+                snprintf (summary, PATH_MAX+256,
+                          "no_of_files=%"PRIu64":current_file=%s",
+                          number_files, filename);
         }
+        snprintf (dict_str, PATH_MAX+256, "status=%d:%s",
+                  (pump_priv->pump_finished)?1:0, summary);
 
         dict = dict_new ();
 
-        ret = dict_set_str (dict, PUMP_CMD_STATUS, dict_str);
+        ret = dict_set_dynstr (dict, RB_PUMP_CMD_STATUS, dict_str);
         if (ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
-                        "dict_set_str returned negative value");
+                        "dict_set_dynstr returned negative value");
+        } else {
+                dict_str = NULL;
         }
 
         op_ret = 0;
 
 out:
 
-        AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+        AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, NULL);
+
+        if (dict)
+                dict_unref (dict);
 
-        dict_unref (dict);
         GF_FREE (dict_str);
 
         return 0;
@@ -990,7 +1027,7 @@ pump_execute_start (call_frame_t *frame, xlator_t *this)
         afr_local_t   *local = NULL;
 
         int ret = 0;
-        loc_t loc;
+        loc_t loc = {0};
 
         priv = this->private;
         local = frame->local;
@@ -1005,14 +1042,14 @@ pump_execute_start (call_frame_t *frame, xlator_t *this)
 
         GF_ASSERT (priv->root_inode);
 
-        build_root_loc (priv->root_inode, &loc);
+        afr_build_root_loc (this, &loc);
 
 	STACK_WIND (frame,
 		    pump_cmd_start_getxattr_cbk,
 		    PUMP_SOURCE_CHILD(this),
 		    PUMP_SOURCE_CHILD(this)->fops->getxattr,
 		    &loc,
-		    PUMP_PATH);
+		    PUMP_PATH, NULL);
 
         ret = 0;
 
@@ -1022,22 +1059,98 @@ out:
                 pump_command_reply (frame, this);
         }
 
+        loc_wipe (&loc);
 	return 0;
 }
 
+static int
+pump_cleanup_helper (void *data) {
+        call_frame_t *frame = data;
+
+        pump_xattr_cleaner (frame, 0, frame->this, 0, 0, NULL);
+
+        return 0;
+}
+
+static int
+pump_cleanup_done (int ret, call_frame_t *sync_frame, void *data)
+{
+        STACK_DESTROY (sync_frame->root);
+
+        return 0;
+}
+
+int
+pump_execute_commit (call_frame_t *frame, xlator_t *this)
+{
+        afr_private_t  *priv       = NULL;
+        pump_private_t *pump_priv  = NULL;
+        afr_local_t    *local      = NULL;
+        call_frame_t   *sync_frame = NULL;
+        int             ret        = 0;
+
+        priv      = this->private;
+        pump_priv = priv->pump_private;
+        local     = frame->local;
+
+        local->op_ret = 0;
+        if (pump_priv->pump_finished) {
+                pump_change_state (this, PUMP_STATE_COMMIT);
+                sync_frame = create_frame (this, this->ctx->pool);
+                ret = synctask_new (pump_priv->env, pump_cleanup_helper,
+                                    pump_cleanup_done, sync_frame, frame);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_DEBUG, "Couldn't create "
+                                "synctask for cleaning up xattrs.");
+                }
+
+        } else {
+                gf_log (this->name, GF_LOG_ERROR, "Commit can't proceed. "
+                        "Migration in progress");
+                local->op_ret = -1;
+                local->op_errno = EINPROGRESS;
+                pump_command_reply (frame, this);
+        }
+
+        return 0;
+}
 int
 pump_execute_abort (call_frame_t *frame, xlator_t *this)
 {
-        afr_private_t *priv = NULL;
-        afr_local_t   *local = NULL;
+        afr_private_t  *priv       = NULL;
+        pump_private_t *pump_priv  = NULL;
+        afr_local_t    *local      = NULL;
+        call_frame_t   *sync_frame = NULL;
+        int             ret        = 0;
 
-        priv = this->private;
-        local = frame->local;
+        priv      = this->private;
+        pump_priv = priv->pump_private;
+        local     = frame->local;
 
         pump_change_state (this, PUMP_STATE_ABORT);
 
+        LOCK (&pump_priv->resume_path_lock);
+        {
+                pump_priv->number_files_pumped = 0;
+                pump_priv->current_file[0] = '\0';
+        }
+        UNLOCK (&pump_priv->resume_path_lock);
+
         local->op_ret = 0;
-        pump_command_reply (frame, this);
+        if (pump_priv->pump_finished) {
+                sync_frame = create_frame (this, this->ctx->pool);
+                ret = synctask_new (pump_priv->env, pump_cleanup_helper,
+                                    pump_cleanup_done, sync_frame, frame);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_DEBUG, "Couldn't create "
+                                "synctask for cleaning up xattrs.");
+                }
+
+        } else {
+                pump_priv->cleaner = fop_setxattr_cbk_stub (frame,
+                                                            pump_xattr_cleaner,
+                                                            0, 0, NULL);
+        }
 
         return 0;
 }
@@ -1049,7 +1162,7 @@ pump_command_status (xlator_t *this, dict_t *dict)
         int dict_ret = -1;
         int ret = _gf_true;
 
-        dict_ret = dict_get_str (dict, PUMP_CMD_STATUS, &cmd);
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_STATUS, &cmd);
         if (dict_ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "Not a pump status command");
@@ -1073,7 +1186,7 @@ pump_command_pause (xlator_t *this, dict_t *dict)
         int dict_ret = -1;
         int ret = _gf_true;
 
-        dict_ret = dict_get_str (dict, PUMP_CMD_PAUSE, &cmd);
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_PAUSE, &cmd);
         if (dict_ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "Not a pump pause command");
@@ -1091,13 +1204,37 @@ out:
 }
 
 gf_boolean_t
+pump_command_commit (xlator_t *this, dict_t *dict)
+{
+        char *cmd = NULL;
+        int dict_ret = -1;
+        int ret = _gf_true;
+
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_COMMIT, &cmd);
+        if (dict_ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Not a pump commit command");
+                ret = _gf_false;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Hit a pump command - commit");
+        ret = _gf_true;
+
+out:
+        return ret;
+
+}
+
+gf_boolean_t
 pump_command_abort (xlator_t *this, dict_t *dict)
 {
         char *cmd = NULL;
         int dict_ret = -1;
         int ret = _gf_true;
 
-        dict_ret = dict_get_str (dict, PUMP_CMD_ABORT, &cmd);
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_ABORT, &cmd);
         if (dict_ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "Not a pump abort command");
@@ -1121,7 +1258,7 @@ pump_command_start (xlator_t *this, dict_t *dict)
         int dict_ret = -1;
         int ret = _gf_true;
 
-        dict_ret = dict_get_str (dict, PUMP_CMD_START, &cmd);
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_START, &cmd);
         if (dict_ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "Not a pump start command");
@@ -1143,7 +1280,7 @@ struct _xattr_key {
         struct list_head list;
 };
 
-static void
+static int
 __gather_xattr_keys (dict_t *dict, char *key, data_t *value,
                      void *data)
 {
@@ -1155,13 +1292,14 @@ __gather_xattr_keys (dict_t *dict, char *key, data_t *value,
 
                 xkey = GF_CALLOC (1, sizeof (*xkey), gf_afr_mt_xattr_key);
                 if (!xkey)
-                        return;
+                        return -1;
 
                 xkey->key = key;
                 INIT_LIST_HEAD (&xkey->list);
 
                 list_add_tail (&xkey->list, list);
         }
+        return 0;
 }
 
 static void
@@ -1189,16 +1327,17 @@ __filter_xattrs (dict_t *dict)
 int32_t
 pump_getxattr_cbk (call_frame_t *frame, void *cookie,
 		  xlator_t *this, int32_t op_ret, int32_t op_errno,
-		  dict_t *dict)
+		  dict_t *dict, dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
+	afr_private_t   *priv           = NULL;
+	afr_local_t     *local          = NULL;
+	xlator_t        **children      = NULL;
+	int             unwind          = 1;
+        int32_t         *last_index     = NULL;
+        int32_t         next_call_child = -1;
+        int32_t         read_child      = -1;
+        int32_t         *fresh_children = NULL;
 
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-        int read_child = -1;
 
 	priv     = this->private;
 	children = priv->children;
@@ -1208,25 +1347,22 @@ pump_getxattr_cbk (call_frame_t *frame, void *cookie,
         read_child = (long) cookie;
 
 	if (op_ret == -1) {
-        retry:
-		last_tried = local->cont.getxattr.last_tried;
-
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try = ++local->cont.getxattr.last_tried;
-
-                if (this_try == read_child) {
-                        goto retry;
-                }
+		last_index = &local->cont.getxattr.last_index;
+                fresh_children = local->fresh_children;
+                next_call_child = afr_next_call_child (fresh_children,
+                                                       local->child_up,
+                                                       priv->child_count,
+                                                       last_index, read_child);
+                if (next_call_child < 0)
+                        goto out;
 
 		unwind = 0;
 		STACK_WIND_COOKIE (frame, pump_getxattr_cbk,
 				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->getxattr,
+				   children[next_call_child],
+				   children[next_call_child]->fops->getxattr,
 				   &local->loc,
-				   local->cont.getxattr.name);
+				   local->cont.getxattr.name, NULL);
 	}
 
 out:
@@ -1234,7 +1370,7 @@ out:
                 if (op_ret >= 0 && dict)
                         __filter_xattrs (dict);
 
-		AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+		AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, NULL);
 	}
 
 	return 0;
@@ -1242,17 +1378,15 @@ out:
 
 int32_t
 pump_getxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, const char *name)
+	      loc_t *loc, const char *name, dict_t *xdata)
 {
 	afr_private_t *   priv       = NULL;
 	xlator_t **       children   = NULL;
 	int               call_child = 0;
-	afr_local_t     * local      = NULL;
-
-        int               read_child = -1;
-
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+	afr_local_t       *local     = NULL;
+	int32_t           ret     = -1;
+	int32_t           op_errno   = 0;
+        uint64_t          read_child = 0;
 
 
 	VALIDATE_OR_GOTO (frame, out);
@@ -1263,9 +1397,21 @@ pump_getxattr (call_frame_t *frame, xlator_t *this,
 	VALIDATE_OR_GOTO (priv->children, out);
 
 	children = priv->children;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame, default_getxattr_cbk,
+                            FIRST_CHILD (this),
+                            (FIRST_CHILD (this))->fops->getxattr,
+                            loc, name, xdata);
+                return 0;
+        }
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-	frame->local = local;
+
+	AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+	local = frame->local;
+
+        ret = afr_local_init (local, priv, &op_errno);
+        if (ret < 0)
+                goto out;
 
         if (name) {
                 if (!strncmp (name, AFR_XATTR_PREFIX,
@@ -1275,42 +1421,33 @@ pump_getxattr (call_frame_t *frame, xlator_t *this,
                         goto out;
                 }
 
-                if (!strcmp (name, PUMP_CMD_STATUS)) {
+                if (!strcmp (name, RB_PUMP_CMD_STATUS)) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "Hit pump command - status");
                         pump_execute_status (frame, this);
-                        op_ret = 0;
+                        ret = 0;
                         goto out;
                 }
         }
 
-        if (!priv->use_afr_in_pump) {
-                STACK_WIND (frame, default_getxattr_cbk,
-                            FIRST_CHILD (this),
-                            (FIRST_CHILD (this))->fops->getxattr,
-                            loc, name);
-                return 0;
+        local->fresh_children = GF_CALLOC (priv->child_count,
+                                          sizeof (*local->fresh_children),
+                                          gf_afr_mt_int32_t);
+        if (!local->fresh_children) {
+                ret = -1;
+                op_errno = ENOMEM;
+                goto out;
         }
 
-        read_child = afr_read_child (this, loc->inode);
-
-        if (read_child >= 0) {
-                call_child = read_child;
-
-                local->cont.getxattr.last_tried = -1;
-        } else {
-                call_child = afr_first_up_child (priv);
-
-                if (call_child == -1) {
-                        op_errno = ENOTCONN;
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "no child is up");
-                        goto out;
-                }
-
-                local->cont.getxattr.last_tried = call_child;
+        read_child = afr_inode_get_read_ctx (this, loc->inode, local->fresh_children);
+        ret = afr_get_call_child (this, local->child_up, read_child,
+                                     local->fresh_children,
+                                     &call_child,
+                                     &local->cont.getxattr.last_index);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto out;
         }
-
 	loc_copy (&local->loc, loc);
 	if (name)
 	  local->cont.getxattr.name       = gf_strdup (name);
@@ -1318,13 +1455,12 @@ pump_getxattr (call_frame_t *frame, xlator_t *this,
 	STACK_WIND_COOKIE (frame, pump_getxattr_cbk,
 			   (void *) (long) call_child,
 			   children[call_child], children[call_child]->fops->getxattr,
-			   loc, name);
+			   loc, name, xdata);
 
-	op_ret = 0;
+	ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, NULL);
-	}
+	if (ret < 0)
+		AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
 	return 0;
 }
 
@@ -1332,11 +1468,9 @@ static int
 afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
 {
 	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
 	call_frame_t   *main_frame = NULL;
 
 	local = frame->local;
-	priv  = this->private;
 
 	LOCK (&frame->lock);
 	{
@@ -1348,14 +1482,14 @@ afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
 
 	if (main_frame) {
 		AFR_STACK_UNWIND (setxattr, main_frame,
-                                  local->op_ret, local->op_errno)
+                                  local->op_ret, local->op_errno, NULL);
 	}
 	return 0;
 }
 
 static int
 afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno)
+		       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
 	afr_local_t *   local = NULL;
 	afr_private_t * priv  = NULL;
@@ -1407,7 +1541,7 @@ afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
 	local = frame->local;
 	priv = this->private;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+	call_count = afr_up_children_count (local->child_up, priv->child_count);
 
 	if (call_count == 0) {
 		local->transaction.resume (frame, this);
@@ -1424,7 +1558,7 @@ afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
 					   priv->children[i]->fops->setxattr,
 					   &local->loc,
 					   local->cont.setxattr.dict,
-					   local->cont.setxattr.flags);
+					   local->cont.setxattr.flags, NULL);
 
 			if (!--call_count)
 				break;
@@ -1452,11 +1586,9 @@ pump_setxattr_cbk (call_frame_t *frame,
 		      void *cookie,
 		      xlator_t *this,
 		      int32_t op_ret,
-		      int32_t op_errno)
+		      int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno);
+	AFR_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
 	return 0;
 }
 
@@ -1468,18 +1600,16 @@ pump_command_reply (call_frame_t *frame, xlator_t *this)
         local = frame->local;
 
         if (local->op_ret < 0)
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "Command failed");
         else
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "Command succeeded");
 
-        dict_unref (local->dict);
-
         AFR_STACK_UNWIND (setxattr,
                           frame,
                           local->op_ret,
-                          local->op_errno);
+                          local->op_errno, NULL);
 
         return 0;
 }
@@ -1505,56 +1635,64 @@ pump_parse_command (call_frame_t *frame, xlator_t *this,
                 frame->local = local;
                 local->dict = dict_ref (dict);
                 ret = pump_execute_abort (frame, this);
+
+        } else if (pump_command_commit (this, dict)) {
+                frame->local = local;
+                local->dict = dict_ref (dict);
+                ret = pump_execute_commit (frame, this);
         }
         return ret;
 }
 
 int
 pump_setxattr (call_frame_t *frame, xlator_t *this,
-               loc_t *loc, dict_t *dict, int32_t flags)
+               loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata)
 {
 	afr_private_t * priv  = NULL;
 	afr_local_t   * local = NULL;
 	call_frame_t   *transaction_frame = NULL;
-
 	int ret = -1;
-
-	int op_ret   = -1;
 	int op_errno = 0;
 
 	VALIDATE_OR_GOTO (frame, out);
 	VALIDATE_OR_GOTO (this, out);
 	VALIDATE_OR_GOTO (this->private, out);
 
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.pump*", dict,
+                                   op_errno, out);
+
 	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame, default_setxattr_cbk,
+                            FIRST_CHILD (this),
+                            (FIRST_CHILD (this))->fops->setxattr,
+                            loc, dict, flags, xdata);
+                return 0;
+        }
+
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+	AFR_LOCAL_ALLOC_OR_GOTO (local, out);
 
-	ret = AFR_LOCAL_INIT (local, priv);
+	ret = afr_local_init (local, priv, &op_errno);
 	if (ret < 0) {
-		op_errno = -ret;
+                afr_local_cleanup (local, this);
 		goto out;
-	}
+        }
 
         ret = pump_parse_command (frame, this,
                                   local, dict);
         if (ret >= 0) {
-                op_ret = 0;
+                ret = 0;
                 goto out;
         }
 
-        if (!priv->use_afr_in_pump) {
-                STACK_WIND (frame, default_setxattr_cbk,
-                            FIRST_CHILD (this),
-                            (FIRST_CHILD (this))->fops->setxattr,
-                            loc, dict, flags);
-                return 0;
-        }
-
 	transaction_frame = copy_frame (frame);
 	if (!transaction_frame) {
 		gf_log (this->name, GF_LOG_ERROR,
 			"Out of memory.");
+                op_errno = ENOMEM;
+                ret = -1;
+                afr_local_cleanup (local, this);
 		goto out;
 	}
 
@@ -1577,12 +1715,12 @@ pump_setxattr (call_frame_t *frame, xlator_t *this,
 
 	afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
 
-	op_ret = 0;
+	ret = 0;
 out:
-	if (op_ret == -1) {
+	if (ret < 0) {
 		if (transaction_frame)
 			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (setxattr, frame, op_ret, op_errno);
+		AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
 	}
 
 	return 0;
@@ -1616,7 +1754,7 @@ static int32_t
 pump_truncate (call_frame_t *frame,
                xlator_t *this,
                loc_t *loc,
-               off_t offset)
+               off_t offset, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1626,11 +1764,11 @@ pump_truncate (call_frame_t *frame,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->truncate,
                             loc,
-                            offset);
+                            offset, xdata);
                 return 0;
         }
 
-        afr_truncate (frame, this, loc, offset);
+        afr_truncate (frame, this, loc, offset, xdata);
         return 0;
 }
 
@@ -1639,7 +1777,7 @@ static int32_t
 pump_ftruncate (call_frame_t *frame,
                 xlator_t *this,
                 fd_t *fd,
-                off_t offset)
+                off_t offset, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1649,11 +1787,11 @@ pump_ftruncate (call_frame_t *frame,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->ftruncate,
                             fd,
-                            offset);
+                            offset, xdata);
                 return 0;
         }
 
-        afr_ftruncate (frame, this, fd, offset);
+        afr_ftruncate (frame, this, fd, offset, xdata);
         return 0;
 }
 
@@ -1662,7 +1800,7 @@ pump_ftruncate (call_frame_t *frame,
 
 int
 pump_mknod (call_frame_t *frame, xlator_t *this,
-            loc_t *loc, mode_t mode, dev_t rdev, dict_t *parms)
+            loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1670,10 +1808,10 @@ pump_mknod (call_frame_t *frame, xlator_t *this,
                 STACK_WIND (frame, default_mknod_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->mknod,
-                            loc, mode, rdev, parms);
+                            loc, mode, rdev, umask, xdata);
                 return 0;
         }
-        afr_mknod (frame, this, loc, mode, rdev, parms);
+        afr_mknod (frame, this, loc, mode, rdev, umask, xdata);
         return 0;
 
 }
@@ -1682,7 +1820,7 @@ pump_mknod (call_frame_t *frame, xlator_t *this,
 
 int
 pump_mkdir (call_frame_t *frame, xlator_t *this,
-            loc_t *loc, mode_t mode, dict_t *params)
+            loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1690,10 +1828,10 @@ pump_mkdir (call_frame_t *frame, xlator_t *this,
                 STACK_WIND (frame, default_mkdir_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->mkdir,
-                            loc, mode, params);
+                            loc, mode, umask, xdata);
                 return 0;
         }
-        afr_mkdir (frame, this, loc, mode, params);
+        afr_mkdir (frame, this, loc, mode, umask, xdata);
         return 0;
 
 }
@@ -1702,7 +1840,7 @@ pump_mkdir (call_frame_t *frame, xlator_t *this,
 static int32_t
 pump_unlink (call_frame_t *frame,
              xlator_t *this,
-             loc_t *loc)
+             loc_t *loc, int xflag, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1711,31 +1849,32 @@ pump_unlink (call_frame_t *frame,
                             default_unlink_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->unlink,
-                            loc);
+                            loc, xflag, xdata);
                 return 0;
         }
-        afr_unlink (frame, this, loc);
+        afr_unlink (frame, this, loc, xflag, xdata);
         return 0;
 
 }
 
 
-static int32_t
-pump_rmdir (call_frame_t *frame,
-            xlator_t *this,
-            loc_t *loc)
+static int
+pump_rmdir (call_frame_t *frame, xlator_t *this,
+            loc_t *loc, int flags, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
+
 	priv = this->private;
+
         if (!priv->use_afr_in_pump) {
-                STACK_WIND (frame,
-                            default_rmdir_cbk,
+                STACK_WIND (frame, default_rmdir_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->rmdir,
-                            loc);
+                            loc, flags, xdata);
                 return 0;
         }
-        afr_rmdir (frame, this, loc);
+
+        afr_rmdir (frame, this, loc, flags, xdata);
         return 0;
 
 }
@@ -1744,7 +1883,7 @@ pump_rmdir (call_frame_t *frame,
 
 int
 pump_symlink (call_frame_t *frame, xlator_t *this,
-              const char *linkpath, loc_t *loc, dict_t *params)
+              const char *linkpath, loc_t *loc, mode_t umask, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1752,10 +1891,10 @@ pump_symlink (call_frame_t *frame, xlator_t *this,
                 STACK_WIND (frame, default_symlink_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->symlink,
-                            linkpath, loc, params);
+                            linkpath, loc, umask, xdata);
                 return 0;
         }
-        afr_symlink (frame, this, linkpath, loc, params);
+        afr_symlink (frame, this, linkpath, loc, umask, xdata);
         return 0;
 
 }
@@ -1765,7 +1904,7 @@ static int32_t
 pump_rename (call_frame_t *frame,
              xlator_t *this,
              loc_t *oldloc,
-             loc_t *newloc)
+             loc_t *newloc, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1774,10 +1913,10 @@ pump_rename (call_frame_t *frame,
                             default_rename_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->rename,
-                            oldloc, newloc);
+                            oldloc, newloc, xdata);
                 return 0;
         }
-        afr_rename (frame, this, oldloc, newloc);
+        afr_rename (frame, this, oldloc, newloc, xdata);
         return 0;
 
 }
@@ -1787,7 +1926,7 @@ static int32_t
 pump_link (call_frame_t *frame,
            xlator_t *this,
            loc_t *oldloc,
-           loc_t *newloc)
+           loc_t *newloc, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1796,10 +1935,10 @@ pump_link (call_frame_t *frame,
                             default_link_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->link,
-                            oldloc, newloc);
+                            oldloc, newloc, xdata);
                 return 0;
         }
-        afr_link (frame, this, oldloc, newloc);
+        afr_link (frame, this, oldloc, newloc, xdata);
         return 0;
 
 }
@@ -1808,7 +1947,7 @@ pump_link (call_frame_t *frame,
 static int32_t
 pump_create (call_frame_t *frame, xlator_t *this,
              loc_t *loc, int32_t flags, mode_t mode,
-             fd_t *fd, dict_t *params)
+             mode_t umask, fd_t *fd, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1816,10 +1955,10 @@ pump_create (call_frame_t *frame, xlator_t *this,
                 STACK_WIND (frame, default_create_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->create,
-                            loc, flags, mode, fd, params);
+                            loc, flags, mode, umask, fd, xdata);
                 return 0;
         }
-        afr_create (frame, this, loc, flags, mode, fd, params);
+        afr_create (frame, this, loc, flags, mode, umask, fd, xdata);
         return 0;
 
 }
@@ -1829,8 +1968,7 @@ static int32_t
 pump_open (call_frame_t *frame,
            xlator_t *this,
            loc_t *loc,
-           int32_t flags, fd_t *fd,
-           int32_t wbflags)
+           int32_t flags, fd_t *fd, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1839,10 +1977,10 @@ pump_open (call_frame_t *frame,
                             default_open_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->open,
-                            loc, flags, fd, wbflags);
+                            loc, flags, fd, xdata);
                 return 0;
         }
-        afr_open (frame, this, loc, flags, fd, wbflags);
+        afr_open (frame, this, loc, flags, fd, xdata);
         return 0;
 
 }
@@ -1854,8 +1992,8 @@ pump_writev (call_frame_t *frame,
              fd_t *fd,
              struct iovec *vector,
              int32_t count,
-             off_t off,
-             struct iobref *iobref)
+             off_t off, uint32_t flags,
+             struct iobref *iobref, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1867,20 +2005,20 @@ pump_writev (call_frame_t *frame,
                             fd,
                             vector,
                             count,
-                            off,
-                            iobref);
+                            off, flags,
+                            iobref, xdata);
                 return 0;
         }
-        afr_writev (frame, this, fd, vector, count, off, iobref);
-        return 0;
 
+        afr_writev (frame, this, fd, vector, count, off, flags, iobref, xdata);
+        return 0;
 }
 
 
 static int32_t
 pump_flush (call_frame_t *frame,
             xlator_t *this,
-            fd_t *fd)
+            fd_t *fd, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1889,10 +2027,10 @@ pump_flush (call_frame_t *frame,
                             default_flush_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->flush,
-                            fd);
+                            fd, xdata);
                 return 0;
         }
-        afr_flush (frame, this, fd);
+        afr_flush (frame, this, fd, xdata);
         return 0;
 
 }
@@ -1902,7 +2040,7 @@ static int32_t
 pump_fsync (call_frame_t *frame,
             xlator_t *this,
             fd_t *fd,
-            int32_t flags)
+            int32_t flags, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1912,10 +2050,10 @@ pump_fsync (call_frame_t *frame,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->fsync,
                             fd,
-                            flags);
+                            flags, xdata);
                 return 0;
         }
-        afr_fsync (frame, this, fd, flags);
+        afr_fsync (frame, this, fd, flags, xdata);
         return 0;
 
 }
@@ -1924,7 +2062,7 @@ pump_fsync (call_frame_t *frame,
 static int32_t
 pump_opendir (call_frame_t *frame,
               xlator_t *this,
-              loc_t *loc, fd_t *fd)
+              loc_t *loc, fd_t *fd, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1933,10 +2071,10 @@ pump_opendir (call_frame_t *frame,
                             default_opendir_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->opendir,
-                            loc, fd);
+                            loc, fd, xdata);
                 return 0;
         }
-        afr_opendir (frame, this, loc, fd);
+        afr_opendir (frame, this, loc, fd, xdata);
         return 0;
 
 }
@@ -1946,7 +2084,7 @@ static int32_t
 pump_fsyncdir (call_frame_t *frame,
                xlator_t *this,
                fd_t *fd,
-               int32_t flags)
+               int32_t flags, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1956,10 +2094,10 @@ pump_fsyncdir (call_frame_t *frame,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->fsyncdir,
                             fd,
-                            flags);
+                            flags, xdata);
                 return 0;
         }
-        afr_fsyncdir (frame, this, fd, flags);
+        afr_fsyncdir (frame, this, fd, flags, xdata);
         return 0;
 
 }
@@ -1970,7 +2108,7 @@ pump_xattrop (call_frame_t *frame,
               xlator_t *this,
               loc_t *loc,
               gf_xattrop_flags_t flags,
-              dict_t *dict)
+              dict_t *dict, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1981,10 +2119,10 @@ pump_xattrop (call_frame_t *frame,
                             FIRST_CHILD(this)->fops->xattrop,
                             loc,
                             flags,
-                            dict);
+                            dict, xdata);
                 return 0;
         }
-        afr_xattrop (frame, this, loc, flags, dict);
+        afr_xattrop (frame, this, loc, flags, dict, xdata);
         return 0;
 
 }
@@ -1994,7 +2132,7 @@ pump_fxattrop (call_frame_t *frame,
                xlator_t *this,
                fd_t *fd,
                gf_xattrop_flags_t flags,
-               dict_t *dict)
+               dict_t *dict, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2005,10 +2143,10 @@ pump_fxattrop (call_frame_t *frame,
                             FIRST_CHILD(this)->fops->fxattrop,
                             fd,
                             flags,
-                            dict);
+                            dict, xdata);
                 return 0;
         }
-        afr_fxattrop (frame, this, fd, flags, dict);
+        afr_fxattrop (frame, this, fd, flags, dict, xdata);
         return 0;
 
 }
@@ -2018,9 +2156,17 @@ static int32_t
 pump_removexattr (call_frame_t *frame,
                   xlator_t *this,
                   loc_t *loc,
-                  const char *name)
+                  const char *name, dict_t *xdata)
 {
-        afr_private_t *priv  = NULL;
+        afr_private_t *priv     = NULL;
+        int            op_errno = -1;
+
+        VALIDATE_OR_GOTO (this, out);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.pump*",
+                                 name, op_errno, out);
+
+        op_errno = 0;
 	priv = this->private;
         if (!priv->use_afr_in_pump) {
                 STACK_WIND (frame,
@@ -2028,10 +2174,14 @@ pump_removexattr (call_frame_t *frame,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->removexattr,
                             loc,
-                            name);
+                            name, xdata);
                 return 0;
         }
-        afr_removexattr (frame, this, loc, name);
+        afr_removexattr (frame, this, loc, name, xdata);
+
+ out:
+        if (op_errno)
+                AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
         return 0;
 
 }
@@ -2043,7 +2193,7 @@ pump_readdir (call_frame_t *frame,
               xlator_t *this,
               fd_t *fd,
               size_t size,
-              off_t off)
+              off_t off, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2052,21 +2202,18 @@ pump_readdir (call_frame_t *frame,
                             default_readdir_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->readdir,
-                            fd, size, off);
+                            fd, size, off, xdata);
                 return 0;
         }
-        afr_readdir (frame, this, fd, size, off);
+        afr_readdir (frame, this, fd, size, off, xdata);
         return 0;
 
 }
 
 
 static int32_t
-pump_readdirp (call_frame_t *frame,
-               xlator_t *this,
-               fd_t *fd,
-               size_t size,
-               off_t off)
+pump_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd,
+               size_t size, off_t off, dict_t *dict)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2075,10 +2222,10 @@ pump_readdirp (call_frame_t *frame,
                             default_readdirp_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->readdirp,
-                            fd, size, off);
+                            fd, size, off, dict);
                 return 0;
         }
-        afr_readdirp (frame, this, fd, size, off);
+        afr_readdirp (frame, this, fd, size, off, dict);
         return 0;
 
 }
@@ -2109,13 +2256,24 @@ pump_release (xlator_t *this,
 
 }
 
+static int32_t
+pump_forget (xlator_t *this, inode_t *inode)
+{
+        afr_private_t  *priv  = NULL;
+
+        priv = this->private;
+        if (priv->use_afr_in_pump)
+                afr_forget (this, inode);
+
+        return 0;
+}
 
 static int32_t
 pump_setattr (call_frame_t *frame,
               xlator_t *this,
               loc_t *loc,
               struct iatt *stbuf,
-              int32_t valid)
+              int32_t valid, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2124,10 +2282,10 @@ pump_setattr (call_frame_t *frame,
                             default_setattr_cbk,
                             FIRST_CHILD (this),
                             FIRST_CHILD (this)->fops->setattr,
-                            loc, stbuf, valid);
+                            loc, stbuf, valid, xdata);
                 return 0;
         }
-        afr_setattr (frame, this, loc, stbuf, valid);
+        afr_setattr (frame, this, loc, stbuf, valid, xdata);
         return 0;
 
 }
@@ -2138,7 +2296,7 @@ pump_fsetattr (call_frame_t *frame,
                xlator_t *this,
                fd_t *fd,
                struct iatt *stbuf,
-               int32_t valid)
+               int32_t valid, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2147,10 +2305,10 @@ pump_fsetattr (call_frame_t *frame,
                             default_fsetattr_cbk,
                             FIRST_CHILD (this),
                             FIRST_CHILD (this)->fops->fsetattr,
-                            fd, stbuf, valid);
+                            fd, stbuf, valid, xdata);
                 return 0;
         }
-        afr_fsetattr (frame, this, fd, stbuf, valid);
+        afr_fsetattr (frame, this, fd, stbuf, valid, xdata);
         return 0;
 
 }
@@ -2199,7 +2357,7 @@ notify (xlator_t *this, int32_t event,
 
         child_xl = (xlator_t *) data;
 
-        ret = afr_notify (this, event, data);
+        ret = afr_notify (this, event, data, NULL);
 
 	switch (event) {
 	case GF_EVENT_CHILD_DOWN:
@@ -2210,6 +2368,8 @@ notify (xlator_t *this, int32_t event,
         case GF_EVENT_CHILD_UP:
                 if (is_xlator_pump_sink (child_xl))
                         if (is_pump_start_pending (this)) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "about to start synctask");
                                 ret = pump_start_synctask (this);
                                 if (ret < 0)
                                         gf_log (this->name, GF_LOG_DEBUG,
@@ -2232,7 +2392,7 @@ init (xlator_t *this)
 	xlator_list_t * trav        = NULL;
 	int             i           = 0;
 	int             ret         = -1;
-	int             op_errno    = 0;
+	GF_UNUSED int   op_errno    = 0;
 
         int source_child = 0;
 
@@ -2248,25 +2408,42 @@ init (xlator_t *this)
 			"Volume is dangling.");
 	}
 
-	ALLOC_OR_GOTO (this->private, afr_private_t, out);
+	this->private = GF_CALLOC (1, sizeof (afr_private_t),
+                                   gf_afr_mt_afr_private_t);
+        if (!this->private)
+                goto out;
 
 	priv = this->private;
+        LOCK_INIT (&priv->lock);
+        LOCK_INIT (&priv->read_child_lock);
+        //lock recovery is not done in afr
+        pthread_mutex_init (&priv->mutex, NULL);
+        INIT_LIST_HEAD (&priv->saved_fds);
+
+        child_count = xlator_subvolume_count (this);
+        if (child_count != 2) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "There should be exactly 2 children - one source "
+                        "and one sink");
+                return -1;
+        }
+	priv->child_count = child_count;
 
         priv->read_child = source_child;
         priv->favorite_child = source_child;
         priv->background_self_heal_count = 0;
 
-	priv->data_self_heal     = 1;
+	priv->data_self_heal     = "on";
 	priv->metadata_self_heal = 1;
 	priv->entry_self_heal    = 1;
 
-        priv->data_self_heal_algorithm = "";
-
         priv->data_self_heal_window_size = 16;
 
 	priv->data_change_log     = 1;
 	priv->metadata_change_log = 1;
 	priv->entry_change_log    = 1;
+        priv->use_afr_in_pump = 1;
+        priv->sh_readdir_size = 65536;
 
 	/* Locking options */
 
@@ -2275,31 +2452,9 @@ init (xlator_t *this)
            and the sink.
         */
 
-	priv->data_lock_server_count = 2;
-	priv->metadata_lock_server_count = 2;
-	priv->entry_lock_server_count = 2;
-
 	priv->strict_readdir = _gf_false;
 
-        trav = this->children;
-        while (trav) {
-                child_count++;
-                trav = trav->next;
-        }
-
 	priv->wait_count = 1;
-
-        if (child_count != 2) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "There should be exactly 2 children - one source "
-                        "and one sink");
-                return -1;
-        }
-	priv->child_count = child_count;
-
-	LOCK_INIT (&priv->lock);
-        LOCK_INIT (&priv->read_child_lock);
-
 	priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
                                  gf_afr_mt_char);
 	if (!priv->child_up) {
@@ -2333,8 +2488,8 @@ init (xlator_t *this)
 	while (i < child_count) {
 		priv->children[i] = trav->xlator;
 
-                ret = asprintf (&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX,
-                                trav->xlator->name);
+                ret = gf_asprintf (&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX,
+                                   trav->xlator->name);
                 if (-1 == ret) {
                         gf_log (this->name, GF_LOG_ERROR,
                                 "asprintf failed to set pending key");
@@ -2346,9 +2501,22 @@ init (xlator_t *this)
 		i++;
 	}
 
+        ret = gf_asprintf (&priv->sh_domain, "%s-self-heal", this->name);
+        if (-1 == ret) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+
         priv->first_lookup = 1;
         priv->root_inode = NULL;
 
+        priv->last_event = GF_CALLOC (child_count, sizeof (*priv->last_event),
+                                      gf_afr_mt_int32_t);
+        if (!priv->last_event) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
 	pump_priv = GF_CALLOC (1, sizeof (*pump_priv),
                             gf_afr_mt_pump_priv);
 	if (!pump_priv) {
@@ -2370,7 +2538,7 @@ init (xlator_t *this)
                 goto out;
         }
 
-	pump_priv->env = syncenv_new (0);
+	pump_priv->env = this->ctx->env;
         if (!pump_priv->env) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "Could not create new sync-environment");
@@ -2378,6 +2546,15 @@ init (xlator_t *this)
                 goto out;
         }
 
+        /* keep more local here as we may need them for self-heal etc */
+        this->local_pool = mem_pool_new (afr_local_t, 128);
+        if (!this->local_pool) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto out;
+        }
+
 	priv->pump_private = pump_priv;
 
         pump_change_state (this, PUMP_STATE_ABORT);
@@ -2390,6 +2567,25 @@ out:
 int
 fini (xlator_t *this)
 {
+        afr_private_t * priv        = NULL;
+        pump_private_t *pump_priv = NULL;
+
+        priv      = this->private;
+        this->private = NULL;
+        if (!priv)
+                goto out;
+
+        pump_priv = priv->pump_private;
+        if (!pump_priv)
+                goto afr_priv;
+
+        GF_FREE (pump_priv->resume_path);
+        LOCK_DESTROY (&pump_priv->resume_path_lock);
+        LOCK_DESTROY (&pump_priv->pump_state_lock);
+        GF_FREE (pump_priv);
+afr_priv:
+        afr_priv_destroy (priv);
+out:
 	return 0;
 }
 
@@ -2402,6 +2598,7 @@ struct xlator_fops fops = {
 	.fsyncdir    = pump_fsyncdir,
 	.xattrop     = pump_xattrop,
 	.fxattrop    = pump_fxattrop,
+        .getxattr    = pump_getxattr,
 
 	/* inode write */
 	.writev      = pump_writev,
@@ -2436,6 +2633,7 @@ struct xlator_dumpops dumpops = {
 struct xlator_cbks cbks = {
 	.release     = pump_release,
 	.releasedir  = pump_releasedir,
+        .forget      = pump_forget,
 };
 
 struct volume_options options[] = {
diff --git a/xlators/cluster/afr/src/pump.h b/xlators/cluster/afr/src/pump.h
index 65cfcb55b..bc4c31a78 100644
--- a/xlators/cluster/afr/src/pump.h
+++ b/xlators/cluster/afr/src/pump.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __PUMP_H__
@@ -26,18 +17,6 @@
 #define CLIENT_CMD_CONNECT "trusted.glusterfs.client-connect"
 #define CLIENT_CMD_DISCONNECT "trusted.glusterfs.client-disconnect"
 
-#define PUMP_PID 696969
-#define PUMP_LK_OWNER 696969
-
-#define IS_ROOT_PATH(path) (!strcmp (path, "/"))
-#define IS_ENTRY_CWD(entry) (!strcmp (entry, "."))
-#define IS_ENTRY_PARENT(entry) (!strcmp (entry, ".."))
-
-#define PUMP_CMD_START  "trusted.glusterfs.pump.start"
-#define PUMP_CMD_ABORT  "trusted.glusterfs.pump.abort"
-#define PUMP_CMD_PAUSE  "trusted.glusterfs.pump.pause"
-#define PUMP_CMD_STATUS "trusted.glusterfs.pump.status"
-
 #define PUMP_SOURCE_COMPLETE "trusted.glusterfs.pump-source-complete"
 #define PUMP_SINK_COMPLETE "trusted.glusterfs.pump-sink-complete"
 
@@ -51,11 +30,12 @@ typedef enum {
         PUMP_STATE_RESUME,              /* Pump is resuming from a previous pause */
         PUMP_STATE_PAUSE,               /* Pump is paused */
         PUMP_STATE_ABORT,               /* Pump is aborted */
+        PUMP_STATE_COMMIT,              /* Pump is commited */
 } pump_state_t;
 
 typedef struct _pump_private {
 	struct syncenv *env;            /* The env pointer to the pump synctask */
-        const char *resume_path;        /* path to resume from the last pause */
+        char *resume_path;              /* path to resume from the last pause */
         gf_lock_t resume_path_lock;     /* Synchronize resume_path changes */
         gf_lock_t pump_state_lock;      /* Synchronize pump_state changes */
         pump_state_t pump_state;        /* State of pump */
@@ -64,6 +44,7 @@ typedef struct _pump_private {
         gf_boolean_t pump_finished;     /* Boolean to indicate pump termination */
         char pump_start_pending;        /* Boolean to mark start pending until
                                            CHILD_UP */
+        call_stub_t *cleaner;
 } pump_private_t;
 
 void
diff --git a/xlators/cluster/dht/src/Makefile.am b/xlators/cluster/dht/src/Makefile.am
index 4b69aa071..174bea841 100644
--- a/xlators/cluster/dht/src/Makefile.am
+++ b/xlators/cluster/dht/src/Makefile.am
@@ -2,33 +2,37 @@
 xlator_LTLIBRARIES = dht.la nufa.la switch.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
 
+dht_common_source = dht-layout.c dht-helper.c dht-linkfile.c dht-rebalance.c \
+	dht-selfheal.c dht-rename.c dht-hashfn.c dht-diskusage.c \
+	dht-common.c dht-inode-write.c dht-inode-read.c dht-shared.c \
+	$(top_builddir)/xlators/lib/src/libxlator.c
 
-dht_common_source = dht-layout.c dht-helper.c dht-linkfile.c \
-		dht-selfheal.c dht-rename.c dht-hashfn.c dht-diskusage.c
-
-dht_la_SOURCES = $(dht_common_source) dht.c 
+dht_la_SOURCES = $(dht_common_source) dht.c
 
 nufa_la_SOURCES = $(dht_common_source) nufa.c
 switch_la_SOURCES = $(dht_common_source) switch.c
 
-dht_la_LDFLAGS = -module -avoidversion
+dht_la_LDFLAGS = -module -avoid-version
 dht_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-nufa_la_LDFLAGS = -module -avoidversion
+nufa_la_LDFLAGS = -module -avoid-version
 nufa_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-switch_la_LDFLAGS = -module -avoidversion
+switch_la_LDFLAGS = -module -avoid-version
 switch_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = dht-common.h dht-common.c dht-mem-types.h
+noinst_HEADERS = dht-common.h dht-mem-types.h \
+	$(top_builddir)/xlators/lib/src/libxlator.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+	-I$(top_srcdir)/xlators/lib/src
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
-CLEANFILES = 
+CLEANFILES =
 
 uninstall-local:
 	rm -f $(DESTDIR)$(xlatordir)/distribute.so
 
 install-data-hook:
-	ln -sf dht.so $(DESTDIR)$(xlatordir)/distribute.so
-\ No newline at end of file
+	ln -sf dht.so $(DESTDIR)$(xlatordir)/distribute.so
diff --git a/xlators/cluster/dht/src/dht-common.c b/xlators/cluster/dht/src/dht-common.c
index bec412c41..8f61339e6 100644
--- a/xlators/cluster/dht/src/dht-common.c
+++ b/xlators/cluster/dht/src/dht-common.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2009-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -27,12 +18,88 @@
 
 #include "glusterfs.h"
 #include "xlator.h"
+#include "libxlator.h"
 #include "dht-common.h"
 #include "defaults.h"
+#include "byte-order.h"
+#include "glusterfs-acl.h"
 
 #include <sys/time.h>
 #include <libgen.h>
 
+int
+dht_aggregate (dict_t *this, char *key, data_t *value, void *data)
+{
+        dict_t  *dst  = NULL;
+        int64_t *ptr  = 0, *size = NULL;
+        int32_t  ret  = -1;
+        data_t  *dict_data = NULL;
+
+        dst = data;
+
+        if (strcmp (key, GF_XATTR_QUOTA_SIZE_KEY) == 0) {
+                ret = dict_get_bin (dst, key, (void **)&size);
+                if (ret < 0) {
+                        size = GF_CALLOC (1, sizeof (int64_t),
+                                          gf_common_mt_char);
+                        if (size == NULL) {
+                                gf_log ("dht", GF_LOG_WARNING,
+                                        "memory allocation failed");
+                                return -1;
+                        }
+                        ret = dict_set_bin (dst, key, size, sizeof (int64_t));
+                        if (ret < 0) {
+                                gf_log ("dht", GF_LOG_WARNING,
+                                        "dht aggregate dict set failed");
+                                GF_FREE (size);
+                                return -1;
+                        }
+                }
+
+                ptr = data_to_bin (value);
+                if (ptr == NULL) {
+                        gf_log ("dht", GF_LOG_WARNING, "data to bin failed");
+                        return -1;
+                }
+
+                *size = hton64 (ntoh64 (*size) + ntoh64 (*ptr));
+
+        } else if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0) {
+                ret = gf_get_min_stime (THIS, dst, key, value);
+                if (ret < 0)
+                        return ret;
+        } else {
+                /* compare user xattrs only */
+                if (!strncmp (key, "user.", strlen ("user."))) {
+                        ret = dict_lookup (dst, key, &dict_data);
+                        if (!ret && dict_data && value) {
+                                ret = is_data_equal (dict_data, value);
+                                if (!ret)
+                                        gf_log ("dht", GF_LOG_DEBUG,
+                                                "xattr mismatch for %s", key);
+                        }
+                }
+                ret = dict_set (dst, key, value);
+                if (ret)
+                        gf_log ("dht", GF_LOG_WARNING, "xattr dict set failed");
+        }
+
+        return 0;
+}
+
+
+void
+dht_aggregate_xattr (dict_t *dst, dict_t *src)
+{
+        if ((dst == NULL) || (src == NULL)) {
+                goto out;
+        }
+
+        dict_foreach (src, dht_aggregate, dst);
+out:
+        return;
+}
+
 /* TODO:
    - use volumename in xattr instead of "dht"
    - use NS locks
@@ -43,40 +110,289 @@
 
 int
 dht_lookup_selfheal_cbk (call_frame_t *frame, void *cookie,
-			 xlator_t *this,
-			 int op_ret, int op_errno)
+                         xlator_t *this,
+                         int op_ret, int op_errno, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	dht_layout_t *layout = NULL;
-	int           ret = 0;
+        dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+	int           ret = -1;
 
-	local = frame->local;
-	ret = op_ret;
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
 
-        dht_frame_su_undo (frame);
+        local = frame->local;
+        ret = op_ret;
 
-	if (ret == 0) {
-		layout = local->selfheal.layout;
-		ret = dht_layout_set (this, local->inode, layout);
+        FRAME_SU_UNDO (frame, dht_local_t);
 
-		if (local->ia_ino) {
-			local->stbuf.ia_ino = local->ia_ino;
-		} else {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not find hashed subvolume for %s",
-				local->loc.path);
-		}
+        if (ret == 0) {
+                layout = local->selfheal.layout;
+                ret = dht_layout_set (this, local->inode, layout);
+        }
 
-                if (local->loc.parent)
-                        local->postparent.ia_ino = local->loc.parent->ino;
-	}
+        if (local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           &local->postparent, 1);
+        }
 
-        WIPE (&local->postparent);
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
 
-	DHT_STACK_UNWIND (lookup, frame, ret, local->op_errno, local->inode,
-			  &local->stbuf, local->xattr, &local->postparent);
+        DHT_STACK_UNWIND (lookup, frame, ret, local->op_errno, local->inode,
+                          &local->stbuf, local->xattr, &local->postparent);
 
-	return 0;
+out:
+        return ret;
+}
+
+
+int
+dht_discover_complete (xlator_t *this, call_frame_t *discover_frame)
+{
+        dht_local_t     *local = NULL;
+        call_frame_t    *main_frame = NULL;
+        int              op_errno = 0;
+        int              ret = -1;
+        dht_layout_t    *layout = NULL;
+        dht_conf_t      *conf = NULL;
+
+        local = discover_frame->local;
+        layout = local->layout;
+        conf = this->private;
+
+        LOCK(&discover_frame->lock);
+        {
+                main_frame = local->main_frame;
+                local->main_frame = NULL;
+        }
+        UNLOCK(&discover_frame->lock);
+
+        if (!main_frame)
+                return 0;
+
+        if (local->file_count && local->dir_count) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "path %s exists as a file on one subvolume "
+                        "and directory on another. "
+                        "Please fix it manually",
+                        local->loc.path);
+                op_errno = EIO;
+                goto out;
+        }
+
+        if (local->cached_subvol) {
+                ret = dht_layout_preset (this, local->cached_subvol,
+                                         local->inode);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "failed to set layout for subvolume %s",
+                                local->cached_subvol ? local->cached_subvol->name : "<nil>");
+                        op_errno = EINVAL;
+                        goto out;
+                }
+        } else {
+                ret = dht_layout_normalize (this, &local->loc, layout);
+                if ((ret < 0) || ((ret > 0) && (local->op_ret != 0))) {
+                        /* either the layout is incorrect or the directory is
+                         * not found even in one subvolume.
+                         */
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "normalizing failed on %s "
+                                "(overlaps/holes present: %s, "
+                                "ENOENT errors: %d)", local->loc.path,
+                                (ret < 0) ? "yes" : "no", (ret > 0) ? ret : 0);
+                        if ((ret > 0) && (ret == conf->subvolume_cnt)) {
+                                op_errno = ESTALE;
+                                goto out;
+                        }
+                }
+
+                if (local->inode)
+                        dht_layout_set (this, local->inode, layout);
+        }
+
+        DHT_STACK_UNWIND (lookup, main_frame, local->op_ret, local->op_errno,
+                          local->inode, &local->stbuf, local->xattr,
+                          &local->postparent);
+        return 0;
+out:
+        DHT_STACK_UNWIND (lookup, main_frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL);
+
+        return ret;
+}
+
+
+int
+dht_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int op_ret, int op_errno,
+                  inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                  struct iatt *postparent)
+{
+        dht_local_t  *local                   = NULL;
+        int           this_call_cnt           = 0;
+        call_frame_t *prev                    = NULL;
+        dht_layout_t *layout                  = NULL;
+        int           ret                     = -1;
+        int           is_dir                  = 0;
+        int           is_linkfile             = 0;
+        int           attempt_unwind          = 0;
+        dht_conf_t   *conf                    = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev  = cookie;
+        conf = this->private;
+
+        layout = local->layout;
+
+        /* Check if the gfid is different for file from other node */
+        if (!op_ret && uuid_compare (local->gfid, stbuf->ia_gfid)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: gfid different on %s",
+                        local->loc.path, prev->this->name);
+        }
+
+
+        LOCK (&frame->lock);
+        {
+                /* TODO: assert equal mode on stbuf->st_mode and
+                   local->stbuf->st_mode
+
+                   else mkdir/chmod/chown and fix
+                */
+                ret = dht_layout_merge (this, layout, prev->this,
+                                        op_ret, op_errno, xattr);
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to merge layouts", local->loc.path);
+
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "lookup of %s on %s returned error (%s)",
+                                local->loc.path, prev->this->name,
+                                strerror (op_errno));
+
+                        goto unlock;
+                }
+
+                is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                                 conf->link_xattr_name);
+                is_dir = check_is_dir (inode, stbuf, xattr);
+
+                if (is_dir) {
+                        local->dir_count ++;
+                } else {
+                        local->file_count ++;
+
+                        if (!is_linkfile) {
+                                /* real file */
+                                local->cached_subvol = prev->this;
+                                attempt_unwind = 1;
+                        } else {
+                                goto unlock;
+                        }
+                }
+
+                local->op_ret = 0;
+
+                if (local->xattr == NULL) {
+                        local->xattr = dict_ref (xattr);
+                } else {
+                        dht_aggregate_xattr (local->xattr, xattr);
+                }
+
+                if (local->inode == NULL)
+                        local->inode = inode_ref (inode);
+
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                dht_iatt_merge (this, &local->postparent, postparent,
+                                prev->this);
+        }
+unlock:
+        UNLOCK (&frame->lock);
+out:
+        this_call_cnt = dht_frame_return (frame);
+
+        if (is_last_call (this_call_cnt) || attempt_unwind) {
+                dht_discover_complete (this, frame);
+        }
+
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_DESTROY (frame);
+
+        return 0;
+}
+
+
+int
+dht_discover (call_frame_t *frame, xlator_t *this, loc_t *loc)
+{
+        int          ret;
+        dht_local_t *local = NULL;
+        dht_conf_t  *conf = NULL;
+        int          call_cnt = 0;
+        int          op_errno = EINVAL;
+        int          i = 0;
+        call_frame_t *discover_frame = NULL;
+
+        conf = this->private;
+        local = frame->local;
+
+        ret = dict_set_uint32 (local->xattr_req, conf->xattr_name, 4 * 4);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set '%s' key",
+                        loc->path, conf->xattr_name);
+
+        ret = dict_set_uint32 (local->xattr_req, conf->link_xattr_name, 256);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set '%s' key",
+                        loc->path, conf->link_xattr_name);
+
+        call_cnt        = conf->subvolume_cnt;
+        local->call_cnt = call_cnt;
+
+        local->layout = dht_layout_new (this, conf->subvolume_cnt);
+
+        if (!local->layout) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        uuid_copy (local->gfid, loc->gfid);
+
+        discover_frame = copy_frame (frame);
+        if (!discover_frame) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        discover_frame->local = local;
+        frame->local = NULL;
+        local->main_frame = frame;
+
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (discover_frame, dht_discover_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->lookup,
+                            &local->loc, local->xattr_req);
+        }
+
+        return 0;
+
+err:
+        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL);
+
+        return 0;
 }
 
 
@@ -86,64 +402,77 @@ dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     inode_t *inode, struct iatt *stbuf, dict_t *xattr,
                     struct iatt *postparent)
 {
-	dht_conf_t   *conf                    = NULL;
         dht_local_t  *local                   = NULL;
         int           this_call_cnt           = 0;
         call_frame_t *prev                    = NULL;
-	dht_layout_t *layout                  = NULL;
-	int           ret                     = 0;
-	int           is_dir                  = 0;
+        dht_layout_t *layout                  = NULL;
+        int           ret                     = -1;
+        int           is_dir                  = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
 
-	conf  = this->private;
         local = frame->local;
         prev  = cookie;
 
-	layout = local->layout;
+        layout = local->layout;
+
+        if (!op_ret && uuid_is_null (local->gfid))
+                memcpy (local->gfid, stbuf->ia_gfid, 16);
+
+        /* Check if the gfid is different for file from other node */
+        if (!op_ret && uuid_compare (local->gfid, stbuf->ia_gfid)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: gfid different on %s",
+                        local->loc.path, prev->this->name);
+        }
 
         LOCK (&frame->lock);
         {
                 /* TODO: assert equal mode on stbuf->st_mode and
-		   local->stbuf->st_mode
+                   local->stbuf->st_mode
 
-		   else mkdir/chmod/chown and fix
-		*/
-		ret = dht_layout_merge (this, layout, prev->this,
-					op_ret, op_errno, xattr);
+                   else mkdir/chmod/chown and fix
+                */
+                ret = dht_layout_merge (this, layout, prev->this,
+                                        op_ret, op_errno, xattr);
 
-		if (op_ret == -1) {
-			local->op_errno = ENOENT;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"lookup of %s on %s returned error (%s)",
-				local->loc.path, prev->this->name,
-				strerror (op_errno));
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "lookup of %s on %s returned error (%s)",
+                                local->loc.path, prev->this->name,
+                                strerror (op_errno));
 
-			goto unlock;
-		}
+                        goto unlock;
+                }
 
- 		is_dir = check_is_dir (inode, stbuf, xattr);
- 		if (!is_dir) {
+                is_dir = check_is_dir (inode, stbuf, xattr);
+                if (!is_dir) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "lookup of %s on %s returned non dir 0%o",
                                 local->loc.path, prev->this->name,
                                 stbuf->ia_type);
                         local->need_selfheal = 1;
- 			goto unlock;
+                        goto unlock;
                 }
 
- 		local->op_ret = 0;
- 		if (local->xattr == NULL)
- 			local->xattr = dict_ref (xattr);
- 		if (local->inode == NULL)
- 			local->inode = inode_ref (inode);
+                local->op_ret = 0;
+                if (local->xattr == NULL) {
+                        local->xattr = dict_ref (xattr);
+                } else {
+                        dht_aggregate_xattr (local->xattr, xattr);
+                }
 
-		dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                if (local->inode == NULL)
+                        local->inode = inode_ref (inode);
+
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
                 dht_iatt_merge (this, &local->postparent, postparent,
                                 prev->this);
-
-                if (prev->this == dht_first_up_subvol (this)) {
-			local->ia_ino = local->stbuf.ia_ino;
-                }
-
         }
 unlock:
         UNLOCK (&frame->lock);
@@ -158,44 +487,39 @@ unlock:
                         return 0;
                 }
 
-		if (local->op_ret == 0) {
-			ret = dht_layout_normalize (this, &local->loc, layout);
+                if (local->op_ret == 0) {
+                        ret = dht_layout_normalize (this, &local->loc, layout);
 
-			if (ret != 0) {
-				gf_log (this->name, GF_LOG_DEBUG,
-					"fixing assignment on %s",
-					local->loc.path);
-				goto selfheal;
-			}
-
-			dht_layout_set (this, local->inode, layout);
+                        if (ret != 0) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "fixing assignment on %s",
+                                        local->loc.path);
+                                goto selfheal;
+                        }
 
-			if (local->ia_ino) {
-				local->stbuf.ia_ino = local->ia_ino;
-			} else {
-				gf_log (this->name, GF_LOG_DEBUG,
-					"could not find hashed subvol for %s",
-					local->loc.path);
-			}
+                        dht_layout_set (this, local->inode, layout);
+                }
 
-                        if (local->loc.parent)
-                                local->postparent.ia_ino =
-                                        local->loc.parent->ino;
-		}
+                if (local->loc.parent) {
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
+                }
 
-		DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-				  local->inode, &local->stbuf, local->xattr,
+                DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+                DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+                                  local->inode, &local->stbuf, local->xattr,
                                   &local->postparent);
         }
 
-	return 0;
+        return 0;
 
 selfheal:
-        dht_frame_su_do (frame);
-	ret = dht_selfheal_directory (frame, dht_lookup_selfheal_cbk,
-				      &local->loc, layout);
-
-	return 0;
+        FRAME_SU_DO (frame, dht_local_t);
+        uuid_copy (local->loc.gfid, local->gfid);
+        ret = dht_selfheal_directory (frame, dht_lookup_selfheal_cbk,
+                                      &local->loc, layout);
+out:
+        return ret;
 }
 
 int
@@ -207,137 +531,224 @@ dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         dht_local_t  *local         = NULL;
         int           this_call_cnt = 0;
         call_frame_t *prev          = NULL;
-	dht_layout_t *layout        = NULL;
-	dht_conf_t   *conf          = NULL;
-	int           ret  = -1;
-	int           is_dir = 0;
-	int           is_linkfile = 0;
+        dht_layout_t *layout        = NULL;
+        dht_conf_t   *conf          = NULL;
+        int           ret  = -1;
+        int           is_dir = 0;
+        int           is_linkfile = 0;
+        call_frame_t *copy          = NULL;
+        dht_local_t  *copy_local    = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, err);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, err);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, err);
 
         local = frame->local;
         prev  = cookie;
-	conf = this->private;
+        conf = this->private;
+        if (!conf)
+                goto out;
 
         LOCK (&frame->lock);
         {
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
 
-			if ((op_errno != ENOTCONN) 
+                        if ((op_errno != ENOTCONN)
                             && (op_errno != ENOENT)
                             && (op_errno != ESTALE)) {
-				gf_log (this->name, GF_LOG_DEBUG,
-					"subvolume %s returned -1 (%s)",
-					prev->this->name, strerror (op_errno));
+				gf_log (this->name, GF_LOG_INFO,
+					"subvolume %s for %s returned -1 (%s)",
+					prev->this->name, local->loc.path,
+                                        strerror (op_errno));
 			}
-                        
                         if (op_errno == ESTALE) {
-                                /* propogate the ESTALE to parent. 
-                                 * setting local->layout_mismatch would send
+                                /* propagate the ESTALE to parent.
+                                 * setting local->return_estale would send
                                  * ESTALE to parent. */
-                                local->layout_mismatch = 1;
+                                local->return_estale = 1;
                         }
 
-			goto unlock;
-		}
+                        /* if it is ENOENT, we may have to do a
+                         * 'lookup_everywhere()' to make sure
+                         * the file is not migrated */
+                        if (op_errno == ENOENT) {
+                                if (IA_ISREG (local->loc.inode->ia_type)) {
+                                        local->need_lookup_everywhere = 1;
+                                }
+                        }
+                        goto unlock;
+                }
 
-		if (stbuf->ia_type != local->inode->ia_type) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"mismatching filetypes 0%o v/s 0%o for %s",
-				(stbuf->ia_type), (local->inode->ia_type),
-				local->loc.path);
+                if (stbuf->ia_type != local->inode->ia_type) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "mismatching filetypes 0%o v/s 0%o for %s",
+                                (stbuf->ia_type), (local->inode->ia_type),
+                                local->loc.path);
 
-			local->op_ret = -1;
-			local->op_errno = EINVAL;
+                        local->op_ret = -1;
+                        local->op_errno = EINVAL;
 
-			goto unlock;
-		}
+                        goto unlock;
+                }
 
-		layout = local->layout;
-		
-		is_dir = check_is_dir (inode, stbuf, xattr);
-		is_linkfile = check_is_linkfile (inode, stbuf, xattr);
-		
-		if (is_linkfile) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"linkfile found in revalidate for %s",
-				local->loc.path);
-			local->layout_mismatch = 1;
-
-			goto unlock;
-		}
+                layout = local->layout;
 
-		if (is_dir) {
-			ret = dht_layout_dir_mismatch (this, layout,
-						       prev->this, &local->loc,
-						       xattr);
-			if (ret != 0) {
-				gf_log (this->name, GF_LOG_DEBUG,
-					"mismatching layouts for %s", 
-					local->loc.path);
-			
-				local->layout_mismatch = 1;
-
-				goto unlock;
-			}
-		} 
-		
-		dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                is_dir = check_is_dir (inode, stbuf, xattr);
+                is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                                 conf->link_xattr_name);
+
+                if (is_linkfile) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "linkfile found in revalidate for %s",
+                                local->loc.path);
+                        local->return_estale = 1;
+
+                        goto unlock;
+                }
+
+                if (is_dir) {
+                        ret = dht_dir_has_layout (xattr, conf->xattr_name);
+                        if (ret >= 0) {
+                                if (is_greater_time(local->stbuf.ia_ctime,
+                                                    local->stbuf.ia_ctime_nsec,
+                                                    stbuf->ia_ctime,
+                                                    stbuf->ia_ctime_nsec)) {
+                                        local->prebuf.ia_gid = stbuf->ia_gid;
+                                        local->prebuf.ia_uid = stbuf->ia_uid;
+                                }
+                        }
+                        if (local->stbuf.ia_type != IA_INVAL)
+                        {
+                                if ((local->stbuf.ia_gid != stbuf->ia_gid) ||
+                                    (local->stbuf.ia_uid != stbuf->ia_uid)) {
+                                        local->need_selfheal = 1;
+                                }
+                        }
+                        ret = dht_layout_dir_mismatch (this, layout,
+                                                       prev->this, &local->loc,
+                                                       xattr);
+                        if (ret != 0) {
+                                gf_log (this->name, GF_LOG_INFO,
+                                        "mismatching layouts for %s",
+                                        local->loc.path);
+
+                                local->layout_mismatch = 1;
+
+                                goto unlock;
+                        }
+                }
+
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
                 dht_iatt_merge (this, &local->postparent, postparent,
                                 prev->this);
-		
-		local->op_ret = 0;
-		local->stbuf.ia_ino = local->ia_ino;
 
-                if (local->loc.parent)
-                        local->postparent.ia_ino = local->loc.parent->ino;
+                local->op_ret = 0;
 
-		if (!local->xattr)
-			local->xattr = dict_ref (xattr);
-	}
+                if (!local->xattr) {
+                        local->xattr = dict_ref (xattr);
+                } else if (is_dir) {
+                        dht_aggregate_xattr (local->xattr, xattr);
+                }
+        }
 unlock:
-	UNLOCK (&frame->lock);
-
+        UNLOCK (&frame->lock);
+out:
         this_call_cnt = dht_frame_return (frame);
 
         if (is_last_call (this_call_cnt)) {
-		if (!IA_ISDIR (local->stbuf.ia_type)
-		    && (local->hashed_subvol != local->cached_subvol)
-		    && (local->stbuf.ia_nlink == 1)
-		    && (conf->unhashed_sticky_bit)) {
-			local->stbuf.ia_prot.sticky = 1;
-		}
-
+                if (!IA_ISDIR (local->stbuf.ia_type)
+                    && (local->hashed_subvol != local->cached_subvol)
+                    && (local->stbuf.ia_nlink == 1)
+                    && (conf && conf->unhashed_sticky_bit)) {
+                        local->stbuf.ia_prot.sticky = 1;
+                }
+                if (local->need_selfheal) {
+                        local->need_selfheal = 0;
+                        uuid_copy (local->gfid, local->stbuf.ia_gfid);
+                        local->stbuf.ia_gid = local->prebuf.ia_gid;
+                        local->stbuf.ia_uid = local->prebuf.ia_uid;
+                        copy = create_frame (this, this->ctx->pool);
+                        if (copy) {
+                                copy_local = dht_local_init (copy, &local->loc,
+                                                             NULL, 0);
+                                if (!copy_local)
+                                        goto cont;
+                                copy_local->stbuf = local->stbuf;
+                                copy->local = copy_local;
+                                FRAME_SU_DO (copy, dht_local_t);
+                                ret = synctask_new (this->ctx->env,
+                                                    dht_dir_attr_heal,
+                                                    dht_dir_attr_heal_done,
+                                                    copy, copy);
+                        }
+                }
+cont:
 		if (local->layout_mismatch) {
-			local->op_ret = -1;
-			local->op_errno = ESTALE;
-		}
+                        /* Found layout mismatch in the directory, need to
+                           fix this in the inode context */
+                        dht_layout_unref (this, local->layout);
+                        local->layout = NULL;
+                        dht_lookup_directory (frame, this, &local->loc);
+                        return 0;
+                }
 
-                WIPE (&local->postparent);
+                if (local->need_lookup_everywhere) {
+                        /* As the current layout gave ENOENT error, we would
+                           need a new layout */
+                        dht_layout_unref (this, local->layout);
+                        local->layout = NULL;
+
+                        /* We know that current cached subvol is no more
+                           valid, get the new one */
+                        local->cached_subvol = NULL;
+                        dht_lookup_everywhere (frame, this, &local->loc);
+                        return 0;
+                }
+                if (local->return_estale) {
+                        local->op_ret = -1;
+                        local->op_errno = ESTALE;
+                }
+
+                if (local->loc.parent) {
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
+                }
 
-		DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-				  local->inode, &local->stbuf, local->xattr,
+                DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+                DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+                                  local->inode, &local->stbuf, local->xattr,
                                   &local->postparent);
-	}
+        }
 
-        return 0;
+err:
+        return ret;
 }
 
 
 int
 dht_lookup_linkfile_create_cbk (call_frame_t *frame, void *cookie,
-				xlator_t *this,
-				int32_t op_ret, int32_t op_errno,
+                                xlator_t *this,
+                                int32_t op_ret, int32_t op_errno,
                                 inode_t *inode, struct iatt *stbuf,
-                                struct iatt *preparent, struct iatt *postparent)
+                                struct iatt *preparent, struct iatt *postparent,
+                                dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	xlator_t     *cached_subvol = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_local_t  *local = NULL;
+        xlator_t     *cached_subvol = NULL;
+        dht_conf_t   *conf = NULL;
         int           ret = -1;
 
-	local = frame->local;
-	cached_subvol = local->cached_subvol;
-	conf = this->private;
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        cached_subvol = local->cached_subvol;
+        conf = this->private;
 
         ret = dht_layout_preset (this, local->cached_subvol, inode);
         if (ret < 0) {
@@ -349,72 +760,232 @@ dht_lookup_linkfile_create_cbk (call_frame_t *frame, void *cookie,
                 goto unwind;
         }
 
-	local->op_ret = 0;
-	if ((local->stbuf.ia_nlink == 1)
-	    && (conf->unhashed_sticky_bit)) {
-		local->stbuf.ia_prot.sticky = 1;
-	}
+        local->op_ret = 0;
+        if ((local->stbuf.ia_nlink == 1)
+            && (conf && conf->unhashed_sticky_bit)) {
+                local->stbuf.ia_prot.sticky = 1;
+        }
 
-        if (local->loc.parent)
-                local->postparent.ia_ino = local->loc.parent->ino;
+        if (local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
+        }
 
 unwind:
-        WIPE (&local->postparent);
+        if (local->linked == _gf_true)
+                dht_linkfile_attr_heal (frame, this);
 
-	DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-			  local->inode, &local->stbuf, local->xattr,
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+        DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+                          local->inode, &local->stbuf, local->xattr,
                           &local->postparent);
-	return 0;
+out:
+        return ret;
+}
+
+
+int
+dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this)
+{
+        int           ret = 0;
+        dht_local_t  *local = NULL;
+        xlator_t     *hashed_subvol = NULL;
+        xlator_t     *cached_subvol = NULL;
+        dht_layout_t *layout = NULL;
+
+        local = frame->local;
+        hashed_subvol = local->hashed_subvol;
+        cached_subvol = local->cached_subvol;
+
+        if (local->file_count && local->dir_count) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "path %s exists as a file on one subvolume "
+                        "and directory on another. "
+                        "Please fix it manually",
+                        local->loc.path);
+                DHT_STACK_UNWIND (lookup, frame, -1, EIO, NULL, NULL, NULL,
+                                  NULL);
+                return 0;
+        }
+
+        if (local->dir_count) {
+                dht_lookup_directory (frame, this, &local->loc);
+                return 0;
+        }
+
+        if (!cached_subvol) {
+                DHT_STACK_UNWIND (lookup, frame, -1, ENOENT, NULL, NULL, NULL,
+                                  NULL);
+                return 0;
+        }
+
+        if (local->need_lookup_everywhere) {
+                if (uuid_compare (local->gfid, local->inode->gfid)) {
+                        /* GFID different, return error */
+                        DHT_STACK_UNWIND (lookup, frame, -1, ENOENT, NULL,
+                                          NULL, NULL, NULL);
+                        return 0;
+                }
+                local->op_ret = 0;
+                local->op_errno = 0;
+                layout = dht_layout_for_subvol (this, cached_subvol);
+                if (!layout) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "%s: no pre-set layout for subvolume %s",
+                                local->loc.path, (cached_subvol ?
+                                                  cached_subvol->name :
+                                                  "<nil>"));
+                }
+
+                ret = dht_layout_set (this, local->inode, layout);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "%s: failed to set layout for subvol %s",
+                                local->loc.path, (cached_subvol ?
+                                                  cached_subvol->name :
+                                                  "<nil>"));
+                }
+
+                if (local->loc.parent) {
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
+                }
+
+                DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+                DHT_STACK_UNWIND (lookup, frame, local->op_ret,
+                                  local->op_errno, local->inode,
+                                  &local->stbuf, local->xattr,
+                                  &local->postparent);
+                return 0;
+        }
+
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "cannot create linkfile file for %s on %s: "
+                        "hashed subvolume cannot be found.",
+                        local->loc.path, cached_subvol->name);
+
+                local->op_ret = 0;
+                local->op_errno = 0;
+
+                ret = dht_layout_preset (frame->this, cached_subvol,
+                                         local->inode);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "failed to set layout for subvol %s",
+                                cached_subvol ? cached_subvol->name :
+                                "<nil>");
+                        local->op_ret = -1;
+                        local->op_errno = EINVAL;
+                }
+
+                if (local->loc.parent) {
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
+                }
+
+                DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+                DHT_STACK_UNWIND (lookup, frame, local->op_ret,
+                                  local->op_errno, local->inode,
+                                  &local->stbuf, local->xattr,
+                                  &local->postparent);
+                return 0;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "linking file %s existing on %s to %s (hash)",
+                local->loc.path, cached_subvol->name,
+                hashed_subvol->name);
+
+        ret = dht_linkfile_create (frame,
+                                   dht_lookup_linkfile_create_cbk, this,
+                                   cached_subvol, hashed_subvol, &local->loc);
+
+        return ret;
+}
+
+
+int
+dht_lookup_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int op_ret, int op_errno,
+                       struct iatt *preparent, struct iatt *postparent,
+                       dict_t *xdata)
+{
+        int  this_call_cnt = 0;
+
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                dht_lookup_everywhere_done (frame, this);
+        }
+
+        return 0;
 }
 
 
 int
 dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			   int32_t op_ret, int32_t op_errno,
+                           int32_t op_ret, int32_t op_errno,
                            inode_t *inode, struct iatt *buf, dict_t *xattr,
                            struct iatt *postparent)
 {
-	dht_conf_t   *conf          = NULL;
         dht_local_t  *local         = NULL;
         int           this_call_cnt = 0;
         call_frame_t *prev          = NULL;
-	int           is_linkfile   = 0;
-	int           is_dir        = 0;
-	xlator_t     *subvol        = NULL;
-	loc_t        *loc           = NULL;
-	xlator_t     *link_subvol   = NULL;
-	xlator_t     *hashed_subvol = NULL;
-	xlator_t     *cached_subvol = NULL;
-        int           ret = -1;
+        int           is_linkfile   = 0;
+        int           is_dir        = 0;
+        xlator_t     *subvol        = NULL;
+        loc_t        *loc           = NULL;
+        xlator_t     *link_subvol   = NULL;
+        int           ret           = -1;
+        int32_t       fd_count      = 0;
+        dht_conf_t   *conf          = NULL;
 
-	conf   = this->private;
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
 
-	local  = frame->local;
-	loc    = &local->loc;
+        local  = frame->local;
+        loc    = &local->loc;
+        conf   = this->private;
 
-	prev   = cookie;
-	subvol = prev->this;
+        prev   = cookie;
+        subvol = prev->this;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			if (op_errno != ENOENT)
-				local->op_errno = op_errno;
-			goto unlock;
-		}
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        if (op_errno != ENOENT)
+                                local->op_errno = op_errno;
+                        goto unlock;
+                }
 
-		is_linkfile = check_is_linkfile (inode, buf, xattr);
-		is_dir = check_is_dir (inode, buf, xattr);
-
-		if (is_linkfile) {
-			link_subvol = dht_linkfile_subvol (this, inode, buf,
-							   xattr);
-			gf_log (this->name, GF_LOG_DEBUG,
-				"found on %s linkfile %s (-> %s)",
-				subvol->name, loc->path,
-				link_subvol ? link_subvol->name : "''");
-			goto unlock;
-		}
+                if (uuid_is_null (local->gfid))
+                        uuid_copy (local->gfid, buf->ia_gfid);
+
+                if (uuid_compare (local->gfid, buf->ia_gfid)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: gfid differs on subvolume %s",
+                                loc->path, prev->this->name);
+                }
+
+                is_linkfile = check_is_linkfile (inode, buf, xattr,
+                                                 conf->link_xattr_name);
+                is_dir = check_is_dir (inode, buf, xattr);
+
+                if (is_linkfile) {
+                        link_subvol = dht_linkfile_subvol (this, inode, buf,
+                                                           xattr);
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "found on %s linkfile %s (-> %s)",
+                                subvol->name, loc->path,
+                                link_subvol ? link_subvol->name : "''");
+                        goto unlock;
+                }
+
+                /* non linkfile GFID takes precedence */
+                uuid_copy (local->gfid, buf->ia_gfid);
 
                 if (is_dir) {
                         local->dir_count++;
@@ -434,126 +1005,82 @@ dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 gf_log (this->name, GF_LOG_DEBUG,
                                         "found on %s file %s",
                                         subvol->name, loc->path);
-                                
+
                                 dht_iatt_merge (this, &local->postparent,
                                                 postparent, subvol);
                         } else {
-                                gf_log (this->name, GF_LOG_DEBUG,
+                                /* This is where we need 'rename' both entries logic */
+                                gf_log (this->name, GF_LOG_WARNING,
                                         "multiple subvolumes (%s and %s) have "
-                                        "file %s", local->cached_subvol->name,
+                                        "file %s (preferably rename the file "
+                                        "in the backend, and do a fresh lookup)",
+                                        local->cached_subvol->name,
                                         subvol->name, local->loc.path);
                         }
                 }
-	}
+        }
 unlock:
-	UNLOCK (&frame->lock);
-
-	if (is_linkfile) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"deleting stale linkfile %s on %s",
-			loc->path, subvol->name);
-		dht_linkfile_unlink (frame, this, subvol, loc);
-	}
-
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-		hashed_subvol = local->hashed_subvol;
-		cached_subvol = local->cached_subvol;
-
-                if (local->file_count && local->dir_count) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "path %s exists as a file on one subvolume " 
-                                "and directory on another. "
-                                "Please fix it manually",
-                                loc->path);
-                        DHT_STACK_UNWIND (lookup, frame, -1, EIO, NULL, NULL, NULL,
-                                          NULL);
-                        return 0;
-                }
-
-                if (local->dir_count) {
-                        dht_lookup_directory (frame, this, &local->loc);
-                        return 0;
-                }
-
-		if (!cached_subvol) {
-			DHT_STACK_UNWIND (lookup, frame, -1, ENOENT, NULL, NULL, NULL,
-                                          NULL);
-			return 0;
-		}
-
-                if (!hashed_subvol) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "cannot create linkfile file for %s on %s: "
-                                "hashed subvolume cannot be found.",
-                                loc->path, cached_subvol->name);
-                        
-                        local->op_ret = 0;
-                        local->op_errno = 0;
-
-                        ret = dht_layout_preset (frame->this, cached_subvol,
-                                                 local->inode);
-                        if (ret < 0) {
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "failed to set layout for subvol %s",
-                                        cached_subvol ? cached_subvol->name :
-                                        "<nil>");
-                                local->op_ret = -1;
-                                local->op_errno = EINVAL;
-                        }
-
-                        if (local->loc.parent)
-                                local->postparent.ia_ino =
-                                        local->loc.parent->ino;
-
-                        WIPE (&local->postparent);
+        UNLOCK (&frame->lock);
 
-                        DHT_STACK_UNWIND (lookup, frame, local->op_ret,
-                                          local->op_errno, local->inode,
-                                          &local->stbuf, local->xattr,
-                                          &local->postparent);
+        if (is_linkfile) {
+                ret = dict_get_int32 (xattr, GLUSTERFS_OPEN_FD_COUNT, &fd_count);
+                /* Delete the linkfile only if there are no open fds on it.
+                   if there is a open-fd, it may be in migration */
+                if (!ret && (fd_count == 0)) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "deleting stale linkfile %s on %s",
+                                loc->path, subvol->name);
+                        STACK_WIND (frame, dht_lookup_unlink_cbk,
+                                    subvol, subvol->fops->unlink, loc, 0, NULL);
                         return 0;
                 }
+        }
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "linking file %s existing on %s to %s (hash)",
-                        loc->path, cached_subvol->name,
-                        hashed_subvol->name);
-                        
-                dht_linkfile_create (frame, 
-                                     dht_lookup_linkfile_create_cbk,
-                                     cached_subvol, hashed_subvol, loc);
-	}
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                dht_lookup_everywhere_done (frame, this);
+        }
 
-	return 0;
+out:
+        return ret;
 }
 
 
 int
 dht_lookup_everywhere (call_frame_t *frame, xlator_t *this, loc_t *loc)
 {
-	dht_conf_t     *conf = NULL;
-	dht_local_t    *local = NULL;
-	int             i = 0;
-	int             call_cnt = 0;
+        dht_conf_t     *conf = NULL;
+        dht_local_t    *local = NULL;
+        int             i = 0;
+        int             call_cnt = 0;
 
-	conf = this->private;
-	local = frame->local;
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", loc, out);
+
+        conf = this->private;
+        local = frame->local;
 
-	call_cnt = conf->subvolume_cnt;
-	local->call_cnt = call_cnt;
+        call_cnt = conf->subvolume_cnt;
+        local->call_cnt = call_cnt;
 
-	if (!local->inode)
-		local->inode = inode_ref (loc->inode);
+        if (!local->inode)
+                local->inode = inode_ref (loc->inode);
 
-	for (i = 0; i < call_cnt; i++) {
-		STACK_WIND (frame, dht_lookup_everywhere_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->lookup,
-			    loc, local->xattr_req);
-	}
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_lookup_everywhere_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->lookup,
+                            loc, local->xattr_req);
+        }
 
-	return 0;
+        return 0;
+out:
+        DHT_STACK_UNWIND (lookup, frame, -1, EINVAL, NULL, NULL, NULL, NULL);
+err:
+        return -1;
 }
 
 
@@ -564,60 +1091,82 @@ dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
                          struct iatt *postparent)
 {
         call_frame_t *prev          = NULL;
-	dht_local_t  *local         = NULL;
-	xlator_t     *subvol        = NULL;
-	loc_t        *loc           = NULL;
-	dht_conf_t   *conf          = NULL;
+        dht_local_t  *local         = NULL;
+        xlator_t     *subvol        = NULL;
+        loc_t        *loc           = NULL;
+        dht_conf_t   *conf          = NULL;
         int           ret           = 0;
 
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, unwind);
+
         prev   = cookie;
-	subvol = prev->this;
-	conf   = this->private;
-	local  = frame->local;
-	loc    = &local->loc;
+        subvol = prev->this;
+        conf   = this->private;
+        local  = frame->local;
+        loc    = &local->loc;
 
         if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"lookup of %s on %s (following linkfile) failed (%s)",
-			local->loc.path, subvol->name, strerror (op_errno));
-                goto err;
-	}
+                gf_log (this->name, GF_LOG_INFO,
+                        "lookup of %s on %s (following linkfile) failed (%s)",
+                        local->loc.path, subvol->name, strerror (op_errno));
+
+                /* If cached subvol returned ENOTCONN, do not do
+                lookup_everywhere. We need to make sure linkfile does not get
+                removed, which can take away the namespace, and subvol is
+                anyways down. */
+
+                if (op_errno != ENOTCONN)
+                        goto err;
+                else
+                        goto unwind;
+        }
 
         if (check_is_dir (inode, stbuf, xattr)) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_INFO,
                         "lookup of %s on %s (following linkfile) reached dir",
                         local->loc.path, subvol->name);
                 goto err;
         }
 
-        if (check_is_linkfile (inode, stbuf, xattr)) {
-                gf_log (this->name, GF_LOG_DEBUG,
+        if (check_is_linkfile (inode, stbuf, xattr, conf->link_xattr_name)) {
+                gf_log (this->name, GF_LOG_INFO,
                         "lookup of %s on %s (following linkfile) reached link",
                         local->loc.path, subvol->name);
                 goto err;
         }
 
-	if ((stbuf->ia_nlink == 1)
-	    && (conf->unhashed_sticky_bit)) {
-		stbuf->ia_prot.sticky = 1;
-	}
-        dht_itransform (this, prev->this, stbuf->ia_ino, &stbuf->ia_ino);
-        if (local->loc.parent)
-                postparent->ia_ino = local->loc.parent->ino;
-        
-	ret = dht_layout_preset (this, prev->this, inode);
-	if (ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+        if (uuid_compare (local->gfid, stbuf->ia_gfid)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: gfid different on data file on %s",
+                        local->loc.path, subvol->name);
+                goto err;
+        }
+
+        if ((stbuf->ia_nlink == 1)
+            && (conf && conf->unhashed_sticky_bit)) {
+                stbuf->ia_prot.sticky = 1;
+        }
+
+        ret = dht_layout_preset (this, prev->this, inode);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_INFO,
                         "failed to set layout for subvolume %s",
                         prev->this->name);
-		op_ret   = -1;
-		op_errno = EINVAL;
-		goto out;
-	}
+                op_ret   = -1;
+                op_errno = EINVAL;
+        }
 
-out:
-        WIPE (postparent);
+        if (local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
+        }
 
+unwind:
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
         DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
                           postparent);
 
@@ -625,7 +1174,7 @@ out:
 
 err:
         dht_lookup_everywhere (frame, this, loc);
-
+out:
         return 0;
 }
 
@@ -637,21 +1186,38 @@ dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc)
         int           i = 0;
         dht_conf_t   *conf = NULL;
         dht_local_t  *local = NULL;
+        int           ret = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", loc, unwind);
 
         conf = this->private;
         local = frame->local;
 
         call_cnt        = conf->subvolume_cnt;
         local->call_cnt = call_cnt;
-		
+
         local->layout = dht_layout_new (this, conf->subvolume_cnt);
         if (!local->layout) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                DHT_STACK_UNWIND (lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
-                return 0;
+                goto unwind;
+        }
+
+        if (local->xattr != NULL) {
+                dict_unref (local->xattr);
+                local->xattr = NULL;
+        }
+
+        if (!uuid_is_null (local->gfid)) {
+                ret = dict_set_static_bin (local->xattr_req, "gfid-req",
+                                           local->gfid, 16);
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set gfid", local->loc.path);
         }
-		
+
         for (i = 0; i < call_cnt; i++) {
                 STACK_WIND (frame, dht_lookup_dir_cbk,
                             conf->subvolumes[i],
@@ -659,6 +1225,11 @@ dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc)
                             &local->loc, local->xattr_req);
         }
         return 0;
+unwind:
+        DHT_STACK_UNWIND (lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
+out:
+        return 0;
+
 }
 
 
@@ -676,9 +1247,14 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         loc_t        *loc           = NULL;
         call_frame_t *prev          = NULL;
         int           ret           = 0;
-        uint64_t      tmp_layout    = 0;
         dht_layout_t *parent_layout = NULL;
 
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+
         conf  = this->private;
 
         prev  = cookie;
@@ -688,99 +1264,137 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         /* This is required for handling stale linkfile deletion,
          * or any more call which happens from this 'loc'.
          */
-        if (uuid_is_null (loc->inode->gfid) && !op_ret)
-                memcpy (loc->inode->gfid, stbuf->ia_gfid, 16);
+        if (!op_ret && uuid_is_null (local->gfid))
+                memcpy (local->gfid, stbuf->ia_gfid, 16);
 
-	if (ENTRY_MISSING (op_ret, op_errno)) {
+        if (ENTRY_MISSING (op_ret, op_errno)) {
+                gf_log (this->name, GF_LOG_TRACE, "Entry %s missing on subvol"
+                        " %s", loc->path, prev->this->name);
                 if (conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_ON) {
-			local->op_errno = ENOENT;
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
-		}
+                        local->op_errno = ENOENT;
+                        dht_lookup_everywhere (frame, this, loc);
+                        return 0;
+                }
                 if ((conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_AUTO) &&
                     (loc->parent)) {
-                        ret = inode_ctx_get (loc->parent, this, &tmp_layout);
-                        parent_layout = (dht_layout_t *)(long)tmp_layout;
+                        ret = dht_inode_ctx_layout_get (loc->parent, this,
+                                                        &parent_layout);
+                        if (ret || !parent_layout)
+                                goto out;
                         if (parent_layout->search_unhashed) {
                                 local->op_errno = ENOENT;
                                 dht_lookup_everywhere (frame, this, loc);
                                 return 0;
                         }
                 }
-	}
+        }
 
- 	if (op_ret == 0) {
- 		is_dir      = check_is_dir (inode, stbuf, xattr);
- 		if (is_dir) {
- 			local->inode = inode_ref (inode);
- 			local->xattr = dict_ref (xattr);
- 		}
- 	}
+        if (op_ret == 0) {
+                is_dir      = check_is_dir (inode, stbuf, xattr);
+                if (is_dir) {
+                        local->inode = inode_ref (inode);
+                        local->xattr = dict_ref (xattr);
+                }
+        }
 
- 	if (is_dir || (op_ret == -1 && op_errno == ENOTCONN)) {
+        if (is_dir || (op_ret == -1 && op_errno == ENOTCONN)) {
                 dht_lookup_directory (frame, this, &local->loc);
                 return 0;
- 	}
- 
-        if (op_ret == -1)
+        }
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "Lookup of %s for subvolume"
+                       " %s failed with error %s", loc->path, prev->this->name,
+                       strerror (op_errno));
                 goto out;
+        }
 
-        is_linkfile = check_is_linkfile (inode, stbuf, xattr);
-        is_dir      = check_is_dir (inode, stbuf, xattr);
+        is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                         conf->link_xattr_name);
 
-        if (!is_dir && !is_linkfile) {
+        if (!is_linkfile) {
                 /* non-directory and not a linkfile */
 
-		dht_itransform (this, prev->this, stbuf->ia_ino,
-				&stbuf->ia_ino);
-                if (loc->parent)
-                        postparent->ia_ino = loc->parent->ino;
-
-		ret = dht_layout_preset (this, prev->this, inode);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not set pre-set layout for subvolume %s",
-				prev->this->name);
-			op_ret   = -1;
-			op_errno = EINVAL;
-			goto out;
-		}
-                goto out; 
-	}
-
-        if (is_linkfile) {
-                subvol = dht_linkfile_subvol (this, inode, stbuf, xattr);
-
-                if (!subvol) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "linkfile not having link subvolume. path=%s",
-                                loc->path);
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
+                ret = dht_layout_preset (this, prev->this, inode);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "could not set pre-set layout for subvolume %s",
+                                prev->this->name);
+                        op_ret   = -1;
+                        op_errno = EINVAL;
+                        goto out;
                 }
+                goto out;
+        }
 
-		STACK_WIND (frame, dht_lookup_linkfile_cbk,
-			    subvol, subvol->fops->lookup,
-			    &local->loc, local->xattr_req);
+        subvol = dht_linkfile_subvol (this, inode, stbuf, xattr);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "linkfile not having link subvolume. path=%s",
+                        loc->path);
+                dht_lookup_everywhere (frame, this, loc);
+                return 0;
         }
 
+        STACK_WIND (frame, dht_lookup_linkfile_cbk,
+                    subvol, subvol->fops->lookup,
+                    &local->loc, local->xattr_req);
+
         return 0;
 
 out:
-        /* 
-         * FIXME: postparent->ia_size and postparent->st_blocks do not have 
-         * correct values. since, postparent corresponds to a directory these 
+        /*
+         * FIXME: postparent->ia_size and postparent->st_blocks do not have
+         * correct values. since, postparent corresponds to a directory these
          * two members should have values equal to sum of corresponding values
          * from each of the subvolume. See dht_iatt_merge for reference.
          */
 
-        WIPE (postparent);
+        if (!op_ret && local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
+        }
 
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
         DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
                           postparent);
+err:
         return 0;
 }
 
+/* For directories, check if acl xattrs have been requested (by the acl xlator),
+ * if not, request for them. These xattrs are needed for dht dir self-heal to
+ * perform proper self-healing of dirs
+ */
+void
+dht_check_and_set_acl_xattr_req (inode_t *inode, dict_t *xattr_req)
+{
+        int     ret = 0;
+
+        GF_ASSERT (inode);
+        GF_ASSERT (xattr_req);
+
+        if (inode->ia_type != IA_IFDIR)
+                return;
+
+        if (!dict_get (xattr_req, POSIX_ACL_ACCESS_XATTR)) {
+                ret = dict_set_int8 (xattr_req, POSIX_ACL_ACCESS_XATTR, 0);
+                if (ret)
+                        gf_log (THIS->name, GF_LOG_WARNING,
+                                "failed to set key %s",
+                                POSIX_ACL_ACCESS_XATTR);
+        }
+
+        if (!dict_get (xattr_req, POSIX_ACL_DEFAULT_XATTR)) {
+                ret = dict_set_int8 (xattr_req, POSIX_ACL_DEFAULT_XATTR, 0);
+                if (ret)
+                        gf_log (THIS->name, GF_LOG_WARNING,
+                                "failed to set key %s",
+                                POSIX_ACL_DEFAULT_XATTR);
+        }
+
+        return;
+}
 
 int
 dht_lookup (call_frame_t *frame, xlator_t *this,
@@ -788,34 +1402,40 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
 {
         xlator_t     *subvol = NULL;
         xlator_t     *hashed_subvol = NULL;
-        xlator_t     *cached_subvol = NULL;
         dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_conf_t   *conf = NULL;
         int           ret    = -1;
         int           op_errno = -1;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-	int           call_cnt = 0;
-
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           call_cnt = 0;
+        loc_t         new_loc = {0,};
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
 
-	conf = this->private;
+        conf = this->private;
+        if (!conf)
+                goto err;
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-        if (!dht_filter_loc_subvol_key (this, loc, &local->loc,
-                                        &hashed_subvol)) {
-                ret = loc_dup (loc, &local->loc);
+        local = dht_local_init (frame, loc, NULL, GF_FOP_LOOKUP);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        ret = dht_filter_loc_subvol_key (this, loc, &new_loc,
+                                         &hashed_subvol);
+        if (ret) {
+                loc_wipe (&local->loc);
+                ret = loc_dup (&new_loc, &local->loc);
+
+                /* we no more need 'new_loc' entries */
+                loc_wipe (&new_loc);
+
+                /* check if loc_dup() is successful */
                 if (ret == -1) {
                         op_errno = errno;
                         gf_log (this->name, GF_LOG_DEBUG,
@@ -825,22 +1445,25 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
                 }
         }
 
-	if (xattr_req) {
-		local->xattr_req = dict_ref (xattr_req);
-	} else {
-		local->xattr_req = dict_new ();
-	}
+        if (xattr_req) {
+                local->xattr_req = dict_ref (xattr_req);
+        } else {
+                local->xattr_req = dict_new ();
+        }
+
+        if (uuid_is_null (loc->pargfid) && !uuid_is_null (loc->gfid) &&
+            !__is_root_gfid (loc->inode->gfid)) {
+                local->cached_subvol = NULL;
+                dht_discover (frame, this, loc);
+                return 0;
+        }
 
         if (!hashed_subvol)
                 hashed_subvol = dht_subvol_get_hashed (this, loc);
-	cached_subvol = dht_subvol_get_cached (this, loc->inode);
-
-	local->cached_subvol = cached_subvol;
-	local->hashed_subvol = hashed_subvol;
+        local->hashed_subvol = hashed_subvol;
 
         if (is_revalidate (loc)) {
-		local->layout = layout = dht_layout_get (this, loc->inode);
-
+                layout = local->layout;
                 if (!layout) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "revalidate without cache. path=%s",
@@ -849,71 +1472,93 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		if (layout->gen && (layout->gen < conf->gen)) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"incomplete layout failure for path=%s",
-				loc->path);
+                if (layout->gen && (layout->gen < conf->gen)) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "incomplete layout failure for path=%s",
+                                loc->path);
 
                         dht_layout_unref (this, local->layout);
                         local->layout = NULL;
-			goto do_fresh_lookup;
-		}
+                        local->cached_subvol = NULL;
+                        goto do_fresh_lookup;
+                }
+
+                local->inode = inode_ref (loc->inode);
+
+                /* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
+                 *       revalidates directly go to the cached-subvolume.
+                 */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
+
+                if (IA_ISDIR (local->inode->ia_type)) {
+                        local->call_cnt = call_cnt = conf->subvolume_cnt;
+                        for (i = 0; i < call_cnt; i++) {
+                                STACK_WIND (frame, dht_revalidate_cbk,
+                                            conf->subvolumes[i],
+                                            conf->subvolumes[i]->fops->lookup,
+                                            loc, local->xattr_req);
+                        }
+                        return 0;
+                }
+
+                call_cnt = local->call_cnt = layout->cnt;
+
+                /* need it for self-healing linkfiles which is
+                   'in-migration' state */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       GLUSTERFS_OPEN_FD_COUNT, 4);
 
-		local->inode    = inode_ref (loc->inode);
-		local->ia_ino   = loc->inode->ino;
-		
-		local->call_cnt = layout->cnt;
-		call_cnt = local->call_cnt;
-		
-		/* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
-		 *       revalidates directly go to the cached-subvolume.
-		 */
-		ret = dict_set_uint32 (local->xattr_req, 
-				       "trusted.glusterfs.dht", 4 * 4);
-
-		for (i = 0; i < layout->cnt; i++) {
+                /* need it for dir self-heal */
+                dht_check_and_set_acl_xattr_req (loc->inode, local->xattr_req);
+
+		for (i = 0; i < call_cnt; i++) {
 			subvol = layout->list[i].xlator;
-			
+
 			STACK_WIND (frame, dht_revalidate_cbk,
 				    subvol, subvol->fops->lookup,
 				    &local->loc, local->xattr_req);
 
-			if (!--call_cnt)
-				break;
 		}
         } else {
         do_fresh_lookup:
-		/* TODO: remove the hard-coding */
-		ret = dict_set_uint32 (local->xattr_req, 
-				       "trusted.glusterfs.dht", 4 * 4);
+                /* TODO: remove the hard-coding */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
+
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->link_xattr_name, 256);
+
+                /* need it for self-healing linkfiles which is
+                   'in-migration' state */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       GLUSTERFS_OPEN_FD_COUNT, 4);
 
-		ret = dict_set_uint32 (local->xattr_req, 
-				       "trusted.glusterfs.dht.linkto", 256);
+                /* need it for dir self-heal */
+                dht_check_and_set_acl_xattr_req (loc->inode, local->xattr_req);
 
                 if (!hashed_subvol) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"no subvolume in layout for path=%s, "
-				"checking on all the subvols to see if "
-				"it is a directory", loc->path);
- 			call_cnt        = conf->subvolume_cnt;
- 			local->call_cnt = call_cnt;
- 			
- 			local->layout = dht_layout_new (this,
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "no subvolume in layout for path=%s, "
+                                "checking on all the subvols to see if "
+                                "it is a directory", loc->path);
+                        call_cnt        = conf->subvolume_cnt;
+                        local->call_cnt = call_cnt;
+
+                        local->layout = dht_layout_new (this,
                                                         conf->subvolume_cnt);
- 			if (!local->layout) {
- 				op_errno = ENOMEM;
- 				gf_log (this->name, GF_LOG_ERROR,
- 					"Out of memory");
- 				goto err;
- 			}
-
-			for (i = 0; i < call_cnt; i++) {
- 				STACK_WIND (frame, dht_lookup_dir_cbk,
- 					    conf->subvolumes[i],
- 					    conf->subvolumes[i]->fops->lookup,
- 					    &local->loc, local->xattr_req);
- 			}
- 			return 0;
+                        if (!local->layout) {
+                                op_errno = ENOMEM;
+                                goto err;
+                        }
+
+                        for (i = 0; i < call_cnt; i++) {
+                                STACK_WIND (frame, dht_lookup_dir_cbk,
+                                            conf->subvolumes[i],
+                                            conf->subvolumes[i]->fops->lookup,
+                                            &local->loc, local->xattr_req);
+                        }
+                        return 0;
                 }
 
                 STACK_WIND (frame, dht_lookup_cbk,
@@ -924,335 +1569,52 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
         return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL);
         return 0;
 }
 
 
 int
-dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int op_ret, int op_errno, struct iatt *prebuf,
-              struct iatt *postbuf)
-{
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
-
-
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
-
-                dht_iatt_merge (this, &local->prebuf, prebuf, prev->this);
-		dht_iatt_merge (this, &local->stbuf, postbuf, prev->this);
-
-		if (local->inode) {
-			local->stbuf.ia_ino = local->inode->ino;
-                        local->prebuf.ia_ino = local->inode->ino;
-                }
-
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
-
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (truncate, frame, local->op_ret, local->op_errno,
-				  &local->prebuf, &local->stbuf);
-
-        return 0;
-}
-
-
-
-int
-dht_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int op_ret, int op_errno, struct iatt *stbuf)
-{
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
-
-
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
-
-		dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-		
-		if (local->inode)
-			local->stbuf.ia_ino = local->inode->ino;
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
-
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (stat, frame, local->op_ret, local->op_errno,
-				  &local->stbuf);
-
-        return 0;
-}
-
-
-int
-dht_stat (call_frame_t *frame, xlator_t *this,
-	  loc_t *loc)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->layout = layout = dht_layout_get (this, loc->inode);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no layout for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = layout->cnt;
-
-	for (i = 0; i < layout->cnt; i++) {
-		subvol = layout->list[i].xlator;
-
-		STACK_WIND (frame, dht_attr_cbk,
-			    subvol, subvol->fops->stat,
-			    loc);
-	}
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL);
-
-	return 0;
-}
-
-
-int
-dht_fstat (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->layout = layout = dht_layout_get (this, fd->inode);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no layout for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local->inode    = inode_ref (fd->inode);
-	local->call_cnt = layout->cnt;;
-
-	for (i = 0; i < layout->cnt; i++) {
-		subvol = layout->list[i].xlator;
-		STACK_WIND (frame, dht_attr_cbk,
-			    subvol, subvol->fops->fstat,
-			    fd);
-	}
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fstat, frame, -1, op_errno, NULL);
-
-	return 0;
-}
-
-
-int
-dht_truncate (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, off_t offset)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
-
-	STACK_WIND (frame, dht_truncate_cbk,
-		    subvol, subvol->fops->truncate,
-		    loc, offset);
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL);
-
-	return 0;
-}
-
-
-int
-dht_ftruncate (call_frame_t *frame, xlator_t *this,
-	       fd_t *fd, off_t offset)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
-
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->inode = inode_ref (fd->inode);
-	local->call_cnt = 1;
-
-	STACK_WIND (frame, dht_truncate_cbk,
-		    subvol, subvol->fops->ftruncate,
-		    fd, offset);
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL);
-
-	return 0;
-}
-
-
-int
 dht_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 int op_ret, int op_errno, struct iatt *preparent,
-                struct iatt *postparent)
+                struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev  = NULL;
+        dht_local_t  *local = NULL;
+        call_frame_t *prev  = NULL;
 
-	local = frame->local;
-	prev  = cookie;
+        local = frame->local;
+        prev  = cookie;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
                         local->op_ret   = -1;
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
 
-                preparent->ia_ino = local->loc.parent->ino;
-                postparent->ia_ino = local->loc.parent->ino;
-		local->op_ret = 0;
+                local->op_ret = 0;
 
                 local->postparent = *postparent;
                 local->preparent = *preparent;
 
-                WIPE (&local->postparent);
-                WIPE (&local->preparent);
-	}
+                if (local->loc.parent) {
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->preparent, 0);
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
+                }
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
         DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
-                          &local->preparent, &local->postparent);
+                          &local->preparent, &local->postparent, NULL);
 
         return 0;
 }
@@ -1261,32 +1623,33 @@ unlock:
 int
 dht_unlink_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int op_ret, int op_errno, struct iatt *preparent,
-                         struct iatt *postparent)
+                         struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
 
         xlator_t *cached_subvol = NULL;
 
-	local = frame->local;
-	prev  = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
+        local = frame->local;
+        prev  = cookie;
 
-		local->op_ret = 0;
-	}
+        LOCK (&frame->lock);
+        {
+                if ((op_ret == -1) && !((op_errno == ENOENT) ||
+                                        (op_errno == ENOTCONN))) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
+
+                local->op_ret = 0;
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
-        if (op_ret == -1)
+        if (local->op_ret == -1)
                 goto err;
 
         cached_subvol = dht_subvol_get_cached (this, local->loc.inode);
@@ -1300,321 +1663,311 @@ unlock:
 
         STACK_WIND (frame, dht_unlink_cbk,
                     cached_subvol, cached_subvol->fops->unlink,
-                    &local->loc);
+                    &local->loc, local->flags, NULL);
 
         return 0;
 
 err:
         DHT_STACK_UNWIND (unlink, frame, -1, local->op_errno,
-                          NULL, NULL);
+                          NULL, NULL, NULL);
         return 0;
 }
 
-
 int
-dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
-               int op_errno, struct iatt *prebuf, struct iatt *postbuf)
+dht_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+             int op_ret, int op_errno, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
 
+        local = frame->local;
+        prev = cookie;
 
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
 
-		local->op_ret = 0;
-	}
+                local->op_ret = 0;
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
-        if (local && (op_ret == 0)) {
-                prebuf->ia_ino = local->ia_ino;
-                postbuf->ia_ino = local->ia_ino;
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                DHT_STACK_UNWIND (setxattr, frame, local->op_ret,
+                                  local->op_errno, NULL);
         }
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (fsync, frame, local->op_ret, local->op_errno,
-                                  prebuf, postbuf);
-
         return 0;
 }
 
-
-
-int
-dht_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	     int op_ret, int op_errno)
+static void
+fill_layout_info (dht_layout_t *layout, char *buf)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
-
-
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
-
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
+        int i = 0;
+        char tmp_buf[128] = {0,};
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-		DHT_STACK_UNWIND (setxattr, frame, local->op_ret, local->op_errno);
+        for (i = 0; i < layout->cnt; i++) {
+                snprintf (tmp_buf, 128, "(%s %u %u)",
+                          layout->list[i].xlator->name,
+                          layout->list[i].start,
+                          layout->list[i].stop);
+                if (i)
+                        strcat (buf, " ");
+                strcat (buf, tmp_buf);
         }
-
-        return 0;
 }
 
+void
+dht_fill_pathinfo_xattr (xlator_t *this, dht_local_t *local,
+                         char *xattr_buf, int32_t alloc_len,
+                         int flag, char *layout_buf)
+{
+        if (flag && local->xattr_val)
+                snprintf (xattr_buf, alloc_len,
+                          "((<"DHT_PATHINFO_HEADER"%s> %s) (%s-layout %s))",
+                          this->name, local->xattr_val, this->name,
+                          layout_buf);
+        else if (local->xattr_val)
+                snprintf (xattr_buf, alloc_len,
+                          "(<"DHT_PATHINFO_HEADER"%s> %s)",
+                          this->name, local->xattr_val);
+        else if (flag)
+                snprintf (xattr_buf, alloc_len, "(%s-layout %s)",
+                          this->name, layout_buf);
+}
 
 int
-dht_access (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t mask)
+dht_vgetxattr_alloc_and_fill (dht_local_t *local, dict_t *xattr, xlator_t *this,
+                              int op_errno)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        int      ret       = -1;
+        char    *value     = NULL;
+        int32_t  plen      = 0;
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        ret = dict_get_str (xattr, local->xsel, &value);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Subvolume %s returned -1 (%s)", this->name,
+                        strerror (op_errno));
+                local->op_ret = -1;
+                local->op_errno = op_errno;
+                goto out;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local->alloc_len += strlen(value);
 
-	local->call_cnt = 1;
+        if (!local->xattr_val) {
+                local->alloc_len += (strlen (DHT_PATHINFO_HEADER) + 10);
+                local->xattr_val = GF_CALLOC (local->alloc_len, sizeof (char),
+                                              gf_common_mt_char);
+                if (!local->xattr_val) {
+                        ret = -1;
+                        goto out;
+                }
+        }
 
-	STACK_WIND (frame, dht_err_cbk,
-		    subvol, subvol->fops->access,
-		    loc, mask);
+        if (local->xattr_val) {
+                plen = strlen (local->xattr_val);
+                if (plen) {
+                        /* extra byte(s) for \0 to be safe */
+                        local->alloc_len += (plen + 2);
+                        local->xattr_val = GF_REALLOC (local->xattr_val,
+                                                       local->alloc_len);
+                        if (!local->xattr_val) {
+                                ret = -1;
+                                goto out;
+                        }
+                }
 
-	return 0;
+                (void) strcat (local->xattr_val, value);
+                (void) strcat (local->xattr_val, " ");
+                local->op_ret = 0;
+        }
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (access, frame, -1, op_errno);
+        ret = 0;
 
-	return 0;
+ out:
+        return ret;
 }
 
-
 int
-dht_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int op_ret, int op_errno, const char *path, struct iatt *sbuf)
+dht_vgetxattr_fill_and_set (dht_local_t *local, dict_t **dict, xlator_t *this,
+                            gf_boolean_t flag)
 {
-        dht_local_t *local = NULL;
+        int   ret             = -1;
+        char *xattr_buf       = NULL;
+        char layout_buf[8192] = {0,};
 
-        local = frame->local;
-        if (op_ret == -1)
-                goto err;
+        if (flag)
+                fill_layout_info (local->layout, layout_buf);
 
-        if (local) {
-                sbuf->ia_ino = local->ia_ino;
+        *dict = dict_new ();
+        if (!*dict)
+                goto out;
+
+        local->xattr_val[strlen (local->xattr_val) - 1] = '\0';
+
+        /* we would need max this many bytes to create xattr string
+         * extra 40 bytes is just an estimated amount of additional
+         * space required as we include translator name and some
+         * spaces, brackets etc. when forming the pathinfo string.
+         *
+         * For node-uuid we just don't have all the pretty formatting,
+         * but since this is a generic routine for pathinfo & node-uuid
+         * we dont have conditional space allocation and try to be
+         * generic
+         */
+        local->alloc_len += (2 * strlen (this->name))
+                + strlen (layout_buf)
+                + 40;
+        xattr_buf = GF_CALLOC (local->alloc_len, sizeof (char),
+                               gf_common_mt_char);
+        if (!xattr_buf)
+                goto out;
+
+        if (XATTR_IS_PATHINFO (local->xsel)) {
+                (void) dht_fill_pathinfo_xattr (this, local, xattr_buf,
+                                                local->alloc_len, flag,
+                                                layout_buf);
+        } else if (XATTR_IS_NODE_UUID (local->xsel)) {
+                (void) snprintf (xattr_buf, local->alloc_len, "%s",
+                                 local->xattr_val);
         } else {
-                op_ret = -1;
-                op_errno = EINVAL;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Unknown local->xsel (%s)", local->xsel);
+                goto out;
         }
 
-err:
-        DHT_STACK_UNWIND (readlink, frame, op_ret, op_errno, path, sbuf);
+        ret = dict_set_dynstr (*dict, local->xsel, xattr_buf);
+        GF_FREE (local->xattr_val);
 
-        return 0;
+ out:
+        return ret;
 }
 
-
 int
-dht_readlink (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, size_t size)
+dht_vgetxattr_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-        dht_local_t  *local = NULL;
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        int          ret           = 0;
+        dht_local_t *local         = NULL;
+        int          this_call_cnt = 0;
+        dict_t      *dict          = NULL;
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (frame->local, out);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = frame->local;
 
-        local->ia_ino = loc->inode->ino;
- 
-	STACK_WIND (frame, dht_readlink_cbk,
-		    subvol, subvol->fops->readlink,
-		    loc, size);
+        LOCK (&frame->lock);
+        {
+                this_call_cnt = --local->call_cnt;
+                if (op_ret < 0) {
+                        if (op_errno != ENOTCONN) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "getxattr err (%s) for dir",
+                                        strerror (op_errno));
+				local->op_ret = -1;
+				local->op_errno = op_errno;
+                        }
 
-	return 0;
+                        goto unlock;
+                }
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL);
+                ret = dht_vgetxattr_alloc_and_fill (local, xattr, this,
+                                                    op_errno);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "alloc or fill failure");
+        }
+ unlock:
+        UNLOCK (&frame->lock);
 
-	return 0;
-}
+        if (!is_last_call (this_call_cnt))
+                goto out;
 
+        /* -- last call: do patch ups -- */
 
-int
-dht_fix_layout_cbk (call_frame_t *frame, void *cookie,
-                    xlator_t *this, int32_t op_ret, int32_t op_errno)
-{
-        DHT_STACK_UNWIND (getxattr, frame, -1, ENODATA, NULL);
+        if (local->op_ret == -1) {
+                goto unwind;
+        }
 
-        return 0;
-}
+        ret = dht_vgetxattr_fill_and_set (local, &dict, this, _gf_true);
+        if (ret)
+                goto unwind;
 
-static void
-fill_layout_info (dht_layout_t *layout, char *buf)
-{
-        int i = 0;
-        char tmp_buf[128] = {0,};
+        DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+        goto cleanup;
 
-        for (i = 0; i < layout->cnt; i++) {
-                snprintf (tmp_buf, 128, "(%s %u %u)",
-                          layout->list[i].xlator->name,
-                          layout->list[i].start,
-                          layout->list[i].stop);
-                if (i)
-                        strcat (buf, " ");
-                strcat (buf, tmp_buf);
-        }
+ unwind:
+        DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno, NULL, NULL);
+ cleanup:
+        if (dict)
+                dict_unref (dict);
+ out:
+        return 0;
 }
 
 int
-dht_pathinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                           int op_ret, int op_errno, dict_t *xattr)
+dht_vgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-        int ret = 0;
-        int flag = 0;
-        int this_call_cnt = 0;
-        char *value_got = NULL;
-        char  layout_buf[8192] = {0,};
-        char  xattr_buf[8192 + 1024] = {0,};
-        dict_t *dict = NULL;
+        dht_local_t  *local         = NULL;
+        int           ret           = 0;
+        dict_t       *dict          = NULL;
+        call_frame_t *prev          = NULL;
+        gf_boolean_t  flag          = _gf_true;
 
         local = frame->local;
+        prev = cookie;
 
-        if (op_ret != -1) {
-                ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &value_got);
-                if (!ret) {
-                        if (!local->pathinfo)
-                                local->pathinfo = GF_CALLOC (8192, sizeof (char),
-                                                             gf_common_mt_char);
-                        if (local->pathinfo)
-                                strcat (local->pathinfo, value_got);
-                }
+        if (op_ret < 0) {
+                local->op_ret = -1;
+                local->op_errno = op_errno;
+                gf_log (this->name, GF_LOG_ERROR, "Subvolume %s returned -1 "
+                        "(%s)", prev->this->name, strerror (op_errno));
+                goto unwind;
         }
 
-        this_call_cnt = dht_frame_return (frame);
-        if (is_last_call (this_call_cnt)) {
-                if (local->layout->cnt > 1) {
-                        /* Set it for directory */
-                        fill_layout_info (local->layout, layout_buf);
-                        flag = 1;
-                }
-
-                dict = dict_new ();
-
-                if (flag && local->pathinfo)
-                        snprintf (xattr_buf, 9216, "((%s %s) (%s-layout %s))",
-                                  this->name, local->pathinfo, this->name,
-                                  layout_buf);
-                else if (local->pathinfo)
-                        snprintf (xattr_buf, 9216, "(%s %s)",
-                                  this->name, local->pathinfo);
-                else if (flag)
-                        snprintf (xattr_buf, 9216, "(%s-layout %s)",
-                                  this->name, layout_buf);
-
-                ret = dict_set_str (dict, GF_XATTR_PATHINFO_KEY,
-                                    xattr_buf);
-
-                if (local->pathinfo)
-                        GF_FREE (local->pathinfo);
-                GF_FREE (local->key);
-
-                DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+        ret = dht_vgetxattr_alloc_and_fill (local, xattr, this,
+                                            op_errno);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "alloc or fill failure");
+                goto unwind;
+        }
 
-                if (dict)
-                        dict_unref (dict);
+        flag = (local->layout->cnt > 1) ? _gf_true : _gf_false;
 
-                return 0;
-        }
+        ret = dht_vgetxattr_fill_and_set (local, &dict, this, flag);
+        if (ret)
+                goto unwind;
 
-        if (local->pathinfo)
-                strcat (local->pathinfo, " Link: ");
-        if (local->hashed_subvol) {
-                /* This will happen if there pending */
-                STACK_WIND (frame, dht_pathinfo_getxattr_cbk, local->hashed_subvol,
-                            local->hashed_subvol->fops->getxattr,
-                            &local->loc, local->key);
+        DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+        goto cleanup;
 
-                return 0;
-        }
+ unwind:
+        DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno,
+                          NULL, NULL);
+ cleanup:
+        if (dict)
+                dict_unref (dict);
 
-        gf_log ("this->name", GF_LOG_ERROR, "Unable to find hashed_subvol for path"
-                " %s", local->pathinfo);
-      
-        DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, dict);
         return 0;
 }
 
 int
 dht_linkinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                           int op_ret, int op_errno, dict_t *xattr)
+                           int op_ret, int op_errno, dict_t *xattr,
+                           dict_t *xdata)
 {
         int   ret   = 0;
         char *value = NULL;
@@ -1629,92 +1982,234 @@ dht_linkinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 }
         }
 
-        DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr);
+        DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
 
         return 0;
 }
 
 int
 dht_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int op_ret, int op_errno, dict_t *xattr)
+                  int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
 {
-        if (op_ret != -1) {
-                if (dict_get (xattr, "trusted.glusterfs.dht")) {
-                        dict_del (xattr, "trusted.glusterfs.dht");
-                }
+        int             this_call_cnt = 0;
+        dht_local_t     *local = NULL;
+        dht_conf_t      *conf = NULL;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (frame->local, out);
+        VALIDATE_OR_GOTO (this->private, out);
+
+        conf = this->private;
+        local = frame->local;
+
+        this_call_cnt = dht_frame_return (frame);
+
+        if (!xattr || (op_ret == -1))
+                goto out;
+
+        if (dict_get (xattr, conf->xattr_name)) {
+                dict_del (xattr, conf->xattr_name);
         }
+        local->op_ret = 0;
 
-        DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr);
+        if (!local->xattr) {
+                local->xattr = dict_copy_with_ref (xattr, NULL);
+        } else {
+                /* first aggregate everything into xattr and then copy into
+                 * local->xattr. This is required as we want to have
+                 * 'local->xattr' as the proper final dictionary passed above
+                 * distribute xlator.
+                 */
+                dht_aggregate_xattr (xattr, local->xattr);
+                local->xattr = dict_copy (xattr, local->xattr);
+        }
+out:
+        if (is_last_call (this_call_cnt)) {
+                DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
+                                  local->xattr, NULL);
+        }
+        return 0;
+}
 
+int32_t
+dht_getxattr_unwind (call_frame_t *frame,
+                     int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+{
+        DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
         return 0;
 }
 
 
 int
+dht_getxattr_get_real_filename_cbk (call_frame_t *frame, void *cookie,
+				    xlator_t *this, int op_ret, int op_errno,
+				    dict_t *xattr, dict_t *xdata)
+{
+        int             this_call_cnt = 0;
+        dht_local_t     *local = NULL;
+
+
+        local = frame->local;
+
+	if (op_ret != -1) {
+		if (local->xattr)
+			dict_unref (local->xattr);
+		local->xattr = dict_ref (xattr);
+
+		if (local->xattr_req)
+			dict_unref (local->xattr_req);
+		local->xattr_req = dict_ref (xdata);
+	}
+
+	this_call_cnt = dht_frame_return (frame);
+	if (is_last_call (this_call_cnt)) {
+		DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
+				  local->xattr, local->xattr_req);
+	}
+
+	return 0;
+}
+
+
+int
+dht_getxattr_get_real_filename (call_frame_t *frame, xlator_t *this,
+				loc_t *loc, const char *key, dict_t *xdata)
+{
+	dht_local_t     *local = NULL;
+	int              i = 0;
+	dht_layout_t    *layout = NULL;
+	int              cnt = 0;
+	xlator_t        *subvol = NULL;
+
+
+	local = frame->local;
+	layout = local->layout;
+
+	cnt = local->call_cnt = layout->cnt;
+
+	local->op_ret = -1;
+	local->op_errno = ENODATA;
+
+	for (i = 0; i < cnt; i++) {
+		subvol = layout->list[i].xlator;
+		STACK_WIND (frame, dht_getxattr_get_real_filename_cbk,
+			    subvol, subvol->fops->getxattr,
+			    loc, key, xdata);
+	}
+
+	return 0;
+}
+
+
+int
 dht_getxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, const char *key)
+              loc_t *loc, const char *key, dict_t *xdata)
+#define DHT_IS_DIR(layout)  (layout->cnt > 1)
 {
-	xlator_t     *subvol        = NULL;
-	xlator_t     *hashed_subvol = NULL;
-	xlator_t     *cached_subvol = NULL;
-	dht_conf_t   *conf          = NULL;
-	dht_local_t  *local         = NULL;
+
+        xlator_t     *subvol        = NULL;
+        xlator_t     *hashed_subvol = NULL;
+        xlator_t     *cached_subvol = NULL;
+        dht_conf_t   *conf          = NULL;
+        dht_local_t  *local         = NULL;
         dht_layout_t *layout        = NULL;
+        xlator_t     **sub_volumes  = NULL;
         int           op_errno      = -1;
-        int           ret           = 0;
-        int           flag          = 0;
         int           i             = 0;
+        int           cnt           = 0;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
         conf   = this->private;
-        layout = dht_layout_get (this, loc->inode);
-        if (key && (strcmp (key, GF_XATTR_PATHINFO_KEY) == 0)) {
-                hashed_subvol = dht_subvol_get_hashed (this, loc);
-                cached_subvol = dht_subvol_get_cached (this, loc->inode);
 
-                local = dht_local_init (frame);
-                if (!local) {
-                        op_errno = ENOMEM;
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        goto err;
-                }
+        local = dht_local_init (frame, loc, NULL, GF_FOP_GETXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
 
-                ret = loc_dup (loc, &local->loc);
-                if (ret == -1) {
-                        op_errno = ENOMEM;
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        goto err;
-                }
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "layout is NULL");
+                op_errno = ENOENT;
+                goto err;
+        }
+
+        if (key) {
                 local->key = gf_strdup (key);
                 if (!local->key) {
                         op_errno = ENOMEM;
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
                         goto err;
                 }
-                local->layout = layout;
+        }
 
-                local->call_cnt = 1;
-                if (hashed_subvol != cached_subvol) {
-                        local->call_cnt = 2;
-                        local->hashed_subvol = hashed_subvol;
+	if (key &&
+	    (strncmp (key, GF_XATTR_GET_REAL_FILENAME_KEY,
+		      strlen (GF_XATTR_GET_REAL_FILENAME_KEY)) == 0)
+	    && DHT_IS_DIR(layout)) {
+		dht_getxattr_get_real_filename (frame, this, loc, key, xdata);
+		return 0;
+	}
+
+        /* for file use cached subvolume (obviously!): see if {}
+         * below
+         * for directory:
+         *  wind to all subvolumes and exclude subvolumes which
+         *  return ENOTCONN (in callback)
+         *
+         * NOTE: Don't trust inode here, as that may not be valid
+         *       (until inode_link() happens)
+         */
+        if (key && DHT_IS_DIR(layout) &&
+            ((strcmp (key, GF_XATTR_PATHINFO_KEY) == 0)
+             || (strcmp (key, GF_XATTR_NODE_UUID_KEY) == 0))) {
+                (void) strncpy (local->xsel, key, 256);
+                cnt = local->call_cnt = layout->cnt;
+                for (i = 0; i < cnt; i++) {
+                        subvol = layout->list[i].xlator;
+                        STACK_WIND (frame, dht_vgetxattr_dir_cbk,
+                                    subvol, subvol->fops->getxattr,
+                                    loc, key, NULL);
                 }
+                return 0;
+        }
 
-                STACK_WIND (frame, dht_pathinfo_getxattr_cbk, cached_subvol,
-                            cached_subvol->fops->getxattr, loc, key);
+        /* node-uuid or pathinfo for files */
+        if (key && ((strcmp (key, GF_XATTR_NODE_UUID_KEY) == 0)
+                    || (strcmp (key, GF_XATTR_PATHINFO_KEY) == 0))) {
+                cached_subvol = local->cached_subvol;
+                (void) strncpy (local->xsel, key, 256);
+
+                local->call_cnt = 1;
+                STACK_WIND (frame, dht_vgetxattr_cbk, cached_subvol,
+                            cached_subvol->fops->getxattr, loc, key, NULL);
 
                 return 0;
         }
+
         if (key && (strcmp (key, GF_XATTR_LINKINFO_KEY) == 0)) {
                 hashed_subvol = dht_subvol_get_hashed (this, loc);
+                if (!hashed_subvol) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get"
+                                "hashed subvol for %s", loc->path);
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
                 cached_subvol = dht_subvol_get_cached (this, loc->inode);
+                if (!cached_subvol) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get"
+                                "cached subvol for %s", loc->path);
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
                 if (hashed_subvol == cached_subvol) {
                         op_errno = ENODATA;
                         goto err;
@@ -1722,703 +2217,845 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
                 if (hashed_subvol) {
                         STACK_WIND (frame, dht_linkinfo_getxattr_cbk, hashed_subvol,
                                     hashed_subvol->fops->getxattr, loc,
-                                    GF_XATTR_PATHINFO_KEY);
+                                    GF_XATTR_PATHINFO_KEY, NULL);
                         return 0;
                 }
                 op_errno = ENODATA;
                 goto err;
         }
-        if (key && (strcmp (key, GF_XATTR_FIX_LAYOUT_KEY) == 0)) {
-                for (i = 0; i < layout->cnt; i++) {
-                        if (layout->list[i].start == layout->list[i].stop) {
-                                flag = 1;
-                                break;
-                        }
+
+        if (key && (!strcmp (GF_XATTR_MARKER_KEY, key))
+            && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+                if (DHT_IS_DIR(layout)) {
+                        cnt = layout->cnt;
+                } else {
+                        cnt = 1;
                 }
-                if ((layout->cnt < conf->subvolume_cnt) || flag) {
-                        gf_log (this->name, GF_LOG_INFO,
-                                "expanding layout of %s from %d to %d",
-                                loc->path, layout->cnt, conf->subvolume_cnt);
-                        local = dht_local_init (frame);
-                        if (!local) {
-                                op_errno = ENOMEM;
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory");
-                                goto err;
-                        }
 
-                        ret = loc_dup (loc, &local->loc);
-                        if (ret == -1) {
-                                op_errno = ENOMEM;
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory");
+                sub_volumes = alloca ( cnt * sizeof (xlator_t *));
+                for (i = 0; i < cnt; i++)
+                        *(sub_volumes + i) = layout->list[i].xlator;
+
+                if (cluster_getmarkerattr (frame, this, loc, key,
+                                           local, dht_getxattr_unwind,
+                                           sub_volumes, cnt,
+                                           MARKER_UUID_TYPE, marker_uuid_default_gauge,
+                                           conf->vol_uuid)) {
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
+                return 0;
+        }
+
+        if (key && *conf->vol_uuid) {
+                if ((match_uuid_local (key, conf->vol_uuid) == 0) &&
+                    (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+                        if (DHT_IS_DIR(layout)) {
+                                cnt = layout->cnt;
+                        } else {
+                                cnt = 1;
+                        }
+                        sub_volumes = alloca ( cnt * sizeof (xlator_t *));
+                        for (i = 0; i < cnt; i++)
+                                sub_volumes[i] = layout->list[i].xlator;
+
+                        if (cluster_getmarkerattr (frame, this, loc, key,
+                                                   local, dht_getxattr_unwind,
+                                                   sub_volumes, cnt,
+                                                   MARKER_XTIME_TYPE,
+                                                   marker_xtime_default_gauge,
+                                                   conf->vol_uuid)) {
+                                op_errno = EINVAL;
                                 goto err;
                         }
-                        local->layout = layout;
 
-                        dht_selfheal_new_directory (frame, dht_fix_layout_cbk,
-                                                    layout);
                         return 0;
                 }
-                op_errno = ENODATA;
-                goto err;
         }
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
 
-	STACK_WIND (frame, dht_getxattr_cbk,
-		    subvol, subvol->fops->getxattr,
-		    loc, key);
+        if (DHT_IS_DIR(layout)) {
+                cnt = local->call_cnt = layout->cnt;
+        } else {
+                cnt = local->call_cnt  = 1;
+        }
 
-	return 0;
+        for (i = 0; i < cnt; i++) {
+                subvol = layout->list[i].xlator;
+                STACK_WIND (frame, dht_getxattr_cbk,
+                            subvol, subvol->fops->getxattr,
+                            loc, key, NULL);
+        }
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
-
+#undef DHT_IS_DIR
 
 int
-dht_setxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, dict_t *xattr, int flags)
+dht_fgetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, const char *key, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
+        xlator_t     *subvol        = NULL;
+        dht_local_t  *local         = NULL;
+        dht_layout_t *layout        = NULL;
+        int           op_errno      = -1;
+        int           i             = 0;
+        int           cnt           = 0;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FGETXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+                goto err;
+        }
 
-	local->call_cnt = 1;
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "layout is NULL");
+                op_errno = ENOENT;
+                goto err;
+        }
+
+        if (key) {
+                local->key = gf_strdup (key);
+                if (!local->key) {
+                        op_errno = ENOMEM;
+                        goto err;
+                }
+        }
 
-	STACK_WIND (frame, dht_err_cbk,
-		    subvol, subvol->fops->setxattr,
-		    loc, xattr, flags);
+        if ((fd->inode->ia_type == IA_IFDIR)
+            && (strncmp (key, GF_XATTR_LOCKINFO_KEY,
+                         strlen (GF_XATTR_LOCKINFO_KEY) != 0))) {
+                cnt = local->call_cnt = layout->cnt;
+        } else {
+                cnt = local->call_cnt  = 1;
+        }
 
-	return 0;
+        for (i = 0; i < cnt; i++) {
+                subvol = layout->list[i].xlator;
+                STACK_WIND (frame, dht_getxattr_cbk,
+                            subvol, subvol->fops->fgetxattr,
+                            fd, key, NULL);
+        }
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (setxattr, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fgetxattr, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
-
 int
-dht_removexattr (call_frame_t *frame, xlator_t *this,
-		 loc_t *loc, const char *key)
+dht_fsetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, dict_t *xattr, int flags, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
+        xlator_t     *subvol   = NULL;
+        dht_local_t  *local    = NULL;
+        int           op_errno = EINVAL;
+        dht_conf_t   *conf     = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
+                                   op_errno, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FSETXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_err_cbk,
-		    subvol, subvol->fops->removexattr,
-		    loc, key);
+        local->call_cnt = 1;
 
-	return 0;
+        STACK_WIND (frame, dht_err_cbk, subvol, subvol->fops->fsetxattr,
+                    fd, xattr, flags, NULL);
+
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (removexattr, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
+static int
+dht_common_setxattr_cbk (call_frame_t *frame, void *cookie,
+                         xlator_t *this, int32_t op_ret, int32_t op_errno,
+                         dict_t *xdata)
+{
+        DHT_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
 int
-dht_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int op_ret, int op_errno, fd_t *fd)
+dht_checking_pathinfo_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int op_ret, int op_errno, dict_t *xattr,
+                           dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
+        int           i     = -1;
+        int           ret   = -1;
+        char         *value = NULL;
+        dht_local_t  *local = NULL;
+        dht_conf_t   *conf  = NULL;
+        call_frame_t *prev  = NULL;
+        int           this_call_cnt = 0;
 
+        local = frame->local;
+        prev = cookie;
+        conf = this->private;
 
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
+        if (op_ret == -1)
+                goto out;
 
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
-				  local->fd);
+        ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &value);
+        if (ret)
+                goto out;
 
+        if (!strcmp (value, local->key)) {
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (conf->subvolumes[i] == prev->this)
+                                conf->decommissioned_bricks[i] = prev->this;
+                }
+        }
+
+out:
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                DHT_STACK_UNWIND (setxattr, frame, local->op_ret, ENOTSUP, NULL);
+        }
         return 0;
-}
 
+}
 
 int
-dht_open (call_frame_t *frame, xlator_t *this,
-	  loc_t *loc, int flags, fd_t *fd, int wbflags)
+dht_setxattr (call_frame_t *frame, xlator_t *this,
+              loc_t *loc, dict_t *xattr, int flags, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-	int           ret = -1;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
+        xlator_t     *subvol   = NULL;
+        dht_local_t  *local    = NULL;
+        dht_conf_t   *conf     = NULL;
+        dht_layout_t *layout   = NULL;
+        int           i        = 0;
+        int           op_errno = EINVAL;
+        int           ret      = -1;
+        data_t       *tmp      = NULL;
+        uint32_t      dir_spread = 0;
+        char          value[4096] = {0,};
+        gf_dht_migrate_data_type_t forced_rebalance = GF_DHT_MIGRATE_DATA;
+        int           call_cnt = 0;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        conf   = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
+                                   op_errno, err);
 
-	local->fd = fd_ref (fd);
-	ret = loc_dup (loc, &local->loc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_SETXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_fd_cbk,
-		    subvol, subvol->fops->open,
-		    loc, flags, fd, wbflags);
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	return 0;
+        local->call_cnt = call_cnt = layout->cnt;
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (open, frame, -1, op_errno, NULL);
+        tmp = dict_get (xattr, "distribute.migrate-data");
+        if (tmp) {
+                if (IA_ISDIR (loc->inode->ia_type)) {
+                        op_errno = ENOTSUP;
+                        goto err;
+                }
 
-	return 0;
-}
+                /* TODO: need to interpret the 'value' for more meaning
+                   (ie, 'target' subvolume given there, etc) */
+                memcpy (value, tmp->data, tmp->len);
+                if (strcmp (value, "force") == 0)
+                       forced_rebalance =
+                                GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS;
 
+                if (conf->decommission_in_progress)
+                        forced_rebalance = GF_DHT_MIGRATE_HARDLINK;
 
-int
-dht_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	       int op_ret, int op_errno,
-	       struct iovec *vector, int count, struct iatt *stbuf,
-               struct iobref *iobref)
-{
-        dht_local_t     *local = frame->local;
+                local->rebalance.target_node = dht_subvol_get_hashed (this, loc);
+                if (!local->rebalance.target_node) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                "hashed subvol for %s", loc->path);
+                        op_errno = EINVAL;
+                        goto err;
+                }
 
-        if (!local) {
-                op_ret = -1;
+                local->rebalance.from_subvol = local->cached_subvol;
+
+                if (local->rebalance.target_node == local->rebalance.from_subvol) {
+                        op_errno = EEXIST;
+                        goto err;
+                }
+                if (local->rebalance.target_node) {
+                        local->flags = forced_rebalance;
+
+                        ret = dht_start_rebalance_task (this, frame);
+                        if (!ret)
+                                return 0;
+
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: failed to create a new synctask",
+                                loc->path);
+                }
                 op_errno = EINVAL;
-                goto out;
-        }
+                goto err;
 
-        if (op_ret != -1)
-                stbuf->ia_ino = local->ia_ino;
-out:
-        DHT_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count, stbuf,
-                          iobref);
+        }
 
-        return 0;
-}
+        tmp = dict_get (xattr, "decommission-brick");
+        if (tmp) {
+                /* This operation should happen only on '/' */
+                if (!__is_root_gfid (loc->inode->gfid)) {
+                        op_errno = ENOTSUP;
+                        goto err;
+                }
 
+                memcpy (value, tmp->data, ((tmp->len < 4095) ? tmp->len : 4095));
+                local->key = gf_strdup (value);
+                local->call_cnt = conf->subvolume_cnt;
 
-int
-dht_readv (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd, size_t size, off_t off)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-        dht_local_t  *local = NULL;
+                for (i = 0 ; i < conf->subvolume_cnt; i++) {
+                        /* Get the pathinfo, and then compare */
+                        STACK_WIND (frame, dht_checking_pathinfo_cbk,
+                                    conf->subvolumes[i],
+                                    conf->subvolumes[i]->fops->getxattr,
+                                    loc, GF_XATTR_PATHINFO_KEY, NULL);
+                }
+                return 0;
+        }
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        tmp = dict_get (xattr, GF_XATTR_FIX_LAYOUT_KEY);
+        if (tmp) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "fixing the layout of %s", loc->path);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+                ret = dht_fix_directory_layout (frame, dht_common_setxattr_cbk,
+                                                layout);
+                if (ret) {
+                        op_errno = ENOTCONN;
+                        goto err;
+                }
+                return ret;
+        }
 
-        local = dht_local_init (frame);
-        if (!local) {
-                gf_log (this->name, GF_LOG_ERROR, "Out of memory");
-                op_errno = ENOMEM;
+        tmp = dict_get (xattr, "distribute.directory-spread-count");
+        if (tmp) {
+                /* Setxattr value is packed as 'binary', not string */
+                memcpy (value, tmp->data, ((tmp->len < 4095)?tmp->len:4095));
+                ret = gf_string2uint32 (value, &dir_spread);
+                if (!ret && ((dir_spread <= conf->subvolume_cnt) &&
+                             (dir_spread > 0))) {
+                        layout->spread_cnt = dir_spread;
+
+                        ret = dht_fix_directory_layout (frame,
+                                                        dht_common_setxattr_cbk,
+                                                        layout);
+                        if (ret) {
+                                op_errno = ENOTCONN;
+                                goto err;
+                        }
+                        return ret;
+                }
+                gf_log (this->name, GF_LOG_ERROR,
+                        "wrong 'directory-spread-count' value (%s)", value);
+                op_errno = ENOTSUP;
                 goto err;
         }
 
-        local->ia_ino = fd->inode->ino;
-	STACK_WIND (frame, dht_readv_cbk,
-		    subvol, subvol->fops->readv,
-		    fd, size, off);
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_err_cbk,
+                            layout->list[i].xlator,
+                            layout->list[i].xlator->fops->setxattr,
+                            loc, xattr, flags, xdata);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int op_ret, int op_errno, struct iatt *prebuf,
-                struct iatt *postbuf)
+dht_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int op_ret, int op_errno, dict_t *xdata)
 {
-        dht_local_t *local = NULL;
-
-        if (op_ret == -1) {
-                goto out;
-        }
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
 
         local = frame->local;
-        if (!local) {
-                op_ret = -1;
-                op_errno = EINVAL;
-                goto out;
-        } 
-        
-        prebuf->ia_ino = local->ia_ino;
-        postbuf->ia_ino = local->ia_ino;
+        prev = cookie;
 
-out:
-        DHT_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf);
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
+
+                local->op_ret = 0;
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                DHT_STACK_UNWIND (removexattr, frame, local->op_ret,
+                                  local->op_errno, NULL);
+        }
 
         return 0;
 }
 
 
 int
-dht_writev (call_frame_t *frame, xlator_t *this,
-	    fd_t *fd, struct iovec *vector, int count, off_t off,
-            struct iobref *iobref)
+dht_removexattr (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, const char *key, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        xlator_t     *subvol = NULL;
         int           op_errno = -1;
         dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+        int           call_cnt = 0;
+        dht_conf_t   *conf = NULL;
+
+        int i;
 
-        VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        conf = this->private;
+
+        GF_IF_NATIVE_XATTR_GOTO (conf->wild_xattr_name, key, op_errno, err);
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
 
-        local = dht_local_init (frame);
+        local = dht_local_init (frame, loc, NULL, GF_FOP_REMOVEXATTR);
         if (!local) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
                 op_errno = ENOMEM;
                 goto err;
         }
 
-        local->ia_ino = fd->inode->ino;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!local->layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->call_cnt = call_cnt = layout->cnt;
+        local->key = gf_strdup (key);
 
-	STACK_WIND (frame, dht_writev_cbk,
-		    subvol, subvol->fops->writev,
-		    fd, vector, count, off, iobref);
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_removexattr_cbk,
+                            layout->list[i].xlator,
+                            layout->list[i].xlator->fops->removexattr,
+                            loc, key, NULL);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
-
 int
-dht_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
+dht_fremovexattr (call_frame_t *frame, xlator_t *this,
+                  fd_t *fd, const char *key, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        xlator_t     *subvol = NULL;
         int           op_errno = -1;
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+        int           call_cnt = 0;
+        dht_conf_t   *conf = 0;
 
+        int i;
 
-        VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
-
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->fd = fd_ref (fd);
-	local->call_cnt = 1;
-
-	STACK_WIND (frame, dht_err_cbk,
-		    subvol, subvol->fops->flush, fd);
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (flush, frame, -1, op_errno);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	return 0;
-}
-
-
-int
-dht_fsync (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd, int datasync)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
+        conf = this->private;
 
+        GF_IF_NATIVE_XATTR_GOTO (conf->wild_xattr_name, key, op_errno, err);
 
         VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FREMOVEXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-	local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for inode=%s",
+                        uuid_utoa (fd->inode->gfid));
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!local->layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for inode=%s", uuid_utoa (fd->inode->gfid));
+                op_errno = EINVAL;
+                goto err;
+        }
 
-        local->ia_ino = fd->inode->ino;
+        local->call_cnt = call_cnt = layout->cnt;
+        local->key = gf_strdup (key);
 
-	STACK_WIND (frame, dht_fsync_cbk,
-		    subvol, subvol->fops->fsync,
-		    fd, datasync);
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_removexattr_cbk,
+                            layout->list[i].xlator,
+                            layout->list[i].xlator->fops->fremovexattr,
+                            fd, key, NULL);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fremovexattr, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	    int op_ret, int op_errno, struct flock *flock)
+dht_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+            int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
 {
-        DHT_STACK_UNWIND (lk, frame, op_ret, op_errno, flock);
-
-        return 0;
-}
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
 
+        local = frame->local;
+        prev = cookie;
 
-int
-dht_lk (call_frame_t *frame, xlator_t *this,
-	fd_t *fd, int cmd, struct flock *flock)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
 
+                local->op_ret = 0;
+        }
+unlock:
+        UNLOCK (&frame->lock);
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
+                                  local->fd, NULL);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        return 0;
+}
 
-	STACK_WIND (frame, dht_lk_cbk,
-		    subvol, subvol->fops->lk,
-		    fd, cmd, flock);
+/*
+ * dht_normalize_stats -
+ */
+static void
+dht_normalize_stats (struct statvfs *buf, unsigned long bsize,
+                     unsigned long frsize)
+{
+        double factor = 0;
 
-	return 0;
+        if (buf->f_bsize != bsize) {
+                buf->f_bsize = bsize;
+        }
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (lk, frame, -1, op_errno, NULL);
+        if (buf->f_frsize != frsize) {
+                factor = ((double) buf->f_frsize) / frsize;
+                buf->f_frsize = frsize;
+                buf->f_blocks = (fsblkcnt_t) (factor * buf->f_blocks);
+                buf->f_bfree  = (fsblkcnt_t) (factor * buf->f_bfree);
+                buf->f_bavail = (fsblkcnt_t) (factor * buf->f_bavail);
 
-	return 0;
+        }
 }
 
-
 int
 dht_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int op_ret, int op_errno, struct statvfs *statvfs)
+                int op_ret, int op_errno, struct statvfs *statvfs, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
+        dht_local_t *local         = NULL;
+        int          this_call_cnt = 0;
+        int          bsize         = 0;
+        int          frsize        = 0;
 
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			goto unlock;
-		}
-		local->op_ret = 0;
-
-		/* TODO: normalize sizes */
-		local->statvfs.f_bsize    = statvfs->f_bsize;
-		local->statvfs.f_frsize   = statvfs->f_frsize;
-
-		local->statvfs.f_blocks  += statvfs->f_blocks;
-		local->statvfs.f_bfree   += statvfs->f_bfree;
-		local->statvfs.f_bavail  += statvfs->f_bavail;
-		local->statvfs.f_files   += statvfs->f_files;
-		local->statvfs.f_ffree   += statvfs->f_ffree;
-		local->statvfs.f_favail  += statvfs->f_favail;
-		local->statvfs.f_fsid     = statvfs->f_fsid;
-		local->statvfs.f_flag     = statvfs->f_flag;
-		local->statvfs.f_namemax  = statvfs->f_namemax;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        goto unlock;
+                }
+                local->op_ret = 0;
 
-	}
+                if (local->statvfs.f_bsize != 0) {
+                        bsize = max(local->statvfs.f_bsize, statvfs->f_bsize);
+                        frsize = max(local->statvfs.f_frsize, statvfs->f_frsize);
+                        dht_normalize_stats(&local->statvfs, bsize, frsize);
+                        dht_normalize_stats(statvfs, bsize, frsize);
+                } else {
+                        local->statvfs.f_bsize    = statvfs->f_bsize;
+                        local->statvfs.f_frsize   = statvfs->f_frsize;
+                }
+
+                local->statvfs.f_blocks  += statvfs->f_blocks;
+                local->statvfs.f_bfree   += statvfs->f_bfree;
+                local->statvfs.f_bavail  += statvfs->f_bavail;
+                local->statvfs.f_files   += statvfs->f_files;
+                local->statvfs.f_ffree   += statvfs->f_ffree;
+                local->statvfs.f_favail  += statvfs->f_favail;
+                local->statvfs.f_fsid     = statvfs->f_fsid;
+                local->statvfs.f_flag     = statvfs->f_flag;
+                local->statvfs.f_namemax  = statvfs->f_namemax;
+
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
-				  &local->statvfs);
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
+                                  &local->statvfs, xdata);
 
         return 0;
 }
 
 
 int
-dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
+dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        xlator_t     *subvol = NULL;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
-	int           i = -1;
-
+        int           i = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	local->call_cnt = conf->subvolume_cnt;
+        local = dht_local_init (frame, NULL, NULL, GF_FOP_STATFS);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_statfs_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->statfs, loc);
-	}
+        if (IA_ISDIR (loc->inode->ia_type)) {
+                local->call_cnt = conf->subvolume_cnt;
 
-	return 0;
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        STACK_WIND (frame, dht_statfs_cbk,
+                                    conf->subvolumes[i],
+                                    conf->subvolumes[i]->fops->statfs, loc,
+                                    xdata);
+                }
+                return 0;
+        }
+
+        subvol = dht_subvol_get_cached (this, loc->inode);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->call_cnt = 1;
+
+        STACK_WIND (frame, dht_statfs_cbk,
+                    subvol, subvol->fops->statfs, loc, xdata);
+
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (statfs, frame, -1, op_errno, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
+dht_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+             dict_t *xdata)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
-	int           ret = -1;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
-	int           i = -1;
-
+        int           i = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, loc, fd, GF_FOP_OPENDIR);
+        if (!local) {
+                op_errno = ENOMEM;
 
-	local->fd = fd_ref (fd);
-	ret = loc_dup (loc, &local->loc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+                goto err;
+        }
 
-	local->call_cnt = conf->subvolume_cnt;
+        local->call_cnt = conf->subvolume_cnt;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_fd_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->opendir,
-			    loc, fd);
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                STACK_WIND (frame, dht_fd_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->opendir,
+                            loc, fd, xdata);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (opendir, frame, -1, op_errno, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (opendir, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
-                  int op_errno, gf_dirent_t *orig_entries)
+                  int op_errno, gf_dirent_t *orig_entries, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	gf_dirent_t   entries;
-	gf_dirent_t  *orig_entry = NULL;
-	gf_dirent_t  *entry = NULL;
-	call_frame_t *prev = NULL;
-	xlator_t     *next_subvol = NULL;
+        dht_local_t  *local = NULL;
+        gf_dirent_t   entries;
+        gf_dirent_t  *orig_entry = NULL;
+        gf_dirent_t  *entry = NULL;
+        call_frame_t *prev = NULL;
+        xlator_t     *next_subvol = NULL;
         off_t         next_offset = 0;
-	int           count = 0;
+        int           count = 0;
         dht_layout_t *layout = 0;
         dht_conf_t   *conf   = NULL;
         xlator_t     *subvol = 0;
+        int           ret    = 0;
 
-	INIT_LIST_HEAD (&entries.list);
-	prev = cookie;
-	local = frame->local;
-	conf  = this->private;
+        INIT_LIST_HEAD (&entries.list);
+        prev = cookie;
+        local = frame->local;
+        conf  = this->private;
 
-	if (op_ret < 0)
-		goto done;
+        if (op_ret < 0)
+                goto done;
 
         if (!local->layout)
                 local->layout = dht_layout_get (this, local->fd->inode);
 
         layout = local->layout;
 
-	list_for_each_entry (orig_entry, (&orig_entries->list), list) {
+        list_for_each_entry (orig_entry, (&orig_entries->list), list) {
                 next_offset = orig_entry->d_off;
-
-                if (check_is_linkfile (NULL, (&orig_entry->d_stat), NULL)
-                    || (check_is_dir (NULL, (&orig_entry->d_stat), NULL)
-                        && (prev->this != dht_first_up_subvol (this)))) {
+                if (check_is_dir (NULL, (&orig_entry->d_stat), NULL) &&
+                    (prev->this != local->first_up_subvol)) {
+                        continue;
+                }
+                if (check_is_linkfile (NULL, (&orig_entry->d_stat),
+                                       orig_entry->dict,
+                                       conf->link_xattr_name)) {
                         continue;
                 }
 
                 entry = gf_dirent_for_name (orig_entry->d_name);
                 if (!entry) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
+
                         goto unwind;
                 }
 
@@ -2428,25 +3065,41 @@ dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
                                                     orig_entry->d_name);
                         if (!subvol || (subvol != prev->this)) {
                                 /* TODO: Count the number of entries which need
-                                   linkfile to prove its existance in fs */
+                                   linkfile to prove its existence in fs */
                                 layout->search_unhashed++;
                         }
                 }
-                entry->d_stat = orig_entry->d_stat;
 
-                dht_itransform (this, prev->this, orig_entry->d_ino,
-                                &entry->d_ino);
                 dht_itransform (this, prev->this, orig_entry->d_off,
                                 &entry->d_off);
 
-                entry->d_stat.ia_ino = entry->d_ino;
+                entry->d_stat = orig_entry->d_stat;
+                entry->d_ino  = orig_entry->d_ino;
                 entry->d_type = orig_entry->d_type;
                 entry->d_len  = orig_entry->d_len;
 
+                if (orig_entry->dict)
+                        entry->dict = dict_ref (orig_entry->dict);
+
+                /* making sure we set the inode ctx right with layout,
+                   currently possible only for non-directories, so for
+                   directories don't set entry inodes */
+                if (!IA_ISDIR(entry->d_stat.ia_type)) {
+                        ret = dht_layout_preset (this, prev->this,
+                                                 orig_entry->inode);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "failed to link the layout in inode");
+                        entry->inode = inode_ref (orig_entry->inode);
+                } else if (orig_entry->inode) {
+                        dht_inode_ctx_time_update (orig_entry->inode, this,
+                                                   &entry->d_stat, 1);
+                }
+
                 list_add_tail (&entry->list, &entries.list);
                 count++;
-	}
-	op_ret = count;
+        }
+        op_ret = count;
         /* We need to ensure that only the last subvolume's end-of-directory
          * notification is respected so that directory reading does not stop
          * before all subvolumes have been read. That could happen because the
@@ -2458,7 +3111,7 @@ dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
                 op_errno = 0;
 
 done:
-	if (count == 0) {
+        if (count == 0) {
                 /* non-zero next_offset means that
                    EOF is not yet hit on the current subvol
                 */
@@ -2468,23 +3121,37 @@ done:
                         next_subvol = prev->this;
                 }
 
-		if (!next_subvol) {
-			goto unwind;
-		}
+                if (!next_subvol) {
+                        goto unwind;
+                }
 
-		STACK_WIND (frame, dht_readdirp_cbk,
-			    next_subvol, next_subvol->fops->readdirp,
-			    local->fd, local->size, next_offset);
-		return 0;
-	}
+		if (conf->readdir_optimize == _gf_true) {
+                        if (next_subvol != local->first_up_subvol) {
+                                ret = dict_set_int32 (local->xattr,
+                                                      GF_READDIR_SKIP_DIRS, 1);
+                                if (ret)
+                                        gf_log (this->name, GF_LOG_ERROR,
+					         "dict set failed");
+		        } else {
+                                 dict_del (local->xattr,
+                                           GF_READDIR_SKIP_DIRS);
+                        }
+                }
+
+                STACK_WIND (frame, dht_readdirp_cbk,
+                            next_subvol, next_subvol->fops->readdirp,
+                            local->fd, local->size, next_offset,
+                            local->xattr);
+                return 0;
+        }
 
 unwind:
-	if (op_ret < 0)
-		op_ret = 0;
+        if (op_ret < 0)
+                op_ret = 0;
 
-	DHT_STACK_UNWIND (readdirp, frame, op_ret, op_errno, &entries);
+        DHT_STACK_UNWIND (readdirp, frame, op_ret, op_errno, &entries, NULL);
 
-	gf_dirent_free (&entries);
+        gf_dirent_free (&entries);
 
         return 0;
 }
@@ -2493,34 +3160,33 @@ unwind:
 
 int
 dht_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int op_ret, int op_errno, gf_dirent_t *orig_entries)
+                 int op_ret, int op_errno, gf_dirent_t *orig_entries,
+                 dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	gf_dirent_t   entries;
-	gf_dirent_t  *orig_entry = NULL;
-	gf_dirent_t  *entry = NULL;
-	call_frame_t *prev = NULL;
-	xlator_t     *next_subvol = NULL;
+        dht_local_t  *local = NULL;
+        gf_dirent_t   entries;
+        gf_dirent_t  *orig_entry = NULL;
+        gf_dirent_t  *entry = NULL;
+        call_frame_t *prev = NULL;
+        xlator_t     *next_subvol = NULL;
         off_t         next_offset = 0;
-	int           count = 0;
+        int           count = 0;
         dht_layout_t *layout = 0;
-        dht_conf_t   *conf   = NULL;
         xlator_t     *subvol = 0;
 
-	INIT_LIST_HEAD (&entries.list);
-	prev = cookie;
-	local = frame->local;
-	conf  = this->private;
+        INIT_LIST_HEAD (&entries.list);
+        prev = cookie;
+        local = frame->local;
 
-	if (op_ret < 0)
-		goto done;
+        if (op_ret < 0)
+                goto done;
 
         if (!local->layout)
                 local->layout = dht_layout_get (this, local->fd->inode);
 
         layout = local->layout;
 
-	list_for_each_entry (orig_entry, (&orig_entries->list), list) {
+        list_for_each_entry (orig_entry, (&orig_entries->list), list) {
                 next_offset = orig_entry->d_off;
 
                 subvol = dht_layout_search (this, layout, orig_entry->d_name);
@@ -2533,11 +3199,10 @@ dht_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 goto unwind;
                         }
 
-                        dht_itransform (this, prev->this, orig_entry->d_ino,
-                                        &entry->d_ino);
                         dht_itransform (this, prev->this, orig_entry->d_off,
                                         &entry->d_off);
 
+                        entry->d_ino  = orig_entry->d_ino;
                         entry->d_type = orig_entry->d_type;
                         entry->d_len  = orig_entry->d_len;
 
@@ -2545,7 +3210,7 @@ dht_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         count++;
                 }
         }
-	op_ret = count;
+        op_ret = count;
         /* We need to ensure that only the last subvolume's end-of-directory
          * notification is respected so that directory reading does not stop
          * before all subvolumes have been read. That could happen because the
@@ -2557,7 +3222,7 @@ dht_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 op_errno = 0;
 
 done:
-	if (count == 0) {
+        if (count == 0) {
                 /* non-zero next_offset means that
                    EOF is not yet hit on the current subvol
                 */
@@ -2567,23 +3232,23 @@ done:
                         next_subvol = prev->this;
                 }
 
-		if (!next_subvol) {
-			goto unwind;
-		}
+                if (!next_subvol) {
+                        goto unwind;
+                }
 
-		STACK_WIND (frame, dht_readdir_cbk,
-			    next_subvol, next_subvol->fops->readdir,
-			    local->fd, local->size, next_offset);
-		return 0;
-	}
+                STACK_WIND (frame, dht_readdir_cbk,
+                            next_subvol, next_subvol->fops->readdir,
+                            local->fd, local->size, next_offset, NULL);
+                return 0;
+        }
 
 unwind:
-	if (op_ret < 0)
-		op_ret = 0;
+        if (op_ret < 0)
+                op_ret = 0;
 
-	DHT_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries);
+        DHT_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries, NULL);
 
-	gf_dirent_free (&entries);
+        gf_dirent_free (&entries);
 
         return 0;
 }
@@ -2591,61 +3256,92 @@ unwind:
 
 int
 dht_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-                off_t yoff, int whichop)
+                off_t yoff, int whichop, dict_t *dict)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_local_t  *local  = NULL;
         int           op_errno = -1;
-	xlator_t     *xvol = NULL;
-	off_t         xoff = 0;
-
+        xlator_t     *xvol = NULL;
+        off_t         xoff = 0;
+        int           ret = 0;
+        dht_conf_t   *conf = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+        local = dht_local_init (frame, NULL, NULL, whichop);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->fd = fd_ref (fd);
-	local->size = size;
+        local->fd = fd_ref (fd);
+        local->size = size;
+        local->xattr_req = (dict)? dict_ref (dict) : NULL;
+        local->first_up_subvol = dht_first_up_subvol (this);
 
-	dht_deitransform (this, yoff, &xvol, (uint64_t *)&xoff);
+        dht_deitransform (this, yoff, &xvol, (uint64_t *)&xoff);
+
+        /* TODO: do proper readdir */
+        if (whichop == GF_FOP_READDIRP) {
+                if (dict)
+                        local->xattr = dict_ref (dict);
+                else
+                        local->xattr = dict_new ();
+
+                if (local->xattr) {
+                        ret = dict_set_uint32 (local->xattr,
+                                               conf->link_xattr_name, 256);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "failed to set '%s' key",
+                                        conf->link_xattr_name);
+			if (conf->readdir_optimize == _gf_true) {
+                                if (xvol != local->first_up_subvol) {
+				        ret = dict_set_int32 (local->xattr,
+			                               GF_READDIR_SKIP_DIRS, 1);
+				        if (ret)
+					        gf_log (this->name,
+                                                        GF_LOG_ERROR,
+						        "Dict set failed");
+                                } else {
+                                        dict_del (local->xattr,
+                                                  GF_READDIR_SKIP_DIRS);
+                                }
+			}
+                }
 
-	/* TODO: do proper readdir */
-        if (whichop == GF_FOP_READDIR)
-                STACK_WIND (frame, dht_readdir_cbk, xvol, xvol->fops->readdir,
-                            fd, size, xoff);
-        else
                 STACK_WIND (frame, dht_readdirp_cbk, xvol, xvol->fops->readdirp,
-                            fd, size, xoff);
+                            fd, size, xoff, local->xattr);
+        } else {
+                STACK_WIND (frame, dht_readdir_cbk, xvol, xvol->fops->readdir,
+                            fd, size, xoff, local->xattr);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (readdir, frame, -1, op_errno, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-             off_t yoff)
+             off_t yoff, dict_t *xdata)
 {
         int          op = GF_FOP_READDIR;
         dht_conf_t  *conf = NULL;
         int          i = 0;
 
         conf = this->private;
+        if (!conf)
+                goto out;
 
         for (i = 0; i < conf->subvolume_cnt; i++) {
                 if (!conf->subvolume_status[i]) {
@@ -2657,15 +3353,16 @@ dht_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
         if (conf->use_readdirp)
                 op = GF_FOP_READDIRP;
 
-        dht_do_readdir (frame, this, fd, size, yoff, op);
+out:
+        dht_do_readdir (frame, this, fd, size, yoff, op, 0);
         return 0;
 }
 
 int
 dht_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-              off_t yoff)
+              off_t yoff, dict_t *dict)
 {
-        dht_do_readdir (frame, this, fd, size, yoff, GF_FOP_READDIRP);
+        dht_do_readdir (frame, this, fd, size, yoff, GF_FOP_READDIRP, dict);
         return 0;
 }
 
@@ -2673,88 +3370,88 @@ dht_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
 
 int
 dht_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int op_ret, int op_errno)
+                  int op_ret, int op_errno, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
 
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1)
-			local->op_errno = op_errno;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1)
+                        local->op_errno = op_errno;
 
-		if (op_ret == 0)
-			local->op_ret = 0;
-	}
-	UNLOCK (&frame->lock);
+                if (op_ret == 0)
+                        local->op_ret = 0;
+        }
+        UNLOCK (&frame->lock);
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (fsyncdir, frame, local->op_ret, local->op_errno);
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_UNWIND (fsyncdir, frame, local->op_ret,
+                                  local->op_errno, xdata);
 
         return 0;
 }
 
 
 int
-dht_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync)
+dht_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
+              int datasync, dict_t *xdata)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
-	int           i = -1;
-
+        int           i = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, NULL, NULL, GF_FOP_FSYNCDIR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->fd = fd_ref (fd);
-	local->call_cnt = conf->subvolume_cnt;
+        local->fd = fd_ref (fd);
+        local->call_cnt = conf->subvolume_cnt;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_fsyncdir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->fsyncdir,
-			    fd, datasync);
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                STACK_WIND (frame, dht_fsyncdir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->fsyncdir,
+                            fd, datasync, xdata);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fsyncdir, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int op_ret, int op_errno,
+                 int op_ret, int op_errno,
                  inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
-                 struct iatt *postparent)
+                 struct iatt *postparent, dict_t *xdata)
 {
-	call_frame_t *prev = NULL;
-	int           ret = -1;
+        xlator_t     *prev = NULL;
+        int           ret = -1;
         dht_local_t  *local = NULL;
 
 
-	if (op_ret == -1)
-		goto out;
+        if (op_ret == -1)
+                goto out;
 
         local = frame->local;
         if (!local) {
@@ -2763,38 +3460,39 @@ dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto out;
         }
 
-	prev = cookie;
+        prev = cookie;
 
-	dht_itransform (this, prev->this, stbuf->ia_ino, &stbuf->ia_ino);
         if (local->loc.parent) {
-                preparent->ia_ino = local->loc.parent->ino;
-                postparent->ia_ino = local->loc.parent->ino;
 
-                WIPE (preparent);
-                WIPE (postparent);
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           preparent, 0);
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
         }
 
-        ret = dht_layout_preset (this, prev->this, inode);
-	if (ret < 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"could not set pre-set layout for subvolume %s",
-			prev->this->name);
-		op_ret   = -1;
-		op_errno = EINVAL;
-		goto out;
-	}
+        ret = dht_layout_preset (this, prev, inode);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "could not set pre-set layout for subvolume %s",
+                        prev? prev->name: NULL);
+                op_ret   = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
+        if (local->linked == _gf_true)
+                dht_linkfile_attr_heal (frame, this);
 out:
-        /* 
-         * FIXME: ia_size and st_blocks of preparent and postparent do not have 
+        /*
+         * FIXME: ia_size and st_blocks of preparent and postparent do not have
          * correct values. since, preparent and postparent buffers correspond
          * to a directory these two members should have values equal to sum of
          * corresponding values from each of the subvolume.
          * See dht_iatt_merge for reference.
-         */ 
-
-	DHT_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, stbuf, preparent,
-                          postparent);
-	return 0;
+         */
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, stbuf,
+                          preparent, postparent, xdata);
+        return 0;
 }
 
 int
@@ -2802,405 +3500,376 @@ dht_mknod_linkfile_create_cbk (call_frame_t *frame, void *cookie,
                                xlator_t *this,
                                int32_t op_ret, int32_t op_errno,
                                inode_t *inode, struct iatt *stbuf,
-                               struct iatt *preparent, struct iatt *postparent)
+                               struct iatt *preparent, struct iatt *postparent,
+                               dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	xlator_t     *cached_subvol = NULL;
+        dht_local_t  *local = NULL;
+        xlator_t     *cached_subvol = NULL;
 
         if (op_ret == -1)
                 goto err;
 
-	local = frame->local;
-	cached_subvol = local->cached_subvol;
+        local = frame->local;
+        if (!local || !local->cached_subvol) {
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        cached_subvol = local->cached_subvol;
 
-        STACK_WIND (frame, dht_newfile_cbk,
-                    cached_subvol, cached_subvol->fops->mknod,
-                    &local->loc, local->mode, local->rdev,
-                    local->params);
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)cached_subvol,
+                           cached_subvol, cached_subvol->fops->mknod,
+                           &local->loc, local->mode, local->rdev, local->umask,
+                           local->params);
 
         return 0;
- err:
- 	DHT_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL);	
- 	return 0;
+err:
+        DHT_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+                          NULL);
+        return 0;
 }
 
 int
 dht_mknod (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dev_t rdev, dict_t *params)
+           loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *params)
 {
-	xlator_t    *subvol = NULL;
-	int          op_errno = -1;
-        int          ret = -1;
+        xlator_t    *subvol = NULL;
+        int          op_errno = -1;
         xlator_t    *avail_subvol = NULL;
-	dht_conf_t  *conf = NULL;
-	dht_local_t *local = NULL;
-
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        dht_local_t *local = NULL;
 
-	conf = this->private;
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
         dht_get_du_info (frame, this, loc);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
-
-        ret = loc_dup (loc, &local->loc);
-        if (ret == -1) {
+        local = dht_local_init (frame, loc, NULL, GF_FOP_MKNOD);
+        if (!local) {
                 op_errno = ENOMEM;
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
+                goto err;
+        }
+
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
                 goto err;
         }
 
         if (!dht_is_subvol_filled (this, subvol)) {
                 gf_log (this->name, GF_LOG_TRACE,
                         "creating %s on %s", loc->path, subvol->name);
-                
-                STACK_WIND (frame, dht_newfile_cbk,
-                            subvol, subvol->fops->mknod,
-                            loc, mode, rdev, params);
+
+                STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol,
+                                   subvol, subvol->fops->mknod, loc, mode,
+                                   rdev, umask, params);
         } else {
-                avail_subvol = dht_free_disk_available_subvol (this, subvol);
+
+                avail_subvol = dht_free_disk_available_subvol (this, subvol,
+                                                               local);
                 if (avail_subvol != subvol) {
-                        /* Choose the minimum filled volume, and create the 
+                        /* Choose the minimum filled volume, and create the
                            files there */
 
                         local->params = dict_ref (params);
                         local->cached_subvol = avail_subvol;
-                        local->mode = mode; 
+                        local->mode = mode;
                         local->rdev = rdev;
-                        
-                        dht_linkfile_create (frame, 
+                        local->umask = umask;
+                        dht_linkfile_create (frame,
                                              dht_mknod_linkfile_create_cbk,
-                                             avail_subvol, subvol, loc);
+                                             this, avail_subvol, subvol, loc);
                 } else {
                         gf_log (this->name, GF_LOG_TRACE,
                                 "creating %s on %s", loc->path, subvol->name);
-                        
-                        STACK_WIND (frame, dht_newfile_cbk,
-                                    subvol, subvol->fops->mknod,
-                                    loc, mode, rdev, params);
+
+                        STACK_WIND_COOKIE (frame, dht_newfile_cbk,
+                                           (void *)subvol, subvol,
+                                           subvol->fops->mknod, loc, mode,
+                                           rdev, umask, params);
                 }
         }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_symlink (call_frame_t *frame, xlator_t *this,
-	     const char *linkname, loc_t *loc, dict_t *params)
+             const char *linkname, loc_t *loc, mode_t umask, dict_t *params)
 {
-	xlator_t    *subvol = NULL;
-	int          op_errno = -1;
+        xlator_t    *subvol = NULL;
+        int          op_errno = -1;
         dht_local_t *local = NULL;
-        int          ret = -1;
-
-
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
-        ret = loc_copy (&local->loc, loc);
-        if (ret == -1) {
-                gf_log (this->name, GF_LOG_TRACE, "Failed to copy loc");
+        local = dht_local_init (frame, loc, NULL, GF_FOP_SYMLINK);
+        if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"creating %s on %s", loc->path, subvol->name);
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
+                goto err;
+        }
+
+        gf_log (this->name, GF_LOG_TRACE,
+                "creating %s on %s", loc->path, subvol->name);
 
-	STACK_WIND (frame, dht_newfile_cbk,
-		    subvol, subvol->fops->symlink,
-		    linkname, loc, params);
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+                           subvol->fops->symlink, linkname, loc, umask,
+                           params);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (link, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (link, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
+dht_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+            dict_t *xdata)
 {
-	xlator_t    *cached_subvol = NULL;
-	xlator_t    *hashed_subvol = NULL;
-        int          ret = -1;
-	int          op_errno = -1;
-	dht_local_t *local = NULL;
-
+        xlator_t    *cached_subvol = NULL;
+        xlator_t    *hashed_subvol = NULL;
+        int          op_errno = -1;
+        dht_local_t *local = NULL;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
         if (dht_filter_loc_subvol_key (this, loc, &local->loc,
                                        &cached_subvol)) {
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "unlinking %s on %s (given path %s)",
                         local->loc.path, cached_subvol->name, loc->path);
                 STACK_WIND (frame, dht_unlink_cbk,
                             cached_subvol, cached_subvol->fops->unlink,
-                            &local->loc);
+                            &local->loc, xflag, xdata);
                 goto done;
         }
 
-	cached_subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!cached_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_UNLINK);
+        if (!local) {
+                op_errno = ENOMEM;
 
-	hashed_subvol = dht_subvol_get_hashed (this, loc);
-	if (!hashed_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+                goto err;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        hashed_subvol = dht_subvol_get_hashed (this, loc);
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	ret = loc_copy (&local->loc, loc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        cached_subvol = local->cached_subvol;
+        if (!cached_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	if (hashed_subvol != cached_subvol) {
-		STACK_WIND (frame, dht_unlink_linkfile_cbk,
-			    hashed_subvol, hashed_subvol->fops->unlink, loc);
+        local->flags = xflag;
+        if (hashed_subvol != cached_subvol) {
+                STACK_WIND (frame, dht_unlink_linkfile_cbk,
+                            hashed_subvol, hashed_subvol->fops->unlink, loc,
+                            xflag, xdata);
         } else {
                 STACK_WIND (frame, dht_unlink_cbk,
-                            cached_subvol, cached_subvol->fops->unlink, loc);
+                            cached_subvol, cached_subvol->fops->unlink, loc,
+                            xflag, xdata);
         }
 done:
-	return 0;
+        return 0;
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int op_ret, int op_errno,
+              int op_ret, int op_errno,
               inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
-              struct iatt *postparent)
+              struct iatt *postparent, dict_t *xdata)
 {
         call_frame_t *prev = NULL;
-	dht_layout_t *layout = NULL;
-	dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+        dht_local_t  *local = NULL;
 
         prev = cookie;
-	local = frame->local;
+
+        local = frame->local;
 
         if (op_ret == -1)
                 goto out;
 
-	layout = dht_layout_for_subvol (this, prev->this);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no pre-set layout for subvolume %s",
-			prev->this->name);
-		op_ret   = -1;
-		op_errno = EINVAL;
-		goto out;
-	}
-
-	stbuf->ia_ino = local->loc.inode->ino;
-
-        preparent->ia_ino = local->loc2.parent->ino;
-        postparent->ia_ino = local->loc2.parent->ino;
-
-        WIPE (preparent);
-        WIPE (postparent);
+        layout = dht_layout_for_subvol (this, prev->this);
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no pre-set layout for subvolume %s",
+                        prev->this->name);
+                op_ret   = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
 
+        if (local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           preparent, 0);
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
+        }
+        if (local->linked == _gf_true) {
+                local->stbuf = *stbuf;
+                dht_linkfile_attr_heal (frame, this);
+        }
 out:
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
         DHT_STACK_UNWIND (link, frame, op_ret, op_errno, inode, stbuf, preparent,
-                          postparent);
+                          postparent, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_link_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int op_ret, int op_errno,
+                       int op_ret, int op_errno,
                        inode_t *inode, struct iatt *stbuf,
-                       struct iatt *preparent, struct iatt *postparent)
+                       struct iatt *preparent, struct iatt *postparent,
+                       dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	xlator_t     *srcvol = NULL;
-
+        dht_local_t  *local = NULL;
+        xlator_t     *srcvol = NULL;
 
-	if (op_ret == -1)
-		goto err;
+        if (op_ret == -1)
+                goto err;
 
-	local = frame->local;
-	srcvol = local->linkfile.srcvol;
+        local = frame->local;
+        srcvol = local->linkfile.srcvol;
 
-	STACK_WIND (frame, dht_link_cbk,
-		    srcvol, srcvol->fops->link,
-		    &local->loc, &local->loc2);
+        STACK_WIND (frame, dht_link_cbk, srcvol, srcvol->fops->link,
+                    &local->loc, &local->loc2, xdata);
 
-	return 0;
+        return 0;
 
 err:
-	DHT_STACK_UNWIND (link, frame, op_ret, op_errno, inode, stbuf, preparent,
-                          postparent);
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (link, frame, op_ret, op_errno, inode, stbuf, preparent,
+                          postparent, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_link (call_frame_t *frame, xlator_t *this,
-	  loc_t *oldloc, loc_t *newloc)
+          loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
-	xlator_t    *cached_subvol = NULL;
-	xlator_t    *hashed_subvol = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
-	dht_local_t *local = NULL;
-
-
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (oldloc, err);
-	VALIDATE_OR_GOTO (newloc, err);
-
-	cached_subvol = dht_subvol_get_cached (this, oldloc->inode);
-	if (!cached_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", oldloc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        xlator_t    *cached_subvol = NULL;
+        xlator_t    *hashed_subvol = NULL;
+        int          op_errno = -1;
+        int          ret = -1;
+        dht_local_t *local = NULL;
 
-	hashed_subvol = dht_subvol_get_hashed (this, newloc);
-	if (!hashed_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			newloc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (oldloc, err);
+        VALIDATE_OR_GOTO (newloc, err);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, oldloc, NULL, GF_FOP_LINK);
+        if (!local) {
+                op_errno = ENOMEM;
 
-	ret = loc_copy (&local->loc, oldloc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+                goto err;
+        }
 
-	ret = loc_copy (&local->loc2, newloc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        cached_subvol = local->cached_subvol;
+        if (!cached_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", oldloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	if (hashed_subvol != cached_subvol) {
-		dht_linkfile_create (frame, dht_link_linkfile_cbk,
-				     cached_subvol, hashed_subvol, newloc);
-	} else {
-		STACK_WIND (frame, dht_link_cbk,
-			    cached_subvol, cached_subvol->fops->link,
-			    oldloc, newloc);
-	}
+        hashed_subvol = dht_subvol_get_hashed (this, newloc);
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        newloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	return 0;
+        ret = loc_copy (&local->loc2, newloc);
+        if (ret == -1) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        if (hashed_subvol != cached_subvol) {
+                uuid_copy (local->gfid, oldloc->inode->gfid);
+                dht_linkfile_create (frame, dht_link_linkfile_cbk, this,
+                                     cached_subvol, hashed_subvol, newloc);
+        } else {
+                STACK_WIND (frame, dht_link_cbk,
+                            cached_subvol, cached_subvol->fops->link,
+                            oldloc, newloc, xdata);
+        }
+
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int op_ret, int op_errno,
-		 fd_t *fd, inode_t *inode, struct iatt *stbuf,
-                 struct iatt *preparent, struct iatt *postparent)
+                int op_ret, int op_errno,
+                fd_t *fd, inode_t *inode, struct iatt *stbuf,
+                struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-	call_frame_t *prev = NULL;
-	int           ret = -1;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
         dht_local_t  *local = NULL;
 
-	if (op_ret == -1)
-		goto out;
+        if (op_ret == -1)
+                goto out;
 
         local = frame->local;
         if (!local) {
@@ -3209,107 +3878,99 @@ dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto out;
         }
 
-	prev = cookie;
+        prev = cookie;
 
-	dht_itransform (this, prev->this, stbuf->ia_ino, &stbuf->ia_ino);
         if (local->loc.parent) {
-                preparent->ia_ino = local->loc.parent->ino;
-                postparent->ia_ino = local->loc.parent->ino;
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           preparent, 0);
 
-                WIPE (preparent);
-                WIPE (postparent);
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
         }
 
         ret = dht_layout_preset (this, prev->this, inode);
-	if (ret != 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"could not set preset layout for subvol %s",
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "could not set preset layout for subvol %s",
                         prev->this->name);
-		op_ret   = -1;
-		op_errno = EINVAL;
-		goto out;
-	}
-
+                op_ret   = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
+        if (local->linked == _gf_true) {
+                local->stbuf = *stbuf;
+                dht_linkfile_attr_heal (frame, this);
+        }
 out:
-	DHT_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, stbuf, preparent,
-                          postparent);
-	return 0;
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, stbuf, preparent,
+                          postparent, NULL);
+        return 0;
 }
 
 
 int
 dht_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
-				xlator_t *this,
-				int32_t op_ret, int32_t op_errno,
+                                xlator_t *this,
+                                int32_t op_ret, int32_t op_errno,
                                 inode_t *inode, struct iatt *stbuf,
-                                struct iatt *preparent, struct iatt *postparent)
+                                struct iatt *preparent, struct iatt *postparent,
+                                dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	xlator_t     *cached_subvol = NULL;
+        dht_local_t  *local = NULL;
+        xlator_t     *cached_subvol = NULL;
 
         if (op_ret == -1)
                 goto err;
 
-	local = frame->local;
-	cached_subvol = local->cached_subvol;
+        local = frame->local;
+        cached_subvol = local->cached_subvol;
 
         STACK_WIND (frame, dht_create_cbk,
                     cached_subvol, cached_subvol->fops->create,
                     &local->loc, local->flags, local->mode,
-                    local->fd, local->params);
+                    local->umask, local->fd, local->params);
 
         return 0;
- err:
- 	DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- 	return 0;
+err:
+        DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL, NULL, NULL);
+        return 0;
 }
 
 int
 dht_create (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t flags, mode_t mode,
-            fd_t *fd, dict_t *params)
+            loc_t *loc, int32_t flags, mode_t mode,
+            mode_t umask, fd_t *fd, dict_t *params)
 {
-	int          op_errno = -1;
-        int          ret = -1;
-	xlator_t    *subvol = NULL;
-	dht_conf_t  *conf = NULL;
+        int          op_errno = -1;
+        xlator_t    *subvol = NULL;
         dht_local_t *local = NULL;
         xlator_t    *avail_subvol = NULL;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
-
-	conf = this->private;
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
         dht_get_du_info (frame, this, loc);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, fd, GF_FOP_CREATE);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
         if (dht_filter_loc_subvol_key (this, loc, &local->loc,
                                        &subvol)) {
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "creating %s on %s (got create on %s)",
                         local->loc.path, subvol->name, loc->path);
                 STACK_WIND (frame, dht_create_cbk,
                             subvol, subvol->fops->create,
-                            &local->loc, flags, mode, fd, params);
+                            &local->loc, flags, mode, umask, fd, params);
                 goto done;
         }
 
-        ret = loc_dup (loc, &local->loc);
-        if (ret == -1) {
-                op_errno = ENOMEM;
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
         subvol = dht_subvol_get_hashed (this, loc);
         if (!subvol) {
                 gf_log (this->name, GF_LOG_DEBUG,
@@ -3324,154 +3985,158 @@ dht_create (call_frame_t *frame, xlator_t *this,
                         "creating %s on %s", loc->path, subvol->name);
                 STACK_WIND (frame, dht_create_cbk,
                             subvol, subvol->fops->create,
-                            loc, flags, mode, fd, params);
+                            loc, flags, mode, umask, fd, params);
                 goto done;
         }
-        /* Choose the minimum filled volume, and create the 
+        /* Choose the minimum filled volume, and create the
            files there */
-        /* TODO */
-        avail_subvol = dht_free_disk_available_subvol (this, subvol);
+        avail_subvol = dht_free_disk_available_subvol (this, subvol, local);
         if (avail_subvol != subvol) {
-                local->fd = fd_ref (fd);
                 local->params = dict_ref (params);
                 local->flags = flags;
                 local->mode = mode;
-
+                local->umask = umask;
                 local->cached_subvol = avail_subvol;
                 local->hashed_subvol = subvol;
                 gf_log (this->name, GF_LOG_TRACE,
                         "creating %s on %s (link at %s)", loc->path,
                         avail_subvol->name, subvol->name);
-                dht_linkfile_create (frame,
-                                     dht_create_linkfile_create_cbk,
-                                     avail_subvol, subvol, loc);
+                dht_linkfile_create (frame, dht_create_linkfile_create_cbk,
+                                     this, avail_subvol, subvol, loc);
                 goto done;
         }
         gf_log (this->name, GF_LOG_TRACE,
                 "creating %s on %s", loc->path, subvol->name);
         STACK_WIND (frame, dht_create_cbk,
                     subvol, subvol->fops->create,
-                    loc, flags, mode, fd, params);
+                    loc, flags, mode, umask, fd, params);
 done:
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_mkdir_selfheal_cbk (call_frame_t *frame, void *cookie,
-			xlator_t *this,
-			int32_t op_ret, int32_t op_errno)
+                        xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	dht_local_t   *local = NULL;
-	dht_layout_t  *layout = NULL;
-
+        dht_local_t   *local = NULL;
+        dht_layout_t  *layout = NULL;
 
-	local = frame->local;
-	layout = local->selfheal.layout;
+        local = frame->local;
+        layout = local->selfheal.layout;
 
-	if (op_ret == 0) {
+        if (op_ret == 0) {
                 dht_layout_set (this, local->inode, layout);
-		local->stbuf.ia_ino = local->ia_ino;
                 if (local->loc.parent) {
-                        local->preparent.ia_ino = local->loc.parent->ino;
-                        local->postparent.ia_ino = local->loc.parent->ino;
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->preparent, 0);
 
-                        WIPE (&local->preparent);
-                        WIPE (&local->postparent);
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
                 }
-	}
+        }
 
-	DHT_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
-			  local->inode, &local->stbuf, &local->preparent,
-                          &local->postparent);
+        DHT_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
+                          local->inode, &local->stbuf, &local->preparent,
+                          &local->postparent, NULL);
 
-	return 0;
+        return 0;
 }
 
 int
 dht_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                int op_ret, int op_errno, inode_t *inode, struct iatt *stbuf,
-               struct iatt *preparent, struct iatt *postparent)
+               struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	int           ret = -1;
-        int           subvol_filled = 0;
-	call_frame_t *prev = NULL;
-	dht_layout_t *layout = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        int           ret = -1;
+        gf_boolean_t subvol_filled = _gf_false;
+        call_frame_t *prev = NULL;
+        dht_layout_t *layout = NULL;
 
-	conf = this->private;
-	local = frame->local;
-	prev  = cookie;
-	layout = local->layout;
+        local = frame->local;
+        prev  = cookie;
+        layout = local->layout;
 
         subvol_filled = dht_is_subvol_filled (this, prev->this);
 
-	LOCK (&frame->lock);
-	{
+        LOCK (&frame->lock);
+        {
                 if (subvol_filled && (op_ret != -1)) {
                         ret = dht_layout_merge (this, layout, prev->this,
                                                 -1, ENOSPC, NULL);
                 } else {
+			if (op_ret == -1 && op_errno == EEXIST)
+				/* Very likely just a race between mkdir and
+				   self-heal (from lookup of a concurrent mkdir
+				   attempt).
+				   Ignore error for now. layout setting will
+				   anyways fail if this was a different (old)
+				   pre-existing different directory.
+				*/
+				op_ret = 0;
                         ret = dht_layout_merge (this, layout, prev->this,
                                                 op_ret, op_errno, NULL);
                 }
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to merge layouts", local->loc.path);
 
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			goto unlock;
-		}
-		dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        goto unlock;
+                }
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
                 dht_iatt_merge (this, &local->preparent, preparent, prev->this);
                 dht_iatt_merge (this, &local->postparent, postparent,
                                 prev->this);
-
-                if (prev->this == dht_first_up_subvol (this)) {
-                        local->ia_ino = local->stbuf.ia_ino;
-                }
-
-	}
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-		dht_selfheal_new_directory (frame, dht_mkdir_selfheal_cbk,
-					    layout);
-	}
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                dht_selfheal_new_directory (frame, dht_mkdir_selfheal_cbk,
+                                            layout);
+        }
 
         return 0;
 }
 
 int
-dht_mkdir_hashed_cbk (call_frame_t *frame, void *cookie, 
-		      xlator_t *this, int op_ret, int op_errno,
+dht_mkdir_hashed_cbk (call_frame_t *frame, void *cookie,
+                      xlator_t *this, int op_ret, int op_errno,
                       inode_t *inode, struct iatt *stbuf,
-                      struct iatt *preparent, struct iatt *postparent)
+                      struct iatt *preparent, struct iatt *postparent,
+                      dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           ret = -1;
-	call_frame_t *prev = NULL;
-	dht_layout_t *layout = NULL;
-	dht_conf_t   *conf = NULL;
-	int           i = 0;
-	xlator_t     *hashed_subvol = NULL;
+        dht_local_t  *local = NULL;
+        int           ret = -1;
+        call_frame_t *prev = NULL;
+        dht_layout_t *layout = NULL;
+        dht_conf_t   *conf = NULL;
+        int           i = 0;
+        xlator_t     *hashed_subvol = NULL;
 
-	local = frame->local;
-	prev  = cookie;
-	layout = local->layout;
-	conf = this->private;
-	hashed_subvol = local->hashed_subvol;
+        VALIDATE_OR_GOTO (this->private, err);
+
+        local = frame->local;
+        prev  = cookie;
+        layout = local->layout;
+        conf = this->private;
+        hashed_subvol = local->hashed_subvol;
 
-        if (uuid_is_null (local->loc.inode->gfid) && !op_ret)
-                memcpy (local->loc.inode->gfid, stbuf->ia_gfid, 16);
+        if (uuid_is_null (local->loc.gfid) && !op_ret)
+                uuid_copy (local->loc.gfid, stbuf->ia_gfid);
 
         if (dht_is_subvol_filled (this, hashed_subvol))
                 ret = dht_layout_merge (this, layout, prev->this,
@@ -3479,49 +4144,55 @@ dht_mkdir_hashed_cbk (call_frame_t *frame, void *cookie,
         else
                 ret = dht_layout_merge (this, layout, prev->this,
                                         op_ret, op_errno, NULL);
-        
-	if (op_ret == -1) {
-		local->op_errno = op_errno;
-		goto err;
-	}
-	local->op_ret = 0;
 
-	dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+        /* TODO: we may have to return from the function
+           if layout merge fails. For now, lets just log an error */
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to merge layouts", local->loc.path);
+
+        if (op_ret == -1) {
+                local->op_errno = op_errno;
+                goto err;
+        }
+        local->op_ret = 0;
+
+        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
         dht_iatt_merge (this, &local->preparent, preparent, prev->this);
         dht_iatt_merge (this, &local->postparent, postparent, prev->this);
 
-	local->ia_ino = local->stbuf.ia_ino;
+        local->call_cnt = conf->subvolume_cnt - 1;
 
-	local->call_cnt = conf->subvolume_cnt - 1;
-	
-	if (local->call_cnt == 0) {
-		dht_selfheal_directory (frame, dht_mkdir_selfheal_cbk,
-					&local->loc, layout);
-	}
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (conf->subvolumes[i] == hashed_subvol)
-			continue;
-		STACK_WIND (frame, dht_mkdir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->mkdir,
-			    &local->loc, local->mode, local->params);
-	}
-	return 0;
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, stbuf->ia_gfid);
+        if (local->call_cnt == 0) {
+                dht_selfheal_directory (frame, dht_mkdir_selfheal_cbk,
+                                        &local->loc, layout);
+        }
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == hashed_subvol)
+                        continue;
+                STACK_WIND (frame, dht_mkdir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->mkdir, &local->loc,
+                            local->mode, local->umask, local->params);
+        }
+        return 0;
 err:
-	DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL, NULL);
         return 0;
 }
 
 
-int
+ int
 dht_mkdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dict_t *params)
+           loc_t *loc, mode_t mode, mode_t umask, dict_t *params)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
-	int           ret = -1;
-	xlator_t     *hashed_subvol = NULL;
+        xlator_t     *hashed_subvol = NULL;
 
 
         VALIDATE_OR_GOTO (frame, err);
@@ -3529,149 +4200,234 @@ dht_mkdir (call_frame_t *frame, xlator_t *this,
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
         VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
+        conf = this->private;
 
         dht_get_du_info (frame, this, loc);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	hashed_subvol = dht_subvol_get_hashed (this, loc);
+        local = dht_local_init (frame, loc, NULL, GF_FOP_MKDIR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	if (hashed_subvol == NULL) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"hashed subvol not found for %s",
+        hashed_subvol = dht_subvol_get_hashed (this, loc);
+        if (hashed_subvol == NULL) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "hashed subvol not found for %s",
                         loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local->hashed_subvol = hashed_subvol;
-	local->inode = inode_ref (loc->inode);
-	ret = loc_copy (&local->loc, loc);
-	local->mode = mode;
-
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+                op_errno = EINVAL;
+                goto err;
+        }
 
+        local->hashed_subvol = hashed_subvol;
+        local->mode = mode;
+        local->umask = umask;
         local->params = dict_ref (params);
+        local->inode  = inode_ref (loc->inode);
 
-	local->layout = dht_layout_new (this, conf->subvolume_cnt);
-	if (!local->layout) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+        local->layout = dht_layout_new (this, conf->subvolume_cnt);
+        if (!local->layout) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_mkdir_hashed_cbk,
-		    hashed_subvol,
-		    hashed_subvol->fops->mkdir,
-		    loc, mode, params);
+        STACK_WIND (frame, dht_mkdir_hashed_cbk,
+                    hashed_subvol,
+                    hashed_subvol->fops->mkdir,
+                    loc, mode, umask, params);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rmdir_selfheal_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int op_ret, int op_errno)
+                        int op_ret, int op_errno, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
+
+        DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
+                          &local->preparent, &local->postparent, NULL);
+
+        return 0;
+}
+
+
+int
+dht_rmdir_hashed_subvol_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int op_ret, int op_errno, struct iatt *preparent,
+               struct iatt *postparent, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
+
+        local = frame->local;
+        prev  = cookie;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        local->op_ret   = -1;
+                        if (op_errno != ENOENT && op_errno != EACCES) {
+                                local->need_selfheal = 1;
+                        }
+
+
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "rmdir on %s for %s failed (%s)",
+                                prev->this->name, local->loc.path,
+                                strerror (op_errno));
+                        goto unlock;
+                }
+
+                dht_iatt_merge (this, &local->preparent, preparent, prev->this);
+                dht_iatt_merge (this, &local->postparent, postparent,
+                                prev->this);
 
-        if (local->loc.parent) {
-                local->preparent.ia_ino = local->loc.parent->ino;
-                local->postparent.ia_ino = local->loc.parent->ino;
         }
+unlock:
+        UNLOCK (&frame->lock);
 
-	DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
-                          &local->preparent, &local->postparent);
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+               if (local->need_selfheal) {
+                        local->layout =
+                                dht_layout_get (this, local->loc.inode);
 
-	return 0;
+                        /* TODO: neater interface needed below */
+                        local->stbuf.ia_type = local->loc.inode->ia_type;
+
+                        uuid_copy (local->gfid, local->loc.inode->gfid);
+                        dht_selfheal_restore (frame, dht_rmdir_selfheal_cbk,
+                                              &local->loc, local->layout);
+               } else {
+
+                        if (local->loc.parent) {
+                                dht_inode_ctx_time_update (local->loc.parent,
+                                                           this,
+                                                           &local->preparent,
+                                                           0);
+
+                                dht_inode_ctx_time_update (local->loc.parent,
+                                                           this,
+                                                           &local->postparent,
+                                                           1);
+                        }
+
+                        DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
+                                          local->op_errno, &local->preparent,
+                                          &local->postparent, NULL);
+               }
+        }
+
+        return 0;
 }
 
 
 int
 dht_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	       int op_ret, int op_errno, struct iatt *preparent,
-               struct iatt *postparent)
+               int op_ret, int op_errno, struct iatt *preparent,
+               struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
+        int           done = 0;
 
-	local = frame->local;
-	prev  = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			local->op_ret   = -1;
-
-			if (op_errno != ENOENT)
-				local->need_selfheal = 1;
-
-			gf_log (this->name, GF_LOG_DEBUG,
-				"rmdir on %s for %s failed (%s)",
-				prev->this->name, local->loc.path,
-				strerror (op_errno));
-			goto unlock;
-		}
+        local = frame->local;
+        prev  = cookie;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        local->op_ret   = -1;
 
+                        if (op_errno != ENOENT && op_errno != EACCES) {
+                                local->need_selfheal = 1;
+                        }
+
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "rmdir on %s for %s failed (%s)",
+                                prev->this->name, local->loc.path,
+                                strerror (op_errno));
+                        goto unlock;
+                }
+
+                /* Track if rmdir succeeded on atleast one subvol*/
+                local->fop_succeeded = 1;
                 dht_iatt_merge (this, &local->preparent, preparent, prev->this);
                 dht_iatt_merge (this, &local->postparent, postparent,
                                 prev->this);
-	}
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-		if (local->need_selfheal) {
+        this_call_cnt = dht_frame_return (frame);
+
+        /* if local->hashed_subvol, we are yet to wind to hashed_subvol. */
+        if (local->hashed_subvol && (this_call_cnt == 1)) {
+                done = 1;
+        } else if (!local->hashed_subvol && !this_call_cnt) {
+                done = 1;
+        }
+
+
+        if (done) {
+                if (local->need_selfheal && local->fop_succeeded) {
                         local->layout =
                                 dht_layout_get (this, local->loc.inode);
 
-			/* TODO: neater interface needed below */
-			local->stbuf.ia_type = local->loc.inode->ia_type;
+                        /* TODO: neater interface needed below */
+                        local->stbuf.ia_type = local->loc.inode->ia_type;
+
+                        uuid_copy (local->gfid, local->loc.inode->gfid);
+                        dht_selfheal_restore (frame, dht_rmdir_selfheal_cbk,
+                                              &local->loc, local->layout);
+                } else if (this_call_cnt) {
+                        /* If non-hashed subvol's have responded, proceed */
+
+                        local->need_selfheal = 0;
+                        STACK_WIND (frame, dht_rmdir_hashed_subvol_cbk,
+                                    local->hashed_subvol,
+                                    local->hashed_subvol->fops->rmdir,
+                                    &local->loc, local->flags, NULL);
+                } else if (!this_call_cnt) {
+                        /* All subvol's have responded, proceed */
 
-			dht_selfheal_restore (frame, dht_rmdir_selfheal_cbk,
-					      &local->loc, local->layout);
-		} else {
                         if (local->loc.parent) {
-                                local->preparent.ia_ino =
-                                        local->loc.parent->ino;
-                                local->postparent.ia_ino =
-                                        local->loc.parent->ino;
 
-                                WIPE (&local->preparent);
-                                WIPE (&local->postparent);
+                                dht_inode_ctx_time_update (local->loc.parent,
+                                                           this,
+                                                           &local->preparent,
+                                                           0);
+
+                                dht_inode_ctx_time_update (local->loc.parent,
+                                                           this,
+                                                           &local->postparent,
+                                                           1);
+
                         }
 
-			DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
-					  local->op_errno, &local->preparent,
-                                          &local->postparent);
-		}
-	}
+                        DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
+                                          local->op_errno, &local->preparent,
+                                          &local->postparent, NULL);
+                }
+        }
 
         return 0;
 }
@@ -3680,38 +4436,64 @@ unlock:
 int
 dht_rmdir_do (call_frame_t *frame, xlator_t *this)
 {
-	dht_local_t  *local = NULL;
-	dht_conf_t   *conf = NULL;
-	int           i = 0;
+        dht_local_t  *local = NULL;
+        dht_conf_t   *conf = NULL;
+        int           i = 0;
+        xlator_t     *hashed_subvol = NULL;
 
-	conf = this->private;
-	local = frame->local;
+        VALIDATE_OR_GOTO (this->private, err);
 
-	if (local->op_ret == -1)
-		goto err;
+        conf = this->private;
+        local = frame->local;
 
-	local->call_cnt = conf->subvolume_cnt;
+        if (local->op_ret == -1)
+                goto err;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_rmdir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->rmdir,
-			    &local->loc);
-	}
+        local->call_cnt = conf->subvolume_cnt;
 
-	return 0;
+        /* first remove from non-hashed_subvol */
+        hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
+
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_WARNING, "failed to get hashed "
+                        "subvol for %s",local->loc.path);
+        } else {
+                local->hashed_subvol = hashed_subvol;
+        }
+
+        /* When DHT has only 1 child */
+        if (conf->subvolume_cnt == 1) {
+                STACK_WIND (frame, dht_rmdir_hashed_subvol_cbk,
+                            conf->subvolumes[0],
+                            conf->subvolumes[0]->fops->rmdir,
+                            &local->loc, local->flags, NULL);
+                return 0;
+        }
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (hashed_subvol &&
+                    (hashed_subvol == conf->subvolumes[i]))
+                        continue;
+
+                STACK_WIND (frame, dht_rmdir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->rmdir,
+                            &local->loc, local->flags, NULL);
+        }
+
+        return 0;
 
 err:
-	DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
-                          &local->preparent, &local->postparent);
-	return 0;
+        DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
+                          &local->preparent, &local->postparent, NULL);
+        return 0;
 }
 
 
 int
 dht_rmdir_linkfile_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                int op_ret, int op_errno, struct iatt *preparent,
-                               struct iatt *postparent)
+                               struct iatt *postparent, dict_t *xdata)
 {
         dht_local_t    *local = NULL;
         call_frame_t   *prev = NULL;
@@ -3759,6 +4541,7 @@ dht_rmdir_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         call_frame_t   *main_frame = NULL;
         dht_local_t    *main_local = NULL;
         int             this_call_cnt = 0;
+        dht_conf_t     *conf = this->private;
 
         local = frame->local;
         prev  = cookie;
@@ -3770,7 +4553,7 @@ dht_rmdir_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         if (op_ret != 0)
                 goto err;
 
-        if (check_is_linkfile (inode, stbuf, xattr) == 0) {
+        if (!check_is_linkfile (inode, stbuf, xattr, conf->link_xattr_name)) {
                 main_local->op_ret  = -1;
                 main_local->op_errno = ENOTEMPTY;
 
@@ -3781,7 +4564,7 @@ dht_rmdir_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         }
 
         STACK_WIND (frame, dht_rmdir_linkfile_unlink_cbk,
-                    src, src->fops->unlink, &local->loc);
+                    src, src->fops->unlink, &local->loc, 0, NULL);
         return 0;
 err:
 
@@ -3804,6 +4587,8 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
         call_frame_t       *lookup_frame = NULL;
         dht_local_t        *lookup_local = NULL;
         dht_local_t        *local = NULL;
+        dict_t             *xattrs = NULL;
+        dht_conf_t         *conf = this->private;
 
         local = frame->local;
 
@@ -3812,7 +4597,8 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
                         continue;
                 if (strcmp (trav->d_name, "..") == 0)
                         continue;
-                if (check_is_linkfile (NULL, (&trav->d_stat), NULL) == 1) {
+                if (check_is_linkfile (NULL, (&trav->d_stat), trav->dict,
+                                              conf->link_xattr_name)) {
                         ret++;
                         continue;
                 }
@@ -3824,6 +4610,21 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
                 return 0;
         }
 
+        xattrs = dict_new ();
+        if (!xattrs) {
+                gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
+                return -1;
+        }
+
+        ret = dict_set_uint32 (xattrs, conf->link_xattr_name, 256);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to set linkto key"
+                        " in dict");
+                if (xattrs)
+                        dict_unref (xattrs);
+                return -1;
+        }
+
         list_for_each_entry (trav, &entries->list, list) {
                 if (strcmp (trav->d_name, ".") == 0)
                         continue;
@@ -3835,18 +4636,13 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
 
                 lookup_frame = copy_frame (frame);
                 if (!lookup_frame) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of Memory");
                         /* out of memory, let the rmdir fail
                            (as non-empty, unfortunately) */
                         goto err;
                 }
 
-                lookup_local = GF_CALLOC (sizeof (*local), 1,
-                                          gf_dht_mt_dht_local_t);
+                lookup_local = mem_get0 (this->local_pool);
                 if (!lookup_local) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of Memory");
                         goto err;
                 }
 
@@ -3858,6 +4654,8 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
                 if (build_ret != 0)
                         goto err;
 
+                uuid_copy (lookup_local->loc.gfid, trav->d_stat.ia_gfid);
+
                 gf_log (this->name, GF_LOG_TRACE,
                         "looking up %s on %s",
                         lookup_local->loc.path, src->name);
@@ -3870,12 +4668,18 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
 
                 STACK_WIND (lookup_frame, dht_rmdir_lookup_cbk,
                             src, src->fops->lookup,
-                            &lookup_local->loc, NULL);
+                            &lookup_local->loc, xattrs);
                 ret++;
         }
 
+        if (xattrs)
+                dict_unref (xattrs);
+
         return ret;
 err:
+        if (xattrs)
+                dict_unref (xattrs);
+
         DHT_STACK_DESTROY (lookup_frame);
         return 0;
 }
@@ -3883,19 +4687,20 @@ err:
 
 int
 dht_rmdir_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int op_ret, int op_errno, gf_dirent_t *entries)
+                        int op_ret, int op_errno, gf_dirent_t *entries,
+                        dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = -1;
-	call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = -1;
+        call_frame_t *prev = NULL;
         xlator_t     *src = NULL;
         int           ret = 0;
 
-	local = frame->local;
-	prev  = cookie;
+        local = frame->local;
+        prev  = cookie;
         src   = prev->this;
 
-	if (op_ret > 2) {
+        if (op_ret > 2) {
                 ret = dht_rmdir_is_subvol_empty (frame, this, entries, src);
 
                 switch (ret) {
@@ -3913,724 +4718,495 @@ dht_rmdir_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 prev->this->name, local->loc.path, ret);
                         break;
                 }
-	}
+        }
 
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (is_last_call (this_call_cnt)) {
-		dht_rmdir_do (frame, this);
-	}
+        if (is_last_call (this_call_cnt)) {
+                dht_rmdir_do (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rmdir_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int op_ret, int op_errno, fd_t *fd)
-{
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = -1;
-	call_frame_t *prev = NULL;
-
-
-	local = frame->local;
-	prev  = cookie;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"opendir on %s for %s failed (%s)",
-			prev->this->name, local->loc.path,
-			strerror (op_errno));
-		goto err;
-	}
-
-	STACK_WIND (frame, dht_rmdir_readdirp_cbk,
-		    prev->this, prev->this->fops->readdirp,
-		    local->fd, 4096, 0);
-
-	return 0;
-
-err:
-	this_call_cnt = dht_frame_return (frame);
-
-	if (is_last_call (this_call_cnt)) {
-		dht_rmdir_do (frame, this);
-	}
-
-	return 0;
-}
-
-
-int
-dht_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
+                       int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
-        int           op_errno = -1;
-	int           i = -1;
-	int           ret = -1;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-
-	conf = this->private;
-
-	local = dht_local_init (frame);
-	if (!local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	local->call_cnt = conf->subvolume_cnt;
-	local->op_ret   = 0;
-
-	ret = loc_copy (&local->loc, loc);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	local->fd = fd_create (local->loc.inode, frame->root->pid);
-	if (!local->fd) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_rmdir_opendir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->opendir,
-			    loc, local->fd);
-	}
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (rmdir, frame, -1, op_errno,
-                          NULL, NULL);
-
-	return 0;
-}
-
-
-int
-dht_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int32_t op_ret, int32_t op_errno, dict_t *dict)
-{
-	DHT_STACK_UNWIND (xattrop, frame, op_ret, op_errno, dict);
-	return 0;
-}
-
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = -1;
+        call_frame_t *prev = NULL;
+        dict_t       *dict = NULL;
+        int           ret = 0;
+        dht_conf_t   *conf = this->private;
 
-int
-dht_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
-	     gf_xattrop_flags_t flags, dict_t *dict)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
+        local = frame->local;
+        prev  = cookie;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "opendir on %s for %s failed (%s)",
+                        prev->this->name, local->loc.path,
+                        strerror (op_errno));
+                if (op_errno != ENOENT) {
+                        local->op_ret = -1;
+                        local->op_errno = op_errno;
+                }
+                goto err;
+        }
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        dict = dict_new ();
+        if (!dict) {
+                local->op_ret = -1;
+                local->op_errno = ENOMEM;
+                goto err;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        ret = dict_set_uint32 (dict, conf->link_xattr_name, 256);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set '%s' key",
+                        local->loc.path, conf->link_xattr_name);
 
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
+        STACK_WIND (frame, dht_rmdir_readdirp_cbk,
+                    prev->this, prev->this->fops->readdirp,
+                    local->fd, 4096, 0, dict);
 
-	STACK_WIND (frame,
-		    dht_xattrop_cbk,
-		    subvol, subvol->fops->xattrop,
-		    loc, flags, dict);
+        if (dict)
+                dict_unref (dict);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL);
-
-	return 0;
-}
-
-
-int
-dht_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int32_t op_ret, int32_t op_errno, dict_t *dict)
-{
-	DHT_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, dict);
-	return 0;
-}
-
-
-int
-dht_fxattrop (call_frame_t *frame, xlator_t *this,
-	      fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
-
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	STACK_WIND (frame,
-		    dht_fxattrop_cbk,
-		    subvol, subvol->fops->fxattrop,
-		    fd, flags, dict);
-
-	return 0;
+        this_call_cnt = dht_frame_return (frame);
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL);
+        if (is_last_call (this_call_cnt)) {
+                dht_rmdir_do (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_inodelk_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno)
-
-{
-	DHT_STACK_UNWIND (inodelk, frame, op_ret, op_errno);
-	return 0;
-}
-
-
-int32_t
-dht_inodelk (call_frame_t *frame, xlator_t *this,
-	     const char *volume, loc_t *loc, int32_t cmd, struct flock *lock)
+dht_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+           dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
+        int           i = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
         VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
-
-	STACK_WIND (frame,
-		    dht_inodelk_cbk,
-		    subvol, subvol->fops->inodelk,
-		    volume, loc, cmd, lock);
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (inodelk, frame, -1, op_errno);
-
-	return 0;
-}
-
-
-int
-dht_finodelk_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno)
-
-{
-	DHT_STACK_UNWIND (finodelk, frame, op_ret, op_errno);
-	return 0;
-}
+        conf = this->private;
 
+        local = dht_local_init (frame, loc, NULL, GF_FOP_RMDIR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-int
-dht_finodelk (call_frame_t *frame, xlator_t *this,
-	      const char *volume, fd_t *fd, int32_t cmd, struct flock *lock)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
+        local->call_cnt = conf->subvolume_cnt;
+        local->op_ret   = 0;
+        local->fop_succeeded = 0;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        local->flags = flags;
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local->fd = fd_create (local->loc.inode, frame->root->pid);
+        if (!local->fd) {
 
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	STACK_WIND (frame,
-		    dht_finodelk_cbk,
-		    subvol, subvol->fops->finodelk,
-		    volume, fd, cmd, lock);
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                STACK_WIND (frame, dht_rmdir_opendir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->opendir,
+                            loc, local->fd, NULL);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (finodelk, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (rmdir, frame, -1, op_errno,
+                          NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
-
 int
 dht_entrylk_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno)
+                 xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
 
 {
-	DHT_STACK_UNWIND (entrylk, frame, op_ret, op_errno);
-	return 0;
+        DHT_STACK_UNWIND (entrylk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
 dht_entrylk (call_frame_t *frame, xlator_t *this,
-	     const char *volume, loc_t *loc, const char *basename,
-	     entrylk_cmd cmd, entrylk_type type)
+             const char *volume, loc_t *loc, const char *basename,
+             entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        xlator_t     *subvol = NULL;
         int           op_errno = -1;
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_ENTRYLK);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
+        local->call_cnt = 1;
 
-	STACK_WIND (frame, dht_entrylk_cbk,
-		    subvol, subvol->fops->entrylk,
-		    volume, loc, basename, cmd, type);
+        STACK_WIND (frame, dht_entrylk_cbk,
+                    subvol, subvol->fops->entrylk,
+                    volume, loc, basename, cmd, type, xdata);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (entrylk, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (entrylk, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_fentrylk_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno)
+                  xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
 
 {
-	DHT_STACK_UNWIND (fentrylk, frame, op_ret, op_errno);
-	return 0;
+        DHT_STACK_UNWIND (fentrylk, frame, op_ret, op_errno, NULL);
+        return 0;
 }
 
 
 int
 dht_fentrylk (call_frame_t *frame, xlator_t *this,
-	      const char *volume, fd_t *fd, const char *basename,
-	      entrylk_cmd cmd, entrylk_type type)
+              const char *volume, fd_t *fd, const char *basename,
+              entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        xlator_t     *subvol = NULL;
         int           op_errno = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        subvol = dht_subvol_get_cached (this, fd->inode);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_fentrylk_cbk,
-		    subvol, subvol->fops->fentrylk,
-		    volume, fd, basename, cmd, type);
+        STACK_WIND (frame, dht_fentrylk_cbk,
+                    subvol, subvol->fops->fentrylk,
+                    volume, fd, basename, cmd, type, xdata);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fentrylk, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fentrylk, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int op_ret, int op_errno, struct iatt *statpre,
-                 struct iatt *statpost)
+dht_forget (xlator_t *this, inode_t *inode)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
-
+        uint64_t        ctx_int = 0;
+        dht_inode_ctx_t *ctx    = NULL;
+        dht_layout_t *layout = NULL;
 
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
+        inode_ctx_del (inode, this, &ctx_int);
 
-		dht_iatt_merge (this, &local->prebuf, statpre, prev->this);
-                dht_iatt_merge (this, &local->stbuf, statpost, prev->this);
-		
-		if (local->inode) {
-			local->prebuf.ia_ino = local->inode->ino;
-                        local->stbuf.ia_ino = local->inode->ino;
-                }
+        if (!ctx_int)
+                return 0;
 
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
+        ctx = (dht_inode_ctx_t *) (long) ctx_int;
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (setattr, frame, local->op_ret, local->op_errno,
-				  &local->prebuf, &local->stbuf);
+        layout = ctx->layout;
+        ctx->layout = NULL;
+        dht_layout_unref (this, layout);
+        GF_FREE (ctx);
 
         return 0;
 }
 
 
 int
-dht_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-             struct iatt *stbuf, int32_t valid)
+dht_notify (xlator_t *this, int event, void *data, ...)
 {
-	dht_layout_t *layout = NULL;
-	dht_local_t  *local  = NULL;
-        int           op_errno = -1;
-	int           i = -1;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"memory allocation failed :(");
-		goto err;
-	}
-
-	local->layout = layout = dht_layout_get (this, loc->inode);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no layout for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	if (!layout_is_sane (layout)) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"layout is not sane for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = layout->cnt;
+        xlator_t                *subvol = NULL;
+        int                      cnt    = -1;
+        int                      i      = -1;
+        dht_conf_t              *conf   = NULL;
+        int                      ret    = -1;
+        int                      propagate = 0;
+
+        int                      had_heard_from_all = 0;
+        int                      have_heard_from_all = 0;
+        struct timeval           time = {0,};
+        gf_defrag_info_t        *defrag = NULL;
+        dict_t                  *dict   = NULL;
+        gf_defrag_type           cmd    = 0;
+        dict_t                  *output = NULL;
+        va_list                  ap;
 
-	for (i = 0; i < layout->cnt; i++) {
-		STACK_WIND (frame, dht_setattr_cbk,
-			    layout->list[i].xlator,
-			    layout->list[i].xlator->fops->setattr,
-			    loc, stbuf, valid);
-	}
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
-}
+        conf = this->private;
+        if (!conf)
+                return ret;
 
+        /* had all subvolumes reported status once till now? */
+        had_heard_from_all = 1;
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (!conf->last_event[i]) {
+                        had_heard_from_all = 0;
+                }
+        }
 
-int
-dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
-              int32_t valid)
-{
-	dht_layout_t *layout = NULL;
-	dht_local_t  *local  = NULL;
-        int           op_errno = -1;
-	int           i = -1;
+        switch (event) {
+        case GF_EVENT_CHILD_UP:
+                subvol = data;
 
+                conf->gen++;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (subvol == conf->subvolumes[i]) {
+                                cnt = i;
+                                break;
+                        }
+                }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+                if (cnt == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "got GF_EVENT_CHILD_UP bad subvolume %s",
+                                subvol->name);
+                        break;
+                }
 
-	local->layout = layout = dht_layout_get (this, fd->inode);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no layout for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+                gettimeofday (&time, NULL);
+                LOCK (&conf->subvolume_lock);
+                {
+                        conf->subvolume_status[cnt] = 1;
+                        conf->last_event[cnt] = event;
+                        conf->subvol_up_time[cnt] = time.tv_sec;
+                }
+                UNLOCK (&conf->subvolume_lock);
 
-	if (!layout_is_sane (layout)) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"layout is not sane for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+                /* one of the node came back up, do a stat update */
+                dht_get_du_info_for_subvol (this, cnt);
 
-	local->inode = inode_ref (fd->inode);
-	local->call_cnt = layout->cnt;
+                break;
 
-	for (i = 0; i < layout->cnt; i++) {
-		STACK_WIND (frame, dht_setattr_cbk,
-			    layout->list[i].xlator,
-			    layout->list[i].xlator->fops->fsetattr,
-			    fd, stbuf, valid);
-	}
+        case GF_EVENT_CHILD_MODIFIED:
+                subvol = data;
 
-	return 0;
+                conf->gen++;
+                propagate = 1;
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL);
+                break;
 
-	return 0;
-}
+        case GF_EVENT_CHILD_DOWN:
+                subvol = data;
 
+                if (conf->assert_no_child_down) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Received CHILD_DOWN. Exiting");
+                        if (conf->defrag) {
+                                gf_defrag_stop (conf->defrag, NULL);
+                        } else {
+                                kill (getpid(), SIGTERM);
+                        }
+                }
 
-int
-dht_forget (xlator_t *this, inode_t *inode)
-{
-	uint64_t      tmp_layout = 0;
-	dht_layout_t *layout = NULL;
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (subvol == conf->subvolumes[i]) {
+                                cnt = i;
+                                break;
+                        }
+                }
 
-	inode_ctx_get (inode, this, &tmp_layout);
+                if (cnt == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "got GF_EVENT_CHILD_DOWN bad subvolume %s",
+                                subvol->name);
+                        break;
+                }
 
-	if (!tmp_layout)
-		return 0;
+                LOCK (&conf->subvolume_lock);
+                {
+                        conf->subvolume_status[cnt] = 0;
+                        conf->last_event[cnt] = event;
+                        conf->subvol_up_time[cnt] = 0;
+                }
+                UNLOCK (&conf->subvolume_lock);
 
-	layout = (dht_layout_t *)(long)tmp_layout;
-        dht_layout_unref (this, layout);
+                break;
 
-	return 0;
-}
+        case GF_EVENT_CHILD_CONNECTING:
+                subvol = data;
 
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (subvol == conf->subvolumes[i]) {
+                                cnt = i;
+                                break;
+                        }
+                }
 
+                if (cnt == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "got GF_EVENT_CHILD_CONNECTING bad subvolume %s",
+                                subvol->name);
+                        break;
+                }
 
-int
-dht_init_subvolumes (xlator_t *this, dht_conf_t *conf)
-{
-        xlator_list_t *subvols = NULL;
-        int            cnt = 0;
+                LOCK (&conf->subvolume_lock);
+                {
+                        conf->last_event[cnt] = event;
+                }
+                UNLOCK (&conf->subvolume_lock);
 
+                break;
+        case GF_EVENT_VOLUME_DEFRAG:
+        {
+                if (!conf->defrag) {
+                        return ret;
+                }
+                defrag = conf->defrag;
 
-        for (subvols = this->children; subvols; subvols = subvols->next)
-                cnt++;
+                dict = data;
+                va_start (ap, data);
+                output = va_arg (ap, dict_t*);
 
-        conf->subvolumes = GF_CALLOC (cnt, sizeof (xlator_t *),
-                                      gf_dht_mt_xlator_t);
-        if (!conf->subvolumes) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                return -1;
+                ret = dict_get_int32 (dict, "rebalance-command",
+                                      (int32_t*)&cmd);
+                if (ret)
+                        return ret;
+                LOCK (&defrag->lock);
+                {
+                        if (defrag->is_exiting)
+                                goto unlock;
+                        if (cmd == GF_DEFRAG_CMD_STATUS)
+                                gf_defrag_status_get (defrag, output);
+                        else if (cmd == GF_DEFRAG_CMD_STOP)
+                                gf_defrag_stop (defrag, output);
+                }
+unlock:
+                UNLOCK (&defrag->lock);
+                return 0;
+                break;
         }
-        conf->subvolume_cnt = cnt;
-
-        cnt = 0;
-        for (subvols = this->children; subvols; subvols = subvols->next)
-                conf->subvolumes[cnt++] = subvols->xlator;
-
-	conf->subvolume_status = GF_CALLOC (cnt, sizeof (char),
-                                            gf_dht_mt_char);
-	if (!conf->subvolume_status) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		return -1;
-	}
-
-        return 0;
-}
 
+        default:
+                propagate = 1;
+                break;
+        }
 
-int
-dht_notify (xlator_t *this, int event, void *data, ...)
-{
-	xlator_t   *subvol = NULL;
-	int         cnt    = -1;
-	int         i      = -1;
-	dht_conf_t *conf   = NULL;
-	int         ret    = -1;
-
-        /* Notify is on by default for all events except CHILD_UP. Do not
-         * notify parent for CHILD_UP till all distribute children have come up.
-         */
-        int         notify = 1;
 
-	conf = this->private;
+        /* have all subvolumes reported status once by now? */
+        have_heard_from_all = 1;
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (!conf->last_event[i])
+                        have_heard_from_all = 0;
+        }
 
-	switch (event) {
-	case GF_EVENT_CHILD_UP:
-		subvol = data;
+        /* if all subvols have reported status, no need to hide anything
+           or wait for anything else. Just propagate blindly */
+        if (have_heard_from_all) {
+                propagate = 1;
 
-		conf->gen++;
+        }
 
-		for (i = 0; i < conf->subvolume_cnt; i++) {
-			if (subvol == conf->subvolumes[i]) {
-				cnt = i;
-				break;
-			}
-		}
 
-		if (cnt == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"got GF_EVENT_CHILD_UP bad subvolume %s",
-				subvol->name);
-			break;
-		}
+        if (!had_heard_from_all && have_heard_from_all) {
+                /* This is the first event which completes aggregation
+                   of events from all subvolumes. If at least one subvol
+                   had come up, propagate CHILD_UP, but only this time
+                */
+                event = GF_EVENT_CHILD_DOWN;
 
-		LOCK (&conf->subvolume_lock);
-		{
-                        if (!conf->subvolume_status[cnt]) {
-                                conf->subvolume_status[cnt] = 1;
-                                ++conf->upsubvols;
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (conf->last_event[i] == GF_EVENT_CHILD_UP) {
+                                event = GF_EVENT_CHILD_UP;
+                                break;
                         }
 
-                        if (conf->upsubvols != conf->subvolume_cnt)
-                                notify = 0;
-		}
-		UNLOCK (&conf->subvolume_lock);
-
-                /* one of the node came back up, do a stat update */
-                dht_get_du_info_for_subvol (this, cnt);
-
-		break;
+                        if (conf->last_event[i] == GF_EVENT_CHILD_CONNECTING) {
+                                event = GF_EVENT_CHILD_CONNECTING;
+                                /* continue to check other events for CHILD_UP */
+                        }
+                }
 
-	case GF_EVENT_CHILD_DOWN:
-		subvol = data;
+                /* rebalance is started with assert_no_child_down. So we do
+                 * not need to handle CHILD_DOWN event here.
+                 */
+                if (conf->defrag) {
+                        ret = gf_thread_create (&conf->defrag->th, NULL,
+						gf_defrag_start, this);
+                        if (ret) {
+                                conf->defrag = NULL;
+                                GF_FREE (conf->defrag);
+                                kill (getpid(), SIGTERM);
+                        }
+                }
+        }
 
-		for (i = 0; i < conf->subvolume_cnt; i++) {
-			if (subvol == conf->subvolumes[i]) {
-				cnt = i;
-				break;
-			}
-		}
+        ret = 0;
+        if (propagate)
+                ret = default_notify (this, event, data);
 
-		if (cnt == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"got GF_EVENT_CHILD_DOWN bad subvolume %s",
-				subvol->name);
-			break;
-		}
+        return ret;
+}
 
-		LOCK (&conf->subvolume_lock);
-		{
-                        if (conf->subvolume_status[cnt]) {
-                                conf->subvolume_status[cnt] = 0;
-                                --conf->upsubvols;
-                        }
-		}
-		UNLOCK (&conf->subvolume_lock);
+int
+dht_inode_ctx_layout_get (inode_t *inode, xlator_t *this, dht_layout_t **layout)
+{
+        dht_inode_ctx_t         *ctx            = NULL;
+        int                      ret            = -1;
 
-		break;
-	}
+        ret = dht_inode_ctx_get (inode, this, &ctx);
 
-        if (notify)
-                ret = default_notify (this, event, data);
+        if (!ret && ctx) {
+                if (ctx->layout) {
+                        if (layout)
+                                *layout = ctx->layout;
+                        ret = 0;
+                } else {
+                        ret = -1;
+                }
+        }
 
-	return ret;
+        return ret;
 }
-
diff --git a/xlators/cluster/dht/src/dht-common.h b/xlators/cluster/dht/src/dht-common.h
index 918eb566b..5ccd66799 100644
--- a/xlators/cluster/dht/src/dht-common.h
+++ b/xlators/cluster/dht/src/dht-common.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -22,281 +13,775 @@
 #include "config.h"
 #endif
 
+#include <regex.h>
+
 #include "dht-mem-types.h"
+#include "libxlator.h"
+#include "syncop.h"
 
 #ifndef _DHT_H
 #define _DHT_H
 
-#define GF_XATTR_FIX_LAYOUT_KEY   "trusted.distribute.fix.layout"
+#define GF_XATTR_FIX_LAYOUT_KEY     "distribute.fix.layout"
 #define GF_DHT_LOOKUP_UNHASHED_ON   1
 #define GF_DHT_LOOKUP_UNHASHED_AUTO 2
+#define DHT_PATHINFO_HEADER         "DISTRIBUTE:"
 
 #include <fnmatch.h>
 
 typedef int (*dht_selfheal_dir_cbk_t) (call_frame_t *frame, void *cookie,
-				       xlator_t *this,
-				       int32_t op_ret, int32_t op_errno);
+                                       xlator_t     *this,
+                                       int32_t       op_ret, int32_t op_errno,
+                                       dict_t *xdata);
+typedef int (*dht_defrag_cbk_fn_t) (xlator_t        *this, call_frame_t *frame,
+                                    int              ret);
 
 
 struct dht_layout {
-        int               cnt;
-	int               preset;
-        int               gen;
-	int               type;
-        int               ref;   /* use with dht_conf_t->layout_lock */
-        int               search_unhashed;
+        int                spread_cnt;  /* layout spread count per directory,
+                                           is controlled by 'setxattr()' with
+                                           special key */
+        int                cnt;
+        int                preset;
+        int                gen;
+        int                type;
+        int                ref; /* use with dht_conf_t->layout_lock */
+        int                search_unhashed;
         struct {
-		int       err;   /* 0 = normal
-				   -1 = dir exists and no xattr
-				   >0 = dir lookup failed with errno
-				 */
-                uint32_t  start;
-                uint32_t  stop;
-                xlator_t *xlator;
-        } list[0];
+                int        err;   /* 0 = normal
+                                     -1 = dir exists and no xattr
+                                     >0 = dir lookup failed with errno
+                                  */
+                uint32_t   start;
+                uint32_t   stop;
+                xlator_t  *xlator;
+        } list[];
+};
+typedef struct dht_layout  dht_layout_t;
+
+struct dht_stat_time {
+        uint32_t        atime;
+        uint32_t        atime_nsec;
+        uint32_t        ctime;
+        uint32_t        ctime_nsec;
+        uint32_t        mtime;
+        uint32_t        mtime_nsec;
+};
+
+typedef struct dht_stat_time dht_stat_time_t;
+
+struct dht_inode_ctx {
+        dht_layout_t    *layout;
+        dht_stat_time_t  time;
 };
-typedef struct dht_layout dht_layout_t;
+
+typedef struct dht_inode_ctx dht_inode_ctx_t;
 
 
 typedef enum {
-	DHT_HASH_TYPE_DM,
+        DHT_HASH_TYPE_DM,
+        DHT_HASH_TYPE_DM_USER,
 } dht_hashfn_type_t;
 
+/* rebalance related */
+struct dht_rebalance_ {
+        xlator_t            *from_subvol;
+        xlator_t            *target_node;
+        off_t                offset;
+        size_t               size;
+        int32_t              flags;
+        int                  count;
+        struct iobref       *iobref;
+        struct iovec        *vector;
+        struct iatt          stbuf;
+        dht_defrag_cbk_fn_t  target_op_fn;
+        dict_t              *xdata;
+};
 
 struct dht_local {
-	int                      call_cnt;
-	loc_t                    loc;
-	loc_t                    loc2;
-	int                      op_ret;
-	int                      op_errno;
-	int                      layout_mismatch;
+        int                      call_cnt;
+        loc_t                    loc;
+        loc_t                    loc2;
+        int                      op_ret;
+        int                      op_errno;
+        int                      layout_mismatch;
         /* Use stbuf as the postbuf, when we require both
          * pre and post attrs */
-	struct iatt              stbuf;
+        struct iatt              stbuf;
         struct iatt              prebuf;
         struct iatt              preoldparent;
         struct iatt              postoldparent;
         struct iatt              preparent;
         struct iatt              postparent;
-	struct statvfs           statvfs;
-	fd_t                    *fd;
-	inode_t                 *inode;
+        struct statvfs           statvfs;
+        fd_t                    *fd;
+        inode_t                 *inode;
         dict_t                  *params;
-	dict_t                  *xattr;
-	dict_t                  *xattr_req;
-	dht_layout_t            *layout;
-	size_t                   size;
-	ino_t                    ia_ino;
-	xlator_t                *src_hashed, *src_cached;
-	xlator_t                *dst_hashed, *dst_cached;
-	xlator_t                *cached_subvol;
-	xlator_t                *hashed_subvol;
-	char                     need_selfheal;
+        dict_t                  *xattr;
+        dict_t                  *xattr_req;
+        dht_layout_t            *layout;
+        size_t                   size;
+        ino_t                    ia_ino;
+        xlator_t                *src_hashed, *src_cached;
+        xlator_t                *dst_hashed, *dst_cached;
+        xlator_t                *cached_subvol;
+        xlator_t                *hashed_subvol;
+        char                     need_selfheal;
         int                      file_count;
         int                      dir_count;
         call_frame_t            *main_frame;
-	struct {
-		fop_mknod_cbk_t  linkfile_cbk;
-		struct iatt      stbuf;
-		loc_t            loc;
-		inode_t         *inode;
-		dict_t          *xattr;
-		xlator_t        *srcvol;
-	} linkfile;
-	struct {
-		uint32_t         hole_cnt;
-		uint32_t         overlaps_cnt;
-		uint32_t         missing;
-		uint32_t         down;
-		uint32_t         misc;
-		dht_selfheal_dir_cbk_t   dir_cbk;
-		dht_layout_t    *layout;
-	} selfheal;
+        int                      fop_succeeded;
+        struct {
+                fop_mknod_cbk_t  linkfile_cbk;
+                struct iatt      stbuf;
+                loc_t            loc;
+                inode_t         *inode;
+                dict_t          *xattr;
+                xlator_t        *srcvol;
+        } linkfile;
+        struct {
+                uint32_t         hole_cnt;
+                uint32_t         overlaps_cnt;
+                uint32_t         down;
+                uint32_t         misc;
+                dht_selfheal_dir_cbk_t   dir_cbk;
+                dht_layout_t    *layout;
+        } selfheal;
         uint32_t                 uid;
         uint32_t                 gid;
 
-	/* needed by nufa */
-	int32_t flags;
-	mode_t  mode;
-	dev_t   rdev;
+        /* needed by nufa */
+        int32_t flags;
+        mode_t  mode;
+        dev_t   rdev;
+        mode_t  umask;
 
         /* need for file-info */
-        char   *pathinfo;
+        char   *xattr_val;
         char   *key;
 
+        /* which xattr request? */
+        char xsel[256];
+        int32_t alloc_len;
+
         char   *newpath;
+
+        /* gfid related */
+        uuid_t  gfid;
+
+        /*Marker Related*/
+        struct marker_str    marker;
+
+        /* flag used to make sure we need to return estale in
+           {lookup,revalidate}_cbk */
+        char return_estale;
+        char need_lookup_everywhere;
+
+        glusterfs_fop_t      fop;
+
+        gf_boolean_t     linked;
+        xlator_t        *link_subvol;
+
+        struct dht_rebalance_ rebalance;
+        xlator_t        *first_up_subvol;
+
 };
 typedef struct dht_local dht_local_t;
 
 /* du - disk-usage */
 struct dht_du {
         double   avail_percent;
+	double   avail_inodes;
         uint64_t avail_space;
         uint32_t log;
 };
 typedef struct dht_du dht_du_t;
 
+enum gf_defrag_type {
+        GF_DEFRAG_CMD_START = 1,
+        GF_DEFRAG_CMD_STOP = 1 + 1,
+        GF_DEFRAG_CMD_STATUS = 1 + 2,
+        GF_DEFRAG_CMD_START_LAYOUT_FIX = 1 + 3,
+        GF_DEFRAG_CMD_START_FORCE = 1 + 4,
+};
+typedef enum gf_defrag_type gf_defrag_type;
+
+enum gf_defrag_status_t {
+        GF_DEFRAG_STATUS_NOT_STARTED,
+        GF_DEFRAG_STATUS_STARTED,
+        GF_DEFRAG_STATUS_STOPPED,
+        GF_DEFRAG_STATUS_COMPLETE,
+        GF_DEFRAG_STATUS_FAILED,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_STOPPED,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_FAILED,
+};
+typedef enum gf_defrag_status_t gf_defrag_status_t;
+
+typedef struct gf_defrag_pattern_list gf_defrag_pattern_list_t;
+
+struct gf_defrag_pattern_list {
+        char                       path_pattern[256];
+        uint64_t                   size;
+        gf_defrag_pattern_list_t  *next;
+};
+
+struct gf_defrag_info_ {
+        uint64_t                     total_files;
+        uint64_t                     total_data;
+        uint64_t                     num_files_lookedup;
+        uint64_t                     total_failures;
+        uint64_t                     skipped;
+        gf_lock_t                    lock;
+        int                          cmd;
+        pthread_t                    th;
+        gf_defrag_status_t           defrag_status;
+        struct rpc_clnt             *rpc;
+        uint32_t                     connected;
+        uint32_t                     is_exiting;
+        pid_t                        pid;
+        inode_t                     *root_inode;
+        uuid_t                       node_uuid;
+        struct timeval               start_time;
+        gf_boolean_t                 stats;
+        gf_defrag_pattern_list_t    *defrag_pattern;
+};
+
+typedef struct gf_defrag_info_ gf_defrag_info_t;
+
 struct dht_conf {
-	gf_lock_t      subvolume_lock;
+        gf_lock_t      subvolume_lock;
         int            subvolume_cnt;
         xlator_t     **subvolumes;
-	char          *subvolume_status;
-	dht_layout_t **file_layouts;
-	dht_layout_t **dir_layouts;
-	dht_layout_t  *default_dir_layout;
-	gf_boolean_t   search_unhashed;
-	int            gen;
+        char          *subvolume_status;
+        int           *last_event;
+        dht_layout_t **file_layouts;
+        dht_layout_t **dir_layouts;
+        gf_boolean_t   search_unhashed;
+        int            gen;
         dht_du_t      *du_stats;
-        uint64_t       min_free_disk;
+        double         min_free_disk;
+	double         min_free_inodes;
         char           disk_unit;
         int32_t        refresh_interval;
         gf_boolean_t   unhashed_sticky_bit;
-	struct timeval last_stat_fetch;
+        struct timeval last_stat_fetch;
         gf_lock_t      layout_lock;
         void          *private;     /* Can be used by wrapper xlators over
                                        dht */
         gf_boolean_t   use_readdirp;
-        int             upsubvols;      /* Count the number of online children*/
+        char           vol_uuid[UUID_SIZE + 1];
+        gf_boolean_t   assert_no_child_down;
+        time_t        *subvol_up_time;
+
+        /* This is the count used as the distribute layout for a directory */
+        /* Will be a global flag to control the layout spread count */
+        uint32_t       dir_spread_cnt;
+
+        /* to keep track of nodes which are decomissioned */
+        xlator_t     **decommissioned_bricks;
+        int            decommission_in_progress;
+        int            decommission_subvols_cnt;
+
+        /* defrag related */
+        gf_defrag_info_t *defrag;
+
+        /* Request to filter directory entries in readdir request */
+
+        gf_boolean_t    readdir_optimize;
+
+        /* Support regex-based name reinterpretation. */
+        regex_t         rsync_regex;
+        gf_boolean_t    rsync_regex_valid;
+        regex_t         extra_regex;
+        gf_boolean_t    extra_regex_valid;
+
+        /* Support variable xattr names. */
+        char            *xattr_name;
+        char            *link_xattr_name;
+        char            *wild_xattr_name;
 };
 typedef struct dht_conf dht_conf_t;
 
 
 struct dht_disk_layout {
-	uint32_t           cnt;
-	uint32_t           type;
-	struct {
-		uint32_t   start;
-		uint32_t   stop;
-	} list[1];
+        uint32_t           cnt;
+        uint32_t           type;
+        struct {
+                uint32_t   start;
+                uint32_t   stop;
+        } list[1];
 };
 typedef struct dht_disk_layout dht_disk_layout_t;
 
-#define WIPE(statp) do { typeof(*statp) z = {0,}; if (statp) *statp = z; } while (0)
+typedef enum {
+        GF_DHT_MIGRATE_DATA,
+        GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS,
+        GF_DHT_MIGRATE_HARDLINK,
+        GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS
+} gf_dht_migrate_data_type_t;
 
 #define ENTRY_MISSING(op_ret, op_errno) (op_ret == -1 && op_errno == ENOENT)
 
-#define is_fs_root(loc) (strcmp (loc->path, "/") == 0)
-
-#define is_revalidate(loc) (inode_ctx_get (loc->inode, this, NULL) == 0)
+#define is_revalidate(loc) (dht_inode_ctx_layout_get (loc->inode, this, NULL) == 0)
 
 #define is_last_call(cnt) (cnt == 0)
 
-#define DHT_LINKFILE_MODE (S_ISVTX)
-#define check_is_linkfile(i,s,x) (                                      \
-                ((st_mode_from_ia (s->ia_prot, s->ia_type) & ~S_IFMT)   \
-                 == DHT_LINKFILE_MODE) &&                               \
-                (s->ia_size == 0))
+#define DHT_MIGRATION_IN_PROGRESS 1
+#define DHT_MIGRATION_COMPLETED   2
+
+#define DHT_LINKFILE_MODE        (S_ISVTX)
+
+#define check_is_linkfile(i,s,x,n) (                                      \
+                ((st_mode_from_ia ((s)->ia_prot, (s)->ia_type) & ~S_IFMT) \
+                 == DHT_LINKFILE_MODE) &&                                 \
+                dict_get (x, n))
+
+#define IS_DHT_MIGRATION_PHASE2(buf)  (                                 \
+                IA_ISREG ((buf)->ia_type) &&                            \
+                ((st_mode_from_ia ((buf)->ia_prot, (buf)->ia_type) &    \
+                  ~S_IFMT) == DHT_LINKFILE_MODE))
+
+#define IS_DHT_MIGRATION_PHASE1(buf)  (                                 \
+                IA_ISREG ((buf)->ia_type) &&                            \
+                ((buf)->ia_prot.sticky == 1) &&                         \
+                ((buf)->ia_prot.sgid == 1))
+
+#define DHT_STRIP_PHASE1_FLAGS(buf)  do {                       \
+                if ((buf) && IS_DHT_MIGRATION_PHASE1(buf)) {    \
+                        (buf)->ia_prot.sticky = 0;              \
+                        (buf)->ia_prot.sgid = 0;                \
+                }                                               \
+        } while (0)
 
 #define check_is_dir(i,s,x) (IA_ISDIR(s->ia_type))
 
 #define layout_is_sane(layout) ((layout) && (layout->cnt > 0))
 
 #define DHT_STACK_UNWIND(fop, frame, params ...) do {           \
-		dht_local_t *__local = NULL;                    \
-                xlator_t *__xl = NULL;                          \
+                dht_local_t *__local = NULL;                    \
+                xlator_t    *__xl    = NULL;                    \
                 if (frame) {                                    \
-                        __xl = frame->this;                     \
-                        __local = frame->local;                 \
+                        __xl         = frame->this;             \
+                        __local      = frame->local;            \
                         frame->local = NULL;                    \
                 }                                               \
-		STACK_UNWIND_STRICT (fop, frame, params);       \
-		dht_local_wipe (__xl, __local);                 \
-	} while (0)
-
-#define DHT_STACK_DESTROY(frame) do {		       \
-		dht_local_t *__local = NULL;           \
-                xlator_t *__xl = NULL;                 \
-                __xl = frame->this;                    \
-		__local = frame->local;                \
-		frame->local = NULL;		       \
-		STACK_DESTROY (frame->root);	       \
-		dht_local_wipe (__xl, __local);	       \
-	} while (0)
-
-dht_layout_t *dht_layout_new (xlator_t *this, int cnt);
-dht_layout_t *dht_layout_get (xlator_t *this, inode_t *inode);
-dht_layout_t *dht_layout_for_subvol (xlator_t *this, xlator_t *subvol);
-xlator_t *dht_layout_search (xlator_t *this, dht_layout_t *layout,
-			     const char *name);
-int dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout);
-int dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
-			  uint32_t *holes_p, uint32_t *overlaps_p,
-			  uint32_t *missing_p, uint32_t *down_p,
-			  uint32_t *misc_p);
-int dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout,
-			     xlator_t *subvol, loc_t *loc, dict_t *xattr);
+                STACK_UNWIND_STRICT (fop, frame, params);       \
+                dht_local_wipe (__xl, __local);                 \
+        } while (0)
+
+#define DHT_STACK_DESTROY(frame) do {                   \
+                dht_local_t *__local = NULL;            \
+                xlator_t    *__xl    = NULL;            \
+                __xl                 = frame->this;     \
+                __local              = frame->local;    \
+                frame->local         = NULL;            \
+                STACK_DESTROY (frame->root);            \
+                dht_local_wipe (__xl, __local);         \
+        } while (0)
+
+#define DHT_UPDATE_TIME(ctx_sec, ctx_nsec, new_sec, new_nsec, inode, post) do {\
+                int32_t sec = 0;                                        \
+                sec = new_sec;                                          \
+                LOCK (&inode->lock);                                    \
+                {                                                       \
+                        new_sec = max(new_sec, ctx_sec);                \
+                        if (sec < new_sec)                              \
+                                new_nsec = ctx_nsec;                    \
+                        if (sec == new_sec)                             \
+                                new_nsec = max (new_nsec, ctx_nsec);    \
+                        if (post) {                                     \
+                                ctx_sec = new_sec;                      \
+                                ctx_nsec = new_nsec;                    \
+                        }                                               \
+                }                                                       \
+                UNLOCK (&inode->lock);                                  \
+        } while (0)
+
+#define is_greater_time(a, an, b, bn) (((a) < (b)) || (((a) == (b)) && ((an) < (bn))))
+dht_layout_t                            *dht_layout_new (xlator_t *this, int cnt);
+dht_layout_t                            *dht_layout_get (xlator_t *this, inode_t *inode);
+dht_layout_t                            *dht_layout_for_subvol (xlator_t *this, xlator_t *subvol);
+xlator_t *dht_layout_search (xlator_t   *this, dht_layout_t *layout,
+                             const char *name);
+int                                      dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout);
+int dht_layout_anomalies (xlator_t      *this, loc_t *loc, dht_layout_t *layout,
+                          uint32_t      *holes_p, uint32_t *overlaps_p,
+                          uint32_t      *missing_p, uint32_t *down_p,
+                          uint32_t      *misc_p, uint32_t *no_space_p);
+int dht_layout_dir_mismatch (xlator_t   *this, dht_layout_t *layout,
+                             xlator_t   *subvol, loc_t *loc, dict_t *xattr);
 
 xlator_t *dht_linkfile_subvol (xlator_t *this, inode_t *inode,
-			       struct iatt *buf, dict_t *xattr);
-int dht_linkfile_unlink (call_frame_t *frame, xlator_t *this,
-			 xlator_t *subvol, loc_t *loc);
+                               struct iatt *buf, dict_t *xattr);
+int dht_linkfile_unlink (call_frame_t      *frame, xlator_t *this,
+                         xlator_t          *subvol, loc_t *loc);
 
 int dht_layouts_init (xlator_t *this, dht_conf_t *conf);
 int dht_layout_merge (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
-		      int op_ret, int op_errno, dict_t *xattr);
+                      int       op_ret, int op_errno, dict_t *xattr);
 
 int dht_disk_layout_extract (xlator_t *this, dht_layout_t *layout,
-			     int pos, int32_t **disk_layout_p);
-int dht_disk_layout_merge (xlator_t *this, dht_layout_t *layout,
-			   int pos, void *disk_layout_raw);
+                             int       pos, int32_t **disk_layout_p);
+int dht_disk_layout_merge (xlator_t   *this, dht_layout_t *layout,
+                           int         pos, void *disk_layout_raw, int disk_layout_len);
 
 
 int dht_frame_return (call_frame_t *frame);
 
-int dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y);
+int                             dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y);
 int dht_deitransform (xlator_t *this, uint64_t y, xlator_t **subvol,
-		      uint64_t *x);
+                      uint64_t *x);
 
 void dht_local_wipe (xlator_t *this, dht_local_t *local);
-dht_local_t *dht_local_init (call_frame_t *frame);
-int dht_iatt_merge (xlator_t *this, struct iatt *to, struct iatt *from,
-		    xlator_t *subvol);
+dht_local_t *dht_local_init (call_frame_t    *frame, loc_t *loc, fd_t *fd,
+                             glusterfs_fop_t  fop);
+int dht_iatt_merge (xlator_t                 *this, struct iatt *to, struct iatt *from,
+                    xlator_t                 *subvol);
 
 xlator_t *dht_subvol_get_hashed (xlator_t *this, loc_t *loc);
 xlator_t *dht_subvol_get_cached (xlator_t *this, inode_t *inode);
 xlator_t *dht_subvol_next (xlator_t *this, xlator_t *prev);
-int dht_subvol_cnt (xlator_t *this, xlator_t *subvol);
+xlator_t *dht_subvol_next_available (xlator_t *this, xlator_t *prev);
+int       dht_subvol_cnt (xlator_t *this, xlator_t *subvol);
 
-int dht_hash_compute (int type, const char *name, uint32_t *hash_p);
+int dht_hash_compute (xlator_t *this, int type, const char *name, uint32_t *hash_p);
 
-int dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
-			 xlator_t *tovol, xlator_t *fromvol, loc_t *loc);
-int dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc);
-int dht_lookup_everywhere (call_frame_t *frame, xlator_t *this, loc_t *loc);
+int dht_linkfile_create (call_frame_t    *frame, fop_mknod_cbk_t linkfile_cbk,
+                         xlator_t        *this, xlator_t *tovol,
+                         xlator_t        *fromvol, loc_t *loc);
+int                                       dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc);
+int                                       dht_lookup_everywhere (call_frame_t *frame, xlator_t *this, loc_t *loc);
 int
-dht_selfheal_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
-			loc_t *loc, dht_layout_t *layout);
+dht_selfheal_directory (call_frame_t     *frame, dht_selfheal_dir_cbk_t cbk,
+                        loc_t            *loc, dht_layout_t *layout);
 int
 dht_selfheal_new_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
-			    dht_layout_t *layout);
+                            dht_layout_t *layout);
 int
-dht_selfheal_restore (call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
-		      loc_t *loc, dht_layout_t *layout);
+dht_selfheal_restore (call_frame_t       *frame, dht_selfheal_dir_cbk_t cbk,
+                      loc_t              *loc, dht_layout_t *layout);
 int
 dht_layout_sort_volname (dht_layout_t *layout);
 
-int dht_rename (call_frame_t *frame, xlator_t *this,
-		loc_t *oldloc, loc_t *newloc);
-
 int dht_get_du_info (call_frame_t *frame, xlator_t *this, loc_t *loc);
 
-int dht_is_subvol_filled (xlator_t *this, xlator_t *subvol);
-xlator_t *dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol);
-int dht_get_du_info_for_subvol (xlator_t *this, int subvol_idx);
+gf_boolean_t dht_is_subvol_filled (xlator_t *this, xlator_t *subvol);
+xlator_t *dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol,
+                                          dht_local_t *layout);
+int       dht_get_du_info_for_subvol (xlator_t *this, int subvol_idx);
 
 int dht_layout_preset (xlator_t *this, xlator_t *subvol, inode_t *inode);
-int dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout);
-void dht_layout_unref (xlator_t *this, dht_layout_t *layout);
+int           dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout);;
+void          dht_layout_unref (xlator_t *this, dht_layout_t *layout);
 dht_layout_t *dht_layout_ref (xlator_t *this, dht_layout_t *layout);
-xlator_t *dht_first_up_subvol (xlator_t *this);
-xlator_t *dht_last_up_subvol (xlator_t *this);
-int dht_frame_su_do (call_frame_t *frame);
-int dht_frame_su_undo (call_frame_t *frame);
+xlator_t     *dht_first_up_subvol (xlator_t *this);
+xlator_t     *dht_last_up_subvol (xlator_t *this);
 
 int dht_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name);
 
-int dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
+int dht_filter_loc_subvol_key (xlator_t  *this, loc_t *loc, loc_t *new_loc,
                                xlator_t **subvol);
 
+int                                     dht_rename_cleanup (call_frame_t *frame);
+int dht_rename_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int32_t           op_ret, int32_t op_errno,
+                          inode_t          *inode, struct iatt *stbuf,
+                          struct iatt      *preparent, struct iatt *postparent,
+                          dict_t *xdata);
+
+int dht_fix_directory_layout (call_frame_t *frame,
+                              dht_selfheal_dir_cbk_t  dir_cbk,
+                              dht_layout_t           *layout);
+
+int dht_init_subvolumes (xlator_t *this, dht_conf_t *conf);
+
+/* migration/rebalance */
+int dht_start_rebalance_task (xlator_t *this, call_frame_t *frame);
+
+int dht_rebalance_in_progress_check (xlator_t *this, call_frame_t *frame);
+int dht_rebalance_complete_check (xlator_t *this, call_frame_t *frame);
+
+
+/* FOPS */
+int32_t dht_lookup (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc,
+                    dict_t   *xattr_req);
+
+int32_t dht_stat (call_frame_t *frame,
+                  xlator_t *this,
+                  loc_t    *loc, dict_t *xdata);
+
+int32_t dht_fstat (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd, dict_t *xdata);
+
+int32_t dht_truncate (call_frame_t *frame,
+                      xlator_t *this,
+                      loc_t    *loc,
+                      off_t     offset, dict_t *xdata);
+
+int32_t dht_ftruncate (call_frame_t *frame,
+                       xlator_t *this,
+                       fd_t     *fd,
+                       off_t     offset, dict_t *xdata);
+
+int32_t dht_access (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc,
+                    int32_t   mask, dict_t *xdata);
+
+int32_t dht_readlink (call_frame_t *frame,
+                      xlator_t *this,
+                      loc_t    *loc,
+                      size_t    size, dict_t *xdata);
+
+int32_t dht_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                   mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata);
+
+int32_t dht_mkdir (call_frame_t *frame, xlator_t *this,
+                   loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata);
+
+int32_t dht_unlink (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc, int xflag, dict_t *xdata);
+
+int32_t dht_rmdir (call_frame_t *frame, xlator_t *this,
+                   loc_t    *loc, int flags, dict_t *xdata);
+
+int32_t dht_symlink (call_frame_t   *frame, xlator_t *this,
+                     const char *linkpath, loc_t *loc, mode_t umask,
+                     dict_t *xdata);
+
+int32_t dht_rename (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *oldloc,
+                    loc_t    *newloc, dict_t *xdata);
+
+int32_t dht_link (call_frame_t *frame,
+                  xlator_t *this,
+                  loc_t    *oldloc,
+                  loc_t    *newloc, dict_t *xdata);
+
+int32_t dht_create (call_frame_t *frame, xlator_t *this,
+                    loc_t    *loc, int32_t flags, mode_t mode,
+                    mode_t umask, fd_t     *fd, dict_t *params);
+
+int32_t dht_open (call_frame_t *frame,
+                  xlator_t *this,
+                  loc_t    *loc,
+                  int32_t   flags, fd_t *fd, dict_t *xdata);
+
+int32_t dht_readv (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd,
+                   size_t    size,
+                   off_t     offset, uint32_t flags, dict_t *xdata);
+
+int32_t dht_writev (call_frame_t      *frame,
+                    xlator_t      *this,
+                    fd_t          *fd,
+                    struct iovec  *vector,
+                    int32_t        count,
+                    off_t          offset,
+                    uint32_t       flags,
+                    struct iobref *iobref, dict_t *xdata);
+
+int32_t dht_flush (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd, dict_t *xdata);
+
+int32_t dht_fsync (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd,
+                   int32_t   datasync, dict_t *xdata);
+
+int32_t dht_opendir (call_frame_t *frame,
+                     xlator_t *this,
+                     loc_t    *loc, fd_t *fd, dict_t *xdata);
+
+int32_t dht_fsyncdir (call_frame_t *frame,
+                      xlator_t *this,
+                      fd_t     *fd,
+                      int32_t   datasync, dict_t *xdata);
+
+int32_t dht_statfs (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc, dict_t *xdata);
+
+int32_t dht_setxattr (call_frame_t *frame,
+                      xlator_t *this,
+                      loc_t    *loc,
+                      dict_t   *dict,
+                      int32_t   flags, dict_t *xdata);
+
+int32_t dht_getxattr (call_frame_t   *frame,
+                      xlator_t   *this,
+                      loc_t      *loc,
+                      const char *name, dict_t *xdata);
+
+int32_t dht_fsetxattr (call_frame_t *frame,
+                       xlator_t *this,
+                       fd_t     *fd,
+                       dict_t   *dict,
+                       int32_t   flags, dict_t *xdata);
+
+int32_t dht_fgetxattr (call_frame_t   *frame,
+                       xlator_t   *this,
+                       fd_t       *fd,
+                       const char *name, dict_t *xdata);
+
+int32_t dht_removexattr (call_frame_t   *frame,
+                         xlator_t   *this,
+                         loc_t      *loc,
+                         const char *name, dict_t *xdata);
+int32_t dht_fremovexattr (call_frame_t   *frame,
+                          xlator_t   *this,
+                          fd_t      *fd,
+                          const char *name, dict_t *xdata);
+
+int32_t dht_lk (call_frame_t        *frame,
+                xlator_t        *this,
+                fd_t            *fd,
+                int32_t          cmd,
+                struct gf_flock *flock, dict_t *xdata);
+
+int32_t dht_inodelk (call_frame_t *frame, xlator_t *this,
+                     const char      *volume, loc_t *loc, int32_t cmd,
+                     struct gf_flock *flock, dict_t *xdata);
+
+int32_t dht_finodelk (call_frame_t        *frame, xlator_t *this,
+                      const char      *volume, fd_t *fd, int32_t cmd,
+                      struct gf_flock *flock, dict_t *xdata);
+
+int32_t dht_entrylk (call_frame_t    *frame, xlator_t *this,
+                     const char  *volume, loc_t *loc, const char *basename,
+                     entrylk_cmd  cmd, entrylk_type type, dict_t *xdata);
+
+int32_t dht_fentrylk (call_frame_t    *frame, xlator_t *this,
+                      const char  *volume, fd_t *fd, const char *basename,
+                      entrylk_cmd  cmd, entrylk_type type, dict_t *xdata);
+
+int32_t dht_readdir (call_frame_t  *frame,
+                     xlator_t *this,
+                     fd_t     *fd,
+                     size_t    size, off_t off, dict_t *xdata);
+
+int32_t dht_readdirp (call_frame_t *frame,
+                      xlator_t *this,
+                      fd_t     *fd,
+                      size_t    size, off_t off, dict_t *dict);
+
+int32_t dht_xattrop (call_frame_t           *frame,
+                     xlator_t           *this,
+                     loc_t              *loc,
+                     gf_xattrop_flags_t  flags,
+                     dict_t             *dict, dict_t *xdata);
+
+int32_t dht_fxattrop (call_frame_t *frame,
+                      xlator_t           *this,
+                      fd_t               *fd,
+                      gf_xattrop_flags_t  flags,
+                      dict_t             *dict, dict_t *xdata);
+
+int32_t dht_forget (xlator_t *this, inode_t *inode);
+int32_t dht_setattr (call_frame_t  *frame, xlator_t *this, loc_t *loc,
+                     struct iatt   *stbuf, int32_t valid, dict_t *xdata);
+int32_t dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                      struct iatt  *stbuf, int32_t valid, dict_t *xdata);
+int32_t dht_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd,
+		      int32_t mode, off_t offset, size_t len, dict_t *xdata);
+int32_t dht_discard(call_frame_t *frame, xlator_t *this, fd_t *fd,
+		    off_t offset, size_t len, dict_t *xdata);
+int32_t dht_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd,
+                    off_t offset, size_t len, dict_t *xdata);
+
+int32_t dht_init (xlator_t *this);
+void    dht_fini (xlator_t *this);
+int     dht_reconfigure (xlator_t *this, dict_t *options);
+int32_t dht_notify (xlator_t *this, int32_t event, void *data, ...);
+
+/* definitions for nufa/switch */
+int dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int op_ret, int op_errno, inode_t *inode,
+                        struct iatt *stbuf, dict_t *xattr,
+                        struct iatt *postparent);
+int dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int op_ret, int op_errno, inode_t *inode,
+                        struct iatt *stbuf, dict_t *xattr,
+                        struct iatt *postparent);
+int dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
+                             xlator_t *this, int op_ret, int op_errno,
+                             inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                             struct iatt *postparent);
+int dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int op_ret, int op_errno,
+                    inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                    struct iatt *postparent);
+int dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int op_ret, int op_errno,
+                    fd_t *fd, inode_t *inode, struct iatt *stbuf,
+                    struct iatt *preparent, struct iatt *postparent,
+                    dict_t *xdata);
+int dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int op_ret, int op_errno,
+                     inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata);
+
+int
+gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict);
+
+int
+gf_defrag_stop (gf_defrag_info_t *defrag, dict_t *output);
+
+void*
+gf_defrag_start (void *this);
+
+int32_t
+gf_defrag_handle_hardlink (xlator_t *this, loc_t *loc, dict_t  *xattrs,
+                           struct iatt *stbuf);
+int
+dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
+                 int flag);
+int
+dht_inode_ctx_layout_get (inode_t *inode, xlator_t *this,
+                          dht_layout_t **layout_int);
+int
+dht_inode_ctx_layout_set (inode_t *inode, xlator_t *this,
+                          dht_layout_t* layout_int);
+int
+dht_inode_ctx_time_update (inode_t *inode, xlator_t *this, struct iatt *stat,
+                           int32_t update_ctx);
+
+int dht_inode_ctx_get (inode_t *inode, xlator_t *this, dht_inode_ctx_t **ctx);
+int dht_inode_ctx_set (inode_t *inode, xlator_t *this, dht_inode_ctx_t *ctx);
+int
+dht_dir_attr_heal (void *data);
+int
+dht_dir_attr_heal_done (int ret, call_frame_t *sync_frame, void *data);
+int
+dht_dir_has_layout (dict_t *xattr, char *name);
+gf_boolean_t
+dht_is_subvol_in_layout (dht_layout_t *layout, xlator_t *xlator);
+xlator_t *
+dht_subvol_with_free_space_inodes (xlator_t *this, xlator_t *subvol,
+                                   dht_layout_t *layout);
+xlator_t *
+dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol,
+                                  dht_layout_t *layout);
+int
+dht_linkfile_attr_heal (call_frame_t *frame, xlator_t *this);
+
+void
+dht_layout_dump (dht_layout_t  *layout, const char *prefix);
+int32_t
+dht_priv_dump (xlator_t *this);
+int32_t
+dht_inodectx_dump (xlator_t *this, inode_t *inode);
+
+int
+dht_inode_ctx_get1 (xlator_t *this, inode_t *inode, xlator_t **subvol);
 
-#endif /* _DHT_H */
+#endif/* _DHT_H */
diff --git a/xlators/cluster/dht/src/dht-diskusage.c b/xlators/cluster/dht/src/dht-diskusage.c
index ee009d341..fe3955ecb 100644
--- a/xlators/cluster/dht/src/dht-diskusage.c
+++ b/xlators/cluster/dht/src/dht-diskusage.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -33,50 +24,70 @@
 #include <sys/time.h>
 
 
-int 
+int
 dht_du_info_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int op_ret, int op_errno, struct statvfs *statvfs)
+		 int op_ret, int op_errno, struct statvfs *statvfs,
+                 dict_t *xdata)
 {
 	dht_conf_t    *conf         = NULL;
-        call_frame_t  *prev          = NULL;
+	call_frame_t  *prev          = NULL;
 	int            this_call_cnt = 0;
-        int            i = 0;
-        double         percent = 0;
-        uint64_t       bytes = 0;
-
-        conf = this->private;
-        prev = cookie;
-
-        if (op_ret == -1) 
-                goto out;
+	int            i = 0;
+	double         percent = 0;
+	double         percent_inodes = 0;
+	uint64_t       bytes = 0;
 
-        if (statvfs && statvfs->f_blocks) {
-                percent = (statvfs->f_bfree * 100) / statvfs->f_blocks;
-                bytes = (statvfs->f_bfree * statvfs->f_bsize);
-        }
-        
-        LOCK (&conf->subvolume_lock);
-        {
-                for (i = 0; i < conf->subvolume_cnt; i++)
-                        if (prev->this == conf->subvolumes[i]) {
-                                conf->du_stats[i].avail_percent = percent;
-                                conf->du_stats[i].avail_space   = bytes;
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "on subvolume '%s': avail_percent is: "
-                                        "%.2f and avail_space is: %"PRIu64"",
-                                        prev->this->name, 
-                                        conf->du_stats[i].avail_percent,
-                                        conf->du_stats[i].avail_space);
-                        }
-        }
-        UNLOCK (&conf->subvolume_lock);
-
- out:
+	conf = this->private;
+	prev = cookie;
+
+	if (op_ret == -1) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"failed to get disk info from %s", prev->this->name);
+		goto out;
+	}
+
+	if (statvfs && statvfs->f_blocks) {
+		percent = (statvfs->f_bavail * 100) / statvfs->f_blocks;
+		bytes = (statvfs->f_bavail * statvfs->f_frsize);
+	}
+
+	if (statvfs && statvfs->f_files) {
+		percent_inodes = (statvfs->f_ffree * 100) / statvfs->f_files;
+	} else {
+		/* set percent inodes to 100 for dynamically allocated inode filesystems
+		   this logic holds good so that, distribute has nothing to worry about
+		   total inodes rather let the 'create()' to be scheduled on the hashed
+		   subvol regardless of the total inodes. since we have no awareness on
+		   loosing inodes this logic fits well
+		*/
+		percent_inodes = 100;
+	}
+
+	LOCK (&conf->subvolume_lock);
+	{
+		for (i = 0; i < conf->subvolume_cnt; i++)
+			if (prev->this == conf->subvolumes[i]) {
+				conf->du_stats[i].avail_percent = percent;
+				conf->du_stats[i].avail_space   = bytes;
+				conf->du_stats[i].avail_inodes  = percent_inodes;
+				gf_log (this->name, GF_LOG_DEBUG,
+					"on subvolume '%s': avail_percent is: "
+					"%.2f and avail_space is: %"PRIu64" "
+					"and avail_inodes is: %.2f",
+					prev->this->name,
+					conf->du_stats[i].avail_percent,
+					conf->du_stats[i].avail_space,
+					conf->du_stats[i].avail_inodes);
+			}
+	}
+	UNLOCK (&conf->subvolume_lock);
+
+out:
 	this_call_cnt = dht_frame_return (frame);
 	if (is_last_call (this_call_cnt))
 		DHT_STACK_DESTROY (frame);
 
-        return 0;
+	return 0;
 }
 
 int
@@ -85,179 +96,319 @@ dht_get_du_info_for_subvol (xlator_t *this, int subvol_idx)
 	dht_conf_t    *conf         = NULL;
 	call_frame_t  *statfs_frame = NULL;
 	dht_local_t   *statfs_local = NULL;
-        call_pool_t   *pool         = NULL;
+	call_pool_t   *pool         = NULL;
+	loc_t          tmp_loc      = {0,};
 
 	conf = this->private;
-        pool = this->ctx->pool;
-
-        statfs_frame = create_frame (this, pool);
-        if (!statfs_frame) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
-        
-        statfs_local = dht_local_init (statfs_frame);
-        if (!statfs_local) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
-        
-        loc_t tmp_loc = { .inode = NULL,
-                          .path = "/",
-        };
-        
-        statfs_local->call_cnt = 1;
-        STACK_WIND (statfs_frame, dht_du_info_cbk,
-                    conf->subvolumes[subvol_idx],
-                    conf->subvolumes[subvol_idx]->fops->statfs,
-                    &tmp_loc);
-
-        return 0;
- err:
+	pool = this->ctx->pool;
+
+	statfs_frame = create_frame (this, pool);
+	if (!statfs_frame) {
+		goto err;
+	}
+
+	/* local->fop value is not used in this case */
+	statfs_local = dht_local_init (statfs_frame, NULL, NULL,
+				       GF_FOP_MAXVALUE);
+	if (!statfs_local) {
+		goto err;
+	}
+
+        /* make it root gfid, should be enough to get the proper info back */
+        tmp_loc.gfid[15] = 1;
+
+	statfs_local->call_cnt = 1;
+	STACK_WIND (statfs_frame, dht_du_info_cbk,
+		    conf->subvolumes[subvol_idx],
+		    conf->subvolumes[subvol_idx]->fops->statfs,
+		    &tmp_loc, NULL);
+
+	return 0;
+err:
 	if (statfs_frame)
 		DHT_STACK_DESTROY (statfs_frame);
-        
-        return -1;
+
+	return -1;
 }
 
 int
 dht_get_du_info (call_frame_t *frame, xlator_t *this, loc_t *loc)
 {
-        int            i = 0;
+	int            i            = 0;
 	dht_conf_t    *conf         = NULL;
 	call_frame_t  *statfs_frame = NULL;
 	dht_local_t   *statfs_local = NULL;
-        struct timeval tv = {0,};
+	struct timeval tv           = {0,};
+        loc_t          tmp_loc      = {0,};
 
 	conf  = this->private;
 
 	gettimeofday (&tv, NULL);
-	if (tv.tv_sec > (conf->refresh_interval 
-			 + conf->last_stat_fetch.tv_sec)) {
 
-                statfs_frame = copy_frame (frame);
-                if (!statfs_frame) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        goto err;
-                }
+        /* make it root gfid, should be enough to get the proper
+           info back */
+        tmp_loc.gfid[15] = 1;
 
-                statfs_local = dht_local_init (statfs_frame);
-                if (!statfs_local) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        goto err;
-                }
-
-                loc_copy (&statfs_local->loc, loc);
-                loc_t tmp_loc = { .inode = NULL,
-                              .path = "/",
-                };
-                
-                statfs_local->call_cnt = conf->subvolume_cnt;
-                for (i = 0; i < conf->subvolume_cnt; i++) {
-                        STACK_WIND (statfs_frame, dht_du_info_cbk,
-                                    conf->subvolumes[i],
-                                    conf->subvolumes[i]->fops->statfs,
-                                    &tmp_loc);
-                }
+	if (tv.tv_sec > (conf->refresh_interval
+			 + conf->last_stat_fetch.tv_sec)) {
 
-                conf->last_stat_fetch.tv_sec = tv.tv_sec;
-        }
-        return 0;
+		statfs_frame = copy_frame (frame);
+		if (!statfs_frame) {
+			goto err;
+		}
+
+		/* In this case, 'local->fop' is not used */
+		statfs_local = dht_local_init (statfs_frame, loc, NULL,
+					       GF_FOP_MAXVALUE);
+		if (!statfs_local) {
+			goto err;
+		}
+
+		statfs_local->call_cnt = conf->subvolume_cnt;
+		for (i = 0; i < conf->subvolume_cnt; i++) {
+			STACK_WIND (statfs_frame, dht_du_info_cbk,
+				    conf->subvolumes[i],
+				    conf->subvolumes[i]->fops->statfs,
+				    &tmp_loc, NULL);
+		}
+
+		conf->last_stat_fetch.tv_sec = tv.tv_sec;
+	}
+	return 0;
 err:
 	if (statfs_frame)
 		DHT_STACK_DESTROY (statfs_frame);
 
-        return -1;
+	return -1;
 }
 
 
-int
+gf_boolean_t
 dht_is_subvol_filled (xlator_t *this, xlator_t *subvol)
 {
-        int         i = 0;
-        int         subvol_filled = 0;
+	int         i = 0;
 	dht_conf_t *conf = NULL;
+	gf_boolean_t subvol_filled_inodes = _gf_false;
+	gf_boolean_t subvol_filled_space = _gf_false;
+	gf_boolean_t is_subvol_filled = _gf_false;
 
-        conf = this->private;
+	conf = this->private;
 
-        /* Check for values above specified percent or free disk */
-        LOCK (&conf->subvolume_lock);
-        {
-                for (i = 0; i < conf->subvolume_cnt; i++) {
-                        if (subvol == conf->subvolumes[i]) {
-                                if (conf->disk_unit == 'p') {
-                                        if (conf->du_stats[i].avail_percent <
-                                            conf->min_free_disk) {
-                                                subvol_filled = 1;
-                                                break;
-                                        }
-                                } else {
-                                        if (conf->du_stats[i].avail_space <
-                                            conf->min_free_disk) {
-                                                subvol_filled = 1;
-                                                break;
-                                        }
-                                }
-                        }
+	/* Check for values above specified percent or free disk */
+	LOCK (&conf->subvolume_lock);
+	{
+		for (i = 0; i < conf->subvolume_cnt; i++) {
+			if (subvol == conf->subvolumes[i]) {
+				if (conf->disk_unit == 'p') {
+					if (conf->du_stats[i].avail_percent <
+					    conf->min_free_disk) {
+						subvol_filled_space = _gf_true;
+						break;
+					}
+
+				} else {
+					if (conf->du_stats[i].avail_space <
+					    conf->min_free_disk) {
+						subvol_filled_space = _gf_true;
+						break;
+					}
+				}
+				if (conf->du_stats[i].avail_inodes <
+				    conf->min_free_inodes) {
+					subvol_filled_inodes = _gf_true;
+					break;
+				}
+			}
+		}
+	}
+	UNLOCK (&conf->subvolume_lock);
+
+	if (subvol_filled_space && conf->subvolume_status[i]) {
+		if (!(conf->du_stats[i].log++ % (GF_UNIVERSAL_ANSWER * 10))) {
+			gf_log (this->name, GF_LOG_WARNING,
+				"disk space on subvolume '%s' is getting "
+				"full (%.2f %%), consider adding more nodes",
+				subvol->name,
+				(100 - conf->du_stats[i].avail_percent));
+		}
+	}
+
+	if (subvol_filled_inodes && conf->subvolume_status[i]) {
+		if (!(conf->du_stats[i].log++ % (GF_UNIVERSAL_ANSWER * 10))) {
+			gf_log (this->name, GF_LOG_CRITICAL,
+				"inodes on subvolume '%s' are at "
+				"(%.2f %%), consider adding more nodes",
+				subvol->name,
+				(100 - conf->du_stats[i].avail_inodes));
+		}
+	}
+
+	is_subvol_filled = (subvol_filled_space || subvol_filled_inodes);
+
+	return is_subvol_filled;
+}
+
+
+/*Get the best subvolume to create the file in*/
+xlator_t *
+dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol,
+                                dht_local_t *local)
+{
+	xlator_t   *avail_subvol = NULL;
+	dht_conf_t *conf = NULL;
+        dht_layout_t *layout = NULL;
+        loc_t      *loc = NULL;
+
+	conf = this->private;
+        if (!local)
+                goto out;
+        loc = &local->loc;
+        if (!local->layout) {
+                layout = dht_layout_get (this, loc->parent);
+
+                if (!layout) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "layout missing path=%s parent=%s",
+                                loc->path, uuid_utoa (loc->parent->gfid));
+                        goto out;
                 }
+        } else {
+                layout = dht_layout_ref (this, local->layout);
         }
-        UNLOCK (&conf->subvolume_lock);
-
-        if (subvol_filled && conf->subvolume_status[i]) {
-                if (!(conf->du_stats[i].log++ % (GF_UNIVERSAL_ANSWER * 10))) {
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "disk space on subvolume '%s' is getting "
-                                "full (%.2f %%), consider adding more nodes", 
-                                subvol->name, 
-                                (100 - conf->du_stats[i].avail_percent));
+
+        LOCK (&conf->subvolume_lock);
+	{
+                avail_subvol = dht_subvol_with_free_space_inodes(this, subvol,
+                                                                 layout);
+                if(!avail_subvol)
+                {
+                        avail_subvol = dht_subvol_maxspace_nonzeroinode(this,
+                                                                        subvol,
+                                                                        layout);
                 }
-        }
 
-        return subvol_filled;
+	}
+	UNLOCK (&conf->subvolume_lock);
+out:
+	if (!avail_subvol) {
+		gf_log (this->name,
+                        GF_LOG_DEBUG,
+			"no subvolume has enough free space and/or inodes\
+                         to create");
+                avail_subvol = subvol;
+	}
+
+        if (layout)
+                dht_layout_unref (this, layout);
+	return avail_subvol;
 }
 
+static inline
+int32_t dht_subvol_has_err (xlator_t *this, dht_layout_t *layout)
+{
+        int ret = -1;
+        int i   = 0;
+
+        if (!this || !layout)
+                goto out;
+
+        /* check if subvol has layout errors, before selecting it */
+        for (i = 0; i < layout->cnt; i++) {
+                if (!strcmp (layout->list[i].xlator->name, this->name) &&
+                     (layout->list[i].err != 0)) {
+                        ret = -1;
+                        goto out;
+                }
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+/*Get subvolume which has both space and inodes more than the min criteria*/
 xlator_t *
-dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol) 
+dht_subvol_with_free_space_inodes(xlator_t *this, xlator_t *subvol,
+                                  dht_layout_t *layout)
 {
-        int         i = 0;
-        double      max= 0;
-        xlator_t   *avail_subvol = NULL;
-	dht_conf_t *conf = NULL;
+        int i = 0;
+        double max = 0;
+        double max_inodes = 0;
+        int    ignore_subvol = 0;
+
+        xlator_t *avail_subvol = NULL;
+        dht_conf_t *conf = NULL;
 
         conf = this->private;
 
-        LOCK (&conf->subvolume_lock);
-        {
-                for (i = 0; i < conf->subvolume_cnt; i++) {
-                        if (conf->disk_unit == 'p') {
-                                if (conf->du_stats[i].avail_percent > max) {
-                                        max = conf->du_stats[i].avail_percent;
-                                        avail_subvol = conf->subvolumes[i];
-                                }
-                        } else {
-                                if (conf->du_stats[i].avail_space > max) {
-                                        max = conf->du_stats[i].avail_space;
-                                        avail_subvol = conf->subvolumes[i];
-                                }
+        for(i=0; i < conf->subvolume_cnt; i++) {
+                /* check if subvol has layout errors, before selecting it */
+                ignore_subvol = dht_subvol_has_err (conf->subvolumes[i],
+                                                    layout);
+                if (ignore_subvol)
+                        continue;
+
+                if ((conf->disk_unit == 'p') &&
+                    (conf->du_stats[i].avail_percent > conf->min_free_disk) &&
+                    (conf->du_stats[i].avail_inodes  > conf->min_free_inodes)) {
+                        if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+                            (conf->du_stats[i].avail_percent > max)) {
+                                max = conf->du_stats[i].avail_percent;
+                                max_inodes = conf->du_stats[i].avail_inodes;
+                                avail_subvol = conf->subvolumes[i];
                         }
                 }
-        }
-        UNLOCK (&conf->subvolume_lock);
 
-        if (!avail_subvol) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "no subvolume has enough free space to create");
+                if ((conf->disk_unit != 'p') &&
+                    (conf->du_stats[i].avail_space > conf->min_free_disk) &&
+                    (conf->du_stats[i].avail_inodes  > conf->min_free_inodes)) {
+                        if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+                            (conf->du_stats[i].avail_space > max)) {
+                                max = conf->du_stats[i].avail_space;
+                                max_inodes = conf->du_stats[i].avail_inodes;
+                                avail_subvol = conf->subvolumes[i];
+                        }
+                }
         }
 
-        if (max < conf->min_free_disk)
-                avail_subvol = subvol;
+        return avail_subvol;
+}
+
+
+/* Get subvol which has atleast one inode and maximum space */
+xlator_t *
+dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol,
+                                  dht_layout_t *layout)
+{
+        int         i = 0;
+        double      max = 0;
+        int         ignore_subvol = 0;
+
+        xlator_t   *avail_subvol = NULL;
+        dht_conf_t *conf = NULL;
 
-        if (!avail_subvol)
-                avail_subvol = subvol;                
+        conf = this->private;
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                /* check if subvol has layout errors, before selecting it */
+                ignore_subvol = dht_subvol_has_err (conf->subvolumes[i],
+                                                    layout);
+                if (ignore_subvol)
+                        continue;
+
+                if (conf->disk_unit == 'p') {
+                        if ((conf->du_stats[i].avail_percent > max)
+                            && (conf->du_stats[i].avail_inodes > 0 )) {
+                                max = conf->du_stats[i].avail_percent;
+                                avail_subvol = conf->subvolumes[i];
+                        }
+               } else {
+                         if ((conf->du_stats[i].avail_space > max)
+                            && (conf->du_stats[i].avail_inodes > 0)) {
+                                 max = conf->du_stats[i].avail_space;
+                                 avail_subvol = conf->subvolumes[i];
+                         }
+               }
+        }
 
         return avail_subvol;
 }
diff --git a/xlators/cluster/dht/src/dht-hashfn.c b/xlators/cluster/dht/src/dht-hashfn.c
index dfc1541fa..656cf23a0 100644
--- a/xlators/cluster/dht/src/dht-hashfn.c
+++ b/xlators/cluster/dht/src/dht-hashfn.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -32,50 +23,89 @@
 int
 dht_hash_compute_internal (int type, const char *name, uint32_t *hash_p)
 {
-	int      ret = 0;
-	uint32_t hash = 0;
-
-	switch (type) {
-	case DHT_HASH_TYPE_DM:
-		hash = gf_dm_hashfn (name, strlen (name));
-		break;
-	default:
-		ret = -1;
-		break;
-	}
-
-	if (ret == 0) {
-		*hash_p = hash;
-	}
-
-	return ret;
+        int      ret = 0;
+        uint32_t hash = 0;
+
+        switch (type) {
+        case DHT_HASH_TYPE_DM:
+        case DHT_HASH_TYPE_DM_USER:
+                hash = gf_dm_hashfn (name, strlen (name));
+                break;
+        default:
+                ret = -1;
+                break;
+        }
+
+        if (ret == 0) {
+                *hash_p = hash;
+        }
+
+        return ret;
 }
 
 
-#define MAKE_RSYNC_FRIENDLY_NAME(rsync_frndly_name, name) do {          \
-                rsync_frndly_name = (char *) name;			\
-                if (name[0] == '.') {                                   \
-                        char *dot   = 0;                                \
-                        int namelen = 0;                                \
-                                                                        \
-                        dot = strrchr (name, '.');                      \
-                        if (dot && dot > (name + 1) && *(dot + 1)) {    \
-                                namelen = (dot - name);                 \
-                                rsync_frndly_name = alloca (namelen);   \
-                                strncpy (rsync_frndly_name, name + 1,   \
-                                         namelen);                      \
-                                rsync_frndly_name[namelen - 1] = 0;     \
-                        }                                               \
-                }                                                       \
-        } while (0);
-
+static inline
+gf_boolean_t
+dht_munge_name (const char *original, char *modified, size_t len, regex_t *re)
+{
+        regmatch_t      matches[2];
+        size_t          new_len;
+
+        if (regexec(re,original,2,matches,0) != REG_NOMATCH) {
+                if (matches[1].rm_so != -1) {
+                        new_len = matches[1].rm_eo - matches[1].rm_so;
+                        /* Equal would fail due to the NUL at the end. */
+                        if (new_len < len) {
+                                memcpy (modified,original+matches[1].rm_so,
+                                        new_len);
+                                modified[new_len] = '\0';
+                                return _gf_true;
+                        }
+                }
+        }
+
+        /* This is guaranteed safe because of how the dest was allocated. */
+        strcpy(modified,original);
+        return _gf_false;
+}
 
 int
-dht_hash_compute (int type, const char *name, uint32_t *hash_p)
+dht_hash_compute (xlator_t *this, int type, const char *name, uint32_t *hash_p)
 {
-	char     *rsync_friendly_name = NULL;
-
-	MAKE_RSYNC_FRIENDLY_NAME (rsync_friendly_name, name);
-
-	return dht_hash_compute_internal (type, rsync_friendly_name, hash_p);
+        char            *rsync_friendly_name    = NULL;
+        dht_conf_t      *priv                   = this->private;
+        size_t           len                    = 0;
+        gf_boolean_t     munged                 = _gf_false;
+
+        /*
+         * It wouldn't be safe to use alloca in an inline function that doesn't
+         * actually get inlined, and it wouldn't be efficient to do a real
+         * allocation, so we use alloca here (if needed) and pass that to the
+         * inline.
+         */
+
+        if (priv->extra_regex_valid) {
+                len = strlen(name) + 1;
+                rsync_friendly_name = alloca(len);
+                munged = dht_munge_name (name, rsync_friendly_name, len,
+                                         &priv->extra_regex);
+        }
+
+        if (!munged && priv->rsync_regex_valid) {
+                len = strlen(name) + 1;
+                rsync_friendly_name = alloca(len);
+                gf_log (this->name, GF_LOG_TRACE, "trying regex for %s", name);
+                munged = dht_munge_name (name, rsync_friendly_name, len,
+                                         &priv->rsync_regex);
+                if (munged) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "munged down to %s", rsync_friendly_name);
+                }
+        }
+
+        if (!munged) {
+                rsync_friendly_name = (char *)name;
+        }
+
+        return dht_hash_compute_internal (type, rsync_friendly_name, hash_p);
 }
diff --git a/xlators/cluster/dht/src/dht-helper.c b/xlators/cluster/dht/src/dht-helper.c
index 147cd0fe7..311a48112 100644
--- a/xlators/cluster/dht/src/dht-helper.c
+++ b/xlators/cluster/dht/src/dht-helper.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -27,54 +18,133 @@
 #include "xlator.h"
 #include "dht-common.h"
 
+static inline int
+dht_inode_ctx_set1 (xlator_t *this, inode_t *inode, xlator_t *subvol)
+{
+        uint64_t tmp_subvol = 0;
+
+        tmp_subvol = (long)subvol;
+        return inode_ctx_set1 (inode, this, &tmp_subvol);
+}
+
+int
+dht_inode_ctx_get1 (xlator_t *this, inode_t *inode, xlator_t **subvol)
+{
+        int ret = -1;
+        uint64_t tmp_subvol = 0;
+
+        ret =  inode_ctx_get1 (inode, this, &tmp_subvol);
+        if (tmp_subvol && subvol)
+                *subvol = (xlator_t *)tmp_subvol;
+
+        return ret;
+}
+
 
 int
 dht_frame_return (call_frame_t *frame)
 {
-	dht_local_t *local = NULL;
-	int          this_call_cnt = -1;
+        dht_local_t *local = NULL;
+        int          this_call_cnt = -1;
+
+        if (!frame)
+                return -1;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                this_call_cnt = --local->call_cnt;
+        }
+        UNLOCK (&frame->lock);
 
-	if (!frame)
-		return -1;
+        return this_call_cnt;
+}
 
-	local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		this_call_cnt = --local->call_cnt;
+static uint64_t
+dht_bits_for (uint64_t num)
+{
+	uint64_t bits = 0, ctrl = 1;
+
+	while (ctrl < num) {
+		ctrl *= 2;
+		bits ++;
 	}
-	UNLOCK (&frame->lock);
 
-	return this_call_cnt;
+	return bits;
 }
 
+/*
+ * A slightly "updated" version of the algorithm described in the commit log
+ * is used here.
+ *
+ * The only enhancement is that:
+ *
+ * - The number of bits used by the backend filesystem for HUGE d_off which
+ *   is described as 63, and
+ * - The number of bits used by the d_off presented by the transformation
+ *   upwards which is described as 64, are both made "configurable."
+ */
+
+
+#define BACKEND_D_OFF_BITS 63
+#define PRESENT_D_OFF_BITS 63
+
+#define ONE 1ULL
+#define MASK (~0ULL)
+#define PRESENT_MASK (MASK >> (64 - PRESENT_D_OFF_BITS))
+#define BACKEND_MASK (MASK >> (64 - BACKEND_D_OFF_BITS))
+
+#define TOP_BIT (ONE << (PRESENT_D_OFF_BITS - 1))
+#define SHIFT_BITS (max (0, (BACKEND_D_OFF_BITS - PRESENT_D_OFF_BITS + 1)))
 
 int
 dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y_p)
 {
-	dht_conf_t *conf = NULL;
-	int         cnt = 0;
-	int         max = 0;
-	uint64_t    y = 0;
+        dht_conf_t *conf = NULL;
+        int         cnt = 0;
+        int         max = 0;
+        uint64_t    y = 0;
+        uint64_t    hi_mask = 0;
+        uint64_t    off_mask = 0;
+        int         max_bits = 0;
+
+        if (x == ((uint64_t) -1)) {
+                y = (uint64_t) -1;
+                goto out;
+        }
+
+        conf = this->private;
+        if (!conf)
+                goto out;
 
+        max = conf->subvolume_cnt;
+        cnt = dht_subvol_cnt (this, subvol);
 
-	if (x == ((uint64_t) -1)) {
-		y = (uint64_t) -1;
+	if (max == 1) {
+		y = x;
 		goto out;
 	}
 
-	conf = this->private;
+        max_bits = dht_bits_for (max);
 
-	max = conf->subvolume_cnt;
-	cnt = dht_subvol_cnt (this, subvol);
+        hi_mask = ~(PRESENT_MASK >> (max_bits + 1));
 
-	y = ((x * max) + cnt);
+        if (x & hi_mask) {
+                /* HUGE d_off */
+                off_mask = MASK << max_bits;
+                y = TOP_BIT | ((x >> SHIFT_BITS) & off_mask) | cnt;
+        } else {
+                /* small d_off */
+                y = ((x * max) + cnt);
+        }
 
 out:
-	if (y_p)
-		*y_p = y;
+        if (y_p)
+                *y_p = y;
 
-	return 0;
+        return 0;
 }
 
 int
@@ -88,7 +158,7 @@ dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
         int            ret       = 0; /* not found */
 
         /* Why do other tasks if first required 'char' itself is not there */
-        if (loc->name && !strchr (loc->name, '@'))
+        if (!new_loc || !loc || !loc->name || !strchr (loc->name, '@'))
                 goto out;
 
         trav = this->children;
@@ -116,7 +186,6 @@ dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
                                 new_loc->path   = ((new_path) ? new_path:
                                                    gf_strdup (loc->path));
                                 new_loc->name   = new_name;
-                                new_loc->ino    = loc->ino;
                                 new_loc->inode  = inode_ref (loc->inode);
                                 new_loc->parent = inode_ref (loc->parent);
                         }
@@ -129,152 +198,200 @@ dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
 out:
         if (!ret) {
                 /* !success */
-                if (new_path)
-                        GF_FREE (new_path);
-                if (new_name)
-                        GF_FREE (new_name);
+                GF_FREE (new_path);
+                GF_FREE (new_name);
         }
         return ret;
 }
 
 int
 dht_deitransform (xlator_t *this, uint64_t y, xlator_t **subvol_p,
-		  uint64_t *x_p)
+                  uint64_t *x_p)
 {
-	dht_conf_t *conf = NULL;
-	int         cnt = 0;
-	int         max = 0;
-	uint64_t    x = 0;
-	xlator_t   *subvol = 0;
+        dht_conf_t *conf = NULL;
+        int         cnt = 0;
+        int         max = 0;
+        uint64_t    x = 0;
+        xlator_t   *subvol = 0;
+        int         max_bits = 0;
+        uint64_t    off_mask = 0;
+        uint64_t    host_mask = 0;
+
+        if (!this->private)
+                return -1;
+
+        conf = this->private;
+        max = conf->subvolume_cnt;
 
+	if (max == 1) {
+		x = y;
+		cnt = 0;
+		goto out;
+	}
 
-	conf = this->private;
-	max = conf->subvolume_cnt;
+        if (y & TOP_BIT) {
+                /* HUGE d_off */
+                max_bits = dht_bits_for (max);
+                off_mask = (MASK << max_bits);
+                host_mask = ~(off_mask);
 
-	cnt = y % max;
-	x   = y / max;
+                x = ((y & ~TOP_BIT) & off_mask) << SHIFT_BITS;
 
-	subvol = conf->subvolumes[cnt];
+                cnt = y & host_mask;
+	} else {
+                /* small d_off */
+                cnt = y % max;
+                x = y / max;
+        }
 
-	if (subvol_p)
-		*subvol_p = subvol;
+out:
+        subvol = conf->subvolumes[cnt];
 
-	if (x_p)
-		*x_p = x;
+        if (subvol_p)
+                *subvol_p = subvol;
 
-	return 0;
+        if (x_p)
+                *x_p = x;
+
+        return 0;
 }
 
 
 void
 dht_local_wipe (xlator_t *this, dht_local_t *local)
 {
-	if (!local)
-		return;
+        if (!local)
+                return;
 
-	loc_wipe (&local->loc);
-	loc_wipe (&local->loc2);
+        loc_wipe (&local->loc);
+        loc_wipe (&local->loc2);
 
-	if (local->xattr)
-		dict_unref (local->xattr);
+        if (local->xattr)
+                dict_unref (local->xattr);
 
-	if (local->inode)
-		inode_unref (local->inode);
+        if (local->inode)
+                inode_unref (local->inode);
 
-	if (local->layout) {
-		dht_layout_unref (this, local->layout);
+        if (local->layout) {
+                dht_layout_unref (this, local->layout);
                 local->layout = NULL;
         }
 
-	loc_wipe (&local->linkfile.loc);
+        loc_wipe (&local->linkfile.loc);
 
-	if (local->linkfile.xattr)
-		dict_unref (local->linkfile.xattr);
+        if (local->linkfile.xattr)
+                dict_unref (local->linkfile.xattr);
 
-	if (local->linkfile.inode)
-		inode_unref (local->linkfile.inode);
+        if (local->linkfile.inode)
+                inode_unref (local->linkfile.inode);
 
-	if (local->fd) {
-		fd_unref (local->fd);
-		local->fd = NULL;
-	}
+        if (local->fd) {
+                fd_unref (local->fd);
+                local->fd = NULL;
+        }
 
         if (local->params) {
                 dict_unref (local->params);
                 local->params = NULL;
         }
 
-	if (local->xattr_req)
-		dict_unref (local->xattr_req);
+        if (local->xattr_req)
+                dict_unref (local->xattr_req);
 
         if (local->selfheal.layout) {
                 dht_layout_unref (this, local->selfheal.layout);
                 local->selfheal.layout = NULL;
         }
 
-        if (local->newpath) {
-                GF_FREE (local->newpath);
-        }
+        GF_FREE (local->newpath);
+
+        GF_FREE (local->key);
 
-	GF_FREE (local);
+        GF_FREE (local->rebalance.vector);
+
+        if (local->rebalance.iobref)
+                iobref_unref (local->rebalance.iobref);
+
+        mem_put (local);
 }
 
 
 dht_local_t *
-dht_local_init (call_frame_t *frame)
+dht_local_init (call_frame_t *frame, loc_t *loc, fd_t *fd, glusterfs_fop_t fop)
 {
-	dht_local_t *local = NULL;
-
-	/* TODO: use mem-pool */
-	local = GF_CALLOC (1, sizeof (*local),
-                           gf_dht_mt_dht_local_t);
+        dht_local_t *local = NULL;
+        inode_t     *inode = NULL;
+        int          ret   = 0;
 
-	if (!local)
-		return NULL;
+        local = mem_get0 (THIS->local_pool);
+        if (!local)
+                goto out;
 
-	local->op_ret = -1;
-	local->op_errno = EUCLEAN;
+        if (loc) {
+                ret = loc_copy (&local->loc, loc);
+                if (ret)
+                        goto out;
 
-	frame->local = local;
+                inode = loc->inode;
+        }
 
-	return local;
-}
+        if (fd) {
+                local->fd = fd_ref (fd);
+                if (!inode)
+                        inode = fd->inode;
+        }
 
+        local->op_ret   = -1;
+        local->op_errno = EUCLEAN;
+        local->fop      = fop;
 
-char *
-basestr (const char *str)
-{
-        char *basestr = NULL;
+        if (inode) {
+                local->layout   = dht_layout_get (frame->this, inode);
+                local->cached_subvol = dht_subvol_get_cached (frame->this,
+                                                              inode);
+        }
 
-        basestr = strrchr (str, '/');
-        if (basestr)
-                basestr ++;
+        frame->local = local;
 
-        return basestr;
+out:
+        if (ret) {
+                if (local)
+                        mem_put (local);
+                local = NULL;
+        }
+        return local;
 }
 
-
 xlator_t *
 dht_first_up_subvol (xlator_t *this)
 {
-	dht_conf_t *conf = NULL;
-	xlator_t   *child = NULL;
-	int         i = 0;
-
-	conf = this->private;
-
-	LOCK (&conf->subvolume_lock);
-	{
-		for (i = 0; i < conf->subvolume_cnt; i++) {
-			if (conf->subvolume_status[i]) {
-				child = conf->subvolumes[i];
-				break;
-			}
-		}
-	}
-	UNLOCK (&conf->subvolume_lock);
+        dht_conf_t *conf = NULL;
+        xlator_t   *child = NULL;
+        int         i = 0;
+        time_t      time = 0;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        LOCK (&conf->subvolume_lock);
+        {
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (conf->subvol_up_time[i]) {
+                                if (!time) {
+                                        time = conf->subvol_up_time[i];
+                                        child = conf->subvolumes[i];
+                                } else if (time > conf->subvol_up_time[i]) {
+                                        time  = conf->subvol_up_time[i];
+                                        child = conf->subvolumes[i];
+                                }
+                        }
+                }
+        }
+        UNLOCK (&conf->subvolume_lock);
 
-	return child;
+out:
+        return child;
 }
 
 xlator_t *
@@ -285,6 +402,9 @@ dht_last_up_subvol (xlator_t *this)
         int         i = 0;
 
         conf = this->private;
+        if (!conf)
+                goto out;
+
         LOCK (&conf->subvolume_lock);
         {
                 for (i = conf->subvolume_cnt-1; i >= 0; i--) {
@@ -296,6 +416,7 @@ dht_last_up_subvol (xlator_t *this)
         }
         UNLOCK (&conf->subvolume_lock);
 
+out:
         return child;
 }
 
@@ -305,17 +426,23 @@ dht_subvol_get_hashed (xlator_t *this, loc_t *loc)
         dht_layout_t *layout = NULL;
         xlator_t     *subvol = NULL;
 
-        if (is_fs_root (loc)) {
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, loc, out);
+
+        if (__is_root_gfid (loc->gfid)) {
                 subvol = dht_first_up_subvol (this);
                 goto out;
         }
 
+        GF_VALIDATE_OR_GOTO (this->name, loc->parent, out);
+        GF_VALIDATE_OR_GOTO (this->name, loc->name, out);
+
         layout = dht_layout_get (this, loc->parent);
 
         if (!layout) {
                 gf_log (this->name, GF_LOG_DEBUG,
-                        "layout missing path=%s parent=%"PRId64,
-                        loc->path, loc->parent->ino);
+                        "layout missing path=%s parent=%s",
+                        loc->path, uuid_utoa (loc->parent->gfid));
                 goto out;
         }
 
@@ -343,6 +470,8 @@ dht_subvol_get_cached (xlator_t *this, inode_t *inode)
         dht_layout_t *layout = NULL;
         xlator_t     *subvol = NULL;
 
+        GF_VALIDATE_OR_GOTO (this->name, this, out);
+        GF_VALIDATE_OR_GOTO (this->name, inode, out);
 
         layout = dht_layout_get (this, inode);
 
@@ -350,7 +479,7 @@ dht_subvol_get_cached (xlator_t *this, inode_t *inode)
                 goto out;
         }
 
-	subvol = layout->list[0].xlator;
+        subvol = layout->list[0].xlator;
 
 out:
         if (layout) {
@@ -364,145 +493,655 @@ out:
 xlator_t *
 dht_subvol_next (xlator_t *this, xlator_t *prev)
 {
-	dht_conf_t *conf = NULL;
-	int         i = 0;
-	xlator_t   *next = NULL;
-
-	conf = this->private;
-
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (conf->subvolumes[i] == prev) {
-			if ((i + 1) < conf->subvolume_cnt)
-				next = conf->subvolumes[i + 1];
-			break;
-		}
-	}
+        dht_conf_t *conf = NULL;
+        int         i = 0;
+        xlator_t   *next = NULL;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
 
-	return next;
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == prev) {
+                        if ((i + 1) < conf->subvolume_cnt)
+                                next = conf->subvolumes[i + 1];
+                        break;
+                }
+        }
+
+out:
+        return next;
 }
 
+/* This func wraps around, if prev is actually the last subvol.
+ */
+xlator_t *
+dht_subvol_next_available (xlator_t *this, xlator_t *prev)
+{
+        dht_conf_t *conf = NULL;
+        int         i = 0;
+        xlator_t   *next = NULL;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
 
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == prev) {
+                        /* if prev is last in conf->subvolumes, then wrap
+                         * around.
+                         */
+                        if ((i + 1) < conf->subvolume_cnt) {
+                                next = conf->subvolumes[i + 1];
+                        } else {
+                                next = conf->subvolumes[0];
+                        }
+                        break;
+                }
+        }
+
+out:
+        return next;
+}
 int
 dht_subvol_cnt (xlator_t *this, xlator_t *subvol)
 {
-	int i = 0;
-	int ret = -1;
-	dht_conf_t *conf = NULL;
-
+        int i = 0;
+        int ret = -1;
+        dht_conf_t *conf = NULL;
 
-	conf = this->private;
+        conf = this->private;
+        if (!conf)
+                goto out;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (subvol == conf->subvolumes[i]) {
-			ret = i;
-			break;
-		}
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (subvol == conf->subvolumes[i]) {
+                        ret = i;
+                        break;
+                }
+        }
 
-	return ret;
+out:
+        return ret;
 }
 
 
-#define set_if_greater(a, b) do {		\
-		if ((a) < (b))			\
-			(a) = (b);		\
-	} while (0)
+#define set_if_greater(a, b) do {               \
+                if ((a) < (b))                  \
+                        (a) = (b);              \
+        } while (0)
+
+
+#define set_if_greater_time(a, an, b, bn) do {                          \
+                if (((a) < (b)) || (((a) == (b)) && ((an) < (bn)))){    \
+                        (a) = (b);                                      \
+                        (an) = (bn);                                    \
+                }                                                       \
+        } while (0)                                                     \
+
 
 int
 dht_iatt_merge (xlator_t *this, struct iatt *to,
-		struct iatt *from, xlator_t *subvol)
+                struct iatt *from, xlator_t *subvol)
 {
         if (!from || !to)
                 return 0;
 
-	to->ia_dev      = from->ia_dev;
+        to->ia_dev      = from->ia_dev;
 
         uuid_copy (to->ia_gfid, from->ia_gfid);
 
-	dht_itransform (this, subvol, from->ia_ino, &to->ia_ino);
+        to->ia_ino      = from->ia_ino;
+        to->ia_prot     = from->ia_prot;
+        to->ia_type     = from->ia_type;
+        to->ia_nlink    = from->ia_nlink;
+        to->ia_rdev     = from->ia_rdev;
+        to->ia_size    += from->ia_size;
+        to->ia_blksize  = from->ia_blksize;
+        to->ia_blocks  += from->ia_blocks;
+
+        set_if_greater (to->ia_uid, from->ia_uid);
+        set_if_greater (to->ia_gid, from->ia_gid);
+
+        set_if_greater_time(to->ia_atime, to->ia_atime_nsec,
+                            from->ia_atime, from->ia_atime_nsec);
+        set_if_greater_time (to->ia_mtime, to->ia_mtime_nsec,
+                             from->ia_mtime, from->ia_mtime_nsec);
+        set_if_greater_time (to->ia_ctime, to->ia_ctime_nsec,
+                             from->ia_ctime, from->ia_ctime_nsec);
+
+        return 0;
+}
+
+int
+dht_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
+{
+        if (!child) {
+                goto err;
+        }
+
+        if (strcmp (parent->path, "/") == 0)
+                gf_asprintf ((char **)&child->path, "/%s", name);
+        else
+                gf_asprintf ((char **)&child->path, "%s/%s", parent->path, name);
 
-	to->ia_prot     = from->ia_prot;
-	to->ia_type     = from->ia_type;
-	to->ia_nlink    = from->ia_nlink;
-	to->ia_rdev     = from->ia_rdev;
-	to->ia_size    += from->ia_size;
-	to->ia_blksize  = from->ia_blksize;
-	to->ia_blocks  += from->ia_blocks;
+        if (!child->path) {
+                goto err;
+        }
 
-	set_if_greater (to->ia_uid, from->ia_uid);
-	set_if_greater (to->ia_gid, from->ia_gid);
+        child->name = strrchr (child->path, '/');
+        if (child->name)
+                child->name++;
+
+        child->parent = inode_ref (parent->inode);
+        child->inode = inode_new (parent->inode->table);
 
-	set_if_greater (to->ia_atime, from->ia_atime);
-	set_if_greater (to->ia_mtime, from->ia_mtime);
-	set_if_greater (to->ia_ctime, from->ia_ctime);
+        if (!child->inode) {
+                goto err;
+        }
 
-	return 0;
+        return 0;
+err:
+        loc_wipe (child);
+        return -1;
 }
 
+
+
 int
-dht_frame_su_do (call_frame_t *frame)
+dht_init_subvolumes (xlator_t *this, dht_conf_t *conf)
 {
-        dht_local_t     *local = NULL;
+        xlator_list_t *subvols = NULL;
+        int            cnt = 0;
 
-        local = frame->local;
+        if (!conf)
+                return -1;
+
+        for (subvols = this->children; subvols; subvols = subvols->next)
+                cnt++;
+
+        conf->subvolumes = GF_CALLOC (cnt, sizeof (xlator_t *),
+                                      gf_dht_mt_xlator_t);
+        if (!conf->subvolumes) {
+                return -1;
+        }
+        conf->subvolume_cnt = cnt;
+
+        cnt = 0;
+        for (subvols = this->children; subvols; subvols = subvols->next)
+                conf->subvolumes[cnt++] = subvols->xlator;
 
-        local->uid = frame->root->uid;
-        local->gid = frame->root->gid;
+        conf->subvolume_status = GF_CALLOC (cnt, sizeof (char),
+                                            gf_dht_mt_char);
+        if (!conf->subvolume_status) {
+                return -1;
+        }
+
+        conf->last_event = GF_CALLOC (cnt, sizeof (int),
+                                      gf_dht_mt_char);
+        if (!conf->last_event) {
+                return -1;
+        }
+
+        conf->subvol_up_time = GF_CALLOC (cnt, sizeof (time_t),
+                                          gf_dht_mt_subvol_time);
+        if (!conf->subvol_up_time) {
+                return -1;
+        }
 
-        frame->root->uid = 0;
-        frame->root->gid = 0;
+        conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
+                                    gf_dht_mt_dht_du_t);
+        if (!conf->du_stats) {
+                return -1;
+        }
+
+        conf->decommissioned_bricks = GF_CALLOC (cnt, sizeof (xlator_t *),
+                                                 gf_dht_mt_xlator_t);
+        if (!conf->decommissioned_bricks) {
+                return -1;
+        }
 
         return 0;
 }
 
 
-int
-dht_frame_su_undo (call_frame_t *frame)
+
+
+static int
+dht_migration_complete_check_done (int op_ret, call_frame_t *frame, void *data)
 {
-        dht_local_t     *local = NULL;
+        dht_local_t *local = NULL;
 
         local = frame->local;
 
-        frame->root->uid = local->uid;
-        frame->root->gid = local->gid;
+        local->rebalance.target_op_fn (THIS, frame, op_ret);
 
         return 0;
 }
 
 
 int
-dht_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
+dht_migration_complete_check_task (void *data)
 {
-        if (!child) {
-                goto err;
+        int           ret      = -1;
+        xlator_t     *src_node = NULL;
+        xlator_t     *dst_node = NULL;
+        dht_local_t  *local    = NULL;
+        dict_t       *dict     = NULL;
+        dht_layout_t *layout   = NULL;
+        struct iatt   stbuf    = {0,};
+        xlator_t     *this     = NULL;
+        call_frame_t *frame    = NULL;
+        loc_t         tmp_loc  = {0,};
+        char         *path     = NULL;
+        dht_conf_t   *conf     = NULL;
+        inode_t      *inode    = NULL;
+        fd_t         *iter_fd  = NULL;
+        uint64_t      tmp_subvol = 0;
+        int           open_failed = 0;
+
+        this  = THIS;
+        frame = data;
+        local = frame->local;
+        conf = this->private;
+
+        src_node = local->cached_subvol;
+
+        if (!local->loc.inode && !local->fd)
+                goto out;
+
+        inode = (!local->fd) ? local->loc.inode : local->fd->inode;
+
+        /* getxattr on cached_subvol for 'linkto' value. Do path based getxattr
+         * as root:root. If a fd is already open, access check wont be done*/
+
+        if (!local->loc.inode) {
+                ret = syncop_fgetxattr (src_node, local->fd, &dict,
+                                        conf->link_xattr_name);
+        } else {
+                SYNCTASK_SETID (0, 0);
+                ret = syncop_getxattr (src_node, &local->loc, &dict,
+                                       conf->link_xattr_name);
+                SYNCTASK_SETID (frame->root->uid, frame->root->gid);
         }
 
-        if (strcmp (parent->path, "/") == 0)
-                gf_asprintf ((char **)&child->path, "/%s", name);
-        else
-                gf_asprintf ((char **)&child->path, "%s/%s", parent->path, name);
+        if (!ret)
+                dst_node = dht_linkfile_subvol (this, NULL, NULL, dict);
 
-        if (!child->path) {
+        if (ret) {
+                if ((errno != ENOENT) || (!local->loc.inode)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: failed to get the 'linkto' xattr %s",
+                                local->loc.path, strerror (errno));
+                        goto out;
+                }
+                /* Need to do lookup on hashed subvol, then get the file */
+                ret = syncop_lookup (this, &local->loc, NULL, &stbuf, NULL,
+                                     NULL);
+                if (ret)
+                        goto out;
+                dst_node = dht_subvol_get_cached (this, local->loc.inode);
+        }
+
+        if (!dst_node) {
                 gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
+                        "%s: failed to get the destination node",
+                        local->loc.path);
+                ret = -1;
+                goto out;
         }
 
-        child->name = strrchr (child->path, '/');
-        if (child->name)
-                child->name++;
+        /* lookup on dst */
+        if (local->loc.inode) {
+                ret = syncop_lookup (dst_node, &local->loc, NULL, &stbuf, NULL, NULL);
 
-        child->parent = inode_ref (parent->inode);
-        child->inode = inode_new (parent->inode->table);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: failed to lookup the file on %s",
+                                local->loc.path, dst_node->name);
+                        goto out;
+                }
 
-        if (!child->inode) {
+                if (uuid_compare (stbuf.ia_gfid, local->loc.inode->gfid)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: gfid different on the target file on %s",
+                                local->loc.path, dst_node->name);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        /* update inode ctx (the layout) */
+        dht_layout_unref (this, local->layout);
+
+        ret = dht_layout_preset (this, dst_node, inode);
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "%s: could not set preset layout for subvol %s",
+                        local->loc.path, dst_node->name);
+                ret   = -1;
+                goto out;
+        }
+
+        layout = dht_layout_for_subvol (this, dst_node);
+        if (!layout) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "%s: no pre-set layout for subvolume %s",
+                        local->loc.path, dst_node ? dst_node->name : "<nil>");
+                ret = -1;
+                goto out;
+        }
+
+        ret = dht_layout_set (this, inode, layout);
+        if (ret) {
                 gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
+                        "%s: failed to set the new layout",
+                        local->loc.path);
+                goto out;
         }
 
+        local->cached_subvol = dst_node;
+        ret = 0;
+
+        /* once we detect the migration complete, the inode-ctx2 is no more
+           required.. delete the ctx and also, it means, open() already
+           done on all the fd of inode */
+        ret = inode_ctx_reset1 (inode, this, &tmp_subvol);
+        if (tmp_subvol)
+                goto out;
+
+        if (list_empty (&inode->fd_list))
+                goto out;
+
+        /* perform open as root:root. There is window between linkfile
+         * creation(root:root) and setattr with the correct uid/gid
+         */
+        SYNCTASK_SETID(0, 0);
+
+        /* perform 'open()' on all the fd's present on the inode */
+        tmp_loc.inode = inode;
+        inode_path (inode, NULL, &path);
+        if (path)
+                tmp_loc.path = path;
+        list_for_each_entry (iter_fd, &inode->fd_list, inode_list) {
+                if (fd_is_anonymous (iter_fd))
+                        continue;
+
+                ret = syncop_open (dst_node, &tmp_loc,
+                                   iter_fd->flags, iter_fd);
+                if (ret == -1) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to open "
+                                "the fd (%p, flags=0%o) on file %s @ %s",
+                                iter_fd, iter_fd->flags, path, dst_node->name);
+                        open_failed = 1;
+                }
+        }
+        GF_FREE (path);
+
+        SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+        if (open_failed) {
+                ret = -1;
+                goto out;
+        }
+        ret = 0;
+out:
+
+        return ret;
+}
+
+int
+dht_rebalance_complete_check (xlator_t *this, call_frame_t *frame)
+{
+        int         ret     = -1;
+
+        ret = synctask_new (this->ctx->env, dht_migration_complete_check_task,
+                            dht_migration_complete_check_done,
+                            frame, frame);
+        return ret;
+}
+
+/* During 'in-progress' state, both nodes should have the file */
+static int
+dht_inprogress_check_done (int op_ret, call_frame_t *sync_frame, void *data)
+{
+        dht_local_t *local = NULL;
+
+        local = sync_frame->local;
+
+        local->rebalance.target_op_fn (THIS, sync_frame, op_ret);
+
         return 0;
-err:
-        loc_wipe (child);
-        return -1;
+}
+
+static int
+dht_rebalance_inprogress_task (void *data)
+{
+        int           ret      = -1;
+        xlator_t     *src_node = NULL;
+        xlator_t     *dst_node = NULL;
+        dht_local_t  *local    = NULL;
+        dict_t       *dict     = NULL;
+        call_frame_t *frame    = NULL;
+        xlator_t     *this     = NULL;
+        char         *path     = NULL;
+        struct iatt   stbuf    = {0,};
+        loc_t         tmp_loc  = {0,};
+        dht_conf_t   *conf     = NULL;
+        inode_t      *inode    = NULL;
+        fd_t         *iter_fd  = NULL;
+        int           open_failed = 0;
+
+        this  = THIS;
+        frame = data;
+        local = frame->local;
+        conf = this->private;
+
+        src_node = local->cached_subvol;
+
+        if (!local->loc.inode && !local->fd)
+                goto out;
+
+        inode = (!local->fd) ? local->loc.inode : local->fd->inode;
+
+        /* getxattr on cached_subvol for 'linkto' value. Do path based getxattr
+         * as root:root. If a fd is already open, access check wont be done*/
+        if (local->loc.inode) {
+                SYNCTASK_SETID (0, 0);
+                ret = syncop_getxattr (src_node, &local->loc, &dict,
+                                       conf->link_xattr_name);
+                SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+        } else {
+                ret = syncop_fgetxattr (src_node, local->fd, &dict,
+                                        conf->link_xattr_name);
+        }
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to get the 'linkto' xattr %s",
+                        local->loc.path, strerror (errno));
+                        goto out;
+        }
+
+        dst_node = dht_linkfile_subvol (this, NULL, NULL, dict);
+        if (!dst_node) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to get the 'linkto' xattr from dict",
+                        local->loc.path);
+                ret = -1;
+                goto out;
+        }
+
+        local->rebalance.target_node = dst_node;
+
+        if (local->loc.inode) {
+                /* lookup on dst */
+                ret = syncop_lookup (dst_node, &local->loc, NULL,
+                                     &stbuf, NULL, NULL);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: failed to lookup the file on %s",
+                                local->loc.path, dst_node->name);
+                        goto out;
+                }
+
+                if (uuid_compare (stbuf.ia_gfid, local->loc.inode->gfid)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: gfid different on the target file on %s",
+                                local->loc.path, dst_node->name);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        ret = 0;
+
+        if (list_empty (&inode->fd_list))
+                goto done;
+
+        /* perform open as root:root. There is window between linkfile
+         * creation(root:root) and setattr with the correct uid/gid
+         */
+        SYNCTASK_SETID (0, 0);
+
+        tmp_loc.inode = inode;
+        inode_path (inode, NULL, &path);
+        if (path)
+                tmp_loc.path = path;
+
+        list_for_each_entry (iter_fd, &inode->fd_list, inode_list) {
+                if (fd_is_anonymous (iter_fd))
+                        continue;
+
+                ret = syncop_open (dst_node, &tmp_loc,
+                                   iter_fd->flags, iter_fd);
+                if (ret == -1) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to send open "
+                                "the fd (%p, flags=0%o) on file %s @ %s",
+                                iter_fd, iter_fd->flags, path, dst_node->name);
+                        open_failed = 1;
+                }
+        }
+        GF_FREE (path);
+
+        SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+
+        if (open_failed) {
+                ret = -1;
+                goto out;
+        }
+
+done:
+        ret = dht_inode_ctx_set1 (this, inode, dst_node);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set inode-ctx target file at %s",
+                        local->loc.path, dst_node->name);
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+dht_rebalance_in_progress_check (xlator_t *this, call_frame_t *frame)
+{
+
+        int         ret     = -1;
+
+        ret = synctask_new (this->ctx->env, dht_rebalance_inprogress_task,
+                            dht_inprogress_check_done,
+                            frame, frame);
+        return ret;
+}
+
+int
+dht_inode_ctx_layout_set (inode_t *inode, xlator_t *this,
+                          dht_layout_t *layout_int)
+{
+        dht_inode_ctx_t         *ctx            = NULL;
+        int                      ret            = -1;
+
+        ret = dht_inode_ctx_get (inode, this, &ctx);
+        if (!ret && ctx) {
+                ctx->layout = layout_int;
+        } else {
+                ctx = GF_CALLOC (1, sizeof (*ctx), gf_dht_mt_inode_ctx_t);
+                if (!ctx)
+                        return ret;
+                ctx->layout = layout_int;
+        }
+
+        ret = dht_inode_ctx_set (inode, this, ctx);
+
+        return ret;
+}
+
+int
+dht_inode_ctx_time_update (inode_t *inode, xlator_t *this, struct iatt *stat,
+                           int32_t post)
+{
+        dht_inode_ctx_t         *ctx            = NULL;
+        dht_stat_time_t         *time           = 0;
+        int                      ret            = -1;
+
+        GF_VALIDATE_OR_GOTO (this->name, stat, out);
+        GF_VALIDATE_OR_GOTO (this->name, inode, out);
+
+        ret = dht_inode_ctx_get (inode, this, &ctx);
+
+        if (ret) {
+                ctx = GF_CALLOC (1, sizeof (*ctx), gf_dht_mt_inode_ctx_t);
+                if (!ctx)
+                        return -1;
+        }
+
+        time = &ctx->time;
+
+        DHT_UPDATE_TIME(time->mtime, time->mtime_nsec,
+                        stat->ia_mtime, stat->ia_mtime_nsec, inode, post);
+        DHT_UPDATE_TIME(time->ctime, time->ctime_nsec,
+                        stat->ia_ctime, stat->ia_ctime_nsec, inode, post);
+        DHT_UPDATE_TIME(time->atime, time->atime_nsec,
+                        stat->ia_atime, stat->ia_atime_nsec, inode, post);
+
+        ret = dht_inode_ctx_set (inode, this, ctx);
+out:
+        return 0;
+}
+
+int
+dht_inode_ctx_get (inode_t *inode, xlator_t *this, dht_inode_ctx_t **ctx)
+{
+        int             ret     = -1;
+        uint64_t        ctx_int = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, inode, out);
+
+        ret = inode_ctx_get (inode, this, &ctx_int);
+
+        if (ret)
+                return ret;
+
+        if (ctx)
+                *ctx = (dht_inode_ctx_t *) ctx_int;
+out:
+        return ret;
+}
+
+int dht_inode_ctx_set (inode_t *inode, xlator_t *this, dht_inode_ctx_t *ctx)
+{
+        int             ret = -1;
+        uint64_t        ctx_int = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, inode, out);
+        GF_VALIDATE_OR_GOTO (this->name, ctx, out);
+
+        ctx_int = (long)ctx;
+        ret = inode_ctx_set (inode, this, &ctx_int);
+out:
+        return ret;
 }
diff --git a/xlators/cluster/dht/src/dht-inode-read.c b/xlators/cluster/dht/src/dht-inode-read.c
new file mode 100644
index 000000000..ece84151a
--- /dev/null
+++ b/xlators/cluster/dht/src/dht-inode-read.c
@@ -0,0 +1,1139 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "dht-common.h"
+
+int dht_access2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_readv2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_attr2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_open2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_flush2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_lk2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_fsync2 (xlator_t *this, call_frame_t *frame, int ret);
+
+int
+dht_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+              int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = 0;
+
+        local = frame->local;
+        prev = cookie;
+
+        local->op_errno = op_errno;
+        if ((op_ret == -1) && (op_errno != ENOENT)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+                goto out;
+        }
+
+        if (!op_ret || (local->call_cnt != 1))
+                goto out;
+
+        /* rebalance would have happened */
+        local->rebalance.target_op_fn = dht_open2;
+        ret = dht_rebalance_complete_check (this, frame);
+        if (!ret)
+                return 0;
+
+out:
+        DHT_STACK_UNWIND (open, frame, op_ret, op_errno, local->fd, xdata);
+
+        return 0;
+}
+
+int
+dht_open2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t *local  = NULL;
+        xlator_t    *subvol = NULL;
+        int          op_errno = EINVAL;
+
+        local = frame->local;
+        if (!local)
+                goto out;
+
+        op_errno = ENOENT;
+        if (op_ret)
+                goto out;
+
+        local->call_cnt = 2;
+        subvol = local->cached_subvol;
+
+        STACK_WIND (frame, dht_open_cbk, subvol, subvol->fops->open,
+                    &local->loc, local->rebalance.flags, local->fd,
+                    NULL);
+        return 0;
+
+out:
+        DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
+        return 0;
+}
+
+int
+dht_open (call_frame_t *frame, xlator_t *this,
+          loc_t *loc, int flags, fd_t *fd, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, loc, fd, GF_FOP_OPEN);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->rebalance.flags = flags;
+        local->call_cnt = 1;
+
+        STACK_WIND (frame, dht_open_cbk, subvol, subvol->fops->open,
+                    loc, flags, fd, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (open, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+int
+dht_file_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int op_ret, int op_errno, struct iatt *stbuf, dict_t *xdata)
+{
+        xlator_t     *subvol = 0;
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
+        inode_t      *inode = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev = cookie;
+
+        if ((op_ret == -1) && (op_errno != ENOENT)) {
+                local->op_errno = op_errno;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+                goto out;
+        }
+
+        if (local->call_cnt != 1)
+                goto out;
+
+        local->op_errno = op_errno;
+        /* Check if the rebalance phase2 is true */
+        if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (stbuf)) {
+                inode = (local->fd) ? local->fd->inode : local->loc.inode;
+                ret = dht_inode_ctx_get1 (this, inode, &subvol);
+                if (!subvol) {
+                        /* Phase 2 of migration */
+                        local->rebalance.target_op_fn = dht_attr2;
+                        ret = dht_rebalance_complete_check (this, frame);
+                        if (!ret)
+                                return 0;
+                } else {
+                        /* value is already set in fd_ctx, that means no need
+                           to check for whether its complete or not. */
+                        dht_attr2 (this, frame, 0);
+                        return 0;
+                }
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (stat, frame, op_ret, op_errno, stbuf, xdata);
+err:
+        return 0;
+}
+
+int
+dht_attr2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t *local  = NULL;
+        xlator_t    *subvol = NULL;
+        int          op_errno = EINVAL;
+
+        local = frame->local;
+        if (!local)
+                goto out;
+
+        op_errno = local->op_errno;
+        if (op_ret == -1)
+                goto out;
+
+        subvol = local->cached_subvol;
+        local->call_cnt = 2;
+
+        if (local->fop == GF_FOP_FSTAT) {
+                STACK_WIND (frame, dht_file_attr_cbk, subvol,
+                            subvol->fops->fstat, local->fd, NULL);
+        } else {
+                STACK_WIND (frame, dht_file_attr_cbk, subvol,
+                            subvol->fops->stat, &local->loc, NULL);
+        }
+        return 0;
+
+out:
+        DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
+        return 0;
+}
+
+int
+dht_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+              int op_ret, int op_errno, struct iatt *stbuf, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev = cookie;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+
+                        goto unlock;
+                }
+
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+
+                local->op_ret = 0;
+        }
+unlock:
+        UNLOCK (&frame->lock);
+out:
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                DHT_STACK_UNWIND (stat, frame, local->op_ret, local->op_errno,
+                                  &local->stbuf, xdata);
+        }
+err:
+        return 0;
+}
+
+int
+dht_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           call_cnt = 0;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_STAT);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (IA_ISREG (loc->inode->ia_type)) {
+                local->call_cnt = 1;
+
+                subvol = local->cached_subvol;
+
+                STACK_WIND (frame, dht_file_attr_cbk, subvol,
+                            subvol->fops->stat, loc, xdata);
+
+                return 0;
+        }
+
+        local->call_cnt = call_cnt = layout->cnt;
+
+        for (i = 0; i < call_cnt; i++) {
+                subvol = layout->list[i].xlator;
+
+                STACK_WIND (frame, dht_attr_cbk,
+                            subvol, subvol->fops->stat,
+                            loc, xdata);
+        }
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+dht_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           call_cnt = 0;
+
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FSTAT);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (IA_ISREG (fd->inode->ia_type)) {
+                local->call_cnt = 1;
+
+                subvol = local->cached_subvol;
+
+                STACK_WIND (frame, dht_file_attr_cbk, subvol,
+                            subvol->fops->fstat, fd, xdata);
+
+                return 0;
+        }
+
+        local->call_cnt = call_cnt = layout->cnt;
+
+        for (i = 0; i < call_cnt; i++) {
+                subvol = layout->list[i].xlator;
+                STACK_WIND (frame, dht_attr_cbk,
+                            subvol, subvol->fops->fstat,
+                            fd, xdata);
+        }
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+int
+dht_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int op_ret, int op_errno,
+               struct iovec *vector, int count, struct iatt *stbuf,
+               struct iobref *iobref, dict_t *xdata)
+{
+        dht_local_t *local      = NULL;
+        int          ret        = 0;
+        inode_t     *inode      = NULL;
+        xlator_t    *subvol = 0;
+
+        local = frame->local;
+        if (!local) {
+                op_ret = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
+
+        /* This is already second try, no need for re-check */
+        if (local->call_cnt != 1)
+                goto out;
+
+        if ((op_ret == -1) && (op_errno != ENOENT))
+                goto out;
+
+        local->op_errno = op_errno;
+        if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (stbuf)) {
+                /* File would be migrated to other node */
+                ret = dht_inode_ctx_get1 (this, inode, &subvol);
+                if (!subvol) {
+                        local->rebalance.target_op_fn = dht_readv2;
+                        ret = dht_rebalance_complete_check (this, frame);
+                        if (!ret)
+                                return 0;
+                } else {
+                        /* value is already set in fd_ctx, that means no need
+                           to check for whether its complete or not. */
+                        dht_readv2 (this, frame, 0);
+                        return 0;
+                }
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count, stbuf,
+                          iobref, xdata);
+
+        return 0;
+}
+
+int
+dht_readv2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t *local  = NULL;
+        xlator_t    *subvol = NULL;
+        int          op_errno = EINVAL;
+
+        local = frame->local;
+        if (!local)
+                goto out;
+
+        op_errno = local->op_errno;
+        if (op_ret == -1)
+                goto out;
+
+        local->call_cnt = 2;
+        subvol = local->cached_subvol;
+
+        STACK_WIND (frame, dht_readv_cbk, subvol, subvol->fops->readv,
+                    local->fd, local->rebalance.size, local->rebalance.offset,
+                    local->rebalance.flags, NULL);
+
+        return 0;
+
+out:
+        DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
+        return 0;
+}
+
+int
+dht_readv (call_frame_t *frame, xlator_t *this,
+           fd_t *fd, size_t size, off_t off, uint32_t flags, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_READ);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->rebalance.offset = off;
+        local->rebalance.size   = size;
+        local->rebalance.flags  = flags;
+        local->call_cnt = 1;
+
+        STACK_WIND (frame, dht_readv_cbk,
+                    subvol, subvol->fops->readv,
+                    fd, size, off, flags, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
+
+        return 0;
+}
+
+int
+dht_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int op_ret, int op_errno, dict_t *xdata)
+{
+        int          ret = -1;
+        dht_local_t *local = NULL;
+        xlator_t    *subvol = NULL;
+        call_frame_t *prev = NULL;
+
+        local = frame->local;
+        prev = cookie;
+
+        if (!prev || !prev->this)
+                goto out;
+        if (local->call_cnt != 1)
+                goto out;
+        if ((op_ret == -1) && (op_errno == ENOTCONN) &&
+            IA_ISDIR(local->loc.inode->ia_type)) {
+
+                subvol = dht_subvol_next_available (this, prev->this);
+                if (!subvol)
+                        goto out;
+
+                /* check if we are done with visiting every node */
+                if (subvol == local->cached_subvol) {
+                        goto out;
+                }
+
+                STACK_WIND (frame, dht_access_cbk, subvol, subvol->fops->access,
+                            &local->loc, local->rebalance.flags, NULL);
+                return 0;
+        }
+        if ((op_ret == -1) && (op_errno == ENOENT)) {
+                /* File would be migrated to other node */
+                local->op_errno = op_errno;
+                local->rebalance.target_op_fn = dht_access2;
+                ret = dht_rebalance_complete_check (frame->this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+out:
+        DHT_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+dht_access2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t *local  = NULL;
+        xlator_t    *subvol = NULL;
+        int          op_errno = EINVAL;
+
+        local = frame->local;
+        if (!local)
+                goto out;
+
+        op_errno = local->op_errno;
+        if (op_ret == -1)
+                goto out;
+
+        local->call_cnt = 2;
+        subvol = local->cached_subvol;
+
+        STACK_WIND (frame, dht_access_cbk, subvol, subvol->fops->access,
+                    &local->loc, local->rebalance.flags, NULL);
+
+        return 0;
+
+out:
+        DHT_STACK_UNWIND (access, frame, -1, op_errno, NULL);
+        return 0;
+}
+
+
+int
+dht_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+            dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_ACCESS);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        local->rebalance.flags = mask;
+        local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_access_cbk, subvol, subvol->fops->access,
+                    loc, mask, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (access, frame, -1, op_errno, NULL);
+
+        return 0;
+}
+
+
+int
+dht_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int op_ret, int op_errno, dict_t *xdata)
+{
+        dht_local_t  *local  = NULL;
+        inode_t      *inode  = NULL;
+        xlator_t     *subvol = 0;
+
+        local = frame->local;
+
+        local->op_errno = op_errno;
+
+        if (local->call_cnt != 1)
+                goto out;
+
+        /* If context is set, then send flush() it to the destination */
+        dht_inode_ctx_get1 (this, inode, &subvol);
+        if (subvol) {
+                dht_flush2 (this, frame, 0);
+                return 0;
+        }
+
+out:
+        DHT_STACK_UNWIND (flush, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+int
+dht_flush2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+
+        local = frame->local;
+
+        dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+        STACK_WIND (frame, dht_flush_cbk,
+                    subvol, subvol->fops->flush, local->fd, NULL);
+
+        return 0;
+}
+
+
+int
+dht_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FLUSH);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->call_cnt = 1;
+
+        STACK_WIND (frame, dht_flush_cbk,
+                    subvol, subvol->fops->flush, fd, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
+
+        return 0;
+}
+
+
+int
+dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+               int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+               dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
+        inode_t      *inode = NULL;
+        xlator_t     *subvol = 0;
+
+        local = frame->local;
+        prev = cookie;
+
+        local->op_errno = op_errno;
+        if (op_ret == -1 && (op_errno != ENOENT)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+                goto out;
+        }
+
+        if (local->call_cnt != 1) {
+                if (local->stbuf.ia_blocks) {
+                        dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+                        dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+                }
+                goto out;
+        }
+
+        local->op_errno = op_errno;
+        dht_inode_ctx_get1 (this, inode, &subvol);
+        if (!subvol) {
+                local->rebalance.target_op_fn = dht_fsync2;
+
+                /* Check if the rebalance phase1 is true */
+                if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+                        dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+                        dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+
+                        ret = dht_rebalance_in_progress_check (this, frame);
+                }
+
+                /* Check if the rebalance phase2 is true */
+                if (IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+                        ret = dht_rebalance_complete_check (this, frame);
+                }
+                if (!ret)
+                        return 0;
+        } else {
+                dht_fsync2 (this, frame, 0);
+                return 0;
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (postbuf);
+        DHT_STRIP_PHASE1_FLAGS (prebuf);
+        DHT_STACK_UNWIND (fsync, frame, op_ret, op_errno,
+                          prebuf, postbuf, xdata);
+
+        return 0;
+}
+
+int
+dht_fsync2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+
+        local = frame->local;
+
+        dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+        STACK_WIND (frame, dht_fsync_cbk, subvol, subvol->fops->fsync,
+                    local->fd, local->rebalance.flags, NULL);
+
+        return 0;
+}
+
+int
+dht_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync,
+           dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FSYNC);
+        if (!local) {
+                op_errno = ENOMEM;
+
+                goto err;
+        }
+
+        local->call_cnt = 1;
+        local->rebalance.flags = datasync;
+
+        subvol = local->cached_subvol;
+
+        STACK_WIND (frame, dht_fsync_cbk, subvol, subvol->fops->fsync,
+                    fd, datasync, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+/* TODO: for 'lk()' call, we need some other special error, may be ESTALE to
+   indicate that lock migration happened on the fd, so we can consider it as
+   phase 2 of migration */
+int
+dht_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+            int op_ret, int op_errno, struct gf_flock *flock, dict_t *xdata)
+{
+        DHT_STACK_UNWIND (lk, frame, op_ret, op_errno, flock, xdata);
+
+        return 0;
+}
+
+
+int
+dht_lk (call_frame_t *frame, xlator_t *this,
+        fd_t *fd, int cmd, struct gf_flock *flock, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        subvol = dht_subvol_get_cached (this, fd->inode);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        /* TODO: for rebalance, we need to preserve the fop arguments */
+        STACK_WIND (frame, dht_lk_cbk, subvol, subvol->fops->lk, fd,
+                    cmd, flock, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+/* Symlinks are currently not migrated, so no need for any check here */
+int
+dht_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int op_ret, int op_errno, const char *path,
+                  struct iatt *stbuf, dict_t *xdata)
+{
+        dht_local_t *local = NULL;
+
+        local = frame->local;
+        if (op_ret == -1)
+                goto err;
+
+        if (!local) {
+                op_ret = -1;
+                op_errno = EINVAL;
+        }
+
+err:
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (readlink, frame, op_ret, op_errno, path, stbuf, xdata);
+
+        return 0;
+}
+
+
+int
+dht_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+              dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_READLINK);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_readlink_cbk,
+                    subvol, subvol->fops->readlink,
+                    loc, size, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+/* Currently no translators on top of 'distribute' will be using
+ * below fops, hence not implementing 'migration' related checks
+ */
+
+int
+dht_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        DHT_STACK_UNWIND (xattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+
+int
+dht_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_XATTROP);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->call_cnt = 1;
+
+        STACK_WIND (frame,
+                    dht_xattrop_cbk,
+                    subvol, subvol->fops->xattrop,
+                    loc, flags, dict, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+dht_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        DHT_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+
+int
+dht_fxattrop (call_frame_t *frame, xlator_t *this,
+              fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        subvol = dht_subvol_get_cached (this, fd->inode);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame,
+                    dht_fxattrop_cbk,
+                    subvol, subvol->fops->fxattrop,
+                    fd, flags, dict, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+dht_inodelk_cbk (call_frame_t *frame, void *cookie,
+                 xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
+
+{
+        DHT_STACK_UNWIND (inodelk, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+
+int32_t
+dht_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+             loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_INODELK);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->call_cnt = 1;
+
+        STACK_WIND (frame,
+                    dht_inodelk_cbk,
+                    subvol, subvol->fops->inodelk,
+                    volume, loc, cmd, lock, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (inodelk, frame, -1, op_errno, NULL);
+
+        return 0;
+}
+
+
+int
+dht_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *xdata)
+
+{
+        DHT_STACK_UNWIND (finodelk, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+
+int
+dht_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+              fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        subvol = dht_subvol_get_cached (this, fd->inode);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+
+        STACK_WIND (frame, dht_finodelk_cbk, subvol, subvol->fops->finodelk,
+                    volume, fd, cmd, lock, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (finodelk, frame, -1, op_errno, NULL);
+
+        return 0;
+}
diff --git a/xlators/cluster/dht/src/dht-inode-write.c b/xlators/cluster/dht/src/dht-inode-write.c
new file mode 100644
index 000000000..4b3f3a049
--- /dev/null
+++ b/xlators/cluster/dht/src/dht-inode-write.c
@@ -0,0 +1,1013 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "dht-common.h"
+
+int dht_writev2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_truncate2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_setattr2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_fallocate2(xlator_t *this, call_frame_t *frame, int op_ret);
+int dht_discard2(xlator_t *this, call_frame_t *frame, int op_ret);
+int dht_zerofill2(xlator_t *this, call_frame_t *frame, int op_ret);
+
+int
+dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int op_ret, int op_errno, struct iatt *prebuf,
+                struct iatt *postbuf, dict_t *xdata)
+{
+        dht_local_t *local = NULL;
+        int          ret   = -1;
+        xlator_t    *subvol = NULL;
+
+        if (op_ret == -1 && (op_errno != ENOENT)) {
+                goto out;
+        }
+
+        local = frame->local;
+        if (!local) {
+                op_ret = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
+
+        if (local->call_cnt != 1) {
+                /* preserve the modes of source */
+                if (local->stbuf.ia_blocks) {
+                        dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+                        dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+                }
+                goto out;
+        }
+
+        local->rebalance.target_op_fn = dht_writev2;
+
+        local->op_errno = op_errno;
+        /* Phase 2 of migration */
+        if (IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+                ret = dht_rebalance_complete_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+        /* Check if the rebalance phase1 is true */
+        if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+                dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+                dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+
+                ret = dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+                if (subvol) {
+                        dht_writev2 (this, frame, 0);
+                        return 0;
+                }
+                ret = dht_rebalance_in_progress_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (postbuf);
+        DHT_STRIP_PHASE1_FLAGS (prebuf);
+
+        DHT_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf,
+                          xdata);
+
+        return 0;
+}
+
+int
+dht_writev2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+
+        local = frame->local;
+
+        dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+        STACK_WIND (frame, dht_writev_cbk,
+                    subvol, subvol->fops->writev,
+                    local->fd, local->rebalance.vector, local->rebalance.count,
+                    local->rebalance.offset, local->rebalance.flags,
+                    local->rebalance.iobref, NULL);
+
+        return 0;
+}
+
+int
+dht_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+            struct iovec *vector, int count, off_t off, uint32_t flags,
+            struct iobref *iobref, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_WRITE);
+        if (!local) {
+
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+
+        local->rebalance.vector = iov_dup (vector, count);
+        local->rebalance.offset = off;
+        local->rebalance.count = count;
+        local->rebalance.flags = flags;
+        local->rebalance.iobref = iobref_ref (iobref);
+        local->call_cnt = 1;
+
+        STACK_WIND (frame, dht_writev_cbk,
+                    subvol, subvol->fops->writev,
+                    fd, vector, count, off, flags, iobref, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+
+int
+dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int op_ret, int op_errno, struct iatt *prebuf,
+                  struct iatt *postbuf, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
+        xlator_t    *subvol = NULL;
+        inode_t      *inode = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev = cookie;
+
+        if ((op_ret == -1) && (op_errno != ENOENT)) {
+                local->op_errno = op_errno;
+                local->op_ret = -1;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+
+                goto out;
+        }
+
+        if (local->call_cnt != 1) {
+                if (local->stbuf.ia_blocks) {
+                        dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+                        dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+                }
+                goto out;
+        }
+
+        local->rebalance.target_op_fn = dht_truncate2;
+
+        local->op_errno = op_errno;
+        /* Phase 2 of migration */
+        if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+                ret = dht_rebalance_complete_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+        /* Check if the rebalance phase1 is true */
+        if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+                dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+                dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+                inode = (local->fd) ? local->fd->inode : local->loc.inode;
+                dht_inode_ctx_get1 (this, inode, &subvol);
+                if (subvol) {
+                        dht_truncate2 (this, frame, 0);
+                        return 0;
+                }
+                ret = dht_rebalance_in_progress_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (postbuf);
+        DHT_STRIP_PHASE1_FLAGS (prebuf);
+        DHT_STACK_UNWIND (truncate, frame, op_ret, op_errno,
+                          prebuf, postbuf, xdata);
+err:
+        return 0;
+}
+
+
+int
+dht_truncate2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+        inode_t      *inode = NULL;
+
+        local = frame->local;
+
+        inode = local->fd ? local->fd->inode : local->loc.inode;
+
+        dht_inode_ctx_get1 (this, inode, &subvol);
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+        if (local->fop == GF_FOP_TRUNCATE) {
+                STACK_WIND (frame, dht_truncate_cbk, subvol,
+                            subvol->fops->truncate, &local->loc,
+                            local->rebalance.offset, NULL);
+        } else {
+                STACK_WIND (frame, dht_truncate_cbk, subvol,
+                            subvol->fops->ftruncate, local->fd,
+                            local->rebalance.offset, NULL);
+        }
+
+        return 0;
+}
+
+int
+dht_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+              dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_TRUNCATE);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        local->rebalance.offset = offset;
+        local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_truncate_cbk,
+                    subvol, subvol->fops->truncate,
+                    loc, offset, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+int
+dht_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+               dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FTRUNCATE);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        local->rebalance.offset = offset;
+        local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_truncate_cbk,
+                    subvol, subvol->fops->ftruncate,
+                    fd, offset, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+dht_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                  int op_ret, int op_errno, struct iatt *prebuf,
+                  struct iatt *postbuf, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
+        xlator_t    *subvol = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev = cookie;
+
+        if ((op_ret == -1) && (op_errno != ENOENT)) {
+                local->op_errno = op_errno;
+                local->op_ret = -1;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+
+                goto out;
+        }
+
+        if (local->call_cnt != 1) {
+                if (local->stbuf.ia_blocks) {
+                        dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+                        dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+                }
+                goto out;
+        }
+        local->rebalance.target_op_fn = dht_fallocate2;
+
+        /* Phase 2 of migration */
+        if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+                ret = dht_rebalance_complete_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+        /* Check if the rebalance phase1 is true */
+        if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+                dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+                dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+                dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+                if (subvol) {
+                        dht_fallocate2 (this, frame, 0);
+                        return 0;
+                }
+                ret = dht_rebalance_in_progress_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (postbuf);
+        DHT_STRIP_PHASE1_FLAGS (prebuf);
+        DHT_STACK_UNWIND (fallocate, frame, op_ret, op_errno,
+                          prebuf, postbuf, xdata);
+err:
+        return 0;
+}
+
+int
+dht_fallocate2(xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+
+        local = frame->local;
+
+        dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+	STACK_WIND(frame, dht_fallocate_cbk, subvol, subvol->fops->fallocate,
+		   local->fd, local->rebalance.flags, local->rebalance.offset,
+		   local->rebalance.size, NULL);
+
+        return 0;
+}
+
+int
+dht_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+	      off_t offset, size_t len, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FALLOCATE);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+	local->rebalance.flags = mode;
+        local->rebalance.offset = offset;
+	local->rebalance.size = len;
+
+        local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_fallocate_cbk,
+                    subvol, subvol->fops->fallocate,
+                    fd, mode, offset, len, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+dht_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                int op_ret, int op_errno, struct iatt *prebuf,
+                struct iatt *postbuf, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
+        xlator_t    *subvol = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev = cookie;
+
+        if ((op_ret == -1) && (op_errno != ENOENT)) {
+                local->op_errno = op_errno;
+                local->op_ret = -1;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+
+                goto out;
+        }
+
+        if (local->call_cnt != 1) {
+                if (local->stbuf.ia_blocks) {
+                        dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+                        dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+                }
+                goto out;
+        }
+        local->rebalance.target_op_fn = dht_discard2;
+
+        /* Phase 2 of migration */
+        if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+                ret = dht_rebalance_complete_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+        /* Check if the rebalance phase1 is true */
+        if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+                dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+                dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+                dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+                if (subvol) {
+                        dht_discard2 (this, frame, 0);
+                        return 0;
+                }
+                ret = dht_rebalance_in_progress_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (postbuf);
+        DHT_STRIP_PHASE1_FLAGS (prebuf);
+        DHT_STACK_UNWIND (discard, frame, op_ret, op_errno,
+                          prebuf, postbuf, xdata);
+err:
+        return 0;
+}
+
+int
+dht_discard2(xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+
+        local = frame->local;
+
+        dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+	STACK_WIND(frame, dht_discard_cbk, subvol, subvol->fops->discard,
+		   local->fd, local->rebalance.offset, local->rebalance.size,
+		   NULL);
+
+        return 0;
+}
+
+int
+dht_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	    size_t len, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_DISCARD);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        local->rebalance.offset = offset;
+	local->rebalance.size = len;
+
+        local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_discard_cbk, subvol, subvol->fops->discard,
+                    fd, offset, len, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+int
+dht_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                int op_ret, int op_errno, struct iatt *prebuf,
+                struct iatt *postbuf, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev  = NULL;
+        int           ret   = -1;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev = cookie;
+
+        if ((op_ret == -1) && (op_errno != ENOENT)) {
+                local->op_errno = op_errno;
+                local->op_ret = -1;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+                goto out;
+        }
+
+        if (local->call_cnt != 1) {
+                if (local->stbuf.ia_blocks) {
+                        dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+                        dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+                }
+                goto out;
+        }
+        local->rebalance.target_op_fn = dht_zerofill2;
+        /* Phase 2 of migration */
+        if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+                ret = dht_rebalance_complete_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+        /* Check if the rebalance phase1 is true */
+        if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+                dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+                dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+                ret = fd_ctx_get (local->fd, this, NULL);
+                if (!ret) {
+                        dht_zerofill2 (this, frame, 0);
+                        return 0;
+                }
+                ret = dht_rebalance_in_progress_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (postbuf);
+        DHT_STRIP_PHASE1_FLAGS (prebuf);
+        DHT_STACK_UNWIND (zerofill, frame, op_ret, op_errno,
+                          prebuf, postbuf, xdata);
+err:
+        return 0;
+}
+
+int
+dht_zerofill2(xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local          = NULL;
+        xlator_t     *subvol         = NULL;
+        uint64_t      tmp_subvol     = 0;
+        int           ret            = -1;
+
+        local = frame->local;
+
+        if (local->fd)
+                ret = fd_ctx_get (local->fd, this, &tmp_subvol);
+        if (!ret)
+                subvol = (xlator_t *)(long)tmp_subvol;
+
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+        STACK_WIND(frame, dht_zerofill_cbk, subvol, subvol->fops->zerofill,
+                   local->fd, local->rebalance.offset, local->rebalance.size,
+                   NULL);
+
+        return 0;
+}
+
+int
+dht_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+            size_t len, dict_t *xdata)
+{
+        xlator_t     *subvol       = NULL;
+        int           op_errno     = -1;
+        dht_local_t  *local        = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_ZEROFILL);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        local->rebalance.offset = offset;
+        local->rebalance.size = len;
+
+        local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_zerofill_cbk, subvol, subvol->fops->zerofill,
+                    fd, offset, len, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+
+/* handle cases of migration here for 'setattr()' calls */
+int
+dht_file_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int op_ret, int op_errno, struct iatt *prebuf,
+                      struct iatt *postbuf, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
+
+        local = frame->local;
+        prev = cookie;
+
+        local->op_errno = op_errno;
+        if ((op_ret == -1) && (op_errno != ENOENT)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+                goto out;
+        }
+
+        if (local->call_cnt != 1)
+                goto out;
+
+        local->rebalance.target_op_fn = dht_setattr2;
+
+        /* Phase 2 of migration */
+        if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+                ret = dht_rebalance_complete_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+        /* At the end of the migration process, whatever 'attr' we
+           have on source file will be migrated to destination file
+           in one shot, hence we don't need to check for in progress
+           state here (ie, PHASE1) */
+out:
+        DHT_STRIP_PHASE1_FLAGS (postbuf);
+        DHT_STRIP_PHASE1_FLAGS (prebuf);
+        DHT_STACK_UNWIND (setattr, frame, op_ret, op_errno,
+                          prebuf, postbuf, xdata);
+
+        return 0;
+}
+
+int
+dht_setattr2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+        inode_t      *inode = NULL;
+
+        local = frame->local;
+
+        inode = (local->fd) ? local->fd->inode : local->loc.inode;
+
+        dht_inode_ctx_get1 (this, inode, &subvol);
+
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+        if (local->fop == GF_FOP_SETATTR) {
+                STACK_WIND (frame, dht_file_setattr_cbk, subvol,
+                            subvol->fops->setattr, &local->loc,
+                            &local->rebalance.stbuf, local->rebalance.flags,
+                            NULL);
+        } else {
+                STACK_WIND (frame, dht_file_setattr_cbk, subvol,
+                            subvol->fops->fsetattr, local->fd,
+                            &local->rebalance.stbuf, local->rebalance.flags,
+                            NULL);
+        }
+
+        return 0;
+}
+
+
+/* Keep the existing code same for all the cases other than regular file */
+int
+dht_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int op_ret, int op_errno, struct iatt *statpre,
+                 struct iatt *statpost, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
+
+
+        local = frame->local;
+        prev = cookie;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
+
+                dht_iatt_merge (this, &local->prebuf, statpre, prev->this);
+                dht_iatt_merge (this, &local->stbuf, statpost, prev->this);
+
+                local->op_ret = 0;
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_UNWIND (setattr, frame, local->op_ret, local->op_errno,
+                                  &local->prebuf, &local->stbuf, xdata);
+
+        return 0;
+}
+
+
+int
+dht_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        dht_layout_t *layout = NULL;
+        dht_local_t  *local  = NULL;
+        int           op_errno = -1;
+        int           i = -1;
+        int           call_cnt = 0;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_SETATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (!layout_is_sane (layout)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "layout is not sane for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (IA_ISREG (loc->inode->ia_type)) {
+                /* in the regular file _cbk(), we need to check for
+                   migration possibilities */
+                local->rebalance.stbuf = *stbuf;
+                local->rebalance.flags = valid;
+                local->call_cnt = 1;
+                subvol = local->cached_subvol;
+
+                STACK_WIND (frame, dht_file_setattr_cbk, subvol,
+                            subvol->fops->setattr,
+                            loc, stbuf, valid, xdata);
+
+                return 0;
+        }
+
+        local->call_cnt = call_cnt = layout->cnt;
+
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_setattr_cbk,
+                            layout->list[i].xlator,
+                            layout->list[i].xlator->fops->setattr,
+                            loc, stbuf, valid, xdata);
+        }
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
+              int32_t valid, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        dht_layout_t *layout = NULL;
+        dht_local_t  *local  = NULL;
+        int           op_errno = -1;
+        int           i = -1;
+        int           call_cnt = 0;
+
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FSETATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (!layout_is_sane (layout)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "layout is not sane for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (IA_ISREG (fd->inode->ia_type)) {
+                /* in the regular file _cbk(), we need to check for
+                   migration possibilities */
+                local->rebalance.stbuf = *stbuf;
+                local->rebalance.flags = valid;
+                local->call_cnt = 1;
+                subvol = local->cached_subvol;
+
+                STACK_WIND (frame, dht_file_setattr_cbk, subvol,
+                            subvol->fops->fsetattr,
+                            fd, stbuf, valid, xdata);
+
+                return 0;
+        }
+
+        local->call_cnt = call_cnt = layout->cnt;
+
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_setattr_cbk,
+                            layout->list[i].xlator,
+                            layout->list[i].xlator->fops->fsetattr,
+                            fd, stbuf, valid, xdata);
+        }
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
diff --git a/xlators/cluster/dht/src/dht-layout.c b/xlators/cluster/dht/src/dht-layout.c
index 5ff2bdbe8..38e9970a7 100644
--- a/xlators/cluster/dht/src/dht-layout.c
+++ b/xlators/cluster/dht/src/dht-layout.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -38,28 +29,29 @@
 dht_layout_t *
 dht_layout_new (xlator_t *this, int cnt)
 {
-	dht_layout_t *layout = NULL;
+        dht_layout_t *layout = NULL;
         dht_conf_t   *conf = NULL;
 
 
         conf = this->private;
 
-	layout = GF_CALLOC (1, layout_size (cnt),
+        layout = GF_CALLOC (1, layout_size (cnt),
                             gf_dht_mt_dht_layout_t);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto out;
-	}
+        if (!layout) {
+                goto out;
+        }
 
         layout->type = DHT_HASH_TYPE_DM;
-	layout->cnt = cnt;
-        if (conf)
+        layout->cnt = cnt;
+
+        if (conf) {
+                layout->spread_cnt = conf->dir_spread_cnt;
                 layout->gen = conf->gen;
+        }
 
         layout->ref = 1;
 out:
-	return layout;
+        return layout;
 }
 
 
@@ -67,21 +59,22 @@ dht_layout_t *
 dht_layout_get (xlator_t *this, inode_t *inode)
 {
         dht_conf_t   *conf = NULL;
-        uint64_t      layout_int = 0;
         dht_layout_t *layout = NULL;
-        int           ret    = -1;
 
         conf = this->private;
+        if (!conf)
+                goto out;
+
         LOCK (&conf->layout_lock);
         {
-                ret = inode_ctx_get (inode, this, &layout_int);
-                if (ret == 0) {
-                        layout = (dht_layout_t *) (unsigned long) layout_int;
+                dht_inode_ctx_layout_get (inode, this, &layout);
+                if (layout) {
                         layout->ref++;
                 }
         }
         UNLOCK (&conf->layout_lock);
 
+out:
         return layout;
 }
 
@@ -93,24 +86,24 @@ dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout)
         int           oldret = -1;
         int           ret = 0;
         dht_layout_t *old_layout;
-        uint64_t      old_layout_int;
 
         conf = this->private;
+        if (!conf)
+                goto out;
+
         LOCK (&conf->layout_lock);
         {
-                oldret = inode_ctx_get (inode, this, &old_layout_int);
-
+                oldret = dht_inode_ctx_layout_get (inode, this, &old_layout);
                 layout->ref++;
-                ret = inode_ctx_put (inode, this, (uint64_t) (unsigned long)
-                                     layout);
+                dht_inode_ctx_layout_set (inode, this, layout);
         }
         UNLOCK (&conf->layout_lock);
 
-        if (oldret == 0) {
-                old_layout = (dht_layout_t *) (unsigned long) old_layout_int;
+        if (!oldret) {
                 dht_layout_unref (this, old_layout);
         }
 
+out:
         return ret;
 }
 
@@ -121,10 +114,11 @@ dht_layout_unref (xlator_t *this, dht_layout_t *layout)
         dht_conf_t  *conf = NULL;
         int          ref = 0;
 
-        if (layout->preset)
+        if (!layout || layout->preset || !this->private)
                 return;
 
         conf = this->private;
+
         LOCK (&conf->layout_lock);
         {
                 ref = --layout->ref;
@@ -141,7 +135,7 @@ dht_layout_ref (xlator_t *this, dht_layout_t *layout)
 {
         dht_conf_t  *conf = NULL;
 
-        if (layout->preset)
+        if (layout->preset || !this->private)
                 return layout;
 
         conf = this->private;
@@ -158,528 +152,611 @@ dht_layout_ref (xlator_t *this, dht_layout_t *layout)
 xlator_t *
 dht_layout_search (xlator_t *this, dht_layout_t *layout, const char *name)
 {
-	uint32_t   hash = 0;
+        uint32_t   hash = 0;
         xlator_t  *subvol = NULL;
-	int        i = 0;
-	int        ret = 0;
+        int        i = 0;
+        int        ret = 0;
 
 
-	ret = dht_hash_compute (layout->type, name, &hash);
-	if (ret != 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"hash computation failed for type=%d name=%s",
-			layout->type, name);
-		goto out;
-	}
+        ret = dht_hash_compute (this, layout->type, name, &hash);
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "hash computation failed for type=%d name=%s",
+                        layout->type, name);
+                goto out;
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].start <= hash
-		    && layout->list[i].stop >= hash) {
-			subvol = layout->list[i].xlator;
-			break;
-		}
-	}
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].start <= hash
+                    && layout->list[i].stop >= hash) {
+                        subvol = layout->list[i].xlator;
+                        break;
+                }
+        }
 
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume for hash (value) = %u", hash);
-	}
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "no subvolume for hash (value) = %u", hash);
+        }
 
 out:
-	return subvol;
+        return subvol;
 }
 
 
 dht_layout_t *
 dht_layout_for_subvol (xlator_t *this, xlator_t *subvol)
 {
-	dht_conf_t   *conf = NULL;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-
+        dht_conf_t   *conf = NULL;
+        dht_layout_t *layout = NULL;
+        int           i = 0;
 
-	conf = this->private;
+        conf = this->private;
+        if (!conf)
+                goto out;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (conf->subvolumes[i] == subvol) {
-			layout = conf->file_layouts[i];
-			break;
-		}
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == subvol) {
+                        layout = conf->file_layouts[i];
+                        break;
+                }
+        }
 
-	return layout;
+out:
+        return layout;
 }
 
 
 int
 dht_layouts_init (xlator_t *this, dht_conf_t *conf)
 {
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-	int           ret = -1;
-	
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           ret = -1;
 
-	conf->file_layouts = GF_CALLOC (conf->subvolume_cnt,
-				        sizeof (dht_layout_t *),
+        if (!conf)
+                goto out;
+
+        conf->file_layouts = GF_CALLOC (conf->subvolume_cnt,
+                                        sizeof (dht_layout_t *),
                                         gf_dht_mt_dht_layout_t);
-	if (!conf->file_layouts) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto out;
-	}
+        if (!conf->file_layouts) {
+                goto out;
+        }
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		layout = dht_layout_new (this, 1);
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                layout = dht_layout_new (this, 1);
 
-		if (!layout) {
-			goto out;
-		}
+                if (!layout) {
+                        goto out;
+                }
 
-		layout->preset = 1;
+                layout->preset = 1;
 
-		layout->list[0].xlator = conf->subvolumes[i];
+                layout->list[0].xlator = conf->subvolumes[i];
 
-		conf->file_layouts[i] = layout;
-	}
+                conf->file_layouts[i] = layout;
+        }
 
-	ret = 0;
+        ret = 0;
 out:
-	return ret;
+        return ret;
 }
 
 
 int
 dht_disk_layout_extract (xlator_t *this, dht_layout_t *layout,
-			 int pos, int32_t **disk_layout_p)
+                         int pos, int32_t **disk_layout_p)
 {
-	int      ret = -1;
-	int32_t *disk_layout = NULL;
+        int      ret = -1;
+        int32_t *disk_layout = NULL;
 
-	disk_layout = GF_CALLOC (5, sizeof (int),
+        disk_layout = GF_CALLOC (5, sizeof (int),
                                  gf_dht_mt_int32_t);
-	if (!disk_layout) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto out;
-	}
+        if (!disk_layout) {
+                goto out;
+        }
+
+        disk_layout[0] = hton32 (1);
+        disk_layout[1] = hton32 (layout->type);
+        disk_layout[2] = hton32 (layout->list[pos].start);
+        disk_layout[3] = hton32 (layout->list[pos].stop);
 
-	disk_layout[0] = hton32 (1);
-	disk_layout[1] = hton32 (layout->type);
-	disk_layout[2] = hton32 (layout->list[pos].start);
-	disk_layout[3] = hton32 (layout->list[pos].stop);
+        if (disk_layout_p)
+                *disk_layout_p = disk_layout;
+        else
+                GF_FREE (disk_layout);
 
-	if (disk_layout_p)
-		*disk_layout_p = disk_layout;
-	ret = 0;
+        ret = 0;
 
 out:
-	return ret;
+        return ret;
 }
 
 
 int
 dht_disk_layout_merge (xlator_t *this, dht_layout_t *layout,
-		       int pos, void *disk_layout_raw)
+		       int pos, void *disk_layout_raw, int disk_layout_len)
 {
-	int      cnt = 0;
-	int      type = 0;
-	int      start_off = 0;
-	int      stop_off = 0;
+        int      cnt = 0;
+        int      type = 0;
+        int      start_off = 0;
+        int      stop_off = 0;
         int      disk_layout[4];
 
-	/* TODO: assert disk_layout_ptr is of required length */
+	if (!disk_layout_raw) {
+		gf_log (this->name, GF_LOG_CRITICAL,
+                        "error no layout on disk for merge");
+		return -1;
+	}
 
-        memcpy (disk_layout, disk_layout_raw, sizeof (disk_layout));
+	GF_ASSERT (disk_layout_len == sizeof (disk_layout));
 
-	cnt  = ntoh32 (disk_layout[0]);
-	if (cnt != 1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"disk layout has invalid count %d", cnt);
+        memcpy (disk_layout, disk_layout_raw, disk_layout_len);
+
+        cnt  = ntoh32 (disk_layout[0]);
+        if (cnt != 1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "disk layout has invalid count %d", cnt);
+                return -1;
+        }
+
+        type = ntoh32 (disk_layout[1]);
+	switch (type) {
+        case DHT_HASH_TYPE_DM_USER:
+                gf_log (this->name, GF_LOG_DEBUG, "found user-set layout");
+                layout->type = type;
+                /* Fall through. */
+	case DHT_HASH_TYPE_DM:
+		break;
+        default:
+		gf_log (this->name, GF_LOG_CRITICAL,
+			"Catastrophic error layout with unknown type found %d",
+			disk_layout[1]);
 		return -1;
 	}
 
-	/* TODO: assert type is compatible */
-	type      = ntoh32 (disk_layout[1]);
-	start_off = ntoh32 (disk_layout[2]);
-	stop_off  = ntoh32 (disk_layout[3]);
+        start_off = ntoh32 (disk_layout[2]);
+        stop_off  = ntoh32 (disk_layout[3]);
 
-	layout->list[pos].start = start_off;
-	layout->list[pos].stop  = stop_off;
+        layout->list[pos].start = start_off;
+        layout->list[pos].stop  = stop_off;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"merged to layout: %u - %u (type %d) from %s",
-		start_off, stop_off, type,
-		layout->list[pos].xlator->name);
+        gf_log (this->name, GF_LOG_TRACE,
+                "merged to layout: %u - %u (type %d) from %s",
+                start_off, stop_off, type,
+                layout->list[pos].xlator->name);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_layout_merge (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
-		  int op_ret, int op_errno, dict_t *xattr)
+                  int op_ret, int op_errno, dict_t *xattr)
 {
-	int      i     = 0;
-	int      ret   = -1;
-	int      err   = -1;
-	void    *disk_layout_raw = NULL;
-
-
-	if (op_ret != 0) {
-		err = op_errno;
-	}
+        int         i     = 0;
+        int         ret   = -1;
+        int         err   = -1;
+        void       *disk_layout_raw = NULL;
+        int         disk_layout_len = 0;
+        dht_conf_t *conf = this->private;
+
+        if (op_ret != 0) {
+                err = op_errno;
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].xlator == NULL) {
-			layout->list[i].err    = err;
-			layout->list[i].xlator = subvol;
-			break;
-		}
-	}
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].xlator == NULL) {
+                        layout->list[i].err    = err;
+                        layout->list[i].xlator = subvol;
+                        break;
+                }
+        }
 
-	if (op_ret != 0) {
-		ret = 0;
-		goto out;
-	}
+        if (op_ret != 0) {
+                ret = 0;
+                goto out;
+        }
 
-	if (xattr) {
-		/* during lookup and not mkdir */
-		ret = dict_get_ptr (xattr, "trusted.glusterfs.dht",
-				    &disk_layout_raw);
-	}
+        if (xattr) {
+                /* during lookup and not mkdir */
+                ret = dict_get_ptr_and_len (xattr, conf->xattr_name,
+					    &disk_layout_raw, &disk_layout_len);
+        }
 
-	if (ret != 0) {
-		layout->list[i].err = -1;
-		gf_log (this->name, GF_LOG_TRACE,
-			"missing disk layout on %s. err = %d",
-			subvol->name, err);
-		ret = 0;
-		goto out;
-	}
+        if (ret != 0) {
+                layout->list[i].err = 0;
+                gf_log (this->name, GF_LOG_TRACE,
+                        "missing disk layout on %s. err = %d",
+                        subvol->name, err);
+                ret = 0;
+                goto out;
+        }
 
-	ret = dht_disk_layout_merge (this, layout, i, disk_layout_raw);
-	if (ret != 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"layout merge from subvolume %s failed",
-			subvol->name);
-		goto out;
-	}
-	layout->list[i].err = 0;
+        ret = dht_disk_layout_merge (this, layout, i, disk_layout_raw,
+				     disk_layout_len);
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "layout merge from subvolume %s failed",
+                        subvol->name);
+                goto out;
+        }
+        layout->list[i].err = 0;
 
 out:
-	return ret;
+        return ret;
 }
 
 
 void
 dht_layout_entry_swap (dht_layout_t *layout, int i, int j)
 {
-	uint32_t  start_swap = 0;
-	uint32_t  stop_swap = 0;
-	xlator_t *xlator_swap = 0;
-	int       err_swap = 0;
-
-
-	start_swap  = layout->list[i].start;
-	stop_swap   = layout->list[i].stop;
-	xlator_swap = layout->list[i].xlator;
-	err_swap    = layout->list[i].err;
-
-	layout->list[i].start  = layout->list[j].start;
-	layout->list[i].stop   = layout->list[j].stop;
-	layout->list[i].xlator = layout->list[j].xlator;
-	layout->list[i].err    = layout->list[j].err;
-
-	layout->list[j].start  = start_swap;
-	layout->list[j].stop   = stop_swap;
-	layout->list[j].xlator = xlator_swap;
-	layout->list[j].err    = err_swap;
+        uint32_t  start_swap = 0;
+        uint32_t  stop_swap = 0;
+        xlator_t *xlator_swap = 0;
+        int       err_swap = 0;
+
+        start_swap  = layout->list[i].start;
+        stop_swap   = layout->list[i].stop;
+        xlator_swap = layout->list[i].xlator;
+        err_swap    = layout->list[i].err;
+
+        layout->list[i].start  = layout->list[j].start;
+        layout->list[i].stop   = layout->list[j].stop;
+        layout->list[i].xlator = layout->list[j].xlator;
+        layout->list[i].err    = layout->list[j].err;
+
+        layout->list[j].start  = start_swap;
+        layout->list[j].stop   = stop_swap;
+        layout->list[j].xlator = xlator_swap;
+        layout->list[j].err    = err_swap;
+}
+
+void
+dht_layout_range_swap (dht_layout_t *layout, int i, int j)
+{
+        uint32_t  start_swap = 0;
+        uint32_t  stop_swap = 0;
+
+        start_swap  = layout->list[i].start;
+        stop_swap   = layout->list[i].stop;
+
+        layout->list[i].start  = layout->list[j].start;
+        layout->list[i].stop   = layout->list[j].stop;
+
+        layout->list[j].start  = start_swap;
+        layout->list[j].stop   = stop_swap;
 }
 
 int64_t
 dht_layout_entry_cmp_volname (dht_layout_t *layout, int i, int j)
 {
-	return (strcmp (layout->list[i].xlator->name, 
-			layout->list[j].xlator->name));
+        return (strcmp (layout->list[i].xlator->name,
+                        layout->list[j].xlator->name));
+}
+
+
+gf_boolean_t
+dht_is_subvol_in_layout (dht_layout_t *layout, xlator_t *xlator)
+{
+        int i = 0;
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (!strcmp (layout->list[i].xlator->name, xlator->name))
+                        return _gf_true;
+        }
+        return _gf_false;
 }
 
 int64_t
 dht_layout_entry_cmp (dht_layout_t *layout, int i, int j)
 {
-	int64_t diff = 0;
+        int64_t diff = 0;
 
-	if (layout->list[i].err || layout->list[j].err)
-		diff = layout->list[i].err - layout->list[j].err;
-	else
-		diff = (int64_t) layout->list[i].start
-			- (int64_t) layout->list[j].start;
+        /* swap zero'ed out layouts to front, if needed */
+        if (!layout->list[j].start && !layout->list[j].stop) {
+                diff = (int64_t) layout->list[i].stop
+                       - (int64_t) layout->list[j].stop;
+                       goto out;
+        }
+        if (layout->list[i].err || layout->list[j].err)
+                diff = layout->list[i].err - layout->list[j].err;
+        else
+                diff = (int64_t) layout->list[i].start
+                        - (int64_t) layout->list[j].start;
 
-	return diff;
+out:
+        return diff;
 }
 
 
 int
 dht_layout_sort (dht_layout_t *layout)
 {
-	int       i = 0;
-	int       j = 0;
-	int64_t   ret = 0;
-
-	/* TODO: O(n^2) -- bad bad */
-
-	for (i = 0; i < layout->cnt - 1; i++) {
-		for (j = i + 1; j < layout->cnt; j++) {
-			ret = dht_layout_entry_cmp (layout, i, j);
-			if (ret > 0)
-				dht_layout_entry_swap (layout, i, j);
-		}
-	}
+        int       i = 0;
+        int       j = 0;
+        int64_t   ret = 0;
 
-	return 0;
+        /* TODO: O(n^2) -- bad bad */
+
+        for (i = 0; i < layout->cnt - 1; i++) {
+                for (j = i + 1; j < layout->cnt; j++) {
+                        ret = dht_layout_entry_cmp (layout, i, j);
+                        if (ret > 0)
+                                dht_layout_entry_swap (layout, i, j);
+                }
+        }
+
+        return 0;
 }
 
 int
 dht_layout_sort_volname (dht_layout_t *layout)
 {
-	int       i = 0;
-	int       j = 0;
-	int64_t   ret = 0;
-
-	/* TODO: O(n^2) -- bad bad */
-
-	for (i = 0; i < layout->cnt - 1; i++) {
-		for (j = i + 1; j < layout->cnt; j++) {
-			ret = dht_layout_entry_cmp_volname (layout, i, j);
-			if (ret > 0)
-				dht_layout_entry_swap (layout, i, j);
-		}
-	}
+        int       i = 0;
+        int       j = 0;
+        int64_t   ret = 0;
+
+        /* TODO: O(n^2) -- bad bad */
 
-	return 0;
+        for (i = 0; i < layout->cnt - 1; i++) {
+                for (j = i + 1; j < layout->cnt; j++) {
+                        ret = dht_layout_entry_cmp_volname (layout, i, j);
+                        if (ret > 0)
+                                dht_layout_entry_swap (layout, i, j);
+                }
+        }
+
+        return 0;
 }
 
 
 int
 dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
-		      uint32_t *holes_p, uint32_t *overlaps_p,
-		      uint32_t *missing_p, uint32_t *down_p, uint32_t *misc_p)
+                      uint32_t *holes_p, uint32_t *overlaps_p,
+                      uint32_t *missing_p, uint32_t *down_p, uint32_t *misc_p,
+                      uint32_t *no_space_p)
 {
-	uint32_t    overlaps = 0;
-	uint32_t    missing  = 0;
-	uint32_t    down     = 0;
-	uint32_t    misc     = 0;
-	uint32_t    hole_cnt = 0;
-	uint32_t    overlap_cnt = 0;
-	int         i = 0;
-	int         ret = 0;
-	uint32_t    prev_stop = 0;
-	uint32_t    last_stop = 0;
-	char        is_virgin = 1;
-
-	/* TODO: explain WTF is happening */
-
-	last_stop = layout->list[0].start - 1;
-	prev_stop = last_stop;
-
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err) {
-			switch (layout->list[i].err) {
-			case -1:
-			case ENOENT:
-				missing++;
-				break;
-			case ENOTCONN:
-				down++;
-				break;
-			case ENOSPC:
-				down++;
-				break;
-			default:
-				misc++;
-			}
-			continue;
-		}
-
-		is_virgin = 0;
-
-		if ((prev_stop + 1) < layout->list[i].start) {
-			hole_cnt++;
-		}
-
-		if ((prev_stop + 1) > layout->list[i].start) {
-			overlap_cnt++;
-			overlaps += ((prev_stop + 1) - layout->list[i].start);
-		}
-		prev_stop = layout->list[i].stop;
-	}
+        uint32_t    overlaps = 0;
+        uint32_t    missing  = 0;
+        uint32_t    down     = 0;
+        uint32_t    misc     = 0;
+        uint32_t    hole_cnt = 0;
+        uint32_t    overlap_cnt = 0;
+        int         i = 0;
+        int         ret = 0;
+        uint32_t    prev_stop = 0;
+        uint32_t    last_stop = 0;
+        char        is_virgin = 1;
+        uint32_t    no_space  = 0;
+
+        /* TODO: explain what is happening */
+
+        last_stop = layout->list[0].start - 1;
+        prev_stop = last_stop;
+
+        for (i = 0; i < layout->cnt; i++) {
+                switch (layout->list[i].err) {
+                case -1:
+                case ENOENT:
+                        missing++;
+                        continue;
+                case ENOTCONN:
+                        down++;
+                        continue;
+                case ENOSPC:
+                        no_space++;
+                        continue;
+                case 0:
+                        /* if err == 0 and start == stop, then it is a non misc++;
+                         * participating subvolume(spread-cnt). Then, do not
+                         * check for anomalies. If start != stop, then treat it
+                         * as misc err */
+                        if (layout->list[i].start == layout->list[i].stop) {
+                                continue;
+                        }
+                        break;
+                default:
+                        misc++;
+                        continue;
+                 }
+
+                is_virgin = 0;
+
+                if ((prev_stop + 1) < layout->list[i].start) {
+                        hole_cnt++;
+                }
+
+                if ((prev_stop + 1) > layout->list[i].start) {
+                        overlap_cnt++;
+                        overlaps += ((prev_stop + 1) - layout->list[i].start);
+                }
+                prev_stop = layout->list[i].stop;
+        }
+
+        if ((last_stop - prev_stop) || is_virgin)
+                hole_cnt++;
 
-	if ((last_stop - prev_stop) || is_virgin)
-	    hole_cnt++;
+        if (holes_p)
+                *holes_p = hole_cnt;
 
-	if (holes_p)
-		*holes_p = hole_cnt;
+        if (overlaps_p)
+                *overlaps_p = overlap_cnt;
 
-	if (overlaps_p)
-		*overlaps_p = overlap_cnt;
+        if (missing_p)
+                *missing_p = missing;
 
-	if (missing_p)
-		*missing_p = missing;
+        if (down_p)
+                *down_p = down;
 
-	if (down_p)
-		*down_p = down;
+        if (misc_p)
+                *misc_p = misc;
 
-	if (misc_p)
-		*misc_p = misc;
+        if (no_space_p)
+                *no_space_p = no_space;
 
-	return ret;
+        return ret;
 }
 
 
 int
 dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout)
 {
-	int          ret   = 0;
-	int          i = 0;
-	uint32_t     holes = 0;
-	uint32_t     overlaps = 0;
-	uint32_t     missing = 0;
-	uint32_t     down = 0;
-	uint32_t     misc = 0;
-
-
-	ret = dht_layout_sort (layout);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"sort failed?! how the ....");
-		goto out;
-	}
+        int          ret   = 0;
+        int          i = 0;
+        uint32_t     holes = 0;
+        uint32_t     overlaps = 0;
+        uint32_t     missing = 0;
+        uint32_t     down = 0;
+        uint32_t     misc = 0;
+
+        ret = dht_layout_sort (layout);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "sort failed?! how the ....");
+                goto out;
+        }
 
-	ret = dht_layout_anomalies (this, loc, layout,
-				    &holes, &overlaps,
-				    &missing, &down, &misc);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"error while finding anomalies in %s -- not good news",
-			loc->path);
-		goto out;
-	}
+        ret = dht_layout_anomalies (this, loc, layout,
+                                    &holes, &overlaps,
+                                    &missing, &down, &misc, NULL);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "error while finding anomalies in %s -- not good news",
+                        loc->path);
+                goto out;
+        }
 
-	if (holes || overlaps) {
-		if (missing == layout->cnt) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"directory %s looked up first time",
-				loc->path);
-		} else {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"found anomalies in %s. holes=%d overlaps=%d",
-				loc->path, holes, overlaps);
-		}
-		ret = 1;
-	}
+        if (holes || overlaps) {
+                if (missing == layout->cnt) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "directory %s looked up first time",
+                                loc->path);
+                } else {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "found anomalies in %s. holes=%d overlaps=%d",
+                                loc->path, holes, overlaps);
+                }
+                ret = -1;
+        }
+
+        for (i = 0; i < layout->cnt; i++) {
+                /* TODO During DHT selfheal rewrite (almost) find a better place
+                 * to detect this - probably in dht_layout_anomalies()
+                 */
+                if (layout->list[i].err > 0) {
+                        gf_log_callingfn (this->name, GF_LOG_DEBUG,
+                                          "path=%s err=%s on subvol=%s",
+                                          loc->path,
+                                          strerror (layout->list[i].err),
+                                          (layout->list[i].xlator ?
+                                           layout->list[i].xlator->name
+                                           : "<>"));
+                        if ((layout->list[i].err == ENOENT) && (ret >= 0)) {
+                                ret++;
+                        }
+                }
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-	/* TODO During DHT selfheal rewrite (almost) find a better place to 
-	 * detect this - probably in dht_layout_anomalies() 
-	 */
-		if (layout->list[i].err > 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"path=%s err=%s on subvol=%s",
-                                loc->path, strerror (layout->list[i].err),
-                                (layout->list[i].xlator ?
-                                 layout->list[i].xlator->name : "<>"));
-                        if (layout->list[i].err == ENOENT)
-                                ret = 1;
-		}
-	}
 
 out:
-	return ret;
+        return ret;
 }
 
-
 int
-dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
-			 loc_t *loc, dict_t *xattr)
+dht_dir_has_layout (dict_t *xattr, char *name)
 {
-	int       idx = 0;
-	int       pos = -1;
-	int       ret = 0;
-        int       err = 0;
-        int       dict_ret = 0;
-	int32_t   disk_layout[4];
+
         void     *disk_layout_raw = NULL;
-	int32_t   count = -1;
-	uint32_t  start_off = -1;
-	uint32_t  stop_off = -1;
 
+        return dict_get_ptr (xattr, name, &disk_layout_raw);
+}
 
-	for (idx = 0; idx < layout->cnt; idx++) {
-		if (layout->list[idx].xlator == subvol) {
-			pos = idx;
-			break;
-		}
-	}
-	
-	if (pos == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"%s - no layout info for subvolume %s",
-			loc->path, subvol->name);
-		ret = 1;
-		goto out;
-	}
+int
+dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
+                         loc_t *loc, dict_t *xattr)
+{
+        int         idx = 0;
+        int         pos = -1;
+        int         ret = 0;
+        int         err = 0;
+        int         dict_ret = 0;
+        int32_t     disk_layout[4];
+        void       *disk_layout_raw = NULL;
+        int32_t     count = -1;
+        uint32_t    start_off = -1;
+        uint32_t    stop_off = -1;
+        dht_conf_t *conf = this->private;
+
+
+        for (idx = 0; idx < layout->cnt; idx++) {
+                if (layout->list[idx].xlator == subvol) {
+                        pos = idx;
+                        break;
+                }
+        }
+
+        if (pos == -1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "%s - no layout info for subvolume %s",
+                        loc->path, subvol->name);
+                ret = 1;
+                goto out;
+        }
 
         err = layout->list[pos].err;
 
-	if (!xattr) {
+        if (!xattr) {
                 if (err == 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+                        gf_log (this->name, GF_LOG_INFO,
                                 "%s - xattr dictionary is NULL",
                                 loc->path);
                         ret = -1;
                 }
-		goto out;
-	}
+                goto out;
+        }
 
-	dict_ret = dict_get_ptr (xattr, "trusted.glusterfs.dht",
+        dict_ret = dict_get_ptr (xattr, conf->xattr_name,
                                  &disk_layout_raw);
 
-	if (dict_ret < 0) {
+        if (dict_ret < 0) {
                 if (err == 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+                        gf_log (this->name, GF_LOG_INFO,
                                 "%s - disk layout missing", loc->path);
                         ret = -1;
                 }
-		goto out;
-	}
+                goto out;
+        }
 
         memcpy (disk_layout, disk_layout_raw, sizeof (disk_layout));
 
-	count  = ntoh32 (disk_layout[0]);
-	if (count != 1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"%s - disk layout has invalid count %d",
-			loc->path, count);
-		ret = -1;
-		goto out;
-	}
+        count  = ntoh32 (disk_layout[0]);
+        if (count != 1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s - disk layout has invalid count %d",
+                        loc->path, count);
+                ret = -1;
+                goto out;
+        }
 
-	start_off = ntoh32 (disk_layout[2]);
-	stop_off  = ntoh32 (disk_layout[3]);
-	
-	if ((layout->list[pos].start != start_off)
-	    || (layout->list[pos].stop != stop_off)) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"subvol: %s; inode layout - %"PRId32" - %"PRId32"; "
-			"disk layout - %"PRId32" - %"PRId32,
-			layout->list[pos].xlator->name,
-			layout->list[pos].start, layout->list[pos].stop,
-			start_off, stop_off);
-		ret = 1;
-	} else {
-		ret = 0;
-	}
+        start_off = ntoh32 (disk_layout[2]);
+        stop_off  = ntoh32 (disk_layout[3]);
+
+        if ((layout->list[pos].start != start_off)
+            || (layout->list[pos].stop != stop_off)) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "subvol: %s; inode layout - %"PRIu32" - %"PRIu32"; "
+                        "disk layout - %"PRIu32" - %"PRIu32,
+                        layout->list[pos].xlator->name,
+                        layout->list[pos].start, layout->list[pos].stop,
+                        start_off, stop_off);
+                ret = 1;
+        } else {
+                ret = 0;
+        }
 out:
-	return ret;
+        return ret;
 }
 
 
@@ -691,19 +768,21 @@ dht_layout_preset (xlator_t *this, xlator_t *subvol, inode_t *inode)
         dht_conf_t   *conf = NULL;
 
         conf = this->private;
-
-	layout = dht_layout_for_subvol (this, subvol);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no pre-set layout for subvolume %s",
-			subvol ? subvol->name : "<nil>");
-		ret = -1;
-		goto out;
-	}
+        if (!conf)
+                goto out;
+
+        layout = dht_layout_for_subvol (this, subvol);
+        if (!layout) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no pre-set layout for subvolume %s",
+                        subvol ? subvol->name : "<nil>");
+                ret = -1;
+                goto out;
+        }
 
         LOCK (&conf->layout_lock);
         {
-                inode_ctx_put (inode, this, (uint64_t)(long)layout);
+                dht_inode_ctx_layout_set (inode, this, layout);
         }
         UNLOCK (&conf->layout_lock);
 
diff --git a/xlators/cluster/dht/src/dht-linkfile.c b/xlators/cluster/dht/src/dht-linkfile.c
index 1ecae5dfd..dbc9d0b3c 100644
--- a/xlators/cluster/dht/src/dht-linkfile.c
+++ b/xlators/cluster/dht/src/dht-linkfile.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -28,214 +19,310 @@
 #include "compat.h"
 #include "dht-common.h"
 
-
-
 int
-dht_linkfile_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int op_ret, int op_errno)
+dht_linkfile_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int op_ret, int op_errno,
+                         inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                         struct iatt *postparent)
 {
-	dht_local_t *local = NULL;
+        char          is_linkfile   = 0;
+        dht_conf_t   *conf          = NULL;
+        dht_local_t  *local         = NULL;
+        call_frame_t *prev          = NULL;
 
+        local = frame->local;
+        prev = cookie;
+        conf = this->private;
 
-	local = frame->local;
-	local->linkfile.linkfile_cbk (frame, cookie, this, op_ret, op_errno,
-				      local->linkfile.inode,
-				      &local->linkfile.stbuf, NULL, NULL);
+        if (op_ret)
+                goto out;
 
-	return 0;
+        is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                         conf->link_xattr_name);
+        if (!is_linkfile)
+                gf_log (this->name, GF_LOG_WARNING, "got non-linkfile %s:%s",
+                        prev->this->name, local->loc.path);
+out:
+        local->linkfile.linkfile_cbk (frame, cookie, this, op_ret, op_errno,
+                                      inode, stbuf, postparent, postparent,
+                                      xattr);
+        return 0;
 }
 
-
+#define is_equal(a, b) (a == b)
 int
 dht_linkfile_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int op_ret, int op_errno, inode_t *inode,
                          struct iatt *stbuf, struct iatt *preparent,
-                         struct iatt *postparent)
+                         struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev = NULL;
-	dict_t       *xattr = NULL;
-	data_t       *str_data = NULL;
-	int           ret = -1;
-
-	local = frame->local;
-	prev  = cookie;
-
-	if (op_ret == -1)
-		goto err;
-
-	xattr = get_new_dict ();
-	if (!xattr) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	local->linkfile.xattr = dict_ref (xattr);
-	local->linkfile.inode = inode_ref (inode);
-
-	str_data = str_to_data (local->linkfile.srcvol->name);
-	if (!str_data) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	ret = dict_set (xattr, "trusted.glusterfs.dht.linkto", str_data);
-	if (ret < 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"failed to initialize linkfile data");
-	}
-	str_data = NULL;
-
-	local->linkfile.stbuf = *stbuf;
-
-	STACK_WIND (frame, dht_linkfile_xattr_cbk,
-		    prev->this, prev->this->fops->setxattr,
-		    &local->linkfile.loc, local->linkfile.xattr, 0);
-
-	return 0;
-
-err:
-	if (str_data) {
-		data_destroy (str_data);
-		str_data = NULL;
-	}
-
+        dht_local_t  *local = NULL;
+        xlator_t     *subvol = NULL;
+        call_frame_t *prev = NULL;
+        dict_t       *xattrs = NULL;
+        dht_conf_t   *conf = NULL;
+        int           ret = -1;
+
+        local = frame->local;
+
+        if (!op_ret)
+                local->linked = _gf_true;
+
+        FRAME_SU_UNDO (frame, dht_local_t);
+
+        if (op_ret && (op_errno == EEXIST)) {
+                conf = this->private;
+                prev = cookie;
+                subvol = prev->this;
+                if (!subvol)
+                        goto out;
+                xattrs = dict_new ();
+                if (!xattrs)
+                        goto out;
+                ret = dict_set_uint32 (xattrs, conf->link_xattr_name, 256);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to set linkto key");
+                        goto out;
+                }
+
+                STACK_WIND (frame, dht_linkfile_lookup_cbk, subvol,
+                            subvol->fops->lookup, &local->loc, xattrs);
+                if (xattrs)
+                        dict_unref (xattrs);
+                return 0;
+        }
+out:
         local->linkfile.linkfile_cbk (frame, cookie, this, op_ret, op_errno,
-                                      inode, stbuf, preparent, postparent);
-	return 0;
+                                      inode, stbuf, preparent, postparent,
+                                      xdata);
+        if (xattrs)
+                dict_unref (xattrs);
+        return 0;
 }
 
 
 int
 dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
-		     xlator_t *tovol, xlator_t *fromvol, loc_t *loc)
+                     xlator_t *this,
+                     xlator_t *tovol, xlator_t *fromvol, loc_t *loc)
 {
-	dht_local_t *local = NULL;
+        dht_local_t *local = NULL;
         dict_t      *dict = NULL;
+        int          need_unref = 0;
         int          ret = 0;
+        dht_conf_t  *conf = this->private;
+
+        local = frame->local;
+        local->linkfile.linkfile_cbk = linkfile_cbk;
+        local->linkfile.srcvol = tovol;
+
+        local->linked = _gf_false;
+
+        dict = local->params;
+        if (!dict) {
+                dict = dict_new ();
+                if (!dict)
+                        goto out;
+                need_unref = 1;
+        }
+
+        if (!uuid_is_null (local->gfid)) {
+                ret = dict_set_static_bin (dict, "gfid-req", local->gfid, 16);
+                if (ret)
+                        gf_log ("dht-linkfile", GF_LOG_INFO,
+                                "%s: gfid set failed", loc->path);
+        }
+
+        ret = dict_set_str (dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
+        if (ret)
+                gf_log ("dht-linkfile", GF_LOG_INFO,
+                        "%s: internal-fop set failed", loc->path);
 
-	local = frame->local;
-	local->linkfile.linkfile_cbk = linkfile_cbk;
-	local->linkfile.srcvol = tovol;
-	loc_copy (&local->linkfile.loc, loc);
+        ret = dict_set_str (dict, conf->link_xattr_name, tovol->name);
 
-        dict = dict_new ();
-        if (!dict)
+        if (ret < 0) {
+                gf_log (frame->this->name, GF_LOG_INFO,
+                        "%s: failed to initialize linkfile data",
+                        loc->path);
                 goto out;
+        }
 
-        ret = dict_set_static_bin (dict, "gfid-req", loc->inode->gfid, 16);
-        if (ret)
-                gf_log ("dht-linkfile", GF_LOG_DEBUG, "gfid set failed");
+        local->link_subvol = fromvol;
+        /* Always create as root:root. dht_linkfile_attr_heal fixes the
+         * ownsership */
+        FRAME_SU_DO (frame, dht_local_t);
+        STACK_WIND (frame, dht_linkfile_create_cbk,
+                    fromvol, fromvol->fops->mknod, loc,
+                    S_IFREG | DHT_LINKFILE_MODE, 0, 0, dict);
 
-	STACK_WIND (frame, dht_linkfile_create_cbk,
-		    fromvol, fromvol->fops->mknod, loc,
-		    S_IFREG | DHT_LINKFILE_MODE, 0, dict);
-
-        if (dict)
+        if (need_unref && dict)
                 dict_unref (dict);
 
         return 0;
 out:
-	local->linkfile.linkfile_cbk (frame, NULL, frame->this, -1, ENOMEM,
-				      loc->inode, NULL, NULL, NULL);
-	return 0;
+        local->linkfile.linkfile_cbk (frame, NULL, frame->this, -1, ENOMEM,
+                                      loc->inode, NULL, NULL, NULL, NULL);
+
+        if (need_unref && dict)
+                dict_unref (dict);
+
+        return 0;
 }
 
 
 int
 dht_linkfile_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			 int32_t op_ret, int32_t op_errno,
-                         struct iatt *preparent, struct iatt *postparent)
+                         int32_t op_ret, int32_t op_errno,
+                         struct iatt *preparent, struct iatt *postparent,
+                         dict_t *xdata)
 {
-	dht_local_t   *local = NULL;
-	call_frame_t  *prev = NULL;
-	xlator_t      *subvol = NULL;
+        dht_local_t   *local = NULL;
+        call_frame_t  *prev = NULL;
+        xlator_t      *subvol = NULL;
 
-	local = frame->local;
-	prev = cookie;
-	subvol = prev->this;
+        local = frame->local;
+        prev = cookie;
+        subvol = prev->this;
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unlinking linkfile %s on %s failed (%s)",
-			local->loc.path, subvol->name, strerror (op_errno));
-	}
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "unlinking linkfile %s on %s failed (%s)",
+                        local->loc.path, subvol->name, strerror (op_errno));
+        }
 
-	DHT_STACK_DESTROY (frame);
+        DHT_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_linkfile_unlink (call_frame_t *frame, xlator_t *this,
-		     xlator_t *subvol, loc_t *loc)
+                     xlator_t *subvol, loc_t *loc)
 {
-	call_frame_t *unlink_frame = NULL;
-	dht_local_t  *unlink_local = NULL;
-
-	unlink_frame = copy_frame (frame);
-	if (!unlink_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	unlink_local = dht_local_init (unlink_frame);
-	if (!unlink_local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	loc_copy (&unlink_local->loc, loc);
-
-	STACK_WIND (unlink_frame, dht_linkfile_unlink_cbk,
-		    subvol, subvol->fops->unlink,
-		    &unlink_local->loc);
-
-	return 0;
+        call_frame_t *unlink_frame = NULL;
+        dht_local_t  *unlink_local = NULL;
+
+        unlink_frame = copy_frame (frame);
+        if (!unlink_frame) {
+                goto err;
+        }
+
+        /* Using non-fop value here, as anyways, 'local->fop' is not used in
+           this particular case */
+        unlink_local = dht_local_init (unlink_frame, loc, NULL,
+                                       GF_FOP_MAXVALUE);
+        if (!unlink_local) {
+                goto err;
+        }
+
+        STACK_WIND (unlink_frame, dht_linkfile_unlink_cbk,
+                    subvol, subvol->fops->unlink,
+                    &unlink_local->loc, 0, NULL);
+
+        return 0;
 err:
-	if (unlink_frame)
-		DHT_STACK_DESTROY (unlink_frame);
+        if (unlink_frame)
+                DHT_STACK_DESTROY (unlink_frame);
 
-	return -1;
+        return -1;
 }
 
 
 xlator_t *
 dht_linkfile_subvol (xlator_t *this, inode_t *inode, struct iatt *stbuf,
-		     dict_t *xattr)
+                     dict_t *xattr)
 {
-	dht_conf_t *conf = NULL;
-	xlator_t   *subvol = NULL;
-	void       *volname = NULL;
-	int         i = 0, ret = 0;
-
+        dht_conf_t *conf = NULL;
+        xlator_t   *subvol = NULL;
+        void       *volname = NULL;
+        int         i = 0, ret = 0;
 
-	conf = this->private;
+        conf = this->private;
 
-	if (!xattr)
-		goto out;
+        if (!xattr)
+                goto out;
 
-	ret = dict_get_ptr (xattr, "trusted.glusterfs.dht.linkto", &volname);
+        ret = dict_get_ptr (xattr, conf->link_xattr_name, &volname);
 
-	if ((-1 == ret) || !volname)
-		goto out;
+        if ((-1 == ret) || !volname)
+                goto out;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (strcmp (conf->subvolumes[i]->name, (char *)volname) == 0) {
-			subvol = conf->subvolumes[i];
-			break;
-		}
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (strcmp (conf->subvolumes[i]->name, (char *)volname) == 0) {
+                        subvol = conf->subvolumes[i];
+                        break;
+                }
+        }
 
 out:
-	return subvol;
+        return subvol;
 }
 
+int
+dht_linkfile_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int op_ret, int op_errno, struct iatt *statpre,
+                          struct iatt *statpost, dict_t *xdata)
+{
+        dht_local_t *local = NULL;
+        loc_t *loc = NULL;
+
+        local = frame->local;
+        loc = &local->loc;
+
+        if (op_ret)
+                gf_log (this->name, GF_LOG_ERROR, "setattr of uid/gid on %s"
+                        " :<gfid:%s> failed (%s)",
+                        (loc->path? loc->path: "NULL"),
+                        uuid_utoa(local->gfid), strerror(op_errno));
+
+        DHT_STACK_DESTROY (frame);
+
+       return 0;
+}
+
+int
+dht_linkfile_attr_heal (call_frame_t *frame, xlator_t *this)
+{
+        int     ret = -1;
+        call_frame_t *copy = NULL;
+        dht_local_t  *local = NULL;
+        dht_local_t  *copy_local = NULL;
+        xlator_t     *subvol = NULL;
+        struct iatt   stbuf = {0,};
+
+        local = frame->local;
 
+        GF_VALIDATE_OR_GOTO ("dht", local, out);
+        GF_VALIDATE_OR_GOTO ("dht", local->link_subvol, out);
+
+        if (local->stbuf.ia_type == IA_INVAL)
+                return 0;
+
+        uuid_copy (local->loc.gfid, local->stbuf.ia_gfid);
+
+        copy = copy_frame (frame);
+
+        if (!copy)
+                goto out;
+
+       copy_local = dht_local_init (copy, &local->loc, NULL, 0);
+
+       if (!copy_local)
+               goto out;
+
+        stbuf = local->stbuf;
+        subvol = local->link_subvol;
+
+        copy->local = copy_local;
+
+        FRAME_SU_DO (copy, dht_local_t);
+
+        STACK_WIND (copy, dht_linkfile_setattr_cbk, subvol,
+                    subvol->fops->setattr, &copy_local->loc,
+                    &stbuf, (GF_SET_ATTR_UID | GF_SET_ATTR_GID), NULL);
+        ret = 0;
+out:
+        return ret;
+}
diff --git a/xlators/cluster/dht/src/dht-mem-types.h b/xlators/cluster/dht/src/dht-mem-types.h
index 4a7a8bd81..e893eb48f 100644
--- a/xlators/cluster/dht/src/dht-mem-types.h
+++ b/xlators/cluster/dht/src/dht-mem-types.h
@@ -1,21 +1,11 @@
-
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
 
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -29,14 +19,17 @@ enum gf_dht_mem_types_ {
         gf_dht_mt_dht_conf_t,
         gf_dht_mt_char,
         gf_dht_mt_int32_t,
-        gf_dht_mt_dht_local_t,
         gf_dht_mt_xlator_t,
         gf_dht_mt_dht_layout_t,
         gf_switch_mt_dht_conf_t,
         gf_switch_mt_dht_du_t,
         gf_switch_mt_switch_sched_array,
         gf_switch_mt_switch_struct,
+        gf_dht_mt_subvol_time,
+        gf_dht_mt_loc_t,
+        gf_defrag_info_mt,
+        gf_dht_mt_inode_ctx_t,
+        gf_dht_mt_ctx_stat_time_t,
         gf_dht_mt_end
 };
 #endif
-
diff --git a/xlators/cluster/dht/src/dht-rebalance.c b/xlators/cluster/dht/src/dht-rebalance.c
new file mode 100644
index 000000000..bcb19f23e
--- /dev/null
+++ b/xlators/cluster/dht/src/dht-rebalance.c
@@ -0,0 +1,1815 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "dht-common.h"
+#include "xlator.h"
+#include <fnmatch.h>
+
+#define GF_DISK_SECTOR_SIZE             512
+#define DHT_REBALANCE_PID               4242 /* Change it if required */
+#define DHT_REBALANCE_BLKSIZE           (128 * 1024)
+
+static int
+dht_write_with_holes (xlator_t *to, fd_t *fd, struct iovec *vec, int count,
+                      int32_t size, off_t offset, struct iobref *iobref)
+{
+        int i            = 0;
+        int ret          = -1;
+        int start_idx    = 0;
+        int tmp_offset   = 0;
+        int write_needed = 0;
+        int buf_len      = 0;
+        int size_pending = 0;
+        char *buf        = NULL;
+
+        /* loop through each vector */
+        for (i = 0; i < count; i++) {
+                buf = vec[i].iov_base;
+                buf_len = vec[i].iov_len;
+
+                for (start_idx = 0; (start_idx + GF_DISK_SECTOR_SIZE) <= buf_len;
+                     start_idx += GF_DISK_SECTOR_SIZE) {
+
+                        if (mem_0filled (buf + start_idx, GF_DISK_SECTOR_SIZE) != 0) {
+                                write_needed = 1;
+                                continue;
+                        }
+
+                        if (write_needed) {
+                                ret = syncop_write (to, fd, (buf + tmp_offset),
+                                                    (start_idx - tmp_offset),
+                                                    (offset + tmp_offset),
+                                                    iobref, 0);
+                                /* 'path' will be logged in calling function */
+                                if (ret < 0) {
+                                        gf_log (THIS->name, GF_LOG_WARNING,
+                                                "failed to write (%s)",
+                                                strerror (errno));
+                                        goto out;
+                                }
+
+                                write_needed = 0;
+                        }
+                        tmp_offset = start_idx + GF_DISK_SECTOR_SIZE;
+                }
+
+                if ((start_idx < buf_len) || write_needed) {
+                        /* This means, last chunk is not yet written.. write it */
+                        ret = syncop_write (to, fd, (buf + tmp_offset),
+                                            (buf_len - tmp_offset),
+                                            (offset + tmp_offset), iobref, 0);
+                        if (ret < 0) {
+                                /* 'path' will be logged in calling function */
+                                gf_log (THIS->name, GF_LOG_WARNING,
+                                        "failed to write (%s)",
+                                        strerror (errno));
+                                goto out;
+                        }
+                }
+
+                size_pending = (size - buf_len);
+                if (!size_pending)
+                        break;
+        }
+
+        ret = size;
+out:
+        return ret;
+
+}
+
+int32_t
+gf_defrag_handle_hardlink (xlator_t *this, loc_t *loc, dict_t  *xattrs,
+                           struct iatt *stbuf)
+{
+        int32_t                 ret             = -1;
+        xlator_t               *cached_subvol   = NULL;
+        xlator_t               *hashed_subvol   = NULL;
+        xlator_t               *linkto_subvol   = NULL;
+        data_t                 *data            = NULL;
+        struct iatt             iatt            = {0,};
+        int32_t                 op_errno        = 0;
+        dht_conf_t             *conf            = NULL;
+
+        GF_VALIDATE_OR_GOTO ("defrag", loc, out);
+        GF_VALIDATE_OR_GOTO ("defrag", loc->name, out);
+        GF_VALIDATE_OR_GOTO ("defrag", stbuf, out);
+        GF_VALIDATE_OR_GOTO ("defrag", this, out);
+        GF_VALIDATE_OR_GOTO ("defrag", xattrs, out);
+        GF_VALIDATE_OR_GOTO ("defrag", this->private, out);
+
+        conf = this->private;
+
+        if (uuid_is_null (loc->pargfid)) {
+                gf_log ("", GF_LOG_ERROR, "loc->pargfid is NULL for "
+                        "%s", loc->path);
+                goto out;
+        }
+
+        if (uuid_is_null (loc->gfid)) {
+                gf_log ("", GF_LOG_ERROR, "loc->gfid is NULL for "
+                        "%s", loc->path);
+                goto out;
+        }
+
+        cached_subvol = dht_subvol_get_cached (this, loc->inode);
+        if (!cached_subvol) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get cached subvol"
+                        " for %s on %s", loc->name, this->name);
+                goto out;
+        }
+
+        hashed_subvol = dht_subvol_get_hashed (this, loc);
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get hashed subvol"
+                        " for %s on %s", loc->name, this->name);
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_INFO, "Attempting to migrate hardlink %s "
+                "with gfid %s from %s -> %s", loc->name, uuid_utoa (loc->gfid),
+                cached_subvol->name, hashed_subvol->name);
+        data = dict_get (xattrs, conf->link_xattr_name);
+        /* set linkto on cached -> hashed if not present, else link it */
+        if (!data) {
+                ret = dict_set_str (xattrs, conf->link_xattr_name,
+                                    hashed_subvol->name);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+                                "linkto xattr in dict for %s", loc->name);
+                        goto out;
+                }
+
+                ret = syncop_setxattr (cached_subvol, loc, xattrs, 0);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Linkto setxattr "
+                                "failed %s -> %s (%s)", cached_subvol->name,
+                                loc->name, strerror (errno));
+                        goto out;
+                }
+                goto out;
+        } else {
+                linkto_subvol = dht_linkfile_subvol (this, NULL, NULL, xattrs);
+                if (!linkto_subvol) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                "linkto subvol for %s", loc->name);
+                } else {
+                        hashed_subvol = linkto_subvol;
+                }
+
+                ret = syncop_link (hashed_subvol, loc, loc);
+                if  (ret) {
+                        op_errno = errno;
+                        gf_log (this->name, GF_LOG_ERROR, "link of %s -> %s"
+                                " failed on  subvol %s (%s)", loc->name,
+                                uuid_utoa(loc->gfid),
+                                hashed_subvol->name, strerror (op_errno));
+                        if (op_errno != EEXIST)
+                                goto out;
+                }
+        }
+        ret = syncop_lookup (hashed_subvol, loc, NULL, &iatt, NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed lookup %s on %s (%s)"
+                        , loc->name, hashed_subvol->name, strerror (errno));
+                goto out;
+        }
+
+        if (iatt.ia_nlink == stbuf->ia_nlink) {
+                ret = dht_migrate_file (this, loc, cached_subvol, hashed_subvol,
+                                        GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS);
+                if (ret)
+                        goto out;
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+
+static inline int
+__is_file_migratable (xlator_t *this, loc_t *loc,
+                      struct iatt *stbuf, dict_t *xattrs, int flags)
+{
+        int ret = -1;
+
+        if (IA_ISDIR (stbuf->ia_type)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: migrate-file called on directory", loc->path);
+                ret = -1;
+                goto out;
+        }
+
+        if (flags == GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS) {
+                ret = 0;
+                goto out;
+        }
+        if (stbuf->ia_nlink > 1) {
+                /* support for decomission */
+                if (flags == GF_DHT_MIGRATE_HARDLINK) {
+                        ret = gf_defrag_handle_hardlink (this, loc,
+                                                         xattrs, stbuf);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "%s: failed to migrate file with link",
+                                        loc->path);
+                        }
+                } else {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: file has hardlinks", loc->path);
+                }
+                ret = ENOTSUP;
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+static inline int
+__dht_rebalance_create_dst_file (xlator_t *to, xlator_t *from, loc_t *loc, struct iatt *stbuf,
+                                 dict_t *dict, fd_t **dst_fd, dict_t *xattr)
+{
+        xlator_t    *this = NULL;
+        int          ret  = -1;
+        fd_t        *fd   = NULL;
+        struct iatt  new_stbuf = {0,};
+        dht_conf_t  *conf = NULL;
+
+        this = THIS;
+        conf = this->private;
+
+        ret = dict_set_static_bin (dict, "gfid-req", stbuf->ia_gfid, 16);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set gfid in dict for create", loc->path);
+                goto out;
+        }
+
+        ret = dict_set_str (dict, conf->link_xattr_name, from->name);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set gfid in dict for create", loc->path);
+                goto out;
+        }
+
+        fd = fd_create (loc->inode, DHT_REBALANCE_PID);
+        if (!fd) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: fd create failed (destination) (%s)",
+                        loc->path, strerror (errno));
+                ret = -1;
+                goto out;
+        }
+
+        ret = syncop_lookup (to, loc, NULL, &new_stbuf, NULL, NULL);
+        if (!ret) {
+                /* File exits in the destination, check if gfid matches */
+                if (uuid_compare (stbuf->ia_gfid, new_stbuf.ia_gfid) != 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "file %s exits in %s with different gfid",
+                                loc->path, to->name);
+                        fd_unref (fd);
+                        goto out;
+                }
+        }
+        if ((ret == -1) && (errno != ENOENT)) {
+                /* File exists in destination, but not accessible */
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "%s: failed to lookup file (%s)",
+                        loc->path, strerror (errno));
+                goto out;
+        }
+
+        /* Create the destination with LINKFILE mode, and linkto xattr,
+           if the linkfile already exists, it will just open the file */
+        ret = syncop_create (to, loc, O_RDWR, DHT_LINKFILE_MODE, fd,
+                             dict, &new_stbuf);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create %s on %s (%s)",
+                        loc->path, to->name, strerror (errno));
+                goto out;
+        }
+
+        ret = syncop_fsetxattr (to, fd, xattr, 0);
+        if (ret == -1)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set xattr on %s (%s)",
+                        loc->path, to->name, strerror (errno));
+
+        ret = syncop_ftruncate (to, fd, stbuf->ia_size);
+        if (ret < 0)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "ftruncate failed for %s on %s (%s)",
+                        loc->path, to->name, strerror (errno));
+
+        ret = syncop_fsetattr (to, fd, stbuf,
+                               (GF_SET_ATTR_UID | GF_SET_ATTR_GID),
+                                NULL, NULL);
+        if (ret < 0)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "chown failed for %s on %s (%s)",
+                        loc->path, to->name, strerror (errno));
+
+        if (dst_fd)
+                *dst_fd = fd;
+
+        /* success */
+        ret = 0;
+
+out:
+        return ret;
+}
+
+static inline int
+__dht_check_free_space (xlator_t *to, xlator_t *from, loc_t *loc,
+                        struct iatt *stbuf, int flag)
+{
+        struct statvfs  src_statfs = {0,};
+        struct statvfs  dst_statfs = {0,};
+        int             ret        = -1;
+        xlator_t       *this       = NULL;
+
+        uint64_t        src_statfs_blocks = 1;
+        uint64_t        dst_statfs_blocks = 1;
+
+        this = THIS;
+
+        ret = syncop_statfs (from, loc, &src_statfs);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get statfs of %s on %s (%s)",
+                        loc->path, from->name, strerror (errno));
+                goto out;
+        }
+
+        ret = syncop_statfs (to, loc, &dst_statfs);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get statfs of %s on %s (%s)",
+                        loc->path, to->name, strerror (errno));
+                goto out;
+        }
+
+        /* if force option is given, do not check for space @ dst.
+         * Check only if space is avail for the file */
+        if (flag != GF_DHT_MIGRATE_DATA)
+                goto check_avail_space;
+
+        /* Check:
+           During rebalance `migrate-data` - Destination subvol experiences
+           a `reduction` in 'blocks' of free space, at the same time source
+           subvol gains certain 'blocks' of free space. A valid check is
+           necessary here to avoid errorneous move to destination where
+           the space could be scantily available.
+         */
+        if (stbuf) {
+                dst_statfs_blocks = ((dst_statfs.f_bavail *
+                                      dst_statfs.f_bsize) /
+                                     GF_DISK_SECTOR_SIZE);
+                src_statfs_blocks = ((src_statfs.f_bavail *
+                                      src_statfs.f_bsize) /
+                                     GF_DISK_SECTOR_SIZE);
+                if ((dst_statfs_blocks - stbuf->ia_blocks) <
+                    (src_statfs_blocks + stbuf->ia_blocks)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "data movement attempted from node (%s) with"
+                                " higher disk space to a node (%s) with "
+                                "lesser disk space (%s)", from->name,
+                                to->name, loc->path);
+
+                        /* this is not a 'failure', but we don't want to
+                           consider this as 'success' too :-/ */
+                        ret = 1;
+                        goto out;
+                }
+        }
+check_avail_space:
+        if (((dst_statfs.f_bavail * dst_statfs.f_bsize) /
+              GF_DISK_SECTOR_SIZE) < stbuf->ia_blocks) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "data movement attempted from node (%s) with "
+                        "to node (%s) which does not have required free space"
+                        " for %s", from->name, to->name, loc->path);
+                ret = 1;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+static inline int
+__dht_rebalance_migrate_data (xlator_t *from, xlator_t *to, fd_t *src, fd_t *dst,
+                             uint64_t ia_size, int hole_exists)
+{
+        int            ret    = 0;
+        int            count  = 0;
+        off_t          offset = 0;
+        struct iovec  *vector = NULL;
+        struct iobref *iobref = NULL;
+        uint64_t       total  = 0;
+        size_t         read_size = 0;
+
+        /* if file size is '0', no need to enter this loop */
+        while (total < ia_size) {
+                read_size = (((ia_size - total) > DHT_REBALANCE_BLKSIZE) ?
+                             DHT_REBALANCE_BLKSIZE : (ia_size - total));
+                ret = syncop_readv (from, src, read_size,
+                                    offset, 0, &vector, &count, &iobref);
+                if (!ret || (ret < 0)) {
+                        break;
+                }
+
+                if (hole_exists)
+                        ret = dht_write_with_holes (to, dst, vector, count,
+                                                    ret, offset, iobref);
+                else
+                        ret = syncop_writev (to, dst, vector, count,
+                                             offset, iobref, 0);
+                if (ret < 0) {
+                        break;
+                }
+                offset += ret;
+                total += ret;
+
+                GF_FREE (vector);
+                if (iobref)
+                        iobref_unref (iobref);
+                iobref = NULL;
+                vector = NULL;
+        }
+        if (iobref)
+                iobref_unref (iobref);
+        GF_FREE (vector);
+
+        if (ret >= 0)
+                ret = 0;
+
+        return ret;
+}
+
+
+static inline int
+__dht_rebalance_open_src_file (xlator_t *from, xlator_t *to, loc_t *loc,
+                               struct iatt *stbuf, fd_t **src_fd)
+{
+        int          ret  = 0;
+        fd_t        *fd   = NULL;
+        dict_t      *dict = NULL;
+        xlator_t    *this = NULL;
+        struct iatt  iatt = {0,};
+        dht_conf_t  *conf = NULL;
+
+        this = THIS;
+        conf = this->private;
+
+        fd = fd_create (loc->inode, DHT_REBALANCE_PID);
+        if (!fd) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: fd create failed (source)", loc->path);
+                ret = -1;
+                goto out;
+        }
+
+        ret = syncop_open (from, loc, O_RDWR, fd);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to open file %s on %s (%s)",
+                        loc->path, from->name, strerror (errno));
+                goto out;
+        }
+
+        ret = -1;
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        ret = dict_set_str (dict, conf->link_xattr_name, to->name);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to set xattr in dict for %s (linkto:%s)",
+                        loc->path, to->name);
+                goto out;
+        }
+
+        /* Once the migration starts, the source should have 'linkto' key set
+           to show which is the target, so other clients can work around it */
+        ret = syncop_setxattr (from, loc, dict, 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to set xattr on %s in %s (%s)",
+                        loc->path, from->name, strerror (errno));
+                goto out;
+        }
+
+        /* mode should be (+S+T) to indicate migration is in progress */
+        iatt.ia_prot = stbuf->ia_prot;
+        iatt.ia_type = stbuf->ia_type;
+        iatt.ia_prot.sticky = 1;
+        iatt.ia_prot.sgid = 1;
+
+        ret = syncop_setattr (from, loc, &iatt, GF_SET_ATTR_MODE, NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to set mode on %s in %s (%s)",
+                        loc->path, from->name, strerror (errno));
+                goto out;
+        }
+
+        if (src_fd)
+                *src_fd = fd;
+
+        /* success */
+        ret = 0;
+out:
+        if (dict)
+                dict_unref (dict);
+
+        return ret;
+}
+
+int
+migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
+                       struct iatt *buf)
+{
+        int          ret      = -1;
+        dict_t      *rsp_dict = NULL;
+        dict_t      *dict     = NULL;
+        char        *link     = NULL;
+        struct iatt  stbuf    = {0,};
+        dht_conf_t  *conf     = this->private;
+
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        ret = dict_set_int32 (dict, conf->link_xattr_name, 256);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set 'linkto' key in dict", loc->path);
+                goto out;
+        }
+
+        /* check in the destination if the file is link file */
+        ret = syncop_lookup (to, loc, dict, &stbuf, &rsp_dict, NULL);
+        if ((ret == -1) && (errno != ENOENT)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s: lookup failed (%s)",
+                        loc->path, strerror (errno));
+                goto out;
+        }
+
+        /* we no more require this key */
+        dict_del (dict, conf->link_xattr_name);
+
+        /* file exists in target node, only if it is 'linkfile' its valid,
+           otherwise, error out */
+        if (!ret) {
+                if (!check_is_linkfile (loc->inode, &stbuf, rsp_dict,
+                                        conf->link_xattr_name)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: file exists in destination", loc->path);
+                        ret = -1;
+                        goto out;
+                }
+
+                /* as file is linkfile, delete it */
+                ret = syncop_unlink (to, loc);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to delete the linkfile (%s)",
+                                loc->path, strerror (errno));
+                        goto out;
+                }
+        }
+
+        /* Set the gfid of the source file in dict */
+        ret = dict_set_static_bin (dict, "gfid-req", buf->ia_gfid, 16);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set gfid in dict for create", loc->path);
+                goto out;
+        }
+
+        /* Create the file in target */
+        if (IA_ISLNK (buf->ia_type)) {
+                /* Handle symlinks separately */
+                ret = syncop_readlink (from, loc, &link, buf->ia_size);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: readlink on symlink failed (%s)",
+                                loc->path, strerror (errno));
+                        goto out;
+                }
+
+                ret = syncop_symlink (to, loc, link, dict, 0);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: creating symlink failed (%s)",
+                                loc->path, strerror (errno));
+                        goto out;
+                }
+
+                goto done;
+        }
+
+        ret = syncop_mknod (to, loc, st_mode_from_ia (buf->ia_prot,
+                                                      buf->ia_type),
+                            makedev (ia_major (buf->ia_rdev),
+                                     ia_minor (buf->ia_rdev)), dict, 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "%s: mknod failed (%s)",
+                        loc->path, strerror (errno));
+                goto out;
+        }
+
+done:
+        ret = syncop_setattr (to, loc, buf,
+                              (GF_SET_ATTR_UID | GF_SET_ATTR_GID |
+                               GF_SET_ATTR_MODE), NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to perform setattr on %s (%s)",
+                        loc->path, to->name, strerror (errno));
+        }
+
+        ret = syncop_unlink (from, loc);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING, "%s: unlink failed (%s)",
+                        loc->path, strerror (errno));
+
+out:
+        if (dict)
+                dict_unref (dict);
+
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        return ret;
+}
+
+/*
+  return values:
+
+   -1 : failure
+    0 : successfully migrated data
+    1 : not a failure, but we can't migrate data as of now
+*/
+int
+dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
+                  int flag)
+{
+        int             ret            = -1;
+        struct iatt     new_stbuf      = {0,};
+        struct iatt     stbuf          = {0,};
+        struct iatt     empty_iatt     = {0,};
+        ia_prot_t       src_ia_prot    = {0,};
+        fd_t           *src_fd         = NULL;
+        fd_t           *dst_fd         = NULL;
+        dict_t         *dict           = NULL;
+        dict_t         *xattr          = NULL;
+        dict_t         *xattr_rsp      = NULL;
+        int             file_has_holes = 0;
+        dht_conf_t     *conf           = this->private;
+
+        gf_log (this->name, GF_LOG_INFO, "%s: attempting to move from %s to %s",
+                loc->path, from->name, to->name);
+
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        ret = dict_set_int32 (dict, conf->link_xattr_name, 256);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set 'linkto' key in dict", loc->path);
+                goto out;
+        }
+
+        /* Phase 1 - Data migration is in progress from now on */
+        ret = syncop_lookup (from, loc, dict, &stbuf, &xattr_rsp, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s: lookup failed on %s (%s)",
+                        loc->path, from->name, strerror (errno));
+                goto out;
+        }
+
+        /* we no more require this key */
+        dict_del (dict, conf->link_xattr_name);
+
+        /* preserve source mode, so set the same to the destination */
+        src_ia_prot = stbuf.ia_prot;
+
+        /* Check if file can be migrated */
+        ret = __is_file_migratable (this, loc, &stbuf, xattr_rsp, flag);
+        if (ret)
+                goto out;
+
+        /* Take care of the special files */
+        if (!IA_ISREG (stbuf.ia_type)) {
+                /* Special files */
+                ret = migrate_special_files (this, from, to, loc, &stbuf);
+                goto out;
+        }
+
+        /* TODO: move all xattr related operations to fd based operations */
+        ret = syncop_listxattr (from, loc, &xattr);
+        if (ret == -1)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to get xattr from %s (%s)",
+                        loc->path, from->name, strerror (errno));
+
+        /* create the destination, with required modes/xattr */
+        ret = __dht_rebalance_create_dst_file (to, from, loc, &stbuf,
+                                               dict, &dst_fd, xattr);
+        if (ret)
+                goto out;
+
+        ret = __dht_check_free_space (to, from, loc, &stbuf, flag);
+        if (ret) {
+                goto out;
+        }
+
+        /* Open the source, and also update mode/xattr */
+        ret = __dht_rebalance_open_src_file (from, to, loc, &stbuf, &src_fd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to open %s on %s",
+                        loc->path, from->name);
+                goto out;
+        }
+
+
+        ret = syncop_fstat (from, src_fd, &stbuf);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to lookup %s on %s (%s)",
+                        loc->path, from->name, strerror (errno));
+                goto out;
+        }
+
+        /* Try to preserve 'holes' while migrating data */
+        if (stbuf.ia_size > (stbuf.ia_blocks * GF_DISK_SECTOR_SIZE))
+                file_has_holes = 1;
+
+        /* All I/O happens in this function */
+        ret = __dht_rebalance_migrate_data (from, to, src_fd, dst_fd,
+					    stbuf.ia_size, file_has_holes);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s: failed to migrate data",
+                        loc->path);
+                /* reset the destination back to 0 */
+                ret = syncop_ftruncate (to, dst_fd, 0);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: failed to reset target size back to 0 (%s)",
+                                loc->path, strerror (errno));
+                }
+
+                ret = -1;
+                goto out;
+        }
+
+        /* TODO: Sync the locks */
+
+        ret = syncop_fsync (to, dst_fd, 0);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to fsync on %s (%s)",
+                        loc->path, to->name, strerror (errno));
+
+
+        /* Phase 2 - Data-Migration Complete, Housekeeping updates pending */
+
+        ret = syncop_fstat (from, src_fd, &new_stbuf);
+        if (ret < 0) {
+                /* Failed to get the stat info */
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to fstat file %s on %s (%s)",
+                        loc->path, from->name, strerror (errno));
+                goto out;
+        }
+
+        /* source would have both sticky bit and sgid bit set, reset it to 0,
+           and set the source permission on destination, if it was not set
+           prior to setting rebalance-modes in source  */
+        if (!src_ia_prot.sticky)
+                new_stbuf.ia_prot.sticky = 0;
+
+        if (!src_ia_prot.sgid)
+                new_stbuf.ia_prot.sgid = 0;
+
+        /* TODO: if the source actually had sticky bit, or sgid bit set,
+           we are not handling it */
+
+        ret = syncop_fsetattr (to, dst_fd, &new_stbuf,
+                               (GF_SET_ATTR_UID | GF_SET_ATTR_GID |
+                                GF_SET_ATTR_MODE), NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to perform setattr on %s (%s)",
+                        loc->path, to->name, strerror (errno));
+                goto out;
+        }
+
+        /* Because 'futimes' is not portable */
+        ret = syncop_setattr (to, loc, &new_stbuf,
+                              (GF_SET_ATTR_MTIME | GF_SET_ATTR_ATIME),
+                              NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to perform setattr on %s (%s)",
+                        loc->path, to->name, strerror (errno));
+        }
+
+        /* Make the source as a linkfile first before deleting it */
+        empty_iatt.ia_prot.sticky = 1;
+        ret = syncop_fsetattr (from, src_fd, &empty_iatt,
+                               GF_SET_ATTR_MODE, NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,             \
+                        "%s: failed to perform setattr on %s (%s)",
+                        loc->path, from->name, strerror (errno));
+                goto out;
+        }
+
+       /* Free up the data blocks on the source node, as the whole
+           file is migrated */
+        ret = syncop_ftruncate (from, src_fd, 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to perform truncate on %s (%s)",
+                        loc->path, from->name, strerror (errno));
+        }
+
+        /* remove the 'linkto' xattr from the destination */
+        ret = syncop_fremovexattr (to, dst_fd, conf->link_xattr_name);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to perform removexattr on %s (%s)",
+                        loc->path, to->name, strerror (errno));
+        }
+
+        /* Do a stat and check the gfid before unlink */
+        ret = syncop_stat (from, loc, &empty_iatt);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to do a stat on %s (%s)",
+                        loc->path, from->name, strerror (errno));
+                goto out;
+        }
+
+        if (uuid_compare (empty_iatt.ia_gfid, loc->gfid) == 0) {
+                /* take out the source from namespace */
+                ret = syncop_unlink (from, loc);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to perform unlink on %s (%s)",
+                                loc->path, from->name, strerror (errno));
+                        goto out;
+                }
+        }
+
+        ret = syncop_lookup (this, loc, NULL, NULL, NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "%s: failed to lookup the file on subvolumes (%s)",
+                        loc->path, strerror (errno));
+        }
+
+        gf_log (this->name, GF_LOG_INFO,
+                "completed migration of %s from subvolume %s to %s",
+                loc->path, from->name, to->name);
+
+        ret = 0;
+out:
+        if (dict)
+                dict_unref (dict);
+
+        if (xattr)
+                dict_unref (xattr);
+        if (xattr_rsp)
+                dict_unref (xattr_rsp);
+
+        if (dst_fd)
+                syncop_close (dst_fd);
+        if (src_fd)
+                syncop_close (src_fd);
+
+        return ret;
+}
+
+static int
+rebalance_task (void *data)
+{
+        int           ret   = -1;
+        dht_local_t  *local = NULL;
+        call_frame_t *frame = NULL;
+
+        frame = data;
+
+        local = frame->local;
+
+        /* This function is 'synchrounous', hence if it returns,
+           we are done with the task */
+        ret = dht_migrate_file (THIS, &local->loc, local->rebalance.from_subvol,
+                                local->rebalance.target_node, local->flags);
+
+        return ret;
+}
+
+static int
+rebalance_task_completion (int op_ret, call_frame_t *sync_frame, void *data)
+{
+        int           ret        = -1;
+        uint64_t      layout_int = 0;
+        dht_layout_t *layout     = 0;
+        xlator_t     *this       = NULL;
+        dht_local_t  *local      = NULL;
+        int32_t       op_errno   = EINVAL;
+
+        this = THIS;
+        local = sync_frame->local;
+
+        if (!op_ret) {
+                /* Make sure we have valid 'layout' in inode ctx
+                   after the operation */
+                ret = inode_ctx_del (local->loc.inode, this, &layout_int);
+                if (!ret && layout_int) {
+                        layout = (dht_layout_t *)(long)layout_int;
+                        dht_layout_unref (this, layout);
+                }
+
+                ret = dht_layout_preset (this, local->rebalance.target_node,
+                                         local->loc.inode);
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set inode ctx", local->loc.path);
+        }
+
+        if (op_ret == -1) {
+                /* Failure of migration process, mostly due to write process.
+                   as we can't preserve the exact errno, lets say there was
+                   no space to migrate-data
+                */
+                op_errno = ENOSPC;
+        }
+
+        if (op_ret == 1) {
+                /* migration didn't happen, but is not a failure, let the user
+                   understand that he doesn't have permission to migrate the
+                   file.
+                */
+                op_ret = -1;
+                op_errno = EPERM;
+        }
+
+        DHT_STACK_UNWIND (setxattr, sync_frame, op_ret, op_errno, NULL);
+        return 0;
+}
+
+int
+dht_start_rebalance_task (xlator_t *this, call_frame_t *frame)
+{
+        int         ret     = -1;
+
+        ret = synctask_new (this->ctx->env, rebalance_task,
+                            rebalance_task_completion,
+                            frame, frame);
+        return ret;
+}
+
+int
+gf_listener_stop (xlator_t *this)
+{
+        glusterfs_ctx_t  *ctx = NULL;
+        cmd_args_t       *cmd_args = NULL;
+        int              ret = 0;
+
+        ctx = this->ctx;
+        GF_ASSERT (ctx);
+        cmd_args = &ctx->cmd_args;
+        if (cmd_args->sock_file) {
+                ret = unlink (cmd_args->sock_file);
+                if (ret && (ENOENT == errno)) {
+                        ret = 0;
+                }
+        }
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to unlink listener "
+                        "socket %s, error: %s", cmd_args->sock_file,
+                        strerror (errno));
+        }
+        return ret;
+}
+
+void
+dht_build_root_inode (xlator_t *this, inode_t **inode)
+{
+        inode_table_t    *itable        = NULL;
+        uuid_t            root_gfid     = {0, };
+
+        itable = inode_table_new (0, this);
+        if (!itable)
+                return;
+
+        root_gfid[15] = 1;
+        *inode = inode_find (itable, root_gfid);
+}
+
+void
+dht_build_root_loc (inode_t *inode, loc_t *loc)
+{
+        loc->path = "/";
+        loc->inode = inode;
+        loc->inode->ia_type = IA_IFDIR;
+        memset (loc->gfid, 0, 16);
+        loc->gfid[15] = 1;
+}
+
+
+/* return values: 1 -> error, bug ignore and continue
+                  0 -> proceed
+                 -1 -> error, handle it */
+int32_t
+gf_defrag_handle_migrate_error (int32_t op_errno, gf_defrag_info_t *defrag)
+{
+        /* if errno is not ENOSPC or ENOTCONN, we can still continue
+           with rebalance process */
+        if ((errno != ENOSPC) || (errno != ENOTCONN))
+                return 1;
+
+        if (errno == ENOTCONN) {
+                /* Most probably mount point went missing (mostly due
+                   to a brick down), say rebalance failure to user,
+                   let him restart it if everything is fine */
+                defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
+                return -1;
+        }
+
+        if (errno == ENOSPC) {
+                /* rebalance process itself failed, may be
+                   remote brick went down, or write failed due to
+                   disk full etc etc.. */
+                defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
+                return -1;
+        }
+
+        return 0;
+}
+
+static gf_boolean_t
+gf_defrag_pattern_match (gf_defrag_info_t *defrag, char *name, uint64_t size)
+{
+        gf_defrag_pattern_list_t *trav = NULL;
+        gf_boolean_t               match = _gf_false;
+        gf_boolean_t               ret = _gf_false;
+
+        GF_VALIDATE_OR_GOTO ("dht", defrag, out);
+
+        trav = defrag->defrag_pattern;
+        while (trav) {
+                if (!fnmatch (trav->path_pattern, name, FNM_NOESCAPE)) {
+                        match = _gf_true;
+                        break;
+                }
+                trav = trav->next;
+        }
+
+        if ((match == _gf_true) && (size >= trav->size))
+                ret = _gf_true;
+
+ out:
+        return ret;
+}
+
+/* We do a depth first traversal of directories. But before we move into
+ * subdirs, we complete the data migration of those directories whose layouts
+ * have been fixed
+ */
+
+int
+gf_defrag_migrate_data (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
+                        dict_t *migrate_data)
+{
+        int                      ret            = -1;
+        loc_t                    entry_loc      = {0,};
+        fd_t                    *fd             = NULL;
+        gf_dirent_t              entries;
+        gf_dirent_t             *tmp            = NULL;
+        gf_dirent_t             *entry          = NULL;
+        gf_boolean_t             free_entries   = _gf_false;
+        off_t                    offset         = 0;
+        dict_t                  *dict           = NULL;
+        struct iatt              iatt           = {0,};
+        int32_t                  op_errno       = 0;
+        char                    *uuid_str       = NULL;
+        uuid_t                   node_uuid      = {0,};
+        int                      readdir_operrno = 0;
+        struct timeval           dir_start      = {0,};
+        struct timeval           end            = {0,};
+        double                   elapsed        = {0,};
+        struct timeval           start          = {0,};
+        int32_t                  err            = 0;
+
+        gf_log (this->name, GF_LOG_INFO, "migrate data called on %s",
+                loc->path);
+        gettimeofday (&dir_start, NULL);
+
+        fd = fd_create (loc->inode, defrag->pid);
+        if (!fd) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to create fd");
+                goto out;
+        }
+
+        ret = syncop_opendir (this, loc, fd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to open dir %s",
+                        loc->path);
+                goto out;
+        }
+
+        INIT_LIST_HEAD (&entries.list);
+
+        while ((ret = syncop_readdirp (this, fd, 131072, offset, NULL,
+                                       &entries)) != 0) {
+
+                if (ret < 0) {
+
+                        gf_log (this->name, GF_LOG_ERROR, "Readdir returned %s."
+                                " Aborting migrate-data",
+                                strerror(readdir_operrno));
+                        goto out;
+                }
+
+                /* Need to keep track of ENOENT errno, that means, there is no
+                   need to send more readdirp() */
+                readdir_operrno = errno;
+
+                if (list_empty (&entries.list))
+                        break;
+
+                free_entries = _gf_true;
+
+                list_for_each_entry_safe (entry, tmp, &entries.list, list) {
+                        if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
+                                ret = 1;
+                                goto out;
+                        }
+
+                        offset = entry->d_off;
+
+                        if (!strcmp (entry->d_name, ".") ||
+                            !strcmp (entry->d_name, ".."))
+                                continue;
+
+                        if (IA_ISDIR (entry->d_stat.ia_type))
+                                continue;
+
+                        defrag->num_files_lookedup++;
+                        if (defrag->stats == _gf_true) {
+                                gettimeofday (&start, NULL);
+                        }
+                        if (defrag->defrag_pattern &&
+                            (gf_defrag_pattern_match (defrag, entry->d_name,
+                                                      entry->d_stat.ia_size)
+                             == _gf_false)) {
+                                continue;
+                        }
+                        loc_wipe (&entry_loc);
+                        ret =dht_build_child_loc (this, &entry_loc, loc,
+                                                  entry->d_name);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Child loc"
+                                        " build failed");
+                                goto out;
+                        }
+
+                        if (uuid_is_null (entry->d_stat.ia_gfid)) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+                                        " gfid not present", loc->path,
+                                         entry->d_name);
+                                continue;
+                        }
+
+                        uuid_copy (entry_loc.gfid, entry->d_stat.ia_gfid);
+
+                        if (uuid_is_null (loc->gfid)) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+                                        " gfid not present", loc->path,
+                                         entry->d_name);
+                                continue;
+                        }
+
+                        uuid_copy (entry_loc.pargfid, loc->gfid);
+
+                        entry_loc.inode->ia_type = entry->d_stat.ia_type;
+
+                        ret = syncop_lookup (this, &entry_loc, NULL, &iatt,
+                                             NULL, NULL);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s"
+                                        " lookup failed", entry_loc.path);
+                                continue;
+                        }
+
+                        ret = syncop_getxattr (this, &entry_loc, &dict,
+                                               GF_XATTR_NODE_UUID_KEY);
+                        if(ret < 0) {
+                                gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                        "get node-uuid for %s", entry_loc.path);
+                                continue;
+                        }
+
+                        ret = dict_get_str (dict, GF_XATTR_NODE_UUID_KEY,
+                                            &uuid_str);
+                        if(ret < 0) {
+                                gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                        "get node-uuid from dict for %s",
+                                        entry_loc.path);
+                                continue;
+                        }
+
+                        if (uuid_parse (uuid_str, node_uuid)) {
+                                gf_log (this->name, GF_LOG_ERROR, "uuid_parse "
+                                        "failed for %s", entry_loc.path);
+                                continue;
+                        }
+
+                        /* if file belongs to different node, skip migration
+                         * the other node will take responsibility of migration
+                         */
+                        if (uuid_compare (node_uuid, defrag->node_uuid)) {
+                                gf_log (this->name, GF_LOG_TRACE, "%s does not"
+                                        "belong to this node", entry_loc.path);
+                                continue;
+                        }
+
+                        uuid_str = NULL;
+
+                        dict_del (dict, GF_XATTR_NODE_UUID_KEY);
+
+
+                        /* if distribute is present, it will honor this key.
+                         * -1 is returned if distribute is not present or file
+                         * doesn't have a link-file. If file has link-file, the
+                         * path of link-file will be the value, and also that
+                         * guarantees that file has to be mostly migrated */
+
+                        ret = syncop_getxattr (this, &entry_loc, &dict,
+                                               GF_XATTR_LINKINFO_KEY);
+                        if (ret < 0) {
+                                gf_log (this->name, GF_LOG_TRACE, "failed to "
+                                        "get link-to key for %s",
+                                        entry_loc.path);
+                                continue;
+                        }
+
+                        ret = syncop_setxattr (this, &entry_loc, migrate_data,
+                                               0);
+                        if (ret) {
+                                err = op_errno;
+                                /* errno is overloaded. See
+                                 * rebalance_task_completion () */
+                                if (err != ENOSPC) {
+                                        gf_log (this->name, GF_LOG_DEBUG,
+                                                "migrate-data skipped for %s"
+                                                " due to space constraints",
+                                                entry_loc.path);
+                                        defrag->skipped +=1;
+                                } else{
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "migrate-data failed for %s",
+                                                entry_loc.path);
+                                        defrag->total_failures +=1;
+                                }
+                        }
+
+                        if (ret == -1) {
+                                op_errno = errno;
+                                ret = gf_defrag_handle_migrate_error (op_errno,
+                                                                      defrag);
+
+                                if (!ret)
+                                        gf_log (this->name, GF_LOG_DEBUG,
+                                                "migrate-data on %s failed: %s",
+                                                entry_loc.path,
+                                                strerror (op_errno));
+                                else if (ret == 1)
+                                        continue;
+                                else if (ret == -1)
+                                        goto out;
+                        }
+
+                        LOCK (&defrag->lock);
+                        {
+                                defrag->total_files += 1;
+                                defrag->total_data += iatt.ia_size;
+                        }
+                        UNLOCK (&defrag->lock);
+                        if (defrag->stats == _gf_true) {
+                                gettimeofday (&end, NULL);
+                                elapsed = (end.tv_sec - start.tv_sec) * 1e6 +
+                                          (end.tv_usec - start.tv_usec);
+                                gf_log (this->name, GF_LOG_INFO, "Migration of "
+                                        "file:%s size:%"PRIu64" bytes took %.2f"
+                                        "secs", entry_loc.path, iatt.ia_size,
+                                         elapsed/1e6);
+                        }
+                }
+
+                gf_dirent_free (&entries);
+                free_entries = _gf_false;
+                INIT_LIST_HEAD (&entries.list);
+
+                if (readdir_operrno == ENOENT)
+                        break;
+        }
+
+        gettimeofday (&end, NULL);
+        elapsed = (end.tv_sec - dir_start.tv_sec) * 1e6 +
+                  (end.tv_usec - dir_start.tv_usec);
+        gf_log (this->name, GF_LOG_INFO, "Migration operation on dir %s took "
+                "%.2f secs", loc->path, elapsed/1e6);
+        ret = 0;
+out:
+        if (free_entries)
+                gf_dirent_free (&entries);
+
+        loc_wipe (&entry_loc);
+
+        if (dict)
+                dict_unref(dict);
+
+        if (fd)
+                fd_unref (fd);
+        return ret;
+
+}
+
+
+int
+gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
+                  dict_t *fix_layout, dict_t *migrate_data)
+{
+        int                      ret            = -1;
+        loc_t                    entry_loc      = {0,};
+        fd_t                    *fd             = NULL;
+        gf_dirent_t              entries;
+        gf_dirent_t             *tmp            = NULL;
+        gf_dirent_t             *entry          = NULL;
+        gf_boolean_t             free_entries   = _gf_false;
+        dict_t                  *dict           = NULL;
+        off_t                    offset         = 0;
+        struct iatt              iatt           = {0,};
+        int                      readdirp_errno = 0;
+
+        ret = syncop_lookup (this, loc, NULL, &iatt, NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Lookup failed on %s",
+                        loc->path);
+                goto out;
+        }
+
+        if (defrag->cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) {
+                ret = gf_defrag_migrate_data (this, defrag, loc, migrate_data);
+                if (ret)
+                        goto out;
+        }
+
+        gf_log (this->name, GF_LOG_TRACE, "fix layout called on %s", loc->path);
+
+        fd = fd_create (loc->inode, defrag->pid);
+        if (!fd) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to create fd");
+                ret = -1;
+                goto out;
+        }
+
+        ret = syncop_opendir (this, loc, fd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to open dir %s",
+                        loc->path);
+                ret = -1;
+                goto out;
+        }
+
+        INIT_LIST_HEAD (&entries.list);
+        while ((ret = syncop_readdirp (this, fd, 131072, offset, NULL,
+                &entries)) != 0)
+        {
+
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "Readdir returned %s"
+                                ". Aborting fix-layout",strerror(errno));
+                        goto out;
+                }
+
+                /* Need to keep track of ENOENT errno, that means, there is no
+                   need to send more readdirp() */
+                readdirp_errno = errno;
+
+                if (list_empty (&entries.list))
+                        break;
+
+                free_entries = _gf_true;
+
+                list_for_each_entry_safe (entry, tmp, &entries.list, list) {
+                        if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
+                                ret = 1;
+                                goto out;
+                        }
+
+                        offset = entry->d_off;
+
+                        if (!strcmp (entry->d_name, ".") ||
+                            !strcmp (entry->d_name, ".."))
+                                continue;
+
+                        if (!IA_ISDIR (entry->d_stat.ia_type))
+                                continue;
+
+                        loc_wipe (&entry_loc);
+                        ret =dht_build_child_loc (this, &entry_loc, loc,
+                                                  entry->d_name);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Child loc"
+                                        " build failed");
+                                goto out;
+                        }
+
+                        if (uuid_is_null (entry->d_stat.ia_gfid)) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+                                        " gfid not present", loc->path,
+                                         entry->d_name);
+                                continue;
+                        }
+
+                        entry_loc.inode->ia_type = entry->d_stat.ia_type;
+
+                        uuid_copy (entry_loc.gfid, entry->d_stat.ia_gfid);
+                        if (uuid_is_null (loc->gfid)) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+                                        " gfid not present", loc->path,
+                                         entry->d_name);
+                                continue;
+                        }
+
+                        uuid_copy (entry_loc.pargfid, loc->gfid);
+
+                        ret = syncop_lookup (this, &entry_loc, NULL, &iatt,
+                                             NULL, NULL);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s"
+                                        " lookup failed", entry_loc.path);
+                                continue;
+                        }
+
+                        ret = syncop_setxattr (this, &entry_loc, fix_layout,
+                                               0);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Setxattr "
+                                        "failed for %s", entry_loc.path);
+                                defrag->defrag_status =
+                                GF_DEFRAG_STATUS_FAILED;
+                                defrag->total_failures ++;
+                                goto out;
+                        }
+                        ret = gf_defrag_fix_layout (this, defrag, &entry_loc,
+                                                    fix_layout, migrate_data);
+
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Fix layout "
+                                        "failed for %s", entry_loc.path);
+                                defrag->total_failures++;
+                                goto out;
+                        }
+
+                }
+                gf_dirent_free (&entries);
+                free_entries = _gf_false;
+                INIT_LIST_HEAD (&entries.list);
+                if (readdirp_errno == ENOENT)
+                        break;
+        }
+
+        ret = 0;
+out:
+        if (free_entries)
+                gf_dirent_free (&entries);
+
+        loc_wipe (&entry_loc);
+
+        if (dict)
+                dict_unref(dict);
+
+        if (fd)
+                fd_unref (fd);
+
+        return ret;
+
+}
+
+
+int
+gf_defrag_start_crawl (void *data)
+{
+        xlator_t                *this   = NULL;
+        dht_conf_t              *conf   = NULL;
+        gf_defrag_info_t        *defrag = NULL;
+        int                      ret    = -1;
+        loc_t                    loc    = {0,};
+        struct iatt              iatt   = {0,};
+        struct iatt              parent = {0,};
+        dict_t                  *fix_layout = NULL;
+        dict_t                  *migrate_data = NULL;
+        dict_t                  *status = NULL;
+        glusterfs_ctx_t         *ctx = NULL;
+
+        this = data;
+        if (!this)
+                goto out;
+
+        ctx = this->ctx;
+        if (!ctx)
+                goto out;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        defrag = conf->defrag;
+        if (!defrag)
+                goto out;
+
+        gettimeofday (&defrag->start_time, NULL);
+        dht_build_root_inode (this, &defrag->root_inode);
+        if (!defrag->root_inode)
+                goto out;
+
+        dht_build_root_loc (defrag->root_inode, &loc);
+
+        /* fix-layout on '/' first */
+
+        ret = syncop_lookup (this, &loc, NULL, &iatt, NULL, &parent);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "look up on / failed");
+                goto out;
+        }
+
+        fix_layout = dict_new ();
+        if (!fix_layout) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_str (fix_layout, GF_XATTR_FIX_LAYOUT_KEY, "yes");
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set dict str");
+                goto out;
+        }
+
+        ret = syncop_setxattr (this, &loc, fix_layout, 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "fix layout on %s failed",
+                        loc.path);
+                defrag->total_failures++;
+                goto out;
+        }
+
+        if (defrag->cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) {
+                migrate_data = dict_new ();
+                if (!migrate_data) {
+                        ret = -1;
+                        goto out;
+                }
+                if (defrag->cmd == GF_DEFRAG_CMD_START_FORCE)
+                        ret = dict_set_str (migrate_data,
+                                            "distribute.migrate-data", "force");
+                else
+                        ret = dict_set_str (migrate_data,
+                                            "distribute.migrate-data",
+                                            "non-force");
+                if (ret)
+                        goto out;
+        }
+        ret = gf_defrag_fix_layout (this, defrag, &loc, fix_layout,
+                                    migrate_data);
+        if ((defrag->defrag_status != GF_DEFRAG_STATUS_STOPPED) &&
+            (defrag->defrag_status != GF_DEFRAG_STATUS_FAILED)) {
+                defrag->defrag_status = GF_DEFRAG_STATUS_COMPLETE;
+        }
+
+
+
+out:
+        LOCK (&defrag->lock);
+        {
+                status = dict_new ();
+                gf_defrag_status_get (defrag, status);
+                if (ctx->notify)
+                        ctx->notify (GF_EN_DEFRAG_STATUS, status);
+                if (status)
+                        dict_unref (status);
+                defrag->is_exiting = 1;
+        }
+        UNLOCK (&defrag->lock);
+
+        if (defrag) {
+                GF_FREE (defrag);
+                conf->defrag = NULL;
+        }
+
+        return ret;
+}
+
+
+static int
+gf_defrag_done  (int ret, call_frame_t *sync_frame, void *data)
+{
+        gf_listener_stop (sync_frame->this);
+
+        STACK_DESTROY (sync_frame->root);
+        kill (getpid(), SIGTERM);
+        return 0;
+}
+
+void *
+gf_defrag_start (void *data)
+{
+        int                      ret    = -1;
+        call_frame_t            *frame  = NULL;
+        dht_conf_t              *conf   = NULL;
+        gf_defrag_info_t        *defrag = NULL;
+        xlator_t                *this  = NULL;
+
+        this = data;
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        defrag = conf->defrag;
+        if (!defrag)
+                goto out;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame)
+                goto out;
+
+        frame->root->pid = GF_CLIENT_PID_DEFRAG;
+
+        defrag->pid = frame->root->pid;
+
+        defrag->defrag_status = GF_DEFRAG_STATUS_STARTED;
+
+        ret = synctask_new (this->ctx->env, gf_defrag_start_crawl,
+                            gf_defrag_done, frame, this);
+
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "Could not create"
+                        " task for rebalance");
+out:
+        return NULL;
+}
+
+int
+gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict)
+{
+        int      ret    = 0;
+        uint64_t files  = 0;
+        uint64_t size   = 0;
+        uint64_t lookup = 0;
+        uint64_t failures = 0;
+        uint64_t skipped = 0;
+        char     *status = "";
+        double   elapsed = 0;
+        struct timeval end = {0,};
+
+
+        if (!defrag)
+                goto out;
+
+        ret = 0;
+        if (defrag->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED)
+                goto out;
+
+        files  = defrag->total_files;
+        size   = defrag->total_data;
+        lookup = defrag->num_files_lookedup;
+        failures = defrag->total_failures;
+        skipped = defrag->skipped;
+
+        gettimeofday (&end, NULL);
+
+        elapsed = end.tv_sec - defrag->start_time.tv_sec;
+
+        if (!dict)
+                goto log;
+
+        ret = dict_set_uint64 (dict, "files", files);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set file count");
+
+        ret = dict_set_uint64 (dict, "size", size);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set size of xfer");
+
+        ret = dict_set_uint64 (dict, "lookups", lookup);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set lookedup file count");
+
+
+        ret = dict_set_int32 (dict, "status", defrag->defrag_status);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set status");
+        if (elapsed) {
+                ret = dict_set_double (dict, "run-time", elapsed);
+                if (ret)
+                        gf_log (THIS->name, GF_LOG_WARNING,
+                                "failed to set run-time");
+        }
+
+        ret = dict_set_uint64 (dict, "failures", failures);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set failure count");
+
+        ret = dict_set_uint64 (dict, "skipped", skipped);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set skipped file count");
+log:
+        switch (defrag->defrag_status) {
+        case GF_DEFRAG_STATUS_NOT_STARTED:
+                status = "not started";
+                break;
+        case GF_DEFRAG_STATUS_STARTED:
+                status = "in progress";
+                break;
+        case GF_DEFRAG_STATUS_STOPPED:
+                status = "stopped";
+                break;
+        case GF_DEFRAG_STATUS_COMPLETE:
+                status = "completed";
+                break;
+        case GF_DEFRAG_STATUS_FAILED:
+                status = "failed";
+                break;
+        default:
+                break;
+        }
+
+        gf_log (THIS->name, GF_LOG_INFO, "Rebalance is %s. Time taken is %.2f "
+                "secs", status, elapsed);
+        gf_log (THIS->name, GF_LOG_INFO, "Files migrated: %"PRIu64", size: %"
+                PRIu64", lookups: %"PRIu64", failures: %"PRIu64", skipped: "
+                "%"PRIu64, files, size, lookup, failures, skipped);
+
+
+out:
+        return 0;
+}
+
+int
+gf_defrag_stop (gf_defrag_info_t *defrag, dict_t *output)
+{
+        /* TODO: set a variable 'stop_defrag' here, it should be checked
+           in defrag loop */
+        int     ret = -1;
+        GF_ASSERT (defrag);
+
+        if (defrag->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED) {
+                goto out;
+        }
+
+        gf_log ("", GF_LOG_INFO, "Received stop command on rebalance");
+        defrag->defrag_status = GF_DEFRAG_STATUS_STOPPED;
+
+        if (output)
+                gf_defrag_status_get (defrag, output);
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
diff --git a/xlators/cluster/dht/src/dht-rename.c b/xlators/cluster/dht/src/dht-rename.c
index d88fc7445..5d6f4f232 100644
--- a/xlators/cluster/dht/src/dht-rename.c
+++ b/xlators/cluster/dht/src/dht-rename.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 /* TODO: link(oldpath, newpath) fails if newpath already exists. DHT should
@@ -33,232 +24,332 @@
 
 int
 dht_rename_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+                    int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
                     struct iatt *preoldparent, struct iatt *postoldparent,
-                    struct iatt *prenewparent, struct iatt *postnewparent)
+                    struct iatt *prenewparent, struct iatt *postnewparent,
+                    dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
-
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
 
-	local = frame->local;
-	prev = cookie;
+        local = frame->local;
+        prev = cookie;
 
-	if (op_ret == -1) {
-		/* TODO: undo the damage */
-
-		gf_log (this->name, GF_LOG_DEBUG,
-			"rename %s -> %s on %s failed (%s)",
-			local->loc.path, local->loc2.path,
-			prev->this->name, strerror (op_errno));
-
-		local->op_ret   = op_ret;
-		local->op_errno = op_errno;
-	} else {
-		/* TODO: construct proper stbuf for dir */
-                /* 
-                 * FIXME: is this the correct way to build stbuf and
-                 * parent bufs?
-                */
-                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-                dht_iatt_merge (this, &local->preoldparent, preoldparent,
-                                prev->this);
-                dht_iatt_merge (this, &local->postoldparent, postoldparent,
-                                prev->this);
-                dht_iatt_merge (this, &local->preparent, prenewparent,
-                                prev->this);
-                dht_iatt_merge (this, &local->postparent, postnewparent,
-                                prev->this);
-	}
+        if (op_ret == -1) {
+                /* TODO: undo the damage */
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-                local->stbuf.ia_ino = local->loc.inode->ino;
+                gf_log (this->name, GF_LOG_INFO,
+                        "rename %s -> %s on %s failed (%s)",
+                        local->loc.path, local->loc2.path,
+                        prev->this->name, strerror (op_errno));
 
-                local->preoldparent.ia_ino = local->loc.parent->ino;
-                local->postoldparent.ia_ino = local->loc.parent->ino;
+                local->op_ret   = op_ret;
+                local->op_errno = op_errno;
+                goto unwind;
+        }
+        /* TODO: construct proper stbuf for dir */
+        /*
+         * FIXME: is this the correct way to build stbuf and
+         * parent bufs?
+         */
+        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+        dht_iatt_merge (this, &local->preoldparent, preoldparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->postoldparent, postoldparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->preparent, prenewparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->postparent, postnewparent,
+                        prev->this);
 
-                local->preparent.ia_ino = local->loc2.parent->ino;
-                local->postparent.ia_ino = local->loc2.parent->ino;
 
+unwind:
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
                 WIPE (&local->preoldparent);
                 WIPE (&local->postoldparent);
                 WIPE (&local->preparent);
                 WIPE (&local->postparent);
 
-		DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
-				  &local->stbuf, &local->preoldparent,
+                DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+                DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+                                  &local->stbuf, &local->preoldparent,
                                   &local->postoldparent,
-                                  &local->preparent, &local->postparent);
-	}
+                                  &local->preparent, &local->postparent, xdata);
+        }
 
-	return 0;
+        return 0;
 }
 
 
+int
+dht_rename_hashed_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+                           struct iatt *preoldparent,
+                           struct iatt *postoldparent,
+                           struct iatt *prenewparent,
+                           struct iatt *postnewparent, dict_t *xdata)
+{
+        dht_conf_t   *conf = NULL;
+        dht_local_t  *local = NULL;
+        int           call_cnt = 0;
+        call_frame_t *prev = NULL;
+        int           i = 0;
+
+        conf = this->private;
+        local = frame->local;
+        prev = cookie;
+
+        if (op_ret == -1) {
+                /* TODO: undo the damage */
+
+                gf_log (this->name, GF_LOG_INFO,
+                        "rename %s -> %s on %s failed (%s)",
+                        local->loc.path, local->loc2.path,
+                        prev->this->name, strerror (op_errno));
+
+                local->op_ret   = op_ret;
+                local->op_errno = op_errno;
+                goto unwind;
+        }
+        /* TODO: construct proper stbuf for dir */
+        /*
+         * FIXME: is this the correct way to build stbuf and
+         * parent bufs?
+         */
+        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+        dht_iatt_merge (this, &local->preoldparent, preoldparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->postoldparent, postoldparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->preparent, prenewparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->postparent, postnewparent,
+                        prev->this);
+
+        call_cnt = local->call_cnt = conf->subvolume_cnt - 1;
+
+        if (!local->call_cnt)
+                goto unwind;
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == local->dst_hashed)
+                        continue;
+                STACK_WIND (frame, dht_rename_dir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->rename,
+                            &local->loc, &local->loc2, NULL);
+                if (!--call_cnt)
+                        break;
+        }
+
+
+        return 0;
+unwind:
+        WIPE (&local->preoldparent);
+        WIPE (&local->postoldparent);
+        WIPE (&local->preparent);
+        WIPE (&local->postparent);
+
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+        DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+                          &local->stbuf, &local->preoldparent,
+                          &local->postoldparent,
+                          &local->preparent, &local->postparent, NULL);
+
+        return 0;
+}
+
 
 int
 dht_rename_dir_do (call_frame_t *frame, xlator_t *this)
 {
-	dht_local_t  *local = NULL;
-	dht_conf_t   *conf = NULL;
-	int           i = 0;
-
-	conf = this->private;
-	local = frame->local;
+        dht_local_t  *local = NULL;
 
-	if (local->op_ret == -1)
-		goto err;
+        local = frame->local;
 
-	local->call_cnt = conf->subvolume_cnt;
-	local->op_ret = 0;
+        if (local->op_ret == -1)
+                goto err;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_rename_dir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->rename,
-			    &local->loc, &local->loc2);
-	}
+        local->op_ret = 0;
 
-	return 0;
+        STACK_WIND (frame, dht_rename_hashed_dir_cbk,
+                    local->dst_hashed,
+                    local->dst_hashed->fops->rename,
+                    &local->loc, &local->loc2, NULL);
+        return 0;
 
 err:
-	DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno, NULL, NULL,
-                          NULL, NULL, NULL);
-	return 0;
+        DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno, NULL, NULL,
+                          NULL, NULL, NULL, NULL);
+        return 0;
 }
 
 
 int
 dht_rename_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int op_ret, int op_errno, gf_dirent_t *entries)
+                        int op_ret, int op_errno, gf_dirent_t *entries,
+                        dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = -1;
-	call_frame_t *prev = NULL;
-
-	local = frame->local;
-	prev  = cookie;
-
-	if (op_ret > 2) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"readdir on %s for %s returned %d entries",
-			prev->this->name, local->loc.path, op_ret);
-		local->op_ret = -1;
-		local->op_errno = ENOTEMPTY;
-	}
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = -1;
+        call_frame_t *prev = NULL;
+
+        local = frame->local;
+        prev  = cookie;
+
+        if (op_ret > 2) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "readdir on %s for %s returned %d entries",
+                        prev->this->name, local->loc.path, op_ret);
+                local->op_ret = -1;
+                local->op_errno = ENOTEMPTY;
+        }
 
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (is_last_call (this_call_cnt)) {
-		dht_rename_dir_do (frame, this);
-	}
+        if (is_last_call (this_call_cnt)) {
+                dht_rename_dir_do (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rename_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int op_ret, int op_errno, fd_t *fd)
+                        int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = -1;
-	call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = -1;
+        call_frame_t *prev = NULL;
 
 
-	local = frame->local;
-	prev  = cookie;
+        local = frame->local;
+        prev  = cookie;
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"opendir on %s for %s failed (%s)",
-			prev->this->name, local->loc.path,
-			strerror (op_errno));
-		goto err;
-	}
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "opendir on %s for %s failed (%s)",
+                        prev->this->name, local->loc.path,
+                        strerror (op_errno));
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_rename_readdir_cbk,
-		    prev->this, prev->this->fops->readdir,
-		    local->fd, 4096, 0);
+        STACK_WIND (frame, dht_rename_readdir_cbk,
+                    prev->this, prev->this->fops->readdir,
+                    local->fd, 4096, 0, NULL);
 
-	return 0;
+        return 0;
 
 err:
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (is_last_call (this_call_cnt)) {
-		dht_rename_dir_do (frame, this);
-	}
+        if (is_last_call (this_call_cnt)) {
+                dht_rename_dir_do (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rename_dir (call_frame_t *frame, xlator_t *this)
 {
-	dht_conf_t  *conf = NULL;
-	dht_local_t *local = NULL;
-	int          i = 0;
-	int          op_errno = -1;
+        dht_conf_t  *conf = NULL;
+        dht_local_t *local = NULL;
+        int          i = 0;
+        int          op_errno = -1;
 
 
-	conf = frame->this->private;
-	local = frame->local;
+        conf = frame->this->private;
+        local = frame->local;
 
-	local->call_cnt = conf->subvolume_cnt;
+        local->call_cnt = conf->subvolume_cnt;
 
         for (i = 0; i < conf->subvolume_cnt; i++) {
                 if (!conf->subvolume_status[i]) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "one of the subvolumes down (%s)",
+                                conf->subvolumes[i]->name);
                         op_errno = ENOTCONN;
                         goto err;
                 }
         }
 
-	local->fd = fd_create (local->loc.inode, frame->root->pid);
-	if (!local->fd) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+        local->fd = fd_create (local->loc.inode, frame->root->pid);
+        if (!local->fd) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->op_ret = 0;
+        local->op_ret = 0;
 
-	if (!local->dst_cached) {
-		dht_rename_dir_do (frame, this);
-		return 0;
-	}
+        if (!local->dst_cached) {
+                dht_rename_dir_do (frame, this);
+                return 0;
+        }
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_rename_opendir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->opendir,
-			    &local->loc2, local->fd);
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                STACK_WIND (frame, dht_rename_opendir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->opendir,
+                            &local->loc2, local->fd, NULL);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
-	return 0;
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+                          NULL, NULL);
+        return 0;
 }
+#define DHT_MARK_FOP_INTERNAL(xattr) do {                                      \
+                int tmp = -1;                                                  \
+                if (!xattr) {                                                  \
+                        xattr = dict_new ();                                   \
+                        if (!xattr)                                            \
+                                break;                                         \
+                }                                                              \
+                tmp = dict_set_str (xattr, GLUSTERFS_INTERNAL_FOP_KEY, "yes"); \
+                if (tmp) {                                                     \
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to set"      \
+                                " internal dict key for %s", local->loc.path); \
+                }                                                              \
+        }while (0)
+int
+dht_rename_done (call_frame_t *frame, xlator_t *this)
+{
+        dht_local_t     *local = NULL;
+
+        local = frame->local;
 
+        if (local->linked == _gf_true) {
+                local->linked = _gf_false;
+                dht_linkfile_attr_heal (frame, this);
+        }
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+        DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+                          &local->stbuf, &local->preoldparent,
+                          &local->postoldparent, &local->preparent,
+                          &local->postparent, NULL);
+
+        return 0;
+}
 
 int
 dht_rename_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                       struct iatt *postparent)
+                       int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                       struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev = NULL;
-	int           this_call_cnt = 0;
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           this_call_cnt = 0;
 
-	local = frame->local;
-	prev  = cookie;
+        local = frame->local;
+        prev  = cookie;
 
         if (!local) {
                 gf_log (this->name, GF_LOG_ERROR,
@@ -266,82 +357,84 @@ dht_rename_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto out;
         }
 
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unlink on %s failed (%s)",
-			prev->this->name, strerror (op_errno));
-	}
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: unlink on %s failed (%s)",
+                        local->loc.path, prev->this->name, strerror (op_errno));
+        }
 
         WIPE (&local->preoldparent);
         WIPE (&local->postoldparent);
         WIPE (&local->preparent);
         WIPE (&local->postparent);
 
-	if (is_last_call (this_call_cnt)) {
-		DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
-				  &local->stbuf, &local->preoldparent,
-                                  &local->postoldparent, &local->preparent,
-                                  &local->postparent);
+        if (is_last_call (this_call_cnt)) {
+                dht_rename_done (frame, this);
         }
 
 out:
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rename_cleanup (call_frame_t *frame)
 {
-	dht_local_t *local = NULL;
-	xlator_t    *this = NULL;
-	xlator_t    *src_hashed = NULL;
-	xlator_t    *src_cached = NULL;
-	xlator_t    *dst_hashed = NULL;
-	xlator_t    *dst_cached = NULL;
-	int          call_cnt = 0;
-
-
-	local = frame->local;
-	this  = frame->this;
-
-	src_hashed = local->src_hashed;
-	src_cached = local->src_cached;
-	dst_hashed = local->dst_hashed;
-	dst_cached = local->dst_cached;
-
-	if (src_cached == dst_cached)
-		goto nolinks;
+        dht_local_t *local = NULL;
+        xlator_t    *this = NULL;
+        xlator_t    *src_hashed = NULL;
+        xlator_t    *src_cached = NULL;
+        xlator_t    *dst_hashed = NULL;
+        xlator_t    *dst_cached = NULL;
+        int          call_cnt = 0;
+        dict_t      *xattr      = NULL;
+
+        local = frame->local;
+        this  = frame->this;
+
+        src_hashed = local->src_hashed;
+        src_cached = local->src_cached;
+        dst_hashed = local->dst_hashed;
+        dst_cached = local->dst_cached;
+
+        if (src_cached == dst_cached)
+                goto nolinks;
 
-	if (dst_hashed != src_hashed && dst_hashed != src_cached)
-		call_cnt++;
+        if (dst_hashed != src_hashed && dst_hashed != src_cached)
+                call_cnt++;
 
-	if (src_cached != dst_hashed)
-		call_cnt++;
+        if (src_cached != dst_hashed)
+                call_cnt++;
 
-	local->call_cnt = call_cnt;
+        local->call_cnt = call_cnt;
 
         if (!call_cnt)
                 goto nolinks;
 
-	if (dst_hashed != src_hashed && dst_hashed != src_cached) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"unlinking linkfile %s @ %s => %s",
-			local->loc.path, dst_hashed->name, src_cached->name);
+        DHT_MARK_FOP_INTERNAL (xattr);
+
+        if (dst_hashed != src_hashed && dst_hashed != src_cached) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "unlinking linkfile %s @ %s => %s",
+                        local->loc.path, dst_hashed->name, src_cached->name);
                 STACK_WIND (frame, dht_rename_unlink_cbk,
                             dst_hashed, dst_hashed->fops->unlink,
-                            &local->loc);
-	}
+                            &local->loc, 0, xattr);
+        }
 
-	if (src_cached != dst_hashed) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"unlinking link %s => %s (%s)", local->loc.path,
-			local->loc2.path, src_cached->name);
-		STACK_WIND (frame, dht_rename_unlink_cbk,
-			    src_cached, src_cached->fops->unlink,
-			    &local->loc2);
-	}
+        if (src_cached != dst_hashed) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "unlinking link %s => %s (%s)", local->loc.path,
+                        local->loc2.path, src_cached->name);
+                STACK_WIND (frame, dht_rename_unlink_cbk,
+                            src_cached, src_cached->fops->unlink,
+                            &local->loc2, 0, xattr);
+        }
+
+        if (xattr)
+                dict_unref (xattr);
 
         return 0;
 
@@ -351,10 +444,40 @@ nolinks:
         WIPE (&local->preparent);
         WIPE (&local->postparent);
 
-	DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
-			  &local->stbuf, &local->preoldparent,
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+        DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+                          &local->stbuf, &local->preoldparent,
                           &local->postoldparent, &local->preparent,
-                          &local->postparent);
+                          &local->postparent, NULL);
+
+        return 0;
+}
+
+
+int
+dht_rename_links_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                             int32_t op_ret, int32_t op_errno,
+                             inode_t *inode, struct iatt *stbuf,
+                             struct iatt *preparent, struct iatt *postparent,
+                             dict_t *xdata)
+{
+        call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+
+        prev = cookie;
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "link/file %s on %s failed (%s)",
+                        local->loc.path, prev->this->name, strerror (op_errno));
+        }
+
+        if (local->linked == _gf_true) {
+                local->linked = _gf_false;
+                dht_linkfile_attr_heal (frame, this);
+        }
+        DHT_STACK_DESTROY (frame);
 
         return 0;
 }
@@ -362,119 +485,156 @@ nolinks:
 
 int
 dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+                int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
                 struct iatt *preoldparent, struct iatt *postoldparent,
-                struct iatt *prenewparent, struct iatt *postnewparent)
+                struct iatt *prenewparent, struct iatt *postnewparent,
+                dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev = NULL;
-	xlator_t     *src_hashed = NULL;
-	xlator_t     *src_cached = NULL;
-	xlator_t     *dst_hashed = NULL;
-	xlator_t     *dst_cached = NULL;
-	xlator_t     *rename_subvol = NULL;
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        xlator_t     *src_hashed = NULL;
+        xlator_t     *src_cached = NULL;
+        xlator_t     *dst_hashed = NULL;
+        xlator_t     *dst_cached = NULL;
+        xlator_t     *rename_subvol = NULL;
+        call_frame_t *link_frame = NULL;
+        dht_local_t *link_local = NULL;
+        dict_t       *xattr     = NULL;
+
+        local = frame->local;
+        prev = cookie;
+
+        src_hashed = local->src_hashed;
+        src_cached = local->src_cached;
+        dst_hashed = local->dst_hashed;
+        dst_cached = local->dst_cached;
+
+        if (local->linked == _gf_true)
+                FRAME_SU_UNDO (frame, dht_local_t);
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: rename on %s failed (%s)", local->loc.path,
+                        prev->this->name, strerror (op_errno));
+                local->op_ret   = op_ret;
+                local->op_errno = op_errno;
+                goto cleanup;
+        }
 
-	local = frame->local;
-	prev = cookie;
+        if ((src_cached == dst_cached) && (dst_hashed != dst_cached)) {
+                link_frame = copy_frame (frame);
+                if (!link_frame) {
+                        goto err;
+                }
 
-	src_hashed = local->src_hashed;
-	src_cached = local->src_cached;
-	dst_hashed = local->dst_hashed;
-	dst_cached = local->dst_cached;
+                /* fop value sent as maxvalue because it is not used
+                   anywhere in this case */
+                link_local = dht_local_init (link_frame, &local->loc2, NULL,
+                                             GF_FOP_MAXVALUE);
+                if (!link_local) {
+                        goto err;
+                }
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"rename on %s failed (%s)", prev->this->name,
-			strerror (op_errno));
-		local->op_ret   = op_ret;
-		local->op_errno = op_errno;
-		goto cleanup;
-	}
+                if (link_local->loc.inode)
+                        inode_unref (link_local->loc.inode);
+                link_local->loc.inode = inode_ref (local->loc.inode);
+                uuid_copy (link_local->gfid, local->loc.inode->gfid);
 
-        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-        dht_iatt_merge (this, &local->preoldparent, preoldparent, prev->this);
-        dht_iatt_merge (this, &local->postoldparent, postoldparent, prev->this);
-        dht_iatt_merge (this, &local->preparent, prenewparent, prev->this);
-        dht_iatt_merge (this, &local->postparent, postnewparent, prev->this);
+                dht_linkfile_create (link_frame, dht_rename_links_create_cbk,
+                                     this, src_cached, dst_hashed,
+                                     &link_local->loc);
+        }
 
-        local->stbuf.ia_ino = local->loc.inode->ino;
+err:
+        /* Merge attrs only from src_cached. In case there of src_cached !=
+         * dst_hashed, this ignores linkfile attrs. */
+        if (prev->this == src_cached) {
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                dht_iatt_merge (this, &local->preoldparent, preoldparent,
+                                prev->this);
+                dht_iatt_merge (this, &local->postoldparent, postoldparent,
+                                prev->this);
+                dht_iatt_merge (this, &local->preparent, prenewparent,
+                                prev->this);
+                dht_iatt_merge (this, &local->postparent, postnewparent,
+                                prev->this);
+        }
 
-        local->preoldparent.ia_ino = local->loc.parent->ino;
-        local->postoldparent.ia_ino = local->loc.parent->ino;
 
-        local->preparent.ia_ino = local->loc2.parent->ino;
-        local->postparent.ia_ino = local->loc2.parent->ino;
-	
-	/* NOTE: rename_subvol is the same subvolume from which dht_rename_cbk
-	 *       is called. since rename has already happened on rename_subvol,
-	 *       unlink should not be sent for oldpath (either linkfile or cached-file)
-	 *       on rename_subvol. */
-	if (src_cached == dst_cached)
-		rename_subvol = src_cached;
-	else
-		rename_subvol = dst_hashed;
+        /* NOTE: rename_subvol is the same subvolume from which dht_rename_cbk
+         *       is called. since rename has already happened on rename_subvol,
+         *       unlink should not be sent for oldpath (either linkfile or cached-file)
+         *       on rename_subvol. */
+        if (src_cached == dst_cached)
+                rename_subvol = src_cached;
+        else
+                rename_subvol = dst_hashed;
 
-	/* TODO: delete files in background */
+        /* TODO: delete files in background */
 
-	if (src_cached != dst_hashed && src_cached != dst_cached)
-		local->call_cnt++;
+        if (src_cached != dst_hashed && src_cached != dst_cached)
+                local->call_cnt++;
 
-	if (src_hashed != rename_subvol && src_hashed != src_cached)
-		local->call_cnt++;
+        if (src_hashed != rename_subvol && src_hashed != src_cached)
+                local->call_cnt++;
 
-	if (dst_cached && dst_cached != dst_hashed && dst_cached != src_cached)
-		local->call_cnt++;
+        if (dst_cached && dst_cached != dst_hashed && dst_cached != src_cached)
+                local->call_cnt++;
 
-	if (local->call_cnt == 0)
-		goto unwind;
+        if (local->call_cnt == 0)
+                goto unwind;
 
-	if (src_cached != dst_hashed && src_cached != dst_cached) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"deleting old src datafile %s @ %s",
-			local->loc.path, src_cached->name);
+        DHT_MARK_FOP_INTERNAL (xattr);
 
-		STACK_WIND (frame, dht_rename_unlink_cbk,
-			    src_cached, src_cached->fops->unlink,
-			    &local->loc);
-	}
+        if (src_cached != dst_hashed && src_cached != dst_cached) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "deleting old src datafile %s @ %s",
+                        local->loc.path, src_cached->name);
 
-	if (src_hashed != rename_subvol && src_hashed != src_cached) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"deleting old src linkfile %s @ %s",
-			local->loc.path, src_hashed->name);
+                STACK_WIND (frame, dht_rename_unlink_cbk,
+                            src_cached, src_cached->fops->unlink,
+                            &local->loc, 0, xattr);
+        }
 
-		STACK_WIND (frame, dht_rename_unlink_cbk,
-			    src_hashed, src_hashed->fops->unlink,
-			    &local->loc);
-	}
+        if (src_hashed != rename_subvol && src_hashed != src_cached) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "deleting old src linkfile %s @ %s",
+                        local->loc.path, src_hashed->name);
 
-	if (dst_cached
-	    && (dst_cached != dst_hashed)
-	    && (dst_cached != src_cached)) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"deleting old dst datafile %s @ %s",
-			local->loc2.path, dst_cached->name);
+                STACK_WIND (frame, dht_rename_unlink_cbk,
+                            src_hashed, src_hashed->fops->unlink,
+                            &local->loc, 0, xattr);
+        }
 
-		STACK_WIND (frame, dht_rename_unlink_cbk,
-			    dst_cached, dst_cached->fops->unlink,
-			    &local->loc2);
-	}
-	return 0;
+        if (dst_cached
+            && (dst_cached != dst_hashed)
+            && (dst_cached != src_cached)) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "deleting old dst datafile %s @ %s",
+                        local->loc2.path, dst_cached->name);
+
+                STACK_WIND (frame, dht_rename_unlink_cbk,
+                            dst_cached, dst_cached->fops->unlink,
+                            &local->loc2, 0, xattr);
+        }
+        if (xattr)
+                dict_unref (xattr);
+        return 0;
 
 unwind:
         WIPE (&local->preoldparent);
         WIPE (&local->postoldparent);
         WIPE (&local->preparent);
         WIPE (&local->postparent);
+        if (xattr)
+                dict_unref (xattr);
 
-	DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
-			  &local->stbuf, &local->preoldparent,
-                          &local->postoldparent, &local->preparent,
-                          &local->postparent);
+        dht_rename_done (frame, this);
 
-	return 0;
+        return 0;
 
 cleanup:
+        if (xattr)
+                dict_unref (xattr);
         dht_rename_cleanup (frame);
 
         return 0;
@@ -484,67 +644,110 @@ cleanup:
 int
 dht_do_rename (call_frame_t *frame)
 {
-	dht_local_t *local = NULL;
-	xlator_t    *dst_hashed = NULL;
-	xlator_t    *src_cached = NULL;
-	xlator_t    *dst_cached = NULL;
-	xlator_t    *this = NULL;
-	xlator_t    *rename_subvol = NULL;
+        dht_local_t *local = NULL;
+        xlator_t    *dst_hashed = NULL;
+        xlator_t    *src_cached = NULL;
+        xlator_t    *dst_cached = NULL;
+        xlator_t    *this = NULL;
+        xlator_t    *rename_subvol = NULL;
 
 
-	local = frame->local;
-	this  = frame->this;
+        local = frame->local;
+        this  = frame->this;
 
-	dst_hashed = local->dst_hashed;
-	dst_cached = local->dst_cached;
-	src_cached = local->src_cached;
+        dst_hashed = local->dst_hashed;
+        dst_cached = local->dst_cached;
+        src_cached = local->src_cached;
 
-	if (src_cached == dst_cached)
-		rename_subvol = src_cached;
-	else
-		rename_subvol = dst_hashed;
+        if (src_cached == dst_cached)
+                rename_subvol = src_cached;
+        else
+                rename_subvol = dst_hashed;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"renaming %s => %s (%s)",
-		local->loc.path, local->loc2.path, rename_subvol->name);
+        gf_log (this->name, GF_LOG_TRACE,
+                "renaming %s => %s (%s)",
+                local->loc.path, local->loc2.path, rename_subvol->name);
 
-	STACK_WIND (frame, dht_rename_cbk,
-		    rename_subvol, rename_subvol->fops->rename,
-		    &local->loc, &local->loc2);
+        if (local->linked == _gf_true)
+                FRAME_SU_DO (frame, dht_local_t);
+        STACK_WIND (frame, dht_rename_cbk,
+                    rename_subvol, rename_subvol->fops->rename,
+                    &local->loc, &local->loc2, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rename_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		      int32_t op_ret, int32_t op_errno,
+                      int32_t op_ret, int32_t op_errno,
                       inode_t *inode, struct iatt *stbuf,
-                      struct iatt *preparent, struct iatt *postparent)
+                      struct iatt *preparent, struct iatt *postparent,
+                      dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           this_call_cnt  = 0;
+
+
+        local = frame->local;
+        prev = cookie;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "link/file on %s failed (%s)",
+                        prev->this->name, strerror (op_errno));
+                local->op_ret   = -1;
+                if (op_errno != ENOENT)
+                        local->op_errno = op_errno;
+        } else if (local->src_cached == prev->this) {
+                /* merge of attr returned only from linkfile creation */
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+        }
+
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                if (local->op_ret == -1)
+                        goto cleanup;
+
+                dht_do_rename (frame);
+        }
+
+        return 0;
+
+cleanup:
+        dht_rename_cleanup (frame);
+
+        return 0;
+}
+
+
+int
+dht_rename_unlink_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                             int32_t op_ret, int32_t op_errno,
+                             struct iatt *preparent, struct iatt *postparent,
+                             dict_t *xdata)
 {
 	dht_local_t  *local = NULL;
 	call_frame_t *prev = NULL;
-	int           this_call_cnt  = 0;
 
 
 	local = frame->local;
 	prev = cookie;
-	
-	if (op_ret == -1) {
+
+	if ((op_ret == -1) && (op_errno != ENOENT)) {
 		gf_log (this->name, GF_LOG_DEBUG,
-			"link/file on %s failed (%s)",
-			prev->this->name, strerror (op_errno));
+			"unlink of %s on %s failed (%s)",
+			local->loc2.path, prev->this->name,
+                        strerror (op_errno));
 		local->op_ret   = -1;
 		local->op_errno = op_errno;
 	}
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-		if (local->op_ret == -1)
-			goto cleanup;
-		
-		dht_do_rename (frame);
-	}
+        if (local->op_ret == -1)
+                goto cleanup;
+
+        dht_do_rename (frame);
 
 	return 0;
 
@@ -558,39 +761,54 @@ cleanup:
 int
 dht_rename_create_links (call_frame_t *frame)
 {
-	dht_local_t *local = NULL;
-	xlator_t    *this = NULL;
-	xlator_t    *src_hashed = NULL;
-	xlator_t    *src_cached = NULL;
-	xlator_t    *dst_hashed = NULL;
-	xlator_t    *dst_cached = NULL;
-	int          call_cnt = 0;
+        dht_local_t *local = NULL;
+        xlator_t    *this = NULL;
+        xlator_t    *src_hashed = NULL;
+        xlator_t    *src_cached = NULL;
+        xlator_t    *dst_hashed = NULL;
+        xlator_t    *dst_cached = NULL;
+        int          call_cnt = 0;
+        dict_t      *xattr = NULL;
 
 
-	local = frame->local;
-	this  = frame->this;
+        local = frame->local;
+        this  = frame->this;
 
-	src_hashed = local->src_hashed;
-	src_cached = local->src_cached;
-	dst_hashed = local->dst_hashed;
-	dst_cached = local->dst_cached;
+        src_hashed = local->src_hashed;
+        src_cached = local->src_cached;
+        dst_hashed = local->dst_hashed;
+        dst_cached = local->dst_cached;
 
-	if (src_cached == dst_cached)
-		goto nolinks;
+        DHT_MARK_FOP_INTERNAL (xattr);
 
-	if (dst_hashed != src_hashed && dst_hashed != src_cached)
-		call_cnt++;
+        if (src_cached == dst_cached) {
+                if (dst_hashed == dst_cached)
+                        goto nolinks;
 
-	if (src_cached != dst_hashed)
-		call_cnt++;
+		gf_log (this->name, GF_LOG_TRACE,
+			"unlinking dst linkfile %s @ %s",
+			local->loc2.path, dst_hashed->name);
 
-	local->call_cnt = call_cnt;
+		STACK_WIND (frame, dht_rename_unlink_links_cbk,
+			    dst_hashed, dst_hashed->fops->unlink,
+			    &local->loc2, 0, xattr);
+                return 0;
+        }
 
-	if (dst_hashed != src_hashed && dst_hashed != src_cached) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"linkfile %s @ %s => %s",
-			local->loc.path, dst_hashed->name, src_cached->name);
-		dht_linkfile_create (frame, dht_rename_links_cbk,
+        if (dst_hashed != src_hashed && dst_hashed != src_cached)
+                call_cnt++;
+
+        if (src_cached != dst_hashed)
+                call_cnt++;
+
+        local->call_cnt = call_cnt;
+
+        if (dst_hashed != src_hashed && dst_hashed != src_cached) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "linkfile %s @ %s => %s",
+                        local->loc.path, dst_hashed->name, src_cached->name);
+                memcpy (local->gfid, local->loc.inode->gfid, 16);
+		dht_linkfile_create (frame, dht_rename_links_cbk, this,
 				     src_cached, dst_hashed, &local->loc);
 	}
 
@@ -600,113 +818,107 @@ dht_rename_create_links (call_frame_t *frame)
 			local->loc2.path, src_cached->name);
 		STACK_WIND (frame, dht_rename_links_cbk,
 			    src_cached, src_cached->fops->link,
-			    &local->loc, &local->loc2);
+			    &local->loc, &local->loc2, xattr);
 	}
 
 nolinks:
-	if (!call_cnt) {
-		/* skip to next step */
-		dht_do_rename (frame);
-	}
+        if (!call_cnt) {
+                /* skip to next step */
+                dht_do_rename (frame);
+        }
+        if (xattr)
+                dict_unref (xattr);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rename (call_frame_t *frame, xlator_t *this,
-	    loc_t *oldloc, loc_t *newloc)
+            loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
-	xlator_t    *src_cached = NULL;
-	xlator_t    *src_hashed = NULL;
-	xlator_t    *dst_cached = NULL;
-	xlator_t    *dst_hashed = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
-	dht_local_t *local = NULL;
-
-
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (oldloc, err);
-	VALIDATE_OR_GOTO (newloc, err);
-
-	src_hashed = dht_subvol_get_hashed (this, oldloc);
-	if (!src_hashed) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			oldloc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	src_cached = dht_subvol_get_cached (this, oldloc->inode);
-	if (!src_cached) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", oldloc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	dst_hashed = dht_subvol_get_hashed (this, newloc);
-	if (!dst_hashed) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			newloc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        xlator_t    *src_cached = NULL;
+        xlator_t    *src_hashed = NULL;
+        xlator_t    *dst_cached = NULL;
+        xlator_t    *dst_hashed = NULL;
+        int          op_errno = -1;
+        int          ret = -1;
+        dht_local_t *local = NULL;
+
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (oldloc, err);
+        VALIDATE_OR_GOTO (newloc, err);
+
+        src_hashed = dht_subvol_get_hashed (this, oldloc);
+        if (!src_hashed) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no subvolume in layout for path=%s",
+                        oldloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	if (newloc->inode)
-		dst_cached = dht_subvol_get_cached (this, newloc->inode);
+        src_cached = dht_subvol_get_cached (this, oldloc->inode);
+        if (!src_cached) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no cached subvolume for path=%s", oldloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        dst_hashed = dht_subvol_get_hashed (this, newloc);
+        if (!dst_hashed) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no subvolume in layout for path=%s",
+                        newloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	ret = loc_copy (&local->loc, oldloc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        if (newloc->inode)
+                dst_cached = dht_subvol_get_cached (this, newloc->inode);
 
-	ret = loc_copy (&local->loc2, newloc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, oldloc, NULL, GF_FOP_RENAME);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+        /* cached_subvol will be set from dht_local_init, reset it to NULL,
+           as the logic of handling rename is different  */
+        local->cached_subvol = NULL;
+
+        ret = loc_copy (&local->loc2, newloc);
+        if (ret == -1) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->src_hashed = src_hashed;
-	local->src_cached = src_cached;
-	local->dst_hashed = dst_hashed;
-	local->dst_cached = dst_cached;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"renaming %s (hash=%s/cache=%s) => %s (hash=%s/cache=%s)",
-		oldloc->path, src_hashed->name, src_cached->name,
-		newloc->path, dst_hashed->name,
-		dst_cached ? dst_cached->name : "<nul>");
-
-	if (IA_ISDIR (oldloc->inode->ia_type)) {
-		dht_rename_dir (frame, this);
-	} else {
-		local->op_ret = 0;
-		dht_rename_create_links (frame);
-	}
+        local->src_hashed = src_hashed;
+        local->src_cached = src_cached;
+        local->dst_hashed = dst_hashed;
+        local->dst_cached = dst_cached;
+
+        gf_log (this->name, GF_LOG_TRACE,
+                "renaming %s (hash=%s/cache=%s) => %s (hash=%s/cache=%s)",
+                oldloc->path, src_hashed->name, src_cached->name,
+                newloc->path, dst_hashed->name,
+                dst_cached ? dst_cached->name : "<nul>");
+
+        if (IA_ISDIR (oldloc->inode->ia_type)) {
+                dht_rename_dir (frame, this);
+        } else {
+                local->op_ret = 0;
+                dht_rename_create_links (frame);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+                          NULL, NULL);
 
-	return 0;
+        return 0;
 }
diff --git a/xlators/cluster/dht/src/dht-selfheal.c b/xlators/cluster/dht/src/dht-selfheal.c
index 6b3e13b1a..3fe96b1c7 100644
--- a/xlators/cluster/dht/src/dht-selfheal.c
+++ b/xlators/cluster/dht/src/dht-selfheal.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -26,263 +17,512 @@
 #include "glusterfs.h"
 #include "xlator.h"
 #include "dht-common.h"
+#include "glusterfs-acl.h"
+
+#define DHT_SET_LAYOUT_RANGE(layout,i,srt,chunk,cnt,path)    do {       \
+                layout->list[i].start = srt;                            \
+                layout->list[i].stop  = srt + chunk - 1;                \
+                                                                        \
+                gf_log (this->name, GF_LOG_TRACE,                       \
+                        "gave fix: %u - %u on %s for %s",               \
+                        layout->list[i].start, layout->list[i].stop,    \
+                        layout->list[i].xlator->name, path);            \
+        } while (0)
+
+#define DHT_RESET_LAYOUT_RANGE(layout)    do {                          \
+                int cnt = 0;                                            \
+                for (cnt = 0; cnt < layout->cnt; cnt++ ) {              \
+                        layout->list[cnt].start = 0;                    \
+                        layout->list[cnt].stop  = 0;                    \
+                }                                                       \
+        } while (0)
+
+static uint32_t
+dht_overlap_calc (dht_layout_t *old, int o, dht_layout_t *new, int n)
+{
+	if (o >= old->cnt || n >= new->cnt)
+		return 0;
+
+	if (old->list[o].err > 0 || new->list[n].err > 0)
+		return 0;
+
+        if (old->list[o].start == old->list[o].stop) {
+                return 0;
+        }
+
+        if (new->list[n].start == new->list[n].stop) {
+                return 0;
+        }
+
+	if ((old->list[o].start > new->list[n].stop) ||
+	    (old->list[o].stop < new->list[n].start))
+		return 0;
+
+	return min (old->list[o].stop, new->list[n].stop) -
+	        max (old->list[o].start, new->list[n].start) + 1;
+}
 
 
 int
 dht_selfheal_dir_finish (call_frame_t *frame, xlator_t *this, int ret)
 {
-	dht_local_t  *local = NULL;
-
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
-	local->selfheal.dir_cbk (frame, NULL, frame->this, ret,
-				 local->op_errno);
+        local = frame->local;
+        local->selfheal.dir_cbk (frame, NULL, frame->this, ret,
+                                 local->op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_selfheal_dir_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			    int op_ret, int op_errno)
+                            int op_ret, int op_errno, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev = NULL;
-	xlator_t     *subvol = NULL;
-	int           i = 0;
-	dht_layout_t *layout = NULL;
-	int           err = 0;
-	int           this_call_cnt = 0;
-
-	local = frame->local;
-	layout = local->selfheal.layout;
-	prev = cookie;
-	subvol = prev->this;
-
-	if (op_ret == 0)
-		err = 0;
-	else
-		err = op_errno;
-
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].xlator == subvol) {
-			layout->list[i].err = err;
-			break;
-		}
-	}
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        xlator_t     *subvol = NULL;
+        int           i = 0;
+        dht_layout_t *layout = NULL;
+        int           err = 0;
+        int           this_call_cnt = 0;
+
+        local = frame->local;
+        layout = local->selfheal.layout;
+        prev = cookie;
+        subvol = prev->this;
+
+        if (op_ret == 0)
+                err = 0;
+        else
+                err = op_errno;
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].xlator == subvol) {
+                        layout->list[i].err = err;
+                        break;
+                }
+        }
 
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (is_last_call (this_call_cnt)) {
-		dht_selfheal_dir_finish (frame, this, 0);
-	}
+        if (is_last_call (this_call_cnt)) {
+                dht_selfheal_dir_finish (frame, this, 0);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_selfheal_dir_xattr_persubvol (call_frame_t *frame, loc_t *loc,
-				  dht_layout_t *layout, int i)
+                                  dht_layout_t *layout, int i,
+                                  xlator_t *req_subvol)
 {
-	xlator_t          *subvol = NULL;
-	dict_t            *xattr = NULL;
-	int                ret = 0;
-	xlator_t          *this = NULL;
-	int32_t           *disk_layout = NULL;
-
-
-	subvol = layout->list[i].xlator;
-	this = frame->this;
+        xlator_t          *subvol = NULL;
+        dict_t            *xattr = NULL;
+        int                ret = 0;
+        xlator_t          *this = NULL;
+        int32_t           *disk_layout = NULL;
+        dht_local_t       *local = NULL;
+        dht_conf_t        *conf = NULL;
+
+        local = frame->local;
+        if (req_subvol)
+                subvol = req_subvol;
+        else
+                subvol = layout->list[i].xlator;
+        this = frame->this;
+
+        GF_VALIDATE_OR_GOTO ("", this, err);
+        GF_VALIDATE_OR_GOTO (this->name, layout, err);
+        GF_VALIDATE_OR_GOTO (this->name, local, err);
+        GF_VALIDATE_OR_GOTO (this->name, subvol, err);
+        VALIDATE_OR_GOTO (this->private, err);
+
+        conf = this->private;
+
+        xattr = get_new_dict ();
+        if (!xattr) {
+                goto err;
+        }
 
-	xattr = get_new_dict ();
-	if (!xattr) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        ret = dht_disk_layout_extract (this, layout, i, &disk_layout);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: (subvol %s) failed to extract disk layout",
+                        loc->path, subvol->name);
+                goto err;
+        }
 
-	ret = dht_disk_layout_extract (this, layout, i, &disk_layout);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"failed to extract disk layout");
-		goto err;
-	}
+        ret = dict_set_bin (xattr, conf->xattr_name, disk_layout, 4 * 4);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: (subvol %s) failed to set xattr dictionary",
+                        loc->path, subvol->name);
+                goto err;
+        }
+        disk_layout = NULL;
 
-	ret = dict_set_bin (xattr, "trusted.glusterfs.dht",
-			    disk_layout, 4 * 4);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"failed to set xattr dictionary");
-		goto err;
-	}
-	disk_layout = NULL;
+        gf_log (this->name, GF_LOG_TRACE,
+                "setting hash range %u - %u (type %d) on subvolume %s for %s",
+                layout->list[i].start, layout->list[i].stop,
+                layout->type, subvol->name, loc->path);
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"setting hash range %u - %u (type %d) on subvolume %s for %s",
-		layout->list[i].start, layout->list[i].stop,
-		layout->type, subvol->name, loc->path);
+        dict_ref (xattr);
 
-	dict_ref (xattr);
+        if (!uuid_is_null (local->gfid))
+                uuid_copy (loc->gfid, local->gfid);
 
-	STACK_WIND (frame, dht_selfheal_dir_xattr_cbk,
-		    subvol, subvol->fops->setxattr,
-		    loc, xattr, 0);
+        STACK_WIND (frame, dht_selfheal_dir_xattr_cbk,
+                    subvol, subvol->fops->setxattr,
+                    loc, xattr, 0, NULL);
 
-	dict_unref (xattr);
+        dict_unref (xattr);
 
-	return 0;
+        return 0;
 
 err:
-	if (xattr)
-		dict_destroy (xattr);
+        if (xattr)
+                dict_destroy (xattr);
 
-	if (disk_layout)
-		GF_FREE (disk_layout);
+        GF_FREE (disk_layout);
 
-	dht_selfheal_dir_xattr_cbk (frame, subvol, frame->this,
-				    -1, ENOMEM);
-	return 0;
+        dht_selfheal_dir_xattr_cbk (frame, subvol, frame->this,
+                                    -1, ENOMEM, NULL);
+        return 0;
 }
 
+int
+dht_fix_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
+{
+        dht_local_t *local = NULL;
+        int          i = 0;
+        int          count = 0;
+        xlator_t    *this = NULL;
+        dht_conf_t  *conf = NULL;
+        dht_layout_t *dummy = NULL;
+
+        local = frame->local;
+        this = frame->this;
+        conf = this->private;
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "writing the new range for all subvolumes");
+
+        local->call_cnt = count = conf->subvolume_cnt;
+
+        for (i = 0; i < layout->cnt; i++) {
+                dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
+
+                if (--count == 0)
+                        goto out;
+        }
+        /* if we are here, subvolcount > layout_count. subvols-per-directory
+         * option might be set here. We need to clear out layout from the
+         * non-participating subvolumes, else it will result in overlaps */
+        dummy = dht_layout_new (this, 1);
+        if (!dummy)
+                goto out;
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (_gf_false ==
+                    dht_is_subvol_in_layout (layout, conf->subvolumes[i])) {
+                        dht_selfheal_dir_xattr_persubvol (frame, loc, dummy, 0,
+                                                          conf->subvolumes[i]);
+                        if (--count == 0)
+                                break;
+                }
+        }
+
+        dht_layout_unref (this, dummy);
+out:
+        return 0;
+}
 
 int
 dht_selfheal_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
 {
-	dht_local_t *local = NULL;
-	int          missing_xattr = 0;
-	int          i = 0;
-	xlator_t    *this = NULL;
-
-	local = frame->local;
-	this = frame->this;
-
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err != -1 || !layout->list[i].stop) {
-			/* err != -1 would mean xattr present on the directory
-			 * or the directory is itself non existant.
-			 * !layout->list[i].stop would mean layout absent
-			 */
-			continue;
-		}
-		missing_xattr++;
-	}
+        dht_local_t *local = NULL;
+        int          missing_xattr = 0;
+        int          i = 0;
+        xlator_t    *this = NULL;
+        dht_conf_t   *conf = NULL;
+        dht_layout_t *dummy = NULL;
+
+        local = frame->local;
+        this = frame->this;
+        conf = this->private;
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err != -1 || !layout->list[i].stop) {
+                        /* err != -1 would mean xattr present on the directory
+                         * or the directory is non existent.
+                         * !layout->list[i].stop would mean layout absent
+                         */
+
+                        continue;
+                }
+                missing_xattr++;
+        }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"%d subvolumes missing xattr for %s",
-		missing_xattr, loc->path);
+        gf_log (this->name, GF_LOG_TRACE,
+                "%d subvolumes missing xattr for %s",
+                missing_xattr, loc->path);
 
-	if (missing_xattr == 0) {
-		dht_selfheal_dir_finish (frame, this, 0);
-		return 0;
-	}
+        if (missing_xattr == 0) {
+                dht_selfheal_dir_finish (frame, this, 0);
+                return 0;
+        }
 
-	local->call_cnt = missing_xattr;
+        local->call_cnt = missing_xattr;
 
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err != -1 || !layout->list[i].stop)
-			continue;
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err != -1 || !layout->list[i].stop)
+                        continue;
 
-		dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i);
+                dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
 
-		if (--missing_xattr == 0)
-			break;
-	}
-	return 0;
+                if (--missing_xattr == 0)
+                        break;
+        }
+        dummy = dht_layout_new (this, 1);
+        if (!dummy)
+                goto out;
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (_gf_false ==
+                    dht_is_subvol_in_layout (layout, conf->subvolumes[i])) {
+                        dht_selfheal_dir_xattr_persubvol (frame, loc, dummy, 0,
+                                                          conf->subvolumes[i]);
+                }
+        }
+        dht_layout_unref (this, dummy);
+out:
+        return 0;
+}
+
+int
+dht_selfheal_dir_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                              int op_ret, int op_errno, struct iatt *statpre,
+                              struct iatt *statpost, dict_t *xdata)
+{
+        dht_local_t   *local = NULL;
+        dht_layout_t  *layout = NULL;
+        int            this_call_cnt = 0;
+
+        local  = frame->local;
+        layout = local->selfheal.layout;
+
+        this_call_cnt = dht_frame_return (frame);
+
+        if (is_last_call (this_call_cnt)) {
+                dht_selfheal_dir_xattr (frame, &local->loc, layout);
+        }
+
+        return 0;
 }
 
 
 int
+dht_selfheal_dir_setattr (call_frame_t *frame, loc_t *loc, struct iatt *stbuf,
+                          int32_t valid, dht_layout_t *layout)
+{
+        int           missing_attr = 0;
+        int           i     = 0;
+        dht_local_t  *local = NULL;
+        xlator_t     *this = NULL;
+
+        local = frame->local;
+        this = frame->this;
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == -1)
+                        missing_attr++;
+        }
+
+        if (missing_attr == 0) {
+                dht_selfheal_dir_xattr (frame, loc, layout);
+                return 0;
+        }
+
+        if (!uuid_is_null (local->gfid))
+                uuid_copy (loc->gfid, local->gfid);
+
+        local->call_cnt = missing_attr;
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == -1) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "setattr for %s on subvol %s",
+                                loc->path, layout->list[i].xlator->name);
+
+                        STACK_WIND (frame, dht_selfheal_dir_setattr_cbk,
+                                    layout->list[i].xlator,
+                                    layout->list[i].xlator->fops->setattr,
+                                    loc, stbuf, valid, NULL);
+                }
+        }
+
+        return 0;
+}
+
+int
 dht_selfheal_dir_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			    int op_ret, int op_errno,
+                            int op_ret, int op_errno,
                             inode_t *inode, struct iatt *stbuf,
-                            struct iatt *preparent, struct iatt *postparent)
+                            struct iatt *preparent, struct iatt *postparent,
+                            dict_t *xdata)
 {
-	dht_local_t   *local = NULL;
-	dht_layout_t  *layout = NULL;
-	call_frame_t  *prev = NULL;
-	xlator_t      *subvol = NULL;
-	int            i = 0;
-	int            this_call_cnt = 0;
-
-
-	local  = frame->local;
-	layout = local->selfheal.layout;
-	prev   = cookie;
-	subvol = prev->this;
-
-	if ((op_ret == 0) || ((op_ret == -1) && (op_errno == EEXIST))) {
-		for (i = 0; i < layout->cnt; i++) {
-			if (layout->list[i].xlator == subvol) {
-				layout->list[i].err = -1;
-				break;
-			}
-		}
-	}
+        dht_local_t   *local = NULL;
+        dht_layout_t  *layout = NULL;
+        call_frame_t  *prev = NULL;
+        xlator_t      *subvol = NULL;
+        int            i = 0;
+        int            this_call_cnt = 0;
 
-        if (op_ret)
-                goto out;
 
+        local  = frame->local;
+        layout = local->selfheal.layout;
+        prev   = cookie;
+        subvol = prev->this;
 
-        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-        if (prev->this == local->hashed_subvol)
-                local->ia_ino = local->stbuf.ia_ino;
+        if ((op_ret == 0) || ((op_ret == -1) && (op_errno == EEXIST))) {
+                for (i = 0; i < layout->cnt; i++) {
+                        if (layout->list[i].xlator == subvol) {
+                                layout->list[i].err = -1;
+                                break;
+                        }
+                }
+        }
+
+        if (op_ret) {
+                gf_log (this->name, ((op_errno == EEXIST) ? GF_LOG_DEBUG :
+                                     GF_LOG_WARNING),
+                        "selfhealing directory %s failed: %s",
+                        local->loc.path, strerror (op_errno));
+                goto out;
+        }
 
+        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
         dht_iatt_merge (this, &local->preparent, preparent, prev->this);
         dht_iatt_merge (this, &local->postparent, postparent, prev->this);
 
 out:
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (is_last_call (this_call_cnt)) {
-		dht_selfheal_dir_xattr (frame, &local->loc, layout);
-	}
+        if (is_last_call (this_call_cnt)) {
+                dht_selfheal_dir_setattr (frame, &local->loc, &local->stbuf, 0xffffff, layout);
+        }
 
-	return 0;
+        return 0;
 }
 
-
-int
-dht_selfheal_dir_mkdir (call_frame_t *frame, loc_t *loc,
-			dht_layout_t *layout, int force)
+void
+dht_selfheal_dir_mkdir_setacl (dict_t *xattr, dict_t *dict)
 {
-	int           missing_dirs = 0;
-	int           i = 0;
-	dht_local_t  *local = NULL;
-	xlator_t     *this = NULL;
+        data_t          *acl_default = NULL;
+        data_t          *acl_access = NULL;
+        xlator_t        *this = NULL;
+        int     ret = -1;
 
+        GF_ASSERT (xattr);
+        GF_ASSERT (dict);
 
-	local = frame->local;
-	this = frame->this;
+        this = THIS;
+        GF_ASSERT (this);
 
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err == ENOENT || force)
-			missing_dirs++;
-	}
+        acl_default = dict_get (xattr, POSIX_ACL_DEFAULT_XATTR);
 
-	if (missing_dirs == 0) {
-		dht_selfheal_dir_xattr (frame, loc, layout);
-		return 0;
-	}
+        if (!acl_default) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "ACL_DEFAULT xattr not present");
+                goto cont;
+        }
+        ret = dict_set (dict, POSIX_ACL_DEFAULT_XATTR, acl_default);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Could not set ACL_DEFAULT xattr");
+cont:
+        acl_access = dict_get (xattr, POSIX_ACL_ACCESS_XATTR);
+        if (!acl_access) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "ACL_ACCESS xattr not present");
+                goto out;
+        }
+        ret = dict_set (dict, POSIX_ACL_ACCESS_XATTR, acl_access);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Could not set ACL_ACCESS xattr");
+
+out:
+        return;
+}
+
+int
+dht_selfheal_dir_mkdir (call_frame_t *frame, loc_t *loc,
+                        dht_layout_t *layout, int force)
+{
+        int           missing_dirs = 0;
+        int           i     = 0;
+        int           ret   = -1;
+        dht_local_t  *local = NULL;
+        xlator_t     *this = NULL;
+        dict_t       *dict = NULL;
+
+        local = frame->local;
+        this = frame->this;
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == ENOENT || force)
+                        missing_dirs++;
+        }
 
-	local->call_cnt = missing_dirs;
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err == ENOENT || force) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"creating directory %s on subvol %s",
-				loc->path, layout->list[i].xlator->name);
-
-			STACK_WIND (frame, dht_selfheal_dir_mkdir_cbk,
-				    layout->list[i].xlator,
-				    layout->list[i].xlator->fops->mkdir,
-				    loc,
+        if (missing_dirs == 0) {
+                dht_selfheal_dir_setattr (frame, loc, &local->stbuf, 0xffffffff, layout);
+                return 0;
+        }
+
+        local->call_cnt = missing_dirs;
+        if (!uuid_is_null (local->gfid)) {
+                dict = dict_new ();
+                if (!dict)
+                        return -1;
+
+                ret = dict_set_static_bin (dict, "gfid-req", local->gfid, 16);
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set gfid in dict", loc->path);
+        } else if (local->params) {
+                /* Send the dictionary from higher layers directly */
+                dict = dict_ref (local->params);
+        }
+        /* Set acls */
+        if (local->xattr && dict)
+                dht_selfheal_dir_mkdir_setacl (local->xattr, dict);
+
+        if (!dict)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dict is NULL, need to make sure gfids are same");
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == ENOENT || force) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "creating directory %s on subvol %s",
+                                loc->path, layout->list[i].xlator->name);
+
+                        STACK_WIND (frame, dht_selfheal_dir_mkdir_cbk,
+                                    layout->list[i].xlator,
+                                    layout->list[i].xlator->fops->mkdir,
+                                    loc,
                                     st_mode_from_ia (local->stbuf.ia_prot,
                                                      local->stbuf.ia_type),
-                                    NULL);
-		}
-	}
+                                    0, dict);
+                }
+        }
+
+        if (dict)
+                dict_unref (dict);
 
-	return 0;
+        return 0;
 }
 
 
@@ -294,7 +534,7 @@ dht_selfheal_layout_alloc_start (xlator_t *this, loc_t *loc,
         uint32_t      hashval = 0;
         int           ret = 0;
 
-        ret = dht_hash_compute (layout->type, loc->path, &hashval);
+        ret = dht_hash_compute (this, layout->type, loc->path, &hashval);
         if (ret == 0) {
                 start = (hashval % layout->cnt);
         }
@@ -302,235 +542,488 @@ dht_selfheal_layout_alloc_start (xlator_t *this, loc_t *loc,
         return start;
 }
 
-
-void
-dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
-				   dht_layout_t *layout)
+static inline int
+dht_get_layout_count (xlator_t *this, dht_layout_t *layout, int new_layout)
 {
-	xlator_t    *this = NULL;
-	uint32_t     chunk = 0;
-	int          i = 0;
-	uint32_t     start = 0;
-	int          cnt = 0;
-	int          err = 0;
-        int          start_subvol = 0;
-
-	this = frame->this;
+        int i = 0;
+        int j = 0;
+        int err = 0;
+        int count = 0;
+        dht_conf_t *conf = NULL;
+
+        /* Gets in use only for replace-brick, remove-brick */
+        conf = this->private;
+        for (i = 0; i < layout->cnt; i++) {
+                for (j = 0; j < conf->subvolume_cnt; j++) {
+                        if (conf->decommissioned_bricks[j] &&
+                            conf->decommissioned_bricks[j] == layout->list[i].xlator) {
+                                layout->list[i].err = EINVAL;
+                                break;
+                        }
+                }
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-		err = layout->list[i].err;
-		if (err == -1 || err == 0) {
-			layout->list[i].err = -1;
-			cnt++;
-		}
-	}
+        for (i = 0; i < layout->cnt; i++) {
+                err = layout->list[i].err;
+                if (err == -1 || err == 0 || err == ENOENT) {
+			/* Setting list[i].err = -1 is an indication for
+			   dht_selfheal_layout_new_directory() to assign
+			   a range. We set it to -1 based on any one of
+			   the three criteria:
+
+			   - err == -1 already, which means directory
+			     existed but layout was not set on it.
+
+			   - err == 0, which means directory exists and
+			     has an old layout piece which will be
+			     overwritten now.
+
+			   - err == ENOENT, which means directory does
+			     not exist (possibly racing with mkdir or
+			     finishing half done mkdir). The missing
+			     directory will be attempted to be recreated.
+
+			     It is important to note that it is safe
+			     to race with mkdir() as self-heal and
+			     mkdir are idempotent operations. Both will
+			     strive to set the directory and layouts to
+			     the same final state.
+			*/
+                        count++;
+			if (!err)
+				layout->list[i].err = -1;
+                }
+        }
 
         /* no subvolume has enough space, but can't stop directory creation */
-        if (!cnt) {
+        if (!count || !new_layout) {
                 for (i = 0; i < layout->cnt; i++) {
                         err = layout->list[i].err;
                         if (err == ENOSPC) {
                                 layout->list[i].err = -1;
-                                cnt++;
+                                count++;
                         }
                 }
         }
 
-	chunk = ((unsigned long) 0xffffffff) / ((cnt) ? cnt : 1);
+        /* if layout->spread_cnt is set, check if it is <= available
+         * subvolumes (down brick and decommissioned bricks are considered
+         * un-availbale). Else return count (available up bricks) */
+        count = ((layout->spread_cnt &&
+                 (layout->spread_cnt <= count)) ?
+                 layout->spread_cnt : ((count) ? count : 1));
 
-	start_subvol = dht_selfheal_layout_alloc_start (this, loc, layout);
+        return count;
+}
 
-	for (i = start_subvol; i < layout->cnt; i++) {
-		err = layout->list[i].err;
-		if (err == -1) {
-			layout->list[i].start = start;
-			layout->list[i].stop  = start + chunk - 1;
-			
-			start = start + chunk;
 
-			gf_log (this->name, GF_LOG_TRACE,
-				"gave fix: %u - %u on %s for %s",
-				layout->list[i].start, layout->list[i].stop,
-				layout->list[i].xlator->name, loc->path);
-			if (--cnt == 0) {
-				layout->list[i].stop = 0xffffffff;
-				break;
-			}
-		}
-	}
+void dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
+					dht_layout_t *new_layout);
+
+void dht_layout_entry_swap (dht_layout_t *layout, int i, int j);
+void dht_layout_range_swap (dht_layout_t *layout, int i, int j);
+
+/*
+ * It's a bit icky using local variables in a macro, but it makes the rest
+ * of the code a lot clearer.
+ */
+#define OV_ENTRY(x,y)      table[x*new->cnt+y]
+
+void
+dht_selfheal_layout_maximize_overlap (call_frame_t *frame, loc_t *loc,
+				      dht_layout_t *new, dht_layout_t *old)
+{
+	int           i            = 0;
+	int           j            = 0;
+	uint32_t      curr_overlap = 0;
+	uint32_t      max_overlap  = 0;
+	int           max_overlap_idx = -1;
+	uint32_t      overlap      = 0;
+        uint32_t     *table = NULL;
+
+	dht_layout_sort_volname (old);
+	/* Now both old_layout->list[] and new_layout->list[]
+	   are match the same xlators/subvolumes. i.e,
+	   old_layout->[i] and new_layout->[i] are referring
+	   to the same subvolumes
+	*/
+
+        /* Build a table of overlaps between new[i] and old[j]. */
+        table = alloca(sizeof(overlap)*old->cnt*new->cnt);
+        if (!table) {
+                return;
+        }
+        memset(table,0,sizeof(overlap)*old->cnt*new->cnt);
+        for (i = 0; i < new->cnt; ++i) {
+                for (j = 0; j < old->cnt; ++j) {
+                        OV_ENTRY(i,j) = dht_overlap_calc(old,j,new,i);
+                }
+        }
+
+	for (i = 0; i < new->cnt; i++) {
+		if (new->list[i].err > 0) {
+			/* Subvol might be marked for decommission
+			   with EINVAL, or some other serious error
+			   marked with positive errno.
+			*/
+			continue;
+                }
+
+                max_overlap = 0;
+                max_overlap_idx = i;
+                for (j = (i + 1); j < new->cnt; ++j) {
+                        if (new->list[j].err > 0) {
+			        /* Subvol might be marked for decommission
+			        with EINVAL, or some other serious error
+			        marked with positive errno.
+			        */
+			        continue;
+                        }
+                        /* Calculate the overlap now. */
+                        curr_overlap = OV_ENTRY(i,i) + OV_ENTRY(j,j);
+                        /* Calculate the overlap after the proposed swap. */
+                        overlap = OV_ENTRY(i,j) + OV_ENTRY(j,i);
+                        /* Are we better than status quo? */
+                        if (overlap > curr_overlap) {
+                                overlap -= curr_overlap;
+                                /* Are we better than the previous choice? */
+                                if (overlap > max_overlap) {
+                                        max_overlap = overlap;
+                                        max_overlap_idx = j;
+                                }
+                        }
+                }
 
-	for (i = 0; i < start_subvol; i++) {
-		err = layout->list[i].err;
-		if (err == -1) {
-			layout->list[i].start = start;
-			layout->list[i].stop  = start + chunk - 1;
-			
-			start = start + chunk;
-
-			gf_log (this->name, GF_LOG_TRACE,
-				"gave fix: %u - %u on %s for %s",
-				layout->list[i].start, layout->list[i].stop,
-				layout->list[i].xlator->name, loc->path);
-			if (--cnt == 0) {
-				layout->list[i].stop = 0xffffffff;
-				break;
-			}
-		}
+		if (max_overlap_idx != i) {
+ 			dht_layout_range_swap (new, i, max_overlap_idx);
+                        /* Need to swap the table values too. */
+                        for (j = 0; j < old->cnt; ++j) {
+                                overlap = OV_ENTRY(i,j);
+                                OV_ENTRY(i,j) = OV_ENTRY(max_overlap_idx,j);
+                                OV_ENTRY(max_overlap_idx,j) = overlap;
+                        }
+                }
 	}
 }
 
 
+dht_layout_t *
+dht_fix_layout_of_directory (call_frame_t *frame, loc_t *loc,
+                             dht_layout_t *layout)
+{
+        int           i            = 0;
+        xlator_t     *this         = NULL;
+        dht_layout_t *new_layout   = NULL;
+        dht_conf_t   *priv         = NULL;
+        dht_local_t  *local        = NULL;
+        uint32_t      subvol_down  = 0;
+        int           ret          = 0;
+
+        this  = frame->this;
+        priv  = this->private;
+        local = frame->local;
+
+        if (layout->type == DHT_HASH_TYPE_DM_USER) {
+                gf_log (THIS->name, GF_LOG_DEBUG, "leaving %s alone",
+                        loc->path);
+                goto done;
+        }
+
+        new_layout = dht_layout_new (this, priv->subvolume_cnt);
+        if (!new_layout)
+                goto done;
+
+        /* If a subvolume is down, do not re-write the layout. */
+        ret = dht_layout_anomalies (this, loc, layout, NULL, NULL, NULL,
+                                    &subvol_down, NULL, NULL);
+
+        if (subvol_down || (ret == -1)) {
+                gf_log (this->name, GF_LOG_WARNING, "%u subvolume(s) are down"
+                        ". Skipping fix layout.", subvol_down);
+                GF_FREE (new_layout);
+                return NULL;
+        }
+
+        for (i = 0; i < new_layout->cnt; i++) {
+		if (layout->list[i].err != ENOSPC)
+			new_layout->list[i].err = layout->list[i].err;
+		else
+			new_layout->list[i].err = -1;
+
+		new_layout->list[i].xlator = layout->list[i].xlator;
+        }
+
+	/* First give it a layout as though it is a new directory. This
+	   ensures rotation to kick in */
+        dht_layout_sort_volname (new_layout);
+	dht_selfheal_layout_new_directory (frame, loc, new_layout);
+
+	/* Now selectively re-assign ranges only when it helps */
+	dht_selfheal_layout_maximize_overlap (frame, loc, new_layout, layout);
+
+done:
+        if (new_layout) {
+                /* Now that the new layout has all the proper layout, change the
+                   inode context */
+                dht_layout_set (this, loc->inode, new_layout);
+
+                /* Make sure the extra 'ref' for existing layout is removed */
+                dht_layout_unref (this, local->layout);
+
+                local->layout = new_layout;
+        }
+
+        return local->layout;
+}
+
+
+void
+dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
+                                   dht_layout_t *layout)
+{
+        xlator_t    *this = NULL;
+        uint32_t     chunk = 0;
+        int          i = 0;
+        uint32_t     start = 0;
+        int          cnt = 0;
+        int          err = 0;
+        int          start_subvol = 0;
+
+        this = frame->this;
+
+        cnt = dht_get_layout_count (this, layout, 1);
+
+        chunk = ((unsigned long) 0xffffffff) / ((cnt) ? cnt : 1);
+
+        start_subvol = dht_selfheal_layout_alloc_start (this, loc, layout);
+
+        /* clear out the range, as we are re-computing here */
+        DHT_RESET_LAYOUT_RANGE (layout);
+        for (i = start_subvol; i < layout->cnt; i++) {
+                err = layout->list[i].err;
+                if (err == -1 || err == ENOENT) {
+                        DHT_SET_LAYOUT_RANGE(layout, i, start, chunk,
+                                             cnt, loc->path);
+                        if (--cnt == 0) {
+                                layout->list[i].stop = 0xffffffff;
+                                goto done;
+                        }
+                        start += chunk;
+                }
+        }
+
+        for (i = 0; i < start_subvol; i++) {
+                err = layout->list[i].err;
+                if (err == -1 || err == ENOENT) {
+                        DHT_SET_LAYOUT_RANGE(layout, i, start, chunk,
+                                             cnt, loc->path);
+                        if (--cnt == 0) {
+                                layout->list[i].stop = 0xffffffff;
+                                goto done;
+                        }
+                        start += chunk;
+                }
+        }
+
+done:
+        return;
+}
+
 int
 dht_selfheal_dir_getafix (call_frame_t *frame, loc_t *loc,
-			  dht_layout_t *layout)
+                          dht_layout_t *layout)
 {
-	dht_conf_t  *conf = NULL;
-	xlator_t    *this = NULL;
-	dht_local_t *local = NULL;
-	int          missing = -1;
-	int          down = -1;
-	int          holes = -1;
-	int          ret = -1;
-	int          i = -1;
-	int          overlaps = -1;
-
-	this = frame->this;
-	conf = this->private;
-	local = frame->local;
-
-	missing = local->selfheal.missing;
-	down = local->selfheal.down;
-	holes = local->selfheal.hole_cnt;
-	overlaps = local->selfheal.overlaps_cnt;
-
-	if ((missing + down) == conf->subvolume_cnt) {
-		dht_selfheal_layout_new_directory (frame, loc, layout);
-		ret = 0;
-	}
+        dht_local_t *local = NULL;
+        uint32_t     holes = 0;
+        int          ret = -1;
+        int          i = -1;
+        uint32_t     overlaps = 0;
 
-	if (holes <= down) {
-		/* the down subvol might fill up the holes */
-		ret = 0;
-	}
+        local = frame->local;
 
-	if (holes || overlaps) {
-		dht_selfheal_layout_new_directory (frame, loc, layout);
-		ret = 0;
-	}
+        holes = local->selfheal.hole_cnt;
+        overlaps = local->selfheal.overlaps_cnt;
 
-	for (i = 0; i < layout->cnt; i++) {
-		/* directory not present */
-		if (layout->list[i].err == ENOENT) {
-			ret = 0;
-			break;
-		}
-	}
+        if (holes || overlaps) {
+                dht_selfheal_layout_new_directory (frame, loc, layout);
+                ret = 0;
+        }
+
+        for (i = 0; i < layout->cnt; i++) {
+                /* directory not present */
+                if (layout->list[i].err == ENOENT) {
+                        ret = 0;
+                        break;
+                }
+        }
+
+        /* TODO: give a fix to these non-virgins */
+
+        return ret;
+}
+
+int
+dht_selfheal_new_directory (call_frame_t *frame,
+                            dht_selfheal_dir_cbk_t dir_cbk,
+                            dht_layout_t *layout)
+{
+        dht_local_t *local = NULL;
 
-	/* TODO: give a fix to these non-virgins */
+        local = frame->local;
 
-	return ret;
+        local->selfheal.dir_cbk = dir_cbk;
+        local->selfheal.layout = dht_layout_ref (frame->this, layout);
+
+        dht_layout_sort_volname (layout);
+        dht_selfheal_layout_new_directory (frame, &local->loc, layout);
+        dht_selfheal_dir_xattr (frame, &local->loc, layout);
+        return 0;
 }
 
 int
-dht_selfheal_new_directory (call_frame_t *frame, 
-			    dht_selfheal_dir_cbk_t dir_cbk,
-			    dht_layout_t *layout)
+dht_fix_directory_layout (call_frame_t *frame,
+                          dht_selfheal_dir_cbk_t dir_cbk,
+                          dht_layout_t *layout)
 {
-	dht_local_t *local = NULL;
+        dht_local_t  *local = NULL;
+        dht_layout_t *tmp_layout = NULL;
+
+        local = frame->local;
 
-	local = frame->local;
+        local->selfheal.dir_cbk = dir_cbk;
+        local->selfheal.layout = dht_layout_ref (frame->this, layout);
 
-	local->selfheal.dir_cbk = dir_cbk;
-	local->selfheal.layout = dht_layout_ref (frame->this, layout);
+        /* No layout sorting required here */
+        tmp_layout = dht_fix_layout_of_directory (frame, &local->loc, layout);
+        if (!tmp_layout) {
+                return -1;
+        }
+        dht_fix_dir_xattr (frame, &local->loc, tmp_layout);
 
-	dht_layout_sort_volname (layout);
-	dht_selfheal_layout_new_directory (frame, &local->loc, layout);
-	dht_selfheal_dir_xattr (frame, &local->loc, layout);
-	return 0;
+        return 0;
 }
 
 
 int
 dht_selfheal_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
-			loc_t *loc, dht_layout_t *layout)
+                        loc_t *loc, dht_layout_t *layout)
 {
-	dht_local_t *local    = NULL;
-	uint32_t     holes    = 0;
-	uint32_t     down     = 0;
-	uint32_t     misc     = 0;
-	int          ret      = 0;
-	xlator_t    *this     = NULL;
+        dht_local_t *local    = NULL;
+        uint32_t     down     = 0;
+        uint32_t     misc     = 0;
+        int          ret      = 0;
+        xlator_t    *this     = NULL;
 
-	local = frame->local;
-	this = frame->this;
+        local = frame->local;
+        this = frame->this;
 
-	dht_layout_anomalies (this, loc, layout,
+        dht_layout_anomalies (this, loc, layout,
                               &local->selfheal.hole_cnt,
                               &local->selfheal.overlaps_cnt,
-                              &local->selfheal.missing,
-                              &local->selfheal.down,
-                              &local->selfheal.misc);
-
-	holes    = local->selfheal.hole_cnt;
-	down     = local->selfheal.down;
-	misc     = local->selfheal.misc;
-
-	local->selfheal.dir_cbk = dir_cbk;
-	local->selfheal.layout = dht_layout_ref (this, layout);
-
-	if (down) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"%d subvolumes down -- not fixing", down);
-                ret = -1;
-                local->op_errno = ESTALE;
-		goto sorry_no_fix;
-	}
+                              NULL, &local->selfheal.down,
+                              &local->selfheal.misc, NULL);
 
-	if (misc) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"%d subvolumes have unrecoverable errors", misc);
-                ret = -1;
-                local->op_errno = ESTALE;
-		goto sorry_no_fix;
-	}
+        down     = local->selfheal.down;
+        misc     = local->selfheal.misc;
 
-	dht_layout_sort_volname (layout);
-	ret = dht_selfheal_dir_getafix (frame, loc, layout);
+        local->selfheal.dir_cbk = dir_cbk;
+        local->selfheal.layout = dht_layout_ref (this, layout);
 
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"not able to form layout for the directory");
-                ret = -1;
-                local->op_errno = ESTALE;
-		goto sorry_no_fix;
-	}
+        if (down) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%d subvolumes down -- not fixing", down);
+                ret = 0;
+                goto sorry_no_fix;
+        }
 
-	dht_selfheal_dir_mkdir (frame, loc, layout, 0);
+        if (misc) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%d subvolumes have unrecoverable errors", misc);
+                ret = 0;
+                goto sorry_no_fix;
+        }
+
+        dht_layout_sort_volname (layout);
+        ret = dht_selfheal_dir_getafix (frame, loc, layout);
+
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "not able to form layout for the directory");
+                goto sorry_no_fix;
+        }
 
-	return 0;
+        dht_selfheal_dir_mkdir (frame, loc, layout, 0);
+
+        return 0;
 
 sorry_no_fix:
-	/* TODO: need to put appropriate local->op_errno */
-	dht_selfheal_dir_finish (frame, this, ret);
+        /* TODO: need to put appropriate local->op_errno */
+        dht_selfheal_dir_finish (frame, this, ret);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_selfheal_restore (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
-		      loc_t *loc, dht_layout_t *layout)
+                      loc_t *loc, dht_layout_t *layout)
 {
-	int          ret = 0;
-	dht_local_t *local    = NULL;
+        int          ret = 0;
+        dht_local_t *local    = NULL;
+
+        local = frame->local;
 
+        local->selfheal.dir_cbk = dir_cbk;
+        local->selfheal.layout = dht_layout_ref (frame->this, layout);
 
-	local = frame->local;
+        ret = dht_selfheal_dir_mkdir (frame, loc, layout, 1);
 
-	local->selfheal.dir_cbk = dir_cbk;
-	local->selfheal.layout = dht_layout_ref (frame->this, layout);
+        return ret;
+}
 
-	ret = dht_selfheal_dir_mkdir (frame, loc, layout, 1);
+int
+dht_dir_attr_heal (void *data)
+{
+        call_frame_t    *frame = NULL;
+        dht_local_t     *local = NULL;
+        xlator_t        *subvol = NULL;
+        xlator_t        *this  = NULL;
+        dht_conf_t      *conf  = NULL;
+        int              call_cnt = 0;
+        int              ret   = -1;
+        int              i     = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", data, out);
+
+        frame = data;
+        local = frame->local;
+        this = frame->this;
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", local, out);
+        conf = this->private;
+        GF_VALIDATE_OR_GOTO ("dht", conf, out);
+
+        call_cnt = conf->subvolume_cnt;
+
+        for (i = 0; i < call_cnt; i++) {
+                subvol = conf->subvolumes[i];
+                if (!subvol || (subvol == dht_first_up_subvol (this)))
+                        continue;
+                ret = syncop_setattr (subvol, &local->loc, &local->stbuf,
+                                      (GF_SET_ATTR_UID | GF_SET_ATTR_GID),
+                                      NULL, NULL);
+                if (ret)
+                        gf_log ("dht", GF_LOG_ERROR, "Failed to set uid/gid on"
+                                " %s on %s subvol (%s)", local->loc.path,
+                                subvol->name, strerror (errno));
+        }
+out:
+        return 0;
+}
 
-	return ret;
+int
+dht_dir_attr_heal_done (int ret, call_frame_t *sync_frame, void *data)
+{
+        DHT_STACK_DESTROY (sync_frame);
+        return 0;
 }
diff --git a/xlators/cluster/dht/src/dht-shared.c b/xlators/cluster/dht/src/dht-shared.c
new file mode 100644
index 000000000..70aac7710
--- /dev/null
+++ b/xlators/cluster/dht/src/dht-shared.c
@@ -0,0 +1,758 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+/* TODO: add NS locking */
+
+#include "statedump.h"
+#include "dht-common.h"
+
+/* TODO:
+   - use volumename in xattr instead of "dht"
+   - use NS locks
+   - handle all cases in self heal layout reconstruction
+   - complete linkfile selfheal
+*/
+struct volume_options options[];
+
+void
+dht_layout_dump (dht_layout_t  *layout, const char *prefix)
+{
+
+        char    key[GF_DUMP_MAX_BUF_LEN];
+        int     i = 0;
+
+        if (!layout)
+                goto out;
+        if (!prefix)
+                goto out;
+
+        gf_proc_dump_build_key(key, prefix, "cnt");
+        gf_proc_dump_write(key, "%d", layout->cnt);
+        gf_proc_dump_build_key(key, prefix, "preset");
+        gf_proc_dump_write(key, "%d", layout->preset);
+        gf_proc_dump_build_key(key, prefix, "gen");
+        gf_proc_dump_write(key, "%d", layout->gen);
+        if (layout->type != IA_INVAL) {
+                gf_proc_dump_build_key(key, prefix, "inode type");
+                gf_proc_dump_write(key, "%d", layout->type);
+        }
+
+        if  (!IA_ISDIR (layout->type))
+                goto out;
+
+        for (i = 0; i < layout->cnt; i++) {
+                gf_proc_dump_build_key(key, prefix,"list[%d].err", i);
+                gf_proc_dump_write(key, "%d", layout->list[i].err);
+                gf_proc_dump_build_key(key, prefix,"list[%d].start", i);
+                gf_proc_dump_write(key, "%u", layout->list[i].start);
+                gf_proc_dump_build_key(key, prefix,"list[%d].stop", i);
+                gf_proc_dump_write(key, "%u", layout->list[i].stop);
+                if (layout->list[i].xlator) {
+                        gf_proc_dump_build_key(key, prefix,
+                                               "list[%d].xlator.type", i);
+                        gf_proc_dump_write(key, "%s",
+                                           layout->list[i].xlator->type);
+                        gf_proc_dump_build_key(key, prefix,
+                                               "list[%d].xlator.name", i);
+                        gf_proc_dump_write(key, "%s",
+                                           layout->list[i].xlator->name);
+                }
+        }
+
+out:
+        return;
+}
+
+
+int32_t
+dht_priv_dump (xlator_t *this)
+{
+        char            key_prefix[GF_DUMP_MAX_BUF_LEN];
+        char            key[GF_DUMP_MAX_BUF_LEN];
+        int             i = 0;
+        dht_conf_t      *conf = NULL;
+        int             ret = -1;
+
+        if (!this)
+                goto out;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        ret = TRY_LOCK(&conf->subvolume_lock);
+        if (ret != 0) {
+                return ret;
+        }
+
+        gf_proc_dump_add_section("xlator.cluster.dht.%s.priv", this->name);
+        gf_proc_dump_build_key(key_prefix,"xlator.cluster.dht","%s.priv",
+                               this->name);
+        gf_proc_dump_write("subvol_cnt","%d", conf->subvolume_cnt);
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                sprintf (key, "subvolumes[%d]", i);
+                gf_proc_dump_write(key, "%s.%s", conf->subvolumes[i]->type,
+                                   conf->subvolumes[i]->name);
+                if (conf->file_layouts && conf->file_layouts[i]){
+                        sprintf (key, "file_layouts[%d]", i);
+                        dht_layout_dump(conf->file_layouts[i], key);
+                }
+                if (conf->dir_layouts && conf->dir_layouts[i]) {
+                        sprintf (key, "dir_layouts[%d]", i);
+                        dht_layout_dump(conf->dir_layouts[i], key);
+                }
+                if (conf->subvolume_status) {
+
+                        sprintf (key, "subvolume_status[%d]", i);
+                        gf_proc_dump_write(key, "%d",
+                                           (int)conf->subvolume_status[i]);
+                }
+
+        }
+
+        gf_proc_dump_write("search_unhashed", "%d", conf->search_unhashed);
+        gf_proc_dump_write("gen", "%d", conf->gen);
+        gf_proc_dump_write("min_free_disk", "%lf", conf->min_free_disk);
+	gf_proc_dump_write("min_free_inodes", "%lf", conf->min_free_inodes);
+        gf_proc_dump_write("disk_unit", "%c", conf->disk_unit);
+        gf_proc_dump_write("refresh_interval", "%d", conf->refresh_interval);
+        gf_proc_dump_write("unhashed_sticky_bit", "%d", conf->unhashed_sticky_bit);
+        if (conf ->du_stats) {
+                gf_proc_dump_write("du_stats.avail_percent", "%lf",
+                                   conf->du_stats->avail_percent);
+                gf_proc_dump_write("du_stats.avail_space", "%lu",
+                                   conf->du_stats->avail_space);
+		gf_proc_dump_write("du_stats.avail_inodes", "%lf",
+                                   conf->du_stats->avail_inodes);
+                gf_proc_dump_write("du_stats.log", "%lu", conf->du_stats->log);
+        }
+
+        if (conf->last_stat_fetch.tv_sec)
+                gf_proc_dump_write("last_stat_fetch", "%s",
+                                    ctime(&conf->last_stat_fetch.tv_sec));
+
+        UNLOCK(&conf->subvolume_lock);
+
+out:
+        return ret;
+}
+
+int32_t
+dht_inodectx_dump (xlator_t *this, inode_t *inode)
+{
+        int             ret = -1;
+        dht_layout_t    *layout = NULL;
+
+        if (!this)
+                goto out;
+        if (!inode)
+                goto out;
+
+        ret = dht_inode_ctx_layout_get (inode, this, &layout);
+
+        if ((ret != 0) || !layout)
+                return ret;
+
+        gf_proc_dump_add_section("xlator.cluster.dht.%s.inode", this->name);
+        dht_layout_dump(layout, "layout");
+
+out:
+        return ret;
+}
+
+void
+dht_fini (xlator_t *this)
+{
+        int         i = 0;
+        dht_conf_t *conf = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+
+        conf = this->private;
+        this->private = NULL;
+        if (conf) {
+                if (conf->file_layouts) {
+                        for (i = 0; i < conf->subvolume_cnt; i++) {
+                                GF_FREE (conf->file_layouts[i]);
+                        }
+                        GF_FREE (conf->file_layouts);
+                }
+
+                GF_FREE (conf->subvolumes);
+
+                GF_FREE (conf->subvolume_status);
+
+                GF_FREE (conf);
+        }
+out:
+        return;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+
+        ret = xlator_mem_acct_init (this, gf_dht_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                        "failed");
+                return ret;
+        }
+out:
+        return ret;
+}
+
+
+int
+dht_parse_decommissioned_bricks (xlator_t *this, dht_conf_t *conf,
+                                 const char *bricks)
+{
+        int         i  = 0;
+        int         ret  = -1;
+        char       *tmpstr = NULL;
+        char       *dup_brick = NULL;
+        char       *node = NULL;
+
+        if (!conf || !bricks)
+                goto out;
+
+        dup_brick = gf_strdup (bricks);
+        node = strtok_r (dup_brick, ",", &tmpstr);
+        while (node) {
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (!strcmp (conf->subvolumes[i]->name, node)) {
+                                conf->decommissioned_bricks[i] =
+                                        conf->subvolumes[i];
+                                        conf->decommission_subvols_cnt++;
+                                gf_log (this->name, GF_LOG_INFO,
+                                        "decommissioning subvolume %s",
+                                        conf->subvolumes[i]->name);
+                                break;
+                        }
+                }
+                if (i == conf->subvolume_cnt) {
+                        /* Wrong node given. */
+                        goto out;
+                }
+                node = strtok_r (NULL, ",", &tmpstr);
+        }
+
+        ret = 0;
+        conf->decommission_in_progress = 1;
+out:
+        GF_FREE (dup_brick);
+
+        return ret;
+}
+
+
+int
+dht_decommissioned_remove (xlator_t *this, dht_conf_t *conf)
+{
+        int         i  = 0;
+        int         ret  = -1;
+
+        if (!conf)
+                goto out;
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->decommissioned_bricks[i]) {
+                        conf->decommissioned_bricks[i] = NULL;
+                        conf->decommission_subvols_cnt--;
+                }
+        }
+
+        ret = 0;
+out:
+
+        return ret;
+}
+void
+dht_init_regex (xlator_t *this, dict_t *odict, char *name,
+                regex_t *re, gf_boolean_t *re_valid)
+{
+        char    *temp_str;
+
+        if (dict_get_str (odict, name, &temp_str) != 0) {
+                if (strcmp(name,"rsync-hash-regex")) {
+                        return;
+                }
+                temp_str = "^\\.(.+)\\.[^.]+$";
+        }
+
+        if (*re_valid) {
+                regfree(re);
+                *re_valid = _gf_false;
+        }
+
+        if (!strcmp(temp_str,"none")) {
+                return;
+        }
+
+        if (regcomp(re,temp_str,REG_EXTENDED) == 0) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "using regex %s = %s", name, temp_str);
+                *re_valid = _gf_true;
+        }
+        else {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "compiling regex %s failed", temp_str);
+        }
+}
+
+int
+dht_reconfigure (xlator_t *this, dict_t *options)
+{
+        dht_conf_t      *conf = NULL;
+        char            *temp_str = NULL;
+        gf_boolean_t     search_unhashed;
+        int              ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", options, out);
+
+        conf = this->private;
+        if (!conf)
+                return 0;
+
+        if (dict_get_str (options, "lookup-unhashed", &temp_str) == 0) {
+                /* If option is not "auto", other options _should_ be boolean*/
+                if (strcasecmp (temp_str, "auto")) {
+                        if (!gf_string2boolean (temp_str, &search_unhashed)) {
+                                gf_log(this->name, GF_LOG_DEBUG, "Reconfigure:"
+                                       " lookup-unhashed reconfigured (%s)",
+                                       temp_str);
+                                conf->search_unhashed = search_unhashed;
+                        } else {
+                                gf_log(this->name, GF_LOG_ERROR, "Reconfigure:"
+                                       " lookup-unhashed should be boolean,"
+                                       " not (%s), defaulting to (%d)",
+                                       temp_str, conf->search_unhashed);
+                                //return -1;
+                                ret = -1;
+                                goto out;
+                        }
+                } else {
+                        gf_log(this->name, GF_LOG_DEBUG, "Reconfigure:"
+                               " lookup-unhashed reconfigured auto ");
+                        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
+                }
+        }
+
+	GF_OPTION_RECONF ("min-free-disk", conf->min_free_disk, options,
+                          percent_or_size, out);
+        /* option can be any one of percent or bytes */
+        conf->disk_unit = 0;
+        if (conf->min_free_disk < 100.0)
+                conf->disk_unit = 'p';
+
+	GF_OPTION_RECONF ("min-free-inodes", conf->min_free_inodes, options,
+                          percent, out);
+
+        GF_OPTION_RECONF ("directory-layout-spread", conf->dir_spread_cnt,
+                          options, uint32, out);
+
+        GF_OPTION_RECONF ("readdir-optimize", conf->readdir_optimize, options,
+                          bool, out);
+        if (conf->defrag) {
+                GF_OPTION_RECONF ("rebalance-stats", conf->defrag->stats,
+                                  options, bool, out);
+        }
+
+        if (dict_get_str (options, "decommissioned-bricks", &temp_str) == 0) {
+                ret = dht_parse_decommissioned_bricks (this, conf, temp_str);
+                if (ret == -1)
+                        goto out;
+        } else {
+                ret = dht_decommissioned_remove (this, conf);
+                if (ret == -1)
+                        goto out;
+        }
+
+        dht_init_regex (this, options, "rsync-hash-regex",
+                        &conf->rsync_regex, &conf->rsync_regex_valid);
+        dht_init_regex (this, options, "extra-hash-regex",
+                        &conf->extra_regex, &conf->extra_regex_valid);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+static int
+gf_defrag_pattern_list_fill (xlator_t *this, gf_defrag_info_t *defrag, char *data)
+{
+        int                    ret = -1;
+        char                  *tmp_str = NULL;
+        char                  *tmp_str1 = NULL;
+        char                  *dup_str = NULL;
+        char                  *num = NULL;
+        char                  *pattern_str = NULL;
+        char                  *pattern = NULL;
+        gf_defrag_pattern_list_t *temp_list = NULL;
+        gf_defrag_pattern_list_t *pattern_list = NULL;
+
+        if (!this || !defrag || !data)
+                goto out;
+
+        /* Get the pattern for pattern list. "pattern:<optional-size>"
+         * eg: *avi, *pdf:10MB, *:1TB
+         */
+        pattern_str = strtok_r (data, ",", &tmp_str);
+        while (pattern_str) {
+                dup_str = gf_strdup (pattern_str);
+                pattern_list = GF_CALLOC (1, sizeof (gf_defrag_pattern_list_t),
+                                        1);
+                if (!pattern_list) {
+                        goto out;
+                }
+                pattern = strtok_r (dup_str, ":", &tmp_str1);
+                num = strtok_r (NULL, ":", &tmp_str1);
+                if (!pattern)
+                        goto out;
+                if (!num) {
+                        if (gf_string2bytesize(pattern, &pattern_list->size)
+                             == 0) {
+                                pattern = "*";
+                        }
+                } else if (gf_string2bytesize (num, &pattern_list->size) != 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "invalid number format \"%s\"", num);
+                        goto out;
+                }
+                memcpy (pattern_list->path_pattern, pattern, strlen (dup_str));
+
+                if (!defrag->defrag_pattern)
+                        temp_list = NULL;
+                else
+                        temp_list = defrag->defrag_pattern;
+
+                pattern_list->next = temp_list;
+
+                defrag->defrag_pattern = pattern_list;
+                pattern_list = NULL;
+
+                GF_FREE (dup_str);
+                dup_str = NULL;
+
+                pattern_str = strtok_r (NULL, ",", &tmp_str);
+        }
+
+        ret = 0;
+out:
+        if (ret)
+                GF_FREE (pattern_list);
+        GF_FREE (dup_str);
+
+        return ret;
+}
+
+int
+dht_init (xlator_t *this)
+{
+        dht_conf_t                      *conf           = NULL;
+        char                            *temp_str       = NULL;
+        int                              ret            = -1;
+        int                              i              = 0;
+        gf_defrag_info_t                *defrag         = NULL;
+        int                              cmd            = 0;
+        char                            *node_uuid      = NULL;
+
+
+        GF_VALIDATE_OR_GOTO ("dht", this, err);
+
+        if (!this->children) {
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "Distribute needs more than one subvolume");
+                return -1;
+        }
+
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dangling volume. check volfile");
+        }
+
+        conf = GF_CALLOC (1, sizeof (*conf), gf_dht_mt_dht_conf_t);
+        if (!conf) {
+                goto err;
+        }
+
+        ret = dict_get_int32 (this->options, "rebalance-cmd", &cmd);
+
+        if (cmd) {
+                defrag = GF_CALLOC (1, sizeof (gf_defrag_info_t),
+                                    gf_defrag_info_mt);
+
+                GF_VALIDATE_OR_GOTO (this->name, defrag, err);
+
+                LOCK_INIT (&defrag->lock);
+
+                defrag->is_exiting = 0;
+
+                conf->defrag = defrag;
+
+                ret = dict_get_str (this->options, "node-uuid", &node_uuid);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "node-uuid not "
+                                "specified");
+                        goto err;
+                }
+
+                if (uuid_parse (node_uuid, defrag->node_uuid)) {
+                        gf_log (this->name, GF_LOG_ERROR, "Cannot parse "
+                                "glusterd node uuid");
+                        goto err;
+                }
+
+                defrag->cmd = cmd;
+
+                defrag->stats = _gf_false;
+        }
+
+        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
+        if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
+                /* If option is not "auto", other options _should_ be boolean */
+                if (strcasecmp (temp_str, "auto"))
+                        gf_string2boolean (temp_str, &conf->search_unhashed);
+                else
+                        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
+        }
+
+        GF_OPTION_INIT ("unhashed-sticky-bit", conf->unhashed_sticky_bit, bool,
+                        err);
+
+        GF_OPTION_INIT ("use-readdirp", conf->use_readdirp, bool, err);
+
+	GF_OPTION_INIT ("min-free-disk", conf->min_free_disk, percent_or_size,
+			err);
+
+        GF_OPTION_INIT ("min-free-inodes", conf->min_free_inodes, percent,
+                        err);
+
+        conf->dir_spread_cnt = conf->subvolume_cnt;
+        GF_OPTION_INIT ("directory-layout-spread", conf->dir_spread_cnt,
+                        uint32, err);
+
+        GF_OPTION_INIT ("assert-no-child-down", conf->assert_no_child_down,
+                        bool, err);
+
+        GF_OPTION_INIT ("readdir-optimize", conf->readdir_optimize, bool, err);
+
+        if (defrag) {
+                GF_OPTION_INIT ("rebalance-stats", defrag->stats, bool, err);
+                if (dict_get_str (this->options, "rebalance-filter", &temp_str)
+                    == 0) {
+                        if (gf_defrag_pattern_list_fill (this, defrag, temp_str)
+                            == -1) {
+                                gf_log (this->name, GF_LOG_ERROR, "Cannot parse"
+                                        " rebalance-filter (%s)", temp_str);
+                                goto err;
+                        }
+                }
+        }
+
+        /* option can be any one of percent or bytes */
+        conf->disk_unit = 0;
+        if (conf->min_free_disk < 100)
+                conf->disk_unit = 'p';
+
+        ret = dht_init_subvolumes (this, conf);
+        if (ret == -1) {
+                goto err;
+        }
+
+        if (dict_get_str (this->options, "decommissioned-bricks", &temp_str) == 0) {
+                ret = dht_parse_decommissioned_bricks (this, conf, temp_str);
+                if (ret == -1)
+                        goto err;
+        }
+
+        dht_init_regex (this, this->options, "rsync-hash-regex",
+                        &conf->rsync_regex, &conf->rsync_regex_valid);
+        dht_init_regex (this, this->options, "extra-hash-regex",
+                        &conf->extra_regex, &conf->extra_regex_valid);
+
+        ret = dht_layouts_init (this, conf);
+        if (ret == -1) {
+                goto err;
+        }
+
+        LOCK_INIT (&conf->subvolume_lock);
+        LOCK_INIT (&conf->layout_lock);
+
+        conf->gen = 1;
+
+        this->local_pool = mem_pool_new (dht_local_t, 512);
+        if (!this->local_pool) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto err;
+        }
+
+        GF_OPTION_INIT ("xattr-name", conf->xattr_name, str, err);
+        gf_asprintf (&conf->link_xattr_name, "%s.linkto", conf->xattr_name);
+        gf_asprintf (&conf->wild_xattr_name, "%s*", conf->xattr_name);
+        if (!conf->link_xattr_name || !conf->wild_xattr_name) {
+                goto err;
+        }
+
+        this->private = conf;
+
+        return 0;
+
+err:
+        if (conf) {
+                if (conf->file_layouts) {
+                        for (i = 0; i < conf->subvolume_cnt; i++) {
+                                GF_FREE (conf->file_layouts[i]);
+                        }
+                        GF_FREE (conf->file_layouts);
+                }
+
+                GF_FREE (conf->subvolumes);
+
+                GF_FREE (conf->subvolume_status);
+
+                GF_FREE (conf->du_stats);
+
+                GF_FREE (conf->defrag);
+
+                GF_FREE (conf->xattr_name);
+                GF_FREE (conf->link_xattr_name);
+                GF_FREE (conf->wild_xattr_name);
+
+                GF_FREE (conf);
+        }
+
+        return -1;
+}
+
+
+struct volume_options options[] = {
+        { .key  = {"lookup-unhashed"},
+          .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
+                    "on", "off"},
+          .type = GF_OPTION_TYPE_STR,
+          .default_value = "on",
+          .description = "This option if set to ON, does a lookup through "
+          "all the sub-volumes, in case a lookup didn't return any result "
+          "from the hash subvolume. If set to OFF, it does not do a lookup "
+          "on the remaining subvolumes."
+        },
+        { .key  = {"min-free-disk"},
+          .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
+          .default_value = "10%",
+          .description = "Percentage/Size of disk space, after which the "
+          "process starts balancing out the cluster, and logs will appear "
+          "in log files",
+        },
+	{ .key  = {"min-free-inodes"},
+          .type = GF_OPTION_TYPE_PERCENT,
+          .default_value = "5%",
+          .description = "after system has only N% of inodes, warnings "
+          "starts to appear in log files",
+        },
+        { .key = {"unhashed-sticky-bit"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+        },
+        { .key = {"use-readdirp"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "on",
+          .description = "This option if set to ON, forces the use of "
+          "readdirp, and hence also displays the stats of the files."
+        },
+        { .key = {"assert-no-child-down"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "This option if set to ON, in the event of "
+          "CHILD_DOWN, will call exit."
+        },
+        { .key  = {"directory-layout-spread"},
+          .type = GF_OPTION_TYPE_INT,
+          .min  = 1,
+          .validate = GF_OPT_VALIDATE_MIN,
+          .description = "Specifies the directory layout spread."
+        },
+        { .key  = {"decommissioned-bricks"},
+          .type = GF_OPTION_TYPE_ANY,
+          .description = "This option if set to ON, decommissions "
+          "the brick, so that no new data is allowed to be created "
+          "on that brick."
+        },
+        { .key  = {"rebalance-cmd"},
+          .type = GF_OPTION_TYPE_INT,
+        },
+        { .key = {"node-uuid"},
+          .type = GF_OPTION_TYPE_STR,
+        },
+        { .key = {"rebalance-stats"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "This option if set to ON displays and logs the "
+          " time taken for migration of each file, during the rebalance "
+          "process. If set to OFF, the rebalance logs will only display the "
+          "time spent in each directory."
+        },
+        { .key = {"readdir-optimize"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "This option if set to ON enables the optimization "
+          "that allows DHT to requests non-first subvolumes to filter out "
+          "directory entries."
+        },
+        { .key = {"rsync-hash-regex"},
+          .type = GF_OPTION_TYPE_STR,
+          /* Setting a default here doesn't work.  See dht_init_regex. */
+          .description = "Regular expression for stripping temporary-file "
+          "suffix and prefix used by rsync, to prevent relocation when the "
+          "file is renamed."
+        },
+        { .key = {"extra-hash-regex"},
+          .type = GF_OPTION_TYPE_STR,
+          /* Setting a default here doesn't work.  See dht_init_regex. */
+          .description = "Regular expression for stripping temporary-file "
+          "suffix and prefix used by an application, to prevent relocation when "
+          "the file is renamed."
+        },
+        { .key = {"rebalance-filter"},
+          .type = GF_OPTION_TYPE_STR,
+        },
+
+        { .key = {"xattr-name"},
+          .type = GF_OPTION_TYPE_STR,
+          .default_value = "trusted.glusterfs.dht",
+          .description = "Base for extended attributes used by this "
+          "translator instance, to avoid conflicts with others above or "
+          "below it."
+        },
+
+        /* NUFA option */
+        { .key  = {"local-volume-name"},
+          .type = GF_OPTION_TYPE_XLATOR
+        },
+
+        /* switch option */
+        { .key  = {"pattern.switch.case"},
+          .type = GF_OPTION_TYPE_ANY
+        },
+
+        { .key  = {NULL} },
+};
diff --git a/xlators/cluster/dht/src/dht.c b/xlators/cluster/dht/src/dht.c
index bd0258eae..fc0ca2f77 100644
--- a/xlators/cluster/dht/src/dht.c
+++ b/xlators/cluster/dht/src/dht.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -23,460 +14,65 @@
 #include "config.h"
 #endif
 
-/* TODO: add NS locking */
-
 #include "statedump.h"
-#include "dht-common.c"
-
-/* TODO:
-   - use volumename in xattr instead of "dht"
-   - use NS locks
-   - handle all cases in self heal layout reconstruction
-   - complete linkfile selfheal
-*/
-
-
-void
-dht_layout_dump (dht_layout_t  *layout, const char *prefix)
-{
-
-        char    key[GF_DUMP_MAX_BUF_LEN];
-        int     i = 0;
-
-        if (!layout)
-                return;
-
-        gf_proc_dump_build_key(key, prefix, "cnt");
-        gf_proc_dump_write(key, "%d", layout->cnt);
-        gf_proc_dump_build_key(key, prefix, "preset");
-        gf_proc_dump_write(key, "%d", layout->preset);
-        gf_proc_dump_build_key(key, prefix, "gen");
-        gf_proc_dump_write(key, "%d", layout->gen);
-        gf_proc_dump_build_key(key, prefix, "type");
-        gf_proc_dump_write(key, "%d", layout->type);
-
-        for (i = 0; i < layout->cnt; i++) {
-                gf_proc_dump_build_key(key, prefix,"list[%d].err", i);
-                gf_proc_dump_write(key, "%d", layout->list[i].err);
-                gf_proc_dump_build_key(key, prefix,"list[%d].start", i);
-                gf_proc_dump_write(key, "%u", layout->list[i].start);
-                gf_proc_dump_build_key(key, prefix,"list[%d].stop", i);
-                gf_proc_dump_write(key, "%u", layout->list[i].stop);
-                if (layout->list[i].xlator) {
-                        gf_proc_dump_build_key(key, prefix,
-                                               "list[%d].xlator.type", i);
-                        gf_proc_dump_write(key, "%s",
-                                           layout->list[i].xlator->type);
-                        gf_proc_dump_build_key(key, prefix,
-                                               "list[%d].xlator.name", i);
-                        gf_proc_dump_write(key, "%s",
-                                           layout->list[i].xlator->name);
-                }
-        }
-}
-
-
-int32_t
-dht_priv_dump (xlator_t *this)
-{
-	char            key_prefix[GF_DUMP_MAX_BUF_LEN];
-	char            key[GF_DUMP_MAX_BUF_LEN];
-        int             i = 0;
-        dht_conf_t      *conf = NULL;
-        int             ret = 0;
-
-        if (!this)
-                return -1;
-
-        conf = this->private;
-
-        if (!conf)
-                return -1;
-
-        ret = TRY_LOCK(&conf->subvolume_lock);
-
-        if (ret != 0) {
-                gf_log("", GF_LOG_WARNING, "Unable to lock dht subvolume %s",
-                                this->name);
-                return ret;
-        }
-
-        gf_proc_dump_add_section("xlator.cluster.dht.%s.priv", this->name);
-        gf_proc_dump_build_key(key_prefix,"xlator.cluster.dht","%s.priv",
-                               this->name);
-        gf_proc_dump_build_key(key, key_prefix, "subvolume_cnt");
-        gf_proc_dump_write(key,"%d", conf->subvolume_cnt);
-        for (i = 0; i < conf->subvolume_cnt; i++) {
-                gf_proc_dump_build_key(key, key_prefix, "subvolumes[%d]", i);
-                gf_proc_dump_write(key, "%s.%s", conf->subvolumes[i]->type,
-                                   conf->subvolumes[i]->name);
-                if (conf->file_layouts && conf->file_layouts[i]){
-                        gf_proc_dump_build_key(key, key_prefix,
-                                               "file_layouts[%d]",i);
-                        dht_layout_dump(conf->file_layouts[i], key);
-                }
-                if (conf->dir_layouts && conf->dir_layouts[i]) {
-                        gf_proc_dump_build_key(key, key_prefix,
-                                              "dir_layouts[%d]",i);
-                        dht_layout_dump(conf->dir_layouts[i], key);
-                }
-                if (conf->subvolume_status) {
-                        gf_proc_dump_build_key(key, key_prefix,
-                                                "subvolume_status[%d]", i);
-                        gf_proc_dump_write(key, "%d",
-                                               (int)conf->subvolume_status[i]);
-                }
-
-        }
-
-        gf_proc_dump_build_key(key, key_prefix,"default_dir_layout");
-        dht_layout_dump(conf->default_dir_layout, key);
-
-        gf_proc_dump_build_key(key, key_prefix, "search_unhashed");
-        gf_proc_dump_write(key, "%d", conf->search_unhashed);
-        gf_proc_dump_build_key(key, key_prefix, "gen");
-        gf_proc_dump_write(key, "%d", conf->gen);
-        gf_proc_dump_build_key(key, key_prefix, "min_free_disk");
-        gf_proc_dump_write(key, "%lu", conf->min_free_disk);
-        gf_proc_dump_build_key(key, key_prefix, "disk_unit");
-        gf_proc_dump_write(key, "%c", conf->disk_unit);
-        gf_proc_dump_build_key(key, key_prefix, "refresh_interval");
-        gf_proc_dump_write(key, "%d", conf->refresh_interval);
-        gf_proc_dump_build_key(key, key_prefix, "unhashed_sticky_bit");
-        gf_proc_dump_write(key, "%d", conf->unhashed_sticky_bit);
-        if (conf ->du_stats) {
-                gf_proc_dump_build_key(key, key_prefix,
-                                "du_stats.avail_percent");
-                gf_proc_dump_write(key, "%lf", conf->du_stats->avail_percent);
-                gf_proc_dump_build_key(key, key_prefix,
-                                "du_stats.avail_space");
-                gf_proc_dump_write(key, "%lu", conf->du_stats->avail_space);
-                gf_proc_dump_build_key(key, key_prefix,
-                                "du_stats.log");
-                gf_proc_dump_write(key, "%lu", conf->du_stats->log);
-        }
-        gf_proc_dump_build_key(key, key_prefix, "last_stat_fetch");
-        gf_proc_dump_write(key, "%s", ctime(&conf->last_stat_fetch.tv_sec));
-
-        UNLOCK(&conf->subvolume_lock);
-
-        return 0;
-}
-
-int32_t
-dht_inodectx_dump (xlator_t *this, inode_t *inode)
-{
-        int             ret = -1;
-        char            key_prefix[GF_DUMP_MAX_BUF_LEN];
-        dht_layout_t    *layout = NULL;
-	uint64_t        tmp_layout = 0;
-
-        if (!inode)
-                return -1;
-
-	ret = inode_ctx_get (inode, this, &tmp_layout);
-
-        if (ret != 0)
-                return ret;
-
-        layout = (dht_layout_t *)(long)tmp_layout;
-
-        if (!layout)
-                return -1;
-
-        gf_proc_dump_build_key(key_prefix, "xlator.cluster.dht",
-                               "%s.inode.%ld", this->name, inode->ino);
-        dht_layout_dump(layout, key_prefix);
-
-        return 0;
-}
-
-int
-notify (xlator_t *this, int event, void *data, ...)
-{
-	int ret = -1;
-
-	ret = dht_notify (this, event, data);
-
-	return ret;
-}
-
-void
-fini (xlator_t *this)
-{
-        int         i = 0;
-        dht_conf_t *conf = NULL;
-
-	conf = this->private;
-
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                GF_FREE (conf->file_layouts[i]);
-                        }
-                        GF_FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
-
-                GF_FREE (conf);
-        }
-
-	return;
-}
-
-int32_t
-mem_acct_init (xlator_t *this)
-{
-        int     ret = -1;
-
-        if (!this)
-                return ret;
-
-        ret = xlator_mem_acct_init (this, gf_dht_mt_end + 1);
-        
-        if (ret != 0) {
-                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
-                                "failed");
-                return ret;
-        }
-
-        return ret;
-}
-
-int
-reconfigure (xlator_t *this, dict_t *options)
-{
-	dht_conf_t	*conf = NULL;
-	char		*temp_str = NULL;
-	gf_boolean_t     search_unhashed;
-	uint32_t         temp_free_disk = 0;
-	int		 ret = 0;
-
-
-	conf = this->private;
-
-	if (dict_get_str (options, "lookup-unhashed", &temp_str) == 0) {
-                /* If option is not "auto", other options _should_ be boolean*/
-                if (strcasecmp (temp_str, "auto")) {
-                        if (!gf_string2boolean (temp_str, &search_unhashed)) {
-				gf_log(this->name, GF_LOG_DEBUG, "Reconfigure:"
-				       " lookup-unahashed reconfigured (%s)",
-				       temp_str);
-				conf->search_unhashed = search_unhashed;
-			}
-			else {
-				gf_log(this->name, GF_LOG_ERROR, "Reconfigure:"
-				       " lookup-unahashed should be boolean,"
-				        " not (%s), defaulting to (%d)",
-				       temp_str, conf->search_unhashed);
-				//return -1;
-				ret = -1;
-				goto out;
-			}
-		
-		}
-                else {
-			gf_log(this->name, GF_LOG_DEBUG, "Reconfigure:"
-			       " lookup-unahashed reconfigured auto ");
-                        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
-		}
-	}
-
-	if (dict_get_str (options, "min-free-disk", &temp_str) == 0) {
-		if (gf_string2percent (temp_str, &temp_free_disk) == 0) {
-                        if (temp_free_disk > 100) {
-                                gf_string2bytesize (temp_str, 
-                                                    &conf->min_free_disk);
-                                conf->disk_unit = 'b';
-                        } else {
-                                conf->min_free_disk = (uint64_t)temp_free_disk;
-                        }
-                } else {
-                        gf_string2bytesize (temp_str, &conf->min_free_disk);
-                        conf->disk_unit = 'b';
-                }
-
-		gf_log(this->name, GF_LOG_DEBUG, "Reconfigure:"
-			       " min-free-disk reconfigured to ",
-			       temp_str);
-	}
-
-out:
-	return ret;
-}
-
-int
-init (xlator_t *this)
-{
-        dht_conf_t    *conf = NULL;
-	char          *temp_str = NULL;
-        int            ret = -1;
-        int            i = 0;
-        uint32_t       temp_free_disk = 0;
-
-
-	if (!this->children) {
-		gf_log (this->name, GF_LOG_CRITICAL,
-			"Distribute needs more than one subvolume");
-		return -1;
-	}
-  
-	if (!this->parents) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"dangling volume. check volfile");
-	}
-
-        conf = GF_CALLOC (1, sizeof (*conf), gf_dht_mt_dht_conf_t);
-        if (!conf) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
-
-        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
-	if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
-                /* If option is not "auto", other options _should_ be boolean */
-                if (strcasecmp (temp_str, "auto"))
-                        gf_string2boolean (temp_str, &conf->search_unhashed);
-                else
-                        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
-	}
-
-	conf->unhashed_sticky_bit = 0;
-
-	if (dict_get_str (this->options, "unhashed-sticky-bit", 
-                          &temp_str) == 0) {
-	        gf_string2boolean (temp_str, &conf->unhashed_sticky_bit);
-	}
-
-	conf->use_readdirp = 0;
-
-	if (dict_get_str (this->options, "use-readdirp",
-                          &temp_str) == 0) {
-	        gf_string2boolean (temp_str, &conf->use_readdirp);
-	}
-
-        conf->disk_unit = 'p';
-        conf->min_free_disk = 10;
-
-	if (dict_get_str (this->options, "min-free-disk", &temp_str) == 0) {
-		if (gf_string2percent (temp_str, &temp_free_disk) == 0) {
-                        if (temp_free_disk > 100) {
-                                gf_string2bytesize (temp_str, 
-                                                    &conf->min_free_disk);
-                                conf->disk_unit = 'b';
-                        } else {
-                                conf->min_free_disk = (uint64_t)temp_free_disk;
-                        }
-                } else {
-                        gf_string2bytesize (temp_str, &conf->min_free_disk);
-                        conf->disk_unit = 'b';
-                }
-	}
-
-
-        ret = dht_init_subvolumes (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
-
-        ret = dht_layouts_init (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
-
-        conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
-                                    gf_dht_mt_dht_du_t);
-        if (!conf->du_stats) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
-
-	LOCK_INIT (&conf->subvolume_lock);
-	LOCK_INIT (&conf->layout_lock);
-
-	conf->gen = 1;
-
-        this->private = conf;
-
-        return 0;
-
-err:
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                GF_FREE (conf->file_layouts[i]);
-                        }
-                        GF_FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
-
-                if (conf->du_stats)
-                        GF_FREE (conf->du_stats);
-
-                GF_FREE (conf);
-        }
-
-        return -1;
-}
+#include "dht-common.h"
 
+class_methods_t class_methods = {
+        .init           = dht_init,
+        .fini           = dht_fini,
+        .reconfigure    = dht_reconfigure,
+        .notify         = dht_notify
+};
 
 struct xlator_fops fops = {
-	.lookup      = dht_lookup,
-	.mknod       = dht_mknod,
-	.create      = dht_create,
-
-	.stat        = dht_stat,
-	.fstat       = dht_fstat,
-	.truncate    = dht_truncate,
-	.ftruncate   = dht_ftruncate,
-	.access      = dht_access,
-	.readlink    = dht_readlink,
-	.setxattr    = dht_setxattr,
-	.getxattr    = dht_getxattr,
-	.removexattr = dht_removexattr,
-	.open        = dht_open,
-	.readv       = dht_readv,
-	.writev      = dht_writev,
-	.flush       = dht_flush,
-	.fsync       = dht_fsync,
-	.statfs      = dht_statfs,
-	.lk          = dht_lk,
-	.opendir     = dht_opendir,
-	.readdir     = dht_readdir,
-	.readdirp    = dht_readdirp,
-	.fsyncdir    = dht_fsyncdir,
-	.symlink     = dht_symlink,
-	.unlink      = dht_unlink,
-	.link        = dht_link,
-	.mkdir       = dht_mkdir,
-	.rmdir       = dht_rmdir,
-	.rename      = dht_rename,
-	.inodelk     = dht_inodelk,
-	.finodelk    = dht_finodelk,
-	.entrylk     = dht_entrylk,
-	.fentrylk    = dht_fentrylk,
-	.xattrop     = dht_xattrop,
-	.fxattrop    = dht_fxattrop,
-        .setattr     = dht_setattr, 
+        .lookup      = dht_lookup,
+        .mknod       = dht_mknod,
+        .create      = dht_create,
+
+        .open        = dht_open,
+        .statfs      = dht_statfs,
+        .opendir     = dht_opendir,
+        .readdir     = dht_readdir,
+        .readdirp    = dht_readdirp,
+        .fsyncdir    = dht_fsyncdir,
+        .symlink     = dht_symlink,
+        .unlink      = dht_unlink,
+        .link        = dht_link,
+        .mkdir       = dht_mkdir,
+        .rmdir       = dht_rmdir,
+        .rename      = dht_rename,
+        .entrylk     = dht_entrylk,
+        .fentrylk    = dht_fentrylk,
+
+        /* Inode read operations */
+        .stat        = dht_stat,
+        .fstat       = dht_fstat,
+        .access      = dht_access,
+        .readlink    = dht_readlink,
+        .getxattr    = dht_getxattr,
+        .fgetxattr    = dht_fgetxattr,
+        .readv       = dht_readv,
+        .flush       = dht_flush,
+        .fsync       = dht_fsync,
+        .inodelk     = dht_inodelk,
+        .finodelk    = dht_finodelk,
+        .lk          = dht_lk,
+
+        /* Inode write operations */
+        .fremovexattr = dht_fremovexattr,
+        .removexattr = dht_removexattr,
+        .setxattr    = dht_setxattr,
+        .fsetxattr   = dht_fsetxattr,
+        .truncate    = dht_truncate,
+        .ftruncate   = dht_ftruncate,
+        .writev      = dht_writev,
+        .xattrop     = dht_xattrop,
+        .fxattrop    = dht_fxattrop,
+        .setattr     = dht_setattr,
         .fsetattr    = dht_fsetattr,
+	.fallocate   = dht_fallocate,
+	.discard     = dht_discard,
+        .zerofill    = dht_zerofill,
 };
 
 struct xlator_dumpops dumpops = {
@@ -486,26 +82,8 @@ struct xlator_dumpops dumpops = {
 
 
 struct xlator_cbks cbks = {
-//	.release    = dht_release,
+//      .release    = dht_release,
 //      .releasedir = dht_releasedir,
-	.forget     = dht_forget
-};
-
-
-struct volume_options options[] = {
-        { .key  = {"lookup-unhashed"},
-          .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
-                    "on", "off"},
-	  .type = GF_OPTION_TYPE_STR
-	},
-        { .key  = {"min-free-disk"},
-          .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
-        },
-        { .key = {"unhashed-sticky-bit"},
-          .type = GF_OPTION_TYPE_BOOL
-        },
-        { .key = {"use-readdirp"},
-          .type = GF_OPTION_TYPE_BOOL
-        },
-	{ .key  = {NULL} },
+        .forget     = dht_forget
 };
+;
diff --git a/xlators/cluster/dht/src/nufa.c b/xlators/cluster/dht/src/nufa.c
index 190fbae69..e934acdf0 100644
--- a/xlators/cluster/dht/src/nufa.c
+++ b/xlators/cluster/dht/src/nufa.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -23,13 +14,15 @@
 #include "config.h"
 #endif
 
-#include "dht-common.c"
+#include "dht-common.h"
 
 /* TODO: all 'TODO's in dht.c holds good */
 
+extern struct volume_options options[];
+
 int
 nufa_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int op_ret, int op_errno,
+                       int op_ret, int op_errno,
                        inode_t *inode, struct iatt *stbuf, dict_t *xattr,
                        struct iatt *postparent)
 {
@@ -41,67 +34,61 @@ nufa_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         loc_t        *loc         = NULL;
         int           i           = 0;
         call_frame_t *prev        = NULL;
-	int           call_cnt    = 0;
+        int           call_cnt    = 0;
         int           ret         = 0;
 
-
         conf  = this->private;
 
         prev  = cookie;
         local = frame->local;
         loc   = &local->loc;
 
-	if (ENTRY_MISSING (op_ret, op_errno)) {
-		if (conf->search_unhashed) {
-			local->op_errno = ENOENT;
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
-		}
-	}
+        if (ENTRY_MISSING (op_ret, op_errno)) {
+                if (conf->search_unhashed) {
+                        local->op_errno = ENOENT;
+                        dht_lookup_everywhere (frame, this, loc);
+                        return 0;
+                }
+        }
 
         if (op_ret == -1)
                 goto out;
 
-        is_linkfile = check_is_linkfile (inode, stbuf, xattr);
+        is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                         conf->link_xattr_name);
         is_dir      = check_is_dir (inode, stbuf, xattr);
 
         if (!is_dir && !is_linkfile) {
                 /* non-directory and not a linkfile */
-
-		dht_itransform (this, prev->this, stbuf->ia_ino,
-				&stbuf->ia_ino);
-
-		ret = dht_layout_preset (this, prev->this, inode);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not set pre-set layout for subvol %s",
-				prev->this->name);
-			op_ret   = -1;
-			op_errno = EINVAL;
-			goto err;
-		}
+                ret = dht_layout_preset (this, prev->this, inode);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "could not set pre-set layout for subvol %s",
+                                prev->this->name);
+                        op_ret   = -1;
+                        op_errno = EINVAL;
+                        goto err;
+                }
 
                 goto out;
         }
 
         if (is_dir) {
                 call_cnt        = conf->subvolume_cnt;
-		local->call_cnt = call_cnt;
+                local->call_cnt = call_cnt;
 
                 local->inode = inode_ref (inode);
                 local->xattr = dict_ref (xattr);
 
-		local->op_ret = 0;
-		local->op_errno = 0;
+                local->op_ret = 0;
+                local->op_errno = 0;
 
-		local->layout = dht_layout_new (this, conf->subvolume_cnt);
-		if (!local->layout) {
-			op_ret   = -1;
-			op_errno = ENOMEM;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"memory allocation failed :(");
-			goto err;
-		}
+                local->layout = dht_layout_new (this, conf->subvolume_cnt);
+                if (!local->layout) {
+                        op_ret   = -1;
+                        op_errno = ENOMEM;
+                        goto err;
+                }
 
                 for (i = 0; i < call_cnt; i++) {
                         STACK_WIND (frame, dht_lookup_dir_cbk,
@@ -118,53 +105,52 @@ nufa_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "linkfile not having link subvolume. path=%s",
                                 loc->path);
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
+                        dht_lookup_everywhere (frame, this, loc);
+                        return 0;
                 }
 
-		STACK_WIND (frame, dht_lookup_linkfile_cbk,
-			    subvol, subvol->fops->lookup,
-			    &local->loc, local->xattr_req);
+                STACK_WIND (frame, dht_lookup_linkfile_cbk,
+                            subvol, subvol->fops->lookup,
+                            &local->loc, local->xattr_req);
         }
 
         return 0;
 
 out:
-	if (!local->hashed_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			local->loc.path);
+        if (!local->hashed_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        local->loc.path);
                 local->op_errno = ENOENT;
                 dht_lookup_everywhere (frame, this, loc);
                 return 0;
-	}
+        }
 
-	STACK_WIND (frame, dht_lookup_cbk,
-		    local->hashed_subvol, local->hashed_subvol->fops->lookup,
-		    &local->loc, local->xattr_req);
+        STACK_WIND (frame, dht_lookup_cbk,
+                    local->hashed_subvol, local->hashed_subvol->fops->lookup,
+                    &local->loc, local->xattr_req);
 
-	return 0;
+        return 0;
 
- err:
+err:
         DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno,
-                          inode, stbuf, xattr, NULL);
+                          inode, stbuf, xattr, postparent);
         return 0;
 }
 
 int
 nufa_lookup (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, dict_t *xattr_req)
+             loc_t *loc, dict_t *xattr_req)
 {
         xlator_t     *hashed_subvol = NULL;
-        xlator_t     *cached_subvol = NULL;
         xlator_t     *subvol = NULL;
         dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_conf_t   *conf = NULL;
         int           ret    = -1;
         int           op_errno = -1;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-	int           call_cnt = 0;
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           call_cnt = 0;
 
 
         VALIDATE_OR_GOTO (frame, err);
@@ -173,40 +159,26 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
         VALIDATE_OR_GOTO (loc->inode, err);
         VALIDATE_OR_GOTO (loc->path, err);
 
-	conf = this->private;
+        conf = this->private;
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-        ret = loc_dup (loc, &local->loc);
-        if (ret == -1) {
-                op_errno = errno;
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "copying location failed for path=%s",
-                        loc->path);
+        local = dht_local_init (frame, loc, NULL, GF_FOP_LOOKUP);
+        if (!local) {
+                op_errno = ENOMEM;
                 goto err;
         }
 
-	if (xattr_req) {
-		local->xattr_req = dict_ref (xattr_req);
-	} else {
-		local->xattr_req = dict_new ();
-	}
+        if (xattr_req) {
+                local->xattr_req = dict_ref (xattr_req);
+        } else {
+                local->xattr_req = dict_new ();
+        }
 
-	hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
-	cached_subvol = dht_subvol_get_cached (this, local->loc.inode);
+        hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
 
-	local->cached_subvol = cached_subvol;
-	local->hashed_subvol = hashed_subvol;
+        local->hashed_subvol = hashed_subvol;
 
         if (is_revalidate (loc)) {
-		local->layout = layout = dht_layout_get (this, loc->inode);
-
+                layout = local->layout;
                 if (!layout) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "revalidate without cache. path=%s",
@@ -215,25 +187,24 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		if (layout->gen && (layout->gen < conf->gen)) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"incomplete layout failure for path=%s",
-				loc->path);
+                if (layout->gen && (layout->gen < conf->gen)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "incomplete layout failure for path=%s",
+                                loc->path);
                         dht_layout_unref (this, local->layout);
-			goto do_fresh_lookup;
-		}
+                        goto do_fresh_lookup;
+                }
 
-		local->inode    = inode_ref (loc->inode);
-		local->ia_ino   = loc->inode->ino;
+                local->inode = inode_ref (loc->inode);
 
-		local->call_cnt = layout->cnt;
-		call_cnt = local->call_cnt;
+                local->call_cnt = layout->cnt;
+                call_cnt = local->call_cnt;
 
-		/* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
-		 *       revalidates directly go to the cached-subvolume.
-		 */
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht", 4 * 4);
+                /* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
+                 *       revalidates directly go to the cached-subvolume.
+                 */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
                 if (ret < 0) {
                         gf_log (this->name, GF_LOG_ERROR,
                                 "Failed to set dict value.");
@@ -241,20 +212,20 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		for (i = 0; i < layout->cnt; i++) {
-			subvol = layout->list[i].xlator;
+                for (i = 0; i < layout->cnt; i++) {
+                        subvol = layout->list[i].xlator;
 
-			STACK_WIND (frame, dht_revalidate_cbk,
-				    subvol, subvol->fops->lookup,
-				    loc, local->xattr_req);
+                        STACK_WIND (frame, dht_revalidate_cbk,
+                                    subvol, subvol->fops->lookup,
+                                    loc, local->xattr_req);
 
-			if (!--call_cnt)
-				break;
-		}
-	} else {
+                        if (!--call_cnt)
+                                break;
+                }
+        } else {
         do_fresh_lookup:
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht", 4 * 4);
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
                 if (ret < 0) {
                         gf_log (this->name, GF_LOG_ERROR,
                                 "Failed to set dict value.");
@@ -262,8 +233,8 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht.linkto", 256);
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->link_xattr_name, 256);
                 if (ret < 0) {
                         gf_log (this->name, GF_LOG_ERROR,
                                 "Failed to set dict value.");
@@ -271,111 +242,100 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		/* Send it to only local volume */
-		STACK_WIND (frame, nufa_local_lookup_cbk,
-			    (xlator_t *)conf->private,
-			    ((xlator_t *)conf->private)->fops->lookup,
-			    loc, local->xattr_req);
-	}
+                /* Send it to only local volume */
+                STACK_WIND (frame, nufa_local_lookup_cbk,
+                            (xlator_t *)conf->private,
+                            ((xlator_t *)conf->private)->fops->lookup,
+                            loc, local->xattr_req);
+        }
 
         return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
-	return 0;
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL);
+        return 0;
 }
 
 int
 nufa_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
-				 xlator_t *this, int op_ret, int op_errno,
+                                 xlator_t *this, int op_ret, int op_errno,
                                  inode_t *inode, struct iatt *stbuf,
                                  struct iatt *preparent,
-                                 struct iatt *postparent)
+                                 struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	if (op_ret == -1)
-		goto err;
+        if (op_ret == -1)
+                goto err;
 
-	STACK_WIND (frame, dht_create_cbk,
-		    local->cached_subvol, local->cached_subvol->fops->create,
-		    &local->loc, local->flags, local->mode, local->fd,
-                    local->params);
+        STACK_WIND (frame, dht_create_cbk,
+                    local->cached_subvol, local->cached_subvol->fops->create,
+                    &local->loc, local->flags, local->mode, local->umask,
+                    local->fd, local->params);
 
-	return 0;
+        return 0;
 
- err:
-	DHT_STACK_UNWIND (create, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL, NULL);
-	return 0;
+err:
+        DHT_STACK_UNWIND (create, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL, NULL);
+        return 0;
 }
 
 int
 nufa_create (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, int32_t flags, mode_t mode,
-             fd_t *fd, dict_t *params)
+             loc_t *loc, int32_t flags, mode_t mode,
+             mode_t umask, fd_t *fd, dict_t *params)
 {
-	dht_local_t *local = NULL;
-	dht_conf_t  *conf  = NULL;
-	xlator_t    *subvol = NULL;
+        dht_local_t *local = NULL;
+        dht_conf_t  *conf  = NULL;
+        xlator_t    *subvol = NULL;
         xlator_t    *avail_subvol = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
+        int          op_errno = -1;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
-	conf  = this->private;
+        conf  = this->private;
 
         dht_get_du_info (frame, this, loc);
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, fd, GF_FOP_CREATE);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
+                goto err;
+        }
 
         avail_subvol = conf->private;
         if (dht_is_subvol_filled (this, (xlator_t *)conf->private)) {
                 avail_subvol =
                         dht_free_disk_available_subvol (this,
-                                                        (xlator_t *)conf->private);
+                                                        (xlator_t *)conf->private,
+                                                        local);
         }
 
         if (subvol != avail_subvol) {
                 /* create a link file instead of actual file */
-                ret = loc_copy (&local->loc, loc);
-                if (ret == -1) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory");
-                                op_errno = ENOMEM;
-                                goto err;
-                }
-
-                local->fd = fd_ref (fd);
                 local->params = dict_ref (params);
                 local->mode = mode;
                 local->flags = flags;
-
+                local->umask = umask;
                 local->cached_subvol = avail_subvol;
-                dht_linkfile_create (frame,
-                                     nufa_create_linkfile_create_cbk,
-                                     avail_subvol, subvol, loc);
+                dht_linkfile_create (frame, nufa_create_linkfile_create_cbk,
+                                     this, avail_subvol, subvol, loc);
                 return 0;
         }
 
@@ -384,379 +344,333 @@ nufa_create (call_frame_t *frame, xlator_t *this,
 
         STACK_WIND (frame, dht_create_cbk,
                     subvol, subvol->fops->create,
-                    loc, flags, mode, fd, params);
+                    loc, flags, mode, umask, fd, params);
 
         return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (create, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (create, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 int
 nufa_mknod_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int op_ret, int op_errno, inode_t *inode,
                          struct iatt *stbuf, struct iatt *preparent,
-                         struct iatt *postparent)
+                         struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
-
-	if (op_ret >= 0) {
-		STACK_WIND (frame, dht_newfile_cbk,
-			    local->cached_subvol,
-			    local->cached_subvol->fops->mknod,
-			    &local->loc, local->mode, local->rdev,
-                            local->params);
+        local = frame->local;
+        if (!local || !local->cached_subvol) {
+                op_errno = EINVAL;
+                op_ret = -1;
+                goto err;
+        }
 
-		return 0;
-	}
+        if (op_ret >= 0) {
+                STACK_WIND_COOKIE (frame, dht_newfile_cbk,
+                            (void *)local->cached_subvol, local->cached_subvol,
+                            local->cached_subvol->fops->mknod,
+                            &local->loc, local->mode, local->rdev,
+                            local->umask, local->params);
 
+                return 0;
+        }
+err:
         WIPE (postparent);
         WIPE (preparent);
 
-	DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
-                          inode, stbuf, preparent, postparent);
-	return 0;
+        DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
+                          inode, stbuf, preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
 nufa_mknod (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, mode_t mode, dev_t rdev, dict_t *params)
+            loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *params)
 {
-	dht_local_t *local = NULL;
-	dht_conf_t  *conf  = NULL;
-	xlator_t    *subvol = NULL;
+        dht_local_t *local = NULL;
+        dht_conf_t  *conf  = NULL;
+        xlator_t    *subvol = NULL;
         xlator_t    *avail_subvol = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
+        int          op_errno = -1;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
-	conf  = this->private;
+        conf  = this->private;
 
         dht_get_du_info (frame, this, loc);
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_MKNOD);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
+                goto err;
+        }
 
         /* Consider the disksize in consideration */
         avail_subvol = conf->private;
         if (dht_is_subvol_filled (this, (xlator_t *)conf->private)) {
                 avail_subvol =
                         dht_free_disk_available_subvol (this,
-                                                        (xlator_t *)conf->private);
+                                                        (xlator_t *)conf->private,
+                                                        local);
         }
 
-	if (avail_subvol != subvol) {
-		/* Create linkfile first */
-		ret = loc_copy (&local->loc, loc);
-		if (ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"Out of memory");
-			op_errno = ENOMEM;
-			goto err;
-		}
+        if (avail_subvol != subvol) {
+                /* Create linkfile first */
 
                 local->params = dict_ref (params);
-		local->mode = mode;
-		local->rdev = rdev;
-		local->cached_subvol = avail_subvol;
+                local->mode = mode;
+                local->umask = umask;
+                local->rdev = rdev;
+                local->cached_subvol = avail_subvol;
 
-		dht_linkfile_create (frame, nufa_mknod_linkfile_cbk,
+                dht_linkfile_create (frame, nufa_mknod_linkfile_cbk, this,
                                      avail_subvol, subvol, loc);
-		return 0;
-	}
+                return 0;
+        }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"creating %s on %s", loc->path, subvol->name);
+        gf_log (this->name, GF_LOG_TRACE,
+                "creating %s on %s", loc->path, subvol->name);
 
-	STACK_WIND (frame, dht_newfile_cbk,
-		    subvol, subvol->fops->mknod,
-		    loc, mode, rdev, params);
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+                           subvol->fops->mknod, loc, mode, rdev, umask,
+                           params);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
-int
-notify (xlator_t *this, int event, void *data, ...)
+gf_boolean_t
+same_first_part (char *str1, char term1, char *str2, char term2)
 {
-	int ret = -1;
-
-	ret = dht_notify (this, event, data);
-
-	return ret;
+        gf_boolean_t    ended1;
+        gf_boolean_t    ended2;
+
+        for (;;) {
+                ended1 = ((*str1 == '\0') || (*str1 == term1));
+                ended2 = ((*str2 == '\0') || (*str2 == term2));
+                if (ended1 && ended2) {
+                        return _gf_true;
+                }
+                if (ended1 || ended2 || (*str1 != *str2)) {
+                        return _gf_false;
+                }
+                ++str1;
+                ++str2;
+        }
 }
 
-void
-fini (xlator_t *this)
+typedef struct nufa_args {
+        xlator_t    *this;
+        char        *volname;
+        gf_boolean_t addr_match;
+} nufa_args_t;
+
+static void
+nufa_find_local_brick (xlator_t *xl, void *data)
 {
-        int         i = 0;
-        dht_conf_t *conf = NULL;
+        nufa_args_t     *args = data;
+        xlator_t        *this = args->this;
+        char            *local_volname = args->volname;
+        gf_boolean_t    addr_match = args->addr_match;
+        char            *brick_host = NULL;
+        dht_conf_t      *conf = this->private;
+        int             ret = -1;
+
+        /*This means a local subvol was already found. We pick the first brick
+         * that is local*/
+        if (conf->private)
+                return;
+
+        if (strcmp (xl->name, local_volname) == 0) {
+                conf->private = xl;
+                gf_log (this->name, GF_LOG_INFO, "Using specified subvol %s",
+                        local_volname);
+                return;
+        }
 
-	conf = this->private;
+        if (!addr_match)
+                return;
+
+        ret = dict_get_str (xl->options, "remote-host", &brick_host);
+        if ((ret == 0) &&
+            (gf_is_same_address (local_volname, brick_host) ||
+             gf_is_local_addr (brick_host))) {
+                conf->private = xl;
+                gf_log (this->name, GF_LOG_INFO, "Using the first local "
+                        "subvol %s", xl->name);
+                return;
+        }
 
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                GF_FREE (conf->file_layouts[i]);
-                        }
-                        GF_FREE (conf->file_layouts);
-                }
+}
 
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
+static void
+nufa_to_dht (xlator_t *this)
+{
+        GF_ASSERT (this);
+        GF_ASSERT (this->fops);
 
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
+        this->fops->lookup = dht_lookup;
+        this->fops->create = dht_create;
+        this->fops->mknod  = dht_mknod;
+}
 
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
+int
+nufa_find_local_subvol (xlator_t *this,
+                        void (*fn) (xlator_t *each, void* data), void *data)
+{
+        int             ret = -1;
+        dht_conf_t      *conf = this->private;
+        xlator_list_t   *trav = NULL;
+        xlator_t        *parent = NULL;
+        xlator_t        *candidate = NULL;
+
+        xlator_foreach_depth_first (this, fn, data);
+        if (!conf->private) {
+                gf_log (this->name, GF_LOG_ERROR, "Couldn't find a local "
+                        "brick");
+                return -1;
+        }
+
+        candidate = conf->private;
+        trav = candidate->parents;
+        while (trav) {
+
+                parent = trav->xlator;
+                if (strcmp (parent->type, "cluster/nufa") == 0) {
+                        gf_log (this->name, GF_LOG_INFO, "Found local subvol, "
+                                "%s", candidate->name);
+                        ret = 0;
+                        conf->private = candidate;
+                        break;
+                }
 
-                GF_FREE (conf);
+                candidate = parent;
+                trav = parent->parents;
         }
 
-	return;
+        return ret;
 }
 
 int
-init (xlator_t *this)
+nufa_init (xlator_t *this)
 {
-        dht_conf_t    *conf = NULL;
-	xlator_list_t *trav = NULL;
-	data_t        *data = NULL;
-	char          *local_volname = NULL;
-	char          *temp_str = NULL;
+        data_t        *data = NULL;
+        char          *local_volname = NULL;
         int            ret = -1;
-        int            i = 0;
-	char           my_hostname[256];
-	uint32_t       temp_free_disk = 0;
-
-	if (!this->children) {
-		gf_log (this->name, GF_LOG_CRITICAL,
-			"NUFA needs more than one subvolume");
-		return -1;
-	}
-
-	if (!this->parents) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"dangling volume. check volfile");
-	}
-
-        conf = GF_CALLOC (1, sizeof (*conf),
-                          gf_dht_mt_dht_conf_t);
-        if (!conf) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
+        char           my_hostname[256];
+        gf_boolean_t   addr_match = _gf_false;
+        nufa_args_t    args = {0, };
+
+        ret = dht_init(this);
+        if (ret) {
+                return ret;
         }
 
-        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
-	if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
-                /* If option is not "auto", other options _should_ be boolean */
-                if (strcasecmp (temp_str, "auto"))
-                        gf_string2boolean (temp_str, &conf->search_unhashed);
-                else
-                        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
-	}
+        if ((data = dict_get (this->options, "local-volume-name"))) {
+                local_volname = data->data;
 
-        ret = dht_init_subvolumes (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
+        } else {
+                addr_match = _gf_true;
+                local_volname = "localhost";
+                ret = gethostname (my_hostname, 256);
+                if (ret == 0)
+                        local_volname = my_hostname;
 
-        ret = dht_layouts_init (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
+                else
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "could not find hostname (%s)",
+                                strerror (errno));
 
-	LOCK_INIT (&conf->subvolume_lock);
-	LOCK_INIT (&conf->layout_lock);
-
-	conf->gen = 1;
-
-	local_volname = "localhost";
-	ret = gethostname (my_hostname, 256);
-	if (ret < 0) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"could not find hostname (%s)",
-			strerror (errno));
-	}
-
-	if (ret == 0)
-		local_volname = my_hostname;
-
-	data = dict_get (this->options, "local-volume-name");
-	if (data) {
-		local_volname = data->data;
-	}
-
-	trav = this->children;
-	while (trav) {
-		if (strcmp (trav->xlator->name, local_volname) == 0)
-			break;
-		trav = trav->next;
-	}
-
-	if (!trav) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Could not find subvolume named '%s'. "
-			"Please define volume with the name as the hostname "
-			"or override it with 'option local-volume-name'",
-			local_volname);
-		goto err;
-	}
-	/* The volume specified exists */
-	conf->private = trav->xlator;
-
-        conf->min_free_disk = 10;
-        conf->disk_unit = 'p';
-
-        if (dict_get_str (this->options, "min-free-disk",
-                          &temp_str) == 0) {
-                if (gf_string2percent (temp_str,
-                                       &temp_free_disk) == 0) {
-                        if (temp_free_disk > 100) {
-                                gf_string2bytesize (temp_str,
-                                                        &conf->min_free_disk);
-                                conf->disk_unit = 'b';
-                        } else {
-                                conf->min_free_disk = (uint64_t)temp_free_disk;
-                                conf->disk_unit = 'p';
-                        }
-                } else {
-                        gf_string2bytesize (temp_str,
-                                                &conf->min_free_disk);
-                        conf->disk_unit = 'b';
-                }
         }
 
-        conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
-                                    gf_dht_mt_dht_du_t);
-        if (!conf->du_stats) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
+        args.this = this;
+        args.volname = local_volname;
+        args.addr_match = addr_match;
+        ret = nufa_find_local_subvol (this, nufa_find_local_brick, &args);
+        if (ret) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Unable to find local subvolume, switching "
+                        "to dht mode");
+                nufa_to_dht (this);
         }
-
-        this->private = conf;
-
         return 0;
+}
 
-err:
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                GF_FREE (conf->file_layouts[i]);
-                        }
-                        GF_FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
-
-                if (conf->du_stats)
-                        GF_FREE (conf->du_stats);
-
-                GF_FREE (conf);
-        }
 
-        return -1;
-}
+class_methods_t class_methods = {
+        .init           = nufa_init,
+        .fini           = dht_fini,
+        .reconfigure    = dht_reconfigure,
+        .notify         = dht_notify
+};
 
 
 struct xlator_fops fops = {
-	.lookup      = nufa_lookup,
-	.create      = nufa_create,
-	.mknod       = nufa_mknod,
-
-	.stat        = dht_stat,
-	.fstat       = dht_fstat,
-	.truncate    = dht_truncate,
-	.ftruncate   = dht_ftruncate,
-	.access      = dht_access,
-	.readlink    = dht_readlink,
-	.setxattr    = dht_setxattr,
-	.getxattr    = dht_getxattr,
-	.removexattr = dht_removexattr,
-	.open        = dht_open,
-	.readv       = dht_readv,
-	.writev      = dht_writev,
-	.flush       = dht_flush,
-	.fsync       = dht_fsync,
-	.statfs      = dht_statfs,
-	.lk          = dht_lk,
-	.opendir     = dht_opendir,
-	.readdir     = dht_readdir,
-	.readdirp    = dht_readdirp,
-	.fsyncdir    = dht_fsyncdir,
-	.symlink     = dht_symlink,
-	.unlink      = dht_unlink,
-	.link        = dht_link,
-	.mkdir       = dht_mkdir,
-	.rmdir       = dht_rmdir,
-	.rename      = dht_rename,
-	.inodelk     = dht_inodelk,
-	.finodelk    = dht_finodelk,
-	.entrylk     = dht_entrylk,
-	.fentrylk    = dht_fentrylk,
-	.xattrop     = dht_xattrop,
-	.fxattrop    = dht_fxattrop,
+        .lookup      = nufa_lookup,
+        .create      = nufa_create,
+        .mknod       = nufa_mknod,
+
+        .stat        = dht_stat,
+        .fstat       = dht_fstat,
+        .truncate    = dht_truncate,
+        .ftruncate   = dht_ftruncate,
+        .access      = dht_access,
+        .readlink    = dht_readlink,
+        .setxattr    = dht_setxattr,
+        .getxattr    = dht_getxattr,
+        .removexattr = dht_removexattr,
+        .open        = dht_open,
+        .readv       = dht_readv,
+        .writev      = dht_writev,
+        .flush       = dht_flush,
+        .fsync       = dht_fsync,
+        .statfs      = dht_statfs,
+        .lk          = dht_lk,
+        .opendir     = dht_opendir,
+        .readdir     = dht_readdir,
+        .readdirp    = dht_readdirp,
+        .fsyncdir    = dht_fsyncdir,
+        .symlink     = dht_symlink,
+        .unlink      = dht_unlink,
+        .link        = dht_link,
+        .mkdir       = dht_mkdir,
+        .rmdir       = dht_rmdir,
+        .rename      = dht_rename,
+        .inodelk     = dht_inodelk,
+        .finodelk    = dht_finodelk,
+        .entrylk     = dht_entrylk,
+        .fentrylk    = dht_fentrylk,
+        .xattrop     = dht_xattrop,
+        .fxattrop    = dht_fxattrop,
         .setattr     = dht_setattr,
 };
 
 
 struct xlator_cbks cbks = {
-	.forget     = dht_forget
-};
-
-
-struct volume_options options[] = {
-        { .key  = {"lookup-unhashed"},
-          .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
-                    "on", "off"},
-	  .type = GF_OPTION_TYPE_STR
-	},
-	{ .key  = {"local-volume-name"},
-	  .type = GF_OPTION_TYPE_XLATOR
-	},
-        { .key  = {"min-free-disk"},
-          .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
-        },
-	{ .key  = {NULL} },
+        .forget     = dht_forget
 };
diff --git a/xlators/cluster/dht/src/switch.c b/xlators/cluster/dht/src/switch.c
index f3de211fd..d3ea90ba8 100644
--- a/xlators/cluster/dht/src/switch.c
+++ b/xlators/cluster/dht/src/switch.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -23,7 +14,7 @@
 #include "config.h"
 #endif
 
-#include "dht-common.c"
+#include "dht-common.h"
 #include "dht-mem-types.h"
 
 #include <sys/time.h>
@@ -31,21 +22,23 @@
 #include <fnmatch.h>
 #include <string.h>
 
+extern struct volume_options options[];
+
 struct switch_sched_array {
-	xlator_t *xl;
-	int32_t   eligible;
-	int32_t   considered;
+        xlator_t *xl;
+        int32_t   eligible;
+        int32_t   considered;
 };
 
 /* Select one of this struct based on the path's pattern match */
 struct switch_struct {
-	struct switch_struct      *next;
+        struct switch_struct      *next;
         struct switch_sched_array *array;
-	int32_t                    node_index; /* Index of the node in
+        int32_t                    node_index; /* Index of the node in
                                                   this pattern. */
-	int32_t                    num_child;  /* Total num of child nodes
+        int32_t                    num_child;  /* Total num of child nodes
                                                   with this pattern. */
-	char                       path_pattern[256];
+        char                       path_pattern[256];
 };
 
 /* TODO: all 'TODO's in dht.c holds good */
@@ -74,31 +67,39 @@ get_switch_matching_subvol (const char *path, dht_conf_t *conf,
                             xlator_t *hashed_subvol)
 {
         struct switch_struct *cond      = NULL;
-	struct switch_struct *trav      = NULL;
-	char                 *pathname  = NULL;
-	int                   idx     = 0;
+        struct switch_struct *trav      = NULL;
+        char                 *pathname  = NULL;
+        int                   idx       = 0;
+        xlator_t             *subvol    = NULL;
 
         cond = conf->private;
+        subvol = hashed_subvol;
         if (!cond)
-                return hashed_subvol;
+                goto out;
 
-        trav = cond;
         pathname = gf_strdup (path);
-	while (trav) {
-		if (fnmatch (trav->path_pattern,
-			     pathname, FNM_NOESCAPE) == 0) {
+        if (!pathname)
+                goto out;
+
+        trav = cond;
+        while (trav) {
+                if (fnmatch (trav->path_pattern,
+                             pathname, FNM_NOESCAPE) == 0) {
                         for (idx = 0; idx < trav->num_child; idx++) {
                                 if (trav->array[idx].xl == hashed_subvol)
-                                        return hashed_subvol;
+                                        goto out;
                         }
                         idx = trav->node_index++;
                         trav->node_index %= trav->num_child;
-                        return trav->array[idx].xl;
-		}
-		trav = trav->next;
-	}
-	GF_FREE (pathname);
-	return hashed_subvol;
+                        subvol = trav->array[idx].xl;
+                        goto out;
+                }
+                trav = trav->next;
+        }
+out:
+        GF_FREE (pathname);
+
+        return subvol;
 }
 
 
@@ -116,7 +117,7 @@ switch_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         loc_t        *loc         = NULL;
         int           i           = 0;
         call_frame_t *prev        = NULL;
-	int           call_cnt    = 0;
+        int           call_cnt    = 0;
         int           ret         = 0;
 
         conf  = this->private;
@@ -125,57 +126,55 @@ switch_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         local = frame->local;
         loc   = &local->loc;
 
-	if (ENTRY_MISSING (op_ret, op_errno)) {
-		if (conf->search_unhashed) {
-			local->op_errno = ENOENT;
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
-		}
-	}
+        if (ENTRY_MISSING (op_ret, op_errno)) {
+                if (conf->search_unhashed) {
+                        local->op_errno = ENOENT;
+                        dht_lookup_everywhere (frame, this, loc);
+                        return 0;
+                }
+        }
 
         if (op_ret == -1)
                 goto out;
 
-        is_linkfile = check_is_linkfile (inode, stbuf, xattr);
+        is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                         conf->link_xattr_name);
         is_dir      = check_is_dir (inode, stbuf, xattr);
 
         if (!is_dir && !is_linkfile) {
                 /* non-directory and not a linkfile */
 
-		dht_itransform (this, prev->this, stbuf->ia_ino,
-				&stbuf->ia_ino);
-
-		ret = dht_layout_preset (this, prev->this, inode);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not set pre-set layout for subvol %s",
-				prev->this->name);
-			op_ret   = -1;
-			op_errno = EINVAL;
-			goto err;
-		}
+                ret = dht_layout_preset (this, prev->this, inode);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "could not set pre-set layout for subvol %s",
+                                prev->this->name);
+                        op_ret   = -1;
+                        op_errno = EINVAL;
+                        goto err;
+                }
 
                 goto out;
         }
 
         if (is_dir) {
                 call_cnt        = conf->subvolume_cnt;
-		local->call_cnt = call_cnt;
+                local->call_cnt = call_cnt;
 
                 local->inode = inode_ref (inode);
                 local->xattr = dict_ref (xattr);
 
-		local->op_ret = 0;
-		local->op_errno = 0;
+                local->op_ret = 0;
+                local->op_errno = 0;
 
-		local->layout = dht_layout_new (this, conf->subvolume_cnt);
-		if (!local->layout) {
-			op_ret   = -1;
-			op_errno = ENOMEM;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"memory allocation failed :(");
-			goto err;
-		}
+                local->layout = dht_layout_new (this, conf->subvolume_cnt);
+                if (!local->layout) {
+                        op_ret   = -1;
+                        op_errno = ENOMEM;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "memory allocation failed :(");
+                        goto err;
+                }
 
                 for (i = 0; i < call_cnt; i++) {
                         STACK_WIND (frame, dht_lookup_dir_cbk,
@@ -192,34 +191,34 @@ switch_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "linkfile not having link subvolume. path=%s",
                                 loc->path);
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
+                        dht_lookup_everywhere (frame, this, loc);
+                        return 0;
                 }
 
-		STACK_WIND (frame, dht_lookup_linkfile_cbk,
-			    subvol, subvol->fops->lookup,
-			    &local->loc, local->xattr_req);
+                STACK_WIND (frame, dht_lookup_linkfile_cbk,
+                            subvol, subvol->fops->lookup,
+                            &local->loc, local->xattr_req);
         }
 
         return 0;
 
 out:
-	if (!local->hashed_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			local->loc.path);
+        if (!local->hashed_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        local->loc.path);
                 local->op_errno = ENOENT;
                 dht_lookup_everywhere (frame, this, loc);
                 return 0;
-	}
+        }
 
-	STACK_WIND (frame, dht_lookup_cbk,
-		    local->hashed_subvol, local->hashed_subvol->fops->lookup,
-		    &local->loc, local->xattr_req);
+        STACK_WIND (frame, dht_lookup_cbk,
+                    local->hashed_subvol, local->hashed_subvol->fops->lookup,
+                    &local->loc, local->xattr_req);
 
-	return 0;
+        return 0;
 
- err:
+err:
         DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno,
                           inode, stbuf, xattr, NULL);
         return 0;
@@ -227,18 +226,18 @@ out:
 
 int
 switch_lookup (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, dict_t *xattr_req)
+               loc_t *loc, dict_t *xattr_req)
 {
         xlator_t     *hashed_subvol = NULL;
         xlator_t     *cached_subvol = NULL;
         xlator_t     *subvol = NULL;
         dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_conf_t   *conf = NULL;
         int           ret    = -1;
         int           op_errno = -1;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-	int           call_cnt = 0;
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           call_cnt = 0;
 
 
         VALIDATE_OR_GOTO (frame, err);
@@ -247,40 +246,27 @@ switch_lookup (call_frame_t *frame, xlator_t *this,
         VALIDATE_OR_GOTO (loc->inode, err);
         VALIDATE_OR_GOTO (loc->path, err);
 
-	conf = this->private;
+        conf = this->private;
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-        ret = loc_dup (loc, &local->loc);
-        if (ret == -1) {
-                op_errno = errno;
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "copying location failed for path=%s",
-                        loc->path);
+        local = dht_local_init (frame, loc, NULL, GF_FOP_LOOKUP);
+        if (!local) {
+                op_errno = ENOMEM;
                 goto err;
         }
 
-	if (xattr_req) {
-		local->xattr_req = dict_ref (xattr_req);
-	} else {
-		local->xattr_req = dict_new ();
-	}
+        if (xattr_req) {
+                local->xattr_req = dict_ref (xattr_req);
+        } else {
+                local->xattr_req = dict_new ();
+        }
 
-	hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
-	cached_subvol = dht_subvol_get_cached (this, local->loc.inode);
+        hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
+        cached_subvol = local->cached_subvol;
 
-	local->cached_subvol = cached_subvol;
-	local->hashed_subvol = hashed_subvol;
+        local->hashed_subvol = hashed_subvol;
 
         if (is_revalidate (loc)) {
-		local->layout = layout = dht_layout_get (this, loc->inode);
-
+                layout = local->layout;
                 if (!layout) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "revalidate without cache. path=%s",
@@ -289,83 +275,80 @@ switch_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		if (layout->gen && (layout->gen < conf->gen)) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"incomplete layout failure for path=%s",
-				loc->path);
+                if (layout->gen && (layout->gen < conf->gen)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "incomplete layout failure for path=%s",
+                                loc->path);
                         dht_layout_unref (this, local->layout);
-			goto do_fresh_lookup;
-		}
+                        goto do_fresh_lookup;
+                }
 
-		local->inode    = inode_ref (loc->inode);
-		local->ia_ino   = loc->inode->ino;
+                local->inode    = inode_ref (loc->inode);
 
-		local->call_cnt = layout->cnt;
-		call_cnt = local->call_cnt;
+                local->call_cnt = layout->cnt;
+                call_cnt = local->call_cnt;
 
-		/* NOTE: we don't require 'trusted.glusterfs.dht.linkto'
+                /* NOTE: we don't require 'trusted.glusterfs.dht.linkto'
                  * attribute, revalidates directly go to the cached-subvolume.
-		 */
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht", 4 * 4);
+                 */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
                 if (ret < 0)
                         gf_log (this->name, GF_LOG_WARNING,
-                                "failed to set dict value for "
-                                "trusted.glusterfs.dht");
+                                "failed to set dict value for %s",
+                                conf->xattr_name);
 
-		for (i = 0; i < layout->cnt; i++) {
-			subvol = layout->list[i].xlator;
+                for (i = 0; i < layout->cnt; i++) {
+                        subvol = layout->list[i].xlator;
 
-			STACK_WIND (frame, dht_revalidate_cbk,
-				    subvol, subvol->fops->lookup,
-				    loc, local->xattr_req);
+                        STACK_WIND (frame, dht_revalidate_cbk,
+                                    subvol, subvol->fops->lookup,
+                                    loc, local->xattr_req);
 
-			if (!--call_cnt)
-				break;
-		}
-	} else {
+                        if (!--call_cnt)
+                                break;
+                }
+        } else {
         do_fresh_lookup:
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht", 4 * 4);
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
                 if (ret < 0)
                         gf_log (this->name, GF_LOG_WARNING,
-                                "failed to set dict value for "
-                                "trusted.glusterfs.dht");
-                
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht.linkto", 256);
+                                "failed to set dict value for %s",
+                                conf->xattr_name);
+
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->link_xattr_name, 256);
                 if (ret < 0)
-                        gf_log (this->name, GF_LOG_WARNING, 
-                                "failed to set dict value for " 
-                                "trusted.glusterfs.dht.linkto");
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "failed to set dict value for %s",
+                                conf->link_xattr_name);
 
                 if (!hashed_subvol) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"no subvolume in layout for path=%s, "
-				"checking on all the subvols to see if "
-				"it is a directory", loc->path);
- 			call_cnt        = conf->subvolume_cnt;
- 			local->call_cnt = call_cnt;
-
- 			local->layout = dht_layout_new (this,
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "no subvolume in layout for path=%s, "
+                                "checking on all the subvols to see if "
+                                "it is a directory", loc->path);
+                        call_cnt        = conf->subvolume_cnt;
+                        local->call_cnt = call_cnt;
+
+                        local->layout = dht_layout_new (this,
                                                         conf->subvolume_cnt);
- 			if (!local->layout) {
- 				op_errno = ENOMEM;
- 				gf_log (this->name, GF_LOG_ERROR,
- 					"Out of memory");
- 				goto err;
- 			}
-
-			for (i = 0; i < call_cnt; i++) {
- 				STACK_WIND (frame, dht_lookup_dir_cbk,
- 					    conf->subvolumes[i],
- 					    conf->subvolumes[i]->fops->lookup,
- 					    &local->loc, local->xattr_req);
- 			}
- 			return 0;
+                        if (!local->layout) {
+                                op_errno = ENOMEM;
+                                goto err;
+                        }
+
+                        for (i = 0; i < call_cnt; i++) {
+                                STACK_WIND (frame, dht_lookup_dir_cbk,
+                                            conf->subvolumes[i],
+                                            conf->subvolumes[i]->fops->lookup,
+                                            &local->loc, local->xattr_req);
+                        }
+                        return 0;
                 }
 
-		/*  */
+                /*  */
                 cached_subvol = get_switch_matching_subvol (loc->path, conf,
                                                             hashed_subvol);
                 if (cached_subvol == hashed_subvol) {
@@ -384,99 +367,88 @@ switch_lookup (call_frame_t *frame, xlator_t *this,
         return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
-	return 0;
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (lookup, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL);
+        return 0;
 }
 
 int
 switch_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
-				 xlator_t *this, int op_ret, int op_errno,
-                                 inode_t *inode, struct iatt *stbuf,
-                                 struct iatt *preparent,
-                                 struct iatt *postparent)
+                                   xlator_t *this, int op_ret, int op_errno,
+                                   inode_t *inode, struct iatt *stbuf,
+                                   struct iatt *preparent,
+                                   struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	if (op_ret == -1)
-		goto err;
+        if (op_ret == -1)
+                goto err;
 
-	STACK_WIND (frame, dht_create_cbk,
-		    local->cached_subvol, local->cached_subvol->fops->create,
-		    &local->loc, local->flags, local->mode, local->fd,
-                    local->params);
+        STACK_WIND (frame, dht_create_cbk,
+                    local->cached_subvol, local->cached_subvol->fops->create,
+                    &local->loc, local->flags, local->mode, local->umask,
+                    local->fd, local->params);
 
-	return 0;
+        return 0;
 
- err:
-	DHT_STACK_UNWIND (create, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL, NULL);
-	return 0;
+err:
+        DHT_STACK_UNWIND (create, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL, NULL);
+        return 0;
 }
 
 int
 switch_create (call_frame_t *frame, xlator_t *this,
                loc_t *loc, int32_t flags, mode_t mode,
-               fd_t *fd, dict_t *params)
+               mode_t umask, fd_t *fd, dict_t *params)
 {
-	dht_local_t *local = NULL;
-	dht_conf_t  *conf  = NULL;
-	xlator_t    *subvol = NULL;
+        dht_local_t *local = NULL;
+        dht_conf_t  *conf  = NULL;
+        xlator_t    *subvol = NULL;
         xlator_t    *avail_subvol = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
+        int          op_errno = -1;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
-	conf  = this->private;
+        conf  = this->private;
 
         dht_get_du_info (frame, this, loc);
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, fd, GF_FOP_CREATE);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
+                goto err;
+        }
 
         avail_subvol = get_switch_matching_subvol (loc->path, conf, subvol);
         if (dht_is_subvol_filled (this, avail_subvol)) {
                 avail_subvol =
-                        dht_free_disk_available_subvol (this, avail_subvol);
+                        dht_free_disk_available_subvol (this, avail_subvol,
+                                                        local);
         }
 
         if (subvol != avail_subvol) {
                 /* create a link file instead of actual file */
-                ret = loc_copy (&local->loc, loc);
-                if (ret == -1) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory");
-                                op_errno = ENOMEM;
-                                goto err;
-                }
-
-                local->fd = fd_ref (fd);
                 local->mode = mode;
                 local->flags = flags;
-
+                local->umask = umask;
                 local->cached_subvol = avail_subvol;
-                dht_linkfile_create (frame,
-                                     switch_create_linkfile_create_cbk,
-                                     avail_subvol, subvol, loc);
+                dht_linkfile_create (frame, switch_create_linkfile_create_cbk,
+                                     this, avail_subvol, subvol, loc);
                 return 0;
         }
 
@@ -485,176 +457,143 @@ switch_create (call_frame_t *frame, xlator_t *this,
 
         STACK_WIND (frame, dht_create_cbk,
                     subvol, subvol->fops->create,
-                    loc, flags, mode, fd, params);
+                    loc, flags, mode, umask, fd, params);
 
         return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (create, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (create, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 int
 switch_mknod_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                         int op_ret, int op_errno, inode_t *inode,
-                         struct iatt *stbuf, struct iatt *preparent,
-                         struct iatt *postparent)
+                           int op_ret, int op_errno, inode_t *inode,
+                           struct iatt *stbuf, struct iatt *preparent,
+                           struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-
-	local = frame->local;
+        dht_local_t  *local = NULL;
 
-	if (op_ret >= 0) {
-		STACK_WIND (frame, dht_newfile_cbk,
-			    local->cached_subvol,
-			    local->cached_subvol->fops->mknod,
-			    &local->loc, local->mode, local->rdev,
-                            local->params);
+        local = frame->local;
+        if (!local || !local->cached_subvol) {
+                op_errno = EINVAL;
+                op_ret = -1;
+                goto err;
+        }
 
-		return 0;
-	}
+        if (op_ret >= 0) {
+                STACK_WIND_COOKIE (frame, dht_newfile_cbk,
+                            (void *)local->cached_subvol, local->cached_subvol,
+                            local->cached_subvol->fops->mknod,
+                            &local->loc, local->mode, local->rdev,
+                            local->umask, local->params);
 
-	DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
-                          inode, stbuf, preparent, postparent);
-	return 0;
+                return 0;
+        }
+err:
+        DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
+                          inode, stbuf, preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
-switch_mknod (call_frame_t *frame, xlator_t *this,
-              loc_t *loc, mode_t mode, dev_t rdev, dict_t *params)
+switch_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+              dev_t rdev, mode_t umask, dict_t *params)
 {
-	dht_local_t *local = NULL;
-	dht_conf_t  *conf  = NULL;
-	xlator_t    *subvol = NULL;
+        dht_local_t *local = NULL;
+        dht_conf_t  *conf  = NULL;
+        xlator_t    *subvol = NULL;
         xlator_t    *avail_subvol = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
+        int          op_errno = -1;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
-	conf  = this->private;
+        conf  = this->private;
 
         dht_get_du_info (frame, this, loc);
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_MKNOD);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
+                goto err;
+        }
 
         /* Consider the disksize in consideration */
         avail_subvol = get_switch_matching_subvol (loc->path, conf, subvol);
         if (dht_is_subvol_filled (this, avail_subvol)) {
                 avail_subvol =
-                        dht_free_disk_available_subvol (this, avail_subvol);
+                        dht_free_disk_available_subvol (this, avail_subvol,
+                                                        local);
         }
 
-	if (avail_subvol != subvol) {
-		/* Create linkfile first */
-		ret = loc_copy (&local->loc, loc);
-		if (ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"Out of memory");
-			op_errno = ENOMEM;
-			goto err;
-		}
+        if (avail_subvol != subvol) {
+                /* Create linkfile first */
 
                 local->params = dict_ref (params);
-		local->mode = mode;
-		local->rdev = rdev;
-		local->cached_subvol = avail_subvol;
+                local->mode = mode;
+                local->umask = umask;
+                local->rdev = rdev;
+                local->cached_subvol = avail_subvol;
 
-		dht_linkfile_create (frame, switch_mknod_linkfile_cbk,
-                                     avail_subvol, subvol, loc);
-		return 0;
-	}
+                dht_linkfile_create (frame, switch_mknod_linkfile_cbk,
+                                     this, avail_subvol, subvol, loc);
+                return 0;
+        }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"creating %s on %s", loc->path, subvol->name);
+        gf_log (this->name, GF_LOG_TRACE,
+                "creating %s on %s", loc->path, subvol->name);
 
-	STACK_WIND (frame, dht_newfile_cbk,
-		    subvol, subvol->fops->mknod,
-		    loc, mode, rdev, params);
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+                           subvol->fops->mknod, loc, mode, rdev, umask,
+                           params);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
-int
-notify (xlator_t *this, int event, void *data, ...)
-{
-	int ret = -1;
-
-	ret = dht_notify (this, event, data);
-
-	return ret;
-}
-
 void
-fini (xlator_t *this)
+switch_fini (xlator_t *this)
 {
-        int                   i = 0;
         dht_conf_t           *conf = NULL;
         struct switch_struct *trav = NULL;
         struct switch_struct *prev = NULL;
 
-	conf = this->private;
+        conf = this->private;
 
         if (conf) {
                 trav = (struct switch_struct *)conf->private;
                 conf->private = NULL;
                 while (trav) {
-                        if (trav->array)
-                                GF_FREE (trav->array);
+                        GF_FREE (trav->array);
                         prev = trav;
                         trav = trav->next;
                         GF_FREE (prev);
                 }
-
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                GF_FREE (conf->file_layouts[i]);
-                        }
-                        GF_FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
-
-                GF_FREE (conf);
         }
 
-	return;
+        dht_fini(this);
 }
 
 int
@@ -679,32 +618,32 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
         struct switch_struct        *switch_buf = NULL;
         struct switch_struct        *switch_opt = NULL;
         struct switch_struct        *trav = NULL;
-	struct switch_sched_array  *switch_buf_array = NULL;
-	xlator_list_t              *trav_xl = NULL;
+        struct switch_sched_array  *switch_buf_array = NULL;
+        xlator_list_t              *trav_xl = NULL;
 
         trav_xl = this->children;
-	while (trav_xl) {
-		index++;
-		trav_xl = trav_xl->next;
-	}
-	child_count = index;
-	switch_buf_array = GF_CALLOC ((index + 1),
+        while (trav_xl) {
+                index++;
+                trav_xl = trav_xl->next;
+        }
+        child_count = index;
+        switch_buf_array = GF_CALLOC ((index + 1),
                                       sizeof (struct switch_sched_array),
                                       gf_switch_mt_switch_sched_array);
         if (!switch_buf_array)
                 goto err;
 
-	trav_xl = this->children;
-	index = 0;
+        trav_xl = this->children;
+        index = 0;
 
-	while (trav_xl) {
-		switch_buf_array[index].xl = trav_xl->xlator;
-		switch_buf_array[index].eligible = 1;
-		trav_xl = trav_xl->next;
-		index++;
-	}
+        while (trav_xl) {
+                switch_buf_array[index].xl = trav_xl->xlator;
+                switch_buf_array[index].eligible = 1;
+                trav_xl = trav_xl->next;
+                index++;
+        }
 
-	/*  *jpg:child1,child2;*mpg:child3;*:child4,child5,child6 */
+        /*  *jpg:child1,child2;*mpg:child3;*:child4,child5,child6 */
 
         /* Get the pattern for considering switch case.
            "option block-size *avi:10MB" etc */
@@ -714,17 +653,20 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
                 dup_str = gf_strdup (switch_str);
                 switch_opt = GF_CALLOC (1, sizeof (struct switch_struct),
                                         gf_switch_mt_switch_struct);
-                if (!switch_opt)
+                if (!switch_opt) {
+                        GF_FREE (dup_str);
                         goto err;
+                }
 
                 pattern = strtok_r (dup_str, ":", &tmp_str1);
                 childs = strtok_r (NULL, ":", &tmp_str1);
                 if (strncmp (pattern, "*", 2) == 0) {
-                        gf_log ("switch", GF_LOG_NORMAL,
+                        gf_log ("switch", GF_LOG_INFO,
                                 "'*' pattern will be taken by default "
                                 "for all the unconfigured child nodes,"
                                 " hence neglecting current option");
                         switch_str = strtok_r (NULL, ";", &tmp_str);
+                        GF_FREE (switch_opt);
                         GF_FREE (dup_str);
                         continue;
                 }
@@ -749,7 +691,7 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
                         child = strtok_r (childs, ",", &tmp1);
                         switch_opt->num_child = idx;
                         switch_opt->array = GF_CALLOC (1, (idx *
-                                                       sizeof (struct switch_sched_array)),
+                                                           sizeof (struct switch_sched_array)),
                                                        gf_switch_mt_switch_sched_array);
                         if (!switch_opt->array)
                                 goto err;
@@ -785,7 +727,7 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
                         goto err;
                 }
                 GF_FREE (dup_str);
-                
+
                 /* Link it to the main structure */
                 if (switch_buf) {
                         /* there are already few entries */
@@ -797,75 +739,77 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
                         /* First entry */
                         switch_buf = switch_opt;
                 }
+                switch_opt = NULL;
                 switch_str = strtok_r (NULL, ";", &tmp_str);
         }
 
-	/* Now, all the pattern based considerations done, so for all the
-	 * remaining pattern, '*' to all the remaining child nodes
-	 */
-	{
-		for (index=0; index < child_count; index++) {
-			/* check for considered flag */
-			if (switch_buf_array[index].considered)
-				continue;
-			flag++;
-		}
-		if (!flag) {
-			gf_log ("switch", GF_LOG_ERROR,
-				"No nodes left for pattern '*'. Exiting");
-			goto err;
-		}
-		switch_opt = GF_CALLOC (1, sizeof (struct switch_struct),
+        /* Now, all the pattern based considerations done, so for all the
+         * remaining pattern, '*' to all the remaining child nodes
+         */
+        {
+                for (index=0; index < child_count; index++) {
+                        /* check for considered flag */
+                        if (switch_buf_array[index].considered)
+                                continue;
+                        flag++;
+                }
+                if (!flag) {
+                        gf_log ("switch", GF_LOG_ERROR,
+                                "No nodes left for pattern '*'. Exiting");
+                        goto err;
+                }
+                switch_opt = GF_CALLOC (1, sizeof (struct switch_struct),
                                         gf_switch_mt_switch_struct);
                 if (!switch_opt)
                         goto err;
 
-		/* Add the '*' pattern to the array */
-		memcpy (switch_opt->path_pattern, "*", 2);
-		switch_opt->num_child = flag;
-		switch_opt->array =
-			GF_CALLOC (1, 
+                /* Add the '*' pattern to the array */
+                memcpy (switch_opt->path_pattern, "*", 2);
+                switch_opt->num_child = flag;
+                switch_opt->array =
+                        GF_CALLOC (1,
                                    flag * sizeof (struct switch_sched_array),
                                    gf_switch_mt_switch_sched_array);
                 if (!switch_opt->array)
                         goto err;
-		flag = 0;
-		for (index=0; index < child_count; index++) {
-			/* check for considered flag */
-			if (switch_buf_array[index].considered)
-				continue;
-			gf_log ("switch", GF_LOG_DEBUG,
-				"'%s' pattern will be scheduled to \"%s\"",
-				switch_opt->path_pattern,
-				switch_buf_array[index].xl->name);
-			switch_opt->array[flag].xl =
-				switch_buf_array[index].xl;
-			switch_buf_array[index].considered = 1;
-			flag++;
+                flag = 0;
+                for (index=0; index < child_count; index++) {
+                        /* check for considered flag */
+                        if (switch_buf_array[index].considered)
+                                continue;
+                        gf_log ("switch", GF_LOG_DEBUG,
+                                "'%s' pattern will be scheduled to \"%s\"",
+                                switch_opt->path_pattern,
+                                switch_buf_array[index].xl->name);
+                        switch_opt->array[flag].xl =
+                                switch_buf_array[index].xl;
+                        switch_buf_array[index].considered = 1;
+                        flag++;
+                }
+                if (switch_buf) {
+                        /* there are already few entries */
+                        trav = switch_buf;
+                        while (trav->next)
+                                trav = trav->next;
+                        trav->next = switch_opt;
+                } else {
+                        /* First entry */
+                        switch_buf = switch_opt;
                 }
-		if (switch_buf) {
-			/* there are already few entries */
-			trav = switch_buf;
-			while (trav->next)
-				trav = trav->next;
-			trav->next = switch_opt;
-		} else {
-			/* First entry */
-			switch_buf = switch_opt;
-		}
-	}
+                switch_opt = NULL;
+        }
         /* */
         conf->private = switch_buf;
 
         return 0;
 err:
+        GF_FREE (switch_buf_array);
+        GF_FREE (switch_opt);
+
         if (switch_buf) {
-                if (switch_buf_array)
-                        GF_FREE (switch_buf_array);
                 trav = switch_buf;
                 while (trav) {
-                        if (trav->array)
-                                GF_FREE (trav->array);
+                        GF_FREE (trav->array);
                         switch_opt = trav;
                         trav = trav->next;
                         GF_FREE (switch_opt);
@@ -875,192 +819,86 @@ err:
 }
 
 
-int
-init (xlator_t *this)
+int32_t
+switch_init (xlator_t *this)
 {
         dht_conf_t            *conf = NULL;
-	data_t                *data = NULL;
-	char                  *temp_str = NULL;
+        data_t                *data = NULL;
         int                    ret = -1;
-        int                    i = 0;
-	uint32_t               temp_free_disk = 0;
-
-	if (!this->children) {
-		gf_log (this->name, GF_LOG_CRITICAL,
-			"SWITCH needs more than one subvolume");
-		return -1;
-	}
-
-	if (!this->parents) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"dangling volume. check volfile");
-	}
-
-        conf = GF_CALLOC (1, sizeof (*conf), gf_switch_mt_dht_conf_t);
-        if (!conf) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
 
-        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
-	if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
-                /* If option is not "auto", other options _should_ be boolean */
-                if (strcasecmp (temp_str, "auto"))
-                        gf_string2boolean (temp_str, &conf->search_unhashed);
-                else
-                        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
-	}
-
-	conf->unhashed_sticky_bit = 0;
-	if (dict_get_str (this->options, "unhashed-sticky-bit",
-                          &temp_str) == 0) {
-	        gf_string2boolean (temp_str, &conf->unhashed_sticky_bit);
-	}
-
-        conf->min_free_disk = 10;
-        conf->disk_unit = 'p';
-
-        if (dict_get_str (this->options, "min-free-disk",
-                          &temp_str) == 0) {
-                if (gf_string2percent (temp_str,
-                                       &temp_free_disk) == 0) {
-                        if (temp_free_disk > 100) {
-                                gf_string2bytesize (temp_str,
-                                                        &conf->min_free_disk);
-                                conf->disk_unit = 'b';
-                        } else {
-                                conf->min_free_disk = (uint64_t)temp_free_disk;
-                                conf->disk_unit = 'p';
-                        }
-                } else {
-                        gf_string2bytesize (temp_str,
-                                                &conf->min_free_disk);
-                        conf->disk_unit = 'b';
-                }
+        ret = dht_init(this);
+        if (ret) {
+                return ret;
         }
+        conf = this->private;
 
-	data = dict_get (this->options, "pattern.switch.case");
-	if (data) {
+        data = dict_get (this->options, "pattern.switch.case");
+        if (data) {
                 /* TODO: */
                 ret = set_switch_pattern (this, conf, data->data);
                 if (ret) {
                         goto err;
                 }
-	}
-
-        ret = dht_init_subvolumes (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
-
-        ret = dht_layouts_init (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
-
-	LOCK_INIT (&conf->subvolume_lock);
-	LOCK_INIT (&conf->layout_lock);
-
-	conf->gen = 1;
-
-        conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
-                                    gf_switch_mt_dht_du_t);
-        if (!conf->du_stats) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
         }
 
         this->private = conf;
-
         return 0;
 
 err:
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                GF_FREE (conf->file_layouts[i]);
-                        }
-                        GF_FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
-
-                if (conf->du_stats)
-                        GF_FREE (conf->du_stats);
-
-                GF_FREE (conf);
-        }
-
+        dht_fini(this);
         return -1;
 }
 
 
-struct xlator_fops fops = {
-	.lookup      = switch_lookup,
-	.create      = switch_create,
-	.mknod       = switch_mknod,
-
-	.stat        = dht_stat,
-	.fstat       = dht_fstat,
-	.truncate    = dht_truncate,
-	.ftruncate   = dht_ftruncate,
-	.access      = dht_access,
-	.readlink    = dht_readlink,
-	.setxattr    = dht_setxattr,
-	.getxattr    = dht_getxattr,
-	.removexattr = dht_removexattr,
-	.open        = dht_open,
-	.readv       = dht_readv,
-	.writev      = dht_writev,
-	.flush       = dht_flush,
-	.fsync       = dht_fsync,
-	.statfs      = dht_statfs,
-	.lk          = dht_lk,
-	.opendir     = dht_opendir,
-	.readdir     = dht_readdir,
-	.readdirp    = dht_readdirp,
-	.fsyncdir    = dht_fsyncdir,
-	.symlink     = dht_symlink,
-	.unlink      = dht_unlink,
-	.link        = dht_link,
-	.mkdir       = dht_mkdir,
-	.rmdir       = dht_rmdir,
-	.rename      = dht_rename,
-	.inodelk     = dht_inodelk,
-	.finodelk    = dht_finodelk,
-	.entrylk     = dht_entrylk,
-	.fentrylk    = dht_fentrylk,
-	.xattrop     = dht_xattrop,
-	.fxattrop    = dht_fxattrop,
-        .setattr     = dht_setattr,
+class_methods_t class_methods = {
+        .init           = switch_init,
+        .fini           = switch_fini,
+        .reconfigure    = dht_reconfigure,
+        .notify         = dht_notify
 };
 
 
-struct xlator_cbks cbks = {
-	.forget     = dht_forget
+struct xlator_fops fops = {
+        .lookup      = switch_lookup,
+        .create      = switch_create,
+        .mknod       = switch_mknod,
+
+        .stat        = dht_stat,
+        .fstat       = dht_fstat,
+        .truncate    = dht_truncate,
+        .ftruncate   = dht_ftruncate,
+        .access      = dht_access,
+        .readlink    = dht_readlink,
+        .setxattr    = dht_setxattr,
+        .getxattr    = dht_getxattr,
+        .removexattr = dht_removexattr,
+        .open        = dht_open,
+        .readv       = dht_readv,
+        .writev      = dht_writev,
+        .flush       = dht_flush,
+        .fsync       = dht_fsync,
+        .statfs      = dht_statfs,
+        .lk          = dht_lk,
+        .opendir     = dht_opendir,
+        .readdir     = dht_readdir,
+        .readdirp    = dht_readdirp,
+        .fsyncdir    = dht_fsyncdir,
+        .symlink     = dht_symlink,
+        .unlink      = dht_unlink,
+        .link        = dht_link,
+        .mkdir       = dht_mkdir,
+        .rmdir       = dht_rmdir,
+        .rename      = dht_rename,
+        .inodelk     = dht_inodelk,
+        .finodelk    = dht_finodelk,
+        .entrylk     = dht_entrylk,
+        .fentrylk    = dht_fentrylk,
+        .xattrop     = dht_xattrop,
+        .fxattrop    = dht_fxattrop,
+        .setattr     = dht_setattr,
 };
 
 
-struct volume_options options[] = {
-        { .key  = {"lookup-unhashed"},
-          .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
-                    "on", "off"},
-	  .type = GF_OPTION_TYPE_STR
-	},
-	{ .key  = {"pattern.switch.case"},
-	  .type = GF_OPTION_TYPE_ANY
-	},
-        { .key  = {"min-free-disk"},
-          .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
-        },
-	{ .key  = {NULL} },
+struct xlator_cbks cbks = {
+        .forget     = dht_forget
 };
diff --git a/xlators/cluster/ha/src/Makefile.am b/xlators/cluster/ha/src/Makefile.am
index 5f78a2965..5c1364b7f 100644
--- a/xlators/cluster/ha/src/Makefile.am
+++ b/xlators/cluster/ha/src/Makefile.am
@@ -1,15 +1,16 @@
 xlator_LTLIBRARIES = ha.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/testing/cluster
 
-ha_la_LDFLAGS = -module -avoidversion 
+ha_la_LDFLAGS = -module -avoid-version 
 
 ha_la_SOURCES = ha-helpers.c ha.c
 ha_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
 noinst_HEADERS = ha.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	    -I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/cluster/ha/src/ha-helpers.c b/xlators/cluster/ha/src/ha-helpers.c
index fb6593101..19be1ed27 100644
--- a/xlators/cluster/ha/src/ha-helpers.c
+++ b/xlators/cluster/ha/src/ha-helpers.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #include "xlator.h"
 #include "call-stub.h"
 #include "defaults.h"
diff --git a/xlators/cluster/ha/src/ha-mem-types.h b/xlators/cluster/ha/src/ha-mem-types.h
index bdbfcb52b..e5e97d237 100644
--- a/xlators/cluster/ha/src/ha-mem-types.h
+++ b/xlators/cluster/ha/src/ha-mem-types.h
@@ -1,24 +1,13 @@
-
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
 
-
 #ifndef __HA_MEM_TYPES_H__
 #define __HA_MEM_TYPES_H__
 
diff --git a/xlators/cluster/ha/src/ha.c b/xlators/cluster/ha/src/ha.c
index 54b41858e..3eccb516b 100644
--- a/xlators/cluster/ha/src/ha.c
+++ b/xlators/cluster/ha/src/ha.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 /* generate errors randomly, code is simple now, better alogorithm
  * can be written to decide what error to be returned and when
  */
@@ -1876,13 +1866,9 @@ err:
         }
 
         if (hafdp) {
-                if (hafdp->fdstate) {
-                        GF_FREE (hafdp->fdstate);
-                }
+                GF_FREE (hafdp->fdstate);
 
-                if (hafdp->path) {
-                        GF_FREE (hafdp->path);
-                }
+                GF_FREE (hafdp->path);
 
                 GF_FREE (hafdp);
         }
@@ -3066,7 +3052,7 @@ ha_lk_setlk_unlck_cbk (call_frame_t *frame,
 		       xlator_t *this,
 		       int32_t op_ret,
 		       int32_t op_errno,
-		       struct flock *lock)
+		       struct gf_flock *lock)
 {
 	ha_local_t *local = NULL;
 	int cnt = 0;
@@ -3099,7 +3085,7 @@ ha_lk_setlk_cbk (call_frame_t *frame,
 		 xlator_t *this,
 		 int32_t op_ret,
 		 int32_t op_errno,
-		 struct flock *lock)
+		 struct gf_flock *lock)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -3155,7 +3141,7 @@ ha_lk_setlk_cbk (call_frame_t *frame,
 				cnt++;
 		}
 		if (cnt) {
-			struct flock lock;
+			struct gf_flock lock;
 			lock = local->stub->args.lk.lock;
 			for (i = 0; i < child_count; i++) {
 				if (state[i]) {
@@ -3189,7 +3175,7 @@ ha_lk_getlk_cbk (call_frame_t *frame,
 		 xlator_t *this,
 		 int32_t op_ret,
 		 int32_t op_errno,
-		 struct flock *lock)
+		 struct gf_flock *lock)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -3244,7 +3230,7 @@ ha_lk (call_frame_t *frame,
        xlator_t *this,
        fd_t *fd,
        int32_t cmd,
-       struct flock *lock)
+       struct gf_flock *lock)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -3378,7 +3364,7 @@ ha_inodelk (call_frame_t *frame,
             const char *volume,
 	    loc_t *loc,
 	    int32_t cmd,
-	    struct flock *lock)
+	    struct gf_flock *lock)
 {
 	ha_local_t *local = NULL;
 	int op_errno = 0;
diff --git a/xlators/cluster/ha/src/ha.h b/xlators/cluster/ha/src/ha.h
index 5e06b7e02..e2ed7eaa6 100644
--- a/xlators/cluster/ha/src/ha.h
+++ b/xlators/cluster/ha/src/ha.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef __HA_H_
 #define __HA_H_
 
diff --git a/xlators/cluster/map/src/Makefile.am b/xlators/cluster/map/src/Makefile.am
index 26e19137a..a278b05e2 100644
--- a/xlators/cluster/map/src/Makefile.am
+++ b/xlators/cluster/map/src/Makefile.am
@@ -1,15 +1,16 @@
 xlator_LTLIBRARIES = map.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/testing/cluster
 
-map_la_LDFLAGS = -module -avoidversion 
+map_la_LDFLAGS = -module -avoid-version 
 
 map_la_SOURCES = map.c map-helper.c
 map_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
 noinst_HEADERS = map.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	    -I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/cluster/map/src/map-helper.c b/xlators/cluster/map/src/map-helper.c
index ad01b2102..851397b68 100644
--- a/xlators/cluster/map/src/map-helper.c
+++ b/xlators/cluster/map/src/map-helper.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2009-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2009-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
diff --git a/xlators/cluster/map/src/map-mem-types.h b/xlators/cluster/map/src/map-mem-types.h
index f41ab420a..3e89f4736 100644
--- a/xlators/cluster/map/src/map-mem-types.h
+++ b/xlators/cluster/map/src/map-mem-types.h
@@ -1,24 +1,13 @@
-
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
 
-
 #ifndef __MAP_MEM_TYPES_H__
 #define __MAP_MEM_TYPES_H__
 
diff --git a/xlators/cluster/map/src/map.c b/xlators/cluster/map/src/map.c
index dd89d0ebe..6150a33ce 100644
--- a/xlators/cluster/map/src/map.c
+++ b/xlators/cluster/map/src/map.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -438,7 +428,7 @@ map_lk_cbk (call_frame_t *frame,
 		xlator_t *this,
 		int32_t op_ret,
 		int32_t op_errno,
-		struct flock *lock)
+		struct gf_flock *lock)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, lock);
 	return 0;
@@ -1770,7 +1760,7 @@ map_lk (call_frame_t *frame,
 	xlator_t *this,
 	fd_t *fd,
 	int32_t cmd,
-	struct flock *lock)
+	struct gf_flock *lock)
 {
 	int32_t op_errno = 1;
 	xlator_t *subvol   = NULL;
@@ -1799,7 +1789,7 @@ map_lk (call_frame_t *frame,
 
 int32_t
 map_inodelk (call_frame_t *frame, xlator_t *this,
-	     const char *volume, loc_t *loc, int32_t cmd, struct flock *lock)
+	     const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *lock)
 {
 	int32_t op_errno = 1;
 	xlator_t *subvol   = NULL;
@@ -1829,7 +1819,7 @@ map_inodelk (call_frame_t *frame, xlator_t *this,
 
 int32_t
 map_finodelk (call_frame_t *frame, xlator_t *this,
-	      const char *volume, fd_t *fd, int32_t cmd, struct flock *lock)
+	      const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *lock)
 {
 	int32_t op_errno = 1;
 	xlator_t *subvol   = NULL;
@@ -2375,8 +2365,7 @@ fini (xlator_t *this)
 	priv = this->private;
 
 	if (priv) {
-		if (priv->xlarray)
-			GF_FREE (priv->xlarray);
+		GF_FREE (priv->xlarray);
 
 		trav_map = priv->map;
 		while (trav_map) {
diff --git a/xlators/cluster/map/src/map.h b/xlators/cluster/map/src/map.h
index eb549eb06..7703a543e 100644
--- a/xlators/cluster/map/src/map.h
+++ b/xlators/cluster/map/src/map.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __MAP_H__
 #define __MAP_H__
 
diff --git a/xlators/cluster/stripe/src/Makefile.am b/xlators/cluster/stripe/src/Makefile.am
index 6d4fae4d2..2d151422a 100644
--- a/xlators/cluster/stripe/src/Makefile.am
+++ b/xlators/cluster/stripe/src/Makefile.am
@@ -2,15 +2,19 @@
 xlator_LTLIBRARIES = stripe.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
 
-stripe_la_LDFLAGS = -module -avoidversion
+stripe_la_LDFLAGS = -module -avoid-version
+
+stripe_la_SOURCES = stripe.c stripe-helpers.c \
+	$(top_builddir)/xlators/lib/src/libxlator.c
 
-stripe_la_SOURCES = stripe.c
 stripe_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = stripe.h stripe-mem-types.h
+noinst_HEADERS = stripe.h stripe-mem-types.h $(top_builddir)/xlators/lib/src/libxlator.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+	-I$(top_srcdir)/xlators/lib/src
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/cluster/stripe/src/stripe-helpers.c b/xlators/cluster/stripe/src/stripe-helpers.c
new file mode 100644
index 000000000..a83abdc72
--- /dev/null
+++ b/xlators/cluster/stripe/src/stripe-helpers.c
@@ -0,0 +1,675 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#include <fnmatch.h>
+
+#include "stripe.h"
+#include "byte-order.h"
+#include "mem-types.h"
+
+void
+stripe_local_wipe (stripe_local_t *local)
+{
+        if (!local)
+                goto out;
+
+        loc_wipe (&local->loc);
+        loc_wipe (&local->loc2);
+
+        if (local->fd)
+                fd_unref (local->fd);
+
+        if (local->inode)
+                inode_unref (local->inode);
+
+        if (local->xattr)
+                dict_unref (local->xattr);
+
+        if (local->xdata)
+                dict_unref (local->xdata);
+
+out:
+        return;
+}
+
+
+
+int
+stripe_aggregate (dict_t *this, char *key, data_t *value, void *data)
+{
+        dict_t  *dst  = NULL;
+        int64_t *ptr  = 0, *size = NULL;
+        int32_t  ret  = -1;
+
+        dst = data;
+
+        if (strcmp (key, GF_XATTR_QUOTA_SIZE_KEY) == 0) {
+                ret = dict_get_bin (dst, key, (void **)&size);
+                if (ret < 0) {
+                        size = GF_CALLOC (1, sizeof (int64_t),
+                                          gf_common_mt_char);
+                        if (size == NULL) {
+                                gf_log ("stripe", GF_LOG_WARNING,
+                                        "memory allocation failed");
+                                goto out;
+                        }
+                        ret = dict_set_bin (dst, key, size, sizeof (int64_t));
+                        if (ret < 0) {
+                                gf_log ("stripe", GF_LOG_WARNING,
+                                        "stripe aggregate dict set failed");
+                                GF_FREE (size);
+                                goto out;
+                        }
+                }
+
+                ptr = data_to_bin (value);
+                if (ptr == NULL) {
+                        gf_log ("stripe", GF_LOG_WARNING, "data to bin failed");
+                        goto out;
+                }
+
+                *size = hton64 (ntoh64 (*size) + ntoh64 (*ptr));
+        } else if (strcmp (key, GF_CONTENT_KEY)) {
+                /* No need to aggregate 'CONTENT' data */
+                ret = dict_set (dst, key, value);
+                if (ret)
+                        gf_log ("stripe", GF_LOG_WARNING, "xattr dict set failed");
+        }
+
+out:
+        return 0;
+}
+
+
+void
+stripe_aggregate_xattr (dict_t *dst, dict_t *src)
+{
+        if ((dst == NULL) || (src == NULL)) {
+                goto out;
+        }
+
+        dict_foreach (src, stripe_aggregate, dst);
+out:
+        return;
+}
+
+
+int32_t
+stripe_xattr_aggregate (char *buffer, stripe_local_t *local, int32_t *total)
+{
+        int32_t              i     = 0;
+        int32_t              ret   = -1;
+        int32_t              len   = 0;
+        char                *sbuf  = NULL;
+        stripe_xattr_sort_t *xattr = NULL;
+
+        if (!buffer || !local || !local->xattr_list)
+                goto out;
+
+        sbuf = buffer;
+
+        for (i = 0; i < local->nallocs; i++) {
+                xattr = local->xattr_list + i;
+                len = xattr->xattr_len;
+
+                if (len && xattr && xattr->xattr_value) {
+                        memcpy (buffer, xattr->xattr_value, len);
+                        buffer += len;
+                        *buffer++ = ' ';
+                }
+        }
+
+        *--buffer = '\0';
+        if (total)
+                *total = buffer - sbuf;
+        ret = 0;
+
+ out:
+        return ret;
+}
+
+int32_t
+stripe_free_xattr_str (stripe_local_t *local)
+{
+        int32_t              i     = 0;
+        int32_t              ret   = -1;
+        stripe_xattr_sort_t *xattr = NULL;
+
+        if (!local || !local->xattr_list)
+                goto out;
+
+        for (i = 0; i < local->nallocs; i++) {
+                xattr = local->xattr_list + i;
+
+                if (xattr && xattr->xattr_value)
+                        GF_FREE (xattr->xattr_value);
+        }
+
+        ret = 0;
+ out:
+        return ret;
+}
+
+
+int32_t
+stripe_fill_lockinfo_xattr (xlator_t *this, stripe_local_t *local,
+                            void **xattr_serz)
+{
+        int32_t              ret   = -1, i = 0, len = 0;
+        dict_t              *tmp1  = NULL, *tmp2 = NULL;
+        char                *buf   = NULL;
+        stripe_xattr_sort_t *xattr = NULL;
+
+        if (xattr_serz == NULL) {
+                goto out;
+        }
+
+        tmp2 = dict_new ();
+
+        if (tmp2 == NULL) {
+                goto out;
+        }
+
+        for (i = 0; i < local->nallocs; i++) {
+                xattr = local->xattr_list + i;
+                len = xattr->xattr_len;
+
+                if (len && xattr && xattr->xattr_value) {
+                        ret = dict_reset (tmp2);
+                        if (ret < 0) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "dict_reset failed (%s)",
+                                        strerror (-ret));
+                        }
+
+                        ret = dict_unserialize (xattr->xattr_value,
+                                                xattr->xattr_len,
+                                                &tmp2);
+                        if (ret < 0) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "dict_unserialize failed (%s)",
+                                        strerror (-ret));
+                                ret = -1;
+                                goto out;
+                        }
+
+                        tmp1 = dict_copy (tmp2, tmp1);
+                        if (tmp1 == NULL) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "dict_copy failed (%s)",
+                                        strerror (-ret));
+                                ret = -1;
+                                goto out;
+                        }
+                }
+        }
+
+        len = dict_serialized_length (tmp1);
+        if (len > 0) {
+                buf = GF_CALLOC (1, len, gf_common_mt_dict_t);
+                if (buf == NULL) {
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = dict_serialize (tmp1, buf);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "dict_serialize failed (%s)", strerror (-ret));
+                        ret = -1;
+                        goto out;
+                }
+
+                *xattr_serz = buf;
+        }
+
+        ret = 0;
+out:
+        if (tmp1 != NULL) {
+                dict_unref (tmp1);
+        }
+
+        if (tmp2 != NULL) {
+                dict_unref (tmp2);
+        }
+
+        return ret;
+}
+
+
+int32_t
+stripe_fill_pathinfo_xattr (xlator_t *this, stripe_local_t *local,
+                            char **xattr_serz)
+{
+        int      ret             = -1;
+        int32_t  padding         = 0;
+        int32_t  tlen            = 0;
+        char stripe_size_str[20] = {0,};
+        char    *pathinfo_serz   = NULL;
+
+        if (!local) {
+                gf_log (this->name, GF_LOG_ERROR, "Possible NULL deref");
+                goto out;
+        }
+
+        (void) snprintf (stripe_size_str, 20, "%ld",
+                         (local->fctx) ? local->fctx->stripe_size : 0);
+
+        /* extra bytes for decorations (brackets and <>'s) */
+        padding = strlen (this->name) + strlen (STRIPE_PATHINFO_HEADER)
+                + strlen (stripe_size_str) + 7;
+        local->xattr_total_len += (padding + 2);
+
+        pathinfo_serz = GF_CALLOC (local->xattr_total_len, sizeof (char),
+                                   gf_common_mt_char);
+        if (!pathinfo_serz)
+                goto out;
+
+        /* xlator info */
+        (void) sprintf (pathinfo_serz, "(<"STRIPE_PATHINFO_HEADER"%s:[%s]> ",
+                        this->name, stripe_size_str);
+
+        ret = stripe_xattr_aggregate (pathinfo_serz + padding, local, &tlen);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Cannot aggregate pathinfo list");
+                goto out;
+        }
+
+        *(pathinfo_serz + padding + tlen) = ')';
+        *(pathinfo_serz + padding + tlen + 1) = '\0';
+
+        *xattr_serz = pathinfo_serz;
+
+        ret = 0;
+ out:
+        return ret;
+}
+
+/**
+ * stripe_get_matching_bs - Get the matching block size for the given path.
+ */
+int32_t
+stripe_get_matching_bs (const char *path, stripe_private_t *priv)
+{
+        struct stripe_options *trav       = NULL;
+        uint64_t               block_size = 0;
+
+        GF_VALIDATE_OR_GOTO ("stripe", priv, out);
+        GF_VALIDATE_OR_GOTO ("stripe", path, out);
+
+        LOCK (&priv->lock);
+        {
+                block_size = priv->block_size;
+                trav = priv->pattern;
+                while (trav) {
+                        if (!fnmatch (trav->path_pattern, path, FNM_NOESCAPE)) {
+                                block_size = trav->block_size;
+                                break;
+                        }
+                        trav = trav->next;
+                }
+        }
+        UNLOCK (&priv->lock);
+
+out:
+        return block_size;
+}
+
+int32_t
+stripe_ctx_handle (xlator_t *this, call_frame_t *prev, stripe_local_t *local,
+                   dict_t *dict)
+{
+        char            key[256]       = {0,};
+        data_t         *data            = NULL;
+        int32_t         index           = 0;
+        stripe_private_t *priv          = NULL;
+
+        priv = this->private;
+
+
+        if (!local->fctx) {
+                local->fctx =  GF_CALLOC (1, sizeof (stripe_fd_ctx_t),
+                                         gf_stripe_mt_stripe_fd_ctx_t);
+                if (!local->fctx) {
+                        local->op_errno = ENOMEM;
+                        local->op_ret = -1;
+                        goto out;
+                }
+
+                local->fctx->static_array = 0;
+        }
+        /* Stripe block size */
+        sprintf (key, "trusted.%s.stripe-size", this->name);
+        data = dict_get (dict, key);
+        if (!data) {
+                local->xattr_self_heal_needed = 1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get stripe-size");
+                goto out;
+        } else {
+                if (!local->fctx->stripe_size) {
+                        local->fctx->stripe_size =
+                                     data_to_int64 (data);
+                }
+
+                if (local->fctx->stripe_size != data_to_int64 (data)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "stripe-size mismatch in blocks");
+                        local->xattr_self_heal_needed = 1;
+                }
+        }
+
+        /* Stripe count */
+        sprintf (key, "trusted.%s.stripe-count", this->name);
+        data = dict_get (dict, key);
+
+        if (!data) {
+                local->xattr_self_heal_needed = 1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get stripe-count");
+                goto out;
+        }
+        if (!local->fctx->xl_array) {
+                local->fctx->stripe_count = data_to_int32 (data);
+                if (!local->fctx->stripe_count) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "error with stripe-count xattr");
+                        local->op_ret   = -1;
+                        local->op_errno = EIO;
+                        goto out;
+                }
+
+                local->fctx->xl_array = GF_CALLOC (local->fctx->stripe_count,
+                                                   sizeof (xlator_t *),
+                                                   gf_stripe_mt_xlator_t);
+
+                if (!local->fctx->xl_array) {
+                        local->op_errno = ENOMEM;
+                        local->op_ret   = -1;
+                        goto out;
+                }
+        }
+        if (local->fctx->stripe_count != data_to_int32 (data)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "error with stripe-count xattr (%d != %d)",
+                        local->fctx->stripe_count, data_to_int32 (data));
+                local->op_ret   = -1;
+                local->op_errno = EIO;
+                goto out;
+        }
+
+        /* index */
+        sprintf (key, "trusted.%s.stripe-index", this->name);
+        data = dict_get (dict, key);
+        if (!data) {
+                local->xattr_self_heal_needed = 1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get stripe-index");
+                goto out;
+        }
+        index = data_to_int32 (data);
+        if (index > priv->child_count) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "error with stripe-index xattr (%d)", index);
+                local->op_ret   = -1;
+                local->op_errno = EIO;
+                goto out;
+        }
+        if (local->fctx->xl_array) {
+                if (!local->fctx->xl_array[index])
+                        local->fctx->xl_array[index] = prev->this;
+        }
+
+	sprintf(key, "trusted.%s.stripe-coalesce", this->name);
+	data = dict_get(dict, key);
+	if (!data) {
+		/*
+		 * The file was probably created prior to coalesce support.
+		 * Assume non-coalesce mode for this file to maintain backwards
+		 * compatibility.
+		 */
+		gf_log(this->name, GF_LOG_DEBUG, "missing stripe-coalesce "
+			"attr, assume non-coalesce mode");
+		local->fctx->stripe_coalesce = 0;
+	} else {
+		local->fctx->stripe_coalesce = data_to_int32(data);
+	}
+
+
+out:
+        return 0;
+}
+
+int32_t
+stripe_xattr_request_build (xlator_t *this, dict_t *dict, uint64_t stripe_size,
+                            uint32_t stripe_count, uint32_t stripe_index,
+			    uint32_t stripe_coalesce)
+{
+        char            key[256]       = {0,};
+        int32_t         ret             = -1;
+
+        sprintf (key, "trusted.%s.stripe-size", this->name);
+        ret = dict_set_int64 (dict, key, stripe_size);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set %s in xattr_req dict", key);
+                goto out;
+        }
+
+        sprintf (key, "trusted.%s.stripe-count", this->name);
+        ret = dict_set_int32 (dict, key, stripe_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set %s in xattr_req dict", key);
+                goto out;
+        }
+
+        sprintf (key, "trusted.%s.stripe-index", this->name);
+        ret = dict_set_int32 (dict, key, stripe_index);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set %s in xattr_req dict", key);
+                goto out;
+        }
+
+	sprintf(key, "trusted.%s.stripe-coalesce", this->name);
+	ret = dict_set_int32(dict, key, stripe_coalesce);
+	if (ret) {
+		gf_log(this->name, GF_LOG_WARNING,
+			"failed to set %s in xattr_req_dict", key);
+		goto out;
+	}
+out:
+        return ret;
+}
+
+
+static int
+set_default_block_size (stripe_private_t *priv, char *num)
+{
+
+        int             ret = -1;
+        GF_VALIDATE_OR_GOTO ("stripe", THIS, out);
+        GF_VALIDATE_OR_GOTO (THIS->name, priv, out);
+        GF_VALIDATE_OR_GOTO (THIS->name, num, out);
+
+
+        if (gf_string2bytesize (num, &priv->block_size) != 0) {
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "invalid number format \"%s\"", num);
+                goto out;
+        }
+
+        ret = 0;
+
+ out:
+        return ret;
+
+}
+
+
+int
+set_stripe_block_size (xlator_t *this, stripe_private_t *priv, char *data)
+{
+        int                    ret = -1;
+        char                  *tmp_str = NULL;
+        char                  *tmp_str1 = NULL;
+        char                  *dup_str = NULL;
+        char                  *stripe_str = NULL;
+        char                  *pattern = NULL;
+        char                  *num = NULL;
+        struct stripe_options *temp_stripeopt = NULL;
+        struct stripe_options *stripe_opt = NULL;
+
+        if (!this || !priv || !data)
+                goto out;
+
+        /* Get the pattern for striping.
+           "option block-size *avi:10MB" etc */
+        stripe_str = strtok_r (data, ",", &tmp_str);
+        while (stripe_str) {
+                dup_str = gf_strdup (stripe_str);
+                stripe_opt = GF_CALLOC (1, sizeof (struct stripe_options),
+                                        gf_stripe_mt_stripe_options);
+                if (!stripe_opt) {
+                        goto out;
+                }
+
+                pattern = strtok_r (dup_str, ":", &tmp_str1);
+                num = strtok_r (NULL, ":", &tmp_str1);
+                if (!num) {
+                        num = pattern;
+                        pattern = "*";
+                        ret = set_default_block_size (priv, num);
+                        if (ret)
+                                goto out;
+                }
+                if (gf_string2bytesize (num, &stripe_opt->block_size) != 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "invalid number format \"%s\"", num);
+                        goto out;
+                }
+
+                if (stripe_opt->block_size < STRIPE_MIN_BLOCK_SIZE) {
+                        gf_log (this->name, GF_LOG_ERROR, "Invalid Block-size: "
+                                "%s. Should be atleast %llu bytes", num,
+                                STRIPE_MIN_BLOCK_SIZE);
+                        goto out;
+                }
+                if (stripe_opt->block_size % 512) {
+                        gf_log (this->name, GF_LOG_ERROR, "Block-size: %s should"
+                                " be a multiple of 512 bytes", num);
+                        goto out;
+                }
+
+                memcpy (stripe_opt->path_pattern, pattern, strlen (pattern));
+
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "block-size : pattern %s : size %"PRId64,
+                        stripe_opt->path_pattern, stripe_opt->block_size);
+
+                if (priv->pattern)
+                        temp_stripeopt = NULL;
+                else
+                        temp_stripeopt = priv->pattern;
+
+                stripe_opt->next = temp_stripeopt;
+
+                priv->pattern = stripe_opt;
+                stripe_opt = NULL;
+
+                GF_FREE (dup_str);
+                dup_str = NULL;
+
+                stripe_str = strtok_r (NULL, ",", &tmp_str);
+        }
+
+        ret = 0;
+out:
+
+        GF_FREE (dup_str);
+
+        GF_FREE (stripe_opt);
+
+        return ret;
+}
+
+int32_t
+stripe_iatt_merge (struct iatt *from, struct iatt *to)
+{
+        if (to->ia_size < from->ia_size)
+                to->ia_size = from->ia_size;
+        if (to->ia_mtime < from->ia_mtime)
+                to->ia_mtime = from->ia_mtime;
+        if (to->ia_ctime < from->ia_ctime)
+                to->ia_ctime = from->ia_ctime;
+        if (to->ia_atime < from->ia_atime)
+                to->ia_atime = from->ia_atime;
+        return 0;
+}
+
+off_t
+coalesced_offset(off_t offset, uint64_t stripe_size, int stripe_count)
+{
+	size_t line_size = 0;
+	uint64_t stripe_num = 0;
+	off_t coalesced_offset = 0;
+
+	line_size = stripe_size * stripe_count;
+	stripe_num = offset / line_size;
+
+	coalesced_offset = (stripe_num * stripe_size) +
+		(offset % stripe_size);
+
+	return coalesced_offset;
+}
+
+off_t
+uncoalesced_size(off_t size, uint64_t stripe_size, int stripe_count,
+		int stripe_index)
+{
+        uint64_t nr_full_stripe_chunks = 0, mod = 0;
+
+        if (!size)
+                return size;
+
+	/*
+	 * Estimate the number of fully written stripes from the
+	 * local file size. Each stripe_size chunk corresponds to
+	 * a stripe.
+	 */
+        nr_full_stripe_chunks = (size / stripe_size) * stripe_count;
+        mod = size % stripe_size;
+
+        if (!mod) {
+                /*
+		 * There is no remainder, thus we could have overestimated
+		 * the size of the file in terms of chunks. Trim the number
+		 * of chunks by the following stripe members and leave it
+		 * up to those nodes to respond with a larger size (if
+		 * necessary).
+		 */
+                nr_full_stripe_chunks -= stripe_count -
+                        (stripe_index + 1);
+                size = nr_full_stripe_chunks * stripe_size;
+        } else {
+		/*
+		 * There is a remainder and thus we own the last chunk of the
+		 * file. Add the preceding stripe members of the final stripe
+		 * along with the remainder to calculate the exact size.
+		 */
+                nr_full_stripe_chunks += stripe_index;
+                size = nr_full_stripe_chunks * stripe_size + mod;
+        }
+
+        return size;
+}
+
diff --git a/xlators/cluster/stripe/src/stripe-mem-types.h b/xlators/cluster/stripe/src/stripe-mem-types.h
index 06667107d..e9ac9cf46 100644
--- a/xlators/cluster/stripe/src/stripe-mem-types.h
+++ b/xlators/cluster/stripe/src/stripe-mem-types.h
@@ -1,21 +1,11 @@
-
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
 
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -25,15 +15,16 @@
 #include "mem-types.h"
 
 enum gf_stripe_mem_types_ {
-        gf_stripe_mt_stripe_local_t = gf_common_mt_end + 1,
-        gf_stripe_mt_iovec,
-        gf_stripe_mt_readv_replies,
+        gf_stripe_mt_iovec = gf_common_mt_end + 1,
+        gf_stripe_mt_stripe_replies,
         gf_stripe_mt_stripe_fd_ctx_t,
         gf_stripe_mt_char,
         gf_stripe_mt_int8_t,
+        gf_stripe_mt_int32_t,
         gf_stripe_mt_xlator_t,
         gf_stripe_mt_stripe_private_t,
         gf_stripe_mt_stripe_options,
+        gf_stripe_mt_xattr_sort_t,
         gf_stripe_mt_end
 };
 #endif
diff --git a/xlators/cluster/stripe/src/stripe.c b/xlators/cluster/stripe/src/stripe.c
index 00c888f68..69b510e23 100644
--- a/xlators/cluster/stripe/src/stripe.c
+++ b/xlators/cluster/stripe/src/stripe.c
@@ -1,25 +1,16 @@
 /*
-  Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 /**
  * xlators/cluster/stripe:
- *    Stripe translator, stripes the data accross its child nodes,
+ *    Stripe translator, stripes the data across its child nodes,
  *    as per the options given in the volfile. The striping works
  *    fairly simple. It writes files at different offset as per
  *    calculation. So, 'ls -l' output at the real posix level will
@@ -32,65 +23,19 @@
  *  very much necessary, or else, use it in combination with AFR, to have a
  *  backup copy.
  */
+#include <fnmatch.h>
 
 #include "stripe.h"
+#include "libxlator.h"
+#include "byte-order.h"
+#include "statedump.h"
 
-void
-stripe_local_wipe (stripe_local_t *local)
-{
-        if (!local)
-                goto out;
-
-        if (local->loc.path)
-                loc_wipe (&local->loc);
-        if (local->loc2.path)
-                loc_wipe (&local->loc2);
-out:
-        return;
-}
-
-/**
- * stripe_get_matching_bs - Get the matching block size for the given path.
- */
-int32_t
-stripe_get_matching_bs (const char *path, struct stripe_options *opts,
-                        uint64_t default_bs)
-{
-        struct stripe_options *trav       = NULL;
-        char                  *pathname   = NULL;
-        uint64_t               block_size = 0;
-
-        block_size = default_bs;
-
-        if (!path || !opts)
-                goto out;
-
-        /* FIXME: is a strdup really necessary? */
-        pathname = gf_strdup (path);
-        if (!pathname)
-                goto out;
-
-        trav = opts;
-        while (trav) {
-                if (!fnmatch (trav->path_pattern, pathname, FNM_NOESCAPE)) {
-                        block_size = trav->block_size;
-                        break;
-                }
-                trav = trav->next;
-        }
-
-        GF_FREE (pathname);
-
-out:
-        return block_size;
-}
-
-
+struct volume_options options[];
 
 int32_t
 stripe_sh_chown_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno,
-                     struct iatt *preop, struct iatt *postop)
+                     struct iatt *preop, struct iatt *postop, dict_t *xdata)
 {
         int             callcnt = -1;
         stripe_local_t *local   = NULL;
@@ -109,8 +54,7 @@ stripe_sh_chown_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                stripe_local_wipe (local);
-                STACK_DESTROY (frame->root);
+                STRIPE_STACK_DESTROY (frame);
         }
 out:
         return 0;
@@ -120,7 +64,7 @@ int32_t
 stripe_sh_make_entry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                           int32_t op_ret, int32_t op_errno, inode_t *inode,
                           struct iatt *buf, struct iatt *preparent,
-                          struct iatt *postparent)
+                          struct iatt *postparent, dict_t *xdata)
 {
         stripe_local_t *local = NULL;
         call_frame_t    *prev = NULL;
@@ -135,7 +79,7 @@ stripe_sh_make_entry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         STACK_WIND (frame, stripe_sh_chown_cbk, prev->this,
                     prev->this->fops->setattr, &local->loc,
-                    &local->stbuf, (GF_SET_ATTR_UID | GF_SET_ATTR_GID));
+                    &local->stbuf, (GF_SET_ATTR_UID | GF_SET_ATTR_GID), NULL);
 
 out:
         return 0;
@@ -149,7 +93,7 @@ stripe_entry_self_heal (call_frame_t *frame, xlator_t *this,
         call_frame_t     *rframe = NULL;
         stripe_local_t   *rlocal = NULL;
         stripe_private_t *priv   = NULL;
-        dict_t           *dict   = NULL;
+        dict_t           *xdata  = NULL;
         int               ret    = 0;
 
         if (!local || !this || !frame) {
@@ -167,8 +111,7 @@ stripe_entry_self_heal (call_frame_t *frame, xlator_t *this,
         if (!rframe) {
                 goto out;
         }
-        rlocal = GF_CALLOC (1, sizeof (stripe_local_t),
-                            gf_stripe_mt_stripe_local_t);
+        rlocal = mem_get0 (this->local_pool);
         if (!rlocal) {
                 goto out;
         }
@@ -177,14 +120,14 @@ stripe_entry_self_heal (call_frame_t *frame, xlator_t *this,
         loc_copy (&rlocal->loc, &local->loc);
         memcpy (&rlocal->stbuf, &local->stbuf, sizeof (struct iatt));
 
-        dict = dict_new ();
-        if (!dict)
+        xdata = dict_new ();
+        if (!xdata)
                 goto out;
 
-        ret = dict_set_static_bin (dict, "gfid-req", local->stbuf.ia_gfid, 16);
+        ret = dict_set_static_bin (xdata, "gfid-req", local->stbuf.ia_gfid, 16);
         if (ret)
                 gf_log (this->name, GF_LOG_WARNING,
-                        "failed to set gfid-req");
+                        "%s: failed to set gfid-req", local->loc.path);
 
         while (trav) {
                 if (IA_ISREG (local->stbuf.ia_type)) {
@@ -192,36 +135,43 @@ stripe_entry_self_heal (call_frame_t *frame, xlator_t *this,
                                     trav->xlator, trav->xlator->fops->mknod,
                                     &local->loc,
                                     st_mode_from_ia (local->stbuf.ia_prot,
-                                                     local->stbuf.ia_type), 0,
-                                    dict);
+                                                     local->stbuf.ia_type),
+                                    0, 0, xdata);
                 }
                 if (IA_ISDIR (local->stbuf.ia_type)) {
                         STACK_WIND (rframe, stripe_sh_make_entry_cbk,
                                     trav->xlator, trav->xlator->fops->mkdir,
-                                    &local->loc, st_mode_from_ia (local->stbuf.ia_prot,
-                                                                  local->stbuf.ia_type),
-                                    dict);
+                                    &local->loc,
+                                    st_mode_from_ia (local->stbuf.ia_prot,
+                                                     local->stbuf.ia_type),
+                                    0, xdata);
                 }
                 trav = trav->next;
         }
 
+        if (xdata)
+                dict_unref (xdata);
+        return 0;
+
 out:
-        if (dict)
-                dict_unref (dict);
+        if (rframe)
+                STRIPE_STACK_DESTROY (rframe);
+        if (xdata)
+                dict_unref (xdata);
 
         return 0;
 }
 
+
 int32_t
 stripe_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, inode_t *inode,
-                   struct iatt *buf, dict_t *dict, struct iatt *postparent)
+                   struct iatt *buf, dict_t *xdata, struct iatt *postparent)
 {
-        int32_t         callcnt = 0;
-        dict_t         *tmp_dict = NULL;
-        inode_t        *tmp_inode = NULL;
-        stripe_local_t *local = NULL;
-        call_frame_t   *prev = NULL;
+        int32_t         callcnt     = 0;
+        stripe_local_t *local       = NULL;
+        call_frame_t   *prev        = NULL;
+        int             ret         = 0;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -243,7 +193,7 @@ stripe_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                         strerror (op_errno));
                         if (local->op_errno != ESTALE)
                                 local->op_errno = op_errno;
-                        if ((op_errno != ENOENT) ||
+                        if (((op_errno != ENOENT) && (op_errno != ENOTCONN)) ||
                             (prev->this == FIRST_CHILD (this)))
                                 local->failed = 1;
                         if (op_errno == ENOENT)
@@ -252,51 +202,81 @@ stripe_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
                 if (op_ret >= 0) {
                         local->op_ret = 0;
+                        if (IA_ISREG (buf->ia_type)) {
+                                ret = stripe_ctx_handle (this, prev, local,
+                                                         xdata);
+                                if (ret)
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                 "Error getting fctx info from"
+                                                 " dict");
+                        }
 
                         if (FIRST_CHILD(this) == prev->this) {
                                 local->stbuf      = *buf;
                                 local->postparent = *postparent;
                                 local->inode = inode_ref (inode);
-                                local->dict = dict_ref (dict);
+                                if (xdata)
+                                        local->xdata = dict_ref (xdata);
+                                if (local->xattr) {
+                                        stripe_aggregate_xattr (local->xdata,
+                                                                local->xattr);
+                                        dict_unref (local->xattr);
+                                        local->xattr = NULL;
+                                }
+                        }
+
+                        if (!local->xdata && !local->xattr) {
+                                local->xattr = dict_ref (xdata);
+                        } else if (local->xdata) {
+                                stripe_aggregate_xattr (local->xdata, xdata);
+                        } else if (local->xattr) {
+                                stripe_aggregate_xattr (local->xattr, xdata);
                         }
+
                         local->stbuf_blocks      += buf->ia_blocks;
                         local->postparent_blocks += postparent->ia_blocks;
 
+			correct_file_size(buf, local->fctx, prev);
+
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
                         if (local->postparent_size < postparent->ia_size)
                                 local->postparent_size = postparent->ia_size;
+
+                        if (uuid_is_null (local->ia_gfid))
+                                uuid_copy (local->ia_gfid, buf->ia_gfid);
+
+                        /* Make sure the gfid on all the nodes are same */
+                        if (uuid_compare (local->ia_gfid, buf->ia_gfid)) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "%s: gfid different on subvolume %s",
+                                        local->loc.path, prev->this->name);
+                        }
                 }
         }
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                if (local->op_ret == 0 && local->entry_self_heal_needed)
+                if (local->op_ret == 0 && local->entry_self_heal_needed &&
+                    !uuid_is_null (local->loc.inode->gfid))
                         stripe_entry_self_heal (frame, this, local);
 
                 if (local->failed)
                         local->op_ret = -1;
 
-                tmp_dict  = local->dict;
-                tmp_inode = local->inode;
-
                 if (local->op_ret != -1) {
                         local->stbuf.ia_blocks      = local->stbuf_blocks;
                         local->stbuf.ia_size        = local->stbuf_size;
                         local->postparent.ia_blocks = local->postparent_blocks;
                         local->postparent.ia_size   = local->postparent_size;
+                        inode_ctx_put (local->inode, this,
+                                       (uint64_t) (long)local->fctx);
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (lookup, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (lookup, frame, local->op_ret,
                                      local->op_errno, local->inode,
-                                     &local->stbuf, local->dict,
+                                     &local->stbuf, local->xdata,
                                      &local->postparent);
-
-                if (tmp_inode)
-                        inode_unref (tmp_inode);
-                if (tmp_dict)
-                        dict_unref (tmp_dict);
         }
 out:
         return 0;
@@ -304,12 +284,15 @@ out:
 
 int32_t
 stripe_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
-               dict_t *xattr_req)
+               dict_t *xdata)
 {
-        stripe_local_t   *local = NULL;
-        xlator_list_t    *trav = NULL;
-        stripe_private_t *priv = NULL;
+        stripe_local_t   *local    = NULL;
+        xlator_list_t    *trav     = NULL;
+        stripe_private_t *priv     = NULL;
         int32_t           op_errno = EINVAL;
+        int64_t           filesize = 0;
+        int               ret      = 0;
+        uint64_t          tmpctx   = 0;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -321,8 +304,7 @@ stripe_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
         trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -331,26 +313,60 @@ stripe_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
         frame->local = local;
         loc_copy (&local->loc, loc);
 
+        inode_ctx_get (local->inode, this, &tmpctx);
+        if (tmpctx)
+                local->fctx = (stripe_fd_ctx_t*) (long)tmpctx;
+
+        /* quick-read friendly changes */
+        if (xdata && dict_get (xdata, GF_CONTENT_KEY)) {
+                ret = dict_get_int64 (xdata, GF_CONTENT_KEY, &filesize);
+                if (!ret && (filesize > priv->block_size))
+                        dict_del (xdata, GF_CONTENT_KEY);
+        }
+
+        /* get stripe-size xattr on lookup. This would be required for
+         * open/read/write/pathinfo calls. Hence we send down the request
+         * even when type == IA_INVAL */
+
+	/*
+	 * We aren't guaranteed to have xdata here. We need the format info for
+	 * the file, so allocate xdata if necessary.
+	 */
+	if (!xdata)
+		xdata = dict_new();
+	else
+		xdata = dict_ref(xdata);
+
+        if (xdata && (IA_ISREG (loc->inode->ia_type) ||
+            (loc->inode->ia_type == IA_INVAL))) {
+                ret = stripe_xattr_request_build (this, xdata, 8, 4, 4, 0);
+                if (ret)
+                        gf_log (this->name , GF_LOG_ERROR, "Failed to build"
+                                " xattr request for %s", loc->path);
+
+        }
+
         /* Everytime in stripe lookup, all child nodes
            should be looked up */
         local->call_count = priv->child_count;
         while (trav) {
                 STACK_WIND (frame, stripe_lookup_cbk, trav->xlator,
-                            trav->xlator->fops->lookup,
-                            loc, xattr_req);
+                            trav->xlator->fops->lookup, loc, xdata);
                 trav = trav->next;
         }
 
+	dict_unref(xdata);
+
         return 0;
 err:
-        STACK_UNWIND_STRICT (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        STRIPE_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                 int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -385,6 +401,9 @@ stripe_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         }
 
                         local->stbuf_blocks += buf->ia_blocks;
+
+			correct_file_size(buf, local->fctx, prev);
+
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
                 }
@@ -400,20 +419,20 @@ stripe_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->stbuf.ia_blocks = local->stbuf_blocks;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (stat, frame, local->op_ret,
-                                     local->op_errno, &local->stbuf);
+                STRIPE_STACK_UNWIND (stat, frame, local->op_ret,
+                                     local->op_errno, &local->stbuf, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
+stripe_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
+	stripe_fd_ctx_t  *fctx = NULL;
         int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
@@ -431,8 +450,7 @@ stripe_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -441,23 +459,30 @@ stripe_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
         frame->local = local;
         local->call_count = priv->child_count;
 
+	if (IA_ISREG(loc->inode->ia_type)) {
+		inode_ctx_get(loc->inode, this, (uint64_t *) &fctx);
+		if (!fctx)
+			goto err;
+		local->fctx = fctx;
+	}
+
         while (trav) {
                 STACK_WIND (frame, stripe_stat_cbk, trav->xlator,
-                            trav->xlator->fops->stat, loc);
+                            trav->xlator->fops->stat, loc, NULL);
                 trav = trav->next;
         }
 
         return 0;
 
 err:
-        STACK_UNWIND_STRICT (stat, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct statvfs *stbuf)
+                   int32_t op_ret, int32_t op_errno, struct statvfs *stbuf, dict_t *xdata)
 {
         stripe_local_t *local = NULL;
         int32_t         callcnt = 0;
@@ -494,15 +519,15 @@ stripe_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                STACK_UNWIND_STRICT (statfs, frame, local->op_ret,
-                                     local->op_errno, &local->statvfs_buf);
+                STRIPE_STACK_UNWIND (statfs, frame, local->op_ret,
+                                     local->op_errno, &local->statvfs_buf, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
+stripe_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         xlator_list_t    *trav = NULL;
@@ -517,8 +542,7 @@ stripe_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
         priv = this->private;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -530,13 +554,13 @@ stripe_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
         local->call_count = priv->child_count;
         while (trav) {
                 STACK_WIND (frame, stripe_statfs_cbk, trav->xlator,
-                            trav->xlator->fops->statfs, loc);
+                            trav->xlator->fops->statfs, loc, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (statfs, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
@@ -545,7 +569,7 @@ err:
 int32_t
 stripe_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                     struct iatt *postbuf)
+                     struct iatt *postbuf, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -583,6 +607,9 @@ stripe_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->prebuf_blocks  += prebuf->ia_blocks;
                         local->postbuf_blocks += postbuf->ia_blocks;
 
+			correct_file_size(prebuf, local->fctx, prev);
+			correct_file_size(postbuf, local->fctx, prev);
+
                         if (local->prebuf_size < prebuf->ia_size)
                                 local->prebuf_size = prebuf->ia_size;
 
@@ -603,22 +630,23 @@ stripe_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->post_buf.ia_size   = local->postbuf_size;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (truncate, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (truncate, frame, local->op_ret,
                                      local->op_errno, &local->pre_buf,
-                                     &local->post_buf);
+                                     &local->post_buf, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
+stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset, dict_t *xdata)
 {
-        xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
+	stripe_fd_ctx_t  *fctx = NULL;
         int32_t           op_errno = EINVAL;
+	int		  i, eof_idx;
+	off_t		  dest_offset, tmp_offset;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -627,7 +655,6 @@ stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
         VALIDATE_OR_GOTO (loc->inode, err);
 
         priv = this->private;
-        trav = this->children;
 
         if (priv->first_child_down) {
                 op_errno = ENOTCONN;
@@ -635,8 +662,7 @@ stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -645,15 +671,55 @@ stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
         frame->local = local;
         local->call_count = priv->child_count;
 
-        while (trav) {
-                STACK_WIND (frame, stripe_truncate_cbk, trav->xlator,
-                            trav->xlator->fops->truncate, loc, offset);
-                trav = trav->next;
-        }
+	inode_ctx_get(loc->inode, this, (uint64_t *) &fctx);
+	if (!fctx) {
+		gf_log(this->name, GF_LOG_ERROR, "no stripe context");
+		op_errno = EINVAL;
+		goto err;
+	}
+
+	local->fctx = fctx;
+	eof_idx = (offset / fctx->stripe_size) % fctx->stripe_count;
+
+	for (i = 0; i < fctx->stripe_count; i++) {
+		if (!fctx->xl_array[i]) {
+			gf_log(this->name, GF_LOG_ERROR,
+				"no xlator at index %d", i);
+			op_errno = EINVAL;
+			goto err;
+		}
+
+		if (fctx->stripe_coalesce) {
+			/*
+	 		 * The node that owns EOF is truncated to the exact
+	 		 * coalesced offset. Nodes prior to this index should
+	 		 * be rounded up to the size of the complete stripe,
+	 		 * while nodes after this index should be rounded down
+			 * to the size of the previous stripe.
+			 */
+			if (i < eof_idx)
+				tmp_offset = roof(offset, fctx->stripe_size *
+						fctx->stripe_count);
+			else if (i > eof_idx)
+				tmp_offset = floor(offset, fctx->stripe_size *
+						fctx->stripe_count);
+			else
+				tmp_offset = offset;
+
+			dest_offset = coalesced_offset(tmp_offset,
+					fctx->stripe_size, fctx->stripe_count);
+		} else {
+			dest_offset = offset;
+		}
+
+		STACK_WIND(frame, stripe_truncate_cbk, fctx->xl_array[i],
+			fctx->xl_array[i]->fops->truncate, loc, dest_offset,
+			NULL);
+	}
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (truncate, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
@@ -661,7 +727,7 @@ err:
 int32_t
 stripe_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno,
-                    struct iatt *preop, struct iatt *postop)
+                    struct iatt *preop, struct iatt *postop, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -700,6 +766,9 @@ stripe_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->prebuf_blocks  += preop->ia_blocks;
                         local->postbuf_blocks += postop->ia_blocks;
 
+			correct_file_size(preop, local->fctx, prev);
+			correct_file_size(postop, local->fctx, prev);
+
                         if (local->prebuf_size < preop->ia_size)
                                 local->prebuf_size = preop->ia_size;
                         if (local->postbuf_size < postop->ia_size)
@@ -719,10 +788,9 @@ stripe_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->post_buf.ia_size   = local->postbuf_size;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (setattr, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (setattr, frame, local->op_ret,
                                      local->op_errno, &local->pre_buf,
-                                     &local->post_buf);
+                                     &local->post_buf, NULL);
         }
 out:
         return 0;
@@ -731,11 +799,12 @@ out:
 
 int32_t
 stripe_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                struct iatt *stbuf, int32_t valid)
+                struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
+	stripe_fd_ctx_t	 *fctx = NULL;
         int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
@@ -753,33 +822,47 @@ stripe_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         frame->local = local;
-        local->call_count = priv->child_count;
+        if (!IA_ISDIR (loc->inode->ia_type) &&
+            !IA_ISREG (loc->inode->ia_type)) {
+                local->call_count = 1;
+                STACK_WIND (frame, stripe_setattr_cbk, FIRST_CHILD (this),
+                            FIRST_CHILD (this)->fops->setattr,
+                            loc, stbuf, valid, NULL);
+                return 0;
+        }
+
+	if (IA_ISREG(loc->inode->ia_type)) {
+		inode_ctx_get(loc->inode, this, (uint64_t *) &fctx);
+		if (!fctx)
+			goto err;
+		local->fctx = fctx;
+	}
 
+        local->call_count = priv->child_count;
         while (trav) {
                 STACK_WIND (frame, stripe_setattr_cbk,
                             trav->xlator, trav->xlator->fops->setattr,
-                            loc, stbuf, valid);
+                            loc, stbuf, valid, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (setattr, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                 struct iatt *stbuf, int32_t valid)
+                 struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
@@ -795,8 +878,7 @@ stripe_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
         trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -807,13 +889,13 @@ stripe_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
         while (trav) {
                 STACK_WIND (frame, stripe_setattr_cbk, trav->xlator,
-                            trav->xlator->fops->fsetattr, fd, stbuf, valid);
+                            trav->xlator->fops->fsetattr, fd, stbuf, valid, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (fsetattr, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
@@ -821,7 +903,8 @@ int32_t
 stripe_stack_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int32_t op_ret, int32_t op_errno, struct iatt *buf,
                          struct iatt *preoldparent, struct iatt *postoldparent,
-                         struct iatt *prenewparent, struct iatt *postnewparent)
+                         struct iatt *prenewparent, struct iatt *postnewparent,
+                         dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -858,6 +941,8 @@ stripe_stack_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->pre_buf.ia_blocks    += prenewparent->ia_blocks;
                         local->post_buf.ia_blocks   += postnewparent->ia_blocks;
 
+			correct_file_size(buf, local->fctx, prev);
+
                         if (local->stbuf.ia_size < buf->ia_size)
                                 local->stbuf.ia_size =  buf->ia_size;
 
@@ -880,11 +965,10 @@ stripe_stack_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (rename, frame, local->op_ret, local->op_errno,
+                STRIPE_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
                                      &local->stbuf, &local->preparent,
                                      &local->postparent,  &local->pre_buf,
-                                     &local->post_buf);
+                                     &local->post_buf, NULL);
         }
 out:
         return 0;
@@ -894,7 +978,8 @@ int32_t
 stripe_first_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int32_t op_ret, int32_t op_errno, struct iatt *buf,
                          struct iatt *preoldparent, struct iatt *postoldparent,
-                         struct iatt *prenewparent, struct iatt *postnewparent)
+                         struct iatt *prenewparent, struct iatt *postnewparent,
+                         dict_t *xdata)
 {
         stripe_local_t *local = NULL;
         xlator_list_t  *trav = NULL;
@@ -925,25 +1010,25 @@ stripe_first_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         while (trav) {
                 STACK_WIND (frame, stripe_stack_rename_cbk,
                             trav->xlator, trav->xlator->fops->rename,
-                            &local->loc, &local->loc2);
+                            &local->loc, &local->loc2, NULL);
                 trav = trav->next;
         }
         return 0;
 
 unwind:
-        stripe_local_wipe (local);
-        STACK_UNWIND_STRICT (rename, frame, -1, op_errno, buf, preoldparent,
-                             postoldparent, prenewparent, postnewparent);
+        STRIPE_STACK_UNWIND (rename, frame, -1, op_errno, buf, preoldparent,
+                             postoldparent, prenewparent, postnewparent, NULL);
         return 0;
 }
 
 int32_t
 stripe_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
-               loc_t *newloc)
+               loc_t *newloc, dict_t *xdata)
 {
         stripe_private_t *priv = NULL;
         stripe_local_t   *local = NULL;
         xlator_list_t    *trav = NULL;
+	stripe_fd_ctx_t	 *fctx = NULL;
         int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
@@ -963,8 +1048,7 @@ stripe_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -975,24 +1059,67 @@ stripe_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
 
         local->call_count = priv->child_count;
 
+	if (IA_ISREG(oldloc->inode->ia_type)) {
+		inode_ctx_get(oldloc->inode, this, (uint64_t *) &fctx);
+		if (!fctx)
+			goto err;
+		local->fctx = fctx;
+	}
+
         frame->local = local;
 
         STACK_WIND (frame, stripe_first_rename_cbk, trav->xlator,
-                    trav->xlator->fops->rename, oldloc, newloc);
+                    trav->xlator->fops->rename, oldloc, newloc, NULL);
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL, NULL, NULL,
-                             NULL, NULL);
+        STRIPE_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL,
+                             NULL, NULL, NULL);
         return 0;
 }
+int32_t
+stripe_first_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                         struct iatt *postparent, dict_t *xdata)
+{
+        stripe_local_t *local   = NULL;
+        call_frame_t   *prev = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = cookie;
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "%s returned %s",
+                        prev->this->name, strerror (op_errno));
+                goto out;
+        }
+        local->op_ret = 0;
+        local->preparent  = *preparent;
+        local->postparent = *postparent;
+        local->preparent_blocks  += preparent->ia_blocks;
+        local->postparent_blocks += postparent->ia_blocks;
+
+        STRIPE_STACK_UNWIND(unlink, frame, local->op_ret, local->op_errno,
+                            &local->preparent, &local->postparent, xdata);
+        return 0;
+out:
+        STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
 
 
 
 int32_t
 stripe_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                   struct iatt *postparent)
+                   struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local   = NULL;
@@ -1014,50 +1141,33 @@ stripe_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         gf_log (this->name, GF_LOG_DEBUG, "%s returned %s",
                                 prev->this->name, strerror (op_errno));
                         local->op_errno = op_errno;
-                        if ((op_errno != ENOENT) ||
-                            (prev->this == FIRST_CHILD (this)))
+                        if (op_errno != ENOENT) {
                                 local->failed = 1;
-                }
-                if (op_ret >= 0) {
-                        local->op_ret = op_ret;
-                        if (FIRST_CHILD(this) == prev->this) {
-                                local->preparent  = *preparent;
-                                local->postparent = *postparent;
+                                local->op_ret = op_ret;
                         }
-                        local->preparent_blocks  += preparent->ia_blocks;
-                        local->postparent_blocks += postparent->ia_blocks;
-
-                        if (local->preparent_size < preparent->ia_size)
-                                local->preparent_size = preparent->ia_size;
-
-                        if (local->postparent_size < postparent->ia_size)
-                                local->postparent_size = postparent->ia_size;
                 }
         }
         UNLOCK (&frame->lock);
 
-        if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
-
-                if (local->op_ret != -1) {
-                        local->preparent.ia_blocks  = local->preparent_blocks;
-                        local->preparent.ia_size    = local->preparent_size;
-                        local->postparent.ia_blocks = local->postparent_blocks;
-                        local->postparent.ia_size   = local->postparent_size;
+        if (callcnt == 1) {
+                if (local->failed) {
+                        op_errno = local->op_errno;
+                        goto out;
                 }
-
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (unlink, frame, local->op_ret,
-                                     local->op_errno, &local->preparent,
-                                     &local->postparent);
+                STACK_WIND(frame, stripe_first_unlink_cbk, FIRST_CHILD (this),
+                           FIRST_CHILD (this)->fops->unlink, &local->loc,
+                           local->xflag, local->xdata);
         }
+        return 0;
 out:
+        STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
+
         return 0;
 }
 
 int32_t
-stripe_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
+stripe_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
+               int xflag, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
@@ -1085,26 +1195,32 @@ stripe_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
+        loc_copy (&local->loc, loc);
+        local->xflag = xflag;
+
+        if (xdata)
+            local->xdata = dict_ref (xdata);
+
         frame->local = local;
         local->call_count = priv->child_count;
+        trav = trav->next; /* Skip the first child */
 
         while (trav) {
                 STACK_WIND (frame, stripe_unlink_cbk,
                             trav->xlator, trav->xlator->fops->unlink,
-                            loc);
+                            loc, xflag, xdata);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
@@ -1112,10 +1228,8 @@ err:
 int32_t
 stripe_first_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         int32_t op_ret, int32_t op_errno,struct iatt *preparent,
-                        struct iatt *postparent)
-
+                        struct iatt *postparent, dict_t *xdata)
 {
-        xlator_list_t *trav = NULL;
         stripe_local_t *local = NULL;
 
         if (!this || !frame || !frame->local) {
@@ -1128,11 +1242,10 @@ stripe_first_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto err;
         }
 
-        trav = this->children;
         local = frame->local;
+        local->op_ret = 0;
 
         local->call_count--; /* First child successful */
-        trav = trav->next; /* Skip first child */
 
         local->preparent  = *preparent;
         local->postparent = *postparent;
@@ -1141,21 +1254,60 @@ stripe_first_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         local->preparent_blocks  += preparent->ia_blocks;
         local->postparent_blocks += postparent->ia_blocks;
 
-        while (trav) {
-                STACK_WIND (frame, stripe_unlink_cbk, trav->xlator,
-                            trav->xlator->fops->rmdir, &local->loc);
-                trav = trav->next;
-        }
-
+        STRIPE_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
+                             &local->preparent, &local->postparent, xdata);
         return 0;
 err:
-        STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (rmdir, frame, op_ret, op_errno, NULL, NULL, NULL);
         return 0;
 
 }
 
 int32_t
-stripe_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
+stripe_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                   struct iatt *postparent, dict_t *xdata)
+{
+        int32_t         callcnt = 0;
+        stripe_local_t *local   = NULL;
+        call_frame_t   *prev = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = cookie;
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
+
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG, "%s returned %s",
+                                prev->this->name, strerror (op_errno));
+                        if (op_errno != ENOENT)
+                                local->failed = 1;
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        if (callcnt == 1) {
+                if (local->failed)
+                        goto out;
+                STACK_WIND (frame, stripe_first_rmdir_cbk, FIRST_CHILD (this),
+                            FIRST_CHILD (this)->fops->rmdir, &local->loc,
+                            local->flags, NULL);
+        }
+        return 0;
+out:
+        STRIPE_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
+}
+
+int32_t
+stripe_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
@@ -1178,8 +1330,7 @@ stripe_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -1187,14 +1338,19 @@ stripe_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
         local->op_ret = -1;
         frame->local = local;
         loc_copy (&local->loc, loc);
+        local->flags = flags;
         local->call_count = priv->child_count;
+        trav = trav->next; /* skip the first child */
 
-        STACK_WIND (frame, stripe_first_rmdir_cbk,  trav->xlator,
-                    trav->xlator->fops->rmdir, loc);
+        while (trav) {
+                STACK_WIND (frame, stripe_rmdir_cbk,  trav->xlator,
+                            trav->xlator->fops->rmdir, loc, flags, NULL);
+                trav = trav->next;
+        }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (rmdir, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
@@ -1203,7 +1359,7 @@ int32_t
 stripe_mknod_ifreg_fail_unlink_cbk (call_frame_t *frame, void *cookie,
                                     xlator_t *this, int32_t op_ret,
                                     int32_t op_errno, struct iatt *preparent,
-                                    struct iatt *postparent)
+                                    struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -1222,10 +1378,9 @@ stripe_mknod_ifreg_fail_unlink_cbk (call_frame_t *frame, void *cookie,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (mknod, frame, local->op_ret, local->op_errno,
+                STRIPE_STACK_UNWIND (mknod, frame, local->op_ret, local->op_errno,
                                      local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
+                                     &local->preparent, &local->postparent, NULL);
         }
 out:
         return 0;
@@ -1237,7 +1392,7 @@ out:
 int32_t
 stripe_mknod_ifreg_setxattr_cbk (call_frame_t *frame, void *cookie,
                                  xlator_t *this, int32_t op_ret,
-                                 int32_t op_errno)
+                                 int32_t op_errno, dict_t *xdata)
 {
         int32_t           callcnt = 0;
         stripe_local_t   *local = NULL;
@@ -1276,16 +1431,15 @@ stripe_mknod_ifreg_setxattr_cbk (call_frame_t *frame, void *cookie,
                                             stripe_mknod_ifreg_fail_unlink_cbk,
                                             trav->xlator,
                                             trav->xlator->fops->unlink,
-                                            &local->loc);
+                                            &local->loc, 0, NULL);
                                 trav = trav->next;
                         }
                         return 0;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (mknod, frame, local->op_ret, local->op_errno,
+                STRIPE_STACK_UNWIND (mknod, frame, local->op_ret, local->op_errno,
                                      local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
+                                     &local->preparent, &local->postparent, NULL);
         }
 out:
         return 0;
@@ -1295,13 +1449,13 @@ int32_t
 stripe_mknod_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         int32_t op_ret, int32_t op_errno, inode_t *inode,
                         struct iatt *buf, struct iatt *preparent,
-                        struct iatt *postparent)
+                        struct iatt *postparent, dict_t *xdata)
 {
-        int               ret = 0;
         int32_t           callcnt = 0;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
         call_frame_t     *prev = NULL;
+        xlator_list_t    *trav = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -1309,7 +1463,7 @@ stripe_mknod_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         }
 
         prev  = cookie;
-	priv  = this->private;
+        priv  = this->private;
         local = frame->local;
 
         LOCK (&frame->lock);
@@ -1325,20 +1479,24 @@ stripe_mknod_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 local->failed = 1;
                         local->op_errno = op_errno;
                 }
-
                 if (op_ret >= 0) {
                         local->op_ret = op_ret;
 
-                        if (FIRST_CHILD(this) == prev->this) {
-                                local->stbuf      = *buf;
-                                local->preparent  = *preparent;
-                                local->postparent = *postparent;
-                        }
+                        /* Can be used as a mechanism to understand if mknod
+                           was successful in at least one place */
+                        if (uuid_is_null (local->ia_gfid))
+                                uuid_copy (local->ia_gfid, buf->ia_gfid);
+
+			if (stripe_ctx_handle(this, prev, local, xdata))
+				gf_log(this->name, GF_LOG_ERROR,
+					"Error getting fctx info from dict");
 
                         local->stbuf_blocks += buf->ia_blocks;
                         local->preparent_blocks  += preparent->ia_blocks;
                         local->postparent_blocks += postparent->ia_blocks;
 
+			correct_file_size(buf, local->fctx, prev);
+
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
                         if (local->preparent_size < preparent->ia_size)
@@ -1353,6 +1511,23 @@ stripe_mknod_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
+                if ((local->op_ret == -1) && !uuid_is_null (local->ia_gfid)) {
+                        /* ia_gfid set means, at least on one node 'mknod'
+                           is successful */
+                        local->call_count = priv->child_count;
+                        trav = this->children;
+                        while (trav) {
+                                STACK_WIND (frame,
+                                            stripe_mknod_ifreg_fail_unlink_cbk,
+                                            trav->xlator,
+                                            trav->xlator->fops->unlink,
+                                            &local->loc, 0, NULL);
+                                trav = trav->next;
+                        }
+                        return 0;
+                }
+
+
                 if (local->op_ret != -1) {
                         local->preparent.ia_blocks  = local->preparent_blocks;
                         local->preparent.ia_size    = local->preparent_size;
@@ -1360,73 +1535,109 @@ stripe_mknod_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->postparent.ia_size   = local->postparent_size;
                         local->stbuf.ia_size        = local->stbuf_size;
                         local->stbuf.ia_blocks      = local->stbuf_blocks;
+                        inode_ctx_put (local->inode, this,
+                                       (uint64_t)(long) local->fctx);
+
                 }
+                STRIPE_STACK_UNWIND (mknod, frame, local->op_ret, local->op_errno,
+                                     local->inode, &local->stbuf,
+                                     &local->preparent, &local->postparent, NULL);
+        }
+out:
+        return 0;
+}
 
-                if ((local->op_ret != -1) && priv->xattr_supported) {
-                        /* Send a setxattr request to nodes where the
-                           files are created */
-                        int32_t  i              = 0;
-                        char     size_key[256]  = {0,};
-                        char     index_key[256] = {0,};
-                        char     count_key[256] = {0,};
-                        dict_t  *dict           = NULL;
 
-                        sprintf (size_key,
-                                 "trusted.%s.stripe-size", this->name);
-                        sprintf (count_key,
-                                 "trusted.%s.stripe-count", this->name);
-                        sprintf (index_key,
-                                 "trusted.%s.stripe-index", this->name);
+int32_t
+stripe_mknod_first_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, inode_t *inode,
+                        struct iatt *buf, struct iatt *preparent,
+                        struct iatt *postparent, dict_t *xdata)
+{
+        stripe_local_t   *local = NULL;
+        stripe_private_t *priv = NULL;
+        call_frame_t     *prev = NULL;
+        xlator_list_t    *trav = NULL;
+        int               i    = 1;
+        dict_t           *dict           = NULL;
+        int               ret            = 0;
+        int               need_unref     = 0;
 
-                        local->call_count = priv->child_count;
-                        memcpy (local->loc.inode->gfid, local->stbuf.ia_gfid, 16);
-                        for (i = 0; i < priv->child_count; i++) {
-                                dict = get_new_dict ();
-                                if (!dict) {
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "failed to allocate dict");
-                                }
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
 
-                                dict_ref (dict);
-                                /* TODO: check return value */
-                                ret = dict_set_int64 (dict, size_key,
-                                                      local->stripe_size);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set stripe-size failed",
-                                                local->loc.path);
-                                ret = dict_set_int32 (dict, count_key,
-                                                      priv->child_count);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set child_count failed",
-                                                local->loc.path);
-                                ret = dict_set_int32 (dict, index_key, i);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set stripe-index failed",
-                                                local->loc.path);
+        prev  = cookie;
+        priv  = this->private;
+        local = frame->local;
+        trav = this->children;
 
-                                STACK_WIND (frame,
-                                            stripe_mknod_ifreg_setxattr_cbk,
-                                            priv->xl_array[i],
-                                            priv->xl_array[i]->fops->setxattr,
-                                            &local->loc, dict, 0);
+        local->call_count--;
 
-                                dict_unref (dict);
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "%s returned error %s",
+                        prev->this->name, strerror (op_errno));
+                local->failed = 1;
+                local->op_errno = op_errno;
+                goto out;
+        }
+
+        local->op_ret = op_ret;
+
+        local->stbuf      = *buf;
+        local->preparent  = *preparent;
+        local->postparent = *postparent;
+
+        if (uuid_is_null (local->ia_gfid))
+                uuid_copy (local->ia_gfid, buf->ia_gfid);
+        local->preparent.ia_blocks  = local->preparent_blocks;
+        local->preparent.ia_size    = local->preparent_size;
+        local->postparent.ia_blocks = local->postparent_blocks;
+        local->postparent.ia_size   = local->postparent_size;
+        local->stbuf.ia_size        = local->stbuf_size;
+        local->stbuf.ia_blocks      = local->stbuf_blocks;
+
+        trav = trav->next;
+        while (trav) {
+                if (priv->xattr_supported) {
+                        dict = dict_new ();
+                        if (!dict) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to allocate dict %s", local->loc.path);
                         }
-                        return 0;
+                        need_unref = 1;
+
+                        dict_copy (local->xattr, dict);
+
+                        ret = stripe_xattr_request_build (this, dict,
+                                                          local->stripe_size,
+                                                          priv->child_count, i,
+							  priv->coalesce);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to build xattr request");
+
+                } else {
+                        dict = local->xattr;
                 }
 
-                /* Create itself has failed.. so return
-                   without setxattring */
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (mknod, frame, local->op_ret, local->op_errno,
-                                     local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
+                STACK_WIND (frame, stripe_mknod_ifreg_cbk,
+                            trav->xlator, trav->xlator->fops->mknod,
+                            &local->loc, local->mode, local->rdev, 0, dict);
+                trav = trav->next;
+                i++;
+
+                if (dict && need_unref)
+                        dict_unref (dict);
         }
-out:
+
         return 0;
+
+out:
+
+       STRIPE_STACK_UNWIND (mknod, frame, op_ret, op_errno, NULL, NULL, NULL, NULL, NULL);
+       return 0;
 }
 
 
@@ -1434,22 +1645,25 @@ int32_t
 stripe_single_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int32_t op_ret, int32_t op_errno, inode_t *inode,
                          struct iatt *buf, struct iatt *preparent,
-                         struct iatt *postparent)
+                         struct iatt *postparent, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
+        STRIPE_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
         return 0;
 }
 
 
 int
 stripe_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-              dev_t rdev, dict_t *params)
+              dev_t rdev, mode_t umask, dict_t *xdata)
 {
-        stripe_private_t *priv = NULL;
-        stripe_local_t   *local = NULL;
-        xlator_list_t    *trav = NULL;
-        int32_t           op_errno = EINVAL;
+        stripe_private_t *priv           = NULL;
+        stripe_local_t   *local          = NULL;
+        int32_t           op_errno       = EINVAL;
+        int32_t           i              = 0;
+        dict_t           *dict           = NULL;
+        int               ret            = 0;
+        int               need_unref     = 0;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -1458,7 +1672,6 @@ stripe_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
         VALIDATE_OR_GOTO (loc->inode, err);
 
         priv = this->private;
-        trav = this->children;
 
         if (priv->first_child_down) {
                 op_errno = ENOTCONN;
@@ -1477,43 +1690,63 @@ stripe_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
                 }
 
                 /* Initialization */
-                local = GF_CALLOC (1, sizeof (stripe_local_t),
-                                   gf_stripe_mt_stripe_local_t);
+                local = mem_get0 (this->local_pool);
                 if (!local) {
                         op_errno = ENOMEM;
                         goto err;
                 }
                 local->op_ret = -1;
                 local->op_errno = ENOTCONN;
-                local->stripe_size = stripe_get_matching_bs (loc->path,
-                                                             priv->pattern,
-                                                             priv->block_size);
+                local->stripe_size = stripe_get_matching_bs (loc->path, priv);
                 frame->local = local;
-                local->inode = loc->inode;
+                local->inode = inode_ref (loc->inode);
                 loc_copy (&local->loc, loc);
+                local->xattr = dict_copy_with_ref (xdata, NULL);
+                local->mode = mode;
+                local->umask = umask;
+                local->rdev = rdev;
 
                 /* Everytime in stripe lookup, all child nodes should
                    be looked up */
                 local->call_count = priv->child_count;
 
-                while (trav) {
-                        STACK_WIND (frame, stripe_mknod_ifreg_cbk,
-                                    trav->xlator, trav->xlator->fops->mknod,
-                                    loc, mode, rdev, params);
-                        trav = trav->next;
+                if (priv->xattr_supported) {
+                        dict = dict_new ();
+                        if (!dict) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to allocate dict %s", loc->path);
+                        }
+                        need_unref = 1;
+
+                        dict_copy (xdata, dict);
+
+                        ret = stripe_xattr_request_build (this, dict,
+                                                          local->stripe_size,
+                                                          priv->child_count,
+                                                          i, priv->coalesce);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to build xattr request");
+                } else {
+                        dict = xdata;
                 }
 
-                /* This case is handled, no need to continue further. */
+                STACK_WIND (frame, stripe_mknod_first_ifreg_cbk,
+                            FIRST_CHILD (this), FIRST_CHILD (this)->fops->mknod,
+                            loc, mode, rdev, umask, dict);
+
+                        if (dict && need_unref)
+                                dict_unref (dict);
                 return 0;
         }
 
         STACK_WIND (frame, stripe_single_mknod_cbk,
                     FIRST_CHILD(this), FIRST_CHILD(this)->fops->mknod,
-                    loc, mode, rdev, params);
+                    loc, mode, rdev, umask, xdata);
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        STRIPE_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
         return 0;
 }
 
@@ -1522,11 +1755,10 @@ int32_t
 stripe_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno, inode_t *inode,
                   struct iatt *buf, struct iatt *preparent,
-                  struct iatt *postparent)
+                  struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t  *local   = NULL;
-        inode_t         *local_inode = NULL;
         call_frame_t    *prev = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
@@ -1554,12 +1786,6 @@ stripe_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (op_ret >= 0) {
                         local->op_ret = 0;
 
-                        if (FIRST_CHILD(this) == prev->this) {
-                                local->inode      = inode_ref (inode);
-                                local->stbuf      = *buf;
-                                local->postparent = *postparent;
-                                local->preparent  = *preparent;
-                        }
                         local->stbuf_blocks      += buf->ia_blocks;
                         local->preparent_blocks  += preparent->ia_blocks;
                         local->postparent_blocks += postparent->ia_blocks;
@@ -1575,12 +1801,7 @@ stripe_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
-
-                local_inode = local->inode;
-
-                if (local->op_ret != -1) {
+                if (local->failed != -1) {
                         local->preparent.ia_blocks  = local->preparent_blocks;
                         local->preparent.ia_size    = local->preparent_size;
                         local->postparent.ia_blocks = local->postparent_blocks;
@@ -1588,22 +1809,79 @@ stripe_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->stbuf.ia_size        = local->stbuf_size;
                         local->stbuf.ia_blocks      = local->stbuf_blocks;
                 }
-                STACK_UNWIND_STRICT (mkdir, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (mkdir, frame, local->op_ret,
                                      local->op_errno, local->inode,
                                      &local->stbuf, &local->preparent,
-                                     &local->postparent);
+                                     &local->postparent, NULL);
+        }
+out:
+        return 0;
+}
+
 
-                if (local_inode)
-                        inode_unref (local_inode);
+int32_t
+stripe_first_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, inode_t *inode,
+                        struct iatt *buf, struct iatt *preparent,
+                        struct iatt *postparent, dict_t *xdata)
+{
+        stripe_local_t  *local   = NULL;
+        call_frame_t    *prev = NULL;
+        xlator_list_t        *trav = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
         }
+
+        prev  = cookie;
+        local = frame->local;
+        trav = this->children;
+
+        local->call_count--; /* first child is successful */
+        trav = trav->next;   /* skip first child */
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "%s returned error %s",
+                        prev->this->name, strerror (op_errno));
+                local->op_errno = op_errno;
+                goto out;
+        }
+
+        local->op_ret = 0;
+
+        local->inode      = inode_ref (inode);
+        local->stbuf      = *buf;
+        local->postparent = *postparent;
+        local->preparent  = *preparent;
+
+        local->stbuf_blocks      += buf->ia_blocks;
+        local->preparent_blocks  += preparent->ia_blocks;
+        local->postparent_blocks += postparent->ia_blocks;
+
+        local->stbuf_size = buf->ia_size;
+        local->preparent_size = preparent->ia_size;
+        local->postparent_size = postparent->ia_size;
+
+        while (trav) {
+                STACK_WIND (frame, stripe_mkdir_cbk, trav->xlator,
+                            trav->xlator->fops->mkdir, &local->loc, local->mode,
+                            local->umask, local->xdata);
+                trav = trav->next;
+        }
+        return 0;
 out:
+        STRIPE_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
+                             NULL, NULL);
+
         return 0;
+
 }
 
 
 int
 stripe_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-              dict_t *params)
+              mode_t umask, dict_t *xdata)
 {
         stripe_private_t *priv = NULL;
         stripe_local_t   *local = NULL;
@@ -1625,27 +1903,27 @@ stripe_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         local->call_count = priv->child_count;
+        if (xdata)
+                local->xdata = dict_ref (xdata);
+        local->mode  = mode;
+        local->umask = umask;
+        loc_copy (&local->loc, loc);
         frame->local = local;
 
         /* Everytime in stripe lookup, all child nodes should be looked up */
-        while (trav) {
-                STACK_WIND (frame, stripe_mkdir_cbk,
-                            trav->xlator, trav->xlator->fops->mkdir,
-                            loc, mode, params);
-                trav = trav->next;
-        }
+        STACK_WIND (frame, stripe_first_mkdir_cbk, trav->xlator,
+                    trav->xlator->fops->mkdir, loc, mode, umask, xdata);
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        STRIPE_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
         return 0;
 }
 
@@ -1654,12 +1932,12 @@ int32_t
 stripe_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno, inode_t *inode,
                  struct iatt *buf, struct iatt *preparent,
-                 struct iatt *postparent)
+                 struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t  *local   = NULL;
-        inode_t         *local_inode = NULL;
         call_frame_t    *prev = NULL;
+	stripe_fd_ctx_t *fctx = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -1686,6 +1964,16 @@ stripe_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (op_ret >= 0) {
                         local->op_ret = 0;
 
+			if (IA_ISREG(inode->ia_type)) {
+				inode_ctx_get(inode, this, (uint64_t *) &fctx);
+				if (!fctx) {
+					gf_log(this->name, GF_LOG_ERROR,
+						"failed to get stripe context");
+					op_ret = -1;
+					op_errno = EINVAL;
+				}
+			}
+
                         if (FIRST_CHILD(this) == prev->this) {
                                 local->inode      = inode_ref (inode);
                                 local->stbuf      = *buf;
@@ -1696,6 +1984,8 @@ stripe_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->preparent_blocks  += preparent->ia_blocks;
                         local->postparent_blocks += postparent->ia_blocks;
 
+			correct_file_size(buf, fctx, prev);
+
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
                         if (local->preparent_size < preparent->ia_size)
@@ -1710,8 +2000,6 @@ stripe_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                local_inode = local->inode;
-
                 if (local->op_ret != -1) {
                         local->preparent.ia_blocks  = local->preparent_blocks;
                         local->preparent.ia_size    = local->preparent_size;
@@ -1720,20 +2008,17 @@ stripe_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->stbuf.ia_size        = local->stbuf_size;
                         local->stbuf.ia_blocks      = local->stbuf_blocks;
                 }
-                STACK_UNWIND_STRICT (link, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (link, frame, local->op_ret,
                                      local->op_errno, local->inode,
                                      &local->stbuf, &local->preparent,
-                                     &local->postparent);
-
-                if (local_inode)
-                        inode_unref (local_inode);
+                                     &local->postparent, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
+stripe_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
@@ -1756,8 +2041,7 @@ stripe_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -1771,13 +2055,13 @@ stripe_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
         while (trav) {
                 STACK_WIND (frame, stripe_link_cbk,
                             trav->xlator, trav->xlator->fops->link,
-                            oldloc, newloc);
+                            oldloc, newloc, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        STRIPE_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
         return 0;
 }
 
@@ -1785,12 +2069,10 @@ int32_t
 stripe_create_fail_unlink_cbk (call_frame_t *frame, void *cookie,
                                xlator_t *this, int32_t op_ret,
                                int32_t op_errno, struct iatt *preparent,
-                               struct iatt *postparent)
+                               struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
-        fd_t           *lfd = NULL;
         stripe_local_t *local = NULL;
-        inode_t        *local_inode = NULL;
 
         if (!this || !frame || !frame->local) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -1806,18 +2088,9 @@ stripe_create_fail_unlink_cbk (call_frame_t *frame, void *cookie,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                local_inode = local->inode;
-                lfd = local->fd;
-
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (create, frame, local->op_ret, local->op_errno,
+                STRIPE_STACK_UNWIND (create, frame, local->op_ret, local->op_errno,
                                      local->fd, local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
-
-                if (local_inode)
-                        inode_unref (local_inode);
-                if (lfd)
-                        fd_unref (lfd);
+                                     &local->preparent, &local->postparent, NULL);
         }
 out:
         return 0;
@@ -1825,16 +2098,16 @@ out:
 
 
 int32_t
-stripe_create_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                            int32_t op_ret, int32_t op_errno)
+stripe_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, fd_t *fd,
+                   inode_t *inode, struct iatt *buf, struct iatt *preparent,
+                   struct iatt *postparent, dict_t *xdata)
 {
-        inode_t          *local_inode = NULL;
-        fd_t             *lfd = NULL;
+        int32_t           callcnt = 0;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
+        call_frame_t     *prev = NULL;
         xlator_list_t    *trav = NULL;
-        int32_t           callcnt = 0;
-        call_frame_t   *prev = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -1842,7 +2115,7 @@ stripe_create_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         }
 
         prev  = cookie;
-	priv  = this->private;
+        priv  = this->private;
         local = frame->local;
 
         LOCK (&frame->lock);
@@ -1853,13 +2126,40 @@ stripe_create_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "%s returned error %s",
                                 prev->this->name, strerror (op_errno));
-                        local->op_ret = -1;
+                        local->failed = 1;
                         local->op_errno = op_errno;
                 }
+
+                if (op_ret >= 0) {
+			if (IA_ISREG(buf->ia_type)) {
+				if (stripe_ctx_handle(this, prev, local, xdata))
+					gf_log(this->name, GF_LOG_ERROR,
+						"Error getting fctx info from "
+						"dict");
+			}
+
+                        local->op_ret = op_ret;
+
+                        local->stbuf_blocks += buf->ia_blocks;
+                        local->preparent_blocks  += preparent->ia_blocks;
+                        local->postparent_blocks += postparent->ia_blocks;
+
+			correct_file_size(buf, local->fctx, prev);
+
+                        if (local->stbuf_size < buf->ia_size)
+                                local->stbuf_size = buf->ia_size;
+                        if (local->preparent_size < preparent->ia_size)
+                                local->preparent_size = preparent->ia_size;
+                        if (local->postparent_size < postparent->ia_size)
+                                local->postparent_size = postparent->ia_size;
+                }
         }
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
+                if (local->failed)
+                        local->op_ret = -1;
+
                 if (local->op_ret == -1) {
                         local->call_count = priv->child_count;
                         trav = this->children;
@@ -1868,44 +2168,57 @@ stripe_create_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                             stripe_create_fail_unlink_cbk,
                                             trav->xlator,
                                             trav->xlator->fops->unlink,
-                                            &local->loc);
+                                            &local->loc, 0, NULL);
                                 trav = trav->next;
                         }
 
                         return 0;
                 }
 
-                lfd = local->fd;
-                local_inode = local->inode;
+                if (local->op_ret >= 0) {
+                        local->preparent.ia_blocks  = local->preparent_blocks;
+                        local->preparent.ia_size    = local->preparent_size;
+                        local->postparent.ia_blocks = local->postparent_blocks;
+                        local->postparent.ia_size   = local->postparent_size;
+                        local->stbuf.ia_size        = local->stbuf_size;
+                        local->stbuf.ia_blocks      = local->stbuf_blocks;
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (create, frame, local->op_ret, local->op_errno,
-                                     local->fd, local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
+			stripe_copy_xl_array(local->fctx->xl_array,
+				             priv->xl_array,
+					     local->fctx->stripe_count);
+			inode_ctx_put(local->inode, this,
+					(uint64_t) local->fctx);
+                }
 
-                if (local_inode)
-                        inode_unref (local_inode);
-                if (lfd)
-                        fd_unref (lfd);
+                /* Create itself has failed.. so return
+                   without setxattring */
+                STRIPE_STACK_UNWIND (create, frame, local->op_ret,
+                                     local->op_errno, local->fd,
+                                     local->inode, &local->stbuf,
+                                     &local->preparent, &local->postparent, NULL);
         }
+
 out:
         return 0;
 }
 
+
+
 int32_t
-stripe_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+stripe_first_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, fd_t *fd,
                    inode_t *inode, struct iatt *buf, struct iatt *preparent,
-                   struct iatt *postparent)
+                   struct iatt *postparent, dict_t *xdata)
 {
-        int32_t           callcnt = 0;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
-        fd_t             *lfd = NULL;
-        stripe_fd_ctx_t  *fctx = NULL;
-        inode_t          *local_inode = NULL;
         call_frame_t     *prev = NULL;
-        int               ret = 0;
+        xlator_list_t    *trav = NULL;
+        int               i    = 1;
+        dict_t           *dict = NULL;
+        loc_t            *loc  = NULL;
+        int32_t           need_unref = 0;
+        int32_t           ret  = -1;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -1915,150 +2228,89 @@ stripe_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         prev  = cookie;
         priv  = this->private;
         local = frame->local;
+        trav = this->children;
+        loc = &local->loc;
 
-        LOCK (&frame->lock);
-        {
-                callcnt = --local->call_count;
-
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "%s returned error %s",
-                                prev->this->name, strerror (op_errno));
-                        if ((op_errno != ENOENT) ||
-                            (prev->this == FIRST_CHILD (this)))
-                                local->failed = 1;
-                        local->op_errno = op_errno;
-                }
-
-                if (op_ret >= 0) {
-                        local->op_ret = op_ret;
-                        /* Get the mapping in inode private */
-                        /* Get the stat buf right */
-                        if (FIRST_CHILD(this) == prev->this) {
-                                local->stbuf      = *buf;
-                                local->preparent  = *preparent;
-                                local->postparent = *postparent;
-                        }
-
-                        local->stbuf_blocks += buf->ia_blocks;
-                        local->preparent_blocks  += preparent->ia_blocks;
-                        local->postparent_blocks += postparent->ia_blocks;
+        --local->call_count;
 
-                        if (local->stbuf_size < buf->ia_size)
-                                local->stbuf_size = buf->ia_size;
-                        if (local->preparent_size < preparent->ia_size)
-                                local->preparent_size = preparent->ia_size;
-                        if (local->postparent_size < postparent->ia_size)
-                                local->postparent_size = postparent->ia_size;
-                }
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "%s returned error %s",
+                        prev->this->name, strerror (op_errno));
+                 local->failed = 1;
+                 local->op_errno = op_errno;
         }
-        UNLOCK (&frame->lock);
 
-        if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
-
-                if (local->op_ret != -1) {
-                        local->preparent.ia_blocks  = local->preparent_blocks;
-                        local->preparent.ia_size    = local->preparent_size;
-                        local->postparent.ia_blocks = local->postparent_blocks;
-                        local->postparent.ia_size   = local->postparent_size;
-                        local->stbuf.ia_size        = local->stbuf_size;
-                        local->stbuf.ia_blocks      = local->stbuf_blocks;
-                }
-
-                /* */
-                if (local->op_ret >= 0) {
-                        fctx = GF_CALLOC (1, sizeof (stripe_fd_ctx_t),
-                                          gf_stripe_mt_stripe_fd_ctx_t);
-                        if (!fctx) {
-                                local->op_ret = -1;
-                                local->op_errno = ENOMEM;
-                                goto unwind;
-                        }
-
-                        fctx->stripe_size  = local->stripe_size;
-                        fctx->stripe_count = priv->child_count;
-                        fctx->static_array = 1;
-                        fctx->xl_array = priv->xl_array;
-                        fd_ctx_set (local->fd, this,
-                                    (uint64_t)(long)fctx);
-                }
-
-                if ((local->op_ret != -1) &&
-                    local->stripe_size && priv->xattr_supported) {
-                        /* Send a setxattr request to nodes where
-                           the files are created */
-                        int32_t        i = 0;
-                        char           size_key[256] = {0,};
-                        char           index_key[256] = {0,};
-                        char           count_key[256] = {0,};
-                        dict_t        *dict = NULL;
-
-                        sprintf (size_key,
-                                 "trusted.%s.stripe-size", this->name);
-                        sprintf (count_key,
-                                 "trusted.%s.stripe-count", this->name);
-                        sprintf (index_key,
-                                 "trusted.%s.stripe-index", this->name);
+        local->op_ret = 0;
+        /* Get the mapping in inode private */
+        /* Get the stat buf right */
+        local->stbuf      = *buf;
+        local->preparent  = *preparent;
+        local->postparent = *postparent;
 
-                        local->call_count = priv->child_count;
-                        memcpy (local->loc.inode->gfid, local->stbuf.ia_gfid, 16);
-                        for (i = 0; i < priv->child_count; i++) {
-                                dict = get_new_dict ();
-                                if (!dict) {
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "error allocating dict");
-                                }
-                                dict_ref (dict);
+        local->stbuf_blocks += buf->ia_blocks;
+        local->preparent_blocks  += preparent->ia_blocks;
+        local->postparent_blocks += postparent->ia_blocks;
 
-                                /* TODO: check return values */
-                                ret = dict_set_int64 (dict, size_key,
-                                                      local->stripe_size);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set stripe-size failed",
-                                                local->loc.path);
+        if (local->stbuf_size < buf->ia_size)
+              local->stbuf_size = buf->ia_size;
+        if (local->preparent_size < preparent->ia_size)
+              local->preparent_size = preparent->ia_size;
+        if (local->postparent_size < postparent->ia_size)
+              local->postparent_size = postparent->ia_size;
 
-                                ret = dict_set_int32 (dict, count_key,
-                                                      priv->child_count);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set stripe-size failed",
-                                                local->loc.path);
+        if (local->failed)
+                local->op_ret = -1;
 
-                                ret = dict_set_int32 (dict, index_key, i);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set stripe-size failed",
-                                                local->loc.path);
+        if (local->op_ret == -1) {
+                local->call_count = 1;
+                STACK_WIND (frame, stripe_create_fail_unlink_cbk,
+                            FIRST_CHILD (this), FIRST_CHILD (this)->fops->unlink,
+                            &local->loc, 0, NULL);
+                return 0;
+        }
 
-                                STACK_WIND (frame, stripe_create_setxattr_cbk,
-                                            priv->xl_array[i],
-                                            priv->xl_array[i]->fops->setxattr,
-                                            &local->loc, dict, 0);
+        if (local->op_ret >= 0) {
+                local->preparent.ia_blocks  = local->preparent_blocks;
+                local->preparent.ia_size    = local->preparent_size;
+                local->postparent.ia_blocks = local->postparent_blocks;
+                local->postparent.ia_size   = local->postparent_size;
+                local->stbuf.ia_size        = local->stbuf_size;
+                local->stbuf.ia_blocks      = local->stbuf_blocks;
+        }
 
-                                dict_unref (dict);
+        /* Send a setxattr request to nodes where the
+           files are created */
+        trav = trav->next;
+        while (trav) {
+                if (priv->xattr_supported) {
+                        dict = dict_new ();
+                        if (!dict) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to allocate dict %s", loc->path);
                         }
-                        return 0;
-                }
+                        need_unref = 1;
 
-unwind:
-                /* Create itself has failed.. so return
-                   without setxattring */
-                lfd = local->fd;
-                local_inode = local->inode;
+                        dict_copy (local->xattr, dict);
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (create, frame, local->op_ret,
-                                     local->op_errno, local->fd,
-                                     local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
+                        ret = stripe_xattr_request_build (this, dict,
+                                                          local->stripe_size,
+                                                          priv->child_count,
+                                                          i, priv->coalesce);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to build xattr request");
+                } else {
+                        dict = local->xattr;
+                }
 
-                if (local_inode)
-                        inode_unref (local_inode);
-                if (lfd)
-                        fd_unref (lfd);
+                STACK_WIND (frame, stripe_create_cbk, trav->xlator,
+                            trav->xlator->fops->create, &local->loc,
+                            local->flags, local->mode, local->umask, local->fd,
+                            dict);
+                trav = trav->next;
+                if (need_unref && dict)
+                        dict_unref (dict);
+                i++;
         }
 
 out:
@@ -2066,6 +2318,7 @@ out:
 }
 
 
+
 /**
  * stripe_create - If a block-size is specified for the 'name', create the
  *    file in all the child nodes. If not, create it in only first child.
@@ -2074,12 +2327,15 @@ out:
  */
 int32_t
 stripe_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
-               int32_t flags, mode_t mode, fd_t *fd, dict_t *params)
+               int32_t flags, mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
 {
         stripe_private_t *priv = NULL;
         stripe_local_t   *local = NULL;
-        xlator_list_t    *trav = NULL;
         int32_t           op_errno = EINVAL;
+        int               ret            = 0;
+        int               need_unref     = 0;
+        int               i              = 0;
+        dict_t           *dict           = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -2100,46 +2356,71 @@ stripe_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         local->op_errno = ENOTCONN;
-        local->stripe_size = stripe_get_matching_bs (loc->path,
-                                                     priv->pattern,
-                                                     priv->block_size);
+        local->stripe_size = stripe_get_matching_bs (loc->path, priv);
         frame->local = local;
         local->inode = inode_ref (loc->inode);
         loc_copy (&local->loc, loc);
         local->fd = fd_ref (fd);
+        local->flags = flags;
+        local->mode = mode;
+        local->umask = umask;
+        if (xdata)
+                local->xattr = dict_ref (xdata);
 
         local->call_count = priv->child_count;
+        /* Send a setxattr request to nodes where the
+           files are created */
 
-        trav = this->children;
-        while (trav) {
-                STACK_WIND (frame, stripe_create_cbk, trav->xlator,
-                            trav->xlator->fops->create, loc, flags,
-                            mode, fd, params);
-                trav = trav->next;
+        if (priv->xattr_supported) {
+                dict = dict_new ();
+                if (!dict) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to allocate dict %s", loc->path);
+                }
+                need_unref = 1;
+
+                dict_copy (xdata, dict);
+
+                ret = stripe_xattr_request_build (this, dict,
+                                                  local->stripe_size,
+                                                  priv->child_count,
+                                                  i, priv->coalesce);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to build xattr request");
+        } else {
+                        dict = xdata;
         }
 
+
+        STACK_WIND (frame, stripe_first_create_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->create, loc, flags, mode,
+                    umask, fd, dict);
+
+        if (need_unref && dict)
+                dict_unref (dict);
+
+
         return 0;
 err:
-        STACK_UNWIND_STRICT (create, frame, -1, op_errno, NULL, NULL, NULL,
-                             NULL, NULL);
+        STRIPE_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
+                             NULL, NULL, xdata);
         return 0;
 }
 
 int32_t
 stripe_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int32_t op_ret, int32_t op_errno, fd_t *fd)
+                 int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
-        fd_t           *lfd = NULL;
         call_frame_t   *prev = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
@@ -2174,228 +2455,20 @@ stripe_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                if (local->op_ret == -1) {
-                        if (local->fctx) {
-                                if (!local->fctx->static_array)
-                                        GF_FREE (local->fctx->xl_array);
-                                GF_FREE (local->fctx);
-                        }
-                } else {
-                        fd_ctx_set (local->fd, this,
-                                    (uint64_t)(long)local->fctx);
-                }
-
-                lfd = local->fd;
-
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (open, frame, local->op_ret,
-                                     local->op_errno, local->fd);
-                if (lfd)
-                        fd_unref (lfd);
-
+                STRIPE_STACK_UNWIND (open, frame, local->op_ret,
+                                     local->op_errno, local->fd, xdata);
         }
 out:
         return 0;
 }
 
 
-int32_t
-stripe_open_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                          int32_t op_ret, int32_t op_errno, dict_t *dict)
-{
-        int32_t           index = 0;
-        int32_t           callcnt = 0;
-        char              key[256] = {0,};
-        stripe_local_t   *local = NULL;
-        xlator_list_t    *trav = NULL;
-        stripe_private_t *priv = NULL;
-        data_t           *data = NULL;
-        call_frame_t     *prev = NULL;
-        fd_t             *lfd = NULL;
-
-        if (!this || !frame || !frame->local || !cookie) {
-                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
-                goto out;
-        }
-
-        prev  = (call_frame_t *)cookie;
-        priv  = this->private;
-        local = frame->local;
-
-        LOCK (&frame->lock);
-        {
-                callcnt = --local->call_count;
-
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "%s returned error %s",
-                                prev->this->name, strerror (op_errno));
-                        local->op_ret = -1;
-                        if (local->op_errno != EIO)
-                                local->op_errno = op_errno;
-                        if ((op_errno != ENOENT) ||
-                            (prev->this == FIRST_CHILD (this)))
-                                local->failed = 1;
-                        goto unlock;
-                }
-
-                if (!dict)
-                        goto unlock;
-
-                if (!local->fctx) {
-                        local->fctx =  GF_CALLOC (1, sizeof (stripe_fd_ctx_t),
-                                                  gf_stripe_mt_stripe_fd_ctx_t);
-                        if (!local->fctx) {
-                                local->op_errno = ENOMEM;
-                                local->op_ret = -1;
-                                goto unlock;
-                        }
-
-                        local->fctx->static_array = 0;
-                }
-                /* Stripe block size */
-                sprintf (key, "trusted.%s.stripe-size", this->name);
-                data = dict_get (dict, key);
-                if (!data) {
-                        local->xattr_self_heal_needed = 1;
-                } else {
-                        if (!local->fctx->stripe_size) {
-                                local->fctx->stripe_size =
-                                        data_to_int64 (data);
-                        }
-
-                        if (local->fctx->stripe_size != data_to_int64 (data)) {
-                                gf_log (this->name, GF_LOG_WARNING,
-                                        "stripe-size mismatch in blocks");
-                                local->xattr_self_heal_needed = 1;
-                        }
-                }
-                /* Stripe count */
-                sprintf (key, "trusted.%s.stripe-count", this->name);
-                data = dict_get (dict, key);
-                if (!data) {
-                        local->xattr_self_heal_needed = 1;
-                        goto unlock;
-                }
-                if (!local->fctx->xl_array) {
-                        local->fctx->stripe_count = data_to_int32 (data);
-                        if (!local->fctx->stripe_count) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "error with stripe-count xattr");
-                                local->op_ret   = -1;
-                                local->op_errno = EIO;
-                                goto unlock;
-                        }
-
-                        local->fctx->xl_array =
-                                GF_CALLOC (local->fctx->stripe_count,
-                                           sizeof (xlator_t *),
-                                           gf_stripe_mt_xlator_t);
-                        if (!local->fctx->xl_array) {
-                                local->op_errno = ENOMEM;
-                                local->op_ret   = -1;
-                                goto unlock;
-                        }
-                }
-                if (local->fctx->stripe_count != data_to_int32 (data)) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "error with stripe-count xattr (%d != %d)",
-                                local->fctx->stripe_count, data_to_int32 (data));
-                        local->op_ret   = -1;
-                        local->op_errno = EIO;
-                        goto unlock;
-                }
-
-                /* index */
-                sprintf (key, "trusted.%s.stripe-index", this->name);
-                data = dict_get (dict, key);
-                if (!data) {
-                        local->xattr_self_heal_needed = 1;
-                        goto unlock;
-                }
-                index = data_to_int32 (data);
-                if (index > priv->child_count) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "error with stripe-index xattr (%d)", index);
-                        local->op_ret   = -1;
-                        local->op_errno = EIO;
-                        goto unlock;
-                }
-                if (local->fctx->xl_array) {
-                        if (local->fctx->xl_array[index]) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "duplicate entry @ index (%d)", index);
-                                local->op_ret   = -1;
-                                local->op_errno = EIO;
-                                goto unlock;
-                        }
-                        local->fctx->xl_array[index] = prev->this;
-                }
-                local->entry_count++;
-                local->op_ret = 0;
-        }
-unlock:
-        UNLOCK (&frame->lock);
-
-        if (!callcnt) {
-                /* TODO: if self-heal flag is set, do it */
-                if (local->xattr_self_heal_needed) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "%s: stripe info need to be healed",
-                                local->loc.path);
-                }
-
-                if (local->failed)
-                        local->op_ret = -1;
-
-                if (local->op_ret)
-                        goto err;
-
-                if (local->entry_count != local->fctx->stripe_count) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "entry-count (%d) != stripe-count (%d)",
-                                local->entry_count, local->fctx->stripe_count);
-                        local->op_ret = -1;
-                        local->op_errno = EIO;
-                        goto err;
-                }
-                if (!local->fctx->stripe_size) {
-                        gf_log (this->name, GF_LOG_ERROR, "stripe size not set");
-                        local->op_ret = -1;
-                        local->op_errno = EIO;
-                        goto err;
-                }
-
-                local->call_count = local->fctx->stripe_count;
-
-                trav = this->children;
-                while (trav) {
-                        STACK_WIND (frame, stripe_open_cbk, trav->xlator,
-                                    trav->xlator->fops->open, &local->loc,
-                                    local->flags, local->fd, 0);
-                        trav = trav->next;
-                }
-        }
-
-        return 0;
-err:
-        lfd = local->fd;
-
-        stripe_local_wipe (local);
-        STACK_UNWIND_STRICT (open, frame, local->op_ret, local->op_errno,
-                             local->fd);
-        if (lfd)
-                fd_unref (lfd);
-out:
-        return 0;
-}
-
 /**
  * stripe_open -
  */
 int32_t
 stripe_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
-             int32_t flags, fd_t *fd, int32_t wbflags)
+             int32_t flags, fd_t *fd, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
@@ -2417,8 +2490,7 @@ stripe_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2434,52 +2506,28 @@ stripe_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
         /* Striped files */
         local->flags = flags;
         local->call_count = priv->child_count;
-        local->stripe_size = stripe_get_matching_bs (loc->path,
-                                                     priv->pattern,
-                                                     priv->block_size);
-
-        if (priv->xattr_supported) {
-                while (trav) {
-                        STACK_WIND (frame, stripe_open_getxattr_cbk,
-                                    trav->xlator, trav->xlator->fops->getxattr,
-                                    loc, NULL);
-                        trav = trav->next;
-                }
-                return 0;
-        }
-        local->fctx =  GF_CALLOC (1, sizeof (stripe_fd_ctx_t),
-                                  gf_stripe_mt_stripe_fd_ctx_t);
-        if (!local->fctx) {
-                op_errno = ENOMEM;
-                goto err;
-        }
-
-        local->fctx->static_array = 1;
-        local->fctx->stripe_size  = local->stripe_size;
-        local->fctx->stripe_count = priv->child_count;
-        local->fctx->xl_array     = priv->xl_array;
+        local->stripe_size = stripe_get_matching_bs (loc->path, priv);
 
         while (trav) {
                 STACK_WIND (frame, stripe_open_cbk, trav->xlator,
                             trav->xlator->fops->open,
                             &local->loc, local->flags, local->fd,
-                            wbflags);
+                            xdata);
                 trav = trav->next;
         }
         return 0;
 err:
-        STACK_UNWIND_STRICT (open, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (open, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, fd_t *fd)
+                    int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
-        fd_t           *local_fd = NULL;
         call_frame_t   *prev = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
@@ -2508,11 +2556,8 @@ stripe_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                local_fd = local->fd;
-                STACK_UNWIND_STRICT (opendir, frame, local->op_ret,
-                                     local->op_errno, local->fd);
-                if (local_fd)
-                        fd_unref (local_fd);
+                STRIPE_STACK_UNWIND (opendir, frame, local->op_ret,
+                                     local->op_errno, local->fd, NULL);
         }
 out:
         return 0;
@@ -2520,7 +2565,7 @@ out:
 
 
 int32_t
-stripe_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
+stripe_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
@@ -2542,8 +2587,7 @@ stripe_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2554,19 +2598,19 @@ stripe_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
 
         while (trav) {
                 STACK_WIND (frame, stripe_opendir_cbk, trav->xlator,
-                            trav->xlator->fops->opendir, loc, fd);
+                            trav->xlator->fops->opendir, loc, fd, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (opendir, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (opendir, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 int32_t
 stripe_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-               int32_t op_ret, int32_t op_errno, struct flock *lock)
+               int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -2605,8 +2649,8 @@ stripe_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         if (!callcnt) {
                 if (local->failed)
                         local->op_ret = -1;
-                STACK_UNWIND_STRICT (lk, frame, local->op_ret,
-                                     local->op_errno, &local->lock);
+                STRIPE_STACK_UNWIND (lk, frame, local->op_ret,
+                                     local->op_errno, &local->lock, NULL);
         }
 out:
         return 0;
@@ -2614,7 +2658,7 @@ out:
 
 int32_t
 stripe_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
-           struct flock *lock)
+           struct gf_flock *lock, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         xlator_list_t    *trav = NULL;
@@ -2630,8 +2674,7 @@ stripe_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
         priv = this->private;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2642,20 +2685,20 @@ stripe_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
 
         while (trav) {
                 STACK_WIND (frame, stripe_lk_cbk, trav->xlator,
-                            trav->xlator->fops->lk, fd, cmd, lock);
+                            trav->xlator->fops->lk, fd, cmd, lock, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (lk, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno)
+                  int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local   = NULL;
@@ -2691,16 +2734,15 @@ stripe_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (flush, frame, local->op_ret,
-                                     local->op_errno);
+                STRIPE_STACK_UNWIND (flush, frame, local->op_ret,
+                                     local->op_errno, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
+stripe_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
@@ -2720,8 +2762,7 @@ stripe_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
                 goto err;
         }
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2732,13 +2773,13 @@ stripe_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
 
         while (trav) {
                 STACK_WIND (frame, stripe_flush_cbk,  trav->xlator,
-                            trav->xlator->fops->flush, fd);
+                            trav->xlator->fops->flush, fd, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (flush, frame, -1, op_errno);
+        STRIPE_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
         return 0;
 }
 
@@ -2747,7 +2788,7 @@ err:
 int32_t
 stripe_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                  struct iatt *postbuf)
+                  struct iatt *postbuf, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local   = NULL;
@@ -2783,6 +2824,9 @@ stripe_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->prebuf_blocks  += prebuf->ia_blocks;
                         local->postbuf_blocks += postbuf->ia_blocks;
 
+			correct_file_size(prebuf, local->fctx, prev);
+			correct_file_size(postbuf, local->fctx, prev);
+
                         if (local->prebuf_size < prebuf->ia_size)
                                 local->prebuf_size = prebuf->ia_size;
 
@@ -2803,21 +2847,21 @@ stripe_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->post_buf.ia_size   = local->postbuf_size;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (fsync, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (fsync, frame, local->op_ret,
                                      local->op_errno, &local->pre_buf,
-                                     &local->post_buf);
+                                     &local->post_buf, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
+stripe_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
         xlator_list_t    *trav = NULL;
+	stripe_fd_ctx_t  *fctx = NULL;
         int32_t           op_errno = 1;
 
         VALIDATE_OR_GOTO (frame, err);
@@ -2829,31 +2873,38 @@ stripe_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
         trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
+
+	inode_ctx_get(fd->inode, this, (uint64_t *) &fctx);
+	if (!fctx) {
+		op_errno = EINVAL;
+		goto err;
+	}
+	local->fctx = fctx;
+
         local->op_ret = -1;
         frame->local = local;
         local->call_count = priv->child_count;
 
         while (trav) {
                 STACK_WIND (frame, stripe_fsync_cbk, trav->xlator,
-                            trav->xlator->fops->fsync, fd, flags);
+                            trav->xlator->fops->fsync, fd, flags, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (fsync, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 int32_t
 stripe_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                  int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -2888,6 +2939,9 @@ stripe_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 local->stbuf = *buf;
 
                         local->stbuf_blocks += buf->ia_blocks;
+
+			correct_file_size(buf, local->fctx, prev);
+
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
                 }
@@ -2903,9 +2957,8 @@ stripe_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->stbuf.ia_blocks = local->stbuf_blocks;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (fstat, frame, local->op_ret,
-                                     local->op_errno, &local->stbuf);
+                STRIPE_STACK_UNWIND (fstat, frame, local->op_ret,
+                                     local->op_errno, &local->stbuf, NULL);
         }
 
 out:
@@ -2915,11 +2968,12 @@ out:
 int32_t
 stripe_fstat (call_frame_t *frame,
               xlator_t *this,
-              fd_t *fd)
+              fd_t *fd, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
         xlator_list_t    *trav = NULL;
+	stripe_fd_ctx_t  *fctx = NULL;
         int32_t           op_errno = 1;
 
         VALIDATE_OR_GOTO (frame, err);
@@ -2931,8 +2985,7 @@ stripe_fstat (call_frame_t *frame,
         trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2941,26 +2994,35 @@ stripe_fstat (call_frame_t *frame,
         frame->local = local;
         local->call_count = priv->child_count;
 
+	if (IA_ISREG(fd->inode->ia_type)) {
+		inode_ctx_get(fd->inode, this, (uint64_t *) &fctx);
+		if (!fctx)
+			goto err;
+		local->fctx = fctx;
+	}
+
         while (trav) {
                 STACK_WIND (frame, stripe_fstat_cbk, trav->xlator,
-                            trav->xlator->fops->fstat, fd);
+                            trav->xlator->fops->fstat, fd, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (fstat, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 
 int32_t
-stripe_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
+stripe_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
-        xlator_list_t    *trav = NULL;
-        int32_t           op_errno = 1;
+	stripe_fd_ctx_t  *fctx = NULL;
+	int		  i, eof_idx;
+	off_t		  dest_offset, tmp_offset;
+        int32_t		  op_errno = 1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -2968,11 +3030,9 @@ stripe_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
         VALIDATE_OR_GOTO (fd->inode, err);
 
         priv = this->private;
-        trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2981,22 +3041,60 @@ stripe_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
         frame->local = local;
         local->call_count = priv->child_count;
 
-        while (trav) {
-                STACK_WIND (frame, stripe_truncate_cbk, trav->xlator,
-                            trav->xlator->fops->ftruncate, fd, offset);
-                trav = trav->next;
-        }
+	inode_ctx_get(fd->inode, this, (uint64_t *) &fctx);
+	if (!fctx) {
+		gf_log(this->name, GF_LOG_ERROR, "no stripe context");
+		op_errno = EINVAL;
+		goto err;
+	}
+	if (!fctx->stripe_count) {
+		gf_log(this->name, GF_LOG_ERROR, "no stripe count");
+		op_errno = EINVAL;
+		goto err;
+	}
+
+	local->fctx = fctx;
+	eof_idx = (offset / fctx->stripe_size) % fctx->stripe_count;
+
+	for (i = 0; i < fctx->stripe_count; i++) {
+		if (!fctx->xl_array[i]) {
+			gf_log(this->name, GF_LOG_ERROR, "no xlator at index "
+				"%d", i);
+			op_errno = EINVAL;
+			goto err;
+		}
+
+		if (fctx->stripe_coalesce) {
+			if (i < eof_idx)
+				tmp_offset = roof(offset, fctx->stripe_size *
+						fctx->stripe_count);
+			else if (i > eof_idx)
+				tmp_offset = floor(offset, fctx->stripe_size *
+						fctx->stripe_count);
+			else
+				tmp_offset = offset;
+
+			dest_offset = coalesced_offset(tmp_offset,
+				fctx->stripe_size, fctx->stripe_count);
+		} else {
+			dest_offset = offset;
+		}
+
+		STACK_WIND(frame, stripe_truncate_cbk, fctx->xl_array[i],
+			fctx->xl_array[i]->fops->ftruncate, fd, dest_offset,
+			NULL);
+	}
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (ftruncate, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno)
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local   = NULL;
@@ -3032,16 +3130,15 @@ stripe_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (fsyncdir, frame, local->op_ret,
-                                     local->op_errno);
+                STRIPE_STACK_UNWIND (fsyncdir, frame, local->op_ret,
+                                     local->op_errno, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
+stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
@@ -3057,8 +3154,7 @@ stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
         trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -3069,20 +3165,20 @@ stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
 
         while (trav) {
                 STACK_WIND (frame, stripe_fsyncdir_cbk, trav->xlator,
-                            trav->xlator->fops->fsyncdir, fd, flags);
+                            trav->xlator->fops->fsyncdir, fd, flags, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (fsyncdir, frame, -1, op_errno);
+        STRIPE_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                        int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
         int32_t         i = 0;
         int32_t         callcnt = 0;
@@ -3092,6 +3188,7 @@ stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         struct iatt     tmp_stbuf = {0,};
         struct iobref  *tmp_iobref = NULL;
         struct iobuf   *iobuf = NULL;
+	call_frame_t   *prev = NULL;
 
         if (!this || !frame || !frame->local) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -3099,13 +3196,16 @@ stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         }
 
         local = frame->local;
+	prev = cookie;
 
         LOCK (&frame->lock);
         {
                 callcnt = --local->call_count;
-                if (op_ret != -1)
+                if (op_ret != -1) {
+			correct_file_size(buf, local->fctx, prev);
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
+		}
         }
         UNLOCK (&frame->lock);
 
@@ -3134,7 +3234,8 @@ stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 vec[count].iov_len  =
                                         (local->replies[i].requested_size -
                                          local->replies[i].op_ret);
-                                iobuf = iobuf_get (this->ctx->iobuf_pool);
+                                iobuf = iobuf_get2 (this->ctx->iobuf_pool,
+                                                    vec[count].iov_len);
                                 if (!iobuf) {
                                         gf_log (this->name, GF_LOG_ERROR,
                                                 "Out of memory.");
@@ -3143,9 +3244,11 @@ stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                         goto done;
                                 }
                                 memset (iobuf->ptr, 0, vec[count].iov_len);
-                                iobref_add (local->iobref, iobuf);
                                 vec[count].iov_base = iobuf->ptr;
 
+                                iobref_add (local->iobref, iobuf);
+                                iobuf_unref(iobuf);
+
                                 op_ret += vec[count].iov_len;
                                 count++;
                         }
@@ -3162,13 +3265,11 @@ stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         done:
                 GF_FREE (local->replies);
                 tmp_iobref = local->iobref;
-                fd_unref (local->fd);
-                STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vec,
-                                     count, &tmp_stbuf, tmp_iobref);
+                STRIPE_STACK_UNWIND (readv, frame, op_ret, op_errno, vec,
+                                     count, &tmp_stbuf, tmp_iobref, NULL);
 
                 iobref_unref (tmp_iobref);
-                if (vec)
-                        GF_FREE (vec);
+                GF_FREE (vec);
         }
 out:
         return 0;
@@ -3181,7 +3282,7 @@ out:
 int32_t
 stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno, struct iovec *vector,
-                  int32_t count, struct iatt *stbuf, struct iobref *iobref)
+                  int32_t count, struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
 {
         int32_t         index = 0;
         int32_t         callcnt = 0;
@@ -3192,8 +3293,10 @@ stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         stripe_local_t *local = NULL;
         struct iovec   *final_vec = NULL;
         struct iatt     tmp_stbuf = {0,};
+        struct iatt    *tmp_stbuf_p = NULL; //need it for a warning
         struct iobref  *tmp_iobref = NULL;
         stripe_fd_ctx_t  *fctx = NULL;
+	call_frame_t	*prev = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -3202,6 +3305,7 @@ stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         local  = frame->local;
         index  = local->node_index;
+	prev = cookie;
         mframe = local->orig_frame;
         if (!mframe)
                 goto out;
@@ -3222,6 +3326,12 @@ stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         mlocal->replies[index].count  = count;
                         mlocal->replies[index].vector = iov_dup (vector, count);
 
+			correct_file_size(stbuf, fctx, prev);
+
+                        if (local->stbuf_size < stbuf->ia_size)
+                                local->stbuf_size = stbuf->ia_size;
+                        local->stbuf_blocks += stbuf->ia_blocks;
+
                         if (!mlocal->iobref)
                                 mlocal->iobref = iobref_new ();
                         iobref_merge (mlocal->iobref, iobref);
@@ -3278,18 +3388,21 @@ stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  * cause any bugs at higher levels */
                 memcpy (&tmp_stbuf, &mlocal->replies[0].stbuf,
                         sizeof (struct iatt));
+                tmp_stbuf.ia_size = local->stbuf_size;
+                tmp_stbuf.ia_blocks = local->stbuf_blocks;
 
         done:
                 /* */
                 GF_FREE (mlocal->replies);
                 tmp_iobref = mlocal->iobref;
-                fd_unref (mlocal->fd);
-                STACK_UNWIND_STRICT (readv, mframe, op_ret, op_errno, final_vec,
-                                     final_count, &tmp_stbuf, tmp_iobref);
+                /* work around for nfs truncated read. Bug 3774 */
+                tmp_stbuf_p = &tmp_stbuf;
+                WIPE (tmp_stbuf_p);
+                STRIPE_STACK_UNWIND (readv, mframe, op_ret, op_errno, final_vec,
+                                     final_count, &tmp_stbuf, tmp_iobref, NULL);
 
                 iobref_unref (tmp_iobref);
-                if (final_vec)
-                        GF_FREE (final_vec);
+                GF_FREE (final_vec);
         }
 
         goto out;
@@ -3301,11 +3414,11 @@ check_size:
                 STACK_WIND (mframe, stripe_readv_fstat_cbk,
                             (fctx->xl_array[index]),
                             (fctx->xl_array[index])->fops->fstat,
-                            mlocal->fd);
+                            mlocal->fd, NULL);
         }
 
 out:
-        STACK_DESTROY (frame->root);
+        STRIPE_STACK_DESTROY (frame);
 end:
         return 0;
 }
@@ -3313,7 +3426,7 @@ end:
 
 int32_t
 stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
-              size_t size, off_t offset)
+              size_t size, off_t offset, uint32_t flags, dict_t *xdata)
 {
         int32_t           op_errno = EINVAL;
         int32_t           idx = 0;
@@ -3326,6 +3439,7 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
         uint64_t          stripe_size = 0;
         off_t             rounded_start = 0;
         off_t             frame_offset = offset;
+	off_t		  dest_offset = 0;
         stripe_local_t   *local = NULL;
         call_frame_t     *rframe = NULL;
         stripe_local_t   *rlocal = NULL;
@@ -3336,7 +3450,7 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
         VALIDATE_OR_GOTO (fd, err);
         VALIDATE_OR_GOTO (fd->inode, err);
 
-        fd_ctx_get (fd, this, &tmp_fctx);
+        inode_ctx_get (fd->inode, this, &tmp_fctx);
         if (!tmp_fctx) {
                 op_errno = EBADFD;
                 goto err;
@@ -3344,6 +3458,8 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
         fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
         stripe_size = fctx->stripe_size;
 
+        STRIPE_VALIDATE_FCTX (fctx, err);
+
         if (!stripe_size) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "Wrong stripe size for the file");
@@ -3358,8 +3474,7 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
         rounded_end = roof (offset+size, stripe_size);
         num_stripe = (rounded_end- rounded_start)/stripe_size;
 
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -3367,8 +3482,8 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
         frame->local = local;
 
         /* This is where all the vectors should be copied. */
-        local->replies = GF_CALLOC (num_stripe, sizeof (struct readv_replies),
-                                    gf_stripe_mt_readv_replies);
+        local->replies = GF_CALLOC (num_stripe, sizeof (struct stripe_replies),
+                                    gf_stripe_mt_stripe_replies);
         if (!local->replies) {
                 op_errno = ENOMEM;
                 goto err;
@@ -3383,8 +3498,7 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
         for (index = off_index; index < (num_stripe + off_index); index++) {
                 rframe = copy_frame (frame);
-                rlocal = GF_CALLOC (1, sizeof (stripe_local_t),
-                                    gf_stripe_mt_stripe_local_t);
+                rlocal = mem_get0 (this->local_pool);
                 if (!rlocal) {
                         op_errno = ENOMEM;
                         goto err;
@@ -3398,19 +3512,26 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
                 rlocal->readv_size = frame_size;
                 rframe->local = rlocal;
                 idx = (index % fctx->stripe_count);
+
+		if (fctx->stripe_coalesce)
+			dest_offset = coalesced_offset(frame_offset,
+				stripe_size, fctx->stripe_count);
+		else
+			dest_offset = frame_offset;
+
                 STACK_WIND (rframe, stripe_readv_cbk, fctx->xl_array[idx],
                             fctx->xl_array[idx]->fops->readv,
-                            fd, frame_size, frame_offset);
+                            fd, frame_size, dest_offset, flags, xdata);
 
                 frame_offset += frame_size;
         }
 
         return 0;
 err:
-        if (local && local->fd)
-                fd_unref (local->fd);
+        if (rframe)
+                STRIPE_STACK_DESTROY (rframe);
 
-        STACK_UNWIND_STRICT (readv, frame, -1, op_errno, NULL, 0, NULL, NULL);
+        STRIPE_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
         return 0;
 }
 
@@ -3418,11 +3539,15 @@ err:
 int32_t
 stripe_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                   struct iatt *postbuf)
+                   struct iatt *postbuf, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
+	stripe_local_t *mlocal = NULL;
         call_frame_t   *prev = NULL;
+	call_frame_t   *mframe = NULL;
+	struct stripe_replies *reply = NULL;
+	int32_t		i = 0;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -3431,39 +3556,82 @@ stripe_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         prev  = cookie;
         local = frame->local;
+	mframe = local->orig_frame;
+	mlocal = mframe->local;
 
         LOCK(&frame->lock);
         {
-                callcnt = ++local->call_count;
+                callcnt = ++mlocal->call_count;
+
+		mlocal->replies[local->node_index].op_ret = op_ret;
+		mlocal->replies[local->node_index].op_errno = op_errno;
 
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "%s returned error %s",
-                                prev->this->name, strerror (op_errno));
-                        local->op_errno = op_errno;
-                        local->op_ret = -1;
-                }
                 if (op_ret >= 0) {
-                        local->op_ret += op_ret;
-                        local->post_buf = *postbuf;
-                        local->pre_buf = *prebuf;
+                        mlocal->post_buf = *postbuf;
+                        mlocal->pre_buf = *prebuf;
+
+			mlocal->prebuf_blocks  += prebuf->ia_blocks;
+			mlocal->postbuf_blocks += postbuf->ia_blocks;
+
+			correct_file_size(prebuf, mlocal->fctx, prev);
+			correct_file_size(postbuf, mlocal->fctx, prev);
+
+			if (mlocal->prebuf_size < prebuf->ia_size)
+				mlocal->prebuf_size = prebuf->ia_size;
+			if (mlocal->postbuf_size < postbuf->ia_size)
+				mlocal->postbuf_size = postbuf->ia_size;
                 }
         }
         UNLOCK (&frame->lock);
 
-        if ((callcnt == local->wind_count) && local->unwind) {
-                STACK_UNWIND_STRICT (writev, frame, local->op_ret,
-                                     local->op_errno, &local->pre_buf,
-                                     &local->post_buf);
+        if ((callcnt == mlocal->wind_count) && mlocal->unwind) {
+		mlocal->pre_buf.ia_size = mlocal->prebuf_size;
+		mlocal->pre_buf.ia_blocks = mlocal->prebuf_blocks;
+		mlocal->post_buf.ia_size = mlocal->postbuf_size;
+		mlocal->post_buf.ia_blocks = mlocal->postbuf_blocks;
+
+		/*
+		 * Only return the number of consecutively written bytes up until
+		 * the first error. Only return an error if it occurs first.
+		 *
+		 * When a short write occurs, the application should retry at the
+		 * appropriate offset, at which point we'll potentially pass back
+		 * the error.
+		 */
+		for (i = 0, reply = mlocal->replies; i < mlocal->wind_count;
+			i++, reply++) {
+			if (reply->op_ret == -1) {
+				gf_log(this->name, GF_LOG_DEBUG, "reply %d "
+					"returned error %s", i,
+					strerror(reply->op_errno));
+				if (!mlocal->op_ret) {
+					mlocal->op_ret = -1;
+					mlocal->op_errno = reply->op_errno;
+				}
+				break;
+			}
+
+			mlocal->op_ret += reply->op_ret;
+
+			if (reply->op_ret < reply->requested_size)
+				break;
+		}
+
+		GF_FREE(mlocal->replies);
+
+                STRIPE_STACK_UNWIND (writev, mframe, mlocal->op_ret,
+                                     mlocal->op_errno, &mlocal->pre_buf,
+                                     &mlocal->post_buf, NULL);
         }
 out:
+	STRIPE_STACK_DESTROY(frame);
         return 0;
 }
 
 int32_t
 stripe_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
                struct iovec *vector, int32_t count, off_t offset,
-               struct iobref *iobref)
+               uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
         struct iovec     *tmp_vec = NULL;
         stripe_local_t   *local = NULL;
@@ -3477,13 +3645,19 @@ stripe_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
         off_t             fill_size = 0;
         uint64_t          stripe_size = 0;
         uint64_t          tmp_fctx = 0;
+	off_t		  dest_offset = 0;
+	off_t		  rounded_start = 0;
+	off_t		  rounded_end = 0;
+	int32_t		  total_chunks = 0;
+	call_frame_t	  *wframe = NULL;
+	stripe_local_t	  *wlocal = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
         VALIDATE_OR_GOTO (fd->inode, err);
 
-        fd_ctx_get (fd, this, &tmp_fctx);
+        inode_ctx_get (fd->inode, this, &tmp_fctx);
         if (!tmp_fctx) {
                 op_errno = EINVAL;
                 goto err;
@@ -3491,22 +3665,51 @@ stripe_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
         fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
         stripe_size = fctx->stripe_size;
 
+        STRIPE_VALIDATE_FCTX (fctx, err);
+
         /* File has to be stripped across the child nodes */
         for (idx = 0; idx< count; idx ++) {
                 total_size += vector[idx].iov_len;
         }
         remaining_size = total_size;
 
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         frame->local = local;
         local->stripe_size = stripe_size;
+	local->fctx = fctx;
 
+        if (!stripe_size) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Wrong stripe size for the file");
+                op_errno = EINVAL;
+                goto err;
+        }
+
+	rounded_start = floor(offset, stripe_size);
+	rounded_end = roof(offset + total_size, stripe_size);
+	total_chunks = (rounded_end - rounded_start) / stripe_size;
+	local->replies = GF_CALLOC(total_chunks, sizeof(struct stripe_replies),
+				gf_stripe_mt_stripe_replies);
+	if (!local->replies) {
+		op_errno = ENOMEM;
+		goto err;
+	}
+
+	total_chunks = 0;
         while (1) {
+		wframe = copy_frame(frame);
+		wlocal = mem_get0(this->local_pool);
+		if (!wlocal) {
+			op_errno = ENOMEM;
+			goto err;
+		}
+		wlocal->orig_frame = frame;
+		wframe->local = wlocal;
+
                 /* Send striped chunk of the vector to child
                    nodes appropriately. */
                 idx = (((offset + offset_offset) /
@@ -3534,47 +3737,589 @@ stripe_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
                 if (remaining_size == 0)
                         local->unwind = 1;
 
-                STACK_WIND (frame, stripe_writev_cbk, fctx->xl_array[idx],
+		/*
+		 * Store off the request index (with respect to the chunk of the
+		 * initial offset) and the size of the request. This is required
+		 * in the callback to calculate an appropriate return value in
+		 * the event of a write failure in one or more requests.
+		 */
+		wlocal->node_index = total_chunks;
+		local->replies[total_chunks].requested_size = fill_size;
+
+		dest_offset = offset + offset_offset;
+		if (fctx->stripe_coalesce)
+			dest_offset = coalesced_offset(dest_offset,
+					local->stripe_size, fctx->stripe_count);
+
+                STACK_WIND (wframe, stripe_writev_cbk, fctx->xl_array[idx],
                             fctx->xl_array[idx]->fops->writev, fd, tmp_vec,
-                            tmp_count, offset + offset_offset, iobref);
+                            tmp_count, dest_offset, flags, iobref,
+                            xdata);
+
                 GF_FREE (tmp_vec);
                 offset_offset += fill_size;
+		total_chunks++;
                 if (remaining_size == 0)
                         break;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL);
+	if (wframe)
+		STRIPE_STACK_DESTROY(wframe);
+
+        STRIPE_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
 int32_t
-stripe_release (xlator_t *this, fd_t *fd)
+stripe_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
+{
+        int32_t         callcnt = 0;
+        stripe_local_t *local = NULL;
+	stripe_local_t *mlocal = NULL;
+        call_frame_t   *prev = NULL;
+	call_frame_t   *mframe = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = cookie;
+        local = frame->local;
+	mframe = local->orig_frame;
+	mlocal = mframe->local;
+
+        LOCK(&frame->lock);
+        {
+                callcnt = ++mlocal->call_count;
+
+                if (op_ret == 0) {
+                        mlocal->post_buf = *postbuf;
+                        mlocal->pre_buf = *prebuf;
+
+			mlocal->prebuf_blocks  += prebuf->ia_blocks;
+			mlocal->postbuf_blocks += postbuf->ia_blocks;
+
+			correct_file_size(prebuf, mlocal->fctx, prev);
+			correct_file_size(postbuf, mlocal->fctx, prev);
+
+			if (mlocal->prebuf_size < prebuf->ia_size)
+				mlocal->prebuf_size = prebuf->ia_size;
+			if (mlocal->postbuf_size < postbuf->ia_size)
+				mlocal->postbuf_size = postbuf->ia_size;
+                }
+
+		/* return the first failure */
+		if (mlocal->op_ret == 0) {
+			mlocal->op_ret = op_ret;
+			mlocal->op_errno = op_errno;
+		}
+        }
+        UNLOCK (&frame->lock);
+
+        if ((callcnt == mlocal->wind_count) && mlocal->unwind) {
+		mlocal->pre_buf.ia_size = mlocal->prebuf_size;
+		mlocal->pre_buf.ia_blocks = mlocal->prebuf_blocks;
+		mlocal->post_buf.ia_size = mlocal->postbuf_size;
+		mlocal->post_buf.ia_blocks = mlocal->postbuf_blocks;
+
+                STRIPE_STACK_UNWIND (fallocate, mframe, mlocal->op_ret,
+                                     mlocal->op_errno, &mlocal->pre_buf,
+                                     &mlocal->post_buf, NULL);
+        }
+out:
+	STRIPE_STACK_DESTROY(frame);
+        return 0;
+}
+
+int32_t
+stripe_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+		 off_t offset, size_t len, dict_t *xdata)
 {
+        stripe_local_t   *local = NULL;
+        stripe_fd_ctx_t  *fctx = NULL;
+        int32_t           op_errno = 1;
+        int32_t           idx = 0;
+        int32_t           offset_offset = 0;
+        int32_t           remaining_size = 0;
+        off_t             fill_size = 0;
+        uint64_t          stripe_size = 0;
         uint64_t          tmp_fctx = 0;
+	off_t		  dest_offset = 0;
+	call_frame_t	  *fframe = NULL;
+	stripe_local_t	  *flocal = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
+
+        inode_ctx_get (fd->inode, this, &tmp_fctx);
+        if (!tmp_fctx) {
+                op_errno = EINVAL;
+                goto err;
+        }
+        fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
+        stripe_size = fctx->stripe_size;
+
+        STRIPE_VALIDATE_FCTX (fctx, err);
+
+        remaining_size = len;
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+        frame->local = local;
+        local->stripe_size = stripe_size;
+	local->fctx = fctx;
+
+        if (!stripe_size) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Wrong stripe size for the file");
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        while (1) {
+		fframe = copy_frame(frame);
+		flocal = mem_get0(this->local_pool);
+		if (!flocal) {
+			op_errno = ENOMEM;
+			goto err;
+		}
+		flocal->orig_frame = frame;
+		fframe->local = flocal;
+
+		/* send fallocate request to the associated child node */
+                idx = (((offset + offset_offset) /
+                        local->stripe_size) % fctx->stripe_count);
+
+                fill_size = (local->stripe_size -
+                             ((offset + offset_offset) % local->stripe_size));
+                if (fill_size > remaining_size)
+                        fill_size = remaining_size;
+
+                remaining_size -= fill_size;
+
+                local->wind_count++;
+                if (remaining_size == 0)
+                        local->unwind = 1;
+
+		dest_offset = offset + offset_offset;
+		if (fctx->stripe_coalesce)
+			dest_offset = coalesced_offset(dest_offset,
+					local->stripe_size, fctx->stripe_count);
+
+		/*
+		 * TODO: Create a separate handler for coalesce mode that sends a
+		 * single fallocate per-child (since the ranges are linear).
+		 */
+		STACK_WIND(fframe, stripe_fallocate_cbk, fctx->xl_array[idx],
+			   fctx->xl_array[idx]->fops->fallocate, fd, mode,
+			   dest_offset, fill_size, xdata);
+
+                offset_offset += fill_size;
+                if (remaining_size == 0)
+                        break;
+        }
+
+        return 0;
+err:
+	if (fframe)
+		STRIPE_STACK_DESTROY(fframe);
+
+        STRIPE_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+stripe_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                   struct iatt *postbuf, dict_t *xdata)
+{
+        int32_t         callcnt = 0;
+        stripe_local_t *local = NULL;
+	stripe_local_t *mlocal = NULL;
+        call_frame_t   *prev = NULL;
+	call_frame_t   *mframe = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = cookie;
+        local = frame->local;
+	mframe = local->orig_frame;
+	mlocal = mframe->local;
+
+        LOCK(&frame->lock);
+        {
+                callcnt = ++mlocal->call_count;
+
+                if (op_ret == 0) {
+                        mlocal->post_buf = *postbuf;
+                        mlocal->pre_buf = *prebuf;
+
+			mlocal->prebuf_blocks  += prebuf->ia_blocks;
+			mlocal->postbuf_blocks += postbuf->ia_blocks;
+
+			correct_file_size(prebuf, mlocal->fctx, prev);
+			correct_file_size(postbuf, mlocal->fctx, prev);
+
+			if (mlocal->prebuf_size < prebuf->ia_size)
+				mlocal->prebuf_size = prebuf->ia_size;
+			if (mlocal->postbuf_size < postbuf->ia_size)
+				mlocal->postbuf_size = postbuf->ia_size;
+                }
+
+		/* return the first failure */
+		if (mlocal->op_ret == 0) {
+			mlocal->op_ret = op_ret;
+			mlocal->op_errno = op_errno;
+		}
+        }
+        UNLOCK (&frame->lock);
+
+        if ((callcnt == mlocal->wind_count) && mlocal->unwind) {
+		mlocal->pre_buf.ia_size = mlocal->prebuf_size;
+		mlocal->pre_buf.ia_blocks = mlocal->prebuf_blocks;
+		mlocal->post_buf.ia_size = mlocal->postbuf_size;
+		mlocal->post_buf.ia_blocks = mlocal->postbuf_blocks;
+
+                STRIPE_STACK_UNWIND (discard, mframe, mlocal->op_ret,
+                                     mlocal->op_errno, &mlocal->pre_buf,
+                                     &mlocal->post_buf, NULL);
+        }
+out:
+	STRIPE_STACK_DESTROY(frame);
+        return 0;
+}
+
+int32_t
+stripe_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	       size_t len, dict_t *xdata)
+{
+        stripe_local_t   *local = NULL;
         stripe_fd_ctx_t  *fctx = NULL;
+        int32_t           op_errno = 1;
+        int32_t           idx = 0;
+        int32_t           offset_offset = 0;
+        int32_t           remaining_size = 0;
+        off_t             fill_size = 0;
+        uint64_t          stripe_size = 0;
+        uint64_t          tmp_fctx = 0;
+	off_t		  dest_offset = 0;
+	call_frame_t	  *fframe = NULL;
+	stripe_local_t	  *flocal = NULL;
 
+        VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
 
-        fd_ctx_del (fd, this, &tmp_fctx);
+        inode_ctx_get (fd->inode, this, &tmp_fctx);
         if (!tmp_fctx) {
+                op_errno = EINVAL;
+                goto err;
+        }
+        fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
+        stripe_size = fctx->stripe_size;
+
+        STRIPE_VALIDATE_FCTX (fctx, err);
+
+        remaining_size = len;
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
                 goto err;
         }
+        frame->local = local;
+        local->stripe_size = stripe_size;
+	local->fctx = fctx;
+
+        if (!stripe_size) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Wrong stripe size for the file");
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        while (1) {
+		fframe = copy_frame(frame);
+		flocal = mem_get0(this->local_pool);
+		if (!flocal) {
+			op_errno = ENOMEM;
+			goto err;
+		}
+		flocal->orig_frame = frame;
+		fframe->local = flocal;
+
+		/* send discard request to the associated child node */
+                idx = (((offset + offset_offset) /
+                        local->stripe_size) % fctx->stripe_count);
+
+                fill_size = (local->stripe_size -
+                             ((offset + offset_offset) % local->stripe_size));
+                if (fill_size > remaining_size)
+                        fill_size = remaining_size;
+
+                remaining_size -= fill_size;
+
+                local->wind_count++;
+                if (remaining_size == 0)
+                        local->unwind = 1;
+
+		dest_offset = offset + offset_offset;
+		if (fctx->stripe_coalesce)
+			dest_offset = coalesced_offset(dest_offset,
+					local->stripe_size, fctx->stripe_count);
+
+		/*
+		 * TODO: Create a separate handler for coalesce mode that sends a
+		 * single discard per-child (since the ranges are linear).
+		 */
+		STACK_WIND(fframe, stripe_discard_cbk, fctx->xl_array[idx],
+			   fctx->xl_array[idx]->fops->discard, fd, dest_offset,
+			   fill_size, xdata);
+
+                offset_offset += fill_size;
+                if (remaining_size == 0)
+                        break;
+        }
+
+        return 0;
+err:
+	if (fframe)
+		STRIPE_STACK_DESTROY(fframe);
+
+        STRIPE_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
+}
+
+int32_t
+stripe_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                   struct iatt *postbuf, dict_t *xdata)
+{
+        int32_t         callcnt    = 0;
+        stripe_local_t *local      = NULL;
+        stripe_local_t *mlocal     = NULL;
+        call_frame_t   *prev       = NULL;
+        call_frame_t   *mframe     = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = cookie;
+        local = frame->local;
+        mframe = local->orig_frame;
+        mlocal = mframe->local;
+
+        LOCK(&frame->lock);
+        {
+                callcnt = ++mlocal->call_count;
+
+                if (op_ret == 0) {
+                        mlocal->post_buf = *postbuf;
+                        mlocal->pre_buf = *prebuf;
+
+                        mlocal->prebuf_blocks  += prebuf->ia_blocks;
+                        mlocal->postbuf_blocks += postbuf->ia_blocks;
+
+                        correct_file_size(prebuf, mlocal->fctx, prev);
+                        correct_file_size(postbuf, mlocal->fctx, prev);
+
+                        if (mlocal->prebuf_size < prebuf->ia_size)
+                                mlocal->prebuf_size = prebuf->ia_size;
+                        if (mlocal->postbuf_size < postbuf->ia_size)
+                                mlocal->postbuf_size = postbuf->ia_size;
+                }
+
+                /* return the first failure */
+                if (mlocal->op_ret == 0) {
+                        mlocal->op_ret = op_ret;
+                        mlocal->op_errno = op_errno;
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        if ((callcnt == mlocal->wind_count) && mlocal->unwind) {
+                mlocal->pre_buf.ia_size = mlocal->prebuf_size;
+                mlocal->pre_buf.ia_blocks = mlocal->prebuf_blocks;
+                mlocal->post_buf.ia_size = mlocal->postbuf_size;
+                mlocal->post_buf.ia_blocks = mlocal->postbuf_blocks;
 
+                STRIPE_STACK_UNWIND (zerofill, mframe, mlocal->op_ret,
+                                     mlocal->op_errno, &mlocal->pre_buf,
+                                     &mlocal->post_buf, NULL);
+        }
+out:
+        STRIPE_STACK_DESTROY(frame);
+        return 0;
+}
+
+int32_t
+stripe_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+               size_t len, dict_t *xdata)
+{
+        stripe_local_t   *local            = NULL;
+        stripe_fd_ctx_t  *fctx             = NULL;
+        int32_t           op_errno         = 1;
+        int32_t           idx              = 0;
+        int32_t           offset_offset    = 0;
+        int32_t           remaining_size   = 0;
+        off_t             fill_size        = 0;
+        uint64_t          stripe_size      = 0;
+        uint64_t          tmp_fctx         = 0;
+        off_t             dest_offset      = 0;
+        call_frame_t     *fframe           = NULL;
+        stripe_local_t   *flocal           = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
+
+        inode_ctx_get (fd->inode, this, &tmp_fctx);
+        if (!tmp_fctx) {
+                op_errno = EINVAL;
+                goto err;
+        }
         fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
+        stripe_size = fctx->stripe_size;
 
-        if (!fctx->static_array)
-                GF_FREE (fctx->xl_array);
+        STRIPE_VALIDATE_FCTX (fctx, err);
 
-        GF_FREE (fctx);
+        remaining_size = len;
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+        fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
+        stripe_size = fctx->stripe_size;
+
+        STRIPE_VALIDATE_FCTX (fctx, err);
+
+        remaining_size = len;
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+        fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
+        stripe_size = fctx->stripe_size;
+
+        STRIPE_VALIDATE_FCTX (fctx, err);
+
+        remaining_size = len;
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+        frame->local = local;
+        local->stripe_size = stripe_size;
+        local->fctx = fctx;
+
+        if (!stripe_size) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Wrong stripe size for the file");
+                op_errno = EINVAL;
+                goto err;
+        }
 
+        while (1) {
+                fframe = copy_frame(frame);
+                flocal = mem_get0(this->local_pool);
+                if (!flocal) {
+                        op_errno = ENOMEM;
+                        goto err;
+                }
+                flocal->orig_frame = frame;
+                fframe->local = flocal;
+
+                idx = (((offset + offset_offset) /
+                        local->stripe_size) % fctx->stripe_count);
+
+                fill_size = (local->stripe_size -
+                             ((offset + offset_offset) % local->stripe_size));
+                if (fill_size > remaining_size)
+                        fill_size = remaining_size;
+
+                remaining_size -= fill_size;
+
+                local->wind_count++;
+                if (remaining_size == 0)
+                        local->unwind = 1;
+
+                dest_offset = offset + offset_offset;
+                if (fctx->stripe_coalesce)
+                        dest_offset = coalesced_offset(dest_offset,
+                                                       local->stripe_size,
+                                                       fctx->stripe_count);
+
+                STACK_WIND(fframe, stripe_zerofill_cbk, fctx->xl_array[idx],
+                           fctx->xl_array[idx]->fops->zerofill, fd,
+                           dest_offset, fill_size, xdata);
+                offset_offset += fill_size;
+                if (remaining_size == 0)
+                        break;
+        }
+
+        return 0;
 err:
+        if (fframe)
+                STRIPE_STACK_DESTROY(fframe);
+
+        STRIPE_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
+}
+
+int32_t
+stripe_release (xlator_t *this, fd_t *fd)
+{
 	return 0;
 }
 
+int
+stripe_forget (xlator_t *this, inode_t *inode)
+{
+        uint64_t          tmp_fctx = 0;
+        stripe_fd_ctx_t  *fctx = NULL;
+
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (inode, err);
+
+        (void) inode_ctx_del (inode, this, &tmp_fctx);
+        if (!tmp_fctx) {
+                goto err;
+        }
+
+        fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
+
+        if (!fctx->static_array)
+                GF_FREE (fctx->xl_array);
+
+        GF_FREE (fctx);
+err:
+        return 0;
+}
 
 int32_t
 notify (xlator_t *this, int32_t event, void *data, ...)
@@ -3582,6 +4327,7 @@ notify (xlator_t *this, int32_t event, void *data, ...)
         stripe_private_t *priv = NULL;
         int               down_client = 0;
         int               i = 0;
+        gf_boolean_t      heard_from_all_children = _gf_false;
 
         if (!this)
                 return 0;
@@ -3599,23 +4345,28 @@ notify (xlator_t *this, int32_t event, void *data, ...)
                         if (data == priv->xl_array[i])
                                 break;
                 }
-                priv->state[i] = 1;
-                for (i = 0; i < priv->child_count; i++) {
-                        if (!priv->state[i])
-                                down_client++;
+
+                if (priv->child_count == i) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "got GF_EVENT_CHILD_UP bad subvolume %s",
+                                data? ((xlator_t *)data)->name: NULL);
+                        break;
                 }
 
                 LOCK (&priv->lock);
                 {
-                        priv->nodes_down = down_client;
                         if (data == FIRST_CHILD (this))
                                 priv->first_child_down = 0;
-                        if (!priv->nodes_down)
-                                default_notify (this, event, data);
+                        priv->last_event[i] = event;
                 }
                 UNLOCK (&priv->lock);
         }
         break;
+        case GF_EVENT_CHILD_CONNECTING:
+        {
+                // 'CONNECTING' doesn't ensure its CHILD_UP, so do nothing
+                goto out;
+        }
         case GF_EVENT_CHILD_DOWN:
         {
                 /* get an index number to set */
@@ -3623,20 +4374,19 @@ notify (xlator_t *this, int32_t event, void *data, ...)
                         if (data == priv->xl_array[i])
                                 break;
                 }
-                priv->state[i] = 0;
-                for (i = 0; i < priv->child_count; i++) {
-                        if (!priv->state[i])
-                                down_client++;
+
+                if (priv->child_count == i) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "got GF_EVENT_CHILD_DOWN bad subvolume %s",
+                                data? ((xlator_t *)data)->name: NULL);
+                        break;
                 }
 
                 LOCK (&priv->lock);
                 {
-                        priv->nodes_down = down_client;
-
                         if (data == FIRST_CHILD (this))
                                 priv->first_child_down = 1;
-                        if (priv->nodes_down)
-                                default_notify (this, event, data);
+                        priv->last_event[i] = event;
                 }
                 UNLOCK (&priv->lock);
         }
@@ -3646,67 +4396,252 @@ notify (xlator_t *this, int32_t event, void *data, ...)
         {
                 /* */
                 default_notify (this, event, data);
+                goto out;
         }
         break;
         }
 
+        // Consider child as down if it's last_event is not CHILD_UP
+        for (i = 0, down_client = 0; i < priv->child_count; i++)
+                if (priv->last_event[i] != GF_EVENT_CHILD_UP)
+                        down_client++;
+
+        LOCK (&priv->lock);
+        {
+                priv->nodes_down = down_client;
+        }
+        UNLOCK (&priv->lock);
+
+        heard_from_all_children = _gf_true;
+        for (i = 0; i < priv->child_count; i++)
+                if (!priv->last_event[i])
+                        heard_from_all_children = _gf_false;
+
+        if (heard_from_all_children)
+                default_notify (this, event, data);
+out:
         return 0;
 }
 
 int
-set_stripe_block_size (xlator_t *this, stripe_private_t *priv, char *data)
+stripe_setxattr_cbk (call_frame_t *frame, void *cookie,
+                     xlator_t *this, int op_ret, int op_errno, dict_t *xdata)
 {
-        int                    ret = -1;
-        char                  *tmp_str = NULL;
-        char                  *tmp_str1 = NULL;
-        char                  *dup_str = NULL;
-        char                  *stripe_str = NULL;
-        char                  *pattern = NULL;
-        char                  *num = NULL;
-        struct stripe_options *temp_stripeopt = NULL;
-        struct stripe_options *stripe_opt = NULL;
-
-        if (!this || !priv || !data)
-                goto out;
-
-        /* Get the pattern for striping.
-           "option block-size *avi:10MB" etc */
-        stripe_str = strtok_r (data, ",", &tmp_str);
-        while (stripe_str) {
-                dup_str = gf_strdup (stripe_str);
-                stripe_opt = CALLOC (1, sizeof (struct stripe_options));
-                if (!stripe_opt) {
-                        GF_FREE (dup_str);
-                        goto out;
-                }
+        int ret = -1;
+        int call_cnt = 0;
+        stripe_local_t *local = NULL;
 
-                pattern = strtok_r (dup_str, ":", &tmp_str1);
-                num = strtok_r (NULL, ":", &tmp_str1);
-                if (!num) {
-                        num = pattern;
-                        pattern = "*";
+        if (!frame || !frame->local || !this) {
+                gf_log ("", GF_LOG_ERROR, "Possible NULL deref");
+                return ret;
+        }
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                call_cnt = --local->wind_count;
+
+                /**
+                 * We overwrite ->op_* values here for subsequent faliure
+                 * conditions, hence we propogate the last errno down the
+                 * stack.
+                 */
+                if (op_ret < 0) {
+                        local->op_ret = op_ret;
+                        local->op_errno = op_errno;
+                        goto unlock;
                 }
-                if (gf_string2bytesize (num, &stripe_opt->block_size) != 0) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "invalid number format \"%s\"", num);
-                        goto out;
+        }
+
+ unlock:
+        UNLOCK (&frame->lock);
+
+        if (!call_cnt) {
+                STRIPE_STACK_UNWIND (setxattr, frame, local->op_ret,
+                                     local->op_errno, xdata);
+        }
+
+        return 0;
+}
+
+#ifdef HAVE_BD_XLATOR
+int
+stripe_is_bd (dict_t *this, char *key, data_t *value, void *data)
+{
+        gf_boolean_t *is_bd = data;
+
+        if (data == NULL)
+                return 0;
+
+        if (XATTR_IS_BD (key))
+            *is_bd = _gf_true;
+
+        return 0;
+}
+
+inline gf_boolean_t
+stripe_setxattr_is_bd (dict_t *dict)
+{
+        gf_boolean_t is_bd = _gf_false;
+
+        if (dict == NULL)
+                goto out;
+
+        dict_foreach (dict, stripe_is_bd, &is_bd);
+out:
+        return is_bd;
+}
+#else
+#define stripe_setxattr_is_bd(dict) _gf_false
+#endif
+
+int
+stripe_setxattr (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, dict_t *dict, int flags, dict_t *xdata)
+{
+        int32_t           op_errno = EINVAL;
+        xlator_list_t    *trav     = NULL;
+        stripe_private_t *priv     = NULL;
+        stripe_local_t   *local    = NULL;
+        int               i        = 0;
+        gf_boolean_t      is_bd    = _gf_false;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.*stripe*", dict,
+                                   op_errno, err);
+
+        priv = this->private;
+        trav = this->children;
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        frame->local = local;
+        local->wind_count = priv->child_count;
+        local->op_ret = local->op_errno = 0;
+
+        is_bd = stripe_setxattr_is_bd (dict);
+
+        /**
+         * Set xattrs for directories on all subvolumes. Additionally
+         * this power is only given to a special client. Bd xlator
+         * also needs xattrs for regular files (ie LVs)
+         */
+        if (((frame->root->pid == GF_CLIENT_PID_GSYNCD) &&
+             IA_ISDIR (loc->inode->ia_type)) || is_bd) {
+                for (i = 0; i < priv->child_count; i++, trav = trav->next) {
+                        STACK_WIND (frame, stripe_setxattr_cbk,
+                                    trav->xlator, trav->xlator->fops->setxattr,
+                                    loc, dict, flags, xdata);
                 }
-                memcpy (stripe_opt->path_pattern, pattern, strlen (pattern));
+        } else {
+                local->wind_count = 1;
+                STACK_WIND (frame, stripe_setxattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->setxattr,
+                            loc, dict, flags, xdata);
+        }
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "block-size : pattern %s : size %"PRId64,
-                        stripe_opt->path_pattern, stripe_opt->block_size);
+        return 0;
+err:
+        STRIPE_STACK_UNWIND (setxattr, frame, -1,  op_errno, NULL);
+        return 0;
+}
 
-                if (!priv->pattern) {
-                        priv->pattern = stripe_opt;
-                } else {
-                        temp_stripeopt = priv->pattern;
-                        while (temp_stripeopt->next)
-                                temp_stripeopt = temp_stripeopt->next;
-                        temp_stripeopt->next = stripe_opt;
+
+int
+stripe_fsetxattr_cbk (call_frame_t *frame, void *cookie,
+                      xlator_t *this, int op_ret, int op_errno, dict_t *xdata)
+{
+        STRIPE_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+
+int
+stripe_is_special_key (dict_t *this,
+                       char *key,
+                       data_t *value,
+                       void *data)
+{
+        gf_boolean_t *is_special = NULL;
+
+        if (data == NULL) {
+                goto out;
+        }
+
+        is_special = data;
+
+        if (XATTR_IS_LOCKINFO (key) || XATTR_IS_BD (key))
+                *is_special = _gf_true;
+
+out:
+        return 0;
+}
+
+int32_t
+stripe_fsetxattr_everyone_cbk (call_frame_t *frame, void *cookie,
+                               xlator_t *this, int32_t op_ret, int32_t op_errno,
+                               dict_t *xdata)
+{
+        int             call_count = 0;
+        stripe_local_t *local      = NULL;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                call_count = --local->wind_count;
+
+                if (op_ret < 0) {
+                        local->op_ret = op_ret;
+                        local->op_errno = op_errno;
                 }
-                stripe_str = strtok_r (NULL, ",", &tmp_str);
-                GF_FREE (dup_str);
+        }
+        UNLOCK (&frame->lock);
+
+        if (call_count == 0) {
+                STRIPE_STACK_UNWIND (fsetxattr, frame, local->op_ret,
+                                     local->op_errno, NULL);
+        }
+        return 0;
+}
+
+int
+stripe_fsetxattr_to_everyone (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                              dict_t *dict, int flags, dict_t *xdata)
+{
+        xlator_list_t    *trav  = NULL;
+        stripe_private_t *priv  = NULL;
+        int               ret   = -1;
+        stripe_local_t   *local = NULL;
+
+        priv = this->private;
+
+        local = mem_get0 (this->local_pool);
+        if (local == NULL) {
+                goto out;
+        }
+
+        frame->local = local;
+
+        local->wind_count = priv->child_count;
+
+        trav = this->children;
+
+        while (trav) {
+                STACK_WIND (frame, stripe_fsetxattr_everyone_cbk,
+                            trav->xlator, trav->xlator->fops->fsetxattr,
+                            fd, dict, flags, xdata);
+                trav = trav->next;
         }
 
         ret = 0;
@@ -3714,6 +4649,377 @@ out:
         return ret;
 }
 
+inline gf_boolean_t
+stripe_fsetxattr_is_special (dict_t *dict)
+{
+        gf_boolean_t is_spl = _gf_false;
+
+        if (dict == NULL) {
+                goto out;
+        }
+
+        dict_foreach (dict, stripe_is_special_key, &is_spl);
+
+out:
+        return is_spl;
+}
+
+int
+stripe_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                  dict_t *dict, int flags, dict_t *xdata)
+{
+        int32_t      op_ret = -1, ret = -1, op_errno = EINVAL;
+        gf_boolean_t is_spl = _gf_false;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.*stripe*", dict,
+                                   op_errno, err);
+
+        is_spl = stripe_fsetxattr_is_special (dict);
+        if (is_spl) {
+                ret = stripe_fsetxattr_to_everyone (frame, this, fd, dict,
+                                                    flags, xdata);
+                if (ret < 0) {
+                        op_errno = ENOMEM;
+                        goto err;
+                }
+
+                goto out;
+        }
+
+        STACK_WIND (frame, stripe_fsetxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr,
+                    fd, dict, flags, xdata);
+out:
+        return 0;
+err:
+        STRIPE_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, NULL);
+        return 0;
+}
+
+int
+stripe_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        STRIPE_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+stripe_removexattr (call_frame_t *frame, xlator_t *this,
+                    loc_t *loc, const char *name, dict_t *xdata)
+{
+        int32_t         op_errno = EINVAL;
+
+        VALIDATE_OR_GOTO (this, err);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.*stripe*",
+                                 name, op_errno, err);
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (loc, err);
+
+        STACK_WIND (frame, stripe_removexattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->removexattr,
+                    loc, name, xdata);
+        return 0;
+err:
+        STRIPE_STACK_UNWIND (removexattr, frame, -1,  op_errno, NULL);
+        return 0;
+}
+
+
+int
+stripe_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        STRIPE_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+stripe_fremovexattr (call_frame_t *frame, xlator_t *this,
+                     fd_t *fd, const char *name, dict_t *xdata)
+{
+        int32_t         op_ret   = -1;
+        int32_t         op_errno = EINVAL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.*stripe*",
+                                 name, op_errno, err);
+
+        STACK_WIND (frame, stripe_fremovexattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fremovexattr,
+                    fd, name, xdata);
+        return 0;
+ err:
+        STRIPE_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+stripe_readdirp_lookup_cbk (call_frame_t *frame, void *cookie,
+                            xlator_t *this, int op_ret, int op_errno,
+                            inode_t *inode, struct iatt *stbuf,
+                            dict_t *xattr, struct iatt *parent)
+{
+        stripe_local_t          *local          = NULL;
+        call_frame_t            *main_frame     = NULL;
+        stripe_local_t          *main_local     = NULL;
+        gf_dirent_t             *entry          = NULL;
+        call_frame_t            *prev           = NULL;
+        int                      done           = 0;
+
+        local = frame->local;
+        prev = cookie;
+
+        entry = local->dirent;
+
+        main_frame = local->orig_frame;
+        main_local = main_frame->local;
+        LOCK (&frame->lock);
+        {
+
+                local->call_count--;
+                if (!local->call_count)
+                        done = 1;
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        local->op_ret = op_ret;
+                        goto unlock;
+                }
+
+		if (stripe_ctx_handle(this, prev, local, xattr))
+			gf_log(this->name, GF_LOG_ERROR,
+				"Error getting fctx info from dict.");
+
+		correct_file_size(stbuf, local->fctx, prev);
+
+                stripe_iatt_merge (stbuf, &entry->d_stat);
+                local->stbuf_blocks += stbuf->ia_blocks;
+        }
+unlock:
+        UNLOCK(&frame->lock);
+
+        if (done) {
+                inode_ctx_put (entry->inode, this,
+                               (uint64_t) (long)local->fctx);
+
+                done = 0;
+                LOCK (&main_frame->lock);
+                {
+                        main_local->wind_count--;
+                        if (!main_local->wind_count)
+                                done = 1;
+                        if (local->op_ret == -1) {
+                                main_local->op_errno = local->op_errno;
+                                main_local->op_ret = local->op_ret;
+                        }
+                        entry->d_stat.ia_blocks = local->stbuf_blocks;
+                }
+                UNLOCK (&main_frame->lock);
+                if (done) {
+                        main_frame->local = NULL;
+                        STRIPE_STACK_UNWIND (readdir, main_frame,
+                                             main_local->op_ret,
+                                             main_local->op_errno,
+                                             &main_local->entries, NULL);
+                        gf_dirent_free (&main_local->entries);
+                        stripe_local_wipe (main_local);
+                        mem_put (main_local);
+                }
+                frame->local = NULL;
+                stripe_local_wipe (local);
+                mem_put (local);
+                STRIPE_STACK_DESTROY (frame);
+        }
+
+        return 0;
+}
+
+int32_t
+stripe_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno,
+                     gf_dirent_t *orig_entries, dict_t *xdata)
+{
+        stripe_local_t *local = NULL;
+        call_frame_t   *prev = NULL;
+        gf_dirent_t    *local_entry = NULL;
+        gf_dirent_t    *tmp_entry = NULL;
+        xlator_list_t  *trav = NULL;
+        loc_t          loc = {0, };
+        int32_t        count = 0;
+        stripe_private_t *priv = NULL;
+        int32_t        subvols = 0;
+        dict_t         *xattrs = NULL;
+        call_frame_t   *local_frame = NULL;
+        stripe_local_t *local_ent = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+        prev  = cookie;
+        local = frame->local;
+        trav = this->children;
+        priv = this->private;
+
+        subvols = priv->child_count;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s returned error %s",
+                                prev->this->name, strerror (op_errno));
+                        local->op_errno = op_errno;
+                        local->op_ret = op_ret;
+                        goto unlock;
+                } else {
+                        local->op_ret = op_ret;
+                        list_splice_init (&orig_entries->list,
+                                          &local->entries.list);
+                        local->wind_count = op_ret;
+                }
+
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        if (op_ret == -1)
+                goto out;
+
+        xattrs = dict_new ();
+        if (xattrs)
+                (void) stripe_xattr_request_build (this, xattrs, 0, 0, 0, 0);
+        count = op_ret;
+        list_for_each_entry_safe (local_entry, tmp_entry,
+                                  (&local->entries.list), list) {
+
+                if (!local_entry)
+                        break;
+                if (!IA_ISREG (local_entry->d_stat.ia_type)) {
+                        LOCK (&frame->lock);
+                        {
+                                local->wind_count--;
+                                count = local->wind_count;
+                        }
+                        UNLOCK (&frame->lock);
+                        continue;
+                }
+
+                local_frame = copy_frame (frame);
+
+                if (!local_frame) {
+                        op_errno = ENOMEM;
+                        op_ret = -1;
+                        goto out;
+                }
+
+                local_ent = mem_get0 (this->local_pool);
+                if (!local_ent) {
+                        op_errno = ENOMEM;
+                        op_ret = -1;
+                        goto out;
+                }
+
+                loc.inode = inode_ref (local_entry->inode);
+
+                uuid_copy (loc.gfid, local_entry->d_stat.ia_gfid);
+
+                local_ent->orig_frame = frame;
+
+                local_ent->call_count = subvols;
+
+                local_ent->dirent = local_entry;
+
+                local_frame->local = local_ent;
+
+                trav = this->children;
+                while (trav) {
+                        STACK_WIND (local_frame, stripe_readdirp_lookup_cbk,
+                                    trav->xlator, trav->xlator->fops->lookup,
+                                    &loc, xattrs);
+                        trav = trav->next;
+                }
+                loc_wipe (&loc);
+        }
+out:
+        if (!count) {
+                /* all entries are directories */
+                frame->local = NULL;
+                STRIPE_STACK_UNWIND (readdir, frame, local->op_ret,
+                                     local->op_errno, &local->entries, NULL);
+                gf_dirent_free (&local->entries);
+                stripe_local_wipe (local);
+                mem_put (local);
+        }
+        if (xattrs)
+                dict_unref (xattrs);
+        return 0;
+
+}
+int32_t
+stripe_readdirp (call_frame_t *frame, xlator_t *this,
+                 fd_t *fd, size_t size, off_t off, dict_t *xdata)
+{
+        stripe_local_t  *local  = NULL;
+        stripe_private_t *priv = NULL;
+        xlator_list_t   *trav = NULL;
+        int             op_errno = -1;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        priv = this->private;
+        trav = this->children;
+
+        if (priv->first_child_down) {
+                op_errno = ENOTCONN;
+                goto err;
+        }
+
+        /* Initialization */
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        frame->local = local;
+
+        local->fd = fd_ref (fd);
+
+        local->wind_count = 0;
+
+        local->count = 0;
+        local->op_ret = -1;
+        INIT_LIST_HEAD(&local->entries);
+
+        if (!trav)
+                goto err;
+
+        STACK_WIND (frame, stripe_readdirp_cbk, trav->xlator,
+                    trav->xlator->fops->readdirp, fd, size, off, xdata);
+        return 0;
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        STRIPE_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+
+}
+
 int32_t
 mem_acct_init (xlator_t *this)
 {
@@ -3734,34 +5040,86 @@ out:
         return ret;
 }
 
+static int
+clear_pattern_list (stripe_private_t *priv)
+{
+        struct stripe_options *prev = NULL;
+        struct stripe_options *trav = NULL;
+        int                    ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("stripe", priv, out);
+
+        trav = priv->pattern;
+        priv->pattern = NULL;
+        while (trav) {
+                prev = trav;
+                trav = trav->next;
+                GF_FREE (prev);
+        }
+
+        ret = 0;
+ out:
+        return ret;
+
+
+}
+
+
 int
 reconfigure (xlator_t *this, dict_t *options)
 {
 
-	stripe_private_t *priv = NULL;
-	data_t           *data = NULL;
-	int 		  ret = 0;
+        stripe_private_t *priv = NULL;
+        data_t           *data = NULL;
+        int               ret = -1;
+        volume_option_t  *opt = NULL;
 
+        GF_ASSERT (this);
+        GF_ASSERT (this->private);
 
-	priv = this->private;
+        priv = this->private;
 
-	data = dict_get (options, "block-size");
-        if (data) {
-		gf_log (this->name, GF_LOG_TRACE,"Reconfiguring Stripe"
-			" Block-size");
-                ret = set_stripe_block_size (this, priv, data->data);
-                if (ret) {
-			gf_log (this->name, GF_LOG_ERROR,
-                        "Reconfigue: Block-Size reconfiguration failed");
-                        ret = -1;
-			goto out;
-		}
-		gf_log (this->name, GF_LOG_TRACE,
-                        "Reconfigue: Block-Size reconfigured Successfully");
-	}
 
-out:
-	return ret;
+        ret = 0;
+        LOCK (&priv->lock);
+        {
+                ret = clear_pattern_list (priv);
+                if (ret)
+                         goto unlock;
+
+                data = dict_get (options, "block-size");
+                if (data) {
+                        ret = set_stripe_block_size (this, priv, data->data);
+                        if (ret)
+                                goto unlock;
+                } else {
+                        opt = xlator_volume_option_get (this, "block-size");
+                        if (!opt) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "option 'block-size' not found");
+                                ret = -1;
+                                goto unlock;
+                        }
+
+                        if (gf_string2bytesize (opt->default_value, &priv->block_size)){
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to set default block-size ");
+                                ret = -1;
+                                goto unlock;
+                        }
+                }
+
+		GF_OPTION_RECONF("coalesce", priv->coalesce, options, bool,
+				unlock);
+        }
+ unlock:
+        UNLOCK (&priv->lock);
+        if (ret)
+                goto out;
+
+        ret = 0;
+ out:
+        return ret;
 
 }
 
@@ -3774,6 +5132,7 @@ int32_t
 init (xlator_t *this)
 {
         stripe_private_t *priv = NULL;
+        volume_option_t  *opt = NULL;
         xlator_list_t    *trav = NULL;
         data_t           *data = NULL;
         int32_t           count = 0;
@@ -3817,9 +5176,9 @@ init (xlator_t *this)
         if (!priv->xl_array)
                 goto out;
 
-        priv->state = GF_CALLOC (count, sizeof (int8_t),
-                                 gf_stripe_mt_int8_t);
-        if (!priv->state)
+        priv->last_event = GF_CALLOC (count, sizeof (int),
+                                      gf_stripe_mt_int32_t);
+        if (!priv->last_event)
                 goto out;
 
         priv->child_count = count;
@@ -3839,41 +5198,56 @@ init (xlator_t *this)
                 goto out;
         }
 
-        priv->block_size = (128 * GF_UNIT_KB);
-        /* option stripe-pattern *avi:1GB,*pdf:4096 */
-        data = dict_get (this->options, "block-size");
-        if (!data) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "No \"option block-size <x>\" given, defaulting "
-                        "to 128KB");
-        } else {
-                ret = set_stripe_block_size (this, priv, data->data);
-                if (ret)
-                        goto out;
-        }
-
-        priv->xattr_supported = 1;
-        data = dict_get (this->options, "use-xattr");
-        if (data) {
-                if (gf_string2boolean (data->data,
-                                       &priv->xattr_supported) == -1) {
+        ret = 0;
+        LOCK (&priv->lock);
+        {
+                opt = xlator_volume_option_get (this, "block-size");
+                if (!opt) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "option 'block-size' not found");
+                        ret = -1;
+                        goto unlock;
+                }
+                if (gf_string2bytesize (opt->default_value, &priv->block_size)){
                         gf_log (this->name, GF_LOG_ERROR,
-                                "error setting hard check for extended "
-                                "attribute");
-                        //return -1;
+                                "Unable to set default block-size ");
+                        ret = -1;
+                        goto unlock;
+                }
+                /* option stripe-pattern *avi:1GB,*pdf:16K */
+                data = dict_get (this->options, "block-size");
+                if (data) {
+                        ret = set_stripe_block_size (this, priv, data->data);
+                        if (ret)
+                                goto unlock;
                 }
         }
+ unlock:
+        UNLOCK (&priv->lock);
+        if (ret)
+                goto out;
 
+        GF_OPTION_INIT ("use-xattr", priv->xattr_supported, bool, out);
         /* notify related */
         priv->nodes_down = priv->child_count;
+
+	GF_OPTION_INIT("coalesce", priv->coalesce, bool, out);
+
+        this->local_pool = mem_pool_new (stripe_local_t, 128);
+        if (!this->local_pool) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto out;
+        }
+
         this->private = priv;
 
         ret = 0;
 out:
         if (ret) {
                 if (priv) {
-                        if (priv->xl_array)
-                                GF_FREE (priv->xl_array);
+                        GF_FREE (priv->xl_array);
                         GF_FREE (priv);
                 }
         }
@@ -3896,15 +5270,16 @@ fini (xlator_t *this)
 
         priv = this->private;
         if (priv) {
-                if (priv->xl_array)
-                        GF_FREE (priv->xl_array);
+                this->private = NULL;
+                GF_FREE (priv->xl_array);
 
                 trav = priv->pattern;
                 while (trav) {
                         prev = trav;
                         trav = trav->next;
-                        FREE (prev);
+                        GF_FREE (prev);
                 }
+                GF_FREE (priv->last_event);
                 LOCK_DESTROY (&priv->lock);
                 GF_FREE (priv);
         }
@@ -3913,44 +5288,527 @@ out:
         return;
 }
 
+int32_t
+stripe_getxattr_unwind (call_frame_t *frame,
+                        int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+
+{
+        STRIPE_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+int
+stripe_internal_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                              int op_ret, int op_errno, dict_t *xattr,
+                              dict_t *xdata)
+{
+
+        char        size_key[256]  = {0,};
+        char        index_key[256] = {0,};
+        char        count_key[256] = {0,};
+        char        coalesce_key[256] = {0,};
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (frame->local, out);
+
+        if (!xattr || (op_ret == -1))
+            goto out;
+
+        sprintf (size_key, "trusted.%s.stripe-size", this->name);
+        sprintf (count_key, "trusted.%s.stripe-count", this->name);
+        sprintf (index_key, "trusted.%s.stripe-index", this->name);
+	sprintf (coalesce_key, "trusted.%s.stripe-coalesce", this->name);
+
+        dict_del (xattr, size_key);
+        dict_del (xattr, count_key);
+        dict_del (xattr, index_key);
+        dict_del (xattr, coalesce_key);
+
+out:
+        STRIPE_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
+
+        return 0;
+
+}
+
+int
+stripe_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+{
+        int                     call_cnt = 0;
+        stripe_local_t         *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (frame->local, out);
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                call_cnt = --local->wind_count;
+        }
+        UNLOCK (&frame->lock);
+
+        if (!xattr || (op_ret < 0))
+                goto out;
+
+        local->op_ret = 0;
+
+        if (!local->xattr) {
+                local->xattr = dict_ref (xattr);
+        } else {
+                stripe_aggregate_xattr (local->xattr, xattr);
+        }
+
+out:
+        if (!call_cnt) {
+                STRIPE_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
+                                     local->xattr, xdata);
+        }
+
+        return 0;
+}
+
+int32_t
+stripe_vgetxattr_cbk (call_frame_t *frame, void *cookie,
+                      xlator_t *this, int32_t op_ret, int32_t op_errno,
+                      dict_t *dict, dict_t *xdata)
+{
+        stripe_local_t      *local         = NULL;
+        int32_t              callcnt       = 0;
+        int32_t              ret           = -1;
+        long                 cky           = 0;
+        void                *xattr_val     = NULL;
+        void                *xattr_serz    = NULL;
+        stripe_xattr_sort_t *xattr         = NULL;
+        dict_t              *stripe_xattr  = NULL;
+
+        if (!frame || !frame->local || !this) {
+                gf_log ("", GF_LOG_ERROR, "Possible NULL deref");
+                return ret;
+        }
+
+        local = frame->local;
+        cky = (long) cookie;
+
+        if (local->xsel[0] == '\0') {
+                gf_log (this->name, GF_LOG_ERROR, "Empty xattr in cbk");
+                return ret;
+        }
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->wind_count;
+
+                if (!dict || (op_ret < 0))
+                        goto out;
+
+                if (!local->xattr_list)
+                        local->xattr_list = (stripe_xattr_sort_t *)
+                                GF_CALLOC (local->nallocs,
+                                           sizeof (stripe_xattr_sort_t),
+                                           gf_stripe_mt_xattr_sort_t);
+
+                if (local->xattr_list) {
+                        xattr = local->xattr_list + (int32_t) cky;
+
+                        ret = dict_get_ptr_and_len (dict, local->xsel,
+                                                    &xattr_val,
+                                                    &xattr->xattr_len);
+                        if (xattr->xattr_len == 0)
+                                goto out;
+
+                        xattr->pos = cky;
+                        xattr->xattr_value = gf_memdup (xattr_val,
+                                                        xattr->xattr_len);
+
+                        if (xattr->xattr_value != NULL)
+                                local->xattr_total_len += xattr->xattr_len + 1;
+                }
+        }
+ out:
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                if (!local->xattr_total_len)
+                        goto unwind;
+
+                stripe_xattr = dict_new ();
+                if (!stripe_xattr)
+                        goto unwind;
+
+                /* select filler based on ->xsel */
+                if (XATTR_IS_PATHINFO (local->xsel))
+                        ret = stripe_fill_pathinfo_xattr (this, local,
+                                                          (char **)&xattr_serz);
+                else if (XATTR_IS_LOCKINFO (local->xsel)) {
+                        ret = stripe_fill_lockinfo_xattr (this, local,
+                                                          &xattr_serz);
+                } else {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Unknown xattr in xattr request");
+                        goto unwind;
+                }
+
+                if (!ret) {
+                        ret = dict_set_dynptr (stripe_xattr, local->xsel,
+                                               xattr_serz,
+                                               local->xattr_total_len);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Can't set %s key in dict",
+                                        local->xsel);
+                }
+
+        unwind:
+                STRIPE_STACK_UNWIND (getxattr, frame, op_ret, op_errno,
+                                     stripe_xattr, NULL);
+
+                ret = stripe_free_xattr_str (local);
+
+                GF_FREE (local->xattr_list);
+
+                if (stripe_xattr)
+                        dict_unref (stripe_xattr);
+        }
+
+        return ret;
+}
+
+int32_t
+stripe_getxattr (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, const char *name, dict_t *xdata)
+{
+        stripe_local_t    *local    = NULL;
+        xlator_list_t     *trav     = NULL;
+        stripe_private_t  *priv     = NULL;
+        int32_t            op_errno = EINVAL;
+        int                i        = 0;
+        xlator_t         **sub_volumes;
+        int                ret      = 0;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+
+        priv = this->private;
+        trav = this->children;
+
+        /* Initialization */
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+        local->op_ret = -1;
+        frame->local = local;
+        loc_copy (&local->loc, loc);
+
+
+        if (name && (strcmp (GF_XATTR_MARKER_KEY, name) == 0)
+            && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+                local->marker.call_count = priv->child_count;
+
+                sub_volumes = alloca ( priv->child_count *
+                                       sizeof (xlator_t *));
+                for (i = 0, trav = this->children; trav ;
+                     trav = trav->next, i++) {
+
+                        *(sub_volumes + i)  = trav->xlator;
+
+                }
+
+                if (cluster_getmarkerattr (frame, this, loc, name,
+                                           local, stripe_getxattr_unwind,
+                                           sub_volumes, priv->child_count,
+                                           MARKER_UUID_TYPE, marker_uuid_default_gauge,
+                                           priv->vol_uuid)) {
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
+                return 0;
+        }
+
+        if (name && strncmp (name, GF_XATTR_QUOTA_SIZE_KEY,
+                             strlen (GF_XATTR_QUOTA_SIZE_KEY)) == 0) {
+                local->wind_count = priv->child_count;
+
+                for (i = 0, trav=this->children; i < priv->child_count; i++,
+                             trav = trav->next) {
+                        STACK_WIND (frame, stripe_getxattr_cbk,
+                                    trav->xlator, trav->xlator->fops->getxattr,
+                                    loc, name, xdata);
+                }
+
+                return 0;
+        }
+
+        if (name &&
+            ((strncmp (name, GF_XATTR_PATHINFO_KEY,
+                       strlen (GF_XATTR_PATHINFO_KEY)) == 0))) {
+                if (IA_ISREG (loc->inode->ia_type)) {
+                        ret = inode_ctx_get (loc->inode, this,
+                                             (uint64_t *) &local->fctx);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "stripe size unavailable from fctx"
+                                        " relying on pathinfo could lead to"
+                                        " wrong results");
+                }
+
+                local->nallocs = local->wind_count = priv->child_count;
+                (void) strncpy (local->xsel, name, strlen (name));
+
+                /**
+                 * for xattrs that need info from all childs, fill ->xsel
+                 * as above and call the filler function in cbk based on
+                 * it
+                 */
+                for (i = 0, trav = this->children; i < priv->child_count; i++,
+                     trav = trav->next) {
+                        STACK_WIND_COOKIE (frame, stripe_vgetxattr_cbk,
+                                           (void *) (long) i, trav->xlator,
+                                           trav->xlator->fops->getxattr,
+                                           loc, name, xdata);
+                }
+
+                return 0;
+        }
+
+        if (name &&(*priv->vol_uuid)) {
+                if ((match_uuid_local (name, priv->vol_uuid) == 0)
+                    && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+
+                        if (!IA_FILE_OR_DIR (loc->inode->ia_type))
+                                local->marker.call_count = 1;
+                        else
+                                local->marker.call_count = priv->child_count;
+
+                        sub_volumes = alloca (local->marker.call_count *
+                                              sizeof (xlator_t *));
+
+                        for (i = 0, trav = this->children;
+                             i < local->marker.call_count;
+                             i++, trav = trav->next) {
+                                *(sub_volumes + i) = trav->xlator;
+
+                        }
+
+                        if (cluster_getmarkerattr (frame, this, loc, name,
+                                                   local,
+                                                   stripe_getxattr_unwind,
+                                                   sub_volumes,
+                                                   local->marker.call_count,
+                                                   MARKER_XTIME_TYPE,
+                                                   marker_xtime_default_gauge,
+                                                   priv->vol_uuid)) {
+                                op_errno = EINVAL;
+                                goto err;
+                        }
+
+                        return 0;
+                }
+        }
+
+
+        STACK_WIND (frame, stripe_internal_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+
+        return 0;
+
+err:
+        STRIPE_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
+        return 0;
+}
+
+inline gf_boolean_t
+stripe_is_special_xattr (const char *name)
+{
+        gf_boolean_t    is_spl = _gf_false;
+
+        if (!name) {
+                goto out;
+        }
+
+        if (!strncmp (name, GF_XATTR_LOCKINFO_KEY,
+                      strlen (GF_XATTR_LOCKINFO_KEY))
+            || !strncmp (name, GF_XATTR_PATHINFO_KEY,
+                         strlen (GF_XATTR_PATHINFO_KEY)))
+                is_spl = _gf_true;
+out:
+        return is_spl;
+}
+
+int32_t
+stripe_fgetxattr_from_everyone (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                                const char *name, dict_t *xdata)
+{
+        stripe_local_t   *local = NULL;
+        stripe_private_t *priv  = NULL;
+        int32_t           ret   = -1, op_errno = 0;
+        int               i     = 0;
+        xlator_list_t    *trav  = NULL;
+
+        priv = this->private;
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        local->op_ret = -1;
+        frame->local = local;
+
+        strncpy (local->xsel, name, strlen (name));
+        local->nallocs = local->wind_count = priv->child_count;
+
+        for (i = 0, trav = this->children; i < priv->child_count; i++,
+                     trav = trav->next) {
+                STACK_WIND_COOKIE (frame, stripe_vgetxattr_cbk,
+                                   (void *) (long) i, trav->xlator,
+                                   trav->xlator->fops->fgetxattr,
+                                   fd, name, xdata);
+        }
+
+        return 0;
+
+err:
+        STACK_UNWIND_STRICT (fgetxattr, frame, -1, op_errno, NULL, NULL);
+        return ret;
+}
+
+int32_t
+stripe_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                  const char *name, dict_t *xdata)
+{
+        if (stripe_is_special_xattr (name)) {
+                stripe_fgetxattr_from_everyone (frame, this, fd, name, xdata);
+                goto out;
+        }
+
+        STACK_WIND (frame, stripe_internal_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+
+out:
+        return 0;
+}
+
+
+
+int32_t
+stripe_priv_dump (xlator_t *this)
+{
+        char                    key[GF_DUMP_MAX_BUF_LEN];
+        int                     i = 0;
+        stripe_private_t       *priv = NULL;
+        int                     ret = -1;
+        struct stripe_options  *options = NULL;
+
+        GF_VALIDATE_OR_GOTO ("stripe", this, out);
+
+        priv = this->private;
+        if (!priv)
+                goto out;
+
+        ret = TRY_LOCK (&priv->lock);
+        if (ret != 0)
+                goto out;
+
+        gf_proc_dump_add_section("xlator.cluster.stripe.%s.priv", this->name);
+        gf_proc_dump_write("child_count","%d", priv->child_count);
+
+        for (i = 0; i < priv->child_count; i++) {
+                sprintf (key, "subvolumes[%d]", i);
+                gf_proc_dump_write (key, "%s.%s", priv->xl_array[i]->type,
+                                    priv->xl_array[i]->name);
+        }
+
+        options = priv->pattern;
+        while (options != NULL) {
+                gf_proc_dump_write ("path_pattern", "%s", priv->pattern->path_pattern);
+                gf_proc_dump_write ("options_block_size", "%ul", options->block_size);
+
+                options = options->next;
+        }
+
+        gf_proc_dump_write ("block_size", "%ul", priv->block_size);
+        gf_proc_dump_write ("nodes-down", "%d", priv->nodes_down);
+        gf_proc_dump_write ("first-child_down", "%d", priv->first_child_down);
+        gf_proc_dump_write ("xattr_supported", "%d", priv->xattr_supported);
+
+        UNLOCK (&priv->lock);
+
+out:
+        return ret;
+}
 
 struct xlator_fops fops = {
-        .stat        = stripe_stat,
-        .unlink      = stripe_unlink,
-        .rename      = stripe_rename,
-        .link        = stripe_link,
-        .truncate    = stripe_truncate,
-        .create      = stripe_create,
-        .open        = stripe_open,
-        .readv       = stripe_readv,
-        .writev      = stripe_writev,
-        .statfs      = stripe_statfs,
-        .flush       = stripe_flush,
-        .fsync       = stripe_fsync,
-        .ftruncate   = stripe_ftruncate,
-        .fstat       = stripe_fstat,
-        .mkdir       = stripe_mkdir,
-        .rmdir       = stripe_rmdir,
-        .lk          = stripe_lk,
-        .opendir     = stripe_opendir,
-        .fsyncdir    = stripe_fsyncdir,
-        .setattr     = stripe_setattr,
-        .fsetattr    = stripe_fsetattr,
-        .lookup      = stripe_lookup,
-        .mknod       = stripe_mknod,
+        .stat           = stripe_stat,
+        .unlink         = stripe_unlink,
+        .rename         = stripe_rename,
+        .link           = stripe_link,
+        .truncate       = stripe_truncate,
+        .create         = stripe_create,
+        .open           = stripe_open,
+        .readv          = stripe_readv,
+        .writev         = stripe_writev,
+        .statfs         = stripe_statfs,
+        .flush          = stripe_flush,
+        .fsync          = stripe_fsync,
+        .ftruncate      = stripe_ftruncate,
+        .fstat          = stripe_fstat,
+        .mkdir          = stripe_mkdir,
+        .rmdir          = stripe_rmdir,
+        .lk             = stripe_lk,
+        .opendir        = stripe_opendir,
+        .fsyncdir       = stripe_fsyncdir,
+        .setattr        = stripe_setattr,
+        .fsetattr       = stripe_fsetattr,
+        .lookup         = stripe_lookup,
+        .mknod          = stripe_mknod,
+        .setxattr       = stripe_setxattr,
+        .fsetxattr      = stripe_fsetxattr,
+        .getxattr       = stripe_getxattr,
+        .fgetxattr      = stripe_fgetxattr,
+        .removexattr    = stripe_removexattr,
+        .fremovexattr   = stripe_fremovexattr,
+        .readdirp       = stripe_readdirp,
+	.fallocate	= stripe_fallocate,
+	.discard	= stripe_discard,
+        .zerofill       = stripe_zerofill,
 };
 
 struct xlator_cbks cbks = {
         .release = stripe_release,
+        .forget  = stripe_forget,
 };
 
+struct xlator_dumpops dumpops = {
+        .priv = stripe_priv_dump,
+};
 
 struct volume_options options[] = {
         { .key  = {"block-size"},
-          .type = GF_OPTION_TYPE_ANY
+          .type = GF_OPTION_TYPE_SIZE_LIST,
+          .default_value = "128KB",
+          .min = STRIPE_MIN_BLOCK_SIZE,
+          .description = "Size of the stripe unit that would be read "
+                         "from or written to the striped servers."
         },
         { .key  = {"use-xattr"},
-          .type = GF_OPTION_TYPE_BOOL
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "true"
         },
+	{ .key = {"coalesce"},
+	  .type = GF_OPTION_TYPE_BOOL,
+	  .default_value = "true",
+	  .description = "Enable/Disable coalesce mode to flatten striped "
+			 "files as stored on the server (i.e., eliminate holes "
+			 "caused by the traditional format)."
+	},
         { .key  = {NULL} },
 };
diff --git a/xlators/cluster/stripe/src/stripe.h b/xlators/cluster/stripe/src/stripe.h
index 8afc6aa9a..5673d18f3 100644
--- a/xlators/cluster/stripe/src/stripe.h
+++ b/xlators/cluster/stripe/src/stripe.h
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -33,9 +24,59 @@
 #include "compat.h"
 #include "compat-errno.h"
 #include "stripe-mem-types.h"
+#include "libxlator.h"
 #include <fnmatch.h>
 #include <signal.h>
 
+#define STRIPE_PATHINFO_HEADER "STRIPE:"
+#define STRIPE_MIN_BLOCK_SIZE  (16*GF_UNIT_KB)
+
+#define STRIPE_STACK_UNWIND(fop, frame, params ...) do {           \
+                stripe_local_t *__local = NULL;                    \
+                if (frame) {                                       \
+                        __local = frame->local;                    \
+                        frame->local = NULL;                       \
+                }                                                  \
+                STACK_UNWIND_STRICT (fop, frame, params);          \
+                if (__local) {                                     \
+                        stripe_local_wipe(__local);                \
+                        mem_put (__local);       \
+                }                                                  \
+        } while (0)
+
+#define STRIPE_STACK_DESTROY(frame) do {                        \
+                stripe_local_t *__local = NULL;                 \
+                __local = frame->local;                         \
+                frame->local = NULL;                            \
+                STACK_DESTROY (frame->root);                    \
+                if (__local) {                                  \
+                        stripe_local_wipe (__local);            \
+                        mem_put (__local);    \
+                }                                               \
+        } while (0)
+
+#define STRIPE_VALIDATE_FCTX(fctx, label) do {                  \
+        int     idx = 0;                                        \
+        if (!fctx) {                                            \
+                op_errno = EINVAL;                              \
+                goto label;                                     \
+        }                                                       \
+        for (idx = 0; idx < fctx->stripe_count; idx++) {        \
+                if (!fctx->xl_array[idx]) {                     \
+                        gf_log (this->name, GF_LOG_ERROR,       \
+                                "fctx->xl_array[%d] is NULL",   \
+                                idx);                           \
+                        op_errno = ESTALE;                      \
+                        goto label;                             \
+                }                                               \
+        }                                                       \
+       } while (0)
+
+typedef struct stripe_xattr_sort {
+        int   pos;
+        int   xattr_len;
+        char *xattr_value;
+} stripe_xattr_sort_t;
 
 /**
  * struct stripe_options : This keeps the pattern and the block-size
@@ -57,15 +98,17 @@ struct stripe_private {
         gf_lock_t               lock;
         uint8_t                 nodes_down;
         int8_t                  first_child_down;
+        int                    *last_event;
         int8_t                  child_count;
-        int8_t                 *state; /* Current state of child node */
         gf_boolean_t            xattr_supported;  /* default yes */
+	gf_boolean_t		coalesce;
+        char                    vol_uuid[UUID_SIZE + 1];
 };
 
 /**
- * Used to keep info about the replies received from fops->readv calls
+ * Used to keep info about the replies received from readv/writev calls
  */
-struct readv_replies {
+struct stripe_replies {
         struct iovec *vector;
         int32_t       count;    //count of vector
         int32_t       op_ret;   //op_ret of readv
@@ -77,6 +120,7 @@ struct readv_replies {
 typedef struct _stripe_fd_ctx {
         off_t      stripe_size;
         int        stripe_count;
+	int	   stripe_coalesce;
         int        static_array;
         xlator_t **xl_array;
 } stripe_fd_ctx_t;
@@ -112,9 +156,9 @@ struct stripe_local {
         blkcnt_t             preparent_blocks;
         blkcnt_t             postparent_blocks;
 
-        struct readv_replies *replies;
-        struct statvfs       statvfs_buf;
-        dir_entry_t         *entry;
+        struct stripe_replies *replies;
+        struct statvfs        statvfs_buf;
+        dir_entry_t          *entry;
 
         int8_t               revalidate;
         int8_t               failed;
@@ -136,8 +180,17 @@ struct stripe_local {
         loc_t                loc;
         loc_t                loc2;
 
+        mode_t               mode;
+        dev_t                rdev;
         /* For File I/O fops */
-        dict_t              *dict;
+        dict_t              *xdata;
+
+        stripe_xattr_sort_t *xattr_list;
+        int32_t              xattr_total_len;
+        int32_t              nallocs;
+        char xsel[256];
+
+        struct marker_str    marker;
 
         /* General usage */
         off_t                offset;
@@ -147,13 +200,89 @@ struct stripe_local {
         int entry_self_heal_needed;
 
         int8_t              *list;
-        struct flock         lock;
+        struct gf_flock         lock;
         fd_t                *fd;
         void                *value;
         struct iobref       *iobref;
+        gf_dirent_t          entries;
+        gf_dirent_t         *dirent;
+        dict_t              *xattr;
+        uuid_t               ia_gfid;
+
+        int                  xflag;
+        mode_t               umask;
 };
 
 typedef struct stripe_local   stripe_local_t;
 typedef struct stripe_private stripe_private_t;
 
+/*
+ * Determine the stripe index of a particular frame based on the translator.
+ */
+static inline int32_t stripe_get_frame_index(stripe_fd_ctx_t *fctx,
+					     call_frame_t *prev)
+{
+	int32_t i, idx = -1;
+
+	for (i = 0; i < fctx->stripe_count; i++) {
+		if (fctx->xl_array[i] == prev->this) {
+			idx = i;
+			break;
+		}
+	}
+
+	return idx;
+}
+
+static inline void stripe_copy_xl_array(xlator_t **dst, xlator_t **src,
+					int count)
+{
+	int i;
+
+	for (i = 0; i < count; i++)
+		dst[i] = src[i];
+}
+
+void stripe_local_wipe (stripe_local_t *local);
+int32_t stripe_ctx_handle (xlator_t *this, call_frame_t *prev,
+                           stripe_local_t *local, dict_t *dict);
+void stripe_aggregate_xattr (dict_t *dst, dict_t *src);
+int32_t stripe_xattr_request_build (xlator_t *this, dict_t *dict,
+                                    uint64_t stripe_size, uint32_t stripe_count,
+                                    uint32_t stripe_index,
+				    uint32_t stripe_coalesce);
+int32_t stripe_get_matching_bs (const char *path, stripe_private_t *priv);
+int set_stripe_block_size (xlator_t *this, stripe_private_t *priv, char *data);
+int32_t stripe_iatt_merge (struct iatt *from, struct iatt *to);
+int32_t stripe_fill_pathinfo_xattr (xlator_t *this, stripe_local_t *local,
+                                    char **xattr_serz);
+int32_t stripe_free_xattr_str (stripe_local_t *local);
+int32_t stripe_xattr_aggregate (char *buffer, stripe_local_t *local,
+                                int32_t *total);
+off_t coalesced_offset(off_t offset, uint64_t stripe_size, int stripe_count);
+off_t uncoalesced_size(off_t size, uint64_t stripe_size, int stripe_count,
+			int stripe_index);
+int32_t
+stripe_fill_lockinfo_xattr (xlator_t *this, stripe_local_t *local,
+                            void **xattr_serz);
+
+/*
+ * Adjust the size attribute for files if coalesce is enabled.
+ */
+static inline void correct_file_size(struct iatt *buf, stripe_fd_ctx_t *fctx,
+	call_frame_t *prev)
+{
+	int index;
+
+	if (!IA_ISREG(buf->ia_type))
+		return;
+
+	if (!fctx || !fctx->stripe_coalesce)
+		return;
+
+	index = stripe_get_frame_index(fctx, prev);
+	buf->ia_size = uncoalesced_size(buf->ia_size, fctx->stripe_size,
+		fctx->stripe_count, index);
+}
+
 #endif /* _STRIPE_H_ */
diff --git a/xlators/cluster/unify/src/Makefile.am b/xlators/cluster/unify/src/Makefile.am
deleted file mode 100644
index 2a1fe8372..000000000
--- a/xlators/cluster/unify/src/Makefile.am
+++ /dev/null
@@ -1,16 +0,0 @@
-
-xlator_LTLIBRARIES = unify.la
-xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/legacy/cluster
-
-unify_la_LDFLAGS = -module -avoidversion
-
-unify_la_SOURCES = unify.c unify-self-heal.c
-unify_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-
-noinst_HEADERS = unify.h
-
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
-
-CLEANFILES = 
-
diff --git a/xlators/cluster/unify/src/unify-mem-types.h b/xlators/cluster/unify/src/unify-mem-types.h
deleted file mode 100644
index 3b4abc8e9..000000000
--- a/xlators/cluster/unify/src/unify-mem-types.h
+++ /dev/null
@@ -1,41 +0,0 @@
-
-/*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
-*/
-
-
-#ifndef __UNIFY_MEM_TYPES_H__
-#define __UNIFY_MEM_TYPES_H__
-
-#include "mem-types.h"
-
-enum gf_unify_mem_types_ {
-        gf_unify_mt_char = gf_common_mt_end + 1,
-        gf_unify_mt_int16_t,
-        gf_unify_mt_xlator_t,
-        gf_unify_mt_unify_private_t,
-        gf_unify_mt_xlator_list_t,
-        gf_unify_mt_dir_entry_t,
-        gf_unify_mt_off_t,
-        gf_unify_mt_int,
-        gf_unify_mt_unify_self_heal_struct,
-        gf_unify_mt_unify_local_t,
-        gf_unify_mt_end
-};
-#endif
-
diff --git a/xlators/cluster/unify/src/unify-self-heal.c b/xlators/cluster/unify/src/unify-self-heal.c
deleted file mode 100644
index 88145af9a..000000000
--- a/xlators/cluster/unify/src/unify-self-heal.c
+++ /dev/null
@@ -1,1239 +0,0 @@
-/*
-  Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-/**
- * unify-self-heal.c : 
- *   This file implements few functions which enables 'unify' translator 
- *  to be consistent in its behaviour when 
- *     > a node fails, 
- *     > a node gets added, 
- *     > a failed node comes back
- *     > a new namespace server is added (ie, an fresh namespace server).
- * 
- *  This functionality of 'unify' will enable glusterfs to support storage
- *  system failure, and maintain consistancy. This works both ways, ie, when
- *  an entry (either file or directory) is found on namespace server, and not
- *  on storage nodes, its created in storage nodes and vica-versa.
- * 
- *  The two fops, where it can be implemented are 'getdents ()' and 'lookup ()'
- *
- */
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "glusterfs.h"
-#include "unify.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "stack.h"
-#include "common-utils.h"
-
-int32_t
-unify_sh_getdents_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       dir_entry_t *entry,
-		       int32_t count);
-
-int32_t
-unify_sh_ns_getdents_cbk (call_frame_t *frame,
-			  void *cookie,
-			  xlator_t *this,
-			  int32_t op_ret,
-			  int32_t op_errno,
-			  dir_entry_t *entry,
-			  int32_t count);
-
-int32_t
-unify_bgsh_getdents_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 dir_entry_t *entry,
-			 int32_t count);
-
-int32_t
-unify_bgsh_ns_getdents_cbk (call_frame_t *frame,
-			    void *cookie,
-			    xlator_t *this,
-			    int32_t op_ret,
-			    int32_t op_errno,
-			    dir_entry_t *entry,
-			    int32_t count);
-
-/**
- * unify_local_wipe - free all the extra allocation of local->* here.
- */
-static void 
-unify_local_wipe (unify_local_t *local)
-{
-	/* Free the strdup'd variables in the local structure */
-	if (local->name) {
-		GF_FREE (local->name);
-	}
-
-	if (local->sh_struct) {
-		if (local->sh_struct->offset_list)
-			GF_FREE (local->sh_struct->offset_list);
-
-		if (local->sh_struct->entry_list)
-			GF_FREE (local->sh_struct->entry_list);
-
-		if (local->sh_struct->count_list)
-			GF_FREE (local->sh_struct->count_list);
-
-		GF_FREE (local->sh_struct);
-	}
-
-	loc_wipe (&local->loc1);
-	loc_wipe (&local->loc2);
-}
-
-int32_t 
-unify_sh_setdents_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	inode_t *inode = NULL;
-	dict_t *tmp_dict = NULL;
-	dir_entry_t *prev, *entry, *trav;
-
-	LOCK (&frame->lock);
-	{
-		/* if local->call_count == 0, that means, setdents on 
-		 * storagenodes is still pending.
-		 */
-		if (local->call_count)
-			callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-
-	if (callcnt == 0) {
-		if (local->sh_struct->entry_list[0]) {
-			prev = entry = local->sh_struct->entry_list[0];
-			if (!entry)
-				return 0;
-			trav = entry->next;
-			while (trav) {
-				prev->next = trav->next;
-				GF_FREE (trav->name);
-				if (IA_ISLNK (trav->buf.ia_type))
-					GF_FREE (trav->link);
-				GF_FREE (trav);
-				trav = prev->next;
-			}
-			GF_FREE (entry);
-		}
-
-		if (!local->flags) {
-			if (local->sh_struct->count_list[0] >= 
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-				/* count == size, that means, there are more entries
-				   to read from */
-				//local->call_count = 0;
-				local->sh_struct->offset_list[0] += 
-					UNIFY_SELF_HEAL_GETDENTS_COUNT;
-				STACK_WIND (frame,
-					    unify_sh_ns_getdents_cbk,
-					    NS(this),
-					    NS(this)->fops->getdents,
-					    local->fd,
-					    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-					    local->sh_struct->offset_list[0],
-					    GF_GET_DIR_ONLY);
-			}		
-		} else {
-			inode = local->loc1.inode;
-			fd_unref (local->fd);
-			tmp_dict = local->dict;
-
-			unify_local_wipe (local);
-			
-			STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-				      inode, &local->stbuf, local->dict,
-                                      &local->oldpostparent);
-			if (tmp_dict)
-				dict_unref (tmp_dict);
-		}
-	}
-  
-	return 0;
-}
-
-
-int32_t
-unify_sh_ns_getdents_cbk (call_frame_t *frame,
-			  void *cookie,
-			  xlator_t *this,
-			  int32_t op_ret,
-			  int32_t op_errno,
-			  dir_entry_t *entry,
-			  int32_t count)
-{
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = 0;
-	unsigned long final = 0;
-	dir_entry_t *tmp = GF_CALLOC (1, sizeof (dir_entry_t),
-                                      gf_unify_mt_dir_entry_t);
-	
-	local->sh_struct->entry_list[0] = tmp;
-	local->sh_struct->count_list[0] = count;
-	if (entry) {
-		tmp->next = entry->next;
-		entry->next = NULL;
-	}
-
-	if ((count < UNIFY_SELF_HEAL_GETDENTS_COUNT) || !entry) {
-		final = 1;
-	}
-
-	LOCK (&frame->lock);
-	{
-		/* local->call_count will be '0' till now. make it 1 so, it 
-		   can be UNWIND'ed for the last call. */
-		local->call_count = priv->child_count;
-		if (final)
-			local->flags = 1;
-	}
-	UNLOCK (&frame->lock);
-
-	for (index = 0; index < priv->child_count; index++) 
-	{
-		STACK_WIND_COOKIE (frame,
-				   unify_sh_setdents_cbk, 
-				   (void *)index,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->setdents,
-				   local->fd, GF_SET_DIR_ONLY,
-				   local->sh_struct->entry_list[0], count);
-	}
-
-	return 0;
-}
-
-int32_t 
-unify_sh_ns_setdents_cbk (call_frame_t *frame,
-			  void *cookie,
-			  xlator_t *this,
-			  int32_t op_ret,
-			  int32_t op_errno)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = (long)cookie;
-	dir_entry_t *prev, *entry, *trav;
-
-	LOCK (&frame->lock);
-	{
-		if (local->sh_struct->entry_list[index]) {
-			prev = entry = local->sh_struct->entry_list[index];
-			trav = entry->next;
-			while (trav) {
-				prev->next = trav->next;
-				GF_FREE (trav->name);
-				if (IA_ISLNK (trav->buf.ia_type))
-					GF_FREE (trav->link);
-				GF_FREE (trav);
-				trav = prev->next;
-			}
-			GF_FREE (entry);
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	if (local->sh_struct->count_list[index] < 
-	    UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-		LOCK (&frame->lock);
-		{
-			callcnt = --local->call_count;
-		}
-		UNLOCK (&frame->lock);
-	} else {
-		/* count == size, that means, there are more entries 
-		   to read from */
-		local->sh_struct->offset_list[index] += 
-			UNIFY_SELF_HEAL_GETDENTS_COUNT;
-		STACK_WIND_COOKIE (frame,
-				   unify_sh_getdents_cbk,
-				   cookie,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->getdents,
-				   local->fd,
-				   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-				   local->sh_struct->offset_list[index],
-				   GF_GET_ALL);
-    
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"readdir on (%s) with offset %"PRId64"", 
-			priv->xl_array[index]->name, 
-			local->sh_struct->offset_list[index]);
-	}
-
-	if (!callcnt) {
-		/* All storage nodes have done unified setdents on NS node.
-		 * Now, do getdents from NS and do setdents on storage nodes.
-		 */
-    
-		/* sh_struct->offset_list is no longer required for
-		   storage nodes now */
-		local->sh_struct->offset_list[0] = 0; /* reset */
-
-		STACK_WIND (frame,
-			    unify_sh_ns_getdents_cbk,
-			    NS(this),
-			    NS(this)->fops->getdents,
-			    local->fd,
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-			    0, /* In this call, do send '0' as offset */
-			    GF_GET_DIR_ONLY);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_sh_getdents_cbk -
- */
-int32_t
-unify_sh_getdents_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       dir_entry_t *entry,
-		       int32_t count)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = (long)cookie;
-	dir_entry_t *tmp = NULL; 
-
-	if (op_ret >= 0 && count > 0) {
-		/* There is some dentry found, just send the dentry to NS */
-		tmp = GF_CALLOC (1, sizeof (dir_entry_t),
-                                 gf_unify_mt_dir_entry_t);
-		local->sh_struct->entry_list[index] = tmp;
-		local->sh_struct->count_list[index] = count;
-		if (entry) {
-			tmp->next = entry->next;
-			entry->next = NULL;
-		}
-		STACK_WIND_COOKIE (frame,
-				   unify_sh_ns_setdents_cbk,
-				   cookie,
-				   NS(this),
-				   NS(this)->fops->setdents,
-				   local->fd,
-				   GF_SET_IF_NOT_PRESENT,
-				   local->sh_struct->entry_list[index],
-				   count);
-		return 0;
-	}
-  
-	if (count < UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-		LOCK (&frame->lock);
-		{
-			callcnt = --local->call_count;
-		}
-		UNLOCK (&frame->lock);
-	} else {
-		/* count == size, that means, there are more entries 
-		   to read from */
-		local->sh_struct->offset_list[index] += 
-			UNIFY_SELF_HEAL_GETDENTS_COUNT;
-		STACK_WIND_COOKIE (frame,
-				   unify_sh_getdents_cbk,
-				   cookie,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->getdents,
-				   local->fd,
-				   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-				   local->sh_struct->offset_list[index],
-				   GF_GET_ALL);
-    
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"readdir on (%s) with offset %"PRId64"", 
-			priv->xl_array[index]->name, 
-			local->sh_struct->offset_list[index]);
-	}
-
-	if (!callcnt) {
-		/* All storage nodes have done unified setdents on NS node.
-		 * Now, do getdents from NS and do setdents on storage nodes.
-		 */
-    
-		/* sh_struct->offset_list is no longer required for
-		   storage nodes now */
-		local->sh_struct->offset_list[0] = 0; /* reset */
-
-		STACK_WIND (frame,
-			    unify_sh_ns_getdents_cbk,
-			    NS(this),
-			    NS(this)->fops->getdents,
-			    local->fd,
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-			    0, /* In this call, do send '0' as offset */
-			    GF_GET_DIR_ONLY);
-	}
-
-	return 0;
-}
-
-/**
- * unify_sh_opendir_cbk -
- *
- * @cookie: 
- */
-int32_t 
-unify_sh_opendir_cbk (call_frame_t *frame,
-		      void *cookie,
-		      xlator_t *this,
-		      int32_t op_ret,
-		      int32_t op_errno,
-		      fd_t *fd)
-{
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t index = 0;
-	inode_t *inode = NULL;
-	dict_t *tmp_dict = NULL;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret >= 0) {
-			local->op_ret = op_ret;
-		} else {
-			gf_log (this->name, GF_LOG_WARNING, "failed");
-			local->failed = 1;
-		}
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		local->call_count = priv->child_count + 1;
-    
-		if (!local->failed) {
-			/* send getdents() namespace after finishing
-			   storage nodes */
-			local->call_count--; 
-      
-			fd_bind (fd);
-
-			if (local->call_count) {
-				/* Used as the offset index. This list keeps
-				 * track of offset sent to each node during
-				 * STACK_WIND.
-				 */
-				local->sh_struct->offset_list = 
-					GF_CALLOC (priv->child_count, 
-					           sizeof (off_t),
-                                                   gf_unify_mt_off_t);
-				ERR_ABORT (local->sh_struct->offset_list);
-	
-				local->sh_struct->entry_list = 
-					GF_CALLOC (priv->child_count, 
-						   sizeof (dir_entry_t *),
-                                                   gf_unify_mt_dir_entry_t);
-				ERR_ABORT (local->sh_struct->entry_list);
-
-				local->sh_struct->count_list = 
-					GF_CALLOC (priv->child_count, 
-						   sizeof (int),
-                                                   gf_unify_mt_int);
-				ERR_ABORT (local->sh_struct->count_list);
-
-				/* Send getdents on all the fds */
-				for (index = 0; 
-				     index < priv->child_count; index++) {
-					STACK_WIND_COOKIE (frame,
-							   unify_sh_getdents_cbk,
-							   (void *)(long)index,
-							   priv->xl_array[index],
-							   priv->xl_array[index]->fops->getdents,
-							   local->fd,
-							   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-							   0, /* In this call, do send '0' as offset */
-							   GF_GET_ALL);
-				}
-
-				/* did stack wind, so no need to unwind here */
-				return 0;
-			} /* (local->call_count) */
-		} /* (!local->failed) */
-
-		/* Opendir failed on one node. */ 
-		inode = local->loc1.inode;
-		fd_unref (local->fd);
-		tmp_dict = local->dict;
-
-		unify_local_wipe (local);
-		/* Only 'self-heal' failed, lookup() was successful. */
-		local->op_ret = 0;
-
-		/* This is lookup_cbk ()'s UNWIND. */
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, inode,
-			      &local->stbuf, local->dict, &local->oldpostparent);
-		if (tmp_dict)
-			dict_unref (tmp_dict);
-	}
-
-	return 0;
-}
-
-/**
- * gf_sh_checksum_cbk - 
- * 
- * @frame: frame used in lookup. get a copy of it, and use that copy.
- * @this: pointer to unify xlator.
- * @inode: pointer to inode, for which the consistency check is required.
- *
- */
-int32_t 
-unify_sh_checksum_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       uint8_t *file_checksum,
-		       uint8_t *dir_checksum)
-{
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t index = 0;
-	int32_t callcnt = 0;
-	inode_t *inode = NULL;
-	dict_t *tmp_dict = NULL;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret >= 0) {
-			if (NS(this) == (xlator_t *)cookie) {
-				memcpy (local->sh_struct->ns_file_checksum, 
-					file_checksum, NAME_MAX);
-				memcpy (local->sh_struct->ns_dir_checksum, 
-					dir_checksum, NAME_MAX);
-			} else {
-				if (local->entry_count == 0) {
-					/* Initialize the dir_checksum to be 
-					 * used for comparision with other
-					 * storage nodes. Should be done for
-					 * the first successful call *only*. 
-					 */
-                                        /* Using 'entry_count' as a flag */
-					local->entry_count = 1;
-					memcpy (local->sh_struct->dir_checksum,
-						dir_checksum, NAME_MAX);
-				}
-
-				/* Reply from the storage nodes */
-				for (index = 0; 
-				     index < NAME_MAX; index++) {
-					/* Files should be present in
-					   only one node */
-					local->sh_struct->file_checksum[index] ^= file_checksum[index];
-	  
-					/* directory structure should be
-					   same accross */
-					if (local->sh_struct->dir_checksum[index] != dir_checksum[index])
-						local->failed = 1;
-				}
-			}
-		} 
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		for (index = 0; index < NAME_MAX ; index++) {
-			if (local->sh_struct->file_checksum[index] != 
-			    local->sh_struct->ns_file_checksum[index]) {
-				local->failed = 1;
-				break;
-			}
-			if (local->sh_struct->dir_checksum[index] != 
-			    local->sh_struct->ns_dir_checksum[index]) {
-				local->failed = 1;
-				break;
-			}
-		}
-	
-		if (local->failed) {
-			/* Log it, it should be a rare event */
-			gf_log (this->name, GF_LOG_WARNING, 
-				"Self-heal triggered on directory %s", 
-				local->loc1.path);
-
-			/* Any self heal will be done at directory level */
-			local->call_count = 0;
-			local->op_ret = -1;
-			local->failed = 0;
-      
-			local->fd = fd_create (local->loc1.inode, 
-					       frame->root->pid);
-
-			local->call_count = priv->child_count + 1;
-	
-			for (index = 0; 
-			     index < (priv->child_count + 1); index++) {
-				STACK_WIND_COOKIE (frame,
-						   unify_sh_opendir_cbk,
-						   priv->xl_array[index]->name,
-						   priv->xl_array[index],
-						   priv->xl_array[index]->fops->opendir,
-						   &local->loc1,
-						   local->fd);
-			}
-			/* opendir can be done on the directory */
-			return 0;
-		}
-
-		/* no mismatch */
-		inode = local->loc1.inode;
-		tmp_dict = local->dict;
-
-		unify_local_wipe (local);
-
-		/* This is lookup_cbk ()'s UNWIND. */
-		STACK_UNWIND (frame,
-			      local->op_ret,
-			      local->op_errno,
-			      inode,
-			      &local->stbuf,
-			      local->dict, &local->oldpostparent);
-		if (tmp_dict)
-			dict_unref (tmp_dict);
-	}
-
-	return 0;
-}
-
-/* Foreground self-heal part over */
-
-/* Background self-heal part */
-
-int32_t 
-unify_bgsh_setdents_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	dir_entry_t *prev, *entry, *trav;
-
-	LOCK (&frame->lock);
-	{
-		/* if local->call_count == 0, that means, setdents 
-		   on storagenodes is still pending. */
-		if (local->call_count)
-			callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-
-
-	if (callcnt == 0) {
-		if (local->sh_struct->entry_list[0]) {
-			prev = entry = local->sh_struct->entry_list[0];
-			trav = entry->next;
-			while (trav) {
-				prev->next = trav->next;
-				GF_FREE (trav->name);
-				if (IA_ISLNK (trav->buf.ia_type))
-					GF_FREE (trav->link);
-				GF_FREE (trav);
-				trav = prev->next;
-			}
-			GF_FREE (entry);
-		}
-
-		if (!local->flags) {
-			if (local->sh_struct->count_list[0] >= 
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-				/* count == size, that means, there are more
-				   entries to read from */
-				//local->call_count = 0;
-				local->sh_struct->offset_list[0] += 
-					UNIFY_SELF_HEAL_GETDENTS_COUNT;
-				STACK_WIND (frame,
-					    unify_bgsh_ns_getdents_cbk,
-					    NS(this),
-					    NS(this)->fops->getdents,
-					    local->fd,
-					    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-					    local->sh_struct->offset_list[0],
-					    GF_GET_DIR_ONLY);
-			}		
-		} else {
-			fd_unref (local->fd);
-			unify_local_wipe (local);
-			STACK_DESTROY (frame->root);
-		}
-	}
-
-	return 0;
-}
-
-
-int32_t
-unify_bgsh_ns_getdents_cbk (call_frame_t *frame,
-			    void *cookie,
-			    xlator_t *this,
-			    int32_t op_ret,
-			    int32_t op_errno,
-			    dir_entry_t *entry,
-			    int32_t count)
-{
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = 0;
-	unsigned long final = 0;
-	dir_entry_t *tmp = GF_CALLOC (1, sizeof (dir_entry_t),
-                                      gf_unify_mt_dir_entry_t);
-
-	local->sh_struct->entry_list[0] = tmp;
-	local->sh_struct->count_list[0] = count;
-	if (entry) {
-		tmp->next = entry->next;
-		entry->next = NULL;
-	}
-
-	if ((count < UNIFY_SELF_HEAL_GETDENTS_COUNT) || !entry) {
-		final = 1;
-	}
-
-	LOCK (&frame->lock);
-	{
-		/* local->call_count will be '0' till now. make it 1 so, 
-		   it can be UNWIND'ed for the last call. */
-		local->call_count = priv->child_count;
-		if (final)
-			local->flags = 1;
-	}
-	UNLOCK (&frame->lock);
-
-	for (index = 0; index < priv->child_count; index++) 
-	{
-		STACK_WIND_COOKIE (frame,
-				   unify_bgsh_setdents_cbk, 
-				   (void *)index,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->setdents,
-				   local->fd, GF_SET_DIR_ONLY,
-				   local->sh_struct->entry_list[0], count);
-	}
-
-	return 0;
-}
-
-int32_t 
-unify_bgsh_ns_setdents_cbk (call_frame_t *frame,
-			    void *cookie,
-			    xlator_t *this,
-			    int32_t op_ret,
-			    int32_t op_errno)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = (long)cookie;
-	dir_entry_t *prev, *entry, *trav;
-
-	if (local->sh_struct->entry_list[index]) {
-		prev = entry = local->sh_struct->entry_list[index];
-		if (!entry)
-			return 0;
-		trav = entry->next;
-		while (trav) {
-			prev->next = trav->next;
-			GF_FREE (trav->name);
-			if (IA_ISLNK (trav->buf.ia_type))
-				GF_FREE (trav->link);
-			GF_FREE (trav);
-			trav = prev->next;
-		}
-		GF_FREE (entry);
-	}
-
-	if (local->sh_struct->count_list[index] < 
-	    UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-		LOCK (&frame->lock);
-		{
-			callcnt = --local->call_count;
-		}
-		UNLOCK (&frame->lock);
-	} else {
-		/* count == size, that means, there are more entries 
-		   to read from */
-		local->sh_struct->offset_list[index] += 
-			UNIFY_SELF_HEAL_GETDENTS_COUNT;
-		STACK_WIND_COOKIE (frame,
-				   unify_bgsh_getdents_cbk,
-				   cookie,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->getdents,
-				   local->fd,
-				   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-				   local->sh_struct->offset_list[index],
-				   GF_GET_ALL);
-    
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"readdir on (%s) with offset %"PRId64"", 
-			priv->xl_array[index]->name, 
-			local->sh_struct->offset_list[index]);
-	}
-
-	if (!callcnt) {
-		/* All storage nodes have done unified setdents on NS node.
-		 * Now, do getdents from NS and do setdents on storage nodes.
-		 */
-    
-		/* sh_struct->offset_list is no longer required for
-		   storage nodes now */
-		local->sh_struct->offset_list[0] = 0; /* reset */
-
-		STACK_WIND (frame,
-			    unify_bgsh_ns_getdents_cbk,
-			    NS(this),
-			    NS(this)->fops->getdents,
-			    local->fd,
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-			    0, /* In this call, do send '0' as offset */
-			    GF_GET_DIR_ONLY);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_bgsh_getdents_cbk -
- */
-int32_t
-unify_bgsh_getdents_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 dir_entry_t *entry,
-			 int32_t count)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = (long)cookie;
-	dir_entry_t *tmp = NULL; 
-
-	if (op_ret >= 0 && count > 0) {
-		/* There is some dentry found, just send the dentry to NS */
-		tmp = GF_CALLOC (1, sizeof (dir_entry_t),
-                                 gf_unify_mt_dir_entry_t);
-		local->sh_struct->entry_list[index] = tmp;
-		local->sh_struct->count_list[index] = count;
-		if (entry) {
-			tmp->next = entry->next;
-			entry->next = NULL;
-		}
-		STACK_WIND_COOKIE (frame,
-				   unify_bgsh_ns_setdents_cbk,
-				   cookie,
-				   NS(this),
-				   NS(this)->fops->setdents,
-				   local->fd,
-				   GF_SET_IF_NOT_PRESENT,
-				   local->sh_struct->entry_list[index],
-				   count);
-		return 0;
-	}
-  
-	if (count < UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-		LOCK (&frame->lock);
-		{
-			callcnt = --local->call_count;
-		}
-		UNLOCK (&frame->lock);
-	} else {
-		/* count == size, that means, there are more entries to read from */
-		local->sh_struct->offset_list[index] += 
-			UNIFY_SELF_HEAL_GETDENTS_COUNT;
-
-		STACK_WIND_COOKIE (frame,
-				   unify_bgsh_getdents_cbk,
-				   cookie,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->getdents,
-				   local->fd,
-				   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-				   local->sh_struct->offset_list[index],
-				   GF_GET_ALL);
-    
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"readdir on (%s) with offset %"PRId64"", 
-			priv->xl_array[index]->name, 
-			local->sh_struct->offset_list[index]);
-	}
-
-	if (!callcnt) {
-		/* All storage nodes have done unified setdents on NS node.
-		 * Now, do getdents from NS and do setdents on storage nodes.
-		 */
-    
-		/* sh_struct->offset_list is no longer required for 
-		   storage nodes now */
-		local->sh_struct->offset_list[0] = 0; /* reset */
-
-		STACK_WIND (frame,
-			    unify_bgsh_ns_getdents_cbk,
-			    NS(this),
-			    NS(this)->fops->getdents,
-			    local->fd,
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-			    0, /* In this call, do send '0' as offset */
-			    GF_GET_DIR_ONLY);
-	}
-
-	return 0;
-}
-
-/**
- * unify_bgsh_opendir_cbk -
- *
- * @cookie: 
- */
-int32_t 
-unify_bgsh_opendir_cbk (call_frame_t *frame,
-			void *cookie,
-			xlator_t *this,
-			int32_t op_ret,
-			int32_t op_errno,
-			fd_t *fd)
-{
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int32_t callcnt = 0;
-	int16_t index = 0;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret >= 0) {
-			local->op_ret = op_ret;
-		} else {
-			local->failed = 1;
-		}
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		local->call_count = priv->child_count + 1;
-    
-		if (!local->failed) {
-			/* send getdents() namespace after finishing 
-			   storage nodes */
-			local->call_count--; 
-			callcnt = local->call_count;
-      
-			fd_bind (fd);
-
-			if (local->call_count) {
-				/* Used as the offset index. This list keeps 
-				   track of offset sent to each node during 
-				   STACK_WIND. */
-				local->sh_struct->offset_list = 
-					GF_CALLOC (priv->child_count, 
-					           sizeof (off_t),
-                                                   gf_unify_mt_off_t);
-				ERR_ABORT (local->sh_struct->offset_list);
-	
-				local->sh_struct->entry_list = 
-					GF_CALLOC (priv->child_count, 
-						   sizeof (dir_entry_t *),
-                                                   gf_unify_mt_dir_entry_t);
-				ERR_ABORT (local->sh_struct->entry_list);
-
-				local->sh_struct->count_list = 
-					GF_CALLOC (priv->child_count, 
-						   sizeof (int),
-                                                   gf_unify_mt_int);
-				ERR_ABORT (local->sh_struct->count_list);
-
-				/* Send getdents on all the fds */
-				for (index = 0; 
-				     index < priv->child_count; index++) {
-					STACK_WIND_COOKIE (frame,
-							   unify_bgsh_getdents_cbk,
-							   (void *)(long)index,
-							   priv->xl_array[index],
-							   priv->xl_array[index]->fops->getdents,
-							   local->fd,
-							   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-							   0, /* In this call, do send '0' as offset */
-							   GF_GET_ALL);
-				}
-				/* did a stack wind, so no need to unwind here */
-				return 0;
-			} /* (local->call_count) */
-		} /* (!local->failed) */
-
-		/* Opendir failed on one node. 	 */
-		fd_unref (local->fd);
-		
-		unify_local_wipe (local);
-		STACK_DESTROY (frame->root);
-	}
-
-	return 0;
-}
-
-/**
- * gf_bgsh_checksum_cbk - 
- * 
- * @frame: frame used in lookup. get a copy of it, and use that copy.
- * @this: pointer to unify xlator.
- * @inode: pointer to inode, for which the consistency check is required.
- *
- */
-int32_t 
-unify_bgsh_checksum_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 uint8_t *file_checksum,
-			 uint8_t *dir_checksum)
-{
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t index = 0;
-	int32_t callcnt = 0;
-  
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret >= 0) {
-			if (NS(this) == (xlator_t *)cookie) {
-				memcpy (local->sh_struct->ns_file_checksum, 
-					file_checksum, NAME_MAX);
-				memcpy (local->sh_struct->ns_dir_checksum, 
-					dir_checksum, NAME_MAX);
-			} else {
-				if (local->entry_count == 0) {
-					/* Initialize the dir_checksum to be 
-					 * used for comparision with other 
-					 * storage nodes. Should be done for
-					 * the first successful call *only*. 
-					 */
-					/* Using 'entry_count' as a flag */
-					local->entry_count = 1; 
-					memcpy (local->sh_struct->dir_checksum,
-						dir_checksum, NAME_MAX);
-				}
-
-				/* Reply from the storage nodes */
-				for (index = 0; 
-				     index < NAME_MAX; index++) {
-					/* Files should be present in only 
-					   one node */
-					local->sh_struct->file_checksum[index] ^= file_checksum[index];
-	  
-					/* directory structure should be same 
-					   accross */
-					if (local->sh_struct->dir_checksum[index] != dir_checksum[index])
-						local->failed = 1;
-				}
-			}
-		} 
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		for (index = 0; index < NAME_MAX ; index++) {
-			if (local->sh_struct->file_checksum[index] != 
-			    local->sh_struct->ns_file_checksum[index]) {
-				local->failed = 1;
-				break;
-			}
-			if (local->sh_struct->dir_checksum[index] != 
-			    local->sh_struct->ns_dir_checksum[index]) {
-				local->failed = 1;
-				break;
-			}
-		}
-	
-		if (local->failed) {
-			/* Log it, it should be a rare event */
-			gf_log (this->name, GF_LOG_WARNING, 
-				"Self-heal triggered on directory %s", 
-				local->loc1.path);
-
-			/* Any self heal will be done at the directory level */
-			local->op_ret = -1;
-			local->failed = 0;
-      
-			local->fd = fd_create (local->loc1.inode, 
-					       frame->root->pid);
-			local->call_count = priv->child_count + 1;
-	
-			for (index = 0; 
-			     index < (priv->child_count + 1); index++) {
-				STACK_WIND_COOKIE (frame,
-						   unify_bgsh_opendir_cbk,
-						   priv->xl_array[index]->name,
-						   priv->xl_array[index],
-						   priv->xl_array[index]->fops->opendir,
-						   &local->loc1,
-						   local->fd);
-			}
-      
-			/* opendir can be done on the directory */
-			return 0;
-		}
-
-		/* no mismatch */
-		unify_local_wipe (local);
-		STACK_DESTROY (frame->root);
-	}
-
-	return 0;
-}
-
-/* Background self-heal part over */
-
-
-
-
-/**
- * zr_unify_self_heal - 
- * 
- * @frame: frame used in lookup. get a copy of it, and use that copy.
- * @this: pointer to unify xlator.
- * @inode: pointer to inode, for which the consistency check is required.
- *
- */
-int32_t 
-zr_unify_self_heal (call_frame_t *frame,
-		    xlator_t *this,
-		    unify_local_t *local)
-{
-	unify_private_t *priv = this->private;
-	call_frame_t *bg_frame = NULL;
-	unify_local_t *bg_local = NULL;
-	inode_t *tmp_inode = NULL;
-	dict_t *tmp_dict = NULL;
-	int16_t index = 0;
-  
-	if (local->inode_generation < priv->inode_generation) {
-		/* Any self heal will be done at the directory level */
-		/* Update the inode's generation to the current generation
-		   value. */
-		local->inode_generation = priv->inode_generation;
-		inode_ctx_put (local->loc1.inode, this, 
-			  (uint64_t)(long)local->inode_generation);
-
-		if (priv->self_heal == ZR_UNIFY_FG_SELF_HEAL) {
-			local->op_ret = 0;
-			local->failed = 0;
-			local->call_count = priv->child_count + 1;
-			local->sh_struct = 
-	         	     GF_CALLOC (1, sizeof (struct unify_self_heal_struct),
-                                        gf_unify_mt_unify_self_heal_struct);
-      
-			/* +1 is for NS */
-			for (index = 0; 
-			     index < (priv->child_count + 1); index++) {
-				STACK_WIND_COOKIE (frame,
-						   unify_sh_checksum_cbk,
-						   priv->xl_array[index],
-						   priv->xl_array[index],
-						   priv->xl_array[index]->fops->checksum,
-						   &local->loc1,
-						   0);
-			}
-
-			/* Self-heal in foreground, hence no need 
-			   to UNWIND here */
-			return 0;
-		}
-
-		/* Self Heal done in background */
-		bg_frame = copy_frame (frame);
-		INIT_LOCAL (bg_frame, bg_local);
-		loc_copy (&bg_local->loc1, &local->loc1);
-		bg_local->op_ret = 0;
-		bg_local->failed = 0;
-		bg_local->call_count = priv->child_count + 1;
-		bg_local->sh_struct = 
-			GF_CALLOC (1, sizeof (struct unify_self_heal_struct),
-                                   gf_unify_mt_unify_self_heal_struct);
-    
-		/* +1 is for NS */
-		for (index = 0; index < (priv->child_count + 1); index++) {
-			STACK_WIND_COOKIE (bg_frame,
-					   unify_bgsh_checksum_cbk,
-					   priv->xl_array[index],
-					   priv->xl_array[index],
-					   priv->xl_array[index]->fops->checksum,
-					   &bg_local->loc1,
-					   0);
-		}
-	}
-
-	/* generation number matches, self heal already done or
-	 * self heal done in background: just do STACK_UNWIND 
-	 */
-	tmp_inode = local->loc1.inode;
-	tmp_dict = local->dict;
-
-	unify_local_wipe (local);
-
-	/* This is lookup_cbk ()'s UNWIND. */
-	STACK_UNWIND (frame,
-		      local->op_ret,
-		      local->op_errno,
-		      tmp_inode,
-		      &local->stbuf,
-		      local->dict,
-                      &local->oldpostparent);
-
-	if (tmp_dict)
-		dict_unref (tmp_dict);
-
-	return 0;
-}
-
diff --git a/xlators/cluster/unify/src/unify.c b/xlators/cluster/unify/src/unify.c
deleted file mode 100644
index e50d3274f..000000000
--- a/xlators/cluster/unify/src/unify.c
+++ /dev/null
@@ -1,4589 +0,0 @@
-/*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-/**
- * xlators/cluster/unify:
- *     - This xlator is one of the main translator in GlusterFS, which
- *   actually does the clustering work of the file system. One need to 
- *   understand that, unify assumes file to be existing in only one of 
- *   the child node, and directories to be present on all the nodes. 
- *
- * NOTE:
- *   Now, unify has support for global namespace, which is used to keep a 
- * global view of fs's namespace tree. The stat for directories are taken
- * just from the namespace, where as for files, just 'ia_ino' is taken from
- * Namespace node, and other stat info is taken from the actual storage node.
- * Also Namespace node helps to keep consistant inode for files across 
- * glusterfs (re-)mounts.
- */
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "glusterfs.h"
-#include "unify.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "stack.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include <signal.h>
-#include <libgen.h>
-#include "compat-errno.h"
-#include "compat.h"
-
-#define UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR(_loc) do { \
-  if (!(_loc && _loc->inode)) {                            \
-    STACK_UNWIND (frame, -1, EINVAL, NULL, NULL, NULL);    \
-    return 0;                                              \
-  }                                                        \
-} while(0)
-
-
-#define UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR(_fd) do { \
-  if (!(_fd && !fd_ctx_get (_fd, this, NULL))) {       \
-    STACK_UNWIND (frame, -1, EBADFD, NULL, NULL);      \
-    return 0;                                          \
-  }                                                    \
-} while(0)
-
-#define UNIFY_CHECK_FD_AND_UNWIND_ON_ERR(_fd) do { \
-  if (!_fd) {                                      \
-    STACK_UNWIND (frame, -1, EBADFD, NULL, NULL);  \
-    return 0;                                      \
-  }                                                \
-} while(0)
-
-/**
- * unify_local_wipe - free all the extra allocation of local->* here.
- */
-static void 
-unify_local_wipe (unify_local_t *local)
-{
-	/* Free the strdup'd variables in the local structure */
-	if (local->name) {
-		GF_FREE (local->name);
-	}
-	loc_wipe (&local->loc1);
-	loc_wipe (&local->loc2);
-}
-
-
-
-/*
- * unify_normalize_stats -
- */
-void
-unify_normalize_stats (struct statvfs *buf,
-		       unsigned long bsize,
-		       unsigned long frsize)
-{
-	double factor;
-
-	if (buf->f_bsize != bsize) {
-		factor = ((double) buf->f_bsize) / bsize;
-		buf->f_bsize  = bsize;
-		buf->f_bfree  = (fsblkcnt_t) (factor * buf->f_bfree);
-		buf->f_bavail = (fsblkcnt_t) (factor * buf->f_bavail);
-	}
-  
-	if (buf->f_frsize != frsize) {
-		factor = ((double) buf->f_frsize) / frsize;
-		buf->f_frsize = frsize;
-		buf->f_blocks = (fsblkcnt_t) (factor * buf->f_blocks);
-	}
-}
-
-
-xlator_t *
-unify_loc_subvol (loc_t *loc, xlator_t *this)
-{
-	unify_private_t *priv = NULL;
-	xlator_t        *subvol = NULL;
-	int16_t         *list = NULL;
-	long             index = 0;
-	xlator_t        *subvol_i = NULL;
-	int              ret = 0;
-	uint64_t         tmp_list = 0;
-
-	priv   = this->private;
-	subvol = NS (this);
-
-	if (!IA_ISDIR (loc->inode->ia_type)) {
-		ret = inode_ctx_get (loc->inode, this, &tmp_list);
-		list = (int16_t *)(long)tmp_list;
-		if (!list)
-			goto out;
-
-		for (index = 0; list[index] != -1; index++) {
-			subvol_i = priv->xl_array[list[index]];
-			if (subvol_i != NS (this)) {
-				subvol = subvol_i;
-				break;
-			}
-		}
-	}
-out:
-	return subvol;
-}
-
-
-
-/**
- * unify_statfs_cbk -
- */
-int32_t
-unify_statfs_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  struct statvfs *stbuf)
-{
-	int32_t callcnt = 0;
-	struct statvfs *dict_buf = NULL;
-	unsigned long bsize;
-	unsigned long frsize;
-	unify_local_t *local = (unify_local_t *)frame->local;
-	call_frame_t *prev_frame = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret >= 0) {
-			/* when a call is successfull, add it to local->dict */
-			dict_buf = &local->statvfs_buf;
-
-			if (dict_buf->f_bsize != 0) {
-				bsize  = max (dict_buf->f_bsize, 
-					      stbuf->f_bsize);
-
-				frsize = max (dict_buf->f_frsize, 
-					      stbuf->f_frsize);
-				unify_normalize_stats(dict_buf, bsize, frsize);
-				unify_normalize_stats(stbuf, bsize, frsize);
-			} else {
-				dict_buf->f_bsize   = stbuf->f_bsize;
-				dict_buf->f_frsize  = stbuf->f_frsize;
-			}
-      
-			dict_buf->f_blocks += stbuf->f_blocks;
-			dict_buf->f_bfree  += stbuf->f_bfree;
-			dict_buf->f_bavail += stbuf->f_bavail;
-			dict_buf->f_files  += stbuf->f_files;
-			dict_buf->f_ffree  += stbuf->f_ffree;
-			dict_buf->f_favail += stbuf->f_favail;
-			dict_buf->f_fsid    = stbuf->f_fsid;
-			dict_buf->f_flag    = stbuf->f_flag;
-			dict_buf->f_namemax = stbuf->f_namemax;
-			local->op_ret = op_ret;
-		} else {
-			/* fop on storage node has failed due to some error */
-			if (op_errno != ENOTCONN) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"child(%s): %s", 
-					prev_frame->this->name, 
-					strerror (op_errno));
-			}
-			local->op_errno = op_errno;
-		}
-		callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      &local->statvfs_buf);
-	}
-
-	return 0;
-}
-
-/**
- * unify_statfs -
- */
-int32_t
-unify_statfs (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc)
-{
-	unify_local_t *local = NULL;
-	xlator_list_t *trav = this->children;
-
-	INIT_LOCAL (frame, local);
-	local->call_count = ((unify_private_t *)this->private)->child_count;
-
-	while(trav) {
-		STACK_WIND (frame,
-			    unify_statfs_cbk,
-			    trav->xlator,
-			    trav->xlator->fops->statfs,
-			    loc);
-		trav = trav->next;
-	}
-
-	return 0;
-}
-
-/**
- * unify_buf_cbk - 
- */
-int32_t
-unify_buf_cbk (call_frame_t *frame,
-	       void *cookie,
-	       xlator_t *this,
-	       int32_t op_ret,
-	       int32_t op_errno,
-	       struct iatt *buf)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"%s(): child(%s): path(%s): %s", 
-				gf_fop_list[frame->root->op],
-				prev_frame->this->name, 
-				(local->loc1.path)?local->loc1.path:"", 
-				strerror (op_errno));
-
-			local->op_errno = op_errno;
-			if ((op_errno == ENOENT) && priv->optimist) 
-				local->op_ret = 0;
-		}
-
-		if (op_ret >= 0) {
-			local->op_ret = 0;
-
-			if (NS (this) == prev_frame->this) {
-				local->ia_ino = buf->ia_ino;
-				/* If the entry is directory, get the stat
-				   from NS node */
-				if (IA_ISDIR (buf->ia_type) || 
-				    !local->stbuf.ia_blksize) {
-					local->stbuf = *buf;
-				}
-			}
-
-			if ((!IA_ISDIR (buf->ia_type)) && 
-			    (NS (this) != prev_frame->this)) {
-				/* If file, take the stat info from Storage 
-				   node. */
-				local->stbuf = *buf;
-			}
-		}
-	}
-	UNLOCK (&frame->lock);
-    
-	if (!callcnt) {
-		/* If the inode number is not filled, operation should
-		   fail */
-		if (!local->ia_ino)
-			local->op_ret = -1;
-
-		local->stbuf.ia_ino = local->ia_ino;
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      &local->stbuf);
-	}
-
-	return 0;
-}
-
-#define check_if_dht_linkfile(s) \
-        ((st_mode_from_ia (s->ia_prot, s->ia_type) & ~S_IFMT) == S_ISVTX)
-
-/**
- * unify_lookup_cbk - 
- */
-int32_t 
-unify_lookup_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  inode_t *inode,
-		  struct iatt *buf,
-		  dict_t *dict,
-                  struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	inode_t *tmp_inode = NULL;
-	dict_t *local_dict = NULL;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
- 
-		if (op_ret == -1) {
-                        if (local->revalidate &&
-                            (op_errno == ESTALE)) {
-                                /* ESTALE takes priority */
-                                local->op_errno = op_errno;
-                                local->failed = 1;
-                        }
-
-			if ((op_errno != ENOTCONN) 
-                            && (op_errno != ENOENT)
-                            && (local->op_errno != ESTALE)) {
-                                /* if local->op_errno is already ESTALE, then
-                                 * ESTALE has to propogated to the parent first.
-                                 * do not enter here.
-                                 */
-				gf_log (this->name, GF_LOG_ERROR,
-					"child(%s): path(%s): %s", 
-					priv->xl_array[(long)cookie]->name, 
-					local->loc1.path, strerror (op_errno));
-                                local->op_errno = op_errno;
-				local->failed = 1;
-
-			} else if (local->revalidate && 
-                                   (local->op_errno != ESTALE) &&
-				   !(priv->optimist && (op_errno == ENOENT))) {
-
-				gf_log (this->name, 
-					(op_errno == ENOTCONN) ? 
-					GF_LOG_DEBUG:GF_LOG_ERROR,
-					"child(%s): path(%s): %s", 
-					priv->xl_array[(long)cookie]->name, 
-					local->loc1.path, strerror (op_errno));
-				local->op_errno = op_errno;
-				local->failed = 1;
-			}
-		}
-
-		if (op_ret == 0) {
-			local->op_ret = 0; 
-			
-			if (check_if_dht_linkfile(buf)) {
-				gf_log (this->name, GF_LOG_CRITICAL,
-					"file %s may be DHT link file on %s, "
-					"make sure the backend is not shared "
-					"between unify and DHT", 
-					local->loc1.path, 
-					priv->xl_array[(long)cookie]->name);
-			}
-
-			if (local->stbuf.ia_type && local->stbuf.ia_blksize) {
-				/* make sure we already have a stbuf
-				   stored in local->stbuf */
-				if (IA_ISDIR (local->stbuf.ia_type) && 
-				    !IA_ISDIR (buf->ia_type)) {
-					gf_log (this->name, GF_LOG_CRITICAL, 
-						"[CRITICAL] '%s' is directory "
-						"on namespace, non-directory "
-						"on node '%s', returning EIO",
-						local->loc1.path, 
-						priv->xl_array[(long)cookie]->name);
-					local->return_eio = 1;
-				}
-				if (!IA_ISDIR (local->stbuf.ia_type) && 
-				    IA_ISDIR (buf->ia_type)) {
-					gf_log (this->name, GF_LOG_CRITICAL, 
-						"[CRITICAL] '%s' is directory "
-						"on node '%s', non-directory "
-						"on namespace, returning EIO",
-						local->loc1.path, 
-						priv->xl_array[(long)cookie]->name);
-					local->return_eio = 1;
-				}
-			}
-	
-			if (!local->revalidate && !IA_ISDIR (buf->ia_type)) {
-				/* This is the first time lookup on file*/
-				if (!local->list) {
-					/* list is not allocated, allocate 
-					   the max possible range */
-					local->list = GF_CALLOC (1, 2 * (priv->child_count + 2),
-                                                                 gf_unify_mt_int16_t);
-					if (!local->list) {
-						gf_log (this->name, 
-							GF_LOG_CRITICAL, 
-							"Not enough memory");
-						STACK_UNWIND (frame, -1, 
-							      ENOMEM, inode, 
-							      NULL, NULL, NULL);
-						return 0;
-					}
-				}
-				/* update the index of the list */
-				local->list [local->index++] = 
-					(int16_t)(long)cookie;
-			}
-                        
-                        if (!local->revalidate && IA_ISDIR (buf->ia_type)) {
-                                /* fresh lookup of a directory */
-                                inode_ctx_put (local->loc1.inode, this, 
-                                               priv->inode_generation);
-                        }
-
-			if ((!local->dict) && dict &&
-			    (priv->xl_array[(long)cookie] != NS(this)))	{
-				local->dict = dict_ref (dict);
-			}
-
-			/* index of NS node is == total child count */
-			if (priv->child_count == (int16_t)(long)cookie) {
-				/* Take the inode number from namespace */
-				local->ia_ino = buf->ia_ino;
-				if (IA_ISDIR (buf->ia_type) || 
-				    !(local->stbuf.ia_blksize)) {
-					local->stbuf = *buf;
-                                        local->oldpostparent = *postparent;
-				}
-			} else if (!IA_ISDIR (buf->ia_type)) {
-				/* If file, then get the stat from 
-				   storage node */
-				local->stbuf = *buf;
-			}
-
-			if (local->ia_nlink < buf->ia_nlink) {
-				local->ia_nlink = buf->ia_nlink;
-			}
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		local_dict = local->dict;
-		if (local->return_eio) {
-			gf_log (this->name, GF_LOG_CRITICAL, 
-				"[CRITICAL] Unable to fix the path (%s) with "
-				"self-heal, try manual verification. "
-				"returning EIO.", local->loc1.path);
-			unify_local_wipe (local);
-			STACK_UNWIND (frame, -1, EIO, inode, NULL, NULL);
-			if (local_dict)	{
-				dict_unref (local_dict);
-			}
-			return 0;
-		}
-
-		if (!local->stbuf.ia_blksize) {
-			/* Inode not present */
-			local->op_ret = -1;
-		} else {
-			if (!local->revalidate && 
-			    !IA_ISDIR (local->stbuf.ia_type)) { 
-				/* If its a file, big array is useless, 
-				   allocate the smaller one */
-				int16_t *list = NULL;
-				list = GF_CALLOC (1, 2 * (local->index + 1),
-                                                  gf_unify_mt_int16_t);
-				ERR_ABORT (list);
-				memcpy (list, local->list, 2 * local->index);
-				/* Make the end of the list as -1 */
-				GF_FREE (local->list);
-				local->list = list;
-				local->list [local->index] = -1;
-				/* Update the inode's ctx with proper array */
-				/* TODO: log on failure */
-				inode_ctx_put (local->loc1.inode, this, 
-					       (uint64_t)(long)local->list);
-			}
-
-			if (IA_ISDIR(local->loc1.inode->ia_type)) {
-				/* lookup is done for directory */
-				if (local->failed && priv->self_heal) {
-					/* Triggering self-heal */
-                                        /* means, self-heal required for this 
-					   inode */
-					local->inode_generation = 0; 
-					priv->inode_generation++;
-				}
-			} else {
-				local->stbuf.ia_ino = local->ia_ino;
-			}
-	  
-			local->stbuf.ia_nlink = local->ia_nlink;
-		}
-		if (local->op_ret == -1) {
-			if (!local->revalidate && local->list)
-				GF_FREE (local->list);
-		}
-
-		if ((local->op_ret >= 0) && local->failed && 
-		    local->revalidate) {
-			/* Done revalidate, but it failed */
-			if ((op_errno != ENOTCONN) 
-                            && (local->op_errno != ESTALE)) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"Revalidate failed for path(%s): %s", 
-					local->loc1.path, strerror (op_errno));
-			}
-			local->op_ret = -1;
-		}
-
-		if ((priv->self_heal && !priv->optimist) && 
-		    (!local->revalidate && (local->op_ret == 0) && 
-		     IA_ISDIR(local->stbuf.ia_type))) {
-			/* Let the self heal be done here */
-			zr_unify_self_heal (frame, this, local);
-			local_dict = NULL;
-		} else {
-                        if (local->failed) {
-                                /* NOTE: directory lookup is sent to all 
-                                 * subvolumes and success from a subvolume
-                                 * might set local->op_ret to 0 (zero) */
-                                local->op_ret = -1;
-                        }
-
-			/* either no self heal, or op_ret == -1 (failure) */
-			tmp_inode = local->loc1.inode;
-			unify_local_wipe (local);
-			STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-				      tmp_inode, &local->stbuf, local->dict,
-                                      &local->oldpostparent);
-		}
-		if (local_dict) {
-			dict_unref (local_dict);
-		}
-	}
-  
-	return 0;
-}
-
-/**
- * unify_lookup - 
- */
-int32_t 
-unify_lookup (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc,
-	      dict_t *xattr_req)
-{
-	unify_local_t *local = NULL;
-	unify_private_t *priv = this->private;
-	int16_t *list = NULL;
-	long index = 0;
-
-	if (!(loc && loc->inode)) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"%s: Argument not right", loc?loc->path:"(null)");
-		STACK_UNWIND (frame, -1, EINVAL, NULL, NULL, NULL, NULL);
-		return 0;
-	}
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, loc->inode, NULL, NULL, NULL);
-		return 0;
-	}
-        
-        if (inode_ctx_get (loc->inode, this, NULL)
-            && IA_ISDIR (loc->inode->ia_type)) {
-                local->revalidate = 1;
-        }
-
-	if (!inode_ctx_get (loc->inode, this, NULL) && 
-	    loc->inode->ia_type && 
-	    !IA_ISDIR (loc->inode->ia_type)) {
-		uint64_t tmp_list = 0;
-		/* check if revalidate or fresh lookup */
-		inode_ctx_get (loc->inode, this, &tmp_list);
-		local->list = (int16_t *)(long)tmp_list;
-	}
-
-	if (local->list) {
-		list = local->list;
-		for (index = 0; list[index] != -1; index++);
-		if (index != 2) {
-			if (index < 2) {
-				gf_log (this->name, GF_LOG_ERROR,
-					"returning ESTALE for %s: file "
-					"count is %ld", loc->path, index);
-				/* Print where all the file is present */
-				for (index = 0; 
-				     local->list[index] != -1; index++) {
-					gf_log (this->name, GF_LOG_ERROR, 
-						"%s: found on %s", loc->path, 
-						priv->xl_array[list[index]]->name);
-				}
-				unify_local_wipe (local);
-				STACK_UNWIND (frame, -1, ESTALE, 
-					      NULL, NULL, NULL, NULL);
-				return 0;  
-			} else {
-				/* There are more than 2 presences */
-				/* Just log and continue */
-				gf_log (this->name, GF_LOG_ERROR,
-					"%s: file count is %ld", 
-					loc->path, index);
-				/* Print where all the file is present */
-				for (index = 0; 
-				     local->list[index] != -1; index++) {
-					gf_log (this->name, GF_LOG_ERROR, 
-						"%s: found on %s", loc->path, 
-						priv->xl_array[list[index]]->name);
-				}
-			}
-		}
-      
-		/* is revalidate */
-		local->revalidate = 1;
-      
-		for (index = 0; list[index] != -1; index++)
-			local->call_count++;
-      
-		for (index = 0; list[index] != -1; index++) {
-			char need_break = (list[index+1] == -1);
-			STACK_WIND_COOKIE (frame,
-					   unify_lookup_cbk,
-					   (void *)(long)list[index], //cookie
-					   priv->xl_array [list[index]],
-					   priv->xl_array [list[index]]->fops->lookup,
-					   loc,
-					   xattr_req);
-			if (need_break)
-				break;
-		}
-	} else {
-		if (loc->inode->ia_type) {
-			if (inode_ctx_get (loc->inode, this, NULL)) {
-				inode_ctx_get (loc->inode, this, 
-					       &local->inode_generation);
-			}
-		}
-		/* This is first call, there is no list */
-		/* call count should be all child + 1 namespace */
-		local->call_count = priv->child_count + 1;
-      
-		for (index = 0; index <= priv->child_count; index++) {
-			STACK_WIND_COOKIE (frame,
-					   unify_lookup_cbk,
-					   (void *)index, //cookie
-					   priv->xl_array[index],
-					   priv->xl_array[index]->fops->lookup,
-					   loc,
-					   xattr_req);
-		}
-	}
-
-	return 0;
-}
-
-/**
- * unify_stat - if directory, get the stat directly from NameSpace child.
- *     if file, check for a hint and send it only there (also to NS).
- *     if its a fresh stat, then do it on all the nodes.
- *
- * NOTE: for all the call, sending cookie as xlator pointer, which will be 
- *       used in cbk.
- */
-int32_t
-unify_stat (call_frame_t *frame,
-	    xlator_t *this,
-	    loc_t *loc)
-{
-	unify_local_t *local = NULL;
-	unify_private_t *priv = this->private;
-	int16_t index = 0;
-	int16_t *list = NULL;
-	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, NULL);
-		return 0;
-	}
-	local->ia_ino = loc->inode->ino;
-	if (IA_ISDIR (loc->inode->ia_type)) {
-		/* Directory */
-		local->call_count = 1;
-		STACK_WIND (frame, unify_buf_cbk, NS(this),
-			    NS(this)->fops->stat, loc);
-	} else {
-		/* File */
-		inode_ctx_get (loc->inode, this, &tmp_list);
-    		list = (int16_t *)(long)tmp_list;
-
-		for (index = 0; list[index] != -1; index++)
-			local->call_count++;
-    
-		for (index = 0; list[index] != -1; index++) {
-			char need_break = (list[index+1] == -1);
-			STACK_WIND (frame,
-				    unify_buf_cbk,
-				    priv->xl_array[list[index]],
-				    priv->xl_array[list[index]]->fops->stat,
-				    loc);
-			if (need_break)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-/**
- * unify_access_cbk -
- */
-int32_t
-unify_access_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-
-/**
- * unify_access - Send request to only namespace, which has all the 
- *      attributes set for the file.
- */
-int32_t
-unify_access (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc,
-	      int32_t mask)
-{
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	STACK_WIND (frame,
-		    unify_access_cbk,
-		    NS(this),
-		    NS(this)->fops->access,
-		    loc,
-		    mask);
-
-	return 0;
-}
-
-int32_t
-unify_mkdir_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-		 inode_t *inode,
-                 struct iatt *buf,
-                 struct iatt *preparent,
-                 struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	inode_t *tmp_inode = NULL;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-  
-		if ((op_ret == -1) && !(priv->optimist && 
-					(op_errno == ENOENT || 
-					 op_errno == EEXIST))) {
-			/* TODO: Decrement the inode_generation of 
-			 * this->inode's parent inode, hence the missing 
-			 * directory is created properly by self-heal. 
-			 * Currently, there is no way to get the parent 
-			 * inode directly.
-			 */
-			gf_log (this->name, GF_LOG_ERROR,
-				"child(%s): path(%s): %s", 
-				priv->xl_array[(long)cookie]->name, 
-				local->loc1.path, strerror (op_errno));
-			if (op_errno != EEXIST)
-				local->failed = 1;
-			local->op_errno = op_errno;
-		}
-  
-		if (op_ret >= 0)
-			local->op_ret = 0;
-
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		if (!local->failed) {
-			inode_ctx_put (local->loc1.inode, this, 
-				       priv->inode_generation);
-		}
-		
-		tmp_inode = local->loc1.inode;
-		unify_local_wipe (local);
-
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      tmp_inode, &local->stbuf,
-                              &local->oldpreparent, &local->oldpostparent);
-	}
-
-	return 0;
-}
-
-/**
- * unify_ns_mkdir_cbk -
- */
-int32_t
-unify_ns_mkdir_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    inode_t *inode,
-                    struct iatt *buf,
-                    struct iatt *preparent,
-                    struct iatt *postparent)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	long index = 0;
-
-	if (op_ret == -1) {
-		/* No need to send mkdir request to other servers, 
-		 * as namespace action failed 
-		 */
-		gf_log (this->name, GF_LOG_ERROR,
-			"namespace: path(%s): %s", 
-			local->name, strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, inode, NULL,
-                              NULL, NULL);
-		return 0;
-	}
-  
-	/* Create one inode for this entry */
-	local->op_ret = 0;
-	local->stbuf = *buf;
-
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	local->call_count = priv->child_count;
-
-	/* Send mkdir request to all the nodes now */
-	for (index = 0; index < priv->child_count; index++) {
-		STACK_WIND_COOKIE (frame,
-				   unify_mkdir_cbk,
-				   (void *)index, //cookie
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->mkdir,
-				   &local->loc1,
-				   local->mode);
-	}
-  
-	return 0;
-}
-
-
-/**
- * unify_mkdir -
- */
-int32_t
-unify_mkdir (call_frame_t *frame,
-	     xlator_t *this,
-	     loc_t *loc,
-	     mode_t mode)
-{
-	unify_local_t *local = NULL;
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	local->mode = mode;
-
-	loc_copy (&local->loc1, loc);
-
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, NULL, NULL);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_ns_mkdir_cbk,
-		    NS(this),
-		    NS(this)->fops->mkdir,
-		    loc,
-		    mode);
-	return 0;
-}
-
-/**
- * unify_rmdir_cbk -
- */
-int32_t
-unify_rmdir_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-                 struct iatt *preparent,
-                 struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret == 0 || (priv->optimist && (op_errno == ENOENT)))
-			local->op_ret = 0;
-		if (op_ret == -1)
-			local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno,
-                              &local->oldpreparent, &local->oldpostparent);
-	}
-
-	return 0;
-}
-
-/**
- * unify_ns_rmdir_cbk -
- */
-int32_t
-unify_ns_rmdir_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-                    struct iatt *preparent,
-                    struct iatt *postparent)
-{
-	int16_t index = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-  
-	if (op_ret == -1) {
-		/* No need to send rmdir request to other servers, 
-		 * as namespace action failed 
-		 */
-		gf_log (this->name, 
-			((op_errno != ENOTEMPTY) ? 
-			 GF_LOG_ERROR : GF_LOG_DEBUG),
-			"namespace: path(%s): %s", 
-			local->loc1.path, strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, NULL, NULL);
-		return 0;
-	}
-
-	local->call_count = priv->child_count;
-
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	for (index = 0; index < priv->child_count; index++) {
-		STACK_WIND (frame,
-			    unify_rmdir_cbk,
-			    priv->xl_array[index],
-			    priv->xl_array[index]->fops->rmdir,
-			    &local->loc1);
-	}
-
-	return 0;
-}
-
-/**
- * unify_rmdir -
- */
-int32_t
-unify_rmdir (call_frame_t *frame,
-	     xlator_t *this,
-	     loc_t *loc)
-{
-	unify_local_t *local = NULL;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-
-	loc_copy (&local->loc1, loc);
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, NULL, NULL);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_ns_rmdir_cbk,
-		    NS(this),
-		    NS(this)->fops->rmdir,
-		    loc);
-
-	return 0;
-}
-
-/**
- * unify_open_cbk -
- */
-int32_t
-unify_open_cbk (call_frame_t *frame,
-		void *cookie,
-		xlator_t *this,
-		int32_t op_ret,
-		int32_t op_errno,
-		fd_t *fd)
-{
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret >= 0) {
-			local->op_ret = op_ret;
-			if (NS(this) != (xlator_t *)cookie) {
-				/* Store child node's ptr, used in 
-				   all the f*** / FileIO calls */
-				fd_ctx_set (fd, this, (uint64_t)(long)cookie);
-			}
-		}
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			local->failed = 1;
-		}
-		callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		if ((local->failed == 1) && (local->op_ret >= 0)) {
-			local->call_count = 1;
-			/* return -1 to user */
-			local->op_ret = -1;
-			//local->op_errno = EIO; 
-      
-			if (!fd_ctx_get (local->fd, this, NULL)) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"Open success on child node, "
-					"failed on namespace");
-			} else {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"Open success on namespace, "
-					"failed on child node");
-			}
-		}
-
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, 
-			      local->op_errno, local->fd);
-	}
-
-	return 0;
-}
-
-#ifdef GF_DARWIN_HOST_OS
-/**
- * unify_create_lookup_cbk - 
- */
-int32_t 
-unify_open_lookup_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       inode_t *inode,
-		       struct iatt *buf,
-		       dict_t *dict,
-		       struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	int16_t index = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if ((op_ret == -1) && (op_errno != ENOENT)) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"child(%s): path(%s): %s", 
-				priv->xl_array[(long)cookie]->name, 
-				local->loc1.path, strerror (op_errno));
-			local->op_errno = op_errno;
-		}
-    
-		if (op_ret >= 0) {
-			local->op_ret = op_ret; 
-			local->index++;
-			if (NS(this) == priv->xl_array[(long)cookie]) {
-				local->list[0] = (int16_t)(long)cookie;
-			} else {
-				local->list[1] = (int16_t)(long)cookie;
-			}
-			if (IA_ISDIR (buf->ia_type))
-				local->failed = 1;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		int16_t file_list[3] = {0,};
-		local->op_ret = -1;
-
-		file_list[0] = local->list[0];
-		file_list[1] = local->list[1];
-		file_list[2] = -1;
-
-		if (local->index != 2) {
-			/* Lookup failed, can't do open */
-			gf_log (this->name, GF_LOG_ERROR,
-				"%s: present on %d nodes", 
-				local->name, local->index);
-
-			if (local->index < 2) {
-				unify_local_wipe (local);
-				gf_log (this->name, GF_LOG_ERROR,
-					"returning as file found on less "
-					"than 2 nodes");
-				STACK_UNWIND (frame, local->op_ret, 
-					      local->op_errno, local->fd);
-				return 0;
-			}
-		}
-
-		if (local->failed) {
-			/* Open on directory, return EISDIR */
-			unify_local_wipe (local);
-			STACK_UNWIND (frame, -1, EISDIR, local->fd);
-			return 0;
-		}
-
-		/* Everything is perfect :) */    
-		local->call_count = 2;
-    
-		for (index = 0; file_list[index] != -1; index++) {
-			char need_break = (file_list[index+1] == -1);
-			STACK_WIND_COOKIE (frame,
-					   unify_open_cbk,
-					   priv->xl_array[file_list[index]],
-					   priv->xl_array[file_list[index]],
-					   priv->xl_array[file_list[index]]->fops->open,
-					   &local->loc1,
-					   local->flags,
-					   local->fd, local->wbflags);
-			if (need_break)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-
-int32_t
-unify_open_readlink_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 const char *path,
-			 struct iatt *sbuf)
-{
-	int16_t index = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1) {
-		STACK_UNWIND (frame, -1, ENOENT);
-		return 0;
-	}
-
-	if (path[0] == '/') {
-		local->name = gf_strdup (path);
-		ERR_ABORT (local->name);
-	} else {
-		char *tmp_str = gf_strdup (local->loc1.path);
-		char *tmp_base = dirname (tmp_str);
-		local->name = GF_CALLOC (1, ZR_PATH_MAX, gf_unify_mt_char);
-		strcpy (local->name, tmp_base);
-		strncat (local->name, "/", 1);
-		strcat (local->name, path);
-		GF_FREE (tmp_str);
-	}
-  
-	local->list = GF_CALLOC (1, sizeof (int16_t) * 3,
-                                 gf_unify_mt_int16_t);
-	ERR_ABORT (local->list);
-	local->call_count = priv->child_count + 1;
-	local->op_ret = -1;
-	for (index = 0; index <= priv->child_count; index++) {
-		/* Send the lookup to all the nodes including namespace */
-		STACK_WIND_COOKIE (frame,
-				   unify_open_lookup_cbk,
-				   (void *)(long)index,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->lookup,
-				   &local->loc1,
-				   NULL);
-	}
-
-	return 0;
-}
-#endif /* GF_DARWIN_HOST_OS */
-
-/**
- * unify_open - 
- */
-int32_t
-unify_open (call_frame_t *frame,
-	    xlator_t *this,
-	    loc_t *loc,
-	    int32_t flags,
-	    fd_t *fd,
-            int32_t wbflags)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = NULL;
-	int16_t *list = NULL;
-	int16_t index = 0;
-	int16_t file_list[3] = {0,};
-	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Init */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-	local->fd    = fd;
-	local->flags = flags;
-        local->wbflags = wbflags;
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	local->list = list;
-	file_list[0] = priv->child_count; /* Thats namespace */
-	file_list[2] = -1;
-	for (index = 0; list[index] != -1; index++) {
-		local->call_count++;
-		if (list[index] != priv->child_count)
-			file_list[1] = list[index];
-	}
-
-	if (local->call_count != 2) {
-		/* If the lookup was done for file */
-		gf_log (this->name, GF_LOG_ERROR,
-			"%s: entry_count is %d",
-			loc->path, local->call_count);
-		for (index = 0; local->list[index] != -1; index++)
-			gf_log (this->name, GF_LOG_ERROR, "%s: found on %s",
-				loc->path, priv->xl_array[list[index]]->name);
-
-		if (local->call_count < 2) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"returning EIO as file found on onlyone node");
-			STACK_UNWIND (frame, -1, EIO, fd);
-			return 0;
-		}
-	}
-
-#ifdef GF_DARWIN_HOST_OS
-	/* Handle symlink here */
-	if (IA_ISLNK (loc->inode->ia_type)) {
-		/* Callcount doesn't matter here */
-		STACK_WIND (frame,
-			    unify_open_readlink_cbk,
-			    NS(this),
-			    NS(this)->fops->readlink,
-			    loc, ZR_PATH_MAX);
-		return 0;
-	}
-#endif /* GF_DARWIN_HOST_OS */
-
-	local->call_count = 2;
-	for (index = 0; file_list[index] != -1; index++) {
-		char need_break = (file_list[index+1] == -1);
-		STACK_WIND_COOKIE (frame,
-				   unify_open_cbk,
-				   priv->xl_array[file_list[index]], //cookie
-				   priv->xl_array[file_list[index]],
-				   priv->xl_array[file_list[index]]->fops->open,
-				   loc,
-				   flags,
-				   fd, wbflags);
-		if (need_break)
-			break;
-	}
-
-	return 0;
-}
-
-
-int32_t 
-unify_create_unlink_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-                         struct iatt *preparent,
-                         struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-	inode_t *inode = local->loc1.inode;
-
-	unify_local_wipe (local);
-
-	STACK_UNWIND (frame, local->op_ret, local->op_errno, local->fd, 
-		      inode, &local->stbuf,
-                      &local->oldpreparent, &local->oldpostparent);
-  
-	return 0;
-}
-
-/**
- * unify_create_open_cbk -
- */
-int32_t
-unify_create_open_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       fd_t *fd)
-{
-	int ret = 0;
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-	inode_t *inode = NULL;
-	xlator_t *child = NULL;
-	uint64_t tmp_value = 0;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret >= 0) {
-			local->op_ret = op_ret;
-			if (NS(this) != (xlator_t *)cookie) {
-				/* Store child node's ptr, used in all 
-				   the f*** / FileIO calls */
-				/* TODO: log on failure */
-				ret = fd_ctx_get (fd, this, &tmp_value);
-				cookie = (void *)(long)tmp_value;
-			} else {
-				/* NOTE: open successful on namespace.
-				 *       fd's ctx can be used to identify open 
-				 *       failure on storage subvolume. cool 
-				 *       ide ;) */
-				local->failed = 0;
-			}
-		} else {
-			gf_log (this->name, GF_LOG_ERROR,
-				"child(%s): path(%s): %s", 
-				((xlator_t *)cookie)->name,
-				local->loc1.path, strerror (op_errno));
-			local->op_errno = op_errno;
-			local->failed = 1;
-		}
-		callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		if (local->failed == 1 && (local->op_ret >= 0)) {
-			local->call_count = 1;
-			/* return -1 to user */
-			local->op_ret = -1;
-			local->op_errno = EIO;
-			local->fd = fd;
-			local->call_count = 1;
-
-			if (!fd_ctx_get (local->fd, this, &tmp_value)) {
-				child = (xlator_t *)(long)tmp_value;
-
-				gf_log (this->name, GF_LOG_ERROR, 
-					"Create success on child node, "
-					"failed on namespace");
-
-				STACK_WIND (frame,
-					    unify_create_unlink_cbk,
-					    child,
-					    child->fops->unlink,
-					    &local->loc1);
-			} else {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"Create success on namespace, "
-					"failed on child node");
-
-				STACK_WIND (frame,
-					    unify_create_unlink_cbk,
-					    NS(this),
-					    NS(this)->fops->unlink,
-					    &local->loc1);
-			}
-			return 0;
-		}
-		inode = local->loc1.inode;
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, fd,
-			      inode, &local->stbuf,
-                              &local->oldpreparent, &local->oldpostparent);
-	}
-	return 0;
-}
-
-/**
- * unify_create_lookup_cbk - 
- */
-int32_t 
-unify_create_lookup_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 inode_t *inode,
-			 struct iatt *buf,
-			 dict_t *dict,
-                         struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	int16_t index = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"child(%s): path(%s): %s", 
-				priv->xl_array[(long)cookie]->name, 
-				local->loc1.path, strerror (op_errno));
-			local->op_errno = op_errno;
-			local->failed = 1;
-		}
-
-		if (op_ret >= 0) {
-			local->op_ret = op_ret; 
-			local->list[local->index++] = (int16_t)(long)cookie;
-			if (NS(this) == priv->xl_array[(long)cookie]) {
-				local->ia_ino = buf->ia_ino;
-			} else {
-				local->stbuf = *buf;
-			}
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		int16_t *list = local->list;
-		int16_t file_list[3] = {0,};
-		local->op_ret = -1;
-
-		local->list [local->index] = -1;
-		file_list[0] = list[0];
-		file_list[1] = list[1];
-		file_list[2] = -1;
-
-		local->stbuf.ia_ino = local->ia_ino;
-		/* TODO: log on failure */
-		inode_ctx_put (local->loc1.inode, this, 
-			       (uint64_t)(long)local->list);
-
-		if (local->index != 2) {
-			/* Lookup failed, can't do open */
-			gf_log (this->name, GF_LOG_ERROR,
-				"%s: present on %d nodes", 
-				local->loc1.path, local->index);
-			file_list[0] = priv->child_count;
-			for (index = 0; list[index] != -1; index++) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"%s: found on %s", local->loc1.path, 
-					priv->xl_array[list[index]]->name);
-				if (list[index] != priv->child_count)
-					file_list[1] = list[index];
-			}
-
-			if (local->index < 2) {
-				unify_local_wipe (local);
-				gf_log (this->name, GF_LOG_ERROR,
-					"returning EIO as file found on "
-					"only one node");
-				STACK_UNWIND (frame, -1, EIO, 
-					      local->fd, inode, NULL,
-                                              NULL, NULL);
-				return 0;
-			}
-		}
-		/* Everything is perfect :) */    
-		local->call_count = 2;
-    
-		for (index = 0; file_list[index] != -1; index++) {
-			char need_break = (file_list[index+1] == -1);
-			STACK_WIND_COOKIE (frame,
-					   unify_create_open_cbk,
-					   priv->xl_array[file_list[index]],
-					   priv->xl_array[file_list[index]],
-					   priv->xl_array[file_list[index]]->fops->open,
-					   &local->loc1,
-					   local->flags,
-					   local->fd, 0);
-			if (need_break)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_create_cbk -
- */
-int32_t
-unify_create_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  fd_t *fd,
-		  inode_t *inode,
-		  struct iatt *buf,
-                  struct iatt *preparent,
-                  struct iatt *postparent)
-{
-	int ret = 0;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-	inode_t *tmp_inode = NULL;
-
-	if (op_ret == -1) {
-		/* send unlink () on Namespace */
-		local->op_errno = op_errno;
-		local->op_ret = -1;
-		local->call_count = 1;
-		gf_log (this->name, GF_LOG_ERROR,
-			"create failed on %s (file %s, error %s), "
-			"sending unlink to namespace", 
-			prev_frame->this->name, 
-			local->loc1.path, strerror (op_errno));
-
-		STACK_WIND (frame,
-			    unify_create_unlink_cbk,
-			    NS(this),
-			    NS(this)->fops->unlink,
-			    &local->loc1);
-
-		return 0;
-	}
-
-	if (op_ret >= 0) {
-		local->op_ret = op_ret;
-		local->stbuf = *buf;
-		/* Just inode number should be from NS node */
-		local->stbuf.ia_ino = local->ia_ino;
-
-		/* TODO: log on failure */
-		ret = fd_ctx_set (fd, this, (uint64_t)(long)prev_frame->this);
-	}
-  
-	tmp_inode = local->loc1.inode;
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, local->op_ret, local->op_errno, local->fd, 
-		      tmp_inode, &local->stbuf,
-                      &local->oldpreparent, &local->oldpostparent);
-
-	return 0;
-}
-
-/**
- * unify_ns_create_cbk -
- * 
- */
-int32_t
-unify_ns_create_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-		     fd_t *fd,
-		     inode_t *inode,
-		     struct iatt *buf,
-                     struct iatt *preparent,
-                     struct iatt *postparent)
-{
-	struct sched_ops *sched_ops = NULL;
-	xlator_t *sched_xl = NULL;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t *list = NULL;
-	int16_t index = 0;
-
-	if (op_ret == -1) {
-		/* No need to send create request to other servers, as 
-		   namespace action failed. Handle exclusive create here. */
-		if ((op_errno != EEXIST) || 
-		    ((op_errno == EEXIST) && 
-		     ((local->flags & O_EXCL) == O_EXCL))) {
-			/* If its just a create call without O_EXCL, 
-			   don't do this */
-			gf_log (this->name, GF_LOG_ERROR,
-				"namespace: path(%s): %s", 
-				local->loc1.path, strerror (op_errno));
-			unify_local_wipe (local);
-			STACK_UNWIND (frame, op_ret, op_errno, fd, inode, buf,
-                                      preparent, postparent);
-			return 0;
-		}
-	}
-  
-	if (op_ret >= 0) {
-		/* Get the inode number from the NS node */
-		local->ia_ino = buf->ia_ino;
-  
-                local->oldpreparent = *preparent;
-                local->oldpostparent = *postparent;
-
-		local->op_ret = -1;
-
-		/* Start the mapping list */
-		list = GF_CALLOC (1, sizeof (int16_t) * 3,
-                                  gf_unify_mt_int16_t);
-		ERR_ABORT (list);
-		inode_ctx_put (inode, this, (uint64_t)(long)list);
-		list[0] = priv->child_count;
-		list[2] = -1;
-
-		/* This means, file doesn't exist anywhere in the Filesystem */
-		sched_ops = priv->sched_ops;
-
-		/* Send create request to the scheduled node now */
-		sched_xl = sched_ops->schedule (this, local->loc1.path);
-		if (sched_xl == NULL)
-		{
-			/* send unlink () on Namespace */
-			local->op_errno = ENOTCONN;
-			local->op_ret = -1;
-			local->call_count = 1;
-			gf_log (this->name, GF_LOG_ERROR,
-				"no node online to schedule create:(file %s) "
-				"sending unlink to namespace", 
-				(local->loc1.path)?local->loc1.path:"");
-
-			STACK_WIND (frame,
-				    unify_create_unlink_cbk,
-				    NS(this),
-				    NS(this)->fops->unlink,
-				    &local->loc1);
-	
-			return 0;
-		}
-
-		for (index = 0; index < priv->child_count; index++)
-			if (sched_xl == priv->xl_array[index])
-				break;
-		list[1] = index;
-
-		STACK_WIND (frame, unify_create_cbk,
-			    sched_xl, sched_xl->fops->create,
-			    &local->loc1, local->flags, local->mode, fd);
-	} else {
-		/* File already exists, and there is no O_EXCL flag */
-
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"File(%s) already exists on namespace, sending "
-			"open instead", local->loc1.path);
-
-		local->list = GF_CALLOC (1, sizeof (int16_t) * 3,
-                                         gf_unify_mt_int16_t);
-		ERR_ABORT (local->list);
-		local->call_count = priv->child_count + 1;
-		local->op_ret = -1;
-		for (index = 0; index <= priv->child_count; index++) {
-			/* Send lookup() to all nodes including namespace */
-			STACK_WIND_COOKIE (frame,
-					   unify_create_lookup_cbk,
-					   (void *)(long)index,
-					   priv->xl_array[index],
-					   priv->xl_array[index]->fops->lookup,
-					   &local->loc1,
-					   NULL);
-		}
-	}
-	return 0;
-}
-
-/**
- * unify_create - create a file in global namespace first, so other 
- *    clients can see them. Create the file in storage nodes in background.
- */
-int32_t
-unify_create (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc,
-	      int32_t flags,
-	      mode_t mode,
-	      fd_t *fd)
-{
-	unify_local_t *local = NULL;
-  
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	local->mode = mode;
-	local->flags = flags;
-	local->fd = fd;
-
-	loc_copy (&local->loc1, loc);
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, fd, loc->inode, NULL,
-                              NULL, NULL);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_ns_create_cbk,
-		    NS(this),
-		    NS(this)->fops->create,
-		    loc,
-		    flags | O_EXCL,
-		    mode,
-		    fd);
-  
-	return 0;
-}
-
-
-/**
- * unify_opendir_cbk - 
- */
-int32_t
-unify_opendir_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   fd_t *fd)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, fd);
-
-	return 0;
-}
-
-/** 
- * unify_opendir -
- */
-int32_t
-unify_opendir (call_frame_t *frame,
-	       xlator_t *this,
-	       loc_t *loc,
-	       fd_t *fd)
-{
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	STACK_WIND (frame, unify_opendir_cbk,
-		    NS(this), NS(this)->fops->opendir, loc, fd);
-
-	return 0;
-}
-
-
-int32_t
-unify_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *statpre,
-                   struct iatt *statpost)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"%s(): child(%s): path(%s): %s", 
-				gf_fop_list[frame->root->op],
-				prev_frame->this->name, 
-				(local->loc1.path)?local->loc1.path:"", 
-				strerror (op_errno));
-
-			local->op_errno = op_errno;
-			if ((op_errno == ENOENT) && priv->optimist) 
-				local->op_ret = 0;
-		}
-
-		if (op_ret >= 0) {
-			local->op_ret = 0;
-
-			if (NS (this) == prev_frame->this) {
-				local->ia_ino = statpost->ia_ino;
-				/* If the entry is directory, get the stat
-				   from NS node */
-				if (IA_ISDIR (statpost->ia_type) || 
-				    !local->stpost.ia_blksize) {
-					local->stpre = *statpre;
-                                        local->stpost = *statpost;
-				}
-			}
-
-			if ((!IA_ISDIR (statpost->ia_type)) && 
-			    (NS (this) != prev_frame->this)) {
-				/* If file, take the stat info from Storage 
-				   node. */
-				local->stpre = *statpre;
-                                local->stpost = *statpost;
-			}
-		}
-	}
-	UNLOCK (&frame->lock);
-    
-	if (!callcnt) {
-		/* If the inode number is not filled, operation should
-		   fail */
-		if (!local->ia_ino)
-			local->op_ret = -1;
-
-                local->stpre.ia_ino = local->ia_ino;
-		local->stpost.ia_ino = local->ia_ino;
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      &local->stpre, &local->stpost);
-	}
-
-	return 0;
-}
-
-
-int32_t
-unify_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-               struct iatt *stbuf, int32_t valid)
-{
-	unify_local_t *local = NULL;
-	unify_private_t *priv = this->private;
-	int32_t index = 0;
-	int32_t callcnt = 0;
-  	uint64_t tmp_list = 0;
-
-        if (!(loc && loc->inode)) {                     
-                STACK_UNWIND (frame, -1, EINVAL, NULL, NULL);
-                return 0;
-        }                                                    
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-
-	if (IA_ISDIR (loc->inode->ia_type)) {
-		local->call_count = 1;
-      
-                STACK_WIND (frame,
-                            unify_setattr_cbk,
-                            NS (this),
-                            NS (this)->fops->setattr,
-                            loc, stbuf, valid);
-        } else {
-		inode_ctx_get (loc->inode, this, &tmp_list);
-		local->list = (int16_t *)(long)tmp_list;
-
-		for (index = 0; local->list[index] != -1; index++) {
-			local->call_count++;
-			callcnt++;
-		}
-      
-		for (index = 0; local->list[index] != -1; index++) {
-			STACK_WIND (frame,
-				    unify_setattr_cbk,
-				    priv->xl_array[local->list[index]],
-				    priv->xl_array[local->list[index]]->fops->setattr,
-				    loc, stbuf, valid);
-
-			if (!--callcnt)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-
-int32_t
-unify_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                struct iatt *stbuf, int32_t valid)
-{
-	unify_local_t *local = NULL;
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR(fd);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-
-	if (!fd_ctx_get (fd, this, &tmp_child)) {
-		/* If its set, then its file */
-		child = (xlator_t *)(long)tmp_child;		     
-
-		local->call_count = 2;
-
-		STACK_WIND (frame, unify_setattr_cbk, child,
-			    child->fops->fsetattr, fd, stbuf, valid);
-
-		STACK_WIND (frame, unify_setattr_cbk, NS(this),
-			    NS(this)->fops->fsetattr, fd, stbuf, valid);
-	} else {
-		local->call_count = 1;
-    
-		STACK_WIND (frame, unify_setattr_cbk,
-			    NS(this), NS(this)->fops->fsetattr,
-			    fd, stbuf, valid);
-	}
-  
-	return 0;
-}
-
-
-/**
- * unify_truncate_cbk - 
- */
-int32_t
-unify_truncate_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    struct iatt *prebuf,
-                    struct iatt *postbuf)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"child(%s): path(%s): %s", 
-				prev_frame->this->name, 
-				(local->loc1.path)?local->loc1.path:"", 
-				strerror (op_errno));
-			local->op_errno = op_errno;
-			if (!((op_errno == ENOENT) && priv->optimist))
-				local->op_ret = -1;
-		}
-
-		if (op_ret >= 0) {
-			if (NS (this) == prev_frame->this) {
-				local->ia_ino = postbuf->ia_ino;
-				/* If the entry is directory, get the 
-				   stat from NS node */
-				if (IA_ISDIR (postbuf->ia_type) ||
-				    !local->stbuf.ia_blksize) {
-					local->stbuf = *prebuf;
-                                        local->poststbuf = *postbuf;
-				}
-			}
-
-			if ((!IA_ISDIR (postbuf->ia_type)) &&
-			    (NS (this) != prev_frame->this)) {
-				/* If file, take the stat info from 
-				   Storage node. */
-				local->stbuf = *prebuf;
-                                local->poststbuf = *postbuf;
-			}
-		}
-	}
-	UNLOCK (&frame->lock);
-    
-	if (!callcnt) {
-		if (local->ia_ino) {
-			local->stbuf.ia_ino = local->ia_ino;
-                        local->poststbuf.ia_ino = local->ia_ino;
-                } else {
-			local->op_ret = -1;
-                }
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      &local->stbuf, &local->poststbuf);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_truncate - 
- */
-int32_t
-unify_truncate (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		off_t offset)
-{
-	unify_local_t *local = NULL;
-	unify_private_t *priv = this->private;
-	int32_t index = 0;
-	int32_t callcnt = 0;
-  	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-	local->ia_ino = loc->inode->ino;
-
-	if (IA_ISDIR (loc->inode->ia_type)) {
-		local->call_count = 1;
-      
-		STACK_WIND (frame,
-			    unify_truncate_cbk,
-			    NS(this),
-			    NS(this)->fops->truncate,
-			    loc,
-                            0);
-	} else {
-		local->op_ret = 0;
-		inode_ctx_get (loc->inode, this, &tmp_list);
-		local->list = (int16_t *)(long)tmp_list;
-
-		for (index = 0; local->list[index] != -1; index++) {
-			local->call_count++;
-			callcnt++;
-		}
-      
-		/* Don't send offset to NS truncate */
-		STACK_WIND (frame, unify_truncate_cbk, NS(this),
-			    NS(this)->fops->truncate, loc, 0);
-		callcnt--;
-
-		for (index = 0; local->list[index] != -1; index++) {
-			if (NS(this) != priv->xl_array[local->list[index]]) {
-				STACK_WIND (frame,
-					    unify_truncate_cbk,
-					    priv->xl_array[local->list[index]],
-					    priv->xl_array[local->list[index]]->fops->truncate,
-					    loc,
-					    offset);
-				if (!--callcnt)
-					break;
-			}
-		}
-	}
-
-	return 0;
-}
-
-/**
- * unify_readlink_cbk - 
- */
-int32_t
-unify_readlink_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    const char *path,
-                    struct iatt *sbuf)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, path, sbuf);
-	return 0;
-}
-
-/**
- * unify_readlink - Read the link only from the storage node.
- */
-int32_t
-unify_readlink (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		size_t size)
-{
-	unify_private_t *priv = this->private;
-	int32_t entry_count = 0;
-	int16_t *list = NULL;
-	int16_t index = 0;
-  	uint64_t tmp_list = 0;
-  
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-	
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	for (index = 0; list[index] != -1; index++)
-		entry_count++;
-
-	if (entry_count >= 2) {
-		for (index = 0; list[index] != -1; index++) {
-			if (priv->xl_array[list[index]] != NS(this)) {
-				STACK_WIND (frame,
-					    unify_readlink_cbk,
-					    priv->xl_array[list[index]],
-					    priv->xl_array[list[index]]->fops->readlink,
-					    loc,
-					    size);
-				break;
-			}
-		}
-	} else {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"returning ENOENT, no softlink files found "
-			"on storage node");
-		STACK_UNWIND (frame, -1, ENOENT, NULL);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_unlink_cbk - 
- */
-int32_t
-unify_unlink_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-                  struct iatt *preparent,
-                  struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret == 0  || ((op_errno == ENOENT) && priv->optimist))
-			local->op_ret = 0;
-		if (op_ret == -1)
-			local->op_errno = op_errno;
-
-                if (((call_frame_t *)cookie)->this == NS(this)) {
-                        local->oldpreparent = *preparent;
-                        local->oldpostparent = *postparent;
-                }
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno,
-                              &local->oldpreparent, &local->oldpostparent);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_unlink - 
- */
-int32_t
-unify_unlink (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = NULL;
-	int16_t *list = NULL;
-	int16_t index = 0;
-  	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	for (index = 0; list[index] != -1; index++)
-		local->call_count++;
-
-	if (local->call_count) {
-		for (index = 0; list[index] != -1; index++) {
-			char need_break = (list[index+1] == -1);
-			STACK_WIND (frame,
-				    unify_unlink_cbk,
-				    priv->xl_array[list[index]],
-				    priv->xl_array[list[index]]->fops->unlink,
-				    loc);
-			if (need_break)
-				break;
-		}
-	} else {
-		gf_log (this->name, GF_LOG_ERROR,
-			"%s: returning ENOENT", loc->path);
-		STACK_UNWIND (frame, -1, ENOENT, NULL, NULL);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_readv_cbk - 
- */
-int32_t
-unify_readv_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-		 struct iovec *vector,
-		 int32_t count,
-		 struct iatt *stbuf,
-                 struct iobref *iobref)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, vector, count, stbuf, iobref);
-	return 0;
-}
-
-/**
- * unify_readv - 
- */
-int32_t
-unify_readv (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd,
-	     size_t size,
-	     off_t offset)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame,
-		    unify_readv_cbk,
-		    child,
-		    child->fops->readv,
-		    fd,
-		    size,
-		    offset);
-
-
-	return 0;
-}
-
-/**
- * unify_writev_cbk - 
- */
-int32_t
-unify_writev_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-                  struct iatt *prebuf,
-		  struct iatt *postbuf)
-{
-	unify_local_t *local = NULL;
-
-        local = frame->local;
-
-        local->stbuf = *prebuf;
-        local->stbuf.ia_ino = local->ia_ino;
-
-        local->poststbuf = *postbuf;
-        local->poststbuf.ia_ino = local->ia_ino;
-
-	STACK_UNWIND (frame, op_ret, op_errno,
-                      &local->stbuf, &local->poststbuf);
-	return 0;
-}
-
-/**
- * unify_writev - 
- */
-int32_t
-unify_writev (call_frame_t *frame,
-	      xlator_t *this,
-	      fd_t *fd,
-	      struct iovec *vector,
-	      int32_t count,
-	      off_t off,
-              struct iobref *iobref)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-	unify_local_t *local = NULL;
-
-	INIT_LOCAL (frame, local);
-        local->ia_ino = fd->inode->ino;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame,
-		    unify_writev_cbk,
-		    child,
-		    child->fops->writev,
-		    fd,
-		    vector,
-		    count,
-		    off,
-                    iobref);
-
-	return 0;
-}
-
-/**
- * unify_ftruncate -
- */
-int32_t
-unify_ftruncate (call_frame_t *frame,
-		 xlator_t *this,
-		 fd_t *fd,
-		 off_t offset)
-{
-	xlator_t *child = NULL;
-	unify_local_t *local = NULL;
-	uint64_t tmp_child = 0;
-
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR(fd);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	local->op_ret = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	local->call_count = 2;
-  
-	STACK_WIND (frame, unify_truncate_cbk,
-		    child, child->fops->ftruncate,
-		    fd, offset);
-  
-	STACK_WIND (frame, unify_truncate_cbk,
-		    NS(this), NS(this)->fops->ftruncate,
-		    fd, 0);
-  
-	return 0;
-}
-
-
-/**
- * unify_flush_cbk - 
- */
-int32_t
-unify_flush_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-/**
- * unify_flush -
- */
-int32_t
-unify_flush (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_flush_cbk, child, 
-		    child->fops->flush, fd);
-
-	return 0;
-}
-
-
-/**
- * unify_fsync_cbk - 
- */
-int32_t
-unify_fsync_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-                 struct iatt *prebuf,
-                 struct iatt *postbuf)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
-	return 0;
-}
-
-/**
- * unify_fsync - 
- */
-int32_t
-unify_fsync (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd,
-	     int32_t flags)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_fsync_cbk, child,
-		    child->fops->fsync, fd, flags);
-
-	return 0;
-}
-
-/**
- * unify_fstat - Send fstat FOP to Namespace only if its directory, and to 
- *     both namespace and the storage node if its a file.
- */
-int32_t
-unify_fstat (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd)
-{
-	unify_local_t *local = NULL;
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR(fd);
-
-	INIT_LOCAL (frame, local);
-	local->ia_ino = fd->inode->ino;
-
-	if (!fd_ctx_get (fd, this, &tmp_child)) {
-		/* If its set, then its file */
-		child = (xlator_t *)(long)tmp_child;		     
-		local->call_count = 2;
-
-		STACK_WIND (frame, unify_buf_cbk, child,
-			    child->fops->fstat, fd);
-
-		STACK_WIND (frame, unify_buf_cbk, NS(this),
-			    NS(this)->fops->fstat, fd);
-
-	} else {
-		/* this is an directory */
-		local->call_count = 1;
-		STACK_WIND (frame, unify_buf_cbk, NS(this),
-			    NS(this)->fops->fstat, fd);
-	}
-
-	return 0;
-}
-
-/**
- * unify_getdents_cbk - 
- */
-int32_t
-unify_getdents_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    dir_entry_t *entry,
-		    int32_t count)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, entry, count);
-	return 0;
-}
-
-/**
- * unify_getdents - send the FOP request to all the nodes.
- */
-int32_t
-unify_getdents (call_frame_t *frame,
-		xlator_t *this,
-		fd_t *fd,
-		size_t size,
-		off_t offset,
-		int32_t flag)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_getdents_cbk, NS(this),
-		    NS(this)->fops->getdents, fd, size, offset, flag);
-
-	return 0;
-}
-
-
-/**
- * unify_readdir_cbk - 
- */
-int32_t
-unify_readdir_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   gf_dirent_t *buf)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
-
-	return 0;
-}
-
-/**
- * unify_readdir - send the FOP request to all the nodes.
- */
-int32_t
-unify_readdir (call_frame_t *frame,
-	       xlator_t *this,
-	       fd_t *fd,
-	       size_t size,
-	       off_t offset)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_readdir_cbk, NS(this),
-		    NS(this)->fops->readdir, fd, size, offset);
-
-	return 0;
-}
-
-
-int32_t
-unify_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, gf_dirent_t *buf)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
-
-	return 0;
-}
-
-
-int32_t
-unify_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-                off_t offset)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_readdirp_cbk, NS(this),
-		    NS(this)->fops->readdirp, fd, size, offset);
-
-	return 0;
-}
-
-
-/**
- * unify_fsyncdir_cbk - 
- */
-int32_t
-unify_fsyncdir_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-
-	return 0;
-}
-
-/**
- * unify_fsyncdir -
- */
-int32_t
-unify_fsyncdir (call_frame_t *frame,
-		xlator_t *this,
-		fd_t *fd,
-		int32_t flags)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_fsyncdir_cbk,
-		    NS(this), NS(this)->fops->fsyncdir, fd, flags);
-
-	return 0;
-}
-
-/**
- * unify_lk_cbk - UNWIND frame with the proper return arguments.
- */
-int32_t
-unify_lk_cbk (call_frame_t *frame,
-	      void *cookie,
-	      xlator_t *this,
-	      int32_t op_ret,
-	      int32_t op_errno,
-	      struct flock *lock)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, lock);
-	return 0;
-}
-
-/**
- * unify_lk - Send it to all the storage nodes, (should be 1) which has file.
- */
-int32_t
-unify_lk (call_frame_t *frame,
-	  xlator_t *this,
-	  fd_t *fd,
-	  int32_t cmd,
-	  struct flock *lock)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_lk_cbk, child,
-		    child->fops->lk, fd, cmd, lock);
-
-	return 0;
-}
-
-
-int32_t
-unify_setxattr_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno);
-
-static int32_t
-unify_setxattr_file_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno)
-{
-	unify_private_t *private = this->private;
-	unify_local_t *local = frame->local;
-	xlator_t *sched_xl = NULL;
-	struct sched_ops *sched_ops = NULL;
-
-	if (op_ret == -1) {
-		if (!ENOTSUP)
-			gf_log (this->name, GF_LOG_ERROR,
-				"setxattr with XATTR_CREATE on ns: "
-				"path(%s) key(%s): %s",
-				local->loc1.path, local->name, 
-				strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno);
-		return 0;
-	} 
-
-	LOCK (&frame->lock);
-	{
-		local->failed = 0;
-		local->op_ret = 0;
-		local->op_errno = 0;
-		local->call_count = 1;
-	}
-	UNLOCK (&frame->lock);
-
-	/* schedule XATTR_CREATE on one of the child node */
-	sched_ops = private->sched_ops;
-    
-	/* Send create request to the scheduled node now */
-	sched_xl = sched_ops->schedule (this, local->name); 
-	if (!sched_xl) {
-		STACK_UNWIND (frame, -1, ENOTCONN);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_setxattr_cbk,
-		    sched_xl,
-		    sched_xl->fops->setxattr,
-		    &local->loc1,
-		    local->dict,
-		    local->flags);
-	return 0;
-}
-
-/**
- * unify_setxattr_cbk - When all the child nodes return, UNWIND frame.
- */
-int32_t
-unify_setxattr_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno)
-{
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-	dict_t *dict = NULL;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret == -1) {
-			gf_log (this->name, (((op_errno == ENOENT) || 
-					      (op_errno == ENOTSUP))? 
-					     GF_LOG_DEBUG : GF_LOG_ERROR), 
-				"child(%s): path(%s): %s", 
-				prev_frame->this->name, 
-				(local->loc1.path)?local->loc1.path:"", 
-				strerror (op_errno));
-			if (local->failed == -1) {
-				local->failed = 1;
-			}
-			local->op_errno = op_errno;
-		} else {
-			local->failed = 0;
-			local->op_ret = op_ret;
-		}
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		if (local->failed && local->name && 
-		    ZR_FILE_CONTENT_REQUEST(local->name)) {      
-			dict = get_new_dict ();
-			dict_set (dict, local->dict->members_list->key, 
-				  data_from_dynptr(NULL, 0));
-			dict_ref (dict);
-
-			local->call_count = 1;
-
-			STACK_WIND (frame,
-				    unify_setxattr_file_cbk,
-				    NS(this),
-				    NS(this)->fops->setxattr,
-				    &local->loc1,
-				    dict,
-				    XATTR_CREATE);
-
-			dict_unref (dict);
-			return 0;
-		}
-    
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno);
-	}
-
-	return 0;
-}
-
-/**
- * unify_sexattr - This function should be sent to all the storage nodes, 
- *       which contains the file, (excluding namespace).
- */
-int32_t
-unify_setxattr (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		dict_t *dict,
-		int32_t flags)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = NULL;
-	int16_t *list = NULL;
-	int16_t index = 0;
-	int32_t call_count = 0;
-  	uint64_t tmp_list = 0;
-	data_pair_t *trav = dict->members_list;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	local->failed = -1;
-	loc_copy (&local->loc1, loc);
-
-	if (IA_ISDIR (loc->inode->ia_type)) {
-
-		if (trav && trav->key && ZR_FILE_CONTENT_REQUEST(trav->key)) {
-			/* direct the storage xlators to change file 
-			   content only if file exists */
-			local->flags = flags;
-			local->dict = dict;
-			local->name = gf_strdup (trav->key);
-			flags |= XATTR_REPLACE;
-		}
-
-		local->call_count = priv->child_count;
-		for (index = 0; index < priv->child_count; index++) {
-			STACK_WIND (frame,
-				    unify_setxattr_cbk,
-				    priv->xl_array[index],
-				    priv->xl_array[index]->fops->setxattr,
-				    loc, dict, flags);
-		}
-		return 0;
-	}
-
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	for (index = 0; list[index] != -1; index++) {
-		if (NS(this) != priv->xl_array[list[index]]) {
-			local->call_count++;
-			call_count++;
-		}
-	}
-  
-	if (local->call_count) {
-		for (index = 0; list[index] != -1; index++) {
-			if (priv->xl_array[list[index]] != NS(this)) {
-				STACK_WIND (frame,
-					    unify_setxattr_cbk,
-					    priv->xl_array[list[index]],
-					    priv->xl_array[list[index]]->fops->setxattr,
-					    loc,
-					    dict,
-					    flags);
-				if (!--call_count)
-					break;
-			}
-		}
-		return 0;
-	}
-
-	/* No entry in storage nodes */
-	gf_log (this->name, GF_LOG_DEBUG, 
-		"returning ENOENT, file not found on storage node.");
-	STACK_UNWIND (frame, -1, ENOENT);
-
-	return 0;
-}
-
-
-/**
- * unify_getxattr_cbk - This function is called from only one child, so, no
- *     need of any lock or anything else, just send it to above layer 
- */
-int32_t
-unify_getxattr_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    dict_t *value)
-{
-	int32_t callcnt = 0;
-	dict_t *local_value = NULL;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-  
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, 
-				(((op_errno == ENOENT) || 
-				  (op_errno == ENODATA) || 
-				  (op_errno == ENOTSUP)) ? 
-				 GF_LOG_DEBUG : GF_LOG_ERROR), 
-				"child(%s): path(%s): %s", 
-				prev_frame->this->name, 
-				(local->loc1.path)?local->loc1.path:"", 
-				strerror (op_errno));
-		} else {
-			if (!local->dict)
-				local->dict = dict_ref (value);
-			local->op_ret = op_ret;
-		}
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		local_value = local->dict;
-		local->dict = NULL;
-      
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      local_value);
-      
-		if (local_value)
-			dict_unref (local_value);
-	} 
-
-	return 0;
-}
-
-
-/** 
- * unify_getxattr - This FOP is sent to only the storage node.
- */
-int32_t
-unify_getxattr (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		const char *name)
-{
-	unify_private_t *priv = this->private;
-	int16_t *list = NULL;
-	int16_t index = 0;
-	int16_t count = 0;
-	unify_local_t *local = NULL;
-  	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-	INIT_LOCAL (frame, local);
-
-	if (IA_ISDIR (loc->inode->ia_type)) {
-		local->call_count = priv->child_count;
-		for (index = 0; index < priv->child_count; index++)
-			STACK_WIND (frame,
-				    unify_getxattr_cbk,
-				    priv->xl_array[index],
-				    priv->xl_array[index]->fops->getxattr,
-				    loc,
-				    name);
-		return 0;
-	}
-
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	for (index = 0; list[index] != -1; index++) {
-		if (NS(this) != priv->xl_array[list[index]]) {
-			local->call_count++;
-			count++;
-		}
-	}
-
-	if (count) {
-		for (index = 0; list[index] != -1; index++) {
-			if (priv->xl_array[list[index]] != NS(this)) {
-				STACK_WIND (frame,
-					    unify_getxattr_cbk,
-					    priv->xl_array[list[index]],
-					    priv->xl_array[list[index]]->fops->getxattr,
-					    loc,
-					    name);
-				if (!--count)
-					break;
-			}
-		}
-	} else {
-		dict_t *tmp_dict = get_new_dict ();
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"%s: returning ENODATA, no file found on storage node",
-			loc->path);
-		STACK_UNWIND (frame, -1, ENODATA, tmp_dict);
-		dict_destroy (tmp_dict);
-	}
-
-	return 0;
-}
-
-/**
- * unify_removexattr_cbk - Wait till all the child node returns the call
- *      and then UNWIND to above layer.
- */
-int32_t
-unify_removexattr_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno)
-{
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-
-	LOCK (&frame->lock);
-	{ 
-		callcnt = --local->call_count;
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			if (op_errno != ENOTSUP)
-				gf_log (this->name, GF_LOG_ERROR, 
-					"child(%s): path(%s): %s", 
-					prev_frame->this->name, 
-					local->loc1.path, strerror (op_errno));
-		} else {
-			local->op_ret = op_ret;
-		}
-	}
-	UNLOCK (&frame->lock);  
-
-	if (!callcnt) {
-		STACK_UNWIND (frame, local->op_ret, local->op_errno);
-	}
-
-	return 0;
-}
-
-/**
- * unify_removexattr - Send it to all the child nodes which has the files.
- */
-int32_t
-unify_removexattr (call_frame_t *frame,
-		   xlator_t *this,
-		   loc_t *loc,
-		   const char *name)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = NULL;
-	int16_t *list = NULL;
-	int16_t index = 0;
-	int32_t call_count = 0;
-  	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-
-	if (IA_ISDIR (loc->inode->ia_type)) {
-		local->call_count = priv->child_count;
-		for (index = 0; index < priv->child_count; index++)
-			STACK_WIND (frame, 		    
-				    unify_removexattr_cbk,
-				    priv->xl_array[index],
-				    priv->xl_array[index]->fops->removexattr,
-				    loc,
-				    name);
-
-		return 0;
-	}
-
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	for (index = 0; list[index] != -1; index++) {
-		if (NS(this) != priv->xl_array[list[index]]) {
-			local->call_count++;
-			call_count++;
-		}
-	}
-
-	if (local->call_count) {
-		for (index = 0; list[index] != -1; index++) {
-			if (priv->xl_array[list[index]] != NS(this)) {
-				STACK_WIND (frame,
-					    unify_removexattr_cbk,
-					    priv->xl_array[list[index]],
-					    priv->xl_array[list[index]]->fops->removexattr,
-					    loc,
-					    name);
-				if (!--call_count)
-					break;
-			}
-		}
-		return 0;
-	} 
-
-	gf_log (this->name, GF_LOG_DEBUG, 
-		"%s: returning ENOENT, not found on storage node.", loc->path);
-	STACK_UNWIND (frame, -1, ENOENT);
-
-	return 0;
-}
-
-
-int32_t 
-unify_mknod_unlink_cbk (call_frame_t *frame,
-			void *cookie,
-			xlator_t *this,
-			int32_t op_ret,
-			int32_t op_errno,
-                        struct iatt *preparent,
-                        struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1)
-		gf_log (this->name, GF_LOG_ERROR, 
-			"%s: %s", local->loc1.path, strerror (op_errno));
-  
-	unify_local_wipe (local);
-	/* No log required here as this -1 is for mknod call */
-	STACK_UNWIND (frame, -1, local->op_errno, NULL, NULL);
-	return 0;
-}
-
-/**
- * unify_mknod_cbk - 
- */
-int32_t
-unify_mknod_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-		 inode_t *inode,
-                 struct iatt *buf,
-                 struct iatt *preparent,
-                 struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"mknod failed on storage node, sending unlink to "
-			"namespace");
-		local->op_errno = op_errno;
-		STACK_WIND (frame,
-			    unify_mknod_unlink_cbk,
-			    NS(this),
-			    NS(this)->fops->unlink,
-			    &local->loc1);
-		return 0;
-	}
-  
-	local->stbuf = *buf;
-	local->stbuf.ia_ino = local->ia_ino;
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, op_ret, op_errno, inode, &local->stbuf,
-                      &local->oldpreparent, &local->oldpostparent);
-	return 0;
-}
-
-/**
- * unify_ns_mknod_cbk - 
- */
-int32_t
-unify_ns_mknod_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    inode_t *inode,
-                    struct iatt *buf,
-                    struct iatt *preparent,
-                    struct iatt *postparent)
-{
-	struct sched_ops *sched_ops = NULL;
-	xlator_t *sched_xl = NULL;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t *list = NULL;
-	int16_t index = 0;
-	call_frame_t *prev_frame = cookie;
-
-	if (op_ret == -1) {
-		/* No need to send mknod request to other servers, 
-		 * as namespace action failed 
-		 */
-		gf_log (this->name, GF_LOG_ERROR, 
-			"child(%s): path(%s): %s", 
-			prev_frame->this->name, local->loc1.path, 
-			strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                              preparent, postparent);
-		return 0;
-	}
-  
-	/* Create one inode for this entry */
-	local->op_ret = 0;
-	local->stbuf = *buf;
-	local->ia_ino = buf->ia_ino;
-
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	list = GF_CALLOC (1, sizeof (int16_t) * 3, gf_unify_mt_int16_t);
-	ERR_ABORT (list);
-	list[0] = priv->child_count;
-	list[2] = -1;
-	inode_ctx_put (inode, this, (uint64_t)(long)list);
-
-	sched_ops = priv->sched_ops;
-
-	/* Send mknod request to scheduled node now */
-	sched_xl = sched_ops->schedule (this, local->loc1.path); 
-	if (!sched_xl) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"mknod failed on storage node, no node online "
-			"at the moment, sending unlink to NS");
-		local->op_errno = ENOTCONN;
-		STACK_WIND (frame,
-			    unify_mknod_unlink_cbk,
-			    NS(this),
-			    NS(this)->fops->unlink,
-			    &local->loc1);
-      
-		return 0;
-	}
-
-	for (index = 0; index < priv->child_count; index++)
-		if (sched_xl == priv->xl_array[index])
-			break;
-	list[1] = index;
-  
-	STACK_WIND (frame,  unify_mknod_cbk,
-		    sched_xl,  sched_xl->fops->mknod,
-		    &local->loc1, local->mode, local->dev);
-
-	return 0;
-}
-
-/**
- * unify_mknod - Create a device on namespace first, and later create on 
- *       the storage node.
- */
-int32_t
-unify_mknod (call_frame_t *frame,
-	     xlator_t *this,
-	     loc_t *loc,
-	     mode_t mode,
-	     dev_t rdev)
-{
-	unify_local_t *local = NULL;
-  
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	local->mode = mode;
-	local->dev = rdev;
-	loc_copy (&local->loc1, loc);
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, loc->inode, NULL);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_ns_mknod_cbk,
-		    NS(this),
-		    NS(this)->fops->mknod,
-		    loc,
-		    mode,
-		    rdev);
-
-	return 0;
-}
-
-int32_t 
-unify_symlink_unlink_cbk (call_frame_t *frame,
-			  void *cookie,
-			  xlator_t *this,
-			  int32_t op_ret,
-			  int32_t op_errno,
-                          struct iatt *preparent,
-                          struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-	if (op_ret == -1)
-		gf_log (this->name, GF_LOG_ERROR, 
-			"%s: %s", local->loc1.path, strerror (op_errno));
-
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, -1, local->op_errno, NULL, NULL);
-	return 0;
-}
-
-/**
- * unify_symlink_cbk - 
- */
-int32_t
-unify_symlink_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   inode_t *inode,
-                   struct iatt *buf,
-                   struct iatt *preparent,
-                   struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1) {
-		/* Symlink on storage node failed, hence send unlink 
-		   to the NS node */
-		local->op_errno = op_errno;
-		gf_log (this->name, GF_LOG_ERROR, 
-			"symlink on storage node failed, sending unlink "
-			"to namespace");
-
-		STACK_WIND (frame,
-			    unify_symlink_unlink_cbk,
-			    NS(this),
-			    NS(this)->fops->unlink,
-			    &local->loc1);
-    
-		return 0;
-	}
-  
-	local->stbuf = *buf;
-	local->stbuf.ia_ino = local->ia_ino;
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, op_ret, op_errno, inode, &local->stbuf,
-                      &local->oldpreparent, &local->oldpostparent);
-
-	return 0;
-}
-
-/**
- * unify_ns_symlink_cbk - 
- */
-int32_t
-unify_ns_symlink_cbk (call_frame_t *frame,
-		      void *cookie,
-		      xlator_t *this,
-		      int32_t op_ret,
-		      int32_t op_errno,
-		      inode_t *inode,
-                      struct iatt *buf,
-                      struct iatt *preparent,
-                      struct iatt *postparent)
-{
-
-	struct sched_ops *sched_ops = NULL;
-	xlator_t *sched_xl = NULL;
-	int16_t *list = NULL;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t index = 0;
-
-	if (op_ret == -1) {
-		/* No need to send symlink request to other servers, 
-		 * as namespace action failed 
-		 */
-		gf_log (this->name, GF_LOG_ERROR, 
-			"namespace: path(%s): %s", 
-			local->loc1.path, strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, NULL, buf,
-                              preparent, postparent);
-		return 0;
-	}
-  
-	/* Create one inode for this entry */
-	local->op_ret = 0;
-	local->ia_ino = buf->ia_ino;
-  
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	/* Start the mapping list */
-
-	list = GF_CALLOC (1, sizeof (int16_t) * 3, gf_unify_mt_int16_t);
-	ERR_ABORT (list);
-	list[0] = priv->child_count; //namespace's index
-	list[2] = -1;
-	inode_ctx_put (inode, this, (uint64_t)(long)list);
-
-	sched_ops = priv->sched_ops;
-
-	/* Send symlink request to all the nodes now */
-	sched_xl = sched_ops->schedule (this, local->loc1.path); 
-	if (!sched_xl) {
-		/* Symlink on storage node failed, hence send unlink 
-		   to the NS node */
-		local->op_errno = ENOTCONN;
-		gf_log (this->name, GF_LOG_ERROR, 
-			"symlink on storage node failed, no node online, "
-			"sending unlink to namespace");
-      
-		STACK_WIND (frame,
-			    unify_symlink_unlink_cbk,
-			    NS(this),
-			    NS(this)->fops->unlink,
-			    &local->loc1);
-      
-		return 0;
-	}
-
-	for (index = 0; index < priv->child_count; index++)
-		if (sched_xl == priv->xl_array[index])
-			break;
-	list[1] = index;
-	
-	STACK_WIND (frame,
-		    unify_symlink_cbk,
-		    sched_xl,
-		    sched_xl->fops->symlink,
-		    local->name,
-		    &local->loc1);
-
-	return 0;
-}
-
-/**
- * unify_symlink - 
- */
-int32_t
-unify_symlink (call_frame_t *frame,
-	       xlator_t *this,
-	       const char *linkpath,
-	       loc_t *loc)
-{
-	unify_local_t *local = NULL;
-  
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-	local->name = gf_strdup (linkpath);
-
-	if ((local->name == NULL) || 
-	    (local->loc1.path == NULL)) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, loc->inode, NULL);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_ns_symlink_cbk,
-		    NS(this),
-		    NS(this)->fops->symlink,
-		    linkpath,
-		    loc);
-
-	return 0;
-}
-
-
-int32_t 
-unify_rename_unlink_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-                         struct iatt *preparent,
-                         struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-  
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"child(%s): path(%s -> %s): %s", 
-			prev_frame->this->name, 
-			local->loc1.path, local->loc2.path, 
-			strerror (op_errno));
-      
-	}
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		local->stbuf.ia_ino = local->ia_ino;
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      &local->stbuf);
-	}
-	return 0;
-}
-
-int32_t 
-unify_ns_rename_undo_cbk (call_frame_t *frame,
-			  void *cookie,
-			  xlator_t *this,
-			  int32_t op_ret,
-			  int32_t op_errno,
-			  struct iatt *buf,
-                          struct iatt *preoldparent,
-                          struct iatt *postoldparent,
-                          struct iatt *prenewparent,
-                          struct iatt *postnewparent)
-{
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"namespace: path(%s -> %s): %s", 
-			local->loc1.path, local->loc2.path, 
-			strerror (op_errno));
-	}
-
-	local->stbuf.ia_ino = local->ia_ino;
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, local->op_ret, local->op_errno, &local->stbuf);
-	return 0;
-}
-
-int32_t 
-unify_rename_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  struct iatt *buf,
-                  struct iatt *preoldparent,
-                  struct iatt *postoldparent,
-                  struct iatt *prenewparent,
-                  struct iatt *postnewparent)
-{
-	int32_t index = 0;
-	int32_t callcnt = 0;
-	int16_t *list = NULL;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-  
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret >= 0) {
-			if (!IA_ISDIR (buf->ia_type))
-				local->stbuf = *buf;
-			local->op_ret = op_ret;
-		} else {
-			gf_log (this->name, GF_LOG_ERROR, 
-				"child(%s): path(%s -> %s): %s", 
-				prev_frame->this->name, 
-				local->loc1.path, local->loc2.path, 
-				strerror (op_errno));
-			local->op_errno = op_errno;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		local->stbuf.ia_ino = local->ia_ino;
-		if (IA_ISDIR (local->loc1.inode->ia_type)) {
-			unify_local_wipe (local);
-			STACK_UNWIND (frame, local->op_ret, local->op_errno,
-                                      &local->stbuf, &local->oldpreparent,
-                                      &local->oldpostparent, &local->newpreparent,
-                                      &local->newpostparent);
-			return 0;
-		}
-
-		if (local->op_ret == -1) {
-			/* TODO: check this logic */
-
-			/* Rename failed in storage node, successful on NS, 
-			 * hence, rename back the entries in NS */
-			/* NOTE: this will be done only if the destination 
-			 * doesn't exists, if  the destination exists, the 
-			 * job of correcting NS is left to self-heal
-			 */
-			if (!local->index) {
-				loc_t tmp_oldloc = {
-                                        /* its actual 'newloc->path' */
-					.path = local->loc2.path, 
-					.inode = local->loc1.inode,
-					.parent = local->loc2.parent
-				};
-	
-				loc_t tmp_newloc = {
-					/* Actual 'oldloc->path' */
-					.path = local->loc1.path,
-					.parent = local->loc1.parent
-				};
-
-				gf_log (this->name, GF_LOG_ERROR, 
-					"rename succussful on namespace, on "
-					"stroage node failed, reverting back");
-
-				STACK_WIND (frame,
-					    unify_ns_rename_undo_cbk,
-					    NS(this),
-					    NS(this)->fops->rename,
-					    &tmp_oldloc,
-					    &tmp_newloc);
-				return 0;
-			}
-		} else {
-			/* Rename successful on storage nodes */
-
-			int32_t idx = 0;
-			int16_t *tmp_list = NULL;
-			uint64_t tmp_list_int64 = 0;
-			if (local->loc2.inode) {
-				inode_ctx_get (local->loc2.inode, 
-					       this, &tmp_list_int64);
-				list = (int16_t *)(long)tmp_list_int64;
-
-			}
-
-			if (list) {				
-				for (index = 0; list[index] != -1; index++);
-				tmp_list = GF_CALLOC (1, index * 2, 
-                                                      gf_unify_mt_int16_t);
-				memcpy (tmp_list, list, index * 2);
-
-				for (index = 0; list[index] != -1; index++) {
-					/* TODO: Check this logic. */
-					/* If the destination file exists in 
-					 * the same storage node where we sent
-					 * 'rename' call, no need to send 
-					 * unlink 
-					 */
-					for (idx = 0; 
-					     local->list[idx] != -1; idx++) {
-						if (tmp_list[index] == local->list[idx]) {
-							tmp_list[index] = priv->child_count;
-							continue;
-						}
-					}
-	  
-					if (NS(this) != priv->xl_array[tmp_list[index]]) {
-						local->call_count++;
-						callcnt++;
-					}
-				}
-
-				if (local->call_count) {
-					if (callcnt > 1)
-						gf_log (this->name, 
-							GF_LOG_ERROR, 
-							"%s->%s: more (%d) "
-							"subvolumes have the "
-							"newloc entry", 
-							local->loc1.path, 
-							local->loc2.path, 
-							callcnt);
-
-					for (index=0; 
-					     tmp_list[index] != -1; index++) {
-						if (NS(this) != priv->xl_array[tmp_list[index]]) {		    
-							STACK_WIND (frame,
-								    unify_rename_unlink_cbk,
-								    priv->xl_array[tmp_list[index]],
-								    priv->xl_array[tmp_list[index]]->fops->unlink,
-								    &local->loc2);
-							if (!--callcnt)
-								break;
-						}
-					}
-
-					GF_FREE (tmp_list);
-					return 0;
-				}
-				if (tmp_list)
-					GF_FREE (tmp_list);
-			}
-		}
-    
-		/* Need not send 'unlink' to storage node */
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, 
-			      local->op_errno, &local->stbuf,
-                              &local->oldpreparent, &local->oldpostparent,
-                              &local->newpreparent, &local->newpostparent);
-	}
-
-	return 0;
-}
-
-int32_t 
-unify_ns_rename_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-		     struct iatt *buf,
-                     struct iatt *preoldparent,
-                     struct iatt *postoldparent,
-                     struct iatt *prenewparent,
-                     struct iatt *postnewparent)
-{
-	int32_t index = 0;
-	int32_t callcnt = 0;
-	int16_t *list = NULL;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1) {
-		/* Free local->new_inode */
-		gf_log (this->name, GF_LOG_ERROR, 
-			"namespace: path(%s -> %s): %s", 
-			local->loc1.path, local->loc2.path, 
-			strerror (op_errno));
-
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, buf,
-                              preoldparent, postoldparent,
-                              prenewparent, postnewparent);
-		return 0;
-	}
-
-	local->stbuf = *buf;
-	local->ia_ino = buf->ia_ino;
-
-        local->oldpreparent = *preoldparent;
-        local->oldpostparent = *postoldparent;
-        local->newpreparent = *prenewparent;
-        local->newpostparent = *postnewparent;
-
-	/* Everything is fine. */
-	if (IA_ISDIR (buf->ia_type)) {
-		local->call_count = priv->child_count;
-		for (index=0; index < priv->child_count; index++) {
-			STACK_WIND (frame,
-				    unify_rename_cbk,
-				    priv->xl_array[index],
-				    priv->xl_array[index]->fops->rename,
-				    &local->loc1,
-				    &local->loc2);
-		}
-
-		return 0;
-	}
-
-	local->call_count = 0;  
-	/* send rename */
-	list = local->list;
-	for (index=0; list[index] != -1; index++) {
-		if (NS(this) != priv->xl_array[list[index]]) {
-			local->call_count++;
-			callcnt++;
-		}
-	}
-
-	if (local->call_count) {
-		for (index=0; list[index] != -1; index++) {
-			if (NS(this) != priv->xl_array[list[index]]) {
-				STACK_WIND (frame,
-					    unify_rename_cbk,
-					    priv->xl_array[list[index]],
-					    priv->xl_array[list[index]]->fops->rename,
-					    &local->loc1,
-					    &local->loc2);
-				if (!--callcnt)
-					break;
-			}
-		}
-	} else {
-		/* file doesn't seem to be present in storage nodes */
-		gf_log (this->name, GF_LOG_CRITICAL,
-			"CRITICAL: source file not in storage node, "
-			"rename successful on namespace :O");
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, -1, EIO, NULL,
-                              NULL, NULL, /* preoldparent, postoldparent */
-                              NULL, NULL); /* prenewparent, postnewparent */
-	}
-	return 0;
-}
-
-
-/**
- * unify_rename - One of the tricky function. The deadliest of all :O
- */
-int32_t
-unify_rename (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *oldloc,
-	      loc_t *newloc)
-{
-	unify_local_t *local = NULL;
-  	uint64_t tmp_list = 0;
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, oldloc);
-	loc_copy (&local->loc2, newloc);
-
-	if ((local->loc1.path == NULL) || 
-	    (local->loc2.path == NULL)) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, NULL,
-                              NULL, NULL, /* preoldparent, postoldparent */
-                              NULL, NULL); /* prenewparent, postnewparent */
-		return 0;
-	}
-  
-	inode_ctx_get (oldloc->inode, this, &tmp_list);
-	local->list = (int16_t *)(long)tmp_list;
-
-	STACK_WIND (frame,
-		    unify_ns_rename_cbk,
-		    NS(this),
-		    NS(this)->fops->rename,
-		    oldloc,
-		    newloc);
-	return 0;
-}
-
-/**
- * unify_link_cbk -
- */
-int32_t
-unify_link_cbk (call_frame_t *frame,
-		void *cookie,
-		xlator_t *this,
-		int32_t op_ret,
-		int32_t op_errno,
-		inode_t *inode,
-                struct iatt *buf,
-                struct iatt *preparent,
-                struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-
-	if (op_ret >= 0) 
-		local->stbuf = *buf;
-	local->stbuf.ia_ino = local->ia_ino;
-
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, op_ret, op_errno, inode, &local->stbuf,
-                      &local->oldpreparent, &local->oldpostparent);
-
-	return 0;
-}
-
-/**
- * unify_ns_link_cbk - 
- */
-int32_t
-unify_ns_link_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   inode_t *inode,
-                   struct iatt *buf,
-                   struct iatt *preparent,
-                   struct iatt *postparent)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	int16_t *list = local->list;
-	int16_t index = 0;
-
-	if (op_ret == -1) {
-		/* No need to send link request to other servers, 
-		 * as namespace action failed 
-		 */
-		gf_log (this->name, GF_LOG_ERROR, 
-			"namespace: path(%s -> %s): %s", 
-			local->loc1.path, local->loc2.path, 
-			strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                              preparent, postparent);
-		return 0;
-	}
-
-	/* Update inode for this entry */
-	local->op_ret = 0;
-	local->ia_ino = buf->ia_ino;
-
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	/* Send link request to the node now */
-	for (index = 0; list[index] != -1; index++) {
-		char need_break = (list[index+1] == -1);
-		if (priv->xl_array[list[index]] != NS (this)) {
-			STACK_WIND (frame,
-				    unify_link_cbk,
-				    priv->xl_array[list[index]],
-				    priv->xl_array[list[index]]->fops->link,
-				    &local->loc1,
-				    &local->loc2);
-                        break;
-		}
-		if (need_break)
-			break;
-	}
-
-	return 0;
-}
-
-/**
- * unify_link - 
- */
-int32_t
-unify_link (call_frame_t *frame,
-	    xlator_t *this,
-	    loc_t *oldloc,
-	    loc_t *newloc)
-{
-	unify_local_t *local = NULL;
-  	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (oldloc);
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (newloc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-
-	loc_copy (&local->loc1, oldloc);
-	loc_copy (&local->loc2, newloc);
-
-	inode_ctx_get (oldloc->inode, this, &tmp_list);
-	local->list = (int16_t *)(long)tmp_list;
-
-	STACK_WIND (frame,
-		    unify_ns_link_cbk,
-		    NS(this),
-		    NS(this)->fops->link,
-		    oldloc,
-		    newloc);
-
-	return 0;
-}
-
-
-/**
- * unify_checksum_cbk - 
- */
-int32_t
-unify_checksum_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    uint8_t *fchecksum,
-		    uint8_t *dchecksum)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, fchecksum, dchecksum);
-
-	return 0;
-}
-
-/**
- * unify_checksum - 
- */
-int32_t
-unify_checksum (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		int32_t flag)
-{
-	STACK_WIND (frame,
-		    unify_checksum_cbk,
-		    NS(this),
-		    NS(this)->fops->checksum,
-		    loc,
-		    flag);
-
-	return 0;
-}
-
-
-/**
- * unify_finodelk_cbk - 
- */
-int
-unify_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-/**
- * unify_finodelk
- */
-int
-unify_finodelk (call_frame_t *frame, xlator_t *this,
-		const char *volume, fd_t *fd, int cmd, struct flock *flock)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_finodelk_cbk,
-		    child, child->fops->finodelk,
-		    volume, fd, cmd, flock);
-
-	return 0;
-}
-
-
-
-/**
- * unify_fentrylk_cbk - 
- */
-int
-unify_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-/**
- * unify_fentrylk
- */
-int
-unify_fentrylk (call_frame_t *frame, xlator_t *this,
-		const char *volume, fd_t *fd, const char *basename,
-		entrylk_cmd cmd, entrylk_type type)
-		
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_fentrylk_cbk,
-		    child, child->fops->fentrylk,
-		    volume, fd, basename, cmd, type);
-
-	return 0;
-}
-
-
-
-/**
- * unify_fxattrop_cbk - 
- */
-int
-unify_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno, dict_t *xattr)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, xattr);
-	return 0;
-}
-
-/**
- * unify_fxattrop
- */
-int
-unify_fxattrop (call_frame_t *frame, xlator_t *this,
-		fd_t *fd, gf_xattrop_flags_t optype, dict_t *xattr)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_fxattrop_cbk,
-		    child, child->fops->fxattrop,
-		    fd, optype, xattr);
-
-	return 0;
-}
-
-
-/**
- * unify_inodelk_cbk - 
- */
-int
-unify_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-
-/**
- * unify_inodelk
- */
-int
-unify_inodelk (call_frame_t *frame, xlator_t *this,
-	       const char *volume, loc_t *loc, int cmd, struct flock *flock)
-{
-	xlator_t *child = NULL;
-
-	child = unify_loc_subvol (loc, this);
-
-	STACK_WIND (frame, unify_inodelk_cbk,
-		    child, child->fops->inodelk,
-		    volume, loc, cmd, flock);
-
-	return 0;
-}
-
-
-
-/**
- * unify_entrylk_cbk - 
- */
-int
-unify_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-/**
- * unify_entrylk
- */
-int
-unify_entrylk (call_frame_t *frame, xlator_t *this,
-	       const char *volume, loc_t *loc, const char *basename,
-	       entrylk_cmd cmd, entrylk_type type)
-		
-{
-	xlator_t *child = NULL;
-
-	child = unify_loc_subvol (loc, this);
-
-	STACK_WIND (frame, unify_entrylk_cbk,
-		    child, child->fops->entrylk,
-		    volume, loc, basename, cmd, type);
-
-	return 0;
-}
-
-
-
-/**
- * unify_xattrop_cbk - 
- */
-int
-unify_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno, dict_t *xattr)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, xattr);
-	return 0;
-}
-
-/**
- * unify_xattrop
- */
-int
-unify_xattrop (call_frame_t *frame, xlator_t *this,
-		loc_t *loc, gf_xattrop_flags_t optype, dict_t *xattr)
-{
-	xlator_t *child = NULL;
-
-	child = unify_loc_subvol (loc, this);
-
-	STACK_WIND (frame, unify_xattrop_cbk,
-		    child, child->fops->xattrop,
-		    loc, optype, xattr);
-
-	return 0;
-}
-
-int
-unify_forget (xlator_t *this,
-              inode_t *inode)
-{
-        int16_t *list = NULL;
-        uint64_t tmp_list = 0;
-
-        if (inode->ia_type && (!IA_ISDIR(inode->ia_type))) {
-                inode_ctx_get (inode, this, &tmp_list);
-                if (tmp_list) {
-                        list = (int16_t *)(long)tmp_list;
-                        GF_FREE (list);
-                }
-        }
-
-        return 0;
-}
-
-/**
- * notify
- */
-int32_t
-notify (xlator_t *this,
-        int32_t event,
-        void *data,
-        ...)
-{
-	unify_private_t *priv = this->private;
-	struct sched_ops *sched = NULL;
-
-	if (!priv) {
-		return 0;
-	}
-
-	sched = priv->sched_ops;    
-	if (!sched) {
-		gf_log (this->name, GF_LOG_CRITICAL, "No scheduler :O");
-		raise (SIGTERM);
-		return 0;
-	}
-	if (priv->namespace == data) {
-		if (event == GF_EVENT_CHILD_UP) {
-			sched->notify (this, event, data);
-		}
-		return 0;
-	}
-
-	switch (event)
-	{
-	case GF_EVENT_CHILD_UP:
-	{
-		/* Call scheduler's update () to enable it for scheduling */
-		sched->notify (this, event, data);
-	
-		LOCK (&priv->lock);
-		{
-			/* Increment the inode's generation, which is 
-			   used for self_heal */
-			++priv->inode_generation;
-			++priv->num_child_up;
-		}
-		UNLOCK (&priv->lock);
-
-		if (!priv->is_up) {
-			default_notify (this, event, data);
-			priv->is_up = 1;
-		}
-	}
-	break;
-	case GF_EVENT_CHILD_DOWN:
-	{
-		/* Call scheduler's update () to disable the child node 
-		 * for scheduling
-		 */
-		sched->notify (this, event, data);
-		LOCK (&priv->lock);
-		{
-			--priv->num_child_up;
-		}
-		UNLOCK (&priv->lock);
-
-		if (priv->num_child_up == 0) {
-			/* Send CHILD_DOWN to upper layer */
-			default_notify (this, event, data);
-			priv->is_up = 0;
-		}
-	}
-	break;
-
-	default:
-	{
-		default_notify (this, event, data);
-	}
-	break;
-	}
-
-	return 0;
-}
-
-int32_t
-mem_acct_init (xlator_t *this)
-{
-        int     ret = -1;
-
-        if (!this)
-                return ret;
-
-        ret = xlator_mem_acct_init (this, gf_unify_mt_end + 1);
-        
-        if (ret != 0) {
-                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
-                                "failed");
-                return ret;
-        }
-
-        return ret;
-}
-
-/** 
- * init - This function is called first in the xlator, while initializing.
- *   All the config file options are checked and appropriate flags are set.
- *
- * @this - 
- */
-int32_t 
-init (xlator_t *this)
-{
-	int32_t          ret       = 0;
-	int32_t          count     = 0;
-	data_t          *scheduler = NULL;
-	data_t          *data      = NULL;
-	xlator_t        *ns_xl     = NULL;
-	xlator_list_t   *trav      = NULL;
-	xlator_list_t   *xlparent  = NULL;
-	xlator_list_t   *parent    = NULL;
-	unify_private_t *_private  = NULL; 
-
-
-	/* Check for number of child nodes, if there is no child nodes, exit */
-	if (!this->children) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"No child nodes specified. check \"subvolumes \" "
-			"option in volfile");
-		return -1;
-	}
-
-  	if (!this->parents) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"dangling volume. check volfile ");
-	}
-  
-	/* Check for 'scheduler' in volume */
-	scheduler = dict_get (this->options, "scheduler");
-	if (!scheduler) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"\"option scheduler <x>\" is missing in volfile");
-		return -1;
-	}
-
-	/* Setting "option namespace <node>" */
-	data = dict_get (this->options, "namespace");
-	if(!data) {
-		gf_log (this->name, GF_LOG_CRITICAL, 
-			"namespace option not specified, Exiting");
-		return -1;
-	}
-	/* Search namespace in the child node, if found, exit */
-	trav = this->children;
-	while (trav) {
-		if (strcmp (trav->xlator->name, data->data) == 0)
-			break;
-		trav = trav->next;
-	}
-	if (trav) {
-		gf_log (this->name, GF_LOG_CRITICAL, 
-			"namespace node used as a subvolume, Exiting");
-		return -1;
-	}
-	
-	/* Search for the namespace node, if found, continue */
-	ns_xl = this->next;
-	while (ns_xl) {
-		if (strcmp (ns_xl->name, data->data) == 0)
-			break;
-		ns_xl = ns_xl->next;
-	}
-	if (!ns_xl) {
-		gf_log (this->name, GF_LOG_CRITICAL, 
-			"namespace node not found in volfile, Exiting");
-		return -1;
-	}
-	
-	gf_log (this->name, GF_LOG_DEBUG, 
-		"namespace node specified as %s", data->data);
-	
-	_private = GF_CALLOC (1, sizeof (*_private), 
-                              gf_unify_mt_unify_private_t);
-	ERR_ABORT (_private);
-	_private->sched_ops = get_scheduler (this, scheduler->data);
-	if (!_private->sched_ops) {
-		gf_log (this->name, GF_LOG_CRITICAL, 
-			"Error while loading scheduler. Exiting");
-		GF_FREE (_private);
-		return -1;
-	}
-	
-	if (ns_xl->parents) {
-		gf_log (this->name, GF_LOG_CRITICAL,
-			"Namespace node should not be a child of any other node. Exiting");
-		GF_FREE (_private);
-		return -1;
-	}
-
-	_private->namespace = ns_xl;
-	
-	/* update _private structure */
-	{
-		count = 0;
-		trav = this->children;
-		/* Get the number of child count */
-		while (trav) {
-			count++;
-			trav = trav->next;
-		}
-		
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"Child node count is %d", count);    
-
-		_private->child_count = count;
-		if (count == 1) {
-			/* TODO: Should I error out here? */
-			gf_log (this->name, GF_LOG_CRITICAL, 
-				"WARNING: You have defined only one "
-				"\"subvolumes\" for unify volume. It may not "
-				"be the desired config, review your volume "
-				"volfile. If this is how you are testing it,"
-				" you may hit some performance penalty");
-		}
-		
-		_private->xl_array = GF_CALLOC (1, 
-					        sizeof (xlator_t) * (count + 1),
-                                                gf_unify_mt_xlator_t);
-		ERR_ABORT (_private->xl_array);
-		
-		count = 0;
-		trav = this->children;
-		while (trav) {
-			_private->xl_array[count++] = trav->xlator;
-			trav = trav->next;
-		}
-		_private->xl_array[count] = _private->namespace;
-		
-		/* self-heal part, start with generation '1' */
-		_private->inode_generation = 1; 
-                /* Because, Foreground part is tested well */
-		_private->self_heal = ZR_UNIFY_FG_SELF_HEAL; 
-		data = dict_get (this->options, "self-heal");
-		if (data) {
-			if (strcasecmp (data->data, "off") == 0) 
-				_private->self_heal = ZR_UNIFY_SELF_HEAL_OFF;
-
-			if (strcasecmp (data->data, "foreground") == 0)
-				_private->self_heal = ZR_UNIFY_FG_SELF_HEAL;
-
-			if (strcasecmp (data->data, "background") == 0)
-				_private->self_heal = ZR_UNIFY_BG_SELF_HEAL;
-		}
-    
-		/* optimist - ask bulde for more about it */
-		data = dict_get (this->options, "optimist");
-		if (data) {
-			if (gf_string2boolean (data->data, 
-					       &_private->optimist) == -1) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"optimist excepts only boolean "
-					"options");
-			}
-		}
-
-		LOCK_INIT (&_private->lock);
-	}
-
-	/* Now that everything is fine. */
-	this->private = (void *)_private;
-	{
-                ret = _private->sched_ops->mem_acct_init (this);
-
-                if (ret == -1) {
-                        return -1;
-                }
-
-		/* Initialize scheduler, if everything else is successful */
-		ret = _private->sched_ops->init (this); 
-		if (ret == -1) {
-			gf_log (this->name, GF_LOG_CRITICAL,
-				"Initializing scheduler failed, Exiting");
-			GF_FREE (_private);
-			return -1;
-		}
-
-
-		ret = 0;
-
-		/* This section is required because some fops may look 
-		 * for 'xl->parent' variable 
-		 */
-		xlparent = GF_CALLOC (1, sizeof (*xlparent),
-                                      gf_unify_mt_xlator_list_t);
-		xlparent->xlator = this;
-		if (!ns_xl->parents) {
-			ns_xl->parents = xlparent;
-		} else {
-			parent = ns_xl->parents;
-			while (parent->next)
-				parent = parent->next;
-			parent->next = xlparent;
-		}
-	}
-
-	/* Tell namespace node that init is done */
-	xlator_notify (ns_xl, GF_EVENT_PARENT_UP, this);
-
-	return 0;
-}
-
-/** 
- * fini  - Free all the allocated memory 
- */
-void
-fini (xlator_t *this)
-{
-	unify_private_t *priv = this->private;
-	priv->sched_ops->fini (this);
-	this->private = NULL;
-	LOCK_DESTROY (&priv->lock);
-	GF_FREE (priv->xl_array);
-	GF_FREE (priv);
-	return;
-}
-
-
-struct xlator_fops fops = {
-	.stat        = unify_stat,
-	.readlink    = unify_readlink,
-	.mknod       = unify_mknod,
-	.mkdir       = unify_mkdir,
-	.unlink      = unify_unlink,
-	.rmdir       = unify_rmdir,
-	.symlink     = unify_symlink,
-	.rename      = unify_rename,
-	.link        = unify_link,
-	.truncate    = unify_truncate,
-	.create      = unify_create,
-	.open        = unify_open,
-	.readv       = unify_readv,
-	.writev      = unify_writev,
-	.statfs      = unify_statfs,
-	.flush       = unify_flush,
-	.fsync       = unify_fsync,
-	.setxattr    = unify_setxattr,
-	.getxattr    = unify_getxattr,
-	.removexattr = unify_removexattr,
-	.opendir     = unify_opendir,
-	.readdir     = unify_readdir,
-	.readdirp    = unify_readdirp,
-	.fsyncdir    = unify_fsyncdir,
-	.access      = unify_access,
-	.ftruncate   = unify_ftruncate,
-	.fstat       = unify_fstat,
-	.lk          = unify_lk,
-	.lookup      = unify_lookup,
-	.getdents    = unify_getdents,
-	.checksum    = unify_checksum,
-	.inodelk     = unify_inodelk,
-	.finodelk    = unify_finodelk,
-	.entrylk     = unify_entrylk,
-	.fentrylk    = unify_fentrylk,
-	.xattrop     = unify_xattrop,
-	.fxattrop    = unify_fxattrop,
-        .setattr     = unify_setattr, 
-        .fsetattr    = unify_fsetattr,
-};
-
-
-struct xlator_cbks cbks = {
-        .forget  = unify_forget,
-};
-
-struct volume_options options[] = {
-	{ .key   = { "namespace" },  
-	  .type  = GF_OPTION_TYPE_XLATOR 
-	},
-	{ .key   = { "scheduler" },  
-	  .value = { "alu", "rr", "random", "nufa", "switch" },
-	  .type  = GF_OPTION_TYPE_STR
-	},
-	{ .key   = {"self-heal"},  
-	  .value = { "foreground", "background", "off" },
-	  .type  = GF_OPTION_TYPE_STR
-	},
-	/* TODO: remove it some time later */
-	{ .key   = {"optimist"},  
-	  .type  = GF_OPTION_TYPE_BOOL 
-	},
-
-	{ .key   = {NULL} },
-};
diff --git a/xlators/cluster/unify/src/unify.h b/xlators/cluster/unify/src/unify.h
deleted file mode 100644
index 8dcf16598..000000000
--- a/xlators/cluster/unify/src/unify.h
+++ /dev/null
@@ -1,146 +0,0 @@
-/*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#ifndef _UNIFY_H
-#define _UNIFY_H
-
-#include "scheduler.h"
-#include "list.h"
-#include "unify-mem-types.h"
-
-#define MAX_DIR_ENTRY_STRING     (32 * 1024)
-
-#define ZR_UNIFY_SELF_HEAL_OFF 0
-#define ZR_UNIFY_FG_SELF_HEAL  1
-#define ZR_UNIFY_BG_SELF_HEAL  2
-
-/* Sometimes one should use completely random numbers.. its good :p */
-#define UNIFY_SELF_HEAL_GETDENTS_COUNT 512
-
-#define NS(xl)          (((unify_private_t *)xl->private)->namespace)
-
-/* This is used to allocate memory for local structure */
-#define INIT_LOCAL(fr, loc)                   \
-do {                                          \
-  loc = GF_CALLOC (1, sizeof (unify_local_t), gf_unify_mt_unify_local_t);   \
-  ERR_ABORT (loc);			      \
-  if (!loc) {                                 \
-    STACK_UNWIND (fr, -1, ENOMEM);            \
-    return 0;                                 \
-  }                                           \
-  fr->local = loc;                            \
-  loc->op_ret = -1;                           \
-  loc->op_errno = ENOENT;                     \
-} while (0)
-
-
-
-struct unify_private {
-	/* Update this structure depending on requirement */
-	void *scheduler;               /* THIS SHOULD BE THE FIRST VARIABLE, 
-					  if xlator is using scheduler */
-	struct sched_ops *sched_ops;   /* Scheduler options  */
-	xlator_t *namespace;           /* ptr to namespace xlator */
-	xlator_t **xl_array;
-	gf_boolean_t optimist;
-	int16_t child_count;
-	int16_t num_child_up;
-	uint8_t self_heal;
-	uint8_t is_up;
-	uint64_t inode_generation;
-	gf_lock_t lock;
-};
-typedef struct unify_private unify_private_t;
-
-struct unify_self_heal_struct {
-	uint8_t dir_checksum[NAME_MAX];
-	uint8_t ns_dir_checksum[NAME_MAX];
-	uint8_t file_checksum[NAME_MAX];
-	uint8_t ns_file_checksum[NAME_MAX];
-	off_t *offset_list;
-	int   *count_list;
-	dir_entry_t **entry_list;
-};
-
-
-struct _unify_local_t {
-	int32_t call_count;
-	int32_t op_ret;
-	int32_t op_errno;
-	mode_t mode;
-	off_t offset;
-	dev_t dev;
-	uid_t uid;
-	gid_t gid;
-	int32_t flags;
-	int32_t entry_count;
-	int32_t count;    // dir_entry_t count;
-	fd_t *fd;
-	struct iatt stbuf;
-        struct iatt stpre;
-        struct iatt stpost;
-	struct statvfs statvfs_buf;
-	struct timespec tv[2];
-	char *name;
-	int32_t revalidate;
-
-	ino_t ia_ino;
-	nlink_t ia_nlink;
-  
-	dict_t *dict;
-
-	int16_t *list;
-	int16_t *new_list; /* Used only in case of rename */
-	int16_t index;
-
-	int32_t failed;
-	int32_t return_eio;  /* Used in case of different st-mode 
-				present for a given path */
-
-	uint64_t inode_generation; /* used to store the per directory 
-				    * inode_generation. Got from inode's ctx 
-				    * of directory inodes
-				    */
-
-	struct unify_self_heal_struct *sh_struct;
-	loc_t loc1, loc2;
-
-        struct iatt poststbuf;
-        /* When not used for rename, old*
-         * are used as the attrs for the current
-         * parent directory.
-         */
-        struct iatt oldpreparent;
-        struct iatt oldpostparent;
-        struct iatt newpreparent;
-        struct iatt newpostparent;
-        int32_t wbflags;
-};
-typedef struct _unify_local_t unify_local_t;
-
-int32_t zr_unify_self_heal (call_frame_t *frame,
-			    xlator_t *this,
-			    unify_local_t *local);
-
-#endif /* _UNIFY_H */
diff --git a/xlators/debug/error-gen/src/Makefile.am b/xlators/debug/error-gen/src/Makefile.am
index f353b61e6..5075c59a8 100644
--- a/xlators/debug/error-gen/src/Makefile.am
+++ b/xlators/debug/error-gen/src/Makefile.am
@@ -2,15 +2,16 @@
 xlator_LTLIBRARIES = error-gen.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/debug
 
-error_gen_la_LDFLAGS = -module -avoidversion
+error_gen_la_LDFLAGS = -module -avoid-version
 
 error_gen_la_SOURCES = error-gen.c
 error_gen_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = error-gen.h
+noinst_HEADERS = error-gen.h error-gen-mem-types.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/debug/error-gen/src/error-gen-mem-types.h b/xlators/debug/error-gen/src/error-gen-mem-types.h
new file mode 100644
index 000000000..f02280535
--- /dev/null
+++ b/xlators/debug/error-gen/src/error-gen-mem-types.h
@@ -0,0 +1,20 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __ERROR_GEN_MEM_TYPES_H__
+#define __ERROR_GEN_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_error_gen_mem_types_ {
+        gf_error_gen_mt_eg_t = gf_common_mt_end + 1,
+        gf_error_gen_mt_end
+};
+#endif
diff --git a/xlators/debug/error-gen/src/error-gen.c b/xlators/debug/error-gen/src/error-gen.c
index 095b077b4..ec0874b35 100644
--- a/xlators/debug/error-gen/src/error-gen.c
+++ b/xlators/debug/error-gen/src/error-gen.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -24,6 +14,7 @@
 
 #include "xlator.h"
 #include "error-gen.h"
+#include "statedump.h"
 
 sys_error_t error_no_list[] = {
         [GF_FOP_LOOKUP]            = { .error_no_count = 4,
@@ -91,9 +82,10 @@ sys_error_t error_no_list[] = {
         [GF_FOP_READ]              = { .error_no_count = 5,
                                     .error_no = {EINVAL,EBADF,EFAULT,EISDIR,
                                                  ENAMETOOLONG}},
-        [GF_FOP_WRITE]             = { .error_no_count = 5,
+        [GF_FOP_WRITE]             = { .error_no_count = 7,
                                     .error_no = {EINVAL,EBADF,EFAULT,EISDIR,
-                                                 ENAMETOOLONG}},
+                                                 ENAMETOOLONG,ENOSPC,
+						 GF_ERROR_SHORT_WRITE}},
         [GF_FOP_STATFS]            = {.error_no_count = 10,
                                    .error_no = {EACCES,EBADF,EFAULT,EINTR,
                                                 EIO,ENAMETOOLONG,ENOENT,
@@ -113,6 +105,15 @@ sys_error_t error_no_list[] = {
         [GF_FOP_REMOVEXATTR]       = { .error_no_count = 4,
                                     .error_no = {EACCES,EBADF,ENAMETOOLONG,
                                                  EINTR}},
+        [GF_FOP_FSETXATTR]          = { .error_no_count = 4,
+                                        .error_no = {EACCES,EBADF,EINTR,
+                                                     ENAMETOOLONG}},
+        [GF_FOP_FGETXATTR]          = { .error_no_count = 4,
+                                        .error_no = {EACCES,EBADF,ENAMETOOLONG,
+                                                     EINTR}},
+        [GF_FOP_FREMOVEXATTR]       = { .error_no_count = 4,
+                                        .error_no = {EACCES,EBADF,ENAMETOOLONG,
+                                                     EINTR}},
         [GF_FOP_OPENDIR]           = { .error_no_count = 8,
                                     .error_no = {EACCES,EEXIST,EFAULT,
                                                  EISDIR,EMFILE,
@@ -237,6 +238,8 @@ conv_errno_to_int (char **error_no)
                 return EINTR;
         else if (!strcmp ((*error_no), "EFBIG"))
                 return EFBIG;
+	else if (!strcmp((*error_no), "GF_ERROR_SHORT_WRITE"))
+		return GF_ERROR_SHORT_WRITE;
         else
                 return EAGAIN;
 }
@@ -286,6 +289,12 @@ get_fop_int (char **op_no_str)
                 return GF_FOP_GETXATTR;
         else if (!strcmp ((*op_no_str), "removexattr"))
                 return GF_FOP_REMOVEXATTR;
+        else if (!strcmp ((*op_no_str), "fsetxattr"))
+                return GF_FOP_FSETXATTR;
+        else if (!strcmp ((*op_no_str), "fgetxattr"))
+                return GF_FOP_FGETXATTR;
+        else if (!strcmp ((*op_no_str), "fremovexattr"))
+                return GF_FOP_FREMOVEXATTR;
         else if (!strcmp ((*op_no_str), "opendir"))
                 return GF_FOP_OPENDIR;
         else if (!strcmp ((*op_no_str), "readdir"))
@@ -362,7 +371,8 @@ error_gen (xlator_t *this, int op_no)
                                 rand_no = 0;
                         ret = error_no_list[op_no].error_no[rand_no];
                 }
-                egp->failure_iter_no = 3 + (rand () % GF_UNIVERSAL_ANSWER);
+                if (egp->random_failure == _gf_true)
+                        egp->failure_iter_no = 3 + (rand () % GF_UNIVERSAL_ANSWER);
         }
         return ret;
 }
@@ -371,17 +381,17 @@ error_gen (xlator_t *this, int op_no)
 int
 error_gen_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		      int32_t op_ret, int32_t op_errno, inode_t *inode,
-		      struct iatt *buf, dict_t *dict, struct iatt *postparent)
+		      struct iatt *buf, dict_t *xdata, struct iatt *postparent)
 {
 	STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode,
-                             buf, dict, postparent);
-	return 0;
+                             buf, xdata, postparent);
+        return 0;
 }
 
 
 int
 error_gen_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                  dict_t *xattr_req)
+                  dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -396,36 +406,28 @@ error_gen_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
 		STACK_UNWIND_STRICT (lookup, frame, -1, op_errno, NULL, NULL, NULL,
-                              NULL);
-		return 0;
+                                     NULL);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_lookup_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->lookup,
-		    loc, xattr_req);
-	return 0;
-}
-
-
-int
-error_gen_forget (xlator_t *this, inode_t *inode)
-{
-	return 0;
+		    loc, xdata);
+        return 0;
 }
 
 
 int
 error_gen_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno, struct iatt *buf)
+		    int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf);
-
-	return 0;
+	STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
 }
 
 int
-error_gen_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
+error_gen_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -439,32 +441,31 @@ error_gen_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (stat, frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (stat, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_stat_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->stat,
-		    loc);
-	return 0;
+		    loc, xdata);
+        return 0;
 }
 
 
 int
 error_gen_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                        int32_t op_ret, int32_t op_errno,
-                       struct iatt *preop, struct iatt *postop)
+                       struct iatt *preop, struct iatt *postop, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, preop, postop);
-
-	return 0;
+	STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, preop, postop, xdata);
+        return 0;
 }
 
 
 int
 error_gen_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                   struct iatt *stbuf, int32_t valid)
+                   struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -478,21 +479,21 @@ error_gen_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (setattr, frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (setattr, frame, -1, op_errno, NULL, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_setattr_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->setattr,
-		    loc, stbuf, valid);
-	return 0;
+		    loc, stbuf, valid, xdata);
+        return 0;
 }
 
 
 int
 error_gen_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                    struct iatt *stbuf, int32_t valid)
+                    struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
 	int             op_errno = 0;
         eg_t            *egp = NULL;
@@ -506,32 +507,32 @@ error_gen_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (fsetattr, frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (fsetattr, frame, -1, op_errno, NULL, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_setattr_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->fsetattr,
-		    fd, stbuf, valid);
-	return 0;
+		    fd, stbuf, valid, xdata);
+        return 0;
 }
 
 
 int
 error_gen_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 			int32_t op_ret, int32_t op_errno,
-			struct iatt *prebuf, struct iatt *postbuf)
+			struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
 	STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno,
-                             prebuf, postbuf);
-	return 0;
+                             prebuf, postbuf, xdata);
+        return 0;
 }
 
 
 int
 error_gen_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                    off_t offset)
+                    off_t offset, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -546,32 +547,32 @@ error_gen_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
 		STACK_UNWIND_STRICT (truncate, frame, -1, op_errno,
-                                     NULL, NULL);
-		return 0;
+                                     NULL, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_truncate_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->truncate,
-		    loc, offset);
-	return 0;
+		    loc, offset, xdata);
+        return 0;
 }
 
 
 int
 error_gen_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 			 int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                         struct iatt *postbuf)
+                         struct iatt *postbuf, dict_t *xdata)
 {
 	STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno,
-                             prebuf, postbuf);
-	return 0;
+                             prebuf, postbuf, xdata);
+        return 0;
 }
 
 
 int
 error_gen_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd,
-		     off_t offset)
+		     off_t offset, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp =NULL;
@@ -586,31 +587,30 @@ error_gen_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd,
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
 		STACK_UNWIND_STRICT (ftruncate, frame, -1, op_errno,
-                                     NULL, NULL);
-		return 0;
+                                     NULL, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_ftruncate_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->ftruncate,
-		    fd, offset);
-	return 0;
+		    fd, offset, xdata);
+        return 0;
 }
 
 
 int
 error_gen_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		      int32_t op_ret, int32_t op_errno)
+		      int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (access, frame, op_ret, op_errno);
-
-	return 0;
+	STACK_UNWIND_STRICT (access, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
 error_gen_access (call_frame_t *frame, xlator_t *this, loc_t *loc,
-		  int32_t mask)
+		  int32_t mask, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -624,31 +624,31 @@ error_gen_access (call_frame_t *frame, xlator_t *this, loc_t *loc,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (access, frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (access, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_access_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->access,
-		    loc, mask);
-	return 0;
+		    loc, mask, xdata);
+        return 0;
 }
 
 
 int
 error_gen_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 			int32_t op_ret, int32_t op_errno,
-			const char *path, struct iatt *sbuf)
+			const char *path, struct iatt *sbuf, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path, sbuf);
-	return 0;
+	STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path, sbuf, xdata);
+        return 0;
 }
 
 
 int
 error_gen_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                    size_t size)
+                    size_t size, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -662,15 +662,15 @@ error_gen_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (readlink, frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (readlink, frame, -1, op_errno, NULL, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_readlink_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->readlink,
-		    loc, size);
-	return 0;
+		    loc, size, xdata);
+        return 0;
 }
 
 
@@ -678,18 +678,18 @@ int
 error_gen_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		     int32_t op_ret, int32_t op_errno, inode_t *inode,
                      struct iatt *buf, struct iatt *preparent,
-                     struct iatt *postparent)
+                     struct iatt *postparent, dict_t *xdata)
 {
 	STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno,
                              inode, buf,
-                             preparent, postparent);
-	return 0;
+                             preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
 error_gen_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
-		 mode_t mode, dev_t rdev, dict_t *params)
+		 mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -704,15 +704,15 @@ error_gen_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
 		STACK_UNWIND_STRICT (mknod, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL);
-		return 0;
+                                     NULL, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_mknod_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->mknod,
-		    loc, mode, rdev, params);
-	return 0;
+		    loc, mode, rdev, umask, xdata);
+        return 0;
 }
 
 
@@ -720,17 +720,17 @@ int
 error_gen_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		     int32_t op_ret, int32_t op_errno, inode_t *inode,
                      struct iatt *buf, struct iatt *preparent,
-                     struct iatt *postparent)
+                     struct iatt *postparent, dict_t *xdata)
 {
 	STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno,
                              inode, buf,
-                             preparent, postparent);
-	return 0;
+                             preparent, postparent, xdata);
+        return 0;
 }
 
 int
 error_gen_mkdir (call_frame_t *frame, xlator_t *this,
-		 loc_t *loc, mode_t mode, dict_t *params)
+		 loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -745,31 +745,32 @@ error_gen_mkdir (call_frame_t *frame, xlator_t *this,
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
 		STACK_UNWIND_STRICT (mkdir, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL);
-		return 0;
+                                     NULL, NULL, xdata);
+                return 0;
 	}
 
 	STACK_WIND (frame, error_gen_mkdir_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->mkdir,
-		    loc, mode, params);
-	return 0;
+		    loc, mode, umask, xdata);
+        return 0;
 }
 
 
 int
 error_gen_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		      int32_t op_ret, int32_t op_errno,
-                      struct iatt *preparent, struct iatt *postparent)
+                      struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
 	STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
-                             preparent, postparent);
-	return 0;
+                             preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
-error_gen_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
+error_gen_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+                  dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -783,31 +784,34 @@ error_gen_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL, NULL,
+                                     xdata);
+                return 0;
 	}
 
 	STACK_WIND (frame, error_gen_unlink_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->unlink,
-		    loc);
-	return 0;
+		    loc, xflag, xdata);
+        return 0;
 }
 
 
 int
 error_gen_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		     int32_t op_ret, int32_t op_errno,
-                     struct iatt *preparent, struct iatt *postparent)
+                     struct iatt *preparent, struct iatt *postparent,
+                     dict_t *xdata)
 {
 	STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno,
-                             preparent, postparent);
-	return 0;
+                             preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
-error_gen_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
+error_gen_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+                 dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -821,15 +825,15 @@ error_gen_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (rmdir, frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (rmdir, frame, -1, op_errno, NULL, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_rmdir_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->rmdir,
-		    loc);
-	return 0;
+		    loc, flags, xdata);
+        return 0;
 }
 
 
@@ -837,17 +841,17 @@ int
 error_gen_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		       int32_t op_ret, int32_t op_errno, inode_t *inode,
                        struct iatt *buf, struct iatt *preparent,
-                       struct iatt *postparent)
+                       struct iatt *postparent, dict_t *xdata)
 {
 	STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
-	return 0;
+                             preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
 error_gen_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
-		   loc_t *loc, dict_t *params)
+		   loc_t *loc, mode_t umask, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -862,15 +866,15 @@ error_gen_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
 		STACK_UNWIND_STRICT (symlink, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL); /* pre & post parent attr */
+                                     NULL, NULL, NULL); /* pre & post parent attr */
 		return 0;
 	}
 
 	STACK_WIND (frame, error_gen_symlink_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->symlink,
-		    linkpath, loc, params);
-	return 0;
+		    linkpath, loc, umask, xdata);
+        return 0;
 }
 
 
@@ -878,18 +882,19 @@ int
 error_gen_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		      int32_t op_ret, int32_t op_errno, struct iatt *buf,
                       struct iatt *preoldparent, struct iatt *postoldparent,
-                      struct iatt *prenewparent, struct iatt *postnewparent)
+                      struct iatt *prenewparent, struct iatt *postnewparent,
+                      dict_t *xdata)
 {
 	STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf,
                              preoldparent, postoldparent,
-                             prenewparent, postnewparent);
-	return 0;
+                             prenewparent, postnewparent, xdata);
+        return 0;
 }
 
 
 int
 error_gen_rename (call_frame_t *frame, xlator_t *this,
-		  loc_t *oldloc, loc_t *newloc)
+		  loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -904,15 +909,15 @@ error_gen_rename (call_frame_t *frame, xlator_t *this,
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
 		STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL,
-                                     NULL, NULL, NULL, NULL); /* pre & post parent attr */
+                                     NULL, NULL, NULL, NULL, NULL);
 		return 0;
 	}
 
 	STACK_WIND (frame, error_gen_rename_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->rename,
-		    oldloc, newloc);
-	return 0;
+		    oldloc, newloc, xdata);
+        return 0;
 }
 
 
@@ -920,17 +925,17 @@ int
 error_gen_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		    int32_t op_ret, int32_t op_errno, inode_t *inode,
                     struct iatt *buf, struct iatt *preparent,
-                    struct iatt *postparent)
+                    struct iatt *postparent, dict_t *xdata)
 {
 	STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
-	return 0;
+                             preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
 error_gen_link (call_frame_t *frame, xlator_t *this,
-		loc_t *oldloc, loc_t *newloc)
+		loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
 	int             op_errno = 0;
         eg_t            *egp = NULL;
@@ -945,15 +950,15 @@ error_gen_link (call_frame_t *frame, xlator_t *this,
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
 		STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL); /* pre & post parent attr */
+                                     NULL, NULL, NULL);
 		return 0;
 	}
 
 	STACK_WIND (frame, error_gen_link_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->link,
-		    oldloc, newloc);
-	return 0;
+		    oldloc, newloc, xdata);
+        return 0;
 }
 
 
@@ -961,17 +966,18 @@ int
 error_gen_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		      int32_t op_ret, int32_t op_errno,
 		      fd_t *fd, inode_t *inode, struct iatt *buf,
-                      struct iatt *preparent, struct iatt *postparent)
+                      struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
 	STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
-                             preparent, postparent);
-	return 0;
+                             preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
 error_gen_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
-		  int32_t flags, mode_t mode, fd_t *fd, dict_t *params)
+		  int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
+                  dict_t *xdata)
 {
 	int             op_errno = 0;
         eg_t            *egp = NULL;
@@ -986,30 +992,30 @@ error_gen_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
 		STACK_UNWIND_STRICT (create, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL, NULL); /* pre & post attr */
+                                     NULL, NULL, NULL, NULL);
 		return 0;
 	}
 
 	STACK_WIND (frame, error_gen_create_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->create,
-		    loc, flags, mode, fd, params);
-	return 0;
+		    loc, flags, mode, umask, fd, xdata);
+        return 0;
 }
 
 
 int
 error_gen_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno, fd_t *fd)
+		    int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd);
-	return 0;
+	STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+        return 0;
 }
 
 
 int
 error_gen_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
-		int32_t flags, fd_t *fd, int32_t wbflags)
+		int32_t flags, fd_t *fd, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1023,15 +1029,15 @@ error_gen_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (open, frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (open, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_open_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->open,
-		    loc, flags, fd, wbflags);
-	return 0;
+		    loc, flags, fd, xdata);
+        return 0;
 }
 
 
@@ -1039,17 +1045,17 @@ int
 error_gen_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		     int32_t op_ret, int32_t op_errno,
 		     struct iovec *vector, int32_t count,
-		     struct iatt *stbuf, struct iobref *iobref)
+		     struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
 {
 	STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno,
-                             vector, count, stbuf, iobref);
-	return 0;
+                             vector, count, stbuf, iobref, xdata);
+        return 0;
 }
 
 
 int
 error_gen_readv (call_frame_t *frame, xlator_t *this,
-		 fd_t *fd, size_t size, off_t offset)
+		 fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1064,33 +1070,33 @@ error_gen_readv (call_frame_t *frame, xlator_t *this,
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
 		STACK_UNWIND_STRICT (readv, frame, -1, op_errno, NULL, 0,
-                                     NULL, NULL);
-		return 0;
+                                     NULL, NULL, xdata);
+        return 0;
 	}
 
 
 	STACK_WIND (frame, error_gen_readv_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->readv,
-		    fd, size, offset);
-	return 0;
+		    fd, size, offset, flags, xdata);
+        return 0;
 }
 
 
 int
 error_gen_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		      int32_t op_ret, int32_t op_errno,
-                      struct iatt *prebuf, struct iatt *postbuf)
+                      struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf);
-	return 0;
+	STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+        return 0;
 }
 
 
 int
 error_gen_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
 		  struct iovec *vector, int32_t count,
-		  off_t off, struct iobref *iobref)
+		  off_t off, uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1102,31 +1108,47 @@ error_gen_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
         if (enable)
                 op_errno = error_gen (this, GF_FOP_WRITE);
 
-	if (op_errno) {
-		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL);
+	if (op_errno == GF_ERROR_SHORT_WRITE) {
+		struct iovec *shortvec;
+
+		/*
+		 * A short write error returns some value less than what was
+		 * requested from a write. To simulate this, replace the vector
+		 * with one half the size;
+		 */
+		shortvec = iov_dup(vector, 1);
+		shortvec->iov_len /= 2;
+
+		STACK_WIND(frame, error_gen_writev_cbk, FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->writev, fd, shortvec, count,
+			   off, flags, iobref, xdata);
+		GF_FREE(shortvec);
 		return 0;
+	} else if (op_errno) {
+		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
+		STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL, xdata);
+        	return 0;
 	}
 
 	STACK_WIND (frame, error_gen_writev_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->writev,
-		    fd, vector, count, off, iobref);
-	return 0;
+		    fd, vector, count, off, flags, iobref, xdata);
+        return 0;
 }
 
 
 int
 error_gen_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		     int32_t op_ret, int32_t op_errno)
+		     int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno);
-	return 0;
+	STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
-error_gen_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
+error_gen_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1140,15 +1162,15 @@ error_gen_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (flush, frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (flush, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_flush_cbk,
                     FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->flush,
-		    fd);
-	return 0;
+		    fd, xdata);
+        return 0;
 }
 
 
@@ -1156,15 +1178,15 @@ int
 error_gen_fsync_cbk (call_frame_t *frame, void *cookie,
 		     xlator_t *this, int32_t op_ret,
 		     int32_t op_errno, struct iatt *prebuf,
-                     struct iatt *postbuf)
+                     struct iatt *postbuf, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf);
-	return 0;
+	STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+        return 0;
 }
 
 
 int
-error_gen_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
+error_gen_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1178,29 +1200,29 @@ error_gen_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (fsync, frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (fsync, frame, -1, op_errno, NULL, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_fsync_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->fsync,
-		    fd, flags);
-	return 0;
+		    fd, flags, xdata);
+        return 0;
 }
 
 
 int
 error_gen_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                     int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf);
-	return 0;
+	STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
 }
 
 
 int
-error_gen_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd)
+error_gen_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1214,29 +1236,29 @@ error_gen_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd)
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (fstat, frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (fstat, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_fstat_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->fstat,
-		    fd);
-	return 0;
+		    fd, xdata);
+        return 0;
 }
 
 
 int
 error_gen_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno, fd_t *fd)
+		       int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd);
-	return 0;
+	STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
+        return 0;
 }
 
 
 int
-error_gen_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
+error_gen_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1250,29 +1272,29 @@ error_gen_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (opendir, frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (opendir, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_opendir_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->opendir,
-		    loc, fd);
-	return 0;
+		    loc, fd, xdata);
+        return 0;
 }
 
 int
 error_gen_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int32_t op_ret, int32_t op_errno)
+			int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno);
-	return 0;
+	STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
 error_gen_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                    int32_t flags)
+                    int32_t flags, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1286,30 +1308,29 @@ error_gen_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (fsyncdir, frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (fsyncdir, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_fsyncdir_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->fsyncdir,
-		    fd, flags);
-	return 0;
+		    fd, flags, xdata);
+        return 0;
 }
 
 
 int
 error_gen_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		      int32_t op_ret, int32_t op_errno, struct statvfs *buf)
+		      int32_t op_ret, int32_t op_errno, struct statvfs *buf, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf);
-
-	return 0;
+	STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
+        return 0;
 }
 
 
 int
-error_gen_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
+error_gen_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1323,31 +1344,30 @@ error_gen_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (statfs, frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (statfs, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_statfs_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->statfs,
-		    loc);
-	return 0;
+		    loc, xdata);
+        return 0;
 }
 
 
 int
 error_gen_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int32_t op_ret, int32_t op_errno)
+			int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno);
-
-	return 0;
+	STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
 error_gen_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-		    dict_t *dict, int32_t flags)
+		    dict_t *dict, int32_t flags, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1361,30 +1381,30 @@ error_gen_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (setxattr, frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (setxattr, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_setxattr_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->setxattr,
-		    loc, dict, flags);
-	return 0;
+		    loc, dict, flags, xdata);
+        return 0;
 }
 
 
 int
 error_gen_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int32_t op_ret, int32_t op_errno, dict_t *dict)
+			int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict);
-	return 0;
+	STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
 }
 
 
 int
 error_gen_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-		    const char *name)
+		    const char *name, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1398,31 +1418,103 @@ error_gen_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (getxattr, frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (getxattr, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_getxattr_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->getxattr,
-		    loc, name);
-	return 0;
+		    loc, name, xdata);
+        return 0;
+}
+
+int
+error_gen_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
-error_gen_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno, dict_t *dict)
+error_gen_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                     dict_t *dict, int32_t flags, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (xattrop, frame, op_ret, op_errno, dict);
+	int              op_errno = 0;
+        eg_t            *egp = NULL;
+        int              enable = 1;
 
-	return 0;
+        egp = this->private;
+        enable = egp->enable[GF_FOP_FSETXATTR];
+
+        if (enable)
+                op_errno = error_gen (this, GF_FOP_FSETXATTR);
+
+	if (op_errno) {
+		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
+		STACK_UNWIND_STRICT (fsetxattr, frame, -1, op_errno, xdata);
+        return 0;
+	}
+
+	STACK_WIND (frame, error_gen_fsetxattr_cbk,
+		    FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->fsetxattr,
+		    fd, dict, flags, xdata);
+        return 0;
+}
+
+
+int
+error_gen_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+
+int
+error_gen_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                     const char *name, dict_t *xdata)
+{
+	int              op_errno = 0;
+        eg_t            *egp = NULL;
+        int              enable = 1;
+
+        egp = this->private;
+        enable = egp->enable[GF_FOP_FGETXATTR];
+
+        if (enable)
+                op_errno = error_gen (this, GF_FOP_FGETXATTR);
+
+	if (op_errno) {
+		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
+		STACK_UNWIND_STRICT (fgetxattr, frame, -1, op_errno, NULL, xdata);
+        return 0;
+	}
+
+	STACK_WIND (frame, error_gen_fgetxattr_cbk,
+		    FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->fgetxattr,
+		    fd, name, xdata);
+        return 0;
+}
+
+
+int
+error_gen_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (xattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
 }
 
 
 int
 error_gen_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
-		   gf_xattrop_flags_t flags, dict_t *dict)
+		   gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1436,31 +1528,30 @@ error_gen_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (xattrop, frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (xattrop, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_xattrop_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->xattrop,
-		    loc, flags, dict);
-	return 0;
+		    loc, flags, dict, xdata);
+        return 0;
 }
 
 
 int
 error_gen_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int32_t op_ret, int32_t op_errno, dict_t *dict)
+			int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (fxattrop, frame, op_ret, op_errno, dict);
-
-	return 0;
+	STACK_UNWIND_STRICT (fxattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
 }
 
 
 int
 error_gen_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
-		    gf_xattrop_flags_t flags, dict_t *dict)
+		    gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1474,31 +1565,30 @@ error_gen_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (fxattrop, frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (fxattrop, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_fxattrop_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->fxattrop,
-		    fd, flags, dict);
-	return 0;
+		    fd, flags, dict, xdata);
+        return 0;
 }
 
 
 int
 error_gen_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			   int32_t op_ret, int32_t op_errno)
+			   int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno);
-
-	return 0;
+	STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
 error_gen_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-		       const char *name)
+		       const char *name, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1512,30 +1602,66 @@ error_gen_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (removexattr, frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (removexattr, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_removexattr_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->removexattr,
-		    loc, name);
-	return 0;
+		    loc, name, xdata);
+        return 0;
+}
+
+int
+error_gen_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			   int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+
+int
+error_gen_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                        const char *name, dict_t *xdata)
+{
+	int              op_errno = 0;
+        eg_t            *egp = NULL;
+        int              enable = 1;
+
+        egp = this->private;
+        enable = egp->enable[GF_FOP_FREMOVEXATTR];
+
+        if (enable)
+                op_errno = error_gen (this, GF_FOP_FREMOVEXATTR);
+
+	if (op_errno) {
+		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
+		STACK_UNWIND_STRICT (fremovexattr, frame, -1, op_errno, xdata);
+        return 0;
+	}
+
+	STACK_WIND (frame, error_gen_fremovexattr_cbk,
+		    FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->fremovexattr,
+		    fd, name, xdata);
+        return 0;
 }
 
 
 int
 error_gen_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int32_t op_ret, int32_t op_errno, struct flock *lock)
+		  int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock);
-	return 0;
+	STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock, xdata);
+        return 0;
 }
 
 
 int
 error_gen_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
-	      struct flock *lock)
+	      struct gf_flock *lock, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1549,32 +1675,31 @@ error_gen_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (lk, frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (lk, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_lk_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->lk,
-		    fd, cmd, lock);
-	return 0;
+		    fd, cmd, lock, xdata);
+        return 0;
 }
 
 
 int
-error_gen_inodelk_cbk (call_frame_t *frame, void *cookie,
-		       xlator_t *this, int32_t op_ret, int32_t op_errno)
-
+error_gen_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno);
-	return 0;
+	STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
 error_gen_inodelk (call_frame_t *frame, xlator_t *this,
 		   const char *volume, loc_t *loc, int32_t cmd,
-                   struct flock *lock)
+                   struct gf_flock *lock, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1588,32 +1713,31 @@ error_gen_inodelk (call_frame_t *frame, xlator_t *this,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (inodelk, frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (inodelk, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_inodelk_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->inodelk,
-		    volume, loc, cmd, lock);
-	return 0;
+		    volume, loc, cmd, lock, xdata);
+        return 0;
 }
 
 
 int
-error_gen_finodelk_cbk (call_frame_t *frame, void *cookie,
-			xlator_t *this, int32_t op_ret, int32_t op_errno)
-
+error_gen_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (finodelk, frame, op_ret, op_errno);
-	return 0;
+	STACK_UNWIND_STRICT (finodelk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
 error_gen_finodelk (call_frame_t *frame, xlator_t *this,
 		    const char *volume, fd_t *fd, int32_t cmd,
-                    struct flock *lock)
+                    struct gf_flock *lock, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1627,32 +1751,31 @@ error_gen_finodelk (call_frame_t *frame, xlator_t *this,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (finodelk, frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (finodelk, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_finodelk_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->finodelk,
-		    volume, fd, cmd, lock);
-	return 0;
+		    volume, fd, cmd, lock, xdata);
+        return 0;
 }
 
 
 int
-error_gen_entrylk_cbk (call_frame_t *frame, void *cookie,
-		       xlator_t *this, int32_t op_ret, int32_t op_errno)
-
+error_gen_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno);
-	return 0;
+	STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
 error_gen_entrylk (call_frame_t *frame, xlator_t *this,
 		   const char *volume, loc_t *loc, const char *basename,
-		   entrylk_cmd cmd, entrylk_type type)
+		   entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1666,32 +1789,31 @@ error_gen_entrylk (call_frame_t *frame, xlator_t *this,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (entrylk, frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (entrylk, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_entrylk_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->entrylk,
-		    volume, loc, basename, cmd, type);
-	return 0;
+		    volume, loc, basename, cmd, type, xdata);
+        return 0;
 }
 
 
 int
-error_gen_fentrylk_cbk (call_frame_t *frame, void *cookie,
-			xlator_t *this, int32_t op_ret, int32_t op_errno)
-
+error_gen_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (fentrylk, frame, op_ret, op_errno);
-	return 0;
+	STACK_UNWIND_STRICT (fentrylk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
 error_gen_fentrylk (call_frame_t *frame, xlator_t *this,
 		    const char *volume, fd_t *fd, const char *basename,
-		    entrylk_cmd cmd, entrylk_type type)
+		    entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
 	int             op_errno = 0;
         eg_t            *egp = NULL;
@@ -1705,15 +1827,15 @@ error_gen_fentrylk (call_frame_t *frame, xlator_t *this,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (fentrylk, frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (fentrylk, frame, -1, op_errno, xdata);
+                return 0;
 	}
 
 	STACK_WIND (frame, error_gen_fentrylk_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->fentrylk,
-		    volume, fd, basename, cmd, type);
-	return 0;
+		    volume, fd, basename, cmd, type, xdata);
+        return 0;
 }
 
 
@@ -1725,8 +1847,7 @@ error_gen_getspec_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 		       int32_t op_ret, int32_t op_errno, char *spec_data)
 {
 	STACK_UNWIND_STRICT (getspec, frame, op_ret, op_errno, spec_data);
-
-	return 0;
+        return 0;
 }
 
 
@@ -1760,16 +1881,17 @@ error_gen_getspec (call_frame_t *frame, xlator_t *this, const char *key,
 
 int
 error_gen_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno, gf_dirent_t *entries)
+		       int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                       dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, entries);
+	STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, entries, xdata);
 	return 0;
 }
 
 
 int
 error_gen_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
-		   size_t size, off_t off)
+		   size_t size, off_t off, dict_t *xdata)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1783,30 +1905,31 @@ error_gen_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (readdir, frame, -1, op_errno, NULL);
+		STACK_UNWIND_STRICT (readdir, frame, -1, op_errno, NULL, xdata);
 		return 0;
 	}
 
 	STACK_WIND (frame, error_gen_readdir_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->readdir,
-		    fd, size, off);
+		    fd, size, off, xdata);
 	return 0;
 }
 
 
 int
 error_gen_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno, gf_dirent_t *entries)
+                        int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                        dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries);
+	STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
 	return 0;
 }
 
 
 int
 error_gen_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-                    off_t off)
+                    off_t off, dict_t *dict)
 {
 	int              op_errno = 0;
         eg_t            *egp = NULL;
@@ -1820,44 +1943,153 @@ error_gen_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND_STRICT (readdirp, frame, -1, op_errno, NULL);
+		STACK_UNWIND_STRICT (readdirp, frame, -1, op_errno, NULL, NULL);
 		return 0;
 	}
 
 	STACK_WIND (frame, error_gen_readdirp_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readdirp,
-                    fd, size, off);
+                    fd, size, off, dict);
 	return 0;
 }
 
+static void
+error_gen_set_failure (eg_t *pvt, int percent)
+{
+        GF_ASSERT (pvt);
 
-int
-error_gen_closedir (xlator_t *this, fd_t *fd)
+        if (percent)
+                pvt->failure_iter_no = 100/percent;
+        else
+                pvt->failure_iter_no = 100/GF_FAILURE_DEFAULT;
+}
+
+static void
+error_gen_parse_fill_fops (eg_t *pvt, char *enable_fops)
 {
-	return 0;
+        char            *op_no_str = NULL;
+        int              op_no = -1;
+        int              i = 0;
+        xlator_t        *this = THIS;
+        char            *saveptr = NULL;
+
+        GF_ASSERT (pvt);
+        GF_ASSERT (this);
+
+        for (i = 0; i < GF_FOP_MAXVALUE; i++)
+                pvt->enable[i] = 0;
+
+        if (!enable_fops) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "All fops are enabled.");
+                for (i = 0; i < GF_FOP_MAXVALUE; i++)
+                        pvt->enable[i] = 1;
+        } else {
+                op_no_str = strtok_r (enable_fops, ",", &saveptr);
+                while (op_no_str) {
+                        op_no = get_fop_int (&op_no_str);
+                        if (op_no == -1) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "Wrong option value %s", op_no_str);
+                        } else
+                                pvt->enable[op_no] = 1;
+
+                        op_no_str = strtok_r (NULL, ",", &saveptr);
+                }
+        }
 }
 
+int32_t
+error_gen_priv_dump (xlator_t *this)
+{
+        char            key_prefix[GF_DUMP_MAX_BUF_LEN];
+        int             ret = -1;
+        eg_t            *conf = NULL;
 
-int
-error_gen_close (xlator_t *this, fd_t *fd)
+        if (!this)
+            goto out;
+
+        conf = this->private;
+        if (!conf)
+            goto out;
+
+        ret = TRY_LOCK(&conf->lock);
+        if (ret != 0) {
+                return ret;
+        }
+
+        gf_proc_dump_add_section("xlator.debug.error-gen.%s.priv", this->name);
+        gf_proc_dump_build_key(key_prefix,"xlator.debug.error-gen","%s.priv",
+                               this->name);
+
+        gf_proc_dump_write("op_count", "%d", conf->op_count);
+        gf_proc_dump_write("failure_iter_no", "%d", conf->failure_iter_no);
+        gf_proc_dump_write("error_no", "%s", conf->error_no);
+        gf_proc_dump_write("random_failure", "%d", conf->random_failure);
+
+        UNLOCK(&conf->lock);
+out:
+        return ret;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
 {
-	return 0;
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_error_gen_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                        " failed");
+                return ret;
+        }
+
+        return ret;
 }
 
+int
+reconfigure (xlator_t *this, dict_t *options)
+{
+        eg_t            *pvt = NULL;
+        int32_t          ret = 0;
+        char            *error_enable_fops = NULL;
+        int32_t          failure_percent_int = 0;
+
+        if (!this || !this->private)
+                goto out;
+
+        pvt = this->private;
+
+        GF_OPTION_RECONF ("error-no", pvt->error_no, options, str, out);
+
+        GF_OPTION_RECONF ("failure", failure_percent_int, options, int32,
+                          out);
+
+        GF_OPTION_RECONF ("enable", error_enable_fops, options, str, out);
+
+        GF_OPTION_RECONF ("random-failure", pvt->random_failure, options,
+                          bool, out);
+
+        error_gen_parse_fill_fops (pvt, error_enable_fops);
+        error_gen_set_failure (pvt, failure_percent_int);
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "reconfigure returning %d", ret);
+        return ret;
+}
 
 int
 init (xlator_t *this)
 {
         eg_t            *pvt = NULL;
-        data_t          *error_no = NULL;
-        data_t          *failure_percent = NULL;
-        data_t          *enable = NULL;
         int32_t          ret = 0;
         char            *error_enable_fops = NULL;
-        char            *op_no_str = NULL;
-        int              op_no = -1;
-        int              i = 0;
         int32_t          failure_percent_int = 0;
 
         if (!this->children || this->children->next) {
@@ -1872,74 +2104,34 @@ init (xlator_t *this)
                         "dangling volume. check volfile ");
         }
 
-        error_no = dict_get (this->options, "error-no");
-        failure_percent = dict_get (this->options, "failure");
-        enable = dict_get (this->options, "enable");
-
-        pvt = CALLOC (1, sizeof (eg_t));
+        pvt = GF_CALLOC (1, sizeof (eg_t), gf_error_gen_mt_eg_t);
 
         if (!pvt) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "out of memory.");
                 ret = -1;
                 goto out;
         }
 
         LOCK_INIT (&pvt->lock);
 
-        for (i = 0; i < GF_FOP_MAXVALUE; i++)
-                pvt->enable[i] = 0;
-        if (!error_no) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "error-no not specified.");
-        } else {
-                pvt->error_no = data_to_str (error_no);
-        }
+        GF_OPTION_INIT ("error-no", pvt->error_no, str, out);
 
-        if (!failure_percent) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "failure percent not specified.");
-                pvt->failure_iter_no = 100/GF_FAILURE_DEFAULT;
-        } else {
-                failure_percent_int = data_to_int32 (failure_percent);
-                if (failure_percent_int)
-                        pvt->failure_iter_no = 100/failure_percent_int;
-                else
-                        pvt->failure_iter_no = 100/GF_FAILURE_DEFAULT;
-        }
+        GF_OPTION_INIT ("failure", failure_percent_int, int32, out);
+
+        GF_OPTION_INIT ("enable", error_enable_fops, str, out);
+
+        GF_OPTION_INIT ("random-failure", pvt->random_failure, bool, out);
+
+
+        error_gen_parse_fill_fops (pvt, error_enable_fops);
+        error_gen_set_failure (pvt, failure_percent_int);
 
-        if (!enable) {
-                gf_log (this->name, GF_LOG_WARNING,
-                        "All fops are enabled.");
-                for (i = 0; i < GF_FOP_MAXVALUE; i++)
-                        pvt->enable[i] = 1;
-        } else {
-                error_enable_fops = data_to_str (enable);
-                op_no_str = error_enable_fops;
-                while ((*error_enable_fops) != '\0') {
-                        error_enable_fops++;
-                        if (((*error_enable_fops) == ',') ||
-                            ((*error_enable_fops) == '\0')) {
-                                if ((*error_enable_fops) != '\0') {
-                                        (*error_enable_fops) = '\0';
-                                        error_enable_fops++;
-                                }
-                                op_no = get_fop_int (&op_no_str);
-                                if (op_no == -1) {
-                                        gf_log (this->name, GF_LOG_WARNING,
-                                                "Wrong option value %s",
-                                                op_no_str);
-                                } else
-                                        pvt->enable[op_no] = 1;
-                                op_no_str = error_enable_fops;
-                        }
-                }
-        }
         this->private = pvt;
 
         /* Give some seed value here */
         srand (time(NULL));
 out:
+        if (ret)
+                GF_FREE (pvt);
         return ret;
 }
 
@@ -1955,12 +2147,18 @@ fini (xlator_t *this)
 
         if (pvt) {
                 LOCK_DESTROY (&pvt->lock);
-                FREE (pvt);
+                GF_FREE (pvt);
                 gf_log (this->name, GF_LOG_DEBUG, "fini called");
         }
         return;
 }
 
+struct xlator_dumpops dumpops = {
+        .priv = error_gen_priv_dump,
+};
+
+struct xlator_fops cbks;
+
 struct xlator_fops fops = {
 	.lookup      = error_gen_lookup,
 	.stat        = error_gen_stat,
@@ -1983,6 +2181,9 @@ struct xlator_fops fops = {
 	.setxattr    = error_gen_setxattr,
 	.getxattr    = error_gen_getxattr,
 	.removexattr = error_gen_removexattr,
+	.fsetxattr    = error_gen_fsetxattr,
+	.fgetxattr    = error_gen_fgetxattr,
+	.fremovexattr = error_gen_fremovexattr,
 	.opendir     = error_gen_opendir,
 	.readdir     = error_gen_readdir,
 	.readdirp    = error_gen_readdirp,
@@ -2003,22 +2204,29 @@ struct xlator_fops fops = {
 	.getspec     = error_gen_getspec,
 };
 
-struct xlator_cbks cbks = {
-	.release = error_gen_close,
-	.releasedir = error_gen_closedir,
-};
-
 struct volume_options options[] = {
         { .key  = {"failure"},
-          .type = GF_OPTION_TYPE_INT },
+          .type = GF_OPTION_TYPE_INT,
+          .description = "Percentage failure of operations when enabled.",
+        },
+
         { .key  = {"error-no"},
           .value = {"ENOENT","ENOTDIR","ENAMETOOLONG","EACCES","EBADF",
                     "EFAULT","ENOMEM","EINVAL","EIO","EEXIST","ENOSPC",
                     "EPERM","EROFS","EBUSY","EISDIR","ENOTEMPTY","EMLINK"
                     "ENODEV","EXDEV","EMFILE","ENFILE","ENOSYS","EINTR",
-                    "EFBIG","EAGAIN"},
-          .type = GF_OPTION_TYPE_STR },
+                    "EFBIG","EAGAIN","GF_ERROR_SHORT_WRITE"},
+          .type = GF_OPTION_TYPE_STR,
+        },
+
+        { .key = {"random-failure"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+        },
+
         { .key  = {"enable"},
-          .type = GF_OPTION_TYPE_STR },
+          .type = GF_OPTION_TYPE_STR,
+        },
+
         { .key  = {NULL} }
 };
diff --git a/xlators/debug/error-gen/src/error-gen.h b/xlators/debug/error-gen/src/error-gen.h
index fc51fafc3..d92c23062 100644
--- a/xlators/debug/error-gen/src/error-gen.h
+++ b/xlators/debug/error-gen/src/error-gen.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _ERROR_GEN_H
 #define _ERROR_GEN_H
 
@@ -25,13 +15,28 @@
 #include "config.h"
 #endif
 
+#include "error-gen-mem-types.h"
+
 #define GF_FAILURE_DEFAULT 10
 
+/*
+ * Pseudo-errors refer to errors beyond the scope of traditional <-1, op_errno>
+ * returns. This facilitates the ability to return unexpected, but not -1 values
+ * and/or to inject operations that lead to implicit error conditions. The range
+ * for pseudo errors resides at a high value to avoid conflicts with the errno
+ * range.
+ */
+enum GF_PSEUDO_ERRORS {
+	GF_ERROR_SHORT_WRITE = 1000,	/* short writev return value */
+	GF_ERROR_MAX
+};
+
 typedef struct {
         int enable[GF_FOP_MAXVALUE];
         int op_count;
         int failure_iter_no;
         char *error_no;
+        gf_boolean_t random_failure;
         gf_lock_t lock;
 } eg_t;
 
diff --git a/xlators/debug/io-stats/src/Makefile.am b/xlators/debug/io-stats/src/Makefile.am
index b894e79c3..332d79015 100644
--- a/xlators/debug/io-stats/src/Makefile.am
+++ b/xlators/debug/io-stats/src/Makefile.am
@@ -2,14 +2,15 @@
 xlator_LTLIBRARIES = io-stats.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/debug
 
-io_stats_la_LDFLAGS = -module -avoidversion
+io_stats_la_LDFLAGS = -module -avoid-version
 
 io_stats_la_SOURCES = io-stats.c
 io_stats_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
 noinst_HEADERS = io-stats-mem-types.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES =
diff --git a/xlators/debug/io-stats/src/io-stats-mem-types.h b/xlators/debug/io-stats/src/io-stats-mem-types.h
index d9b434d57..c30dfb17e 100644
--- a/xlators/debug/io-stats/src/io-stats-mem-types.h
+++ b/xlators/debug/io-stats/src/io-stats-mem-types.h
@@ -1,24 +1,13 @@
-
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
 
-
 #ifndef __IO_STATS_MEM_TYPES_H__
 #define __IO_STATS_MEM_TYPES_H__
 
@@ -27,6 +16,8 @@
 enum gf_io_stats_mem_types_ {
         gf_io_stats_mt_ios_conf = gf_common_mt_end + 1,
         gf_io_stats_mt_ios_fd,
+        gf_io_stats_mt_ios_stat,
+        gf_io_stats_mt_ios_stat_list,
         gf_io_stats_mt_end
 };
 #endif
diff --git a/xlators/debug/io-stats/src/io-stats.c b/xlators/debug/io-stats/src/io-stats.c
index ee0e93365..7fb697ae4 100644
--- a/xlators/debug/io-stats/src/io-stats.c
+++ b/xlators/debug/io-stats/src/io-stats.c
@@ -1,25 +1,16 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
+#include "xlator.h"
 #endif
 
 /**
@@ -42,7 +33,62 @@
 #include "glusterfs.h"
 #include "xlator.h"
 #include "io-stats-mem-types.h"
+#include <stdarg.h>
+#include "defaults.h"
+#include "logging.h"
+
+#define MAX_LIST_MEMBERS 100
+
+typedef enum {
+        IOS_STATS_TYPE_NONE,
+        IOS_STATS_TYPE_OPEN,
+        IOS_STATS_TYPE_READ,
+        IOS_STATS_TYPE_WRITE,
+        IOS_STATS_TYPE_OPENDIR,
+        IOS_STATS_TYPE_READDIRP,
+        IOS_STATS_TYPE_READ_THROUGHPUT,
+        IOS_STATS_TYPE_WRITE_THROUGHPUT,
+        IOS_STATS_TYPE_MAX
+}ios_stats_type_t;
+
+typedef enum {
+        IOS_STATS_THRU_READ,
+        IOS_STATS_THRU_WRITE,
+        IOS_STATS_THRU_MAX,
+}ios_stats_thru_t;
+
+struct ios_stat_lat {
+        struct timeval  time;
+        double          throughput;
+};
 
+struct ios_stat {
+        gf_lock_t       lock;
+        uuid_t          gfid;
+        char           *filename;
+        uint64_t        counters [IOS_STATS_TYPE_MAX];
+        struct ios_stat_lat thru_counters [IOS_STATS_THRU_MAX];
+        int             refcnt;
+};
+
+struct ios_stat_list {
+        struct list_head  list;
+        struct ios_stat  *iosstat;
+        double            value;
+};
+
+struct ios_stat_head {
+       gf_lock_t                lock;
+       double                   min_cnt;
+       uint64_t                 members;
+       struct ios_stat_list    *iosstats;
+};
+
+struct ios_lat {
+        double  min;
+        double  max;
+        double  avg;
+};
 
 struct ios_global_stats {
         uint64_t        data_written;
@@ -51,6 +97,10 @@ struct ios_global_stats {
         uint64_t        block_count_read[32];
         uint64_t        fop_hits[GF_FOP_MAXVALUE];
         struct timeval  started_at;
+        struct ios_lat  latency[GF_FOP_MAXVALUE];
+        uint64_t        nr_opens;
+        uint64_t        max_nr_opens;
+        struct timeval  max_openfd_time;
 };
 
 
@@ -60,6 +110,10 @@ struct ios_conf {
         uint64_t                  increment;
         struct ios_global_stats   incremental;
         gf_boolean_t              dump_fd_stats;
+        gf_boolean_t              count_fop_hits;
+        gf_boolean_t              measure_latency;
+        struct ios_stat_head      list[IOS_STATS_TYPE_MAX];
+        struct ios_stat_head      thru_list[IOS_STATS_THRU_MAX];
 };
 
 
@@ -72,27 +126,93 @@ struct ios_fd {
         struct timeval  opened_at;
 };
 
+typedef enum {
+        IOS_DUMP_TYPE_NONE = 0,
+        IOS_DUMP_TYPE_FILE = 1,
+        IOS_DUMP_TYPE_DICT = 2,
+        IOS_DUMP_TYPE_MAX  = 3
+} ios_dump_type_t;
+
+struct ios_dump_args {
+        ios_dump_type_t type;
+        union {
+                FILE *logfp;
+                dict_t *dict;
+        } u;
+};
+
+typedef int (*block_dump_func) (xlator_t *, struct ios_dump_args*,
+                                    int , int , uint64_t ) ;
 
 struct ios_local {
         struct timeval  wind_at;
         struct timeval  unwind_at;
 };
 
+struct volume_options options[];
 
-#define BUMP_FOP(op)                                                    \
+inline static int
+is_fop_latency_started (call_frame_t *frame)
+{
+        GF_ASSERT (frame);
+        struct timeval epoch = {0,};
+        return memcmp (&frame->begin, &epoch, sizeof (epoch));
+}
+
+#define END_FOP_LATENCY(frame, op)                                      \
         do {                                                            \
                 struct ios_conf  *conf = NULL;                          \
                                                                         \
                 conf = this->private;                                   \
-                LOCK (&conf->lock);                                     \
-                {                                                       \
-                        conf->cumulative.fop_hits[GF_FOP_##op]++;       \
-                        conf->incremental.fop_hits[GF_FOP_##op]++;      \
+                if (conf && conf->measure_latency) {                    \
+                        gettimeofday (&frame->end, NULL);               \
+                        update_ios_latency (conf, frame, GF_FOP_##op);  \
                 }                                                       \
-                UNLOCK (&conf->lock);                                   \
+        } while (0)
+
+#define START_FOP_LATENCY(frame)                                         \
+        do {                                                             \
+                struct ios_conf  *conf = NULL;                           \
+                                                                         \
+                conf = this->private;                                    \
+                if (conf && conf->measure_latency) {                     \
+                        gettimeofday (&frame->begin, NULL);              \
+                } else {                                                 \
+                        memset (&frame->begin, 0, sizeof (frame->begin));\
+                }                                                        \
         } while (0)
 
 
+#define BUMP_FOP(op)                                                    \
+        do {                                                            \
+                struct ios_conf  *conf = NULL;                          \
+                                                                        \
+                conf = this->private;                                   \
+                if (!conf)                                              \
+                        break;                                          \
+                conf->cumulative.fop_hits[GF_FOP_##op]++;               \
+                conf->incremental.fop_hits[GF_FOP_##op]++;              \
+        } while (0)
+
+#define UPDATE_PROFILE_STATS(frame, op)                                       \
+        do {                                                                  \
+                struct ios_conf  *conf = NULL;                                \
+                                                                              \
+                if (!is_fop_latency_started (frame))                          \
+                        break;                                                \
+                conf = this->private;                                         \
+                LOCK (&conf->lock);                                           \
+                {                                                             \
+                        if (conf && conf->measure_latency &&                  \
+                            conf->count_fop_hits) {                           \
+                                BUMP_FOP(op);                                 \
+                                gettimeofday (&frame->end, NULL);             \
+                                update_ios_latency (conf, frame, GF_FOP_##op);\
+                        }                                                     \
+                }                                                             \
+                UNLOCK (&conf->lock);                                         \
+        } while (0)
+
 #define BUMP_READ(fd, len)                                              \
         do {                                                            \
                 struct ios_conf  *conf = NULL;                          \
@@ -102,6 +222,8 @@ struct ios_local {
                 conf = this->private;                                   \
                 lb2 = log_base2 (len);                                  \
                 ios_fd_ctx_get (fd, this, &iosfd);                      \
+                if (!conf)                                              \
+                        break;                                          \
                                                                         \
                 LOCK (&conf->lock);                                     \
                 {                                                       \
@@ -128,6 +250,8 @@ struct ios_local {
                 conf = this->private;                                   \
                 lb2 = log_base2 (len);                                  \
                 ios_fd_ctx_get (fd, this, &iosfd);                      \
+                if (!conf)                                              \
+                        break;                                          \
                                                                         \
                 LOCK (&conf->lock);                                     \
                 {                                                       \
@@ -145,6 +269,58 @@ struct ios_local {
         } while (0)
 
 
+#define BUMP_STATS(iosstat, type)                                               \
+        do {                                                                    \
+                struct ios_conf         *conf = NULL;                           \
+                uint64_t                 value = 0;                             \
+                                                                                \
+                conf = this->private;                                           \
+                                                                                \
+                LOCK(&iosstat->lock);                                           \
+                {                                                               \
+                        iosstat->counters[type]++;                              \
+                        value = iosstat->counters[type];                        \
+                }                                                               \
+                UNLOCK (&iosstat->lock);                                        \
+                ios_stat_add_to_list (&conf->list[type],                        \
+                                     value, iosstat);                           \
+                                                                                \
+        } while (0)
+
+
+#define BUMP_THROUGHPUT(iosstat, type)						\
+        do {									\
+                struct ios_conf         *conf = NULL;				\
+                double                   elapsed;				\
+                struct timeval          *begin, *end;				\
+                double                   throughput;				\
+               int                      flag = 0;                              \
+                                                                                \
+                begin = &frame->begin;						\
+                end   = &frame->end;						\
+                                                                                \
+                elapsed = (end->tv_sec - begin->tv_sec) * 1e6			\
+                        + (end->tv_usec - begin->tv_usec);			\
+                throughput = op_ret / elapsed;					\
+                                                                                \
+                conf = this->private;						\
+                LOCK(&iosstat->lock);						\
+                {								\
+                        if (iosstat->thru_counters[type].throughput             \
+                                <= throughput) {                                \
+                                iosstat->thru_counters[type].throughput =       \
+                                                                throughput;     \
+                                gettimeofday (&iosstat->                        \
+                                             thru_counters[type].time, NULL);   \
+                               flag = 1;                                       \
+                        }							\
+                }								\
+                UNLOCK (&iosstat->lock);					\
+               if (flag)                                                       \
+                       ios_stat_add_to_list (&conf->thru_list[type],           \
+                                               throughput, iosstat);           \
+        } while (0)
+
 int
 ios_fd_ctx_get (fd_t *fd, xlator_t *this, struct ios_fd **iosfd)
 {
@@ -174,6 +350,183 @@ ios_fd_ctx_set (fd_t *fd, xlator_t *this, struct ios_fd *iosfd)
         return ret;
 }
 
+int
+ios_stat_ref (struct ios_stat *iosstat)
+{
+        LOCK (&iosstat->lock);
+        {
+                iosstat->refcnt++;
+        }
+        UNLOCK (&iosstat->lock);
+
+        return iosstat->refcnt;
+}
+
+int
+ios_stat_unref (struct ios_stat *iosstat)
+{
+        int cleanup = 0;
+        LOCK (&iosstat->lock);
+        {
+                iosstat->refcnt--;
+                if (iosstat->refcnt == 0) {
+                        if (iosstat->filename) {
+                                GF_FREE (iosstat->filename);
+                                iosstat->filename = NULL;
+                        }
+                        cleanup = 1;
+                }
+        }
+        UNLOCK (&iosstat->lock);
+
+        if (cleanup) {
+                GF_FREE (iosstat);
+                iosstat = NULL;
+        }
+
+        return 0;
+}
+
+int
+ios_inode_ctx_set (inode_t *inode, xlator_t *this, struct ios_stat *iosstat)
+{
+        uint64_t   iosstat64 = 0;
+        int        ret     = 0;
+
+        ios_stat_ref (iosstat);
+        iosstat64 = (unsigned long )iosstat;
+        ret = inode_ctx_put (inode, this, iosstat64);
+        return ret;
+}
+
+int
+ios_inode_ctx_get (inode_t *inode, xlator_t *this, struct ios_stat **iosstat)
+{
+        uint64_t      iosstat64 = 0;
+        unsigned long iosstatlong = 0;
+        int           ret = 0;
+
+        ret = inode_ctx_get (inode, this, &iosstat64);
+        iosstatlong = iosstat64;
+        if (ret != -1)
+                *iosstat = (void *) iosstatlong;
+
+        return ret;
+
+}
+
+int
+ios_stat_add_to_list (struct ios_stat_head *list_head, uint64_t value,
+                            struct ios_stat *iosstat)
+{
+        struct ios_stat_list *new = NULL;
+        struct ios_stat_list *entry = NULL;
+        struct ios_stat_list *t = NULL;
+        struct ios_stat_list *list_entry = NULL;
+        struct ios_stat_list *tmp = NULL;
+        struct ios_stat_list *last = NULL;
+        struct ios_stat      *stat = NULL;
+        int                   cnt = 0;
+        int                   found = 0;
+        int                   reposition = 0;
+        double                min_count = 0;
+
+        LOCK (&list_head->lock);
+        {
+
+                if (list_head->min_cnt == 0)
+                        list_head->min_cnt = value;
+                if ((list_head->members == MAX_LIST_MEMBERS) &&
+                     (list_head->min_cnt > value))
+                        goto out;
+
+                list_for_each_entry_safe (entry, t,
+                              &list_head->iosstats->list, list) {
+                        cnt++;
+                        if (cnt == list_head->members)
+                                last = entry;
+
+                        if (!uuid_compare (iosstat->gfid,
+                            entry->iosstat->gfid)) {
+                                list_entry = entry;
+                                found = cnt;
+                                entry->value = value;
+                                if (!reposition) {
+                                        if (cnt == list_head->members)
+                                                list_head->min_cnt = value;
+                                        goto out;
+                                }
+                                break;
+                        } else if (entry->value <= value && !reposition) {
+                                reposition = cnt;
+                                tmp = entry;
+                                if (cnt == list_head->members - 1)
+                                        min_count = entry->value;
+                        }
+                }
+                if (found) {
+                        list_del (&list_entry->list);
+                        list_add_tail (&list_entry->list, &tmp->list);
+                        if (min_count)
+                                list_head->min_cnt = min_count;
+                        goto out;
+                } else if (list_head->members == MAX_LIST_MEMBERS && reposition) {
+                        new = GF_CALLOC (1, sizeof (*new),
+                                        gf_io_stats_mt_ios_stat_list);
+                        new->iosstat = iosstat;
+                        new->value = value;
+                        ios_stat_ref (iosstat);
+                        list_add_tail (&new->list, &tmp->list);
+                        stat = last->iosstat;
+                        last->iosstat = NULL;
+                        ios_stat_unref (stat);
+                        list_del (&last->list);
+                        GF_FREE (last);
+                        if (reposition == MAX_LIST_MEMBERS)
+                                list_head->min_cnt = value;
+                        else if (min_count) {
+                                list_head->min_cnt = min_count;
+                        }
+                } else if (list_head->members < MAX_LIST_MEMBERS) {
+                        new  = GF_CALLOC (1, sizeof (*new),
+                                          gf_io_stats_mt_ios_stat_list);
+                        new->iosstat = iosstat;
+                        new->value = value;
+                        ios_stat_ref (iosstat);
+                        if (reposition) {
+                                list_add_tail (&new->list, &tmp->list);
+                        } else {
+                                list_add_tail (&new->list, &entry->list);
+                        }
+                        list_head->members++;
+                        if (list_head->min_cnt > value)
+                                list_head->min_cnt = value;
+                }
+        }
+out:
+        UNLOCK (&list_head->lock);
+        return 0;
+}
+
+inline int
+ios_stats_cleanup (xlator_t *this, inode_t *inode)
+{
+
+        struct ios_stat *iosstat = NULL;
+        uint64_t         iosstat64 = 0;
+
+        inode_ctx_del (inode, this, &iosstat64);
+        if (!iosstat64) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "could not get inode ctx");
+                return 0;
+        }
+        iosstat = (void *) (long)iosstat64;
+        if (iosstat) {
+                ios_stat_unref (iosstat);
+        }
+        return 0;
+}
 
 #define ios_log(this, logfp, fmt ...)                           \
         do {                                                    \
@@ -181,60 +534,401 @@ ios_fd_ctx_set (fd_t *fd, xlator_t *this, struct ios_fd *iosfd)
                         fprintf (logfp, fmt);                   \
                         fprintf (logfp, "\n");                  \
                 }                                               \
-                gf_log (this->name, GF_LOG_NORMAL, fmt);        \
+                gf_log (this->name, GF_LOG_INFO, fmt);        \
         } while (0)
 
+int
+ios_dump_file_stats (struct ios_stat_head *list_head, xlator_t *this, FILE* logfp)
+{
+        struct ios_stat_list *entry = NULL;
+
+        LOCK (&list_head->lock);
+        {
+                list_for_each_entry (entry, &list_head->iosstats->list, list) {
+                        ios_log (this, logfp, "%-12.0f %s",
+                                entry->value, entry->iosstat->filename);
+                }
+        }
+        UNLOCK (&list_head->lock);
+        return 0;
+}
 
 int
-io_stats_dump_global (xlator_t *this, struct ios_global_stats *stats,
-                      struct timeval *now, int interval, FILE *logfp)
+ios_dump_throughput_stats (struct ios_stat_head *list_head, xlator_t *this,
+                            FILE* logfp, ios_stats_type_t type)
 {
-        int    i = 0;
+        struct ios_stat_list *entry = NULL;
+        struct timeval        time  = {0, };
+        char                  timestr[256] = {0, };
+
+        LOCK (&list_head->lock);
+        {
+                list_for_each_entry (entry, &list_head->iosstats->list, list) {
+                        gf_time_fmt (timestr, sizeof timestr,
+                                     entry->iosstat->thru_counters[type].time.tv_sec,
+                                     gf_timefmt_FT);
+                        snprintf (timestr + strlen (timestr), sizeof timestr - strlen (timestr),
+                          ".%"GF_PRI_SUSECONDS, time.tv_usec);
+
+                        ios_log (this, logfp, "%s \t %-10.2f  \t  %s",
+                                 timestr, entry->value, entry->iosstat->filename);
+                }
+        }
+        UNLOCK (&list_head->lock);
+        return 0;
+}
+
+int
+io_stats_dump_global_to_logfp (xlator_t *this, struct ios_global_stats *stats,
+                               struct timeval *now, int interval, FILE* logfp)
+{
+        int                   i = 0;
+        int                   per_line = 0;
+        int                   index = 0;
+        struct ios_stat_head *list_head = NULL;
+        struct ios_conf      *conf = NULL;
+        char                  timestr[256] = {0, };
+        char                  str_header[128] = {0};
+        char                  str_read[128] = {0};
+        char                  str_write[128] = {0};
+
+        conf = this->private;
 
         if (interval == -1)
-                ios_log (this, logfp, "=== Cumulative stats ===");
+                ios_log (this, logfp, "\n=== Cumulative stats ===");
         else
-                ios_log (this, logfp, "=== Interval %d stats ===",
+                ios_log (this, logfp, "\n=== Interval %d stats ===",
                          interval);
-        ios_log (this, logfp, "      Duration : %"PRId64"secs",
+        ios_log (this, logfp, "      Duration : %"PRId64" secs",
                  (uint64_t) (now->tv_sec - stats->started_at.tv_sec));
         ios_log (this, logfp, "     BytesRead : %"PRId64,
                  stats->data_read);
-        ios_log (this, logfp, "  BytesWritten : %"PRId64,
+        ios_log (this, logfp, "  BytesWritten : %"PRId64"\n",
                  stats->data_written);
 
+        snprintf (str_header, sizeof (str_header), "%-12s %c", "Block Size", ':');
+        snprintf (str_read, sizeof (str_read), "%-12s %c", "Read Count", ':');
+        snprintf (str_write, sizeof (str_write), "%-12s %c", "Write Count", ':');
+        index = 14;
         for (i = 0; i < 32; i++) {
+                if ((stats->block_count_read[i] == 0) &&
+                    (stats->block_count_write[i] == 0))
+                        continue;
+                per_line++;
+
+                snprintf (str_header+index, sizeof (str_header)-index,
+                          "%16dB+", (1<<i));
                 if (stats->block_count_read[i])
-                        ios_log (this, logfp, " Read %06db+ : %"PRId64,
-                                 (1 << i), stats->block_count_read[i]);
+                        snprintf (str_read+index, sizeof (str_read)-index,
+                                  "%18"PRId64, stats->block_count_read[i]);
+                else    snprintf (str_read+index, sizeof (str_read)-index,
+                                  "%18s", "0");
+                if (stats->block_count_write[i])
+                        snprintf (str_write+index, sizeof (str_write)-index,
+                                  "%18"PRId64, stats->block_count_write[i]);
+                else    snprintf (str_write+index, sizeof (str_write)-index,
+                                  "%18s", "0");
+
+                index += 18;
+                if (per_line == 3) {
+                        ios_log (this, logfp, "%s", str_header);
+                        ios_log (this, logfp, "%s", str_read);
+                        ios_log (this, logfp, "%s\n", str_write);
+
+                        memset (str_header, 0, sizeof (str_header));
+                        memset (str_read, 0, sizeof (str_read));
+                        memset (str_write, 0, sizeof (str_write));
+
+                        snprintf (str_header, sizeof (str_header), "%-12s %c",
+                                  "Block Size", ':');
+                        snprintf (str_read, sizeof (str_read), "%-12s %c",
+                                  "Read Count", ':');
+                        snprintf (str_write, sizeof (str_write), "%-12s %c",
+                                  "Write Count", ':');
+
+                        index = 14;
+                        per_line = 0;
+                }
         }
 
+        if (per_line != 0) {
+                ios_log (this, logfp, "%s", str_header);
+                ios_log (this, logfp, "%s", str_read);
+                ios_log (this, logfp, "%s\n", str_write);
+        }
+
+        ios_log (this, logfp, "%-13s %10s %14s %14s %14s", "Fop",
+                 "Call Count", "Avg-Latency", "Min-Latency",
+                 "Max-Latency");
+        ios_log (this, logfp, "%-13s %10s %14s %14s %14s", "---", "----------",
+                 "-----------", "-----------", "-----------");
+
+        for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+                if (stats->fop_hits[i] && !stats->latency[i].avg)
+                        ios_log (this, logfp, "%-13s %10"PRId64" %11s "
+                                 "us %11s us %11s us", gf_fop_list[i],
+                                 stats->fop_hits[i], "0", "0", "0");
+                else if (stats->fop_hits[i] && stats->latency[i].avg)
+                        ios_log (this, logfp, "%-13s %10"PRId64" %11.2lf us "
+                                 "%11.2lf us %11.2lf us", gf_fop_list[i],
+                                 stats->fop_hits[i], stats->latency[i].avg,
+                                 stats->latency[i].min, stats->latency[i].max);
+        }
+        ios_log (this, logfp, "------ ----- ----- ----- ----- ----- ----- ----- "
+                 " ----- ----- ----- -----\n");
+
+        if (interval == -1) {
+                LOCK (&conf->lock);
+                {
+                        gf_time_fmt (timestr, sizeof timestr,
+                                     conf->cumulative.max_openfd_time.tv_sec,
+                                     gf_timefmt_FT);
+                        snprintf (timestr + strlen (timestr), sizeof timestr - strlen (timestr),
+                                  ".%"GF_PRI_SUSECONDS,
+                                  conf->cumulative.max_openfd_time.tv_usec);
+                        ios_log (this, logfp, "Current open fd's: %"PRId64
+                                 " Max open fd's: %"PRId64" time %s",
+                                 conf->cumulative.nr_opens,
+                                 conf->cumulative.max_nr_opens, timestr);
+                }
+                UNLOCK (&conf->lock);
+                ios_log (this, logfp, "\n==========Open File Stats========");
+                ios_log (this, logfp, "\nCOUNT:  \t  FILE NAME");
+                list_head = &conf->list[IOS_STATS_TYPE_OPEN];
+                ios_dump_file_stats (list_head, this, logfp);
+
+
+                ios_log (this, logfp, "\n==========Read File Stats========");
+                ios_log (this, logfp, "\nCOUNT:  \t  FILE NAME");
+                list_head = &conf->list[IOS_STATS_TYPE_READ];
+                ios_dump_file_stats (list_head, this, logfp);
+
+                ios_log (this, logfp, "\n==========Write File Stats========");
+                ios_log (this, logfp, "\nCOUNT:  \t  FILE NAME");
+                list_head = &conf->list[IOS_STATS_TYPE_WRITE];
+                ios_dump_file_stats (list_head, this, logfp);
+
+                ios_log (this, logfp, "\n==========Directory open stats========");
+                ios_log (this, logfp, "\nCOUNT:  \t  DIRECTORY NAME");
+                list_head = &conf->list[IOS_STATS_TYPE_OPENDIR];
+                ios_dump_file_stats (list_head, this, logfp);
+
+                ios_log (this, logfp, "\n========Directory readdirp Stats=======");
+                ios_log (this, logfp, "\nCOUNT:  \t  DIRECTORY NAME");
+                list_head = &conf->list[IOS_STATS_TYPE_READDIRP];
+                ios_dump_file_stats (list_head, this, logfp);
+
+                ios_log (this, logfp, "\n========Read Throughput File Stats=====");
+                ios_log (this, logfp, "\nTIMESTAMP \t\t\t THROUGHPUT(KBPS)"
+                         "\tFILE NAME");
+                list_head = &conf->thru_list[IOS_STATS_THRU_READ];
+                ios_dump_throughput_stats(list_head, this, logfp, IOS_STATS_THRU_READ);
+
+                ios_log (this, logfp, "\n======Write Throughput File Stats======");
+                ios_log (this, logfp, "\nTIMESTAMP \t\t\t THROUGHPUT(KBPS)"
+                         "\tFILE NAME");
+                list_head = &conf->thru_list[IOS_STATS_THRU_WRITE];
+                ios_dump_throughput_stats (list_head, this, logfp, IOS_STATS_THRU_WRITE);
+        }
+        return 0;
+}
+
+int
+io_stats_dump_global_to_dict (xlator_t *this, struct ios_global_stats *stats,
+                              struct timeval *now, int interval, dict_t *dict)
+{
+        int             ret = 0;
+        char            key[256] = {0};
+        uint64_t        sec = 0;
+        int             i = 0;
+        uint64_t        count = 0;
+
+        GF_ASSERT (stats);
+        GF_ASSERT (now);
+        GF_ASSERT (dict);
+        GF_ASSERT (this);
+
+        if (interval == -1)
+                snprintf (key, sizeof (key), "cumulative");
+        else
+                snprintf (key, sizeof (key), "interval");
+        ret = dict_set_int32 (dict, key, interval);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "failed to set "
+                        "interval %d", interval);
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%d-duration", interval);
+        sec = (uint64_t) (now->tv_sec - stats->started_at.tv_sec);
+        ret = dict_set_uint64 (dict, key, sec);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to set "
+                        "duration(%d) - %"PRId64, interval, sec);
+                goto out;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%d-total-read", interval);
+        ret = dict_set_uint64 (dict, key, stats->data_read);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to set total "
+                       "read(%d) - %"PRId64, interval, stats->data_read);
+                goto out;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%d-total-write", interval);
+        ret = dict_set_uint64 (dict, key, stats->data_written);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to set total "
+                        "write(%d) - %"PRId64, interval, stats->data_written);
+                goto out;
+        }
         for (i = 0; i < 32; i++) {
-                if (stats->block_count_write[i])
-                        ios_log (this, logfp, "Write %06db+ : %"PRId64,
-                                 (1 << i), stats->block_count_write[i]);
+                if (stats->block_count_read[i]) {
+                        memset (key, 0, sizeof (key));
+                        snprintf (key, sizeof (key), "%d-read-%d", interval,
+                                  (1 << i));
+                        count = stats->block_count_read[i];
+                        ret = dict_set_uint64 (dict, key, count);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                        "set read-%db+, with: %"PRId64,
+                                        (1<<i), count);
+                                goto out;
+                        }
+                }
         }
 
-        for (i = 0; i < GF_FOP_MAXVALUE; i++)
-                if (stats->fop_hits[i])
-                        ios_log (this, logfp, "%14s : %"PRId64,
-                                 gf_fop_list[i], stats->fop_hits[i]);
+        for (i = 0; i < 32; i++) {
+                if (stats->block_count_write[i]) {
+                        snprintf (key, sizeof (key), "%d-write-%d", interval,
+                                  (1<<i));
+                        count = stats->block_count_write[i];
+                        ret = dict_set_uint64 (dict, key, count);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                        "set write-%db+, with: %"PRId64,
+                                        (1<<i), count);
+                                goto out;
+                        }
+                }
+        }
 
+        for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+                if (stats->fop_hits[i] == 0)
+                        continue;
+                snprintf (key, sizeof (key), "%d-%d-hits", interval, i);
+                ret = dict_set_uint64 (dict, key, stats->fop_hits[i]);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                "set %s-fop-hits: %"PRIu64, gf_fop_list[i],
+                                stats->fop_hits[i]);
+                        goto out;
+                }
 
-        return 0;
+                if (stats->latency[i].avg == 0)
+                        continue;
+                snprintf (key, sizeof (key), "%d-%d-avglatency", interval, i);
+                ret = dict_set_double (dict, key, stats->latency[i].avg);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to set %s "
+                                "avglatency(%d) with %f", gf_fop_list[i],
+                                interval, stats->latency[i].avg);
+                        goto out;
+                }
+                snprintf (key, sizeof (key), "%d-%d-minlatency", interval, i);
+                ret = dict_set_double (dict, key, stats->latency[i].min);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to set %s "
+                                "minlatency(%d) with %f", gf_fop_list[i],
+                                interval, stats->latency[i].min);
+                        goto out;
+                }
+                snprintf (key, sizeof (key), "%d-%d-maxlatency", interval, i);
+                ret = dict_set_double (dict, key, stats->latency[i].max);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to set %s "
+                                "maxlatency(%d) with %f", gf_fop_list[i],
+                                interval, stats->latency[i].max);
+                        goto out;
+                }
+        }
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+int
+io_stats_dump_global (xlator_t *this, struct ios_global_stats *stats,
+                      struct timeval *now, int interval,
+                      struct ios_dump_args *args)
+{
+        int     ret = -1;
+
+        GF_ASSERT (args);
+        GF_ASSERT (now);
+        GF_ASSERT (stats);
+        GF_ASSERT (this);
+
+
+
+        switch (args->type) {
+        case IOS_DUMP_TYPE_FILE:
+                ret = io_stats_dump_global_to_logfp (this, stats, now,
+                                                     interval, args->u.logfp);
+        break;
+        case IOS_DUMP_TYPE_DICT:
+                ret = io_stats_dump_global_to_dict (this, stats, now,
+                                                    interval, args->u.dict);
+        break;
+        default:
+                GF_ASSERT (0);
+                ret = -1;
+        break;
+        }
+        return ret;
 }
 
+int
+ios_dump_args_init (struct ios_dump_args *args, ios_dump_type_t type,
+                    void *output)
+{
+        int             ret = 0;
+
+        GF_ASSERT (args);
+        GF_ASSERT (type > IOS_DUMP_TYPE_NONE && type < IOS_DUMP_TYPE_MAX);
+        GF_ASSERT (output);
+
+        args->type = type;
+        switch (args->type) {
+        case IOS_DUMP_TYPE_FILE:
+                args->u.logfp = output;
+                break;
+        case IOS_DUMP_TYPE_DICT:
+                args->u.dict = output;
+                break;
+        default:
+                GF_ASSERT (0);
+                ret = -1;
+        }
+
+        return ret;
+}
 
 int
-io_stats_dump (xlator_t *this, char *filename, inode_t *inode,
-               const char *path)
+io_stats_dump (xlator_t *this, struct ios_dump_args *args)
 {
         struct ios_conf         *conf = NULL;
         struct ios_global_stats  cumulative = {0, };
         struct ios_global_stats  incremental = {0, };
         int                      increment = 0;
         struct timeval           now;
-        FILE                    *logfp = NULL;
+
+        GF_ASSERT (this);
+        GF_ASSERT (args);
+        GF_ASSERT (args->type > IOS_DUMP_TYPE_NONE);
+        GF_ASSERT (args->type < IOS_DUMP_TYPE_MAX);
 
         conf = this->private;
 
@@ -251,12 +945,9 @@ io_stats_dump (xlator_t *this, char *filename, inode_t *inode,
         }
         UNLOCK (&conf->lock);
 
-        logfp = fopen (filename, "w+");
-        io_stats_dump_global (this, &cumulative, &now, -1, logfp);
-        io_stats_dump_global (this, &incremental, &now, increment, logfp);
+        io_stats_dump_global (this, &cumulative, &now, -1, args);
+        io_stats_dump_global (this, &incremental, &now, increment, args);
 
-        if (logfp)
-                fclose (logfp);
         return 0;
 }
 
@@ -288,53 +979,214 @@ io_stats_dump_fd (xlator_t *this, struct ios_fd *iosfd)
         sec = now.tv_sec - iosfd->opened_at.tv_sec;
         usec = now.tv_usec - iosfd->opened_at.tv_usec;
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "--- fd stats ---");
 
         if (iosfd->filename)
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "      Filename : %s",
                         iosfd->filename);
 
         if (sec)
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "      Lifetime : %"PRId64"secs, %"PRId64"usecs",
                         sec, usec);
 
         if (iosfd->data_read)
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "     BytesRead : %"PRId64" bytes",
                         iosfd->data_read);
 
         if (iosfd->data_written)
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "  BytesWritten : %"PRId64" bytes",
                         iosfd->data_written);
 
         for (i = 0; i < 32; i++) {
                 if (iosfd->block_count_read[i])
-                        gf_log (this->name, GF_LOG_NORMAL,
+                        gf_log (this->name, GF_LOG_INFO,
                                 " Read %06db+ : %"PRId64,
                                 (1 << i), iosfd->block_count_read[i]);
         }
         for (i = 0; i < 32; i++) {
                 if (iosfd->block_count_write[i])
-                        gf_log (this->name, GF_LOG_NORMAL,
+                        gf_log (this->name, GF_LOG_INFO,
                                 "Write %06db+ : %"PRId64,
                                 (1 << i), iosfd->block_count_write[i]);
         }
         return 0;
 }
 
+static void
+update_ios_latency_stats (struct ios_global_stats   *stats, double elapsed,
+                          glusterfs_fop_t op)
+{
+        double avg;
+
+        GF_ASSERT (stats);
+
+        if (!stats->latency[op].min)
+                stats->latency[op].min = elapsed;
+        if (stats->latency[op].min > elapsed)
+                stats->latency[op].min = elapsed;
+        if (stats->latency[op].max < elapsed)
+                stats->latency[op].max = elapsed;
+
+        avg = stats->latency[op].avg;
+
+        stats->latency[op].avg = avg + (elapsed - avg) / stats->fop_hits[op];
+}
+
+int
+update_ios_latency (struct ios_conf *conf, call_frame_t *frame,
+                    glusterfs_fop_t op)
+{
+        double elapsed;
+        struct timeval *begin, *end;
+
+        begin = &frame->begin;
+        end   = &frame->end;
+
+        elapsed = (end->tv_sec - begin->tv_sec) * 1e6
+                + (end->tv_usec - begin->tv_usec);
+
+        update_ios_latency_stats (&conf->cumulative, elapsed, op);
+        update_ios_latency_stats (&conf->incremental, elapsed, op);
+
+        return 0;
+}
+
+int32_t
+io_stats_dump_stats_to_dict (xlator_t *this, dict_t *resp,
+                             ios_stats_type_t flags, int32_t list_cnt)
+{
+        struct ios_conf         *conf = NULL;
+        int                      cnt  = 0;
+        char                     key[256];
+        struct ios_stat_head    *list_head = NULL;
+        struct ios_stat_list    *entry = NULL;
+        int                      ret = -1;
+        ios_stats_thru_t         index = IOS_STATS_THRU_MAX;
+        char                     timestr[256] = {0, };
+        char                    *dict_timestr = NULL;
+
+        conf = this->private;
+
+        switch (flags) {
+                case IOS_STATS_TYPE_OPEN:
+                        list_head = &conf->list[IOS_STATS_TYPE_OPEN];
+                        LOCK (&conf->lock);
+                        {
+                                ret = dict_set_uint64 (resp, "current-open",
+                                                     conf->cumulative.nr_opens);
+                                if (ret)
+                                        goto unlock;
+                                ret = dict_set_uint64 (resp, "max-open",
+                                                       conf->cumulative.max_nr_opens);
+
+                                gf_time_fmt (timestr, sizeof timestr,
+                                             conf->cumulative.max_openfd_time.tv_sec,
+                                             gf_timefmt_FT);
+                                if (conf->cumulative.max_openfd_time.tv_sec)
+                                        snprintf (timestr + strlen (timestr), sizeof timestr - strlen (timestr),
+                                                  ".%"GF_PRI_SUSECONDS,
+                                                  conf->cumulative.max_openfd_time.tv_usec);
+
+                                dict_timestr = gf_strdup (timestr);
+                                if (!dict_timestr)
+                                        goto unlock;
+                                ret = dict_set_dynstr (resp, "max-openfd-time",
+                                                       dict_timestr);
+                                if (ret)
+                                        goto unlock;
+                        }
+        unlock:
+                        UNLOCK (&conf->lock);
+                        /* Do not proceed if we came here because of some error
+                         * during the dict operation */
+                        if (ret)
+                                goto out;
+                        break;
+                case IOS_STATS_TYPE_READ:
+                        list_head = &conf->list[IOS_STATS_TYPE_READ];
+                        break;
+                case IOS_STATS_TYPE_WRITE:
+                        list_head = &conf->list[IOS_STATS_TYPE_WRITE];
+                        break;
+                case IOS_STATS_TYPE_OPENDIR:
+                        list_head = &conf->list[IOS_STATS_TYPE_OPENDIR];
+                        break;
+                case IOS_STATS_TYPE_READDIRP:
+                        list_head = &conf->list[IOS_STATS_TYPE_READDIRP];
+                        break;
+                case IOS_STATS_TYPE_READ_THROUGHPUT:
+                        list_head = &conf->thru_list[IOS_STATS_THRU_READ];
+                        index = IOS_STATS_THRU_READ;
+                        break;
+                case IOS_STATS_TYPE_WRITE_THROUGHPUT:
+                        list_head = &conf->thru_list[IOS_STATS_THRU_WRITE];
+                        index = IOS_STATS_THRU_WRITE;
+                        break;
+
+                default:
+                       goto out;
+        }
+        ret = dict_set_int32 (resp, "top-op", flags);
+        if (!list_cnt)
+                goto out;
+        LOCK (&list_head->lock);
+        {
+                list_for_each_entry (entry, &list_head->iosstats->list, list) {
+
+                        cnt++;
+                        snprintf (key, 256, "%s-%d", "filename", cnt);
+                        ret = dict_set_str (resp, key, entry->iosstat->filename);
+                        if (ret)
+                                goto unlock_list_head;
+                         snprintf (key, 256, "%s-%d", "value",cnt);
+                         ret = dict_set_uint64 (resp, key, entry->value);
+                         if (ret)
+                                 goto unlock_list_head;
+                         if (index != IOS_STATS_THRU_MAX) {
+                                 snprintf (key, 256, "%s-%d", "time-sec", cnt);
+                                 ret = dict_set_int32 (resp, key,
+                                         entry->iosstat->thru_counters[index].time.tv_sec);
+                                 if (ret)
+                                         goto unlock_list_head;
+                                 snprintf (key, 256, "%s-%d", "time-usec", cnt);
+                                 ret = dict_set_int32 (resp, key,
+                                         entry->iosstat->thru_counters[index].time.tv_usec);
+                                 if (ret)
+                                         goto unlock_list_head;
+                         }
+                         if (cnt == list_cnt)
+                                 break;
+
+                }
+        }
+unlock_list_head:
+        UNLOCK (&list_head->lock);
+        /* ret is !=0 if some dict operation in the above critical region
+         * failed. */
+        if (ret)
+                goto out;
+        ret = dict_set_int32 (resp, "members", cnt);
+ out:
+        return ret;
+}
 
 int
 io_stats_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno, fd_t *fd,
                      inode_t *inode, struct iatt *buf,
-                     struct iatt *preparent, struct iatt *postparent)
+                     struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
         struct ios_fd *iosfd = NULL;
         char          *path = NULL;
+        struct ios_stat *iosstat = NULL;
+        struct ios_conf   *conf = NULL;
+
+        conf = this->private;
 
         path = frame->local;
         frame->local = NULL;
@@ -357,21 +1209,44 @@ io_stats_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         gettimeofday (&iosfd->opened_at, NULL);
 
         ios_fd_ctx_set (fd, this, iosfd);
+        LOCK (&conf->lock);
+        {
+                conf->cumulative.nr_opens++;
+                if (conf->cumulative.nr_opens > conf->cumulative.max_nr_opens) {
+                        conf->cumulative.max_nr_opens = conf->cumulative.nr_opens;
+                        conf->cumulative.max_openfd_time = iosfd->opened_at;
+                }
+        }
+        UNLOCK (&conf->lock);
+
+        iosstat = GF_CALLOC (1, sizeof (*iosstat), gf_io_stats_mt_ios_stat);
+        if (!iosstat) {
+                GF_FREE (path);
+                goto unwind;
+        }
+        iosstat->filename = gf_strdup (path);
+        uuid_copy (iosstat->gfid, buf->ia_gfid);
+        LOCK_INIT (&iosstat->lock);
+        ios_inode_ctx_set (fd->inode, this, iosstat);
 
 unwind:
+        UPDATE_PROFILE_STATS (frame, CREATE);
         STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
-                             preparent, postparent);
+                             preparent, postparent, xdata);
         return 0;
 }
 
 
 int
 io_stats_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, fd_t *fd)
+                   int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
         struct ios_fd *iosfd = NULL;
         char          *path = NULL;
+        struct   ios_stat *iosstat = NULL;
+        struct ios_conf   *conf = NULL;
 
+        conf = this->private;
         path = frame->local;
         frame->local = NULL;
 
@@ -394,17 +1269,46 @@ io_stats_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         ios_fd_ctx_set (fd, this, iosfd);
 
+        ios_inode_ctx_get (fd->inode, this, &iosstat);
+        if (!iosstat) {
+                iosstat = GF_CALLOC (1, sizeof (*iosstat),
+                                     gf_io_stats_mt_ios_stat);
+                if (iosstat) {
+                        iosstat->filename = gf_strdup (path);
+                        uuid_copy (iosstat->gfid, fd->inode->gfid);
+                        LOCK_INIT (&iosstat->lock);
+                        ios_inode_ctx_set (fd->inode, this, iosstat);
+                }
+        }
+
+        LOCK (&conf->lock);
+        {
+                conf->cumulative.nr_opens++;
+                if (conf->cumulative.nr_opens > conf->cumulative.max_nr_opens) {
+                        conf->cumulative.max_nr_opens = conf->cumulative.nr_opens;
+                        conf->cumulative.max_openfd_time = iosfd->opened_at;
+                }
+        }
+        UNLOCK (&conf->lock);
+        if (iosstat) {
+              BUMP_STATS (iosstat, IOS_STATS_TYPE_OPEN);
+              iosstat = NULL;
+        }
 unwind:
-        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd);
+        UPDATE_PROFILE_STATS (frame, OPEN);
+
+        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
         return 0;
+
 }
 
 
 int
 io_stats_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                   int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf);
+        UPDATE_PROFILE_STATS (frame, STAT);
+        STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
@@ -413,11 +1317,11 @@ int
 io_stats_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno,
                     struct iovec *vector, int32_t count,
-                    struct iatt *buf, struct iobref *iobref)
+                    struct iatt *buf, struct iobref *iobref, dict_t *xdata)
 {
         int              len = 0;
         fd_t            *fd = NULL;
-
+        struct ios_stat *iosstat = NULL;
 
         fd = frame->local;
         frame->local = NULL;
@@ -427,19 +1331,46 @@ io_stats_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 BUMP_READ (fd, len);
         }
 
+        UPDATE_PROFILE_STATS (frame, READ);
+        ios_inode_ctx_get (fd->inode, this, &iosstat);
+
+        if (iosstat) {
+              BUMP_STATS (iosstat, IOS_STATS_TYPE_READ);
+              BUMP_THROUGHPUT (iosstat, IOS_STATS_THRU_READ);
+              iosstat = NULL;
+        }
+
         STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno,
-                             vector, count, buf, iobref);
+                             vector, count, buf, iobref, xdata);
         return 0;
+
 }
 
 
 int
 io_stats_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno,
-                     struct iatt *prebuf, struct iatt *postbuf)
-{
-        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf);
+                     struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+        struct ios_stat *iosstat = NULL;
+        inode_t         *inode   = NULL;
+
+        UPDATE_PROFILE_STATS (frame, WRITE);
+        if (frame->local){
+                inode = frame->local;
+                frame->local = NULL;
+                ios_inode_ctx_get (inode, this, &iosstat);
+                if (iosstat) {
+                        BUMP_STATS (iosstat, IOS_STATS_TYPE_WRITE);
+                        BUMP_THROUGHPUT (iosstat, IOS_STATS_THRU_WRITE);
+                        inode = NULL;
+                        iosstat = NULL;
+                }
+        }
+
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
         return 0;
+
 }
 
 
@@ -447,18 +1378,33 @@ io_stats_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 int
 io_stats_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno, gf_dirent_t *buf)
+                       int32_t op_ret, int32_t op_errno, gf_dirent_t *buf, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, buf);
+        struct ios_stat *iosstat = NULL;
+        inode_t         *inode   = frame->local;
+
+        frame->local = NULL;
+
+        UPDATE_PROFILE_STATS (frame, READDIRP);
+
+        ios_inode_ctx_get (inode, this, &iosstat);
+
+        if (iosstat) {
+              BUMP_STATS (iosstat, IOS_STATS_TYPE_READDIRP);
+              iosstat = NULL;
+        }
+
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
 
 int
 io_stats_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno, gf_dirent_t *buf)
+                      int32_t op_ret, int32_t op_errno, gf_dirent_t *buf, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, buf);
+        UPDATE_PROFILE_STATS (frame, READDIR);
+        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
@@ -466,9 +1412,10 @@ io_stats_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 int
 io_stats_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno,
-                    struct iatt *prebuf, struct iatt *postbuf)
+                    struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf);
+        UPDATE_PROFILE_STATS (frame, FSYNC);
+        STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
         return 0;
 }
 
@@ -476,9 +1423,10 @@ io_stats_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 int
 io_stats_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                       int32_t op_ret, int32_t op_errno,
-                      struct iatt *preop, struct iatt *postop)
+                      struct iatt *preop, struct iatt *postop, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, preop, postop);
+        UPDATE_PROFILE_STATS (frame, SETATTR);
+        STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, preop, postop, xdata);
         return 0;
 }
 
@@ -486,11 +1434,13 @@ io_stats_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 int
 io_stats_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno,
-                     struct iatt *preparent, struct iatt *postparent)
+                     struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
+        UPDATE_PROFILE_STATS (frame, UNLINK);
         STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
-                             preparent, postparent);
+                             preparent, postparent, xdata);
         return 0;
+
 }
 
 
@@ -498,11 +1448,12 @@ int
 io_stats_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno, struct iatt *buf,
                      struct iatt *preoldparent, struct iatt *postoldparent,
-                     struct iatt *prenewparent, struct iatt *postnewparent)
+                     struct iatt *prenewparent, struct iatt *postnewparent, dict_t *xdata)
 {
+        UPDATE_PROFILE_STATS (frame, RENAME);
         STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf,
                              preoldparent, postoldparent,
-                             prenewparent, postnewparent);
+                             prenewparent, postnewparent, xdata);
         return 0;
 }
 
@@ -510,9 +1461,10 @@ io_stats_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 int
 io_stats_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                        int32_t op_ret, int32_t op_errno, const char *buf,
-                       struct iatt *sbuf)
+                       struct iatt *sbuf, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, buf, sbuf);
+        UPDATE_PROFILE_STATS (frame, READLINK);
+        STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, buf, sbuf, xdata);
         return 0;
 }
 
@@ -521,9 +1473,10 @@ int
 io_stats_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno,
                      inode_t *inode, struct iatt *buf,
-                     dict_t *xattr, struct iatt *postparent)
+                     dict_t *xdata, struct iatt *postparent)
 {
-        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf, xattr,
+        UPDATE_PROFILE_STATS (frame, LOOKUP);
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf, xdata,
                              postparent);
         return 0;
 }
@@ -533,10 +1486,11 @@ int
 io_stats_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                       int32_t op_ret, int32_t op_errno,
                       inode_t *inode, struct iatt *buf,
-                      struct iatt *preparent, struct iatt *postparent)
+                      struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
+        UPDATE_PROFILE_STATS (frame, SYMLINK);
         STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
+                             preparent, postparent, xdata);
         return 0;
 }
 
@@ -545,10 +1499,11 @@ int
 io_stats_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno,
                     inode_t *inode, struct iatt *buf,
-                    struct iatt *preparent, struct iatt *postparent)
+                    struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
+        UPDATE_PROFILE_STATS (frame, MKNOD);
         STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
+                             preparent, postparent, xdata);
         return 0;
 }
 
@@ -557,10 +1512,30 @@ int
 io_stats_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno,
                     inode_t *inode, struct iatt *buf,
-                    struct iatt *preparent, struct iatt *postparent)
+                    struct iatt *preparent, struct iatt *postparent,
+                    dict_t *xdata)
 {
+        struct ios_stat *iosstat = NULL;
+        char   *path = frame->local;
+
+        UPDATE_PROFILE_STATS (frame, MKDIR);
+        if (op_ret < 0)
+                goto unwind;
+
+        iosstat = GF_CALLOC (1, sizeof (*iosstat), gf_io_stats_mt_ios_stat);
+        if (iosstat) {
+                LOCK_INIT (&iosstat->lock);
+                iosstat->filename = gf_strdup(path);
+                uuid_copy (iosstat->gfid, buf->ia_gfid);
+                ios_inode_ctx_set (inode, this, iosstat);
+        }
+
+unwind:
+        /* local is assigned with path */
+        GF_FREE (frame->local);
+        frame->local = NULL;
         STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
+                             preparent, postparent, xdata);
         return 0;
 }
 
@@ -569,31 +1544,44 @@ int
 io_stats_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno,
                    inode_t *inode, struct iatt *buf,
-                   struct iatt *preparent, struct iatt *postparent)
+                   struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
+        UPDATE_PROFILE_STATS (frame, LINK);
         STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
+                             preparent, postparent, xdata);
         return 0;
 }
 
 
 int
 io_stats_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno)
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, FLUSH);
+        STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
 
 int
 io_stats_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno, fd_t *fd)
+                      int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
-        if (op_ret >= 0)
-                ios_fd_ctx_set (fd, this, 0);
+        struct ios_stat *iosstat = NULL;
+        int              ret     = -1;
 
-        STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd);
+        UPDATE_PROFILE_STATS (frame, OPENDIR);
+        if (op_ret < 0)
+                goto unwind;
+
+        ios_fd_ctx_set (fd, this, 0);
+
+        ret = ios_inode_ctx_get (fd->inode, this, &iosstat);
+        if (!ret)
+                BUMP_STATS (iosstat, IOS_STATS_TYPE_OPENDIR);
+
+unwind:
+        STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
         return 0;
 }
 
@@ -601,10 +1589,13 @@ io_stats_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 int
 io_stats_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno,
-                    struct iatt *preparent, struct iatt *postparent)
+                    struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
+
+        UPDATE_PROFILE_STATS (frame, RMDIR);
+
         STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno,
-                             preparent, postparent);
+                             preparent, postparent, xdata);
         return 0;
 }
 
@@ -612,64 +1603,100 @@ io_stats_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 int
 io_stats_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                        int32_t op_ret, int32_t op_errno,
-                       struct iatt *prebuf, struct iatt *postbuf)
+                       struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
+        UPDATE_PROFILE_STATS (frame, TRUNCATE);
         STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno,
-                             prebuf, postbuf);
+                             prebuf, postbuf, xdata);
         return 0;
 }
 
 
 int
 io_stats_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno, struct statvfs *buf)
+                     int32_t op_ret, int32_t op_errno, struct statvfs *buf, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf);
+        UPDATE_PROFILE_STATS (frame, STATFS);
+        STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
 
 int
 io_stats_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno)
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, SETXATTR);
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
 
 int
 io_stats_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno, dict_t *dict)
+                       int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict);
+        UPDATE_PROFILE_STATS (frame, GETXATTR);
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
         return 0;
 }
 
 
 int
 io_stats_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                          int32_t op_ret, int32_t op_errno)
+                          int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        UPDATE_PROFILE_STATS (frame, REMOVEXATTR);
+        STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+io_stats_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, FSETXATTR);
+        STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+
+int
+io_stats_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        UPDATE_PROFILE_STATS (frame, FGETXATTR);
+        STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+
+int
+io_stats_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        UPDATE_PROFILE_STATS (frame, FREMOVEXATTR);
+        STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
 
 int
 io_stats_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno)
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, FSYNCDIR);
+        STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
 
 int
 io_stats_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno)
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (access, frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, ACCESS);
+        STACK_UNWIND_STRICT (access, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
@@ -677,340 +1704,372 @@ io_stats_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 int
 io_stats_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         int32_t op_ret, int32_t op_errno,
-                        struct iatt *prebuf, struct iatt *postbuf)
+                        struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
+        UPDATE_PROFILE_STATS (frame, FTRUNCATE);
         STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno,
-                             prebuf, postbuf);
+                             prebuf, postbuf, xdata);
         return 0;
 }
 
 
 int
 io_stats_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                    int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf);
+        UPDATE_PROFILE_STATS (frame, FSTAT);
+        STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
 
 int
+io_stats_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+		       int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+		       struct iatt *postbuf, dict_t *xdata)
+{
+	UPDATE_PROFILE_STATS(frame, FALLOCATE);
+	STACK_UNWIND_STRICT(fallocate, frame, op_ret, op_errno, prebuf, postbuf,
+			    xdata);
+	return 0;
+}
+
+
+int
+io_stats_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+		     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+		     struct iatt *postbuf, dict_t *xdata)
+{
+	UPDATE_PROFILE_STATS(frame, DISCARD);
+	STACK_UNWIND_STRICT(discard, frame, op_ret, op_errno, prebuf, postbuf,
+			    xdata);
+	return 0;
+}
+
+int
+io_stats_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
+{
+        UPDATE_PROFILE_STATS(frame, ZEROFILL);
+        STACK_UNWIND_STRICT(zerofill, frame, op_ret, op_errno, prebuf, postbuf,
+                            xdata);
+        return 0;
+}
+
+int
 io_stats_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int32_t op_ret, int32_t op_errno, struct flock *lock)
+                 int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock);
+        UPDATE_PROFILE_STATS (frame, LK);
+        STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock, xdata);
         return 0;
 }
 
 
 int
 io_stats_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno)
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, ENTRYLK);
+        STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
 
 int
 io_stats_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno, dict_t *dict)
+                      int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (xattrop, frame, op_ret, op_errno, dict);
+        UPDATE_PROFILE_STATS (frame, XATTROP);
+        STACK_UNWIND_STRICT (xattrop, frame, op_ret, op_errno, dict, xdata);
         return 0;
 }
 
 
 int
 io_stats_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno, dict_t *dict)
+                       int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (fxattrop, frame, op_ret, op_errno, dict);
+        UPDATE_PROFILE_STATS (frame, FXATTROP);
+        STACK_UNWIND_STRICT (fxattrop, frame, op_ret, op_errno, dict, xdata);
         return 0;
 }
 
 
 int
 io_stats_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno)
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, INODELK);
+        STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
-
 int
 io_stats_entrylk (call_frame_t *frame, xlator_t *this,
                   const char *volume, loc_t *loc, const char *basename,
-                  entrylk_cmd cmd, entrylk_type type)
+                  entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-        BUMP_FOP (ENTRYLK);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_entrylk_cbk,
                     FIRST_CHILD (this),
                     FIRST_CHILD (this)->fops->entrylk,
-                    volume, loc, basename, cmd, type);
+                    volume, loc, basename, cmd, type, xdata);
         return 0;
 }
 
 
 int
 io_stats_inodelk (call_frame_t *frame, xlator_t *this,
-                  const char *volume, loc_t *loc, int32_t cmd, struct flock *flock)
+                  const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
 
-        BUMP_FOP (INODELK);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_inodelk_cbk,
                     FIRST_CHILD (this),
                     FIRST_CHILD (this)->fops->inodelk,
-                    volume, loc, cmd, flock);
+                    volume, loc, cmd, flock, xdata);
         return 0;
 }
 
 
 int
 io_stats_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno)
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
 
-        STACK_UNWIND_STRICT (finodelk, frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, FINODELK);
+        STACK_UNWIND_STRICT (finodelk, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
 
 int
-io_stats_finodelk (call_frame_t *frame, xlator_t *this,
-                   const char *volume, fd_t *fd, int32_t cmd, struct flock *flock)
+io_stats_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+                   fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
-        BUMP_FOP (FINODELK);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_finodelk_cbk,
                     FIRST_CHILD (this),
                     FIRST_CHILD (this)->fops->finodelk,
-                    volume, fd, cmd, flock);
+                    volume, fd, cmd, flock, xdata);
         return 0;
 }
 
 
 int
-io_stats_xattrop (call_frame_t *frame, xlator_t *this,
-                  loc_t *loc, gf_xattrop_flags_t flags, dict_t *dict)
+io_stats_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                  gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
 {
-        BUMP_FOP (XATTROP);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_xattrop_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->xattrop,
-                    loc, flags, dict);
-
+                    loc, flags, dict, xdata);
         return 0;
 }
 
 
 int
-io_stats_fxattrop (call_frame_t *frame, xlator_t *this,
-                   fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict)
+io_stats_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                   gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
 {
-        BUMP_FOP (FXATTROP);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_fxattrop_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fxattrop,
-                    fd, flags, dict);
-
+                    fd, flags, dict, xdata);
         return 0;
 }
 
 
 int
 io_stats_lookup (call_frame_t *frame, xlator_t *this,
-                 loc_t *loc, dict_t *xattr_req)
+                 loc_t *loc, dict_t *xdata)
 {
-        BUMP_FOP (LOOKUP);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_lookup_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->lookup,
-                    loc, xattr_req);
-
+                    loc, xdata);
         return 0;
 }
 
 
 int
-io_stats_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
+io_stats_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-        BUMP_FOP (STAT);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_stat_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->stat,
-                    loc);
-
+                    loc, xdata);
         return 0;
 }
 
 
 int
 io_stats_readlink (call_frame_t *frame, xlator_t *this,
-                   loc_t *loc, size_t size)
+                   loc_t *loc, size_t size, dict_t *xdata)
 {
-        BUMP_FOP (READLINK);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_readlink_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readlink,
-                    loc, size);
-
+                    loc, size, xdata);
         return 0;
 }
 
 
 int
-io_stats_mknod (call_frame_t *frame, xlator_t *this,
-                loc_t *loc, mode_t mode, dev_t dev, dict_t *params)
+io_stats_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                mode_t mode, dev_t dev, mode_t umask, dict_t *xdata)
 {
-        BUMP_FOP (MKNOD);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_mknod_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->mknod,
-                    loc, mode, dev, params);
-
+                    loc, mode, dev, umask, xdata);
         return 0;
 }
 
 
 int
 io_stats_mkdir (call_frame_t *frame, xlator_t *this,
-                loc_t *loc, mode_t mode, dict_t *params)
+                loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
 {
-        BUMP_FOP (MKDIR);
+        frame->local = gf_strdup (loc->path);
+
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_mkdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->mkdir,
-                    loc, mode, params);
+                    loc, mode, umask, xdata);
         return 0;
 }
 
 
 int
 io_stats_unlink (call_frame_t *frame, xlator_t *this,
-                 loc_t *loc)
+                 loc_t *loc, int xflag, dict_t *xdata)
 {
-        BUMP_FOP (UNLINK);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_unlink_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->unlink,
-                    loc);
+                    loc, xflag, xdata);
         return 0;
 }
 
 
 int
 io_stats_rmdir (call_frame_t *frame, xlator_t *this,
-                loc_t *loc)
+                loc_t *loc, int flags, dict_t *xdata)
 {
-        BUMP_FOP (RMDIR);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_rmdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->rmdir,
-                    loc);
-
+                    loc, flags, xdata);
         return 0;
 }
 
 
 int
-io_stats_symlink (call_frame_t *frame, xlator_t *this,
-                  const char *linkpath, loc_t *loc, dict_t *params)
+io_stats_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
+                  loc_t *loc, mode_t umask, dict_t *xdata)
 {
-        BUMP_FOP (SYMLINK);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_symlink_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->symlink,
-                    linkpath, loc, params);
-
+                    linkpath, loc, umask, xdata);
         return 0;
 }
 
 
 int
 io_stats_rename (call_frame_t *frame, xlator_t *this,
-                 loc_t *oldloc, loc_t *newloc)
+                 loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
-        BUMP_FOP (RENAME);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_rename_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->rename,
-                    oldloc, newloc);
-
+                    oldloc, newloc, xdata);
         return 0;
 }
 
 
 int
 io_stats_link (call_frame_t *frame, xlator_t *this,
-               loc_t *oldloc, loc_t *newloc)
+               loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
-        BUMP_FOP (LINK);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_link_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->link,
-                    oldloc, newloc);
+                    oldloc, newloc, xdata);
         return 0;
 }
 
 
 int
 io_stats_setattr (call_frame_t *frame, xlator_t *this,
-                  loc_t *loc, struct iatt *stbuf, int32_t valid)
+                  loc_t *loc, struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-        BUMP_FOP (SETATTR);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_setattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->setattr,
-                    loc, stbuf, valid);
-
+                    loc, stbuf, valid, xdata);
         return 0;
 }
 
 
 int
 io_stats_truncate (call_frame_t *frame, xlator_t *this,
-                   loc_t *loc, off_t offset)
+                   loc_t *loc, off_t offset, dict_t *xdata)
 {
-        BUMP_FOP (TRUNCATE);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_truncate_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->truncate,
-                    loc, offset);
-
+                    loc, offset, xdata);
         return 0;
 }
 
 
 int
-io_stats_open (call_frame_t *frame, xlator_t *this,
-               loc_t *loc, int32_t flags, fd_t *fd, int32_t wbflags)
+io_stats_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
+               int32_t flags, fd_t *fd, dict_t *xdata)
 {
-        BUMP_FOP (OPEN);
-
         frame->local = gf_strdup (loc->path);
 
+        START_FOP_LATENCY (frame);
+
         STACK_WIND (frame, io_stats_open_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->open,
-                    loc, flags, fd, wbflags);
+                    loc, flags, fd, xdata);
         return 0;
 }
 
@@ -1018,32 +2077,32 @@ io_stats_open (call_frame_t *frame, xlator_t *this,
 int
 io_stats_create (call_frame_t *frame, xlator_t *this,
                  loc_t *loc, int32_t flags, mode_t mode,
-                 fd_t *fd, dict_t *params)
+                 mode_t umask, fd_t *fd, dict_t *xdata)
 {
-        BUMP_FOP (CREATE);
-
         frame->local = gf_strdup (loc->path);
 
+        START_FOP_LATENCY (frame);
+
         STACK_WIND (frame, io_stats_create_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->create,
-                    loc, flags, mode, fd, params);
+                    loc, flags, mode, umask, fd, xdata);
         return 0;
 }
 
 
 int
 io_stats_readv (call_frame_t *frame, xlator_t *this,
-                fd_t *fd, size_t size, off_t offset)
+                fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
 {
-        BUMP_FOP (READ);
-
         frame->local = fd;
 
+        START_FOP_LATENCY (frame);
+
         STACK_WIND (frame, io_stats_readv_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readv,
-                    fd, size, offset);
+                    fd, size, offset, flags, xdata);
         return 0;
 }
 
@@ -1052,67 +2111,69 @@ int
 io_stats_writev (call_frame_t *frame, xlator_t *this,
                  fd_t *fd, struct iovec *vector,
                  int32_t count, off_t offset,
-                 struct iobref *iobref)
+                 uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
         int                 len = 0;
 
-
+        if (fd->inode)
+                frame->local = fd->inode;
         len = iov_length (vector, count);
 
-        BUMP_FOP (WRITE);
         BUMP_WRITE (fd, len);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_writev_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->writev,
-                    fd, vector, count, offset, iobref);
+                    fd, vector, count, offset, flags, iobref, xdata);
         return 0;
+
 }
 
 
 int
 io_stats_statfs (call_frame_t *frame, xlator_t *this,
-                 loc_t *loc)
+                 loc_t *loc, dict_t *xdata)
 {
-        BUMP_FOP (STATFS);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_statfs_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->statfs,
-                    loc);
+                    loc, xdata);
         return 0;
 }
 
 
 int
 io_stats_flush (call_frame_t *frame, xlator_t *this,
-                fd_t *fd)
+                fd_t *fd, dict_t *xdata)
 {
-        BUMP_FOP (FLUSH);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_flush_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->flush,
-                    fd);
+                    fd, xdata);
         return 0;
 }
 
 
 int
 io_stats_fsync (call_frame_t *frame, xlator_t *this,
-                fd_t *fd, int32_t flags)
+                fd_t *fd, int32_t flags, dict_t *xdata)
 {
-        BUMP_FOP (FSYNC);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_fsync_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fsync,
-                    fd, flags);
+                    fd, flags, xdata);
         return 0;
 }
 
 
-void
+int
 conditional_dump (dict_t *dict, char *key, data_t *value, void *data)
 {
         struct {
@@ -1120,30 +2181,43 @@ conditional_dump (dict_t *dict, char *key, data_t *value, void *data)
                 inode_t        *inode;
                 const char     *path;
         } *stub;
-        xlator_t   *this = NULL;
-        inode_t    *inode = NULL;
-        const char *path = NULL;
-        char       *filename = NULL;
+        xlator_t             *this = NULL;
+        char                 *filename = NULL;
+        FILE                 *logfp = NULL;
+        struct ios_dump_args args = {0};
 
         stub  = data;
         this  = stub->this;
-        inode = stub->inode;
-        path  = stub->path;
 
         filename = alloca (value->len + 1);
         memset (filename, 0, value->len + 1);
         memcpy (filename, data_to_str (value), value->len);
 
         if (fnmatch ("*io*stat*dump", key, 0) == 0) {
-                io_stats_dump (this, filename, inode, path);
+
+                if (!strncmp (filename, "", 1)) {
+                        gf_log (this->name, GF_LOG_ERROR, "No filename given");
+                        return -1;
+                }
+                logfp = fopen (filename, "w+");
+                if (!logfp) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to open %s "
+                                "for writing", filename);
+                        return -1;
+                }
+                (void) ios_dump_args_init (&args, IOS_DUMP_TYPE_FILE,
+                                           logfp);
+                io_stats_dump (this, &args);
+                fclose (logfp);
         }
+        return 0;
 }
 
 
 int
 io_stats_setxattr (call_frame_t *frame, xlator_t *this,
                    loc_t *loc, dict_t *dict,
-                   int32_t flags)
+                   int32_t flags, dict_t *xdata)
 {
         struct {
                 xlator_t     *this;
@@ -1151,175 +2225,255 @@ io_stats_setxattr (call_frame_t *frame, xlator_t *this,
                 const char   *path;
         } stub;
 
-        BUMP_FOP (SETXATTR);
-
         stub.this  = this;
         stub.inode = loc->inode;
         stub.path  = loc->path;
 
         dict_foreach (dict, conditional_dump, &stub);
 
+        START_FOP_LATENCY (frame);
+
         STACK_WIND (frame, io_stats_setxattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->setxattr,
-                    loc, dict, flags);
+                    loc, dict, flags, xdata);
         return 0;
 }
 
 
 int
 io_stats_getxattr (call_frame_t *frame, xlator_t *this,
-                   loc_t *loc, const char *name)
+                   loc_t *loc, const char *name, dict_t *xdata)
 {
-        BUMP_FOP (GETXATTR);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_getxattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->getxattr,
-                    loc, name);
+                    loc, name, xdata);
         return 0;
 }
 
 
 int
 io_stats_removexattr (call_frame_t *frame, xlator_t *this,
-                      loc_t *loc, const char *name)
+                      loc_t *loc, const char *name, dict_t *xdata)
 {
-        BUMP_FOP (REMOVEXATTR);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_removexattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->removexattr,
-                    loc, name);
+                    loc, name, xdata);
+        return 0;
+}
+
+
+int
+io_stats_fsetxattr (call_frame_t *frame, xlator_t *this,
+                    fd_t *fd, dict_t *dict,
+                    int32_t flags, dict_t *xdata)
+{
+        START_FOP_LATENCY (frame);
+
+        STACK_WIND (frame, io_stats_fsetxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr,
+                    fd, dict, flags, xdata);
+        return 0;
+}
+
+
+int
+io_stats_fgetxattr (call_frame_t *frame, xlator_t *this,
+                    fd_t *fd, const char *name, dict_t *xdata)
+{
+        START_FOP_LATENCY (frame);
 
+        STACK_WIND (frame, io_stats_fgetxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fgetxattr,
+                    fd, name, xdata);
+        return 0;
+}
+
+
+int
+io_stats_fremovexattr (call_frame_t *frame, xlator_t *this,
+                       fd_t *fd, const char *name, dict_t *xdata)
+{
+        START_FOP_LATENCY (frame);
+
+        STACK_WIND (frame, io_stats_fremovexattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fremovexattr,
+                    fd, name, xdata);
         return 0;
 }
 
 
 int
 io_stats_opendir (call_frame_t *frame, xlator_t *this,
-                  loc_t *loc, fd_t *fd)
+                  loc_t *loc, fd_t *fd, dict_t *xdata)
 {
-        BUMP_FOP (OPENDIR);
+
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_opendir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->opendir,
-                    loc, fd);
+                    loc, fd, xdata);
         return 0;
 }
 
 int
 io_stats_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-                   off_t offset)
+                   off_t offset, dict_t *dict)
 {
-        BUMP_FOP (READDIRP);
+        frame->local = fd->inode;
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_readdirp_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readdirp,
-                    fd, size, offset);
-
+                    fd, size, offset, dict);
         return 0;
 }
 
 
 int
 io_stats_readdir (call_frame_t *frame, xlator_t *this,
-                  fd_t *fd, size_t size, off_t offset)
+                  fd_t *fd, size_t size, off_t offset, dict_t *xdata)
 {
-        BUMP_FOP (READDIR);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_readdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readdir,
-                    fd, size, offset);
-
+                    fd, size, offset, xdata);
         return 0;
 }
 
 
 int
 io_stats_fsyncdir (call_frame_t *frame, xlator_t *this,
-                   fd_t *fd, int32_t datasync)
+                   fd_t *fd, int32_t datasync, dict_t *xdata)
 {
-        BUMP_FOP (FSYNCDIR);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_fsyncdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fsyncdir,
-                    fd, datasync);
+                    fd, datasync, xdata);
         return 0;
 }
 
 
 int
 io_stats_access (call_frame_t *frame, xlator_t *this,
-                 loc_t *loc, int32_t mask)
+                 loc_t *loc, int32_t mask, dict_t *xdata)
 {
-        BUMP_FOP (ACCESS);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_access_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->access,
-                    loc, mask);
+                    loc, mask, xdata);
         return 0;
 }
 
 
 int
 io_stats_ftruncate (call_frame_t *frame, xlator_t *this,
-                    fd_t *fd, off_t offset)
+                    fd_t *fd, off_t offset, dict_t *xdata)
 {
-        BUMP_FOP (FTRUNCATE);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_ftruncate_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->ftruncate,
-                    fd, offset);
-
+                    fd, offset, xdata);
         return 0;
 }
 
 
 int
 io_stats_fsetattr (call_frame_t *frame, xlator_t *this,
-                   fd_t *fd, struct iatt *stbuf, int32_t valid)
+                   fd_t *fd, struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-        BUMP_FOP (FSETATTR);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_setattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fsetattr,
-                    fd, stbuf, valid);
+                    fd, stbuf, valid, xdata);
         return 0;
 }
 
 
 int
 io_stats_fstat (call_frame_t *frame, xlator_t *this,
-                fd_t *fd)
+                fd_t *fd, dict_t *xdata)
 {
-        BUMP_FOP (FSTAT);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_fstat_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fstat,
-                    fd);
+                    fd, xdata);
+        return 0;
+}
+
+
+int
+io_stats_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+		   off_t offset, size_t len, dict_t *xdata)
+{
+	START_FOP_LATENCY(frame);
+
+	STACK_WIND(frame, io_stats_fallocate_cbk, FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
+		   xdata);
+
+	return 0;
+}
+
+
+int
+io_stats_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+		 size_t len, dict_t *xdata)
+{
+	START_FOP_LATENCY(frame);
+
+	STACK_WIND(frame, io_stats_discard_cbk, FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
+
+	return 0;
+}
+
+int
+io_stats_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+                 size_t len, dict_t *xdata)
+{
+        START_FOP_LATENCY(frame);
+
+        STACK_WIND(frame, io_stats_zerofill_cbk, FIRST_CHILD(this),
+                   FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
+
         return 0;
 }
 
 
 int
 io_stats_lk (call_frame_t *frame, xlator_t *this,
-             fd_t *fd, int32_t cmd, struct flock *lock)
+             fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
 {
-        BUMP_FOP (LK);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_lk_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->lk,
-                    fd, cmd, lock);
+                    fd, cmd, lock, xdata);
         return 0;
 }
 
@@ -1328,15 +2482,23 @@ int
 io_stats_release (xlator_t *this, fd_t *fd)
 {
         struct ios_fd  *iosfd = NULL;
+        struct ios_conf *conf = NULL;
 
         BUMP_FOP (RELEASE);
 
+        conf = this->private;
+
+        LOCK (&conf->lock);
+        {
+                conf->cumulative.nr_opens--;
+        }
+        UNLOCK (&conf->lock);
+
         ios_fd_ctx_get (fd, this, &iosfd);
         if (iosfd) {
                 io_stats_dump_fd (this, iosfd);
 
-                if (iosfd->filename)
-                        GF_FREE (iosfd->filename);
+                GF_FREE (iosfd->filename);
                 GF_FREE (iosfd);
         }
 
@@ -1357,10 +2519,141 @@ int
 io_stats_forget (xlator_t *this, inode_t *inode)
 {
         BUMP_FOP (FORGET);
+        ios_stats_cleanup (this, inode);
+        return 0;
+}
+
+static int
+ios_init_top_stats (struct ios_conf *conf)
+{
+        int     i = 0;
+
+        GF_ASSERT (conf);
+
+        for (i = 0; i <IOS_STATS_TYPE_MAX; i++) {
+                conf->list[i].iosstats = GF_CALLOC (1,
+                                         sizeof(*conf->list[i].iosstats),
+                                         gf_io_stats_mt_ios_stat);
+
+                if (!conf->list[i].iosstats)
+                        return -1;
+
+                INIT_LIST_HEAD(&conf->list[i].iosstats->list);
+                LOCK_INIT (&conf->list[i].lock);
+        }
+
+        for (i = 0; i < IOS_STATS_THRU_MAX; i ++) {
+                conf->thru_list[i].iosstats = GF_CALLOC (1,
+                                 sizeof (*conf->thru_list[i].iosstats),
+                                 gf_io_stats_mt_ios_stat);
+
+                if (!conf->thru_list[i].iosstats)
+                        return -1;
+
+                INIT_LIST_HEAD(&conf->thru_list[i].iosstats->list);
+                LOCK_INIT (&conf->thru_list[i].lock);
+        }
 
         return 0;
 }
 
+static void
+ios_destroy_top_stats (struct ios_conf *conf)
+{
+        int                     i = 0;
+        struct ios_stat_head    *list_head = NULL;
+        struct ios_stat_list    *entry     = NULL;
+        struct ios_stat_list    *tmp       = NULL;
+        struct ios_stat_list    *list      = NULL;
+        struct ios_stat         *stat      = NULL;
+
+        GF_ASSERT (conf);
+
+        LOCK (&conf->lock);
+
+        conf->cumulative.nr_opens = 0;
+        conf->cumulative.max_nr_opens = 0;
+        conf->cumulative.max_openfd_time.tv_sec = 0;
+        conf->cumulative.max_openfd_time.tv_usec = 0;
+
+        for (i = 0; i < IOS_STATS_TYPE_MAX; i++) {
+                list_head = &conf->list[i];
+                if (!list_head)
+                        continue;
+                list_for_each_entry_safe (entry, tmp,
+                                          &list_head->iosstats->list, list) {
+                        list = entry;
+                        stat = list->iosstat;
+                        ios_stat_unref (stat);
+                        list_del (&list->list);
+                        GF_FREE (list);
+                        list_head->members--;
+                }
+        }
+
+        for (i = 0; i < IOS_STATS_THRU_MAX; i++) {
+                list_head = &conf->thru_list[i];
+                if (!list_head)
+                        continue;
+                list_for_each_entry_safe (entry, tmp,
+                                          &list_head->iosstats->list, list) {
+                        list = entry;
+                        stat = list->iosstat;
+                        ios_stat_unref (stat);
+                        list_del (&list->list);
+                        GF_FREE (list);
+                        list_head->members--;
+                }
+        }
+
+        UNLOCK (&conf->lock);
+
+        return;
+}
+
+int
+reconfigure (xlator_t *this, dict_t *options)
+{
+        struct ios_conf    *conf = NULL;
+        int                 ret = -1;
+        char               *sys_log_str = NULL;
+        int                 sys_log_level = -1;
+        char               *log_str = NULL;
+        int                 log_level = -1;
+
+        if (!this || !this->private)
+                goto out;
+
+        conf = this->private;
+
+        GF_OPTION_RECONF ("dump-fd-stats", conf->dump_fd_stats, options, bool,
+                          out);
+
+        GF_OPTION_RECONF ("count-fop-hits", conf->count_fop_hits, options, bool,
+                          out);
+
+        GF_OPTION_RECONF ("latency-measurement", conf->measure_latency,
+                          options, bool, out);
+
+        GF_OPTION_RECONF ("sys-log-level", sys_log_str, options, str, out);
+        if (sys_log_str) {
+                sys_log_level = glusterd_check_log_level (sys_log_str);
+                set_sys_log_level (sys_log_level);
+        }
+
+        GF_OPTION_RECONF ("log-level", log_str, options, str, out);
+        if (log_str) {
+                log_level = glusterd_check_log_level (log_str);
+                gf_log_set_loglevel (log_level);
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "reconfigure returning %d", ret);
+        return ret;
+}
+
+
 int32_t
 mem_acct_init (xlator_t *this)
 {
@@ -1370,7 +2663,7 @@ mem_acct_init (xlator_t *this)
                 return ret;
 
         ret = xlator_mem_acct_init (this, gf_io_stats_mt_end + 1);
-        
+
         if (ret != 0) {
                 gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
                         " failed");
@@ -1383,27 +2676,30 @@ mem_acct_init (xlator_t *this)
 int
 init (xlator_t *this)
 {
-        dict_t             *options = NULL;
         struct ios_conf    *conf = NULL;
-        char               *str = NULL;
-        int                 ret = 0;
+        char               *sys_log_str = NULL;
+        int                 sys_log_level = -1;
+        char               *log_str = NULL;
+        int                 log_level = -1;
+        int                 ret = -1;
 
         if (!this)
                 return -1;
 
-        if (!this->children || this->children->next) {
+        if (!this->children) {
                 gf_log (this->name, GF_LOG_ERROR,
-                        "io_stats translator requires one subvolume");
+                        "io_stats translator requires atleast one subvolume");
                 return -1;
         }
 
         if (!this->parents) {
-                gf_log (this->name, GF_LOG_WARNING,
+                /* This is very much valid as io-stats currently is loaded
+                 * on top of volumes on both client and server, hence this is
+                 * not an warning message */
+                gf_log (this->name, GF_LOG_DEBUG,
                         "dangling volume. check volfile ");
         }
 
-        options = this->options;
-
         conf = GF_CALLOC (1, sizeof(*conf), gf_io_stats_mt_ios_conf);
 
         if (!conf) {
@@ -1417,24 +2713,33 @@ init (xlator_t *this)
         gettimeofday (&conf->cumulative.started_at, NULL);
         gettimeofday (&conf->incremental.started_at, NULL);
 
-        ret = dict_get_str (options, "dump-fd-stats", &str);
-        if (ret == 0) {
-                ret = gf_string2boolean (str, &conf->dump_fd_stats);
-                if (ret == -1) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "'dump-fd-stats' takes only boolean arguments");
-                        return -1;
-                }
+        ret = ios_init_top_stats (conf);
+        if (ret)
+                return -1;
 
-                if (conf->dump_fd_stats) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"enabling dump-fd-stats");
-                }
+        GF_OPTION_INIT ("dump-fd-stats", conf->dump_fd_stats, bool, out);
+
+        GF_OPTION_INIT ("count-fop-hits", conf->count_fop_hits, bool, out);
+
+        GF_OPTION_INIT ("latency-measurement", conf->measure_latency,
+                          bool, out);
+
+        GF_OPTION_INIT ("sys-log-level", sys_log_str, str, out);
+        if (sys_log_str) {
+                sys_log_level = glusterd_check_log_level (sys_log_str);
+                set_sys_log_level (sys_log_level);
         }
 
-        this->private = conf;
+        GF_OPTION_INIT ("log-level", log_str, str, out);
+        if (log_str) {
+                log_level = glusterd_check_log_level (log_str);
+                gf_log_set_loglevel (log_level);
+        }
 
-        return 0;
+        this->private = conf;
+        ret = 0;
+out:
+        return ret;
 }
 
 
@@ -1448,13 +2753,102 @@ fini (xlator_t *this)
 
         conf = this->private;
 
+        if (!conf)
+                return;
+        this->private = NULL;
+
+        ios_destroy_top_stats (conf);
+
         GF_FREE(conf);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "io-stats translator unloaded");
         return;
 }
 
+int
+notify (xlator_t *this, int32_t event, void *data, ...)
+{
+        int          ret = 0;
+        struct ios_dump_args args = {0};
+        dict_t       *output = NULL;
+        dict_t       *dict = NULL;
+        int32_t       top_op = 0;
+        int32_t       list_cnt = 0;
+        double        throughput = 0;
+        double        time = 0;
+        va_list ap;
+
+        dict = data;
+        va_start (ap, data);
+        output = va_arg (ap, dict_t*);
+        va_end (ap);
+        switch (event) {
+        case GF_EVENT_TRANSLATOR_INFO:
+                ret = dict_get_str_boolean (dict, "clear-stats", _gf_false);
+                if (ret) {
+                         ret = dict_set_int32 (output, "top-op", top_op);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set top-op in dict");
+                                goto out;
+                        }
+                        ios_destroy_top_stats (this->private);
+                        ret = ios_init_top_stats (this->private);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to reset top stats");
+                        ret = dict_set_int32 (output, "stats-cleared",
+                                              ret ? 0 : 1);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set stats-cleared"
+                                        " in dict");
+                        goto out;
+                }
+
+                ret = dict_get_int32 (dict, "top-op", &top_op);
+                if (!ret) {
+                        ret = dict_get_int32 (dict, "list-cnt", &list_cnt);
+                        if (top_op > IOS_STATS_TYPE_NONE &&
+                            top_op < IOS_STATS_TYPE_MAX)
+                                ret = io_stats_dump_stats_to_dict (this, output,
+                                                             top_op, list_cnt);
+                        if (top_op == IOS_STATS_TYPE_READ_THROUGHPUT ||
+                                top_op == IOS_STATS_TYPE_WRITE_THROUGHPUT) {
+                                ret = dict_get_double (dict, "throughput",
+                                                        &throughput);
+                                if (!ret) {
+                                        ret = dict_get_double (dict, "time",
+                                                                &time);
+                                        if (ret)
+                                                goto out;
+                                        ret = dict_set_double (output,
+                                                "throughput", throughput);
+                                        if (ret)
+                                                goto out;
+                                        ret = dict_set_double (output, "time",
+                                                        time);
+                                        if (ret)
+                                                goto out;
+                                }
+                                ret = 0;
+
+                        }
+                } else {
+                        (void) ios_dump_args_init (&args, IOS_DUMP_TYPE_DICT,
+                                           output);
+                        ret = io_stats_dump (this, &args);
+                }
+                break;
+        default:
+                default_notify (this, event, data);
+                break;
+
+        }
+out:
+        return ret;
+}
 
 struct xlator_fops fops = {
         .stat        = io_stats_stat,
@@ -1476,6 +2870,9 @@ struct xlator_fops fops = {
         .setxattr    = io_stats_setxattr,
         .getxattr    = io_stats_getxattr,
         .removexattr = io_stats_removexattr,
+        .fsetxattr    = io_stats_fsetxattr,
+        .fgetxattr    = io_stats_fgetxattr,
+        .fremovexattr = io_stats_fremovexattr,
         .opendir     = io_stats_opendir,
         .readdir     = io_stats_readdir,
         .readdirp    = io_stats_readdirp,
@@ -1493,6 +2890,9 @@ struct xlator_fops fops = {
         .fxattrop    = io_stats_fxattrop,
         .setattr     = io_stats_setattr,
         .fsetattr    = io_stats_fsetattr,
+	.fallocate   = io_stats_fallocate,
+	.discard     = io_stats_discard,
+        .zerofill    = io_stats_zerofill,
 };
 
 struct xlator_cbks cbks = {
@@ -1504,6 +2904,47 @@ struct xlator_cbks cbks = {
 struct volume_options options[] = {
         { .key  = {"dump-fd-stats"},
           .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "If on stats related to file-operations would be "
+                         "tracked inside GlusterFS data-structures."
+        },
+        { .key  = { "latency-measurement" },
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "If on stats related to the latency of each operation "
+                         "would be tracked inside GlusterFS data-structures. "
+        },
+        { .key  = {"count-fop-hits"},
+          .type = GF_OPTION_TYPE_BOOL,
+        },
+        { .key = {"log-level"},
+          .type = GF_OPTION_TYPE_STR,
+          .value = { "DEBUG", "WARNING", "ERROR", "INFO",
+                     "CRITICAL", "NONE", "TRACE"}
+        },
+
+        /* These are synthetic entries to assist validation of CLI's  *
+         *  volume set  command                                       */
+        { .key = {"client-log-level"},
+          .type = GF_OPTION_TYPE_STR,
+          .default_value = "INFO",
+          .description = "Changes the log-level of the clients",
+          .value = { "DEBUG", "WARNING", "ERROR", "INFO",
+                     "CRITICAL", "NONE", "TRACE"}
+        },
+        { .key = {"sys-log-level"},
+          .type = GF_OPTION_TYPE_STR,
+          .default_value = "CRITICAL",
+          .description = "Gluster's syslog log-level",
+          .value = { "WARNING", "ERROR", "INFO", "CRITICAL"}
+        },
+        { .key = {"brick-log-level"},
+          .type = GF_OPTION_TYPE_STR,
+          .default_value = "INFO",
+          .description = "Changes the log-level of the bricks",
+          .value = { "DEBUG", "WARNING", "ERROR", "INFO",
+                     "CRITICAL", "NONE", "TRACE"}
         },
         { .key  = {NULL} },
+
 };
diff --git a/xlators/debug/trace/src/Makefile.am b/xlators/debug/trace/src/Makefile.am
index 0f1679a04..7b2597b4d 100644
--- a/xlators/debug/trace/src/Makefile.am
+++ b/xlators/debug/trace/src/Makefile.am
@@ -2,13 +2,15 @@
 xlator_LTLIBRARIES = trace.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/debug
 
-trace_la_LDFLAGS = -module -avoidversion
+trace_la_LDFLAGS = -module -avoid-version
 
 trace_la_SOURCES = trace.c
 trace_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+noinst_HEADERS = trace.h trace-mem-types.h
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/debug/trace/src/trace-mem-types.h b/xlators/debug/trace/src/trace-mem-types.h
new file mode 100644
index 000000000..9fa7d97c2
--- /dev/null
+++ b/xlators/debug/trace/src/trace-mem-types.h
@@ -0,0 +1,21 @@
+/*
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef __TRACE_MEM_TYPES_H__
+#define __TRACE_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_trace_mem_types_ {
+        gf_trace_mt_trace_conf_t = gf_common_mt_end + 1,
+        gf_trace_mt_end
+};
+#endif
diff --git a/xlators/debug/trace/src/trace.c b/xlators/debug/trace/src/trace.c
index f5af77681..c9d839356 100644
--- a/xlators/debug/trace/src/trace.c
+++ b/xlators/debug/trace/src/trace.c
@@ -1,26 +1,15 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
+#include "trace.h"
+#include "trace-mem-types.h"
 
 /**
  * xlators/debug/trace :
@@ -29,1506 +18,2016 @@
  *    Very helpful translator for debugging.
  */
 
-#include <time.h>
-#include <errno.h>
-#include "glusterfs.h"
-#include "xlator.h"
-#include "common-utils.h"
-
-#define ERR_EINVAL_NORETURN(cond)                               \
-        do                                                      \
-        {                                                       \
-                if ((cond))                                     \
-                {                                               \
-                        gf_log ("ERROR", GF_LOG_ERROR,          \
-                                "%s: %s: (%s) is true",         \
-                                __FILE__, __FUNCTION__, #cond); \
-                }                                               \
-        } while (0)
-
-
-typedef struct trace_private {
-        int32_t debug_flag;
-} trace_private_t;
-
-
-struct {
-        char *name;
-        int enabled;
-} trace_fop_names[GF_FOP_MAXVALUE];
-
-int trace_log_level = GF_LOG_NORMAL;
-
-static char *
-trace_stat_to_str (struct iatt *stbuf)
+int
+dump_history_trace (circular_buffer_t *cb, void *data)
 {
-        char    *statstr           = NULL;
-        char     atime_buf[256]    = {0,};
-        char     mtime_buf[256]    = {0,};
-        char     ctime_buf[256]    = {0,};
-        int      asprint_ret_value = 0;
-        uint64_t ia_time           = 0;
+        char    *string      = NULL;
+        struct  tm      *tm  = NULL;
+        char    timestr[256] = {0,};
 
-        ia_time = stbuf->ia_atime;
-        strftime (atime_buf, 256, "[%b %d %H:%M:%S]",
-                  localtime ((time_t *)&ia_time));
+        string = (char *)cb->data;
+        tm    = localtime (&cb->tv.tv_sec);
 
-        ia_time = stbuf->ia_mtime;
-        strftime (mtime_buf, 256, "[%b %d %H:%M:%S]",
-                  localtime ((time_t *)&ia_time));
-
-        ia_time = stbuf->ia_ctime;
-        strftime (ctime_buf, 256, "[%b %d %H:%M:%S]",
-                  localtime ((time_t *)&ia_time));
-
-        asprint_ret_value = gf_asprintf (&statstr,
-                                      "ia_ino=%"PRIu64
-                                      ", st_mode=%o, ia_nlink=%"GF_PRI_NLINK", "
-                                      "ia_uid=%d, ia_gid=%d, ia_size=%"PRId64", ia_blocks=%"PRId64
-                                      ", ia_atime=%s, ia_mtime=%s, ia_ctime=%s",
-                                      stbuf->ia_ino,
-                                      st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type),
-                                      stbuf->ia_nlink, stbuf->ia_uid,
-                                      stbuf->ia_gid, stbuf->ia_size,
-                                      stbuf->ia_blocks, atime_buf,
-                                      mtime_buf, ctime_buf);
+        /* Since we are continuing with adding entries to the buffer even when
+           gettimeofday () fails, it's safe to check tm and then dump the time
+           at which the entry was added to the buffer */
+        if (tm) {
+                strftime (timestr, 256, "%Y-%m-%d %H:%M:%S", tm);
+                snprintf (timestr + strlen (timestr), 256 - strlen (timestr),
+                          ".%"GF_PRI_SUSECONDS, cb->tv.tv_usec);
+                gf_proc_dump_write ("TIME", "%s", timestr);
+        }
 
-        if (asprint_ret_value < 0)
-                statstr = NULL;
+        gf_proc_dump_write ("FOP", "%s\n", string);
 
-        return statstr;
+        return 0;
 }
 
-
 int
 trace_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno, fd_t *fd,
                   inode_t *inode, struct iatt *buf,
-                  struct iatt *preparent, struct iatt *postparent)
+                  struct iatt *preparent, struct iatt *postparent,
+                  dict_t *xdata)
 {
-        char  *statstr = NULL;
-        char  *preparentstr = NULL;
-        char  *postparentstr = NULL;
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_CREATE].enabled) {
+                char  string[4096] = {0,};
                 if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preparentstr = trace_stat_to_str (preparent);
-                        postparentstr = trace_stat_to_str (postparent);
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, fd=%p, ino=%"PRIu64" "
-                                "*stbuf {%s}, *preparent {%s}, *postparent = "
-                                "{%s})",
-                                frame->root->unique, op_ret, fd, inode->ino,
-                                statstr, preparentstr, postparentstr);
-
-                        if (statstr)
-                                GF_FREE (statstr);
-                        if (preparentstr)
-                                GF_FREE (preparentstr);
-                        if (postparentstr)
-                                GF_FREE (postparentstr);
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s (op_ret=%d, fd=%p"
+                                  "*stbuf {%s}, *preparent {%s}, "
+                                  "*postparent = {%s})",
+                                  frame->root->unique,
+                                  uuid_utoa (inode->gfid), op_ret, fd,
+                                  statstr, preparentstr, postparentstr);
+
+                        /* for 'release' log */
+                        fd_ctx_set (fd, this, 0);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, op_errno=%d)",
+                                  frame->root->unique, op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
-                             preparent, postparent);
+out:
+        TRACE_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, buf,
+                            preparent, postparent, xdata);
         return 0;
 }
 
-
 int
 trace_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                int32_t op_ret, int32_t op_errno, fd_t *fd)
+                int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
+        trace_conf_t      *conf = NULL;
 
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_OPEN].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (op_ret=%d, op_errno=%d, *fd=%p)",
-                        frame->root->unique, op_ret, op_errno, fd);
+                char     string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d, "
+                          "*fd=%p", frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno,
+                          fd);
+
+                LOG_ELEMENT (conf, string);
         }
 
-        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd);
+out:
+        /* for 'release' log */
+        if (op_ret >= 0)
+                fd_ctx_set (fd, this, 0);
+
+        TRACE_STACK_UNWIND (open, frame, op_ret, op_errno, fd, xdata);
         return 0;
 }
 
-
 int
 trace_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                dict_t *xdata)
 {
-        uint64_t ia_time = 0;
-        char     atime_buf[256];
-        char     mtime_buf[256];
-        char     ctime_buf[256];
+        char          statstr[4096] = {0, };
+        trace_conf_t  *conf         = NULL;
 
-        if (trace_fop_names[GF_FOP_STAT].enabled) {
-                if (op_ret >= 0) {
-                        ia_time = buf->ia_atime;
-                        strftime (atime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
+        conf = this->private;
 
-                        ia_time = buf->ia_mtime;
-                        strftime (mtime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        ia_time = buf->ia_ctime;
-                        strftime (ctime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, buf {"
-                                "ia_ino=%"PRIu64", st_mode=%o, ia_nlink=%"GF_PRI_NLINK", "
-                                "ia_uid=%d, ia_gid=%d, ia_rdev=%"PRIu64", ia_size=%"PRId64
-                                ", ia_blksize=%"GF_PRI_BLKSIZE", ia_blocks=%"PRId64", "
-                                "ia_atime=%s, ia_mtime=%s, ia_ctime=%s})",
-                                frame->root->unique, op_ret, buf->ia_ino,
-                                st_mode_from_ia (buf->ia_prot, buf->ia_type),
-                                buf->ia_nlink, buf->ia_uid, buf->ia_gid,
-                                buf->ia_rdev, buf->ia_size, buf->ia_blksize,
-                                buf->ia_blocks, atime_buf, mtime_buf, ctime_buf);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_STAT].enabled) {
+                char string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d buf=%s",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  statstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf);
+out:
+        TRACE_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
-
 int
 trace_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno, struct iovec *vector,
-                 int32_t count, struct iatt *buf, struct iobref *iobref)
+                 int32_t count, struct iatt *buf, struct iobref *iobref,
+                 dict_t *xdata)
 {
-        uint64_t ia_time = 0;
-        char     atime_buf[256];
-        char     mtime_buf[256];
-        char     ctime_buf[256];
+        char          statstr[4096] = {0, };
+        trace_conf_t  *conf         = NULL;
+
+        conf = this->private;
 
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_READ].enabled) {
+                char  string[4096] = {0,};
                 if (op_ret >= 0) {
-                        ia_time = buf->ia_atime;
-                        strftime (atime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        ia_time = buf->ia_mtime;
-                        strftime (mtime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        ia_time = buf->ia_ctime;
-                        strftime (ctime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d, *buf {"
-                                "ia_ino=%"PRIu64", st_mode=%o, ia_nlink=%"GF_PRI_NLINK", "
-                                "ia_uid=%d, ia_gid=%d, ia_rdev=%"PRIu64", "
-                                "ia_size=%"PRId64", ia_blksize=%"GF_PRI_BLKSIZE", "
-                                "ia_blocks=%"PRId64", ia_atime=%s, ia_mtime=%s, ia_ctime=%s})",
-                                frame->root->unique, op_ret, op_errno, buf->ia_ino,
-                                st_mode_from_ia (buf->ia_prot, buf->ia_type),
-                                buf->ia_nlink, buf->ia_uid, buf->ia_gid,
-                                buf->ia_rdev, buf->ia_size, buf->ia_blksize, buf->ia_blocks,
-                                atime_buf, mtime_buf, ctime_buf);
+                        trace_stat_to_str (buf, statstr);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d buf=%s",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  statstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector, count,
-                             buf, iobref);
+out:
+        TRACE_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count,
+                            buf, iobref, xdata);
         return 0;
 }
 
-
 int
 trace_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno,
-                  struct iatt *prebuf, struct iatt *postbuf)
+                  struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
-        char  *preopstr = NULL;
-        char  *postopstr = NULL;
+        char         preopstr[4096]  = {0, };
+        char         postopstr[4096] = {0, };
+        trace_conf_t *conf           = NULL;
+
+        conf = this->private;
 
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_WRITE].enabled) {
+                char  string[4096] = {0,};
                 if (op_ret >= 0) {
-                        preopstr = trace_stat_to_str (prebuf);
-                        postopstr = trace_stat_to_str (postbuf);
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, ino = %"PRIu64
-                                ", *prebuf = {%s}, *postbuf = {%s})",
-                                frame->root->unique, op_ret, postbuf->ia_ino,
-                                preopstr, postopstr);
-
-                        if (preopstr)
-                                GF_FREE (preopstr);
-
-                        if (postopstr)
-                                GF_FREE (postopstr);
+                        trace_stat_to_str (prebuf, preopstr);
+                        trace_stat_to_str (postbuf, postopstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s})",
+                                  frame->root->unique, op_ret,
+                                  preopstr, postopstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf);
+out:
+        TRACE_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf,
+                            xdata);
         return 0;
 }
 
-
-
 int
 trace_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, gf_dirent_t *buf)
+                   int32_t op_ret, int32_t op_errno, gf_dirent_t *buf,
+                   dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_READDIR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64" :(op_ret=%d, op_errno=%d)",
-                        frame->root->unique, op_ret, op_errno);
-        }
+                char    string[4096]  = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64" : gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique, uuid_utoa (frame->local),
+                          op_ret, op_errno);
 
-        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, buf);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (readdir, frame, op_ret, op_errno, buf, xdata);
 
         return 0;
 }
 
-
 int
 trace_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, gf_dirent_t *buf)
+                    int32_t op_ret, int32_t op_errno, gf_dirent_t *buf,
+                    dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_READDIRP].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64" :(op_ret=%d, op_errno=%d)",
-                        frame->root->unique, op_ret, op_errno);
-        }
+                char        string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64" : gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique, uuid_utoa (frame->local),
+                          op_ret, op_errno);
 
-        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, buf);
+                LOG_ELEMENT (conf, string);
+        }
 
+out:
+        TRACE_STACK_UNWIND (readdirp, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
-
 int
 trace_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno,
-                 struct iatt *prebuf, struct iatt *postbuf)
+                 struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
-        char  *preopstr = NULL;
-        char  *postopstr = NULL;
-
-        if (trace_fop_names[GF_FOP_FSYNC].enabled) {
-                if (op_ret >= 0) {
-                        preopstr = trace_stat_to_str (prebuf);
-                        postopstr = trace_stat_to_str (postbuf);
+        char          preopstr[4096]  = {0, };
+        char          postopstr[4096] = {0, };
+        trace_conf_t  *conf           = NULL;
 
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, ino = %"PRIu64
-                                ", *prebuf = {%s}, *postbuf = {%s}",
-                                frame->root->unique, op_ret, postbuf->ia_ino,
-                                preopstr, postopstr);
+        conf = this->private;
 
-                        if (preopstr)
-                                GF_FREE (preopstr);
-
-                        if (postopstr)
-                                GF_FREE (postopstr);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSYNC].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (prebuf, preopstr);
+                        trace_stat_to_str (postbuf, postopstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s}",
+                                  frame->root->unique, op_ret,
+                                  preopstr, postopstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf);
+out:
+        TRACE_STACK_UNWIND (fsync, frame, op_ret, op_errno, prebuf, postbuf,
+                            xdata);
 
         return 0;
 }
 
-
 int
 trace_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno,
-                   struct iatt *statpre, struct iatt *statpost)
+                   struct iatt *statpre, struct iatt *statpost, dict_t *xdata)
 {
-        uint64_t ia_time         = 0;
-        char     atime_pre[256]  = {0,};
-        char     mtime_pre[256]  = {0,};
-        char     ctime_pre[256]  = {0,};
-        char     atime_post[256] = {0,};
-        char     mtime_post[256] = {0,};
-        char     ctime_post[256] = {0,};
+        char          preopstr[4096]  = {0, };
+        char          postopstr[4096] = {0, };
+        trace_conf_t  *conf           = NULL;
 
-        if (trace_fop_names[GF_FOP_SETATTR].enabled) {
-                if (op_ret >= 0) {
-                        ia_time = statpre->ia_atime;
-                        strftime (atime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        ia_time = statpre->ia_mtime;
-                        strftime (mtime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
+        conf = this->private;
 
-                        ia_time = statpre->ia_ctime;
-                        strftime (ctime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        ia_time = statpost->ia_atime;
-                        strftime (atime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        ia_time = statpost->ia_mtime;
-                        strftime (mtime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        ia_time = statpost->ia_ctime;
-                        strftime (ctime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, *statpre "
-                                "{ia_ino=%"PRIu64", st_mode=%o, ia_uid=%d, "
-                                "ia_gid=%d, ia_atime=%s, ia_mtime=%s, "
-                                "ia_ctime=%s}, *statpost {ia_ino=%"PRIu64", "
-                                "st_mode=%o, ia_uid=%d, ia_gid=%d, ia_atime=%s,"
-                                " ia_mtime=%s, ia_ctime=%s})",
-                                frame->root->unique, op_ret, statpre->ia_ino,
-                                st_mode_from_ia (statpre->ia_prot, statpre->ia_type),
-                                statpre->ia_uid,
-                                statpre->ia_gid, atime_pre, mtime_pre,
-                                ctime_pre, statpost->ia_ino,
-                                st_mode_from_ia (statpost->ia_prot, statpost->ia_type),
-                                statpost->ia_uid, statpost->ia_gid, atime_post,
-                                mtime_post, ctime_post);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_SETATTR].enabled) {
+                char  string[4096]  = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (statpre, preopstr);
+                        trace_stat_to_str (statpost, postopstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s})",
+                                  frame->root->unique, op_ret,
+                                  preopstr, postopstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, statpre, statpost);
+out:
+        TRACE_STACK_UNWIND (setattr, frame, op_ret, op_errno, statpre,
+                            statpost, xdata);
         return 0;
 }
 
-
 int
 trace_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno,
-                    struct iatt *statpre, struct iatt *statpost)
+                    struct iatt *statpre, struct iatt *statpost, dict_t *xdata)
 {
-        uint64_t ia_time         = 0;
-        char     atime_pre[256]  = {0,};
-        char     mtime_pre[256]  = {0,};
-        char     ctime_pre[256]  = {0,};
-        char     atime_post[256] = {0,};
-        char     mtime_post[256] = {0,};
-        char     ctime_post[256] = {0,};
-
-        if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
-                if (op_ret >= 0) {
-                        ia_time = statpre->ia_atime;
-                        strftime (atime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
+        char          preopstr[4096]  = {0, };
+        char          postopstr[4096] = {0, };
+        trace_conf_t  *conf           = NULL;
 
-                        ia_time = statpre->ia_mtime; 
-                        strftime (mtime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        ia_time = statpre->ia_ctime;
-                        strftime (ctime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        ia_time = statpost->ia_atime;
-                        strftime (atime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
+        conf = this->private;
 
-                        ia_time = statpost->ia_mtime;
-                        strftime (mtime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        ia_time = statpost->ia_ctime;
-                        strftime (ctime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, *statpre "
-                                "{ia_ino=%"PRIu64", st_mode=%o, ia_uid=%d, "
-                                "ia_gid=%d, ia_atime=%s, ia_mtime=%s, "
-                                "ia_ctime=%s}, *statpost {ia_ino=%"PRIu64", "
-                                "st_mode=%o, ia_uid=%d, ia_gid=%d, ia_atime=%s,"
-                                " ia_mtime=%s, ia_ctime=%s})",
-                                frame->root->unique, op_ret, statpre->ia_ino,
-                                st_mode_from_ia (statpre->ia_prot, statpre->ia_type),
-                                statpre->ia_uid,
-                                statpre->ia_gid, atime_pre, mtime_pre,
-                                ctime_pre, statpost->ia_ino,
-                                st_mode_from_ia (statpost->ia_prot, statpost->ia_type),
-                                statpost->ia_uid, statpost->ia_gid, atime_post,
-                                mtime_post, ctime_post);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (statpre, preopstr);
+                        trace_stat_to_str (statpost, postopstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s})",
+                                  frame->root->unique, op_ret,
+                                  preopstr, postopstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, op_errno=%d)",
+                                  frame->root->unique, uuid_utoa (frame->local),
+                                  op_ret, op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (fsetattr, frame, op_ret, op_errno,
-                             statpre, statpost);
+out:
+        TRACE_STACK_UNWIND (fsetattr, frame, op_ret, op_errno,
+                            statpre, statpost, xdata);
         return 0;
 }
 
-
 int
 trace_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno,
-                  struct iatt *preparent, struct iatt *postparent)
+                  struct iatt *preparent, struct iatt *postparent,
+                  dict_t *xdata)
 {
-        char *preparentstr = NULL;
-        char *postparentstr = NULL;
-
-        if (trace_fop_names[GF_FOP_UNLINK].enabled) {
-                if (op_ret >= 0) {
-                        preparentstr = trace_stat_to_str (preparent);
-                        postparentstr = trace_stat_to_str (postparent);
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, *preparent = {%s}, "
-                                "*postparent = {%s})",
-                                frame->root->unique, op_ret, preparentstr,
-                                postparentstr);
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
-                        if (preparentstr)
-                                GF_FREE (preparentstr);
+        conf = this->private;
 
-                        if (postparentstr)
-                                GF_FREE (postparentstr);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_UNLINK].enabled) {
+                char string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  " *preparent = {%s}, "
+                                  "*postparent = {%s})",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, preparentstr,
+                                  postparentstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
-                             preparent, postparent);
+out:
+        TRACE_STACK_UNWIND (unlink, frame, op_ret, op_errno,
+                            preparent, postparent, xdata);
         return 0;
 }
 
-
 int
 trace_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno, struct iatt *buf,
                   struct iatt *preoldparent, struct iatt *postoldparent,
-                  struct iatt *prenewparent, struct iatt *postnewparent)
+                  struct iatt *prenewparent, struct iatt *postnewparent,
+                  dict_t *xdata)
 {
-        char  *statstr = NULL;
-        char  *preoldparentstr = NULL;
-        char  *postoldparentstr = NULL;
-        char  *prenewparentstr = NULL;
-        char  *postnewparentstr = NULL;
-
-        if (trace_fop_names[GF_FOP_RENAME].enabled) {
-                if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preoldparentstr = trace_stat_to_str (preoldparent);
-                        postoldparentstr = trace_stat_to_str (postoldparent);
-
-                        prenewparentstr = trace_stat_to_str (prenewparent);
-                        postnewparentstr = trace_stat_to_str (postnewparent);
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, *stbuf = {%s}, "
-                                "*preoldparent = {%s}, *postoldparent = {%s}"
-                                " *prenewparent = {%s}, *postnewparent = {%s})",
-                                frame->root->unique, op_ret, statstr,
-                                preoldparentstr, postoldparentstr,
-                                prenewparentstr, postnewparentstr);
+        char           statstr[4096]          = {0, };
+        char           preoldparentstr[4096]  = {0, };
+        char           postoldparentstr[4096] = {0, };
+        char           prenewparentstr[4096]  = {0, };
+        char           postnewparentstr[4096] = {0, };
+        trace_conf_t   *conf                  = NULL;
 
-                        if (preoldparentstr)
-                                GF_FREE (preoldparentstr);
+        conf = this->private;
 
-                        if (postoldparentstr)
-                                GF_FREE (postoldparentstr);
-
-                        if (prenewparentstr)
-                                GF_FREE (prenewparentstr);
-
-                        if (postnewparentstr)
-                                GF_FREE (postnewparentstr);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_RENAME].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preoldparent, preoldparentstr);
+                        trace_stat_to_str (postoldparent, postoldparentstr);
+                        trace_stat_to_str (prenewparent, prenewparentstr);
+                        trace_stat_to_str (postnewparent, postnewparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*stbuf = {%s}, *preoldparent = {%s},"
+                                  " *postoldparent = {%s}"
+                                  " *prenewparent = {%s}, "
+                                  "*postnewparent = {%s})",
+                                  frame->root->unique, op_ret, statstr,
+                                  preoldparentstr, postoldparentstr,
+                                  prenewparentstr, postnewparentstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, op_errno);
+
                 }
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (op_ret=%d, op_errno=%d, buf {ia_ino=%"PRIu64"})",
-                        frame->root->unique, op_ret, op_errno,
-                        (buf? buf->ia_ino : 0));
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf,
-                             preoldparent, postoldparent,
-                             prenewparent, postnewparent);
+out:
+        TRACE_STACK_UNWIND (rename, frame, op_ret, op_errno, buf,
+                            preoldparent, postoldparent,
+                            prenewparent, postnewparent, xdata);
         return 0;
 }
 
-
 int
 trace_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno,
-                    const char *buf, struct iatt *stbuf)
+                    const char *buf, struct iatt *stbuf, dict_t *xdata)
 {
-        char *statstr = NULL;
+        char          statstr[4096] = {0, };
+        trace_conf_t  *conf         = NULL;
 
-        if (trace_fop_names[GF_FOP_READLINK].enabled) {
+        conf = this->private;
 
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_READLINK].enabled) {
+                char string[4096] = {0,};
                 if (op_ret == 0) {
-                        statstr = trace_stat_to_str (stbuf);
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d, buf=%s, "
-                                "stbuf = { %s })",
-                                frame->root->unique, op_ret, op_errno, buf,
-                                statstr);
-                } else
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d",
-                                frame->root->unique, op_ret, op_errno);
+                        trace_stat_to_str (stbuf, statstr);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, op_errno=%d,"
+                                  "buf=%s, stbuf = { %s })",
+                                  frame->root->unique, op_ret, op_errno,
+                                  buf, statstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+                }
 
-                if (statstr)
-                        GF_FREE (statstr);
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, buf, stbuf);
+out:
+        TRACE_STACK_UNWIND (readlink, frame, op_ret, op_errno, buf, stbuf,
+                            xdata);
         return 0;
 }
 
-
 int
 trace_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno,
                   inode_t *inode, struct iatt *buf,
-                  dict_t *xattr, struct iatt *postparent)
+                  dict_t *xdata, struct iatt *postparent)
 {
-        char  *statstr = NULL;
-        char  *postparentstr = NULL;
+        char          statstr[4096]       = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
-                if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        postparentstr = trace_stat_to_str (postparent);
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, ino=%"PRIu64", "
-                                "*buf {%s}, *postparent {%s}",
-                                frame->root->unique, op_ret, inode->ino,
-                                statstr, postparentstr);
-
-                        if (statstr)
-                                GF_FREE (statstr);
-                        if (postparentstr)
-                                GF_FREE (postparentstr);
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (postparent, postparentstr);
+                        /* print buf->ia_gfid instead of inode->gfid,
+                         * since if the inode is not yet linked to the
+                         * inode table (fresh lookup) then null gfid
+                         * will be printed.
+                         */
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s (op_ret=%d "
+                                  "*buf {%s}, *postparent {%s}",
+                                  frame->root->unique,
+                                  uuid_utoa (buf->ia_gfid),
+                                  op_ret, statstr, postparentstr);
+
+                        /* For 'forget' */
+                        inode_ctx_put (inode, this, 0);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
-                             xattr, postparent);
+out:
+        TRACE_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, buf,
+                            xdata, postparent);
         return 0;
 }
 
-
 int
 trace_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno,
                    inode_t *inode, struct iatt *buf,
-                   struct iatt *preparent, struct iatt *postparent)
+                   struct iatt *preparent, struct iatt *postparent,
+                   dict_t *xdata)
 {
-        char  *statstr = NULL;
-        char  *preparentstr = NULL;
-        char  *postparentstr = NULL;
-
-        if (trace_fop_names[GF_FOP_SYMLINK].enabled) {
-                if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preparentstr = trace_stat_to_str (preparent);
-                        postparentstr = trace_stat_to_str (postparent);
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, ino=%"PRIu64", "
-                                "*stbuf = {%s}, *preparent = {%s}, "
-                                "*postparent = {%s})",
-                                frame->root->unique, op_ret, inode->ino,
-                                statstr, preparentstr, postparentstr);
-
-                        if (statstr)
-                                GF_FREE (statstr);
-
-                        if (preparentstr)
-                                GF_FREE (preparentstr);
-
-                        if (postparentstr)
-                                GF_FREE (postparentstr);
+        conf = this->private;
 
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_SYMLINK].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s (op_ret=%d "
+                                  "*stbuf = {%s}, *preparent = {%s}, "
+                                  "*postparent = {%s})",
+                                  frame->root->unique,
+                                  uuid_utoa (inode->gfid),
+                                  op_ret, statstr, preparentstr,
+                                  postparentstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": op_ret=%d, op_errno=%d",
+                                  frame->root->unique, op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
+out:
+        TRACE_STACK_UNWIND (symlink, frame, op_ret, op_errno, inode, buf,
+                            preparent, postparent, xdata);
         return 0;
 }
 
-
 int
 trace_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno,
                  inode_t *inode, struct iatt *buf,
-                 struct iatt *preparent, struct iatt *postparent)
+                 struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-        char *statstr = NULL;
-        char *preparentstr = NULL;
-        char *postparentstr = NULL;
-
-        if (trace_fop_names[GF_FOP_MKNOD].enabled) {
-                if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preparentstr = trace_stat_to_str (preparent);
-                        postparentstr = trace_stat_to_str (postparent);
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, ino=%"PRIu64", "
-                                "*stbuf = {%s}, *preparent = {%s}, "
-                                "*postparent = {%s})",
-                                frame->root->unique, op_ret, inode->ino,
-                                statstr, preparentstr, postparentstr);
+        conf = this->private;
 
-                        if (statstr)
-                                GF_FREE (statstr);
-
-                        if (preparentstr)
-                                GF_FREE (preparentstr);
-
-                        if (postparentstr)
-                                GF_FREE (postparentstr);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        char string[4096]  = {0,};
+        if (trace_fop_names[GF_FOP_MKNOD].enabled) {
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s (op_ret=%d "
+                                  "*stbuf = {%s}, *preparent = {%s}, "
+                                  "*postparent = {%s})",
+                                  frame->root->unique,
+                                  uuid_utoa (inode->gfid),
+                                  op_ret, statstr, preparentstr,
+                                  postparentstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, op_errno=%d)",
+                                  frame->root->unique, op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
+out:
+        TRACE_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, buf,
+                            preparent, postparent, xdata);
         return 0;
 }
 
-
 int
 trace_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno,
                  inode_t *inode, struct iatt *buf,
-                 struct iatt *preparent, struct iatt *postparent)
+                 struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-        char  *statstr = NULL;
-        char  *preparentstr = NULL;
-        char  *postparentstr = NULL;
-
-        if (trace_fop_names[GF_FOP_MKDIR].enabled) {
-                if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preparentstr = trace_stat_to_str (preparent);
-                        postparentstr = trace_stat_to_str (postparent);
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, ino = %"PRIu64
-                                ", *stbuf = {%s}, *prebuf = {%s}, "
-                                "*postbuf = {%s} )",
-                                frame->root->unique, op_ret, buf->ia_ino,
-                                statstr, preparentstr, postparentstr);
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
-                        if (statstr)
-                                GF_FREE (statstr);
+        conf = this->private;
 
-                        if (preparentstr)
-                                GF_FREE (preparentstr);
-
-                        if (postparentstr)
-                                GF_FREE (postparentstr);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_MKDIR].enabled) {
+                char  string[4096]  = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s (op_ret=%d "
+                                  ", *stbuf = {%s}, *prebuf = {%s}, "
+                                  "*postbuf = {%s} )",
+                                  frame->root->unique,
+                                  uuid_utoa (inode->gfid),
+                                  op_ret, statstr, preparentstr,
+                                  postparentstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, op_errno=%d)",
+                                  frame->root->unique, op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
+out:
+        TRACE_STACK_UNWIND (mkdir, frame, op_ret, op_errno, inode, buf,
+                            preparent, postparent, xdata);
         return 0;
 }
 
-
 int
 trace_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 int32_t op_ret, int32_t op_errno,
                 inode_t *inode, struct iatt *buf,
-                struct iatt *preparent, struct iatt *postparent)
+                struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-        char  *statstr = NULL;
-        char  *preparentstr = NULL;
-        char  *postparentstr = NULL;
-
-        if (trace_fop_names[GF_FOP_LINK].enabled) {
-                if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preparentstr = trace_stat_to_str (preparent);
-                        postparentstr = trace_stat_to_str (postparent);
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, ino = %"PRIu64
-                                ", *stbuf = {%s}, *prebuf = {%s}, "
-                                "*postbuf = {%s})",
-                                frame->root->unique, op_ret, buf->ia_ino,
-                                statstr, preparentstr, postparentstr);
-
-                        if (statstr)
-                                GF_FREE (statstr);
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
-                        if (preparentstr)
-                                GF_FREE (preparentstr);
+        conf = this->private;
 
-                        if (postparentstr)
-                                GF_FREE (postparentstr);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        char  string[4096]  = {0,};
+        if (trace_fop_names[GF_FOP_LINK].enabled) {
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*stbuf = {%s},  *prebuf = {%s},"
+                                  " *postbuf = {%s})",
+                                  frame->root->unique, op_ret,
+                                  statstr, preparentstr, postparentstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
+out:
+        TRACE_STACK_UNWIND (link, frame, op_ret, op_errno, inode, buf,
+                            preparent, postparent, xdata);
         return 0;
 }
 
-
 int
 trace_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int32_t op_ret, int32_t op_errno)
+                 int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        char      string[4096] = {0,};
         if (trace_fop_names[GF_FOP_FLUSH].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (op_ret=%d, op_errno=%d)",
-                        frame->root->unique, op_ret, op_errno);
-        }
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique, uuid_utoa (frame->local),
+                          op_ret, op_errno);
 
-        STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (flush, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
-
 int
 trace_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, fd_t *fd)
+                   int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        char    string[4096] = {0,};
         if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (op_ret=%d, op_errno=%d, fd=%p)",
-                        frame->root->unique, op_ret, op_errno, fd);
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d,"
+                          " fd=%p",
+                          frame->root->unique, uuid_utoa (frame->local),
+                          op_ret, op_errno, fd);
+
+                LOG_ELEMENT (conf, string);
         }
+out:
+        /* for 'releasedir' log */
+        if (op_ret >= 0)
+                fd_ctx_set (fd, this, 0);
 
-        STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd);
+        TRACE_STACK_UNWIND (opendir, frame, op_ret, op_errno, fd, xdata);
         return 0;
 }
 
-
 int
 trace_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno,
-                 struct iatt *preparent, struct iatt *postparent)
+                 struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-        char  *preparentstr = NULL;
-        char  *postparentstr = NULL;
-
-        if (trace_fop_names[GF_FOP_RMDIR].enabled) {
-                if (op_ret >= 0) {
-                        preparentstr = trace_stat_to_str (preparent);
-                        postparentstr = trace_stat_to_str (postparent);
+        char           preparentstr[4096]  = {0, };
+        char           postparentstr[4096] = {0, };
+        trace_conf_t   *conf               = NULL;
 
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s}",
-                                frame->root->unique, op_ret, preparentstr,
-                                postparentstr);
+        conf = this->private;
 
-                        if (preparentstr)
-                                GF_FREE (preparentstr);
-
-                        if (postparentstr)
-                                GF_FREE (postparentstr);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_RMDIR].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "*prebuf={%s},  *postbuf={%s}",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, preparentstr, postparentstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno,
-                             preparent, postparent);
+out:
+        TRACE_STACK_UNWIND (rmdir, frame, op_ret, op_errno,
+                            preparent, postparent, xdata);
         return 0;
 }
 
-
 int
 trace_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno,
-                    struct iatt *prebuf, struct iatt *postbuf)
+                    struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
-        char  *preopstr = NULL;
-        char  *postopstr = NULL;
+        char           preopstr[4096]  = {0, };
+        char           postopstr[4096] = {0, };
+        trace_conf_t   *conf           = NULL;
 
-        if (trace_fop_names[GF_FOP_TRUNCATE].enabled) {
-                if (op_ret >= 0) {
-                        preopstr = trace_stat_to_str (prebuf);
-                        postopstr = trace_stat_to_str (postbuf);
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s} )",
-                                frame->root->unique, op_ret, preopstr,
-                                postopstr);
+        conf = this->private;
 
-                        if (preopstr)
-                                GF_FREE (preopstr);
-
-                        if (postopstr)
-                                GF_FREE (postopstr);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_TRUNCATE].enabled) {
+                char   string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (prebuf, preopstr);
+                        trace_stat_to_str (postbuf, postopstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s} )",
+                                  frame->root->unique, op_ret,
+                                  preopstr, postopstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, prebuf, postbuf);
+out:
+        TRACE_STACK_UNWIND (truncate, frame, op_ret, op_errno, prebuf,
+                            postbuf, xdata);
         return 0;
 }
 
-
 int
 trace_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno, struct statvfs *buf)
+                  int32_t op_ret, int32_t op_errno, struct statvfs *buf,
+                  dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_STATFS].enabled) {
-                if (op_ret >= 0) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": ({f_bsize=%lu, f_frsize=%lu, f_blocks=%"GF_PRI_FSBLK
-                                ", f_bfree=%"GF_PRI_FSBLK", f_bavail=%"GF_PRI_FSBLK", "
-                                "f_files=%"GF_PRI_FSBLK", f_ffree=%"GF_PRI_FSBLK", f_favail=%"
-                                GF_PRI_FSBLK", f_fsid=%lu, f_flag=%lu, f_namemax=%lu}) => ret=%d",
-                                frame->root->unique, buf->f_bsize, buf->f_frsize, buf->f_blocks,
-                                buf->f_bfree, buf->f_bavail, buf->f_files, buf->f_ffree,
-                                buf->f_favail, buf->f_fsid, buf->f_flag, buf->f_namemax, op_ret);
+                char   string[4096] = {0,};
+                if (op_ret == 0) {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": ({f_bsize=%lu, "
+                                  "f_frsize=%lu, "
+                                  "f_blocks=%"GF_PRI_FSBLK
+                                  ", f_bfree=%"GF_PRI_FSBLK", "
+                                  "f_bavail=%"GF_PRI_FSBLK", "
+                                  "f_files=%"GF_PRI_FSBLK", "
+                                  "f_ffree=%"GF_PRI_FSBLK", "
+                                  "f_favail=%"GF_PRI_FSBLK", "
+                                  "f_fsid=%lu, f_flag=%lu, "
+                                  "f_namemax=%lu}) => ret=%d",
+                                  frame->root->unique, buf->f_bsize,
+                                  buf->f_frsize, buf->f_blocks,
+                                  buf->f_bfree, buf->f_bavail,
+                                  buf->f_files, buf->f_ffree,
+                                  buf->f_favail, buf->f_fsid,
+                                  buf->f_flag, buf->f_namemax, op_ret);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "op_errno=%d)",
+                                  frame->root->unique, op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf);
+out:
+        TRACE_STACK_UNWIND (statfs, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
-
 int
 trace_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno)
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
+        trace_conf_t     *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_SETXATTR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (op_ret=%d, op_errno=%d)",
-                        frame->root->unique, op_ret, op_errno);
-        }
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret,
+                          op_errno);
 
-        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
-
 int
 trace_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, dict_t *dict)
+                    int32_t op_ret, int32_t op_errno, dict_t *dict,
+                    dict_t *xdata)
 {
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_GETXATTR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (op_ret=%d, op_errno=%d, dict=%p)",
-                        frame->root->unique, op_ret, op_errno, dict);
+                char       string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d,"
+                          " dict=%p", frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno,
+                          dict);
+
+                LOG_ELEMENT (conf, string);
         }
+out:
+        TRACE_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+
+        return 0;
+}
 
-        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict);
+int
+trace_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
 
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+                goto out;
+        if (trace_fop_names[GF_FOP_FSETXATTR].enabled) {
+                char    string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
+int
+trace_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *dict,
+                     dict_t *xdata)
+{
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FGETXATTR].enabled) {
+                char      string[4096]  = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d,"
+                          " dict=%p", frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno,
+                          dict);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+
+        return 0;
+}
 
 int
 trace_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno)
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_REMOVEXATTR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (op_ret=%d, op_errno=%d)",
-                        frame->root->unique, op_ret, op_errno);
-        }
+                char       string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
 
-        STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
 
         return 0;
 }
 
-
 int
 trace_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno)
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
+        trace_conf_t     *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_FSYNCDIR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (op_ret=%d, op_errno=%d)",
-                        frame->root->unique, op_ret, op_errno);
-        }
+                char       string[4096]   =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
 
-        STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fsyncdir, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
-
 int
 trace_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno)
+                  int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_ACCESS].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (op_ret=%d, op_errno=%d)",
-                        frame->root->unique, op_ret, op_errno);
-        }
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, "
+                          "op_errno=%d)", frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
 
-        STACK_UNWIND_STRICT (access, frame, op_ret, op_errno);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
-
 int
 trace_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno,
-                     struct iatt *prebuf, struct iatt *postbuf)
+                     struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
-        char  *prebufstr = NULL;
-        char  *postbufstr = NULL;
-
-        if (trace_fop_names[GF_FOP_FTRUNCATE].enabled) {
-                if (op_ret >= 0) {
-                        prebufstr = trace_stat_to_str (prebuf);
-                        postbufstr = trace_stat_to_str (postbuf);
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s} )",
-                                frame->root->unique, op_ret,
-                                prebufstr, postbufstr);
+        char          prebufstr[4096]  = {0, };
+        char          postbufstr[4096] = {0, };
+        trace_conf_t  *conf            = NULL;
 
-                        if (prebufstr)
-                                GF_FREE (prebufstr);
-
-                        if (postbufstr)
-                                GF_FREE (postbufstr);
+        conf = this->private;
 
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FTRUNCATE].enabled) {
+                char  string[4096]  = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (prebuf, prebufstr);
+                        trace_stat_to_str (postbuf, postbufstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s} )",
+                                  frame->root->unique, op_ret,
+                                  prebufstr, postbufstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno, prebuf, postbuf);
+out:
+        TRACE_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, prebuf, postbuf,
+                            xdata);
         return 0;
 }
 
-
 int
 trace_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                 int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
-        uint64_t ia_time        = 0;
-        char     atime_buf[256] = {0, };
-        char     mtime_buf[256] = {0, };
-        char     ctime_buf[256] = {0, };
+        char          statstr[4096] = {0, };
+        trace_conf_t  *conf         = NULL;
 
-        if (trace_fop_names[GF_FOP_FSTAT].enabled) {
-                if (op_ret >= 0) {
-                        ia_time = buf->ia_atime;
-                        strftime (atime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
+        conf = this->private;
 
-                        ia_time = buf->ia_mtime;
-                        strftime (mtime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        ia_time = buf->ia_ctime;
-                        strftime (ctime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime ((time_t *)&ia_time));
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, *buf {"
-                                "ia_ino=%"PRIu64", st_mode=%o, ia_nlink=%"GF_PRI_NLINK", "
-                                "ia_uid=%d, ia_gid=%d, ia_rdev=%"PRIu64", ia_size=%"PRId64", "
-                                "ia_blksize=%"GF_PRI_BLKSIZE", ia_blocks=%"PRId64", ia_atime=%s, "
-                                "ia_mtime=%s, ia_ctime=%s})",
-                                frame->root->unique, op_ret, buf->ia_ino,
-                                st_mode_from_ia (buf->ia_prot, buf->ia_type),
-                                buf->ia_nlink, buf->ia_uid, buf->ia_gid,
-                                buf->ia_rdev, buf->ia_size, buf->ia_blksize,
-                                buf->ia_blocks, atime_buf, mtime_buf, ctime_buf);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSTAT].enabled) {
+                char string[4096]  = {0.};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d "
+                                  "buf=%s", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  statstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
                 }
+                LOG_ELEMENT (conf, string);
         }
-
-        STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf);
+out:
+        TRACE_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
-
 int
 trace_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-              int32_t op_ret, int32_t op_errno, struct flock *lock)
+              int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+              dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_LK].enabled) {
-                if (op_ret >= 0) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, {l_type=%d, l_whence=%d, "
-                                "l_start=%"PRId64", l_len=%"PRId64", l_pid=%u})",
-                                frame->root->unique, op_ret, lock->l_type, lock->l_whence,
-                                lock->l_start, lock->l_len, lock->l_pid);
+                char      string[4096] = {0,};
+                if (op_ret == 0) {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "{l_type=%d, l_whence=%d, "
+                                  "l_start=%"PRId64", "
+                                  "l_len=%"PRId64", l_pid=%u})",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, lock->l_type, lock->l_whence,
+                                  lock->l_start, lock->l_len,
+                                  lock->l_pid);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
                 }
-        }
 
-        STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (lk, frame, op_ret, op_errno, lock, xdata);
         return 0;
 }
 
-
-
 int
 trace_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno)
+                   int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_ENTRYLK].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": op_ret=%d, op_errno=%d",
-                        frame->root->unique, op_ret, op_errno);
-        }
+                char   string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
 
-        STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (entrylk, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
+int
+trace_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FENTRYLK].enabled) {
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fentrylk, frame, op_ret, op_errno, xdata);
+        return 0;
+}
 
 int
 trace_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, dict_t *dict)
+                   int32_t op_ret, int32_t op_errno, dict_t *dict,
+                   dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_XATTROP].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (op_ret=%d, op_errno=%d)",
-                        frame->root->unique, op_ret, op_errno);
-        }
+                char    string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
 
-        STACK_UNWIND_STRICT (xattrop, frame, op_ret, op_errno, dict);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (xattrop, frame, op_ret, op_errno, dict, xdata);
         return 0;
 }
 
-
 int
 trace_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, dict_t *dict)
+                    int32_t op_ret, int32_t op_errno, dict_t *dict,
+                    dict_t *xdata)
 {
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_FXATTROP].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (op_ret=%d, op_errno=%d)",
-                        frame->root->unique, op_ret, op_errno);
-        }
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
 
-        STACK_UNWIND_STRICT (fxattrop, frame, op_ret, op_errno, dict);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, dict, xdata);
         return 0;
 }
 
-
 int
 trace_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno)
+                   int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_INODELK].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": op_ret=%d, op_errno=%d",
-                        frame->root->unique, op_ret, op_errno);
+                char       string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local),op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
         }
+out:
+        TRACE_STACK_UNWIND (inodelk, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+trace_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FINODELK].enabled) {
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
 
-        STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (finodelk, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
+int
+trace_rchecksum_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno,
+                     uint32_t weak_checksum, uint8_t *strong_checksum,
+                     dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+                goto out;
+        if (trace_fop_names[GF_FOP_RCHECKSUM].enabled) {
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        TRACE_STACK_UNWIND (rchecksum, frame, op_ret, op_errno, weak_checksum,
+                            strong_checksum, xdata);
+
+        return 0;
+}
+
+/* *_cbk section over <----------> fop section start */
 
 int
 trace_entrylk (call_frame_t *frame, xlator_t *this,
                const char *volume, loc_t *loc, const char *basename,
-               entrylk_cmd cmd, entrylk_type type)
+               entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_ENTRYLK].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": volume=%s, (loc= {path=%s, ino=%"PRIu64"} basename=%s, cmd=%s, type=%s)",
-                        frame->root->unique, volume, loc->path, loc->inode->ino, basename,
-                        ((cmd == ENTRYLK_LOCK) ? "ENTRYLK_LOCK" : "ENTRYLK_UNLOCK"),
-                        ((type == ENTRYLK_RDLCK) ? "ENTRYLK_RDLCK" : "ENTRYLK_WRLCK"));
+                char     string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s volume=%s, (path=%s "
+                          "basename=%s, cmd=%s, type=%s)",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid),
+                          volume, loc->path, basename,
+                          ((cmd == ENTRYLK_LOCK) ? "ENTRYLK_LOCK" :
+                           "ENTRYLK_UNLOCK"),
+                          ((type == ENTRYLK_RDLCK) ? "ENTRYLK_RDLCK" :
+                           "ENTRYLK_WRLCK"));
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_entrylk_cbk,
                     FIRST_CHILD (this),
                     FIRST_CHILD (this)->fops->entrylk,
-                    volume, loc, basename, cmd, type);
+                    volume, loc, basename, cmd, type, xdata);
         return 0;
 }
 
-
 int
 trace_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
-               loc_t *loc, int32_t cmd, struct flock *flock)
+               loc_t *loc, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
+        char         *cmd_str  = NULL;
+        char         *type_str = NULL;
+        trace_conf_t *conf     = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_INODELK].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": volume=%s, (loc {path=%s, ino=%"PRIu64"}, cmd=%s)",
-                        frame->root->unique, volume, loc->path, loc->inode->ino,
-                        ((cmd == F_SETLK)? "F_SETLK" : "unknown"));
+                char string[4096]  = {0,};
+                switch (cmd) {
+#if F_GETLK != F_GETLK64
+                case F_GETLK64:
+#endif
+                case F_GETLK:
+                        cmd_str = "GETLK";
+                        break;
+
+#if F_SETLK != F_SETLK64
+                case F_SETLK64:
+#endif
+                case F_SETLK:
+                        cmd_str = "SETLK";
+                        break;
+
+#if F_SETLKW != F_SETLKW64
+                case F_SETLKW64:
+#endif
+                case F_SETLKW:
+                        cmd_str = "SETLKW";
+                        break;
+
+                default:
+                        cmd_str = "UNKNOWN";
+                        break;
+                }
+
+                switch (flock->l_type) {
+                case F_RDLCK:
+                        type_str = "READ";
+                        break;
+                case F_WRLCK:
+                        type_str = "WRITE";
+                        break;
+                case F_UNLCK:
+                        type_str = "UNLOCK";
+                        break;
+                default:
+                        type_str = "UNKNOWN";
+                        break;
+                }
+
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s volume=%s, (path=%s "
+                          "cmd=%s, type=%s, start=%llu, len=%llu, "
+                          "pid=%llu)", frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), volume,
+                          loc->path, cmd_str, type_str,
+                          (unsigned long long)flock->l_start,
+                          (unsigned long long) flock->l_len,
+                          (unsigned long long) flock->l_pid);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_inodelk_cbk,
                     FIRST_CHILD (this),
                     FIRST_CHILD (this)->fops->inodelk,
-                    volume, loc, cmd, flock);
+                    volume, loc, cmd, flock, xdata);
         return 0;
 }
 
-
 int
-trace_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno)
+trace_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+                fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
+        char          *cmd_str  = NULL;
+        char          *type_str = NULL;
+        trace_conf_t  *conf     = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_FINODELK].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": op_ret=%d, op_errno=%d",
-                        frame->root->unique, op_ret, op_errno);
-        }
+                char  string[4096] = {0,};
+                switch (cmd) {
+#if F_GETLK != F_GETLK64
+                case F_GETLK64:
+#endif
+                case F_GETLK:
+                        cmd_str = "GETLK";
+                        break;
 
-        STACK_UNWIND_STRICT (finodelk, frame, op_ret, op_errno);
-        return 0;
-}
+#if F_SETLK != F_SETLK64
+                case F_SETLK64:
+#endif
+                case F_SETLK:
+                        cmd_str = "SETLK";
+                        break;
 
+#if F_SETLKW != F_SETLKW64
+                case F_SETLKW64:
+#endif
+                case F_SETLKW:
+                        cmd_str = "SETLKW";
+                        break;
 
-int
-trace_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
-                fd_t *fd, int32_t cmd, struct flock *flock)
-{
-        if (trace_fop_names[GF_FOP_FINODELK].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": volume=%s, (fd=%p, cmd=%s)",
-                        frame->root->unique, volume, fd,
-                        ((cmd == F_SETLK) ? "F_SETLK" : "unknown"));
-        }
+                default:
+                        cmd_str = "UNKNOWN";
+                        break;
+                }
+
+                switch (flock->l_type) {
+                case F_RDLCK:
+                        type_str = "READ";
+                        break;
+                case F_WRLCK:
+                        type_str = "WRITE";
+                        break;
+                case F_UNLCK:
+                        type_str = "UNLOCK";
+                        break;
+                default:
+                        type_str = "UNKNOWN";
+                        break;
+                }
 
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s volume=%s, (fd =%p "
+                          "cmd=%s, type=%s, start=%llu, len=%llu, "
+                          "pid=%llu)", frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), volume, fd,
+                          cmd_str, type_str,
+                          (unsigned long long) flock->l_start,
+                          (unsigned long long) flock->l_len,
+                          (unsigned long long) flock->l_pid);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
         STACK_WIND (frame, trace_finodelk_cbk,
                     FIRST_CHILD (this),
                     FIRST_CHILD (this)->fops->finodelk,
-                    volume, fd, cmd, flock);
+                    volume, fd, cmd, flock, xdata);
         return 0;
 }
 
-
 int
 trace_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
-               gf_xattrop_flags_t flags, dict_t *dict)
+               gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_XATTROP].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (path=%s, ino=%"PRIu64" flags=%d)",
-                        frame->root->unique, loc->path, loc->inode->ino, flags);
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s (path=%s flags=%d)",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          flags);
 
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_xattrop_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->xattrop,
-                    loc, flags, dict);
+                    loc, flags, dict, xdata);
 
         return 0;
 }
 
-
 int
 trace_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                gf_xattrop_flags_t flags, dict_t *dict)
+                gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
 {
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_FXATTROP].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (fd=%p, flags=%d)",
-                        frame->root->unique, fd, flags);
+                char    string[4096]  = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, flags=%d",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, flags);
+
+                frame->local = fd->inode->gfid;
 
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_fxattrop_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fxattrop,
-                    fd, flags, dict);
+                    fd, flags, dict, xdata);
 
         return 0;
 }
 
-
 int
 trace_lookup (call_frame_t *frame, xlator_t *this,
-              loc_t *loc, dict_t *xattr_req)
+              loc_t *loc, dict_t *xdata)
 {
+        trace_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
+                char string[4096] = {0,};
                 /* TODO: print all the keys mentioned in xattr_req */
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"})",
-                        frame->root->unique, loc->path,
-                        loc->inode->ino);
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_lookup_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->lookup,
-                    loc, xattr_req);
+                    loc, xdata);
 
         return 0;
 }
 
-
 int
-trace_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
+trace_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+                goto out;
         if (trace_fop_names[GF_FOP_STAT].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"})",
-                        frame->root->unique, loc->path, loc->inode->ino);
+                char  string[4096]  = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_stat_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->stat,
-                    loc);
+                    loc, xdata);
 
         return 0;
 }
 
-
 int
-trace_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size)
+trace_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+                dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_READLINK].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"}, size=%"GF_PRI_SIZET")",
-                        frame->root->unique, loc->path, loc->inode->ino, size);
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s, "
+                          "size=%"GF_PRI_SIZET")", frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          size);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_readlink_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readlink,
-                    loc, size);
+                    loc, size, xdata);
 
         return 0;
 }
 
-
 int
 trace_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
-             mode_t mode, dev_t dev, dict_t *params)
+             mode_t mode, dev_t dev, mode_t umask, dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+                goto out;
         if (trace_fop_names[GF_FOP_MKNOD].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"}, mode=%d, dev=%"GF_PRI_DEV")",
-                        frame->root->unique, loc->path, loc->inode->ino, mode, dev);
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s mode=%d "
+                          "umask=0%o, dev=%"GF_PRI_DEV")",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          mode, umask, dev);
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_mknod_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->mknod,
-                    loc, mode, dev, params);
+                    loc, mode, dev, umask, xdata);
 
         return 0;
 }
 
-
 int
 trace_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-             dict_t *params)
+             mode_t umask, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_MKDIR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (path=%s, ino=%"PRIu64", mode=%d)",
-                        frame->root->unique, loc->path,
-                        ((loc->inode)? loc->inode->ino : 0), mode);
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s mode=%d"
+                          " umask=0%o", frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          mode, umask);
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_mkdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->mkdir,
-                    loc, mode, params);
+                    loc, mode, umask, xdata);
         return 0;
 }
 
-
 int
-trace_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
+trace_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+              dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_UNLINK].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"})",
-                        frame->root->unique, loc->path, loc->inode->ino);
-        }
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s flag=%d",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          xflag);
+
+                frame->local = loc->inode->gfid;
 
+                LOG_ELEMENT (conf, string);
+        }
+out:
         STACK_WIND (frame, trace_unlink_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->unlink,
-                    loc);
+                    loc, xflag, xdata);
         return 0;
 }
 
-
 int
-trace_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
+trace_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+             dict_t *xdata)
 {
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_RMDIR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"})",
-                        frame->root->unique, loc->path, loc->inode->ino);
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s flags=%d",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          flags);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_rmdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->rmdir,
-                    loc);
+                    loc, flags, xdata);
 
         return 0;
 }
 
-
 int
 trace_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
-               loc_t *loc, dict_t *params)
+               loc_t *loc, mode_t umask, dict_t *xdata)
 {
+        trace_conf_t     *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_SYMLINK].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (linkpath=%s, loc {path=%s, ino=%"PRIu64"})",
-                        frame->root->unique, linkpath, loc->path,
-                        ((loc->inode)? loc->inode->ino : 0));
+                char     string[4096]   =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s linkpath=%s, path=%s"
+                          " umask=0%o", frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), linkpath,
+                          loc->path, umask);
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_symlink_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->symlink,
-                    linkpath, loc, params);
+                    linkpath, loc, umask, xdata);
 
         return 0;
 }
 
-
 int
-trace_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
+trace_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+              dict_t *xdata)
 {
+        char         oldgfid[50] = {0,};
+        char         newgfid[50] = {0,};
+        trace_conf_t *conf       = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_RENAME].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (oldloc {path=%s, ino=%"PRIu64"}, "
-                        "newloc{path=%s, ino=%"PRIu64"})",
-                        frame->root->unique, oldloc->path, oldloc->ino,
-                        newloc->path, newloc->ino);
+                char string[4096] = {0,};
+                if (newloc->inode)
+                        uuid_utoa_r (newloc->inode->gfid, newgfid);
+                else
+                        strcpy (newgfid, "0");
+
+                uuid_utoa_r (oldloc->inode->gfid, oldgfid);
+
+                snprintf (string, sizeof (string),
+                          "%"PRId64": oldgfid=%s oldpath=%s --> "
+                          "newgfid=%s newpath=%s",
+                          frame->root->unique, oldgfid,
+                          oldloc->path, newgfid, newloc->path);
+
+                frame->local = oldloc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_rename_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->rename,
-                    oldloc, newloc);
+                    oldloc, newloc, xdata);
 
         return 0;
 }
 
-
 int
-trace_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
+trace_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+            dict_t *xdata)
 {
+        char         oldgfid[50] = {0,};
+        char         newgfid[50] = {0,};
+        trace_conf_t *conf       = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_LINK].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (oldloc {path=%s, ino=%"PRIu64"}, "
-                        "newloc {path=%s, ino=%"PRIu64"})",
-                        frame->root->unique, oldloc->path, oldloc->inode->ino,
-                        newloc->path, newloc->inode->ino);
+                char string[4096]  = {0,};
+                if (newloc->inode)
+                        uuid_utoa_r (newloc->inode->gfid, newgfid);
+                else
+                        strcpy (newgfid, "0");
+
+                uuid_utoa_r (oldloc->inode->gfid, oldgfid);
+
+                snprintf (string, sizeof (string),
+                          "%"PRId64": oldgfid=%s oldpath=%s --> "
+                          "newgfid=%s newpath=%s", frame->root->unique,
+                          oldgfid, oldloc->path, newgfid,
+                          newloc->path);
+
+                frame->local = oldloc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_link_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->link,
-                    oldloc, newloc);
+                    oldloc, newloc, xdata);
         return 0;
 }
 
-
 int
 trace_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-               struct iatt *stbuf, int32_t valid)
+               struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-        uint64_t ia_time          = 0;
-        char     actime_str[256]  = {0,};
-        char     modtime_str[256] = {0,};
+        uint64_t     ia_time          = 0;
+        char         actime_str[256]  = {0,};
+        char         modtime_str[256] = {0,};
+        trace_conf_t *conf            = NULL;
+
+        conf = this->private;
 
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_SETATTR].enabled) {
+                char     string[4096]  =  {0,};
                 if (valid & GF_SET_ATTR_MODE) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (loc {path=%s, ino=%"PRIu64"},"
-                                " mode=%o)", frame->root->unique, loc->path,
-                                loc->inode->ino,
-                                st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type));
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s path=%s mode=%o)",
+                                  frame->root->unique,
+                                  uuid_utoa (loc->inode->gfid),
+                                  loc->path,
+                                  st_mode_from_ia (stbuf->ia_prot,
+                                                   stbuf->ia_type));
+
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0 , sizeof (string));
                 }
 
                 if (valid & (GF_SET_ATTR_UID | GF_SET_ATTR_GID)) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (loc {path=%s, ino=%"PRIu64"},"
-                                " uid=%o, gid=%o)",
-                                frame->root->unique, loc->path, loc->inode->ino,
-                                stbuf->ia_uid, stbuf->ia_gid);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s path=%s uid=%o,"
+                                  " gid=%o", frame->root->unique,
+                                  uuid_utoa (loc->inode->gfid),
+                                  loc->path, stbuf->ia_uid,
+                                  stbuf->ia_gid);
+
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0 , sizeof (string));
                 }
 
                 if (valid & (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME)) {
@@ -1540,44 +2039,64 @@ trace_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
                         strftime (modtime_str, 256, "[%b %d %H:%M:%S]",
                                   localtime ((time_t *)&ia_time));
 
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (loc {path=%s, ino=%"PRIu64"}, "
-                                "*stbuf=%p {ia_atime=%s, ia_mtime=%s})",
-                                frame->root->unique, loc->path, loc->inode->ino,
-                                stbuf, actime_str, modtime_str);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s path=%s "
+                                  "ia_atime=%s, ia_mtime=%s",
+                                  frame->root->unique,
+                                  uuid_utoa (loc->inode->gfid),
+                                  loc->path, actime_str, modtime_str);
+
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0 , sizeof (string));
                 }
+                frame->local = loc->inode->gfid;
         }
 
+out:
         STACK_WIND (frame, trace_setattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->setattr,
-                    loc, stbuf, valid);
+                    loc, stbuf, valid, xdata);
 
         return 0;
 }
 
-
 int
 trace_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                struct iatt *stbuf, int32_t valid)
+                struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-        uint64_t ia_time          = 0;
-        char     actime_str[256]  = {0,};
-        char     modtime_str[256] = {0,};
+        uint64_t       ia_time          = 0;
+        char           actime_str[256]  = {0,};
+        char           modtime_str[256] = {0,};
+        trace_conf_t   *conf            = NULL;
+
+        conf = this->private;
 
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
+                char     string[4096]  =  {0,};
                 if (valid & GF_SET_ATTR_MODE) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (*fd=%p, mode=%o)",
-                                frame->root->unique, fd,
-                                st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type));
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s fd=%p, mode=%o",
+                                  frame->root->unique,
+                                  uuid_utoa (fd->inode->gfid), fd,
+                                  st_mode_from_ia (stbuf->ia_prot,
+                                                   stbuf->ia_type));
+
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0, sizeof (string));
                 }
 
                 if (valid & (GF_SET_ATTR_UID | GF_SET_ATTR_GID)) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (*fd=%p, uid=%o, gid=%o)",
-                                frame->root->unique, fd,
-                                stbuf->ia_uid, stbuf->ia_gid);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s fd=%p, uid=%o, "
+                                  "gid=%o", frame->root->unique,
+                                  uuid_utoa (fd->inode->gfid),
+                                  fd, stbuf->ia_uid, stbuf->ia_gid);
+
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0, sizeof (string));
                 }
 
                 if (valid & (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME)) {
@@ -1589,372 +2108,816 @@ trace_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
                         strftime (modtime_str, 256, "[%b %d %H:%M:%S]",
                                   localtime ((time_t *)&ia_time));
 
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (*fd=%p"
-                                "*stbuf=%p {ia_atime=%s, ia_mtime=%s})",
-                                frame->root->unique, fd, stbuf, actime_str,
-                                modtime_str);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s fd=%p "
+                                  "ia_atime=%s, ia_mtime=%s",
+                                  frame->root->unique,
+                                  uuid_utoa (fd->inode->gfid),
+                                  fd, actime_str, modtime_str);
+
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0, sizeof (string));
                 }
+                frame->local = fd->inode->gfid;
         }
 
+out:
         STACK_WIND (frame, trace_fsetattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fsetattr,
-                    fd, stbuf, valid);
+                    fd, stbuf, valid, xdata);
 
         return 0;
 }
 
-
 int
 trace_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                off_t offset)
+                off_t offset, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_TRUNCATE].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"}, offset=%"PRId64")",
-                        frame->root->unique, loc->path, loc->inode->ino, offset);
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s, "
+                          "offset=%"PRId64"", frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          offset);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_truncate_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->truncate,
-                    loc, offset);
+                    loc, offset, xdata);
 
         return 0;
 }
 
-
 int
 trace_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
-            int32_t flags, fd_t *fd, int32_t wbflags)
+            int32_t flags, fd_t *fd, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_OPEN].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"}, flags=%d, "
-                        "fd=%p, wbflags=%d)",
-                        frame->root->unique, loc->path, loc->inode->ino, flags,
-                        fd, wbflags);
+                char      string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s flags=%d fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          flags, fd);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_open_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->open,
-                    loc, flags, fd, wbflags);
+                    loc, flags, fd, xdata);
         return 0;
 }
 
-
 int
 trace_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
-              int32_t flags, mode_t mode, fd_t *fd, dict_t *params)
+              int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
+              dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_CREATE].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"}, flags=0%o mode=0%o)",
-                        frame->root->unique, loc->path, loc->inode->ino, flags, mode);
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s, fd=%p, "
+                          "flags=0%o mode=0%o umask=0%o",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          fd, flags, mode, umask);
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_create_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->create,
-                    loc, flags, mode, fd, params);
+                    loc, flags, mode, umask, fd, xdata);
         return 0;
 }
 
-
 int
 trace_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
-             size_t size, off_t offset)
+             size_t size, off_t offset, uint32_t flags, dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_READ].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (*fd=%p, size=%"GF_PRI_SIZET", offset=%"PRId64")",
-                        frame->root->unique, fd, size, offset);
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, size=%"
+                          GF_PRI_SIZET"offset=%"PRId64" flags=0%x)",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, size,
+                          offset, flags);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_readv_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readv,
-                    fd, size, offset);
+                    fd, size, offset, flags, xdata);
         return 0;
 }
 
-
 int
 trace_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
               struct iovec *vector, int32_t count,
-              off_t offset, struct iobref *iobref)
+              off_t offset, uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_WRITE].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (*fd=%p, *vector=%p, count=%d, offset=%"PRId64")",
-                        frame->root->unique, fd, vector, count, offset);
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, count=%d, "
+                          " offset=%"PRId64" flags=0%x)",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, count,
+                          offset, flags);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_writev_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->writev,
-                    fd, vector, count, offset, iobref);
+                    fd, vector, count, offset, flags, iobref, xdata);
         return 0;
 }
 
-
 int
-trace_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
+trace_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_STATFS].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"})",
-                        frame->root->unique, loc->path,
-                        ((loc->inode)? loc->inode->ino : 0));
+                char  string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s",
+                          frame->root->unique, (loc->inode)?
+                          uuid_utoa (loc->inode->gfid):"0", loc->path);
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_statfs_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->statfs,
-                    loc);
+                    loc, xdata);
         return 0;
 }
 
-
 int
-trace_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
+trace_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_FLUSH].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (*fd=%p)",
-                        frame->root->unique, fd);
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_flush_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->flush,
-                    fd);
+                    fd, xdata);
         return 0;
 }
 
-
 int
-trace_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
+trace_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+             dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_FSYNC].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (flags=%d, *fd=%p)",
-                        frame->root->unique, flags, fd);
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s flags=%d fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), flags, fd);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_fsync_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fsync,
-                    fd, flags);
+                    fd, flags, xdata);
         return 0;
 }
 
-
 int
 trace_setxattr (call_frame_t *frame, xlator_t *this,
-                loc_t *loc, dict_t *dict, int32_t flags)
+                loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata)
 {
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_SETXATTR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"}, dict=%p, flags=%d)",
-                        frame->root->unique, loc->path,
-                        ((loc->inode)? loc->inode->ino : 0), dict, flags);
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s flags=%d",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          flags);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_setxattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->setxattr,
-                    loc, dict, flags);
+                    loc, dict, flags, xdata);
         return 0;
 }
 
-
 int
 trace_getxattr (call_frame_t *frame, xlator_t *this,
-                loc_t *loc, const char *name)
+                loc_t *loc, const char *name, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_GETXATTR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"}), name=%s",
-                        frame->root->unique, loc->path,
-                        ((loc->inode)? loc->inode->ino : 0), name);
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s name=%s",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          name);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_getxattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->getxattr,
-                    loc, name);
+                    loc, name, xdata);
         return 0;
 }
 
-
 int
 trace_removexattr (call_frame_t *frame, xlator_t *this,
-                   loc_t *loc, const char *name)
+                   loc_t *loc, const char *name, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_REMOVEXATTR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (loc {path=%s, ino=%"PRIu64"}, name=%s)",
-                        frame->root->unique, loc->path,
-                        ((loc->inode)? loc->inode->ino : 0), name);
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s name=%s",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          name);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_removexattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->removexattr,
-                    loc, name);
+                    loc, name, xdata);
 
         return 0;
 }
 
-
 int
-trace_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
+trace_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+               dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64":( loc {path=%s, ino=%"PRIu64"}, fd=%p)",
-                        frame->root->unique, loc->path, loc->inode->ino, fd);
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path, fd);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_opendir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->opendir,
-                    loc, fd);
+                    loc, fd, xdata);
         return 0;
 }
 
 int
 trace_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-                off_t offset)
+                off_t offset, dict_t *dict)
 {
+        trace_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_READDIRP].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (fd=%p, size=%"GF_PRI_SIZET", offset=%"PRId64")",
-                        frame->root->unique, fd, size, offset);
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, size=%"GF_PRI_SIZET
+                          ", offset=%"PRId64" dict=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, size,
+                          offset, dict);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_readdirp_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readdirp,
-                    fd, size, offset);
+                    fd, size, offset, dict);
 
         return 0;
 }
 
-
 int
 trace_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
-               size_t size, off_t offset)
+               size_t size, off_t offset, dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_READDIR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (fd=%p, size=%"GF_PRI_SIZET", offset=%"PRId64")",
-                        frame->root->unique, fd, size, offset);
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, size=%"GF_PRI_SIZET
+                          ", offset=%"PRId64,
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, size,
+                          offset);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_readdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readdir,
-                    fd, size, offset);
+                    fd, size, offset, xdata);
 
         return 0;
 }
 
-
 int
 trace_fsyncdir (call_frame_t *frame, xlator_t *this,
-                fd_t *fd, int32_t datasync)
+                fd_t *fd, int32_t datasync, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_FSYNCDIR].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (datasync=%d, *fd=%p)",
-                        frame->root->unique, datasync, fd);
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s datasync=%d fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), datasync, fd);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_fsyncdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fsyncdir,
-                    fd, datasync);
+                    fd, datasync, xdata);
         return 0;
 }
 
-
 int
-trace_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask)
+trace_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+              dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_ACCESS].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (*loc {path=%s, ino=%"PRIu64"}, mask=0%o)",
-                        frame->root->unique, loc->path,
-                        ((loc->inode)? loc->inode->ino : 0), mask);
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s mask=0%o",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid),
+                          loc->path, mask);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_access_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->access,
-                    loc, mask);
+                    loc, mask, xdata);
         return 0;
 }
 
+int32_t
+trace_rchecksum (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+                 int32_t len, dict_t *xdata)
+{
+
+        trace_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_RCHECKSUM].enabled) {
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s offset=%"PRId64
+                          "len=%u fd=%p", frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), offset, len, fd);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_rchecksum_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rchecksum,
+                    fd, offset, len, xdata);
+
+        return 0;
+
+}
+
+int32_t
+trace_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume,
+                fd_t *fd, const char *basename, entrylk_cmd cmd,
+                entrylk_type type, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FENTRYLK].enabled) {
+                char      string[4096]   =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s volume=%s, (fd=%p "
+                          "basename=%s, cmd=%s, type=%s)",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), volume, fd,
+                          basename,
+                          ((cmd == ENTRYLK_LOCK) ? "ENTRYLK_LOCK" :
+                           "ENTRYLK_UNLOCK"),
+                          ((type == ENTRYLK_RDLCK) ? "ENTRYLK_RDLCK" :
+                           "ENTRYLK_WRLCK"));
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_fentrylk_cbk,
+                    FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->fentrylk,
+                    volume, fd, basename, cmd, type, xdata);
+        return 0;
+
+}
+
+int32_t
+trace_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 const char *name, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+                goto out;
+        if (trace_fop_names[GF_FOP_FGETXATTR].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p name=%s",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, name);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_fgetxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fgetxattr,
+                    fd, name, xdata);
+        return 0;
+}
+
+int32_t
+trace_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSETXATTR].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p flags=%d",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, flags);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_fsetxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr,
+                    fd, dict, flags, xdata);
+        return 0;
+}
 
 int
 trace_ftruncate (call_frame_t *frame, xlator_t *this,
-                 fd_t *fd, off_t offset)
+                 fd_t *fd, off_t offset, dict_t *xdata)
 {
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_FTRUNCATE].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (offset=%"PRId64", *fd=%p)",
-                        frame->root->unique, offset, fd);
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s offset=%"PRId64" fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), offset, fd);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_ftruncate_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->ftruncate,
-                    fd, offset);
+                    fd, offset, xdata);
 
         return 0;
 }
 
-
 int
-trace_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd)
+trace_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_FSTAT].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (*fd=%p)",
-                        frame->root->unique, fd);
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_fstat_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fstat,
-                    fd);
+                    fd, xdata);
         return 0;
 }
 
-
 int
 trace_lk (call_frame_t *frame, xlator_t *this, fd_t *fd,
-          int32_t cmd, struct flock *lock)
+          int32_t cmd, struct gf_flock *lock, dict_t *xdata)
 {
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
         if (trace_fop_names[GF_FOP_LK].enabled) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "%"PRId64": (*fd=%p, cmd=%d, lock {l_type=%d, l_whence=%d, "
-                        "l_start=%"PRId64", l_len=%"PRId64", l_pid=%u})",
-                        frame->root->unique, fd, cmd, lock->l_type, lock->l_whence,
-                        lock->l_start, lock->l_len, lock->l_pid);
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, cmd=%d, "
+                          "lock {l_type=%d, "
+                          "l_whence=%d, l_start=%"PRId64", "
+                          "l_len=%"PRId64", l_pid=%u})",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, cmd,
+                          lock->l_type, lock->l_whence,
+                          lock->l_start, lock->l_len, lock->l_pid);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
         }
 
+out:
         STACK_WIND (frame, trace_lk_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->lk,
-                    fd, cmd, lock);
+                    fd, cmd, lock, xdata);
         return 0;
 }
 
+int32_t
+trace_forget (xlator_t *this, inode_t *inode)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+        /* If user want to understand when a lookup happens,
+           he should know about 'forget' too */
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "gfid=%s", uuid_utoa (inode->gfid));
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        return 0;
+}
+
+int32_t
+trace_releasedir (xlator_t *this, fd_t *fd)
+{
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "gfid=%s fd=%p",
+                          uuid_utoa (fd->inode->gfid), fd);
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        return 0;
+}
+
+int32_t
+trace_release (xlator_t *this, fd_t *fd)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_OPEN].enabled ||
+            trace_fop_names[GF_FOP_CREATE].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "gfid=%s fd=%p",
+                          uuid_utoa (fd->inode->gfid), fd);
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        return 0;
+}
 
 
 void
@@ -1966,7 +2929,6 @@ enable_all_calls (int enabled)
                 trace_fop_names[i].enabled = enabled;
 }
 
-
 void
 enable_call (const char *name, int enabled)
 {
@@ -1994,6 +2956,105 @@ process_call_list (const char *list, int include)
         }
 }
 
+int32_t
+trace_dump_history (xlator_t *this)
+{
+        int ret = -1;
+        char key_prefix[GF_DUMP_MAX_BUF_LEN] = {0,};
+        trace_conf_t *conf = NULL;
+
+        GF_VALIDATE_OR_GOTO ("trace", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, this->history, out);
+
+        conf = this->private;
+        // Is it ok to return silently if log-history option his off?
+        if (conf && conf->log_history == _gf_true) {
+                gf_proc_dump_build_key (key_prefix, "xlator.debug.trace",
+                                        "history");
+                gf_proc_dump_add_section (key_prefix);
+                eh_dump (this->history, NULL, dump_history_trace);
+        }
+        ret = 0;
+
+out:
+        return ret;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_trace_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                        " failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+int
+reconfigure (xlator_t *this, dict_t *options)
+{
+        int32_t       ret            = -1;
+        trace_conf_t    *conf           = NULL;
+        char            *includes = NULL, *excludes = NULL;
+
+        GF_VALIDATE_OR_GOTO ("quick-read", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, this->private, out);
+        GF_VALIDATE_OR_GOTO (this->name, options, out);
+
+        conf = this->private;
+
+        includes = data_to_str (dict_get (options, "include-ops"));
+        excludes = data_to_str (dict_get (options, "exclude-ops"));
+
+        {
+                int i;
+                for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+                        if (gf_fop_list[i])
+                                strncpy (trace_fop_names[i].name,
+                                         gf_fop_list[i],
+                                         strlen (gf_fop_list[i]));
+                        else
+                                strncpy (trace_fop_names[i].name, ":O",
+                                         strlen (":O"));
+                        trace_fop_names[i].enabled = 1;
+                }
+        }
+
+        if (includes && excludes) {
+                gf_log (this->name,
+                        GF_LOG_ERROR,
+                        "must specify only one of 'include-ops' and "
+                        "'exclude-ops'");
+                goto out;
+        }
+
+        if (includes)
+                process_call_list (includes, 1);
+        if (excludes)
+                process_call_list (excludes, 0);
+
+        /* Should resizing of the event-history be allowed in reconfigure?
+         * for which a new event_history might have to be allocated and the
+         * older history has to be freed.
+         */
+        GF_OPTION_RECONF ("log-file", conf->log_file, options, bool, out);
+
+        GF_OPTION_RECONF ("log-history", conf->log_history, options, bool, out);
+
+        ret = 0;
+
+out:
+        return ret;
+}
 
 int32_t
 init (xlator_t *this)
@@ -2001,6 +3062,10 @@ init (xlator_t *this)
         dict_t *options = NULL;
         char *includes = NULL, *excludes = NULL;
         char *forced_loglevel = NULL;
+        eh_t *history = NULL;
+        int  ret = -1;
+        size_t  history_size = TRACE_DEFAULT_HISTORY_SIZE;
+        trace_conf_t    *conf = NULL;
 
         if (!this)
                 return -1;
@@ -2015,6 +3080,12 @@ init (xlator_t *this)
                         "dangling volume. check volfile ");
         }
 
+        conf = GF_CALLOC (1, sizeof (trace_conf_t), gf_trace_mt_trace_conf_t);
+        if (!conf) {
+                gf_log (this->name, GF_LOG_ERROR, "cannot allocate "
+                        "xl->private");
+                return -1;
+        }
 
         options = this->options;
         includes = data_to_str (dict_get (options, "include-ops"));
@@ -2023,8 +3094,13 @@ init (xlator_t *this)
         {
                 int i;
                 for (i = 0; i < GF_FOP_MAXVALUE; i++) {
-                        trace_fop_names[i].name = (gf_fop_list[i] ?
-                                                   gf_fop_list[i] : ":O");
+                        if (gf_fop_list[i])
+                                strncpy (trace_fop_names[i].name,
+                                         gf_fop_list[i],
+                                         strlen (gf_fop_list[i]));
+                        else
+                                strncpy (trace_fop_names[i].name, ":O",
+                                         strlen (":O"));
                         trace_fop_names[i].enabled = 1;
                 }
         }
@@ -2032,40 +3108,78 @@ init (xlator_t *this)
         if (includes && excludes) {
                 gf_log (this->name,
                         GF_LOG_ERROR,
-                        "must specify only one of 'include-ops' and 'exclude-ops'");
+                        "must specify only one of 'include-ops' and "
+                        "'exclude-ops'");
                 return -1;
         }
+
         if (includes)
                 process_call_list (includes, 1);
         if (excludes)
                 process_call_list (excludes, 0);
 
+
+        GF_OPTION_INIT ("history-size", conf->history_size, size, out);
+
+        gf_log (this->name, GF_LOG_INFO, "history size %"GF_PRI_SIZET,
+                history_size);
+
+        GF_OPTION_INIT ("log-file", conf->log_file, bool, out);
+
+        gf_log (this->name, GF_LOG_INFO, "logging to file %s",
+                (conf->log_file == _gf_true)?"enabled":"disabled");
+
+        GF_OPTION_INIT ("log-history", conf->log_history, bool, out);
+
+        gf_log (this->name, GF_LOG_DEBUG, "logging to history %s",
+                (conf->log_history == _gf_true)?"enabled":"disabled");
+
+        history = eh_new (history_size, _gf_false, NULL);
+        if (!history) {
+                gf_log (this->name, GF_LOG_ERROR, "event history cannot be "
+                        "initialized");
+                return -1;
+        }
+
+        this->history = history;
+
+        conf->trace_log_level = GF_LOG_INFO;
+
         if (dict_get (options, "force-log-level")) {
                 forced_loglevel = data_to_str (dict_get (options,
-                                               "force-log-level"));
+                                                         "force-log-level"));
                 if (!forced_loglevel)
                         goto setloglevel;
 
-                if (strcmp (forced_loglevel, "NORMAL") == 0)
-                        trace_log_level = GF_LOG_NORMAL;
+                if (strcmp (forced_loglevel, "INFO") == 0)
+                        conf->trace_log_level = GF_LOG_INFO;
                 else if (strcmp (forced_loglevel, "TRACE") == 0)
-                        trace_log_level = GF_LOG_TRACE;
+                        conf->trace_log_level = GF_LOG_TRACE;
                 else if (strcmp (forced_loglevel, "ERROR") == 0)
-                        trace_log_level = GF_LOG_ERROR;
+                        conf->trace_log_level = GF_LOG_ERROR;
                 else if (strcmp (forced_loglevel, "DEBUG") == 0)
-                        trace_log_level = GF_LOG_DEBUG;
+                        conf->trace_log_level = GF_LOG_DEBUG;
                 else if (strcmp (forced_loglevel, "WARNING") == 0)
-                        trace_log_level = GF_LOG_WARNING;
+                        conf->trace_log_level = GF_LOG_WARNING;
                 else if (strcmp (forced_loglevel, "CRITICAL") == 0)
-                        trace_log_level = GF_LOG_CRITICAL;
+                        conf->trace_log_level = GF_LOG_CRITICAL;
                 else if (strcmp (forced_loglevel, "NONE") == 0)
-                        trace_log_level = GF_LOG_NONE;
+                        conf->trace_log_level = GF_LOG_NONE;
         }
 
 setloglevel:
-        gf_log_set_loglevel (trace_log_level);
+        gf_log_set_loglevel (conf->trace_log_level);
+        this->private = conf;
+        ret = 0;
+out:
+        if (ret == -1) {
+                if (history)
+                        GF_FREE (history);
+                if (conf)
+                        GF_FREE (conf);
+        }
 
-        return 0;
+        return ret;
 }
 
 void
@@ -2074,7 +3188,10 @@ fini (xlator_t *this)
         if (!this)
                 return;
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        if (this->history)
+                eh_destroy (this->history);
+
+        gf_log (this->name, GF_LOG_INFO,
                 "trace translator unloaded");
         return;
 }
@@ -2098,6 +3215,8 @@ struct xlator_fops fops = {
         .fsync       = trace_fsync,
         .setxattr    = trace_setxattr,
         .getxattr    = trace_getxattr,
+        .fsetxattr   = trace_fsetxattr,
+        .fgetxattr   = trace_fgetxattr,
         .removexattr = trace_removexattr,
         .opendir     = trace_opendir,
         .readdir     = trace_readdir,
@@ -2111,15 +3230,19 @@ struct xlator_fops fops = {
         .inodelk     = trace_inodelk,
         .finodelk    = trace_finodelk,
         .entrylk     = trace_entrylk,
+        .fentrylk    = trace_fentrylk,
         .lookup      = trace_lookup,
+        .rchecksum   = trace_rchecksum,
         .xattrop     = trace_xattrop,
         .fxattrop    = trace_fxattrop,
         .setattr     = trace_setattr,
         .fsetattr    = trace_fsetattr,
 };
 
-
 struct xlator_cbks cbks = {
+        .release     = trace_release,
+        .releasedir  = trace_releasedir,
+        .forget      = trace_forget,
 };
 
 struct volume_options options[] = {
@@ -2131,5 +3254,21 @@ struct volume_options options[] = {
           .type = GF_OPTION_TYPE_STR
           /*.value = { ""} */
         },
+        { .key  = {"history-size"},
+          .type = GF_OPTION_TYPE_SIZET,
+          .default_value = "1024",
+        },
+        { .key  = {"log-file"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "no",
+        },
+        { .key  = {"log-history"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "no",
+        },
         { .key  = {NULL} },
 };
+
+struct xlator_dumpops dumpops = {
+        .history = trace_dump_history
+};
diff --git a/xlators/debug/trace/src/trace.h b/xlators/debug/trace/src/trace.h
new file mode 100644
index 000000000..045eefb36
--- /dev/null
+++ b/xlators/debug/trace/src/trace.h
@@ -0,0 +1,98 @@
+/*
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include <time.h>
+#include <errno.h>
+#include "glusterfs.h"
+#include "xlator.h"
+#include "common-utils.h"
+#include "event-history.h"
+#include "logging.h"
+#include "circ-buff.h"
+#include "statedump.h"
+#include "options.h"
+
+#define TRACE_DEFAULT_HISTORY_SIZE 1024
+
+typedef struct {
+        /* Since the longest fop name is fremovexattr i.e 12 characters, array size
+         * is kept 24, i.e double of the maximum.
+         */
+        char name[24];
+        int enabled;
+} trace_fop_name_t;
+
+trace_fop_name_t trace_fop_names[GF_FOP_MAXVALUE];
+
+typedef struct {
+        gf_boolean_t log_file;
+        gf_boolean_t log_history;
+        size_t       history_size;
+        int trace_log_level;
+} trace_conf_t;
+
+#define TRACE_STACK_UNWIND(op, frame, params ...)                       \
+        do {                                                            \
+                frame->local = NULL;                                    \
+                STACK_UNWIND_STRICT (op, frame, params);                \
+        } while (0);
+
+#define LOG_ELEMENT(_conf, _string)                                     \
+        do {                                                            \
+                if (_conf) {                                            \
+                        if ((_conf->log_history) == _gf_true)           \
+                                gf_log_eh ("%s", _string);              \
+                        if ((_conf->log_file) == _gf_true)              \
+                                gf_log (THIS->name, _conf->trace_log_level, \
+                                        "%s", _string);                 \
+                }                                                       \
+        } while (0);
+
+#define trace_stat_to_str(buf, statstr)                                 \
+        do {                                                            \
+                char     atime_buf[256]    = {0,};                      \
+                char     mtime_buf[256]    = {0,};                      \
+                char     ctime_buf[256]    = {0,};                      \
+                uint64_t ia_time           = 0;                         \
+                                                                        \
+                if (!buf)                                               \
+                        break;                                          \
+                                                                        \
+                ia_time = buf->ia_atime;                                \
+                strftime (atime_buf, 256, "[%b %d %H:%M:%S]",           \
+                          localtime ((time_t *)&ia_time));              \
+                                                                        \
+                ia_time = buf->ia_mtime;                                \
+                strftime (mtime_buf, 256, "[%b %d %H:%M:%S]",           \
+                          localtime ((time_t *)&ia_time));              \
+                                                                        \
+                ia_time = buf->ia_ctime;                                \
+                strftime (ctime_buf, 256, "[%b %d %H:%M:%S]",           \
+                          localtime ((time_t *)&ia_time));              \
+                                                                        \
+                snprintf (statstr, sizeof (statstr),                    \
+                          "gfid=%s ino=%"PRIu64", mode=%o, "            \
+                          "nlink=%"GF_PRI_NLINK", uid=%u, "             \
+                          "gid=%u, size=%"PRIu64", "                    \
+                          "blocks=%"PRIu64", atime=%s, "                \
+                          "mtime=%s, ctime=%s",                         \
+                          uuid_utoa (buf->ia_gfid), buf->ia_ino,        \
+                          st_mode_from_ia (buf->ia_prot,                \
+                                           buf->ia_type),               \
+                          buf->ia_nlink, buf->ia_uid,                   \
+                          buf->ia_gid, buf->ia_size,                    \
+                          buf->ia_blocks, atime_buf,                    \
+                          mtime_buf, ctime_buf);                        \
+        } while (0);
diff --git a/xlators/encryption/Makefile.am b/xlators/encryption/Makefile.am
index 2cbde680f..36efc6698 100644
--- a/xlators/encryption/Makefile.am
+++ b/xlators/encryption/Makefile.am
@@ -1,3 +1,3 @@
-SUBDIRS = rot-13
+SUBDIRS = rot-13 crypt
 
 CLEANFILES = 
diff --git a/xlators/cluster/unify/Makefile.am b/xlators/encryption/crypt/Makefile.am
index d471a3f92..d471a3f92 100644
--- a/xlators/cluster/unify/Makefile.am
+++ b/xlators/encryption/crypt/Makefile.am
diff --git a/xlators/encryption/crypt/src/Makefile.am b/xlators/encryption/crypt/src/Makefile.am
new file mode 100644
index 000000000..faadd117f
--- /dev/null
+++ b/xlators/encryption/crypt/src/Makefile.am
@@ -0,0 +1,24 @@
+if ENABLE_CRYPT_XLATOR
+
+xlator_LTLIBRARIES = crypt.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/encryption
+
+crypt_la_LDFLAGS = -module -avoidversion -lssl -lcrypto
+
+crypt_la_SOURCES = keys.c data.c metadata.c atom.c crypt.c
+crypt_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+noinst_HEADERS = crypt-common.h crypt-mem-types.h crypt.h metadata.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
+
+else
+
+noinst_DIST = keys.c data.c metadata.c atom.c crypt.c
+noinst_HEADERS = crypt-common.h crypt-mem-types.h crypt.h metadata.h
+
+endif
+\ No newline at end of file
diff --git a/xlators/encryption/crypt/src/atom.c b/xlators/encryption/crypt/src/atom.c
new file mode 100644
index 000000000..1ec41495c
--- /dev/null
+++ b/xlators/encryption/crypt/src/atom.c
@@ -0,0 +1,962 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "defaults.h"
+#include "crypt-common.h"
+#include "crypt.h"
+
+/*
+ *                              Glossary
+ *
+ *
+ * cblock            (or cipher block). A logical unit in a file.
+ *                   cblock size is defined as the number of bits
+ *                   in an input (or output) block of the block
+ *                   cipher (*). Cipher block size is a property of
+ *                   cipher algorithm. E.g. cblock size is 64 bits
+ *                   for DES, 128 bits for AES, etc.
+ *
+ * atomic cipher     A cipher algorithm, which requires some chunks of
+ * algorithm         text to be padded at left and(or) right sides before
+ *                   cipher transaform.
+ *
+ *
+ * block (atom)      Minimal chunk of file's data, which doesn't require
+ *                   padding. We'll consider logical units in a file of
+ *                   block size (atom size).
+ *
+ * cipher algorithm  Atomic cipher algorithm, which requires the last
+ * with EOF issue    incomplete cblock in a file to be padded with some
+ *                   data (usually zeros).
+ *
+ *
+ * operation, which  reading/writing from offset, which is not aligned to
+ * forms a gap at    to atom size
+ * the beginning
+ *
+ *
+ * operation, which  reading/writing count bytes starting from offset off,
+ * forms a gap at    so that off+count is not aligned to atom_size
+ * the end
+ *
+ * head block        the first atom affected by an operation, which forms
+ *                   a gap at the beginning, or(and) at the end.
+ *                   Сomment. Head block has at least one gap (either at
+ *                   the beginning, or at the end)
+ *
+ *
+ * tail block        the last atom different from head, affected by an
+ *                   operation, which forms a gap at the end.
+ *                   Сomment: Tail block has exactly one gap (at the end).
+ *
+ *
+ * partial block     head or tail block
+ *
+ *
+ * full block        block without gaps.
+ *
+ *
+ * (*) Recommendation for Block Cipher Modes of Operation
+ *     Methods and Techniques
+ *     NIST Special Publication 800-38A Edition 2001
+ */
+
+/*
+ * atom->offset_at()
+ */
+static off_t offset_at_head(struct avec_config *conf)
+{
+	return conf->aligned_offset;
+}
+
+static off_t offset_at_hole_head(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_head(get_hole_conf(frame));
+}
+
+static off_t offset_at_data_head(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_head(get_data_conf(frame));
+}
+
+
+static off_t offset_at_tail(struct avec_config *conf,
+				    struct object_cipher_info *object)
+{
+	return conf->aligned_offset +
+		(conf->off_in_head ? get_atom_size(object) : 0) +
+		(conf->nr_full_blocks << get_atom_bits(object));
+}
+
+static off_t offset_at_hole_tail(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_tail(get_hole_conf(frame), object);
+}
+
+
+static off_t offset_at_data_tail(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_tail(get_data_conf(frame), object);
+}
+
+static off_t offset_at_full(struct avec_config *conf,
+				    struct object_cipher_info *object)
+{
+	return conf->aligned_offset +
+		(conf->off_in_head ? get_atom_size(object) : 0);
+}
+
+static off_t offset_at_data_full(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_full(get_data_conf(frame), object);
+}
+
+static off_t offset_at_hole_full(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_full(get_hole_conf(frame), object);
+}
+
+/*
+ * atom->io_size_nopad()
+ */
+
+static uint32_t io_size_nopad_head(struct avec_config *conf,
+				   struct object_cipher_info *object)
+{
+	uint32_t gap_at_beg;
+	uint32_t gap_at_end;
+
+	check_head_block(conf);
+
+	gap_at_beg = conf->off_in_head;
+
+	if (has_tail_block(conf) || has_full_blocks(conf) || conf->off_in_tail == 0 )
+		gap_at_end = 0;
+	else
+		gap_at_end = get_atom_size(object) - conf->off_in_tail;
+
+	return get_atom_size(object) - (gap_at_beg + gap_at_end);
+}
+
+static uint32_t io_size_nopad_tail(struct avec_config *conf,
+				   struct object_cipher_info *object)
+{
+	check_tail_block(conf);
+	return conf->off_in_tail;
+}
+
+static uint32_t io_size_nopad_full(struct avec_config *conf,
+				   struct object_cipher_info *object)
+{	
+	check_full_block(conf);
+	return get_atom_size(object);
+}
+
+static uint32_t io_size_nopad_data_head(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_head(get_data_conf(frame), object);
+}
+
+static uint32_t io_size_nopad_hole_head(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_head(get_hole_conf(frame), object);
+}
+
+static uint32_t io_size_nopad_data_tail(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_tail(get_data_conf(frame), object);
+}
+
+static uint32_t io_size_nopad_hole_tail(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_tail(get_hole_conf(frame), object);
+}
+
+static uint32_t io_size_nopad_data_full(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_full(get_data_conf(frame), object);
+}
+
+static uint32_t io_size_nopad_hole_full(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_full(get_hole_conf(frame), object);
+}
+
+static uint32_t offset_in_head(struct avec_config *conf)
+{
+	check_cursor_head(conf);
+
+	return conf->off_in_head;
+}
+
+static uint32_t offset_in_tail(call_frame_t *frame,
+			       struct object_cipher_info *object)
+{
+	return 0;
+}
+
+static uint32_t offset_in_full(struct avec_config *conf,
+			       struct object_cipher_info *object)
+{
+	check_cursor_full(conf);
+
+	if (has_head_block(conf))
+		return (conf->cursor - 1) << get_atom_bits(object);
+	else
+		return conf->cursor << get_atom_bits(object);
+}
+
+static uint32_t offset_in_data_head(call_frame_t *frame,
+				    struct object_cipher_info *object)
+{
+	return offset_in_head(get_data_conf(frame));
+}
+
+static uint32_t offset_in_hole_head(call_frame_t *frame,
+				    struct object_cipher_info *object)
+{
+	return offset_in_head(get_hole_conf(frame));
+}
+
+static uint32_t offset_in_data_full(call_frame_t *frame,
+				    struct object_cipher_info *object)
+{
+	return offset_in_full(get_data_conf(frame), object);
+}
+
+static uint32_t offset_in_hole_full(call_frame_t *frame,
+				    struct object_cipher_info *object)
+{
+	return offset_in_full(get_hole_conf(frame), object);
+}
+
+/*
+ * atom->rmw()
+ */
+/*
+ * Pre-conditions:
+ * @vec contains plain text of the latest
+ * version.
+ *
+ * Uptodate gaps of the @partial block with
+ * this plain text, encrypt the whole block
+ * and write the result to disk.
+ */
+static int32_t rmw_partial_block(call_frame_t *frame,
+				 void *cookie,
+				 xlator_t *this,
+				 int32_t op_ret,
+				 int32_t op_errno,
+				 struct iovec *vec,
+				 int32_t count,
+				 struct iatt *stbuf,
+				 struct iobref *iobref,
+				 struct rmw_atom *atom)
+{
+	size_t was_read = 0;
+	uint64_t file_size;
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	struct iovec *partial = atom->get_iovec(frame, 0);
+	struct avec_config *conf = atom->get_config(frame);
+	end_writeback_handler_t end_writeback_partial_block;
+#if DEBUG_CRYPT
+	gf_boolean_t check_last_cblock = _gf_false;
+#endif
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0)
+		goto exit;
+
+	file_size = local->cur_file_size;
+	was_read = op_ret;
+
+	if (atom->locality == HEAD_ATOM && conf->off_in_head) {
+		/*
+		 * head atom with a non-uptodate gap
+		 * at the beginning
+		 *
+		 * fill the gap with plain text of the
+		 * latest version. Convert a part of hole
+		 * (if any) to zeros.
+		 */
+		int32_t i;
+		int32_t copied = 0;
+		int32_t to_gap; /* amount of data needed to uptodate
+				   the gap at the beginning */
+#if 0
+		int32_t hole = 0; /* The part of the hole which
+				   * got in the head block */
+#endif /* 0 */
+		to_gap = conf->off_in_head;
+
+		if (was_read < to_gap) {
+			if (file_size >
+			    offset_at_head(conf) + was_read) {
+				/*
+				 * It is impossible to uptodate
+				 * head block: too few bytes have
+				 * been read from disk, so that
+				 * partial write is impossible.
+				 *
+				 * It could happen because of many
+				 * reasons: IO errors, (meta)data
+				 * corruption in the local file system,
+				 * etc.
+				 */
+				gf_log(this->name, GF_LOG_WARNING,
+				     "Can not uptodate a gap at the beginning");
+				local->op_ret = -1;
+				local->op_errno = EIO;
+				goto exit;
+			}
+#if 0
+			hole = to_gap - was_read;
+#endif /* 0 */
+			to_gap = was_read;
+		}
+		/*
+		 * uptodate the gap at the beginning
+		 */
+		for (i = 0; i < count && copied < to_gap; i++) {
+			int32_t to_copy;
+
+			to_copy = vec[i].iov_len;
+			if (to_copy > to_gap - copied)
+				to_copy = to_gap - copied;
+
+			memcpy(partial->iov_base, vec[i].iov_base, to_copy);
+			copied += to_copy;
+		}
+#if 0
+		/*
+		 * If possible, convert part of the
+		 * hole, which got in the head block
+		 */
+		ret = TRY_LOCK(&local->hole_lock);
+		if (!ret) {
+			if (local->hole_handled)
+				/*
+				 * already converted by
+				 * crypt_writev_cbk()
+				 */
+				UNLOCK(&local->hole_lock);
+			else {
+				/*
+				 * convert the part of the hole
+				 * which got in the head block
+				 * to zeros.
+				 *
+				 * Update the orig_offset to make
+				 * sure writev_cbk() won't care
+				 * about this part of the hole.
+				 *
+				 */
+				memset(partial->iov_base + to_gap, 0, hole);
+
+				conf->orig_offset -= hole;
+				conf->orig_size += hole;
+				UNLOCK(&local->hole_lock);
+			}
+		}
+		else /*
+		      * conversion is being performed
+		      * by crypt_writev_cbk()
+		      */
+			;
+#endif /* 0 */
+	}
+	if (atom->locality == TAIL_ATOM ||
+	    (!has_tail_block(conf) && conf->off_in_tail)) {
+		/*
+		 * tail atom, or head atom with a non-uptodate
+		 * gap at the end.
+		 *
+		 * fill the gap at the end of the block
+		 * with plain text of the latest version.
+		 * Pad the result, (if needed)
+		 */
+		int32_t i;
+		int32_t to_gap;
+		int copied;
+		off_t off_in_tail;
+		int32_t to_copy;
+
+		off_in_tail = conf->off_in_tail;
+		to_gap = conf->gap_in_tail;
+
+		if (to_gap && was_read < off_in_tail + to_gap) {
+			/*
+			 * It is impossible to uptodate
+			 * the gap at the end: too few bytes
+			 * have been read from disk, so that
+			 * partial write is impossible.
+			 *
+			 * It could happen because of many
+			 * reasons: IO errors, (meta)data
+			 * corruption in the local file system,
+			 * etc.
+			 */
+			gf_log(this->name, GF_LOG_WARNING,
+			       "Can not uptodate a gap at the end");
+			local->op_ret = -1;
+			local->op_errno = EIO;
+			goto exit;
+		}
+		/*
+		 * uptodate the gap at the end
+		 */
+		copied = 0;
+		to_copy = to_gap;
+		for(i = count - 1; i >= 0 && to_copy > 0; i--) {
+			uint32_t from_vec, off_in_vec;
+
+			off_in_vec = 0;
+			from_vec = vec[i].iov_len;
+			if (from_vec > to_copy) {
+				off_in_vec = from_vec - to_copy;
+				from_vec = to_copy;
+			}
+			memcpy(partial->iov_base +
+			       off_in_tail + to_gap - copied - from_vec,
+			       vec[i].iov_base + off_in_vec,
+			       from_vec);
+
+			gf_log(this->name, GF_LOG_DEBUG,
+			       "uptodate %d bytes at tail. Offset at target(source): %d(%d)",
+			       (int)from_vec,
+			       (int)off_in_tail + to_gap - copied - from_vec,
+			       (int)off_in_vec);
+
+			copied += from_vec;
+			to_copy -= from_vec;
+		}
+		partial->iov_len = off_in_tail + to_gap;
+
+		if (object_alg_should_pad(object)) {
+			int32_t resid = 0;
+			resid = partial->iov_len & (object_alg_blksize(object) - 1);
+			if (resid) {
+				/*
+				 * append a new EOF padding
+				 */
+				local->eof_padding_size =
+					object_alg_blksize(object) - resid;
+
+				gf_log(this->name, GF_LOG_DEBUG,
+				       "set padding size %d",
+				       local->eof_padding_size);
+
+				memset(partial->iov_base + partial->iov_len,
+				       1,
+				       local->eof_padding_size);
+				partial->iov_len += local->eof_padding_size;
+#if DEBUG_CRYPT
+				gf_log(this->name, GF_LOG_DEBUG,
+				       "pad cblock with %d zeros:",
+				       local->eof_padding_size);
+				dump_cblock(this,
+					    (unsigned char *)partial->iov_base +
+					    partial->iov_len - object_alg_blksize(object));
+				check_last_cblock = _gf_true;
+#endif
+			}
+		}
+	}
+	/*
+	 * encrypt the whole block
+	 */
+	encrypt_aligned_iov(object,
+			    partial,
+			    1,
+			    atom->offset_at(frame, object));
+#if DEBUG_CRYPT
+	if (check_last_cblock == _gf_true) {
+		gf_log(this->name, GF_LOG_DEBUG,
+		       "encrypt last cblock with offset %llu",
+		       (unsigned long long)atom->offset_at(frame, object));
+		dump_cblock(this, (unsigned char *)partial->iov_base +
+			    partial->iov_len - object_alg_blksize(object));
+	}
+#endif
+	set_local_io_params_writev(frame, object, atom,
+				   atom->offset_at(frame, object),
+				   iovec_get_size(partial, 1));
+	/*
+	 * write the whole block to disk
+	 */
+	end_writeback_partial_block = dispatch_end_writeback(local->fop);
+	conf->cursor ++;
+	STACK_WIND(frame,
+		   end_writeback_partial_block,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->writev,
+		   local->fd,
+		   partial,
+		   1,
+		   atom->offset_at(frame, object),
+		   local->flags,
+		   local->iobref_data,
+		   local->xdata);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "submit partial block: %d bytes from %d offset",
+	       (int)iovec_get_size(partial, 1),
+	       (int)atom->offset_at(frame, object));
+ exit:
+	return 0;
+}
+
+/*
+ * Perform a (read-)modify-write sequence.
+ * This should be performed only after approval
+ * of upper server-side manager, i.e. the caller
+ * needs to make sure this is his turn to rmw.
+ */
+void submit_partial(call_frame_t *frame,
+		    xlator_t *this,
+		    fd_t *fd,
+		    atom_locality_type ltype)
+{
+	int32_t ret;
+	dict_t *dict;
+	struct rmw_atom *atom;
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	atom = atom_by_types(local->active_setup, ltype);
+	/*
+	 * To perform the "read" component of the read-modify-write
+	 * sequence the crypt translator does stack_wind to itself.
+	 * 
+	 * Pass current file size to crypt_readv()
+	 */
+	dict = dict_new();
+	if (!dict) {
+		/*
+		 * FIXME: Handle the error
+		 */
+		gf_log("crypt", GF_LOG_WARNING, "Can not alloc dict");
+		return;
+	}
+	ret = dict_set(dict,
+		       FSIZE_XATTR_PREFIX,
+		       data_from_uint64(local->cur_file_size));
+	if (ret) {
+		/*
+		 * FIXME: Handle the error
+		 */
+		dict_unref(dict);
+		gf_log("crypt", GF_LOG_WARNING, "Can not set dict");
+		goto exit;
+	}
+	STACK_WIND(frame,
+		   atom->rmw,
+		   this,
+		   this->fops->readv, /* crypt_readv */
+		   fd,
+		   atom->count_to_uptodate(frame, object), /* count */
+		   atom->offset_at(frame, object), /* offset to read from */
+		   0,
+		   dict);
+ exit:
+	dict_unref(dict);
+}
+
+/*
+ * submit blocks of FULL_ATOM type
+ */
+void submit_full(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+	struct rmw_atom *atom = atom_by_types(local->active_setup, FULL_ATOM);
+	uint32_t count; /* total number of full blocks to submit */
+	uint32_t granularity; /* number of blocks to submit in one iteration */
+
+	uint64_t off_in_file; /* start offset in the file, bytes */
+	uint32_t off_in_atom; /* start offset in the atom, blocks */
+	uint32_t blocks_written = 0; /* blocks written for this submit */
+
+	struct avec_config *conf = atom->get_config(frame);
+	end_writeback_handler_t end_writeback_full_block;
+	/*
+	 * Write full blocks by groups of granularity size.
+	 */
+	end_writeback_full_block = dispatch_end_writeback(local->fop);
+
+	if (is_ordered_mode(frame)) {
+		uint32_t skip = has_head_block(conf) ? 1 : 0;
+		count = 1;
+		granularity = 1;
+		/*
+		 * calculate start offset using cursor value;
+		 * here we should take into accout head block,
+		 * which corresponds to cursor value 0.
+		 */
+		off_in_file = atom->offset_at(frame, object) +
+			((conf->cursor - skip) << get_atom_bits(object));
+		off_in_atom = conf->cursor - skip;
+	}
+	else {
+		/*
+		 * in parallel mode
+		 */
+		count = conf->nr_full_blocks;
+		granularity = MAX_IOVEC;
+		off_in_file = atom->offset_at(frame, object);
+		off_in_atom = 0;
+	}
+	while (count) {
+		uint32_t blocks_to_write = count;
+
+		if (blocks_to_write > granularity)
+			blocks_to_write = granularity;
+		if (conf->type == HOLE_ATOM)
+			/*
+			 * reset iovec before encryption
+			 */
+			memset(atom->get_iovec(frame, 0)->iov_base,
+			       0,
+			       get_atom_size(object));
+		/*
+		 * encrypt the group
+		 */
+		encrypt_aligned_iov(object,
+				    atom->get_iovec(frame,
+						    off_in_atom +
+						    blocks_written),
+				    blocks_to_write,
+				    off_in_file + (blocks_written <<
+						   get_atom_bits(object)));
+		
+		set_local_io_params_writev(frame, object, atom,
+		        off_in_file + (blocks_written << get_atom_bits(object)),
+			blocks_to_write <<  get_atom_bits(object));
+
+		conf->cursor += blocks_to_write;
+
+		STACK_WIND(frame,
+			   end_writeback_full_block,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->writev,
+			   local->fd,
+			   atom->get_iovec(frame, off_in_atom + blocks_written),
+			   blocks_to_write,
+			   off_in_file + (blocks_written << get_atom_bits(object)),
+			   local->flags,
+			   local->iobref_data ? local->iobref_data : local->iobref,
+			   local->xdata);
+
+		gf_log("crypt", GF_LOG_DEBUG, "submit %d full blocks from %d offset",
+		       blocks_to_write,
+		       (int)(off_in_file + (blocks_written << get_atom_bits(object))));
+
+		count -= blocks_to_write;
+		blocks_written += blocks_to_write;
+	}
+	return;
+}
+
+static int32_t rmw_data_head(call_frame_t *frame,
+			     void *cookie,
+			     xlator_t *this,
+			     int32_t op_ret,
+			     int32_t op_errno,
+			     struct iovec *vec,
+			     int32_t count,
+			     struct iatt *stbuf,
+			     struct iobref *iobref,
+			     dict_t *xdata)
+{
+	return rmw_partial_block(frame,
+				 cookie,
+				 this,
+				 op_ret,
+				 op_errno,
+				 vec,
+				 count,
+				 stbuf,
+				 iobref,
+				 atom_by_types(DATA_ATOM, HEAD_ATOM));
+}
+
+static int32_t rmw_data_tail(call_frame_t *frame,
+			     void *cookie,
+			     xlator_t *this,
+			     int32_t op_ret,
+			     int32_t op_errno,
+			     struct iovec *vec,
+			     int32_t count,
+			     struct iatt *stbuf,
+			     struct iobref *iobref,
+			     dict_t *xdata)
+{
+	return rmw_partial_block(frame,
+				 cookie,
+				 this,
+				 op_ret,
+				 op_errno,
+				 vec,
+				 count,
+				 stbuf,
+				 iobref,
+				 atom_by_types(DATA_ATOM, TAIL_ATOM));
+}
+
+static int32_t rmw_hole_head(call_frame_t *frame,
+			     void *cookie,
+			     xlator_t *this,
+			     int32_t op_ret,
+			     int32_t op_errno,
+			     struct iovec *vec,
+			     int32_t count,
+			     struct iatt *stbuf,
+			     struct iobref *iobref,
+			     dict_t *xdata)
+{
+	return rmw_partial_block(frame,
+				 cookie,
+				 this,
+				 op_ret,
+				 op_errno,
+				 vec,
+				 count,
+				 stbuf,
+				 iobref,
+				 atom_by_types(HOLE_ATOM, HEAD_ATOM));
+}
+
+static int32_t rmw_hole_tail(call_frame_t *frame,
+			     void *cookie,
+			     xlator_t *this,
+			     int32_t op_ret,
+			     int32_t op_errno,
+			     struct iovec *vec,
+			     int32_t count,
+			     struct iatt *stbuf,
+			     struct iobref *iobref,
+			     dict_t *xdata)
+{
+	return rmw_partial_block(frame,
+				 cookie,
+				 this,
+				 op_ret,
+				 op_errno,
+				 vec,
+				 count,
+				 stbuf,
+				 iobref,
+				 atom_by_types(HOLE_ATOM, TAIL_ATOM));
+}
+
+/*
+ * atom->count_to_uptodate()
+ */
+static uint32_t count_to_uptodate_head(struct avec_config *conf,
+				       struct object_cipher_info *object)
+{
+	if (conf->acount == 1 && conf->off_in_tail)
+		return get_atom_size(object);
+	else
+		/* there is no need to read the whole head block */
+		return conf->off_in_head;
+}
+
+static uint32_t count_to_uptodate_tail(struct avec_config *conf,
+				       struct object_cipher_info *object)
+{
+	/* we need to read the whole tail block */
+	return get_atom_size(object);
+}
+
+static uint32_t count_to_uptodate_data_head(call_frame_t *frame,
+					    struct object_cipher_info *object)
+{
+	return count_to_uptodate_head(get_data_conf(frame), object);
+}
+
+static uint32_t count_to_uptodate_data_tail(call_frame_t *frame,
+					    struct object_cipher_info *object)
+{
+	return count_to_uptodate_tail(get_data_conf(frame), object);
+}
+
+static uint32_t count_to_uptodate_hole_head(call_frame_t *frame,
+					    struct object_cipher_info *object)
+{
+	return count_to_uptodate_head(get_hole_conf(frame), object);
+}
+
+static uint32_t count_to_uptodate_hole_tail(call_frame_t *frame,
+					    struct object_cipher_info *object)
+{
+	return count_to_uptodate_tail(get_hole_conf(frame), object);
+}
+
+/* atom->get_config() */
+
+static struct avec_config *get_config_data(call_frame_t *frame)
+{
+	return &((crypt_local_t *)frame->local)->data_conf;
+}
+
+static struct avec_config *get_config_hole(call_frame_t *frame)
+{
+	return &((crypt_local_t *)frame->local)->hole_conf;
+}
+
+/*
+ * atom->get_iovec()
+ */
+static struct iovec *get_iovec_hole_head(call_frame_t *frame,
+					 uint32_t count)
+{
+	struct avec_config *conf = get_hole_conf(frame);
+
+	return conf->avec;
+}
+
+static struct iovec *get_iovec_hole_full(call_frame_t *frame,
+					 uint32_t count)
+{
+	struct avec_config *conf = get_hole_conf(frame);
+
+	return conf->avec + (conf->off_in_head ? 1 : 0);
+}
+
+static inline struct iovec *get_iovec_hole_tail(call_frame_t *frame,
+						uint32_t count)
+{
+	struct avec_config *conf = get_hole_conf(frame);
+
+	return conf->avec + (conf->blocks_in_pool - 1);
+}
+
+static inline struct iovec *get_iovec_data_head(call_frame_t *frame,
+						uint32_t count)
+{
+	struct avec_config *conf = get_data_conf(frame);
+
+	return conf->avec;
+}
+
+static inline struct iovec *get_iovec_data_full(call_frame_t *frame,
+						uint32_t count)
+{
+	struct avec_config *conf = get_data_conf(frame);
+
+	return conf->avec + (conf->off_in_head ? 1 : 0) + count;
+}
+
+static inline struct iovec *get_iovec_data_tail(call_frame_t *frame,
+						uint32_t count)
+{
+	struct avec_config *conf = get_data_conf(frame);
+
+	return conf->avec +
+		(conf->off_in_head ? 1 : 0) +
+		conf->nr_full_blocks;
+}
+
+static struct rmw_atom atoms[LAST_DATA_TYPE][LAST_LOCALITY_TYPE] = {
+	[DATA_ATOM][HEAD_ATOM] =
+	{ .locality = HEAD_ATOM,
+	  .rmw = rmw_data_head,
+	  .offset_at = offset_at_data_head,
+	  .offset_in = offset_in_data_head,
+	  .get_iovec = get_iovec_data_head,
+	  .io_size_nopad = io_size_nopad_data_head,
+	  .count_to_uptodate = count_to_uptodate_data_head,
+	  .get_config = get_config_data
+	},
+	[DATA_ATOM][TAIL_ATOM] =
+	{ .locality = TAIL_ATOM,
+	  .rmw = rmw_data_tail,
+	  .offset_at = offset_at_data_tail,
+	  .offset_in = offset_in_tail,
+	  .get_iovec = get_iovec_data_tail,
+	  .io_size_nopad = io_size_nopad_data_tail,
+	  .count_to_uptodate = count_to_uptodate_data_tail,
+	  .get_config = get_config_data
+	},
+	[DATA_ATOM][FULL_ATOM] =
+	{ .locality = FULL_ATOM,
+	  .offset_at = offset_at_data_full,
+	  .offset_in = offset_in_data_full,
+	  .get_iovec = get_iovec_data_full,
+	  .io_size_nopad = io_size_nopad_data_full,
+	  .get_config = get_config_data
+	},
+	[HOLE_ATOM][HEAD_ATOM] =
+	{ .locality = HEAD_ATOM,
+	  .rmw = rmw_hole_head,
+	  .offset_at = offset_at_hole_head,
+	  .offset_in = offset_in_hole_head,
+	  .get_iovec = get_iovec_hole_head,
+	  .io_size_nopad = io_size_nopad_hole_head,
+	  .count_to_uptodate = count_to_uptodate_hole_head,
+	  .get_config = get_config_hole
+	},
+	[HOLE_ATOM][TAIL_ATOM] =
+	{ .locality = TAIL_ATOM,
+	  .rmw = rmw_hole_tail,
+	  .offset_at = offset_at_hole_tail,
+	  .offset_in = offset_in_tail,
+	  .get_iovec = get_iovec_hole_tail,
+	  .io_size_nopad = io_size_nopad_hole_tail,
+	  .count_to_uptodate = count_to_uptodate_hole_tail,
+	  .get_config = get_config_hole
+	},
+	[HOLE_ATOM][FULL_ATOM] =
+	{ .locality = FULL_ATOM,
+	  .offset_at = offset_at_hole_full,
+	  .offset_in = offset_in_hole_full,
+	  .get_iovec = get_iovec_hole_full,
+	  .io_size_nopad = io_size_nopad_hole_full,
+	  .get_config = get_config_hole
+	}
+};
+
+struct rmw_atom *atom_by_types(atom_data_type data,
+			       atom_locality_type locality)
+{
+	return &atoms[data][locality];
+}
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/crypt-common.h b/xlators/encryption/crypt/src/crypt-common.h
new file mode 100644
index 000000000..7c212ad5d
--- /dev/null
+++ b/xlators/encryption/crypt/src/crypt-common.h
@@ -0,0 +1,141 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef __CRYPT_COMMON_H__
+#define __CRYPT_COMMON_H__
+
+#define INVAL_SUBVERSION_NUMBER (0xff)
+#define CRYPT_INVAL_OP   (GF_FOP_NULL)
+
+#define CRYPTO_FORMAT_PREFIX  "trusted.glusterfs.crypt.att.cfmt"
+#define FSIZE_XATTR_PREFIX    "trusted.glusterfs.crypt.att.size"
+#define SUBREQ_PREFIX         "trusted.glusterfs.crypt.msg.sreq"
+#define FSIZE_MSG_PREFIX      "trusted.glusterfs.crypt.msg.size"
+#define DE_MSG_PREFIX         "trusted.glusterfs.crypt.msg.dent"
+#define REQUEST_ID_PREFIX     "trusted.glusterfs.crypt.msg.rqid"
+#define MSGFLAGS_PREFIX       "trusted.glusterfs.crypt.msg.xfgs"
+
+
+/* messages for crypt_open() */
+#define MSGFLAGS_REQUEST_MTD_RLOCK   1 /* take read lock and don't unlock */
+#define MSGFLAGS_REQUEST_MTD_WLOCK   2 /* take write lock and don't unlock */
+
+#define AES_BLOCK_BITS (4) /* AES_BLOCK_SIZE == 1 << AES_BLOCK_BITS */
+
+#define noop   do {; } while (0)
+#define cassert(cond) ({ switch (-1) { case (cond): case 0: break; } })
+#define __round_mask(x, y) ((__typeof__(x))((y)-1))
+#define round_up(x, y) ((((x)-1) | __round_mask(x, y))+1)
+
+/*
+ * Format of file's metadata
+ */
+struct crypt_format {
+	uint8_t loader_id;  /* version of metadata loader */
+	uint8_t versioned[0]; /* file's metadata of specific version */
+} __attribute__((packed));
+
+typedef enum {
+	AES_CIPHER_ALG,
+	LAST_CIPHER_ALG
+} cipher_alg_t;
+
+typedef enum {
+	XTS_CIPHER_MODE,
+	LAST_CIPHER_MODE
+} cipher_mode_t;
+
+typedef enum {
+	MTD_LOADER_V1,
+	LAST_MTD_LOADER
+} mtd_loader_id;
+
+static inline void msgflags_set_mtd_rlock(uint32_t *flags)
+{
+	*flags |= MSGFLAGS_REQUEST_MTD_RLOCK;
+}
+
+static inline void msgflags_set_mtd_wlock(uint32_t *flags)
+{
+	*flags |= MSGFLAGS_REQUEST_MTD_WLOCK;
+}
+
+static inline gf_boolean_t msgflags_check_mtd_rlock(uint32_t *flags)
+{
+	return *flags & MSGFLAGS_REQUEST_MTD_RLOCK;
+}
+
+static inline gf_boolean_t msgflags_check_mtd_wlock(uint32_t *flags)
+{
+	return *flags & MSGFLAGS_REQUEST_MTD_WLOCK;
+}
+
+static inline gf_boolean_t msgflags_check_mtd_lock(uint32_t *flags)
+{
+	return  msgflags_check_mtd_rlock(flags) ||
+		msgflags_check_mtd_wlock(flags);
+}
+
+/*
+ * returns number of logical blocks occupied
+ * (maybe partially) by @count bytes
+ * at offset @start.
+ */
+static inline off_t logical_blocks_occupied(uint64_t start, off_t count,
+					    int blkbits)
+{
+	return ((start + count - 1) >> blkbits) - (start >> blkbits) + 1;
+}
+
+/*
+ * are two bytes (represented by offsets @off1
+ * and @off2 respectively) in the same logical
+ * block.
+ */
+static inline int in_same_lblock(uint64_t off1, uint64_t off2,
+				 int blkbits)
+{
+	return off1 >> blkbits == off2 >> blkbits;
+}
+
+static inline void dump_cblock(xlator_t *this, unsigned char *buf)
+{
+	gf_log(this->name, GF_LOG_DEBUG,
+	       "dump cblock: %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x",
+	       (buf)[0],
+	       (buf)[1],
+	       (buf)[2],
+	       (buf)[3],
+	       (buf)[4],
+	       (buf)[5],
+	       (buf)[6],
+	       (buf)[7],
+	       (buf)[8],
+	       (buf)[9],
+	       (buf)[10],
+	       (buf)[11],
+	       (buf)[12],
+	       (buf)[13],
+	       (buf)[14],
+	       (buf)[15]);
+}
+
+#endif /* __CRYPT_COMMON_H__ */
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/crypt-mem-types.h b/xlators/encryption/crypt/src/crypt-mem-types.h
new file mode 100644
index 000000000..799727573
--- /dev/null
+++ b/xlators/encryption/crypt/src/crypt-mem-types.h
@@ -0,0 +1,43 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef __CRYPT_MEM_TYPES_H__
+#define __CRYPT_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_crypt_mem_types_ {
+	gf_crypt_mt_priv = gf_common_mt_end + 1,
+	gf_crypt_mt_inode,
+	gf_crypt_mt_data,
+	gf_crypt_mt_mtd,
+	gf_crypt_mt_loc,
+	gf_crypt_mt_iatt,
+	gf_crypt_mt_key,
+	gf_crypt_mt_iovec,
+	gf_crypt_mt_char,
+};
+
+#endif /* __CRYPT_MEM_TYPES_H__ */
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
+
+
+
diff --git a/xlators/encryption/crypt/src/crypt.c b/xlators/encryption/crypt/src/crypt.c
new file mode 100644
index 000000000..db2e6d83c
--- /dev/null
+++ b/xlators/encryption/crypt/src/crypt.c
@@ -0,0 +1,4498 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+#include <ctype.h>
+#include <sys/uio.h>
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "xlator.h"
+#include "logging.h"
+#include "defaults.h"
+
+#include "crypt-common.h"
+#include "crypt.h"
+
+static void init_inode_info_head(struct crypt_inode_info *info, fd_t *fd);
+static int32_t init_inode_info_tail(struct crypt_inode_info *info,
+				    struct master_cipher_info *master);
+static int32_t prepare_for_submit_hole(call_frame_t *frame, xlator_t *this,
+				       uint64_t from, off_t size);
+static int32_t load_file_size(call_frame_t *frame, void *cookie,
+			      xlator_t *this, int32_t op_ret, int32_t op_errno,
+			      dict_t *dict, dict_t *xdata);
+static void do_ordered_submit(call_frame_t *frame, xlator_t *this,
+			      atom_data_type dtype);
+static void do_parallel_submit(call_frame_t *frame, xlator_t *this,
+			       atom_data_type dtype);
+static void put_one_call_open(call_frame_t *frame);
+static void put_one_call_readv(call_frame_t *frame, xlator_t *this);
+static void put_one_call_writev(call_frame_t *frame, xlator_t *this);
+static void put_one_call_ftruncate(call_frame_t *frame, xlator_t *this);
+static void free_avec(struct iovec *avec, char **pool, int blocks_in_pool);
+static void free_avec_data(crypt_local_t *local);
+static void free_avec_hole(crypt_local_t *local);
+
+static crypt_local_t *crypt_alloc_local(call_frame_t *frame, xlator_t *this,
+					glusterfs_fop_t fop)
+{
+	crypt_local_t *local = NULL;
+
+	local = mem_get0(this->local_pool);
+	if (!local) {
+		gf_log(this->name, GF_LOG_ERROR, "out of memory");
+		return NULL;
+	}
+	local->fop = fop;
+	LOCK_INIT(&local->hole_lock);
+	LOCK_INIT(&local->call_lock);
+	LOCK_INIT(&local->rw_count_lock);
+
+	frame->local = local;
+	return local;
+}
+
+struct crypt_inode_info *get_crypt_inode_info(inode_t *inode, xlator_t *this)
+{
+	int ret;
+	uint64_t value = 0;
+	struct crypt_inode_info *info;
+
+	ret = inode_ctx_get(inode, this, &value);
+	if (ret == -1) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"Can not get inode info");
+		return NULL;
+	}
+	info = (struct crypt_inode_info *)(long)value;
+	if (info == NULL) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"Can not obtain inode info");
+		return NULL;
+	}
+	return info;
+}
+
+static struct crypt_inode_info *local_get_inode_info(crypt_local_t *local,
+						      xlator_t *this)
+{
+	if (local->info)
+		return local->info;
+	local->info = get_crypt_inode_info(local->fd->inode, this);
+	return local->info;
+}
+
+static struct crypt_inode_info *alloc_inode_info(crypt_local_t *local,
+						 loc_t *loc)
+{
+	struct crypt_inode_info *info;
+
+	info = GF_CALLOC(1, sizeof(*info), gf_crypt_mt_inode);
+	if (!info) {
+		local->op_ret = -1;
+		local->op_errno = ENOMEM;
+		gf_log ("crypt", GF_LOG_WARNING,
+			"Can not allocate inode info");
+		return NULL;
+	}
+	memset(info, 0, sizeof(*info));
+#if DEBUG_CRYPT
+	info->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
+	if (!info->loc) {
+		gf_log("crypt", GF_LOG_WARNING, "Can not allocate loc");
+		GF_FREE(info);
+		return NULL;
+	}
+	if (loc_copy(info->loc, loc)){
+		GF_FREE(info->loc);
+		GF_FREE(info);
+		return NULL;
+	}
+#endif /* DEBUG_CRYPT */
+
+	local->info = info;
+	return info;
+}
+
+static void free_inode_info(struct crypt_inode_info *info)
+{
+#if DEBUG_CRYPT
+	loc_wipe(info->loc);
+	GF_FREE(info->loc);
+#endif
+	memset(info, 0, sizeof(*info));
+	GF_FREE(info);
+}
+
+int crypt_forget (xlator_t *this, inode_t *inode)
+{
+        uint64_t ctx_addr = 0;
+        if (!inode_ctx_del (inode, this, &ctx_addr))
+                free_inode_info((struct crypt_inode_info *)(long)ctx_addr);
+        return 0;
+}
+
+#if DEBUG_CRYPT
+static void check_read(call_frame_t *frame, xlator_t *this, int32_t read,
+		       struct iovec *vec, int32_t count, struct iatt *stbuf)
+{
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = get_object_cinfo(local->info);
+	struct avec_config *conf = &local->data_conf;
+	uint32_t resid = stbuf->ia_size & (object_alg_blksize(object) - 1);
+
+	if (read <= 0)
+		return;
+	if (read != iovec_get_size(vec, count))
+		gf_log ("crypt", GF_LOG_DEBUG,
+			"op_ret differs from amount of read bytes");
+
+	if (object_alg_should_pad(object) && (read & (object_alg_blksize(object) - 1)))
+		gf_log ("crypt", GF_LOG_DEBUG,
+			"bad amount of read bytes (!= 0 mod(cblock size))");
+
+	if (conf->aligned_offset + read >
+	    stbuf->ia_size + (resid ? object_alg_blksize(object) - resid : 0))
+		gf_log ("crypt", GF_LOG_DEBUG,
+			"bad amount of read bytes (too large))");
+
+}
+
+#define PT_BYTES_TO_DUMP (32)
+static void dump_plain_text(crypt_local_t *local, struct iovec *avec)
+{
+	int32_t to_dump;
+	char str[PT_BYTES_TO_DUMP + 1];
+
+	if (!avec)
+		return;
+	to_dump = avec->iov_len;
+	if (to_dump > PT_BYTES_TO_DUMP)
+		to_dump = PT_BYTES_TO_DUMP;
+	memcpy(str, avec->iov_base, to_dump);
+	memset(str + to_dump, '0', 1);
+	gf_log("crypt", GF_LOG_DEBUG, "Read file: %s", str);
+}
+
+static int32_t data_conf_invariant(struct avec_config *conf)
+{
+	return conf->acount ==
+		!!has_head_block(conf) +
+		!!has_tail_block(conf)+
+		conf->nr_full_blocks;
+}
+
+static int32_t hole_conf_invariant(struct avec_config *conf)
+{
+	return conf->blocks_in_pool ==
+		!!has_head_block(conf) +
+		!!has_tail_block(conf)+
+		!!has_full_blocks(conf);
+}
+
+static void crypt_check_conf(struct avec_config *conf)
+{
+	int32_t ret = 0;
+	const char *msg;
+
+	switch (conf->type) {
+	case DATA_ATOM:
+		msg = "data";
+		ret = data_conf_invariant(conf);
+		break;
+	case HOLE_ATOM:
+		msg = "hole";
+		ret = hole_conf_invariant(conf);
+		break;
+	default:
+		msg = "unknown";
+	}
+	if (!ret)
+		gf_log("crypt", GF_LOG_DEBUG, "bad %s conf", msg);
+}
+
+static void check_buf(call_frame_t *frame, xlator_t *this, struct iatt *buf)
+{
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+	uint64_t local_file_size;
+
+	switch(local->fop) {
+	case GF_FOP_FTRUNCATE:
+		return;
+	case GF_FOP_WRITE:
+		local_file_size = local->new_file_size;
+		break;
+	case GF_FOP_READ:
+		if (parent_is_crypt_xlator(frame, this))
+			return;
+		local_file_size = local->cur_file_size;
+		break;
+	default:
+		gf_log("crypt", GF_LOG_DEBUG, "bad file operation");
+		return;
+	}
+	if (buf->ia_size != round_up(local_file_size,
+					 object_alg_blksize(object)))
+		gf_log("crypt", GF_LOG_DEBUG,
+		       "bad ia_size in buf (%llu), should be %llu",
+		       (unsigned long long)buf->ia_size,
+		       (unsigned long long)round_up(local_file_size,
+						    object_alg_blksize(object)));
+}
+
+#else
+#define check_read(frame, this, op_ret, vec, count, stbuf) noop
+#define dump_plain_text(local, avec) noop
+#define crypt_check_conf(conf) noop
+#define check_buf(frame, this, buf) noop
+#endif /* DEBUG_CRYPT */
+
+/*
+ * Pre-conditions:
+ * @vec represents a ciphertext of expanded size and
+ * aligned offset.
+ *
+ * Compound a temporal vector @avec with block-aligned
+ * components, decrypt and fix it up to represent a chunk
+ * of data corresponding to the original size and offset.
+ * Pass the result to the next translator.
+ */
+int32_t crypt_readv_cbk(call_frame_t *frame,
+			void *cookie,
+			xlator_t *this,
+			int32_t op_ret,
+			int32_t op_errno,
+			struct iovec *vec,
+			int32_t count,
+			struct iatt *stbuf,
+			struct iobref *iobref,
+			dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	struct iovec *avec;
+	uint32_t i;
+	uint32_t to_vec;
+	uint32_t to_user;
+
+	check_buf(frame, this, stbuf);
+	check_read(frame, this, op_ret, vec, count, stbuf);
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+	local->iobref = iobref_ref(iobref);
+
+	local->buf = *stbuf;
+	local->buf.ia_size = local->cur_file_size;
+
+	if (op_ret <= 0 || count == 0 || vec[0].iov_len == 0)
+		goto put_one_call;
+
+	if (conf->orig_offset >= local->cur_file_size) {
+		local->op_ret = 0;
+		goto put_one_call;
+	}
+	/* 
+	 * correct config params with real file size
+	 * and actual amount of bytes read
+	 */
+	set_config_offsets(frame, this,
+			   conf->orig_offset, op_ret, DATA_ATOM, 0);
+
+	if (conf->orig_offset + conf->orig_size > local->cur_file_size)
+		conf->orig_size = local->cur_file_size - conf->orig_offset;
+	/*
+	 * calculate amount of data to be returned
+	 * to user.
+	 */
+	to_user = op_ret;
+	if (conf->aligned_offset + to_user <= conf->orig_offset) {
+		gf_log(this->name, GF_LOG_WARNING, "Incomplete read");
+		local->op_ret = -1;
+		local->op_errno = EIO;
+		goto put_one_call;
+	}
+	to_user -= (conf->aligned_offset - conf->orig_offset);
+
+	if (to_user > conf->orig_size)
+		to_user = conf->orig_size;
+	local->rw_count = to_user;
+
+	op_errno = set_config_avec_data(this, local,
+					conf, object, vec, count);
+	if (op_errno) {
+		local->op_ret = -1;
+		local->op_errno = op_errno;
+		goto put_one_call;
+	}
+	avec = conf->avec;
+#if DEBUG_CRYPT
+	if (conf->off_in_tail != 0 &&
+	    conf->off_in_tail < object_alg_blksize(object) &&
+	    object_alg_should_pad(object))
+		gf_log(this->name, GF_LOG_DEBUG, "Bad offset in tail %d",
+		       conf->off_in_tail);
+	if (iovec_get_size(vec, count) != 0 &&
+	    in_same_lblock(conf->orig_offset + iovec_get_size(vec, count) - 1,
+			   local->cur_file_size - 1,
+			   object_alg_blkbits(object))) {
+		gf_log(this->name, GF_LOG_DEBUG, "Compound last cblock");
+		dump_cblock(this,
+			    (unsigned char *)(avec[conf->acount - 1].iov_base) +
+			    avec[conf->acount - 1].iov_len - object_alg_blksize(object));
+		dump_cblock(this,
+			    (unsigned char *)(vec[count - 1].iov_base) +
+			    vec[count - 1].iov_len - object_alg_blksize(object));
+	}
+#endif
+	decrypt_aligned_iov(object, avec,
+			    conf->acount, conf->aligned_offset);
+	/*
+	 * pass proper plain data to user
+	 */
+	avec[0].iov_base += (conf->aligned_offset - conf->orig_offset);
+	avec[0].iov_len  -= (conf->aligned_offset - conf->orig_offset);
+
+	to_vec = to_user;
+	for (i = 0; i < conf->acount; i++) {
+		if (avec[i].iov_len > to_vec)
+			avec[i].iov_len = to_vec;
+		to_vec -= avec[i].iov_len;
+	}
+ put_one_call:
+	put_one_call_readv(frame, this);
+	return 0;
+}
+
+static int32_t do_readv(call_frame_t *frame,
+			void *cookie,
+			xlator_t *this,
+			int32_t op_ret,
+			int32_t op_errno,
+			dict_t *dict,
+			dict_t *xdata)
+{
+	data_t *data;
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto error;	
+	/*
+	 * extract regular file size
+	 */
+	data = dict_get(dict, FSIZE_XATTR_PREFIX);
+	if (!data) {
+		gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
+		op_errno = EIO;
+		goto error;
+	}
+	local->cur_file_size = data_to_uint64(data);
+	
+	get_one_call(frame);
+	STACK_WIND(frame,
+		   crypt_readv_cbk,
+		   FIRST_CHILD (this),
+		   FIRST_CHILD (this)->fops->readv,
+		   local->fd,
+		   /*
+		    * FIXME: read amount can be reduced
+		    */
+		   local->data_conf.expanded_size,
+		   local->data_conf.aligned_offset,
+		   local->flags,
+		   local->xdata);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+
+	get_one_call(frame);
+	put_one_call_readv(frame, this);
+	return 0;
+}
+
+static int32_t crypt_readv_finodelk_cbk(call_frame_t *frame,
+					void *cookie,
+					xlator_t *this,
+					int32_t op_ret,
+					int32_t op_errno,
+					dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto error;
+	/*
+	 * An access has been granted,
+	 * retrieve file size
+	 */
+	STACK_WIND(frame,
+		   do_readv,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fgetxattr,
+		   local->fd,
+		   FSIZE_XATTR_PREFIX,
+		   NULL);
+	return  0;
+ error:
+	fd_unref(local->fd);
+	if (local->xdata)
+		dict_unref(local->xdata);
+	STACK_UNWIND_STRICT(readv,
+			    frame,
+			    -1,
+			    op_errno,
+			    NULL,
+			    0,
+			    NULL,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+static int32_t readv_trivial_completion(call_frame_t *frame,
+					void *cookie,
+					xlator_t *this,
+					int32_t op_ret,
+					int32_t op_errno,
+					struct iatt *buf,
+					dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "stat failed (%d)", op_errno);
+		goto error;
+	}
+	local->buf = *buf;
+	STACK_WIND(frame,
+		   load_file_size,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->getxattr,
+		   local->loc,
+		   FSIZE_XATTR_PREFIX,
+		   NULL);
+	return 0;	
+ error:
+	STACK_UNWIND_STRICT(readv, frame, op_ret, op_errno,
+			    NULL, 0, NULL, NULL, NULL);
+	return 0;
+}
+
+int32_t crypt_readv(call_frame_t *frame,
+		    xlator_t *this,
+		    fd_t *fd,
+		    size_t size,
+		    off_t offset,
+		    uint32_t flags, dict_t *xdata)
+{
+	int32_t ret;
+	crypt_local_t *local;
+	struct crypt_inode_info *info;
+	struct gf_flock lock = {0, };
+
+#if DEBUG_CRYPT
+	gf_log("crypt", GF_LOG_DEBUG, "reading %d bytes from offset %llu",
+	       (int)size, (long long)offset);
+	if (parent_is_crypt_xlator(frame, this))
+		gf_log("crypt", GF_LOG_DEBUG, "parent is crypt");
+#endif	
+	local = crypt_alloc_local(frame, this, GF_FOP_READ);
+	if (!local) {
+		ret = ENOMEM;
+		goto error;
+	}
+	if (size == 0)
+		goto trivial;
+
+	local->fd = fd_ref(fd);
+	local->flags = flags;
+
+	info = local_get_inode_info(local, this);
+	if (info == NULL) {
+		ret = EINVAL;
+		fd_unref(fd);
+		goto error;
+	}
+	if (!object_alg_atomic(&info->cinfo)) {
+		ret = EINVAL;
+		fd_unref(fd);
+		goto error;
+	}
+	set_config_offsets(frame, this, offset, size,
+			   DATA_ATOM, 0);
+	if (parent_is_crypt_xlator(frame, this)) {
+		data_t *data;
+		/*
+		 * We are called by crypt_writev (or cypt_ftruncate)
+		 * to perform the "read" component of the read-modify-write
+		 * (or read-prune-write) sequence for some atom;
+		 *
+		 * don't ask for access:
+		 * it has already been acquired
+		 *
+		 * Retrieve current file size
+		 */
+		if (!xdata) {
+			gf_log("crypt", GF_LOG_WARNING,
+			       "Regular file size hasn't been passed");
+			ret = EIO;
+			goto error;
+		}
+		data = dict_get(xdata, FSIZE_XATTR_PREFIX);
+		if (!data) {
+			gf_log("crypt", GF_LOG_WARNING,
+			       "Regular file size not found");
+			ret = EIO;
+			goto error;
+		}
+		local->old_file_size =
+			local->cur_file_size = data_to_uint64(data);
+
+		get_one_call(frame);
+		STACK_WIND(frame,
+			   crypt_readv_cbk,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->readv,
+			   local->fd,
+			   /*
+			    * FIXME: read amount can be reduced
+			    */
+			   local->data_conf.expanded_size,
+			   local->data_conf.aligned_offset,
+			   flags,
+			   NULL);
+		return 0;
+	}
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_RDLCK;
+        lock.l_whence = SEEK_SET;
+
+	STACK_WIND(frame,
+		   crypt_readv_finodelk_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ trivial:
+	STACK_WIND(frame,
+		   readv_trivial_completion,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fstat,
+		   fd,
+		   NULL);
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(readv,
+			    frame,
+			    -1,
+			    ret,
+			    NULL,
+			    0,
+			    NULL,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+void set_local_io_params_writev(call_frame_t *frame,
+				struct object_cipher_info *object,
+				struct rmw_atom *atom,
+				off_t io_offset,
+				uint32_t io_size)
+{
+	crypt_local_t *local = frame->local;
+
+	local->io_offset = io_offset;
+	local->io_size = io_size;
+
+	local->io_offset_nopad =
+		atom->offset_at(frame, object) + atom->offset_in(frame, object);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "set nopad offset to %llu",
+	       (unsigned long long)local->io_offset_nopad);
+
+	local->io_size_nopad = atom->io_size_nopad(frame, object);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "set nopad size to %llu",
+	       (unsigned long long)local->io_size_nopad);
+
+	local->update_disk_file_size = 0;
+	/*
+	 * NOTE: eof_padding_size is 0 for all full atoms;
+	 * For head and tail atoms it will be set up at rmw_partial block()
+	 */
+	local->new_file_size = local->cur_file_size;
+
+	if (local->io_offset_nopad + local->io_size_nopad > local->cur_file_size) {
+
+		local->new_file_size = local->io_offset_nopad + local->io_size_nopad;
+
+		gf_log("crypt", GF_LOG_DEBUG,
+		       "set new file size to %llu",
+		       (unsigned long long)local->new_file_size);
+
+		local->update_disk_file_size = 1;
+	}
+}
+
+void set_local_io_params_ftruncate(call_frame_t *frame,
+				   struct object_cipher_info *object)
+{
+	uint32_t resid;
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+
+	resid = conf->orig_offset & (object_alg_blksize(object) - 1);
+	if (resid) {
+		local->eof_padding_size =
+			object_alg_blksize(object) - resid;
+		local->new_file_size = conf->aligned_offset;
+		local->update_disk_file_size = 0;
+		/*
+		 * file size will be updated
+		 * in the ->writev() stack,
+		 * when submitting file tail
+		 */
+	}
+	else {	
+		local->eof_padding_size = 0;
+		local->new_file_size = conf->orig_offset;
+		local->update_disk_file_size = 1;
+		/*
+		 * file size will be updated
+		 * in this ->ftruncate stack
+		 */
+	}
+}
+
+static inline void submit_head(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	submit_partial(frame, this, local->fd, HEAD_ATOM);
+}
+
+static inline void submit_tail(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	submit_partial(frame, this, local->fd, TAIL_ATOM);
+}
+
+static void submit_hole(call_frame_t *frame, xlator_t *this)
+{
+	/*
+	 * hole conversion always means
+	 * appended write and goes in ordered fashion
+	 */
+	do_ordered_submit(frame, this, HOLE_ATOM);
+}
+
+static void submit_data(call_frame_t *frame, xlator_t *this)
+{
+	if (is_ordered_mode(frame)) {
+		do_ordered_submit(frame, this, DATA_ATOM);
+		return;
+	}
+	gf_log("crypt", GF_LOG_WARNING, "Bad submit mode");
+	get_nr_calls(frame, nr_calls_data(frame));
+	do_parallel_submit(frame, this, DATA_ATOM);
+	return;
+}
+
+/*
+ * heplers called by writev_cbk, fruncate_cbk in ordered mode
+ */
+
+static inline int32_t should_submit_hole(crypt_local_t *local)
+{
+	struct avec_config *conf = &local->hole_conf;
+
+	return conf->avec != NULL;
+}
+
+static inline int32_t should_resume_submit_hole(crypt_local_t *local)
+{
+	struct avec_config *conf = &local->hole_conf;
+
+	if (local->fop == GF_FOP_WRITE && has_tail_block(conf))
+		/*
+		 * Don't submit a part of hole, which
+		 * fits into a data block:
+ 		 * this part of hole will be converted
+		 * as a gap filled by zeros in data head
+		 * block.
+		 */
+		return conf->cursor < conf->acount - 1;
+	else
+		return conf->cursor < conf->acount;
+}
+
+static inline int32_t should_resume_submit_data(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+
+	if (is_ordered_mode(frame))
+		return conf->cursor < conf->acount;
+	/*
+	 * parallel writes
+	 */
+	return 0;
+}
+
+static inline int32_t should_submit_data_after_hole(crypt_local_t *local)
+{
+	return local->data_conf.avec != NULL;
+}
+
+static void update_local_file_params(call_frame_t *frame,
+				     xlator_t *this,
+				     struct iatt *prebuf,
+				     struct iatt *postbuf)
+{
+	crypt_local_t *local = frame->local;
+
+	check_buf(frame, this, postbuf);
+
+	local->prebuf = *prebuf;
+	local->postbuf = *postbuf;
+
+	local->prebuf.ia_size  = local->cur_file_size;
+	local->postbuf.ia_size = local->new_file_size;
+
+	local->cur_file_size = local->new_file_size;	
+}
+
+static int32_t end_writeback_writev(call_frame_t *frame,
+				    void *cookie,
+				    xlator_t *this,
+				    int32_t op_ret,
+				    int32_t op_errno,
+				    struct iatt *prebuf,
+				    struct iatt *postbuf,
+				    dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret <= 0) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "writev iteration failed");
+		goto put_one_call;
+	}
+	/*
+	 * op_ret includes paddings (atom's head, atom's tail and EOF)
+	 */
+	if (op_ret < local->io_size) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "Incomplete writev iteration");
+		goto put_one_call;
+	}
+	op_ret -= local->eof_padding_size;
+	local->op_ret = op_ret;
+
+	update_local_file_params(frame, this, prebuf, postbuf);
+
+	if (data_write_in_progress(local)) {
+
+		LOCK(&local->rw_count_lock);
+		local->rw_count += op_ret;
+		UNLOCK(&local->rw_count_lock);
+
+		if (should_resume_submit_data(frame))
+			submit_data(frame, this);
+	}
+	else {
+		/*
+		 * hole conversion is going on;
+		 * don't take into account written zeros
+		 */
+		if (should_resume_submit_hole(local))
+			submit_hole(frame, this);
+
+		else if (should_submit_data_after_hole(local))
+			submit_data(frame, this);
+	}
+ put_one_call:
+	put_one_call_writev(frame, this);
+	return 0;
+}
+
+#define crypt_writev_cbk end_writeback_writev
+
+#define HOLE_WRITE_CHUNK_BITS 12
+#define HOLE_WRITE_CHUNK_SIZE (1 << HOLE_WRITE_CHUNK_BITS)
+
+/*
+ * Convert hole of size @size at offset @off to
+ * zeros and prepare respective iovecs for submit.
+ * The hole lock should be held.
+ *
+ * Pre-conditions:
+ * @local->file_size is set and valid.
+ */
+int32_t prepare_for_submit_hole(call_frame_t *frame, xlator_t *this,
+				uint64_t off, off_t size)
+{
+	int32_t ret;
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	set_config_offsets(frame, this, off, size, HOLE_ATOM, 1);
+
+	ret = set_config_avec_hole(this, local,
+				   &local->hole_conf, object, local->fop);
+	crypt_check_conf(&local->hole_conf);
+
+	return ret;
+}
+
+/*
+ * prepare for submit @count bytes at offset @from
+ */
+int32_t prepare_for_submit_data(call_frame_t *frame, xlator_t *this,
+				off_t from, int32_t size, struct iovec *vec,
+				int32_t vec_count, int32_t setup_gap)
+{
+	uint32_t ret;
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	set_config_offsets(frame, this, from, size,
+			   DATA_ATOM, setup_gap);
+
+	ret = set_config_avec_data(this, local,
+				   &local->data_conf, object, vec, vec_count);
+	crypt_check_conf(&local->data_conf);
+
+	return ret;
+}
+
+static void free_avec(struct iovec *avec,
+		      char **pool, int blocks_in_pool)
+{
+	if (!avec)
+		return;
+	GF_FREE(pool);
+	GF_FREE(avec);
+}
+
+static void free_avec_data(crypt_local_t *local)
+{
+	return free_avec(local->data_conf.avec,
+			 local->data_conf.pool,
+			 local->data_conf.blocks_in_pool);
+}
+
+static void free_avec_hole(crypt_local_t *local)
+{
+	return free_avec(local->hole_conf.avec,
+			 local->hole_conf.pool,
+			 local->hole_conf.blocks_in_pool);
+}
+
+
+static void do_parallel_submit(call_frame_t *frame, xlator_t *this,
+			       atom_data_type dtype)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf;
+
+	local->active_setup = dtype;
+	conf = conf_by_type(frame, dtype);
+
+	if (has_head_block(conf))
+		submit_head(frame, this);
+
+	if (has_full_blocks(conf))
+		submit_full(frame, this);
+
+	if (has_tail_block(conf))
+		submit_tail(frame, this);
+	return;
+}
+
+static void do_ordered_submit(call_frame_t *frame, xlator_t *this,
+			      atom_data_type dtype)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf;
+
+	local->active_setup = dtype;
+	conf = conf_by_type(frame, dtype);
+
+	if (should_submit_head_block(conf)) {
+		get_one_call_nolock(frame);
+		submit_head(frame, this);
+	}
+	else if (should_submit_full_block(conf)) {
+		get_one_call_nolock(frame);
+		submit_full(frame, this);
+	}
+	else if (should_submit_tail_block(conf)) {
+		get_one_call_nolock(frame);
+		submit_tail(frame, this);
+	}
+	else
+		gf_log("crypt", GF_LOG_DEBUG,
+		       "nothing has been submitted in ordered mode");
+	return;
+}
+
+static int32_t do_writev(call_frame_t *frame,
+			 void *cookie,
+			 xlator_t *this,
+			 int32_t op_ret,
+			 int32_t op_errno,
+			 dict_t *dict,
+			 dict_t *xdata)
+{
+	data_t *data;
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+	/*
+	 * extract regular file size
+	 */
+	data = dict_get(dict, FSIZE_XATTR_PREFIX);
+	if (!data) {
+		gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
+		op_ret = -1;
+		op_errno = EIO;
+		goto error;
+	}
+	local->old_file_size = local->cur_file_size = data_to_uint64(data);
+
+	set_gap_at_end(frame, object, &local->data_conf, DATA_ATOM);
+
+	if (local->cur_file_size < local->data_conf.orig_offset) {
+		/*
+		 * Set up hole config
+		 */
+		op_errno = prepare_for_submit_hole(frame,
+			       this,
+			       local->cur_file_size,
+			       local->data_conf.orig_offset - local->cur_file_size);
+		if (op_errno) {
+			local->op_ret = -1;
+			local->op_errno = op_errno;
+			goto error;
+		}
+	}
+	if (should_submit_hole(local))
+		submit_hole(frame, this);
+	else 
+		submit_data(frame, this);
+	return 0;
+ error:
+	get_one_call_nolock(frame);
+	put_one_call_writev(frame, this);
+	return 0;
+}
+
+static int32_t crypt_writev_finodelk_cbk(call_frame_t *frame,
+					 void *cookie,
+					 xlator_t *this,
+					 int32_t op_ret,
+					 int32_t op_errno,
+					 dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0)
+		goto error;
+	/*
+	 * An access has been granted,
+	 * retrieve file size first
+	 */
+	STACK_WIND(frame,
+		   do_writev,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fgetxattr,
+		   local->fd,
+		   FSIZE_XATTR_PREFIX,
+		   NULL);
+	return 0;
+ error:
+	get_one_call_nolock(frame);
+	put_one_call_writev(frame, this);
+	return 0;	
+}
+
+static int32_t writev_trivial_completion(call_frame_t *frame,
+					 void *cookie,
+					 xlator_t *this,
+					 int32_t op_ret,
+					 int32_t op_errno,
+					 struct iatt *buf,
+					 dict_t *dict)
+{
+	crypt_local_t *local = frame->local;
+	
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+	local->prebuf = *buf;
+	local->postbuf = *buf;
+
+	local->prebuf.ia_size = local->cur_file_size;
+	local->postbuf.ia_size = local->cur_file_size;
+
+	get_one_call(frame);
+	put_one_call_writev(frame, this);
+	return 0;
+}
+
+int crypt_writev(call_frame_t *frame,
+		 xlator_t *this,
+		 fd_t *fd,
+		 struct iovec *vec,
+		 int32_t count,
+		 off_t offset,
+		 uint32_t flags,
+		 struct iobref *iobref,
+		 dict_t *xdata)
+{
+	int32_t ret;
+	crypt_local_t *local;
+	struct crypt_inode_info *info;
+	struct gf_flock lock = {0, };
+#if DEBUG_CRYPT
+	gf_log ("crypt", GF_LOG_DEBUG, "writing %d bytes from offset %llu",
+		(int)iovec_get_size(vec, count), (long long)offset);
+#endif
+	local = crypt_alloc_local(frame, this, GF_FOP_WRITE);
+	if (!local) {
+		ret = ENOMEM;
+		goto error;
+	}
+	local->fd = fd_ref(fd);
+
+	if (iobref)
+		local->iobref = iobref_ref(iobref);
+	/*
+	 * to update real file size on the server
+	 */
+	local->xattr = dict_new();
+	if (!local->xattr) {
+		ret = ENOMEM;
+		goto error;
+	}
+	local->flags = flags;
+
+	info = local_get_inode_info(local, this);
+	if (info == NULL) {
+		ret = EINVAL;
+		goto error;
+	}
+	if (!object_alg_atomic(&info->cinfo)) {
+		ret = EINVAL;
+		goto error;
+	}
+	if (iovec_get_size(vec, count) == 0)
+		goto trivial;
+
+	ret = prepare_for_submit_data(frame, this, offset,
+				      iovec_get_size(vec, count),
+				      vec, count, 0 /* don't setup gup
+						       in tail: we don't
+						       know file size yet */);
+	if (ret)
+		goto error;
+
+	if (parent_is_crypt_xlator(frame, this)) {
+		data_t *data;
+		/*
+		 * we are called by shinking crypt_ftruncate(),
+		 * which doesn't perform hole conversion;
+		 *
+		 * don't ask for access:
+		 * it has already been acquired
+		 */
+
+		/*
+		 * extract file size
+		 */
+		if (!xdata) {
+			gf_log("crypt", GF_LOG_WARNING,
+			       "Regular file size hasn't been passed");
+			ret = EIO;
+			goto error;
+		}
+		data = dict_get(xdata, FSIZE_XATTR_PREFIX);
+		if (!data) {
+			gf_log("crypt", GF_LOG_WARNING,
+			       "Regular file size not found");
+			ret = EIO;
+			goto error;
+		}
+		local->old_file_size =
+			local->cur_file_size = data_to_uint64(data);
+
+		submit_data(frame, this);
+		return 0;
+	}
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+	/*
+	 * lock the file and retrieve its size
+	 */
+	lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+	STACK_WIND(frame,
+		   crypt_writev_finodelk_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ trivial:
+	STACK_WIND(frame,
+		   writev_trivial_completion,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fstat,
+		   fd,
+		   NULL);
+	return 0;
+ error:
+	if (local && local->fd)
+		fd_unref(fd);
+	if (local && local->iobref)
+		iobref_unref(iobref);
+	if (local && local->xdata)
+		dict_unref(xdata);
+	if (local && local->xattr)
+		dict_unref(local->xattr);
+	if (local && local->info)
+		free_inode_info(local->info);
+
+	STACK_UNWIND_STRICT(writev, frame, -1, ret, NULL, NULL, NULL);
+	return 0;
+}
+
+int32_t prepare_for_prune(call_frame_t *frame, xlator_t *this, uint64_t offset)
+{
+	set_config_offsets(frame, this,
+			   offset,
+			   0, /* count */
+			   DATA_ATOM,
+			   0 /* since we prune, there is no
+				gap in tail to uptodate */);
+	return 0;
+}
+
+/*
+ * Finish the read-prune-modify sequence
+ *
+ * Can be invoked as
+ * 1) ->ftruncate_cbk() for cblock-aligned, or trivial prune
+ * 2) ->writev_cbk() for non-cblock-aligned prune
+ */
+
+static int32_t prune_complete(call_frame_t *frame,
+			      void *cookie,
+			      xlator_t *this,
+			      int32_t op_ret,
+			      int32_t op_errno,
+			      struct iatt *prebuf,
+			      struct iatt *postbuf,
+			      dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	update_local_file_params(frame, this, prebuf, postbuf);
+
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+/*
+ * This is called as ->ftruncate_cbk()
+ *
+ * Perform the "write" component of the
+ * read-prune-write sequence.
+ *
+ * submuit the rest of the file
+ */
+static int32_t prune_submit_file_tail(call_frame_t *frame,
+				      void *cookie,
+				      xlator_t *this,
+				      int32_t op_ret,
+				      int32_t op_errno,
+				      struct iatt *prebuf,
+				      struct iatt *postbuf,
+				      dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+	dict_t *dict;
+
+	if (op_ret < 0)
+		goto put_one_call;
+
+	if (local->xdata) {
+		dict_unref(local->xdata);
+		local->xdata = NULL;
+	}
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	dict = dict_new();
+	if (!dict) {
+		op_errno = ENOMEM;
+		goto error;
+	}
+
+	update_local_file_params(frame, this, prebuf, postbuf);
+	local->new_file_size = conf->orig_offset;
+
+	/*
+	 * The rest of the file is a partial block and, hence,
+	 * should be written via RMW sequence, so the crypt xlator
+	 * does STACK_WIND to itself.
+	 *
+	 * Pass current file size to crypt_writev()
+	 */
+	op_errno = dict_set(dict,
+			    FSIZE_XATTR_PREFIX,
+			    data_from_uint64(local->cur_file_size));
+	if (op_errno) {
+		gf_log("crypt", GF_LOG_WARNING,
+		       "can not set key to update file size");
+		dict_unref(dict);
+		goto error;
+	}
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "passing current file size (%llu) to crypt_writev",
+	       (unsigned long long)local->cur_file_size);
+	/*
+	 * Padding will be filled with
+	 * zeros by rmw_partial_block()
+	 */
+	STACK_WIND(frame,
+		   prune_complete,
+		   this,
+		   this->fops->writev, /* crypt_writev */
+		   local->fd,
+		   &local->vec,
+		   1,
+		   conf->aligned_offset, /* offset to write from */
+		   0,
+		   local->iobref,
+		   dict);
+
+	dict_unref(dict);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+ put_one_call:
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+/*
+ * This is called as a callback of ->writev() invoked in behalf
+ * of ftruncate(): it can be
+ * 1) ordered writes issued by hole conversion in the case of
+ *    expanded truncate, or
+ * 2) an rmw partial data block issued by non-cblock-aligned
+ * prune.
+ */
+int32_t end_writeback_ftruncate(call_frame_t *frame,
+				void *cookie,
+				xlator_t *this,
+				int32_t op_ret,
+				int32_t op_errno,
+				struct iatt *prebuf,
+				struct iatt *postbuf,
+				dict_t *xdata)
+{
+	crypt_local_t  *local = frame->local;
+	/*
+	 * if nothing has been written,
+	 * then it must be an error
+	 */
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0)
+		goto put_one_call;
+
+	update_local_file_params(frame, this, prebuf, postbuf);
+
+	if (data_write_in_progress(local))
+		/* case (2) */
+		goto put_one_call;
+	/* case (1) */
+	if (should_resume_submit_hole(local))
+		submit_hole(frame, this);
+	/*
+	 * case of hole, when we should't resume
+	 */
+ put_one_call:
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+/*
+ * Perform prune and write components of the
+ * read-prune-write sequence.
+ *
+ * Called as ->readv_cbk()
+ *
+ * Pre-conditions:
+ * @vec contains the latest atom of the file
+ * (plain text)
+ */
+static int32_t prune_write(call_frame_t *frame,
+			   void *cookie,
+			   xlator_t *this,
+			   int32_t op_ret,
+			   int32_t op_errno,
+			   struct iovec *vec,
+			   int32_t count,
+			   struct iatt *stbuf,
+			   struct iobref *iobref,
+			   dict_t *xdata)
+{
+	int32_t i;
+	size_t to_copy;
+	size_t copied = 0;
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+	if (op_ret == -1)
+		goto put_one_call;
+
+	/*
+	 * At first, uptodate head block
+	 */
+	if (iovec_get_size(vec, count) < conf->off_in_head) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "Failed to uptodate head block for prune");
+		local->op_ret = -1;
+		local->op_errno = EIO;
+		goto put_one_call;
+	}
+	local->vec.iov_len = conf->off_in_head;
+	local->vec.iov_base = GF_CALLOC(1, local->vec.iov_len,
+					gf_crypt_mt_data);
+
+	if (local->vec.iov_base == NULL) {
+		local->op_ret = -1;
+		local->op_errno = ENOMEM;
+	}
+	for (i = 0; i < count; i++) {
+		to_copy = vec[i].iov_len;
+		if (to_copy > local->vec.iov_len - copied)
+			to_copy = local->vec.iov_len - copied;
+
+		memcpy((char *)local->vec.iov_base + copied,
+		       vec[i].iov_base,
+		       to_copy);
+		copied += to_copy;
+		if (copied == local->vec.iov_len)
+			break;
+	}
+	/*
+	 * perform prune with aligned offset
+	 * (i.e. at this step we prune a bit
+	 * more then it is needed
+	 */
+	STACK_WIND(frame,
+		   prune_submit_file_tail,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->ftruncate,
+		   local->fd,
+		   conf->aligned_offset,
+		   local->xdata);
+	return 0;
+ put_one_call:
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+/*
+ * Perform a read-prune-write sequence
+ */
+int32_t read_prune_write(call_frame_t *frame, xlator_t *this)
+{
+	int32_t ret = 0;
+	dict_t *dict = NULL;
+	crypt_local_t  *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	set_local_io_params_ftruncate(frame, object);
+	get_one_call_nolock(frame);
+
+	if ((conf->orig_offset & (object_alg_blksize(object) - 1)) == 0) {
+		/*
+		 * cblock-aligned prune:
+		 * we don't need read and write components,
+		 * just cut file body
+		 */
+		gf_log("crypt", GF_LOG_DEBUG,
+		       "prune without RMW (at offset %llu",
+		       (unsigned long long)conf->orig_offset);
+
+		STACK_WIND(frame,
+			   prune_complete,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->ftruncate,
+			   local->fd,
+			   conf->orig_offset,
+			   local->xdata);
+		return 0;
+	}
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "prune with RMW (at offset %llu",
+	       (unsigned long long)conf->orig_offset);	
+	/*
+	 * We are about to perform the "read" component of the
+	 * read-prune-write sequence. It means that we need to
+	 * read encrypted data from disk and decrypt it.
+	 * So, the crypt translator does STACK_WIND to itself.
+	 * 
+	 * Pass current file size to crypt_readv()
+	 
+	 */
+	dict = dict_new();
+	if (!dict) {
+		gf_log("crypt", GF_LOG_WARNING, "Can not alloc dict");
+		ret = ENOMEM;
+		goto exit;
+	}
+	ret = dict_set(dict,
+		       FSIZE_XATTR_PREFIX,
+		       data_from_uint64(local->cur_file_size));
+	if (ret) {
+		gf_log("crypt", GF_LOG_WARNING, "Can not set dict");
+		goto exit;
+	}
+	STACK_WIND(frame,
+		   prune_write,
+		   this,
+		   this->fops->readv, /* crypt_readv */
+		   local->fd,
+		   get_atom_size(object), /* bytes to read */
+		   conf->aligned_offset, /* offset to read from */
+		   0,
+		   dict);
+ exit:
+	if (dict)
+		dict_unref(dict);
+	return ret;
+}
+
+/*
+ * File prune is more complicated than expand.
+ * First we need to read the latest atom to not lose info
+ * needed for proper update. Also we need to make sure that
+ * every component of read-prune-write sequence leaves data
+ * consistent
+ *
+ * Non-cblock aligned prune is performed as read-prune-write
+ * sequence:
+ *
+ * 1) read the latest atom;
+ * 2) perform cblock-aligned prune
+ * 3) issue a write request for the end-of-file
+ */
+int32_t prune_file(call_frame_t *frame, xlator_t *this, uint64_t offset)
+{
+	int32_t ret;
+
+	ret = prepare_for_prune(frame, this, offset);
+	if (ret)
+		return ret;
+	return read_prune_write(frame, this);
+}
+
+int32_t expand_file(call_frame_t *frame, xlator_t *this,
+		    uint64_t offset)
+{
+	int32_t ret;
+	crypt_local_t *local = frame->local;
+
+	ret = prepare_for_submit_hole(frame, this,
+				      local->old_file_size,
+				      offset - local->old_file_size);
+	if (ret)
+		return ret;
+	submit_hole(frame, this);
+	return 0;
+}
+
+static int32_t ftruncate_trivial_completion(call_frame_t *frame,
+					    void *cookie,
+					    xlator_t *this,
+					    int32_t op_ret,
+					    int32_t op_errno,
+					    struct iatt *buf,
+					    dict_t *dict)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+	local->prebuf = *buf;
+	local->postbuf = *buf;
+
+	local->prebuf.ia_size = local->cur_file_size;
+	local->postbuf.ia_size = local->cur_file_size;
+
+	get_one_call(frame);
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+static int32_t do_ftruncate(call_frame_t *frame,
+			    void *cookie,
+			    xlator_t *this,
+			    int32_t op_ret,
+			    int32_t op_errno,
+			    dict_t *dict,
+			    dict_t *xdata)
+{
+	data_t *data;
+	crypt_local_t *local = frame->local;
+
+	if (op_ret)
+		goto error;	
+	/*
+	 * extract regular file size
+	 */
+	data = dict_get(dict, FSIZE_XATTR_PREFIX);
+	if (!data) {
+		gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
+		op_errno = EIO;
+		goto error;
+	}
+	local->old_file_size = local->cur_file_size = data_to_uint64(data);
+
+	if (local->data_conf.orig_offset == local->cur_file_size) {
+#if DEBUG_CRYPT
+		gf_log("crypt", GF_LOG_DEBUG,
+		       "trivial ftruncate (current file size %llu)",
+		       (unsigned long long)local->cur_file_size);
+#endif
+		goto trivial;
+	}
+	else if (local->data_conf.orig_offset < local->cur_file_size) {
+#if DEBUG_CRYPT
+		gf_log("crypt", GF_LOG_DEBUG, "prune from %llu to %llu",
+		       (unsigned long long)local->cur_file_size,
+		       (unsigned long long)local->data_conf.orig_offset);
+#endif
+		op_errno = prune_file(frame,
+				      this,
+				      local->data_conf.orig_offset);
+	}
+	else {
+#if DEBUG_CRYPT
+		gf_log("crypt", GF_LOG_DEBUG, "expand from %llu to %llu",
+		       (unsigned long long)local->cur_file_size,
+		       (unsigned long long)local->data_conf.orig_offset);
+#endif
+		op_errno = expand_file(frame,
+				       this,
+				       local->data_conf.orig_offset);
+	}
+	if (op_errno)
+		goto error;
+	return 0;
+ trivial:
+	STACK_WIND(frame,
+		   ftruncate_trivial_completion,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fstat,
+		   local->fd,
+		   NULL);
+	return 0;
+ error:
+	/*
+	 * finish with ftruncate
+	 */
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+
+	get_one_call_nolock(frame);
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+static int32_t crypt_ftruncate_finodelk_cbk(call_frame_t *frame,
+					    void *cookie,
+					    xlator_t *this,
+					    int32_t op_ret,
+					    int32_t op_errno,
+					    dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0)
+		goto error;
+	/*
+	 * An access has been granted,
+	 * retrieve file size first
+	 */
+	STACK_WIND(frame,
+		   do_ftruncate,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fgetxattr,
+		   local->fd,
+		   FSIZE_XATTR_PREFIX,
+		   NULL);
+	return 0;
+ error:
+	get_one_call_nolock(frame);
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+/*
+ * ftruncate is performed in 2 steps:
+ * . recieve file size;
+ * . expand or prune file.
+ */
+static int32_t crypt_ftruncate(call_frame_t *frame,
+			       xlator_t *this,
+			       fd_t *fd,
+			       off_t offset,
+			       dict_t *xdata)
+{
+	int32_t ret;
+	crypt_local_t *local;
+	struct crypt_inode_info *info;
+	struct gf_flock lock = {0, };
+
+	local = crypt_alloc_local(frame, this, GF_FOP_FTRUNCATE);
+	if (!local) {
+		ret = ENOMEM;
+		goto error;
+	}
+	local->xattr = dict_new();
+	if (!local->xattr) {
+		ret = ENOMEM;
+		goto error;
+	}
+	local->fd = fd_ref(fd);
+	info = local_get_inode_info(local, this);
+	if (info == NULL) {
+		ret = EINVAL;
+		goto error;
+	}
+	if (!object_alg_atomic(&info->cinfo)) {
+		ret = EINVAL;
+		goto error;
+	}
+	local->data_conf.orig_offset = offset;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+	STACK_WIND(frame,
+		   crypt_ftruncate_finodelk_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ error:
+	if (local && local->fd)
+		fd_unref(fd);
+	if (local && local->xdata)
+		dict_unref(xdata);
+	if (local && local->xattr)
+		dict_unref(local->xattr);
+	if (local && local->info)
+		free_inode_info(local->info);
+
+	STACK_UNWIND_STRICT(ftruncate, frame, -1, ret, NULL, NULL, NULL);
+	return 0;
+}
+
+/* ->flush_cbk() */
+int32_t truncate_end(call_frame_t *frame,
+		     void *cookie,
+		     xlator_t *this,
+		     int32_t op_ret,
+		     int32_t op_errno,
+		     dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	STACK_UNWIND_STRICT(truncate,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    &local->prebuf,
+			    &local->postbuf,
+			    local->xdata);
+	return 0;
+}
+
+/* ftruncate_cbk() */
+int32_t truncate_flush(call_frame_t *frame,
+		       void *cookie,
+		       xlator_t *this,
+		       int32_t op_ret,
+		       int32_t op_errno,
+		       struct iatt *prebuf,
+		       struct iatt *postbuf,
+		       dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	fd_t *fd = local->fd;
+	local->prebuf = *prebuf;
+	local->postbuf = *postbuf;
+
+	STACK_WIND(frame,
+		   truncate_end,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->flush,
+		   fd,
+		   NULL);
+	fd_unref(fd);
+	return 0;
+}
+
+/*
+ * is called as ->open_cbk()
+ */
+static int32_t truncate_begin(call_frame_t *frame,
+			      void *cookie,
+			      xlator_t *this,
+			      int32_t op_ret,
+			      int32_t op_errno,
+			      fd_t *fd,
+			      dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0) {
+		fd_unref(fd);
+		STACK_UNWIND_STRICT(truncate,
+				    frame,
+				    op_ret,
+				    op_errno, NULL, NULL, NULL);
+		return 0;
+	}
+	/*
+	 * crypt_truncate() is implemented via crypt_ftruncate(),
+	 * so the crypt xlator does STACK_WIND to itself here
+	 */
+	STACK_WIND(frame,
+		   truncate_flush,
+		   this,
+		   this->fops->ftruncate, /* crypt_ftruncate */
+		   fd,
+		   local->offset,
+		   NULL);
+	return 0;
+}
+
+/*
+ * crypt_truncate() is implemented via crypt_ftruncate() as a
+ * sequence crypt_open() - crypt_ftruncate() - truncate_flush()
+ */
+int32_t crypt_truncate(call_frame_t *frame,
+		       xlator_t *this,
+		       loc_t *loc,
+		       off_t offset,
+		       dict_t *xdata)
+{
+	fd_t *fd;
+	crypt_local_t *local;
+
+#if DEBUG_CRYPT
+	gf_log(this->name, GF_LOG_DEBUG,
+	       "truncate file %s at offset %llu",
+	       loc->path, (unsigned long long)offset);
+#endif
+	local = crypt_alloc_local(frame, this, GF_FOP_TRUNCATE);
+	if (!local)
+		goto error;
+
+	fd = fd_create(loc->inode, frame->root->pid);
+	if (!fd) {
+		gf_log(this->name, GF_LOG_ERROR, "Can not create fd");
+		goto error;
+	}
+	local->fd = fd;
+	local->offset = offset;
+	local->xdata = xdata;
+	STACK_WIND(frame,
+		   truncate_begin,
+		   this,
+		   this->fops->open, /* crypt_open() */
+		   loc,
+		   O_RDWR,
+		   fd,
+		   NULL);
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(truncate, frame, -1, EINVAL, NULL, NULL, NULL);
+	return 0;
+}
+
+end_writeback_handler_t dispatch_end_writeback(glusterfs_fop_t fop)
+{
+	switch (fop) {
+	case GF_FOP_WRITE:
+		return end_writeback_writev;
+	case GF_FOP_FTRUNCATE:
+		return end_writeback_ftruncate;
+	default:
+		gf_log("crypt", GF_LOG_WARNING, "Bad wb operation %d", fop);
+		return NULL;
+	}
+}
+
+/*
+ * true, if the caller needs metadata string
+ */
+static int32_t is_custom_mtd(dict_t *xdata)
+{
+	data_t *data;
+	uint32_t flags;
+
+	if (!xdata)
+		return 0;
+
+	data = dict_get(xdata, MSGFLAGS_PREFIX);
+	if (!data)
+		return 0;
+	if (data->len != sizeof(uint32_t)) {
+		gf_log("crypt", GF_LOG_WARNING,
+		       "Bad msgflags size (%d)", data->len);
+		return -1;
+	}
+	flags = *((uint32_t *)data->data);
+	return msgflags_check_mtd_lock(&flags);
+}
+
+static int32_t crypt_open_done(call_frame_t *frame,
+			       void *cookie,
+			       xlator_t *this,
+			       int32_t op_ret,
+			       int32_t op_errno, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+	if (op_ret < 0)
+		gf_log(this->name, GF_LOG_WARNING, "mtd unlock failed (%d)",
+		       op_errno);
+	put_one_call_open(frame);
+	return 0;
+}
+
+static void crypt_open_tail(call_frame_t *frame, xlator_t *this)
+{
+	struct gf_flock  lock  = {0, };
+	crypt_local_t *local = frame->local;
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+	STACK_WIND(frame,
+		   crypt_open_done,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   local->fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+}
+
+/*
+ * load private inode info at open time
+ * called as ->fgetxattr_cbk()
+ */
+static int load_mtd_open(call_frame_t *frame,
+			 void *cookie,
+			 xlator_t *this,
+			 int32_t op_ret,
+			 int32_t op_errno,
+			 dict_t *dict,
+			 dict_t *xdata)
+{
+	int32_t ret;
+	gf_boolean_t upload_info;
+	data_t *mtd;
+	uint64_t value = 0;
+	struct crypt_inode_info *info;
+	crypt_local_t *local = frame->local;
+	crypt_private_t *priv = this->private;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (local->fd->inode->ia_type == IA_IFLNK)
+		goto exit;
+	if (op_ret < 0)
+		goto exit;
+	/*
+	 * first, check for cached info
+	 */
+	ret = inode_ctx_get(local->fd->inode, this, &value);
+	if (ret != -1) {
+		info = (struct crypt_inode_info *)(long)value;
+		if (info == NULL) {
+			gf_log(this->name, GF_LOG_WARNING,
+			       "Inode info expected, but not found");
+			local->op_ret = -1;
+			local->op_errno = EIO;
+			goto exit;
+		}
+		/*
+		 * info has been found in the cache
+		 */
+		upload_info = _gf_false;
+	}
+	else {
+		/*
+		 * info hasn't been found in the cache.
+		 */
+		info = alloc_inode_info(local, local->loc);
+		if (!info) {
+			local->op_ret = -1;
+			local->op_errno = ENOMEM;
+			goto exit;
+		}
+		init_inode_info_head(info, local->fd);
+		upload_info = _gf_true;
+	}
+	/*
+	 * extract metadata
+	 */
+	mtd = dict_get(dict, CRYPTO_FORMAT_PREFIX);
+	if (!mtd) {
+		local->op_ret = -1;
+		local->op_errno = ENOENT;
+		gf_log (this->name, GF_LOG_WARNING,
+			"Format string wasn't found");
+		goto exit;
+	}
+	/*
+	 * authenticate metadata against the path
+	 */	
+	ret = open_format((unsigned char *)mtd->data,
+			  mtd->len,
+			  local->loc,
+			  info,
+			  get_master_cinfo(priv),
+			  local,
+			  upload_info);
+	if (ret) {
+		local->op_ret = -1;
+		local->op_errno = ret;
+		goto exit;
+	}
+	if (upload_info) {
+		ret = init_inode_info_tail(info, get_master_cinfo(priv));
+		if (ret) {
+			local->op_ret = -1;
+			local->op_errno = ret;
+			goto exit;	
+		}
+		ret = inode_ctx_put(local->fd->inode,
+				    this, (uint64_t)(long)info);
+		if (ret == -1) {
+			local->op_ret = -1;
+			local->op_errno = EIO;
+			goto exit;
+		}
+	}
+	if (local->custom_mtd) {
+		/*
+		 * pass the metadata string to the customer
+		 */
+		ret = dict_set_static_bin(local->xdata,
+					  CRYPTO_FORMAT_PREFIX,
+					  mtd->data,
+					  mtd->len);
+		if (ret) {
+			local->op_ret = -1;
+			local->op_errno = ret;
+			goto exit;
+		}
+	}
+ exit:
+	if (!local->custom_mtd)
+		crypt_open_tail(frame, this);
+	else
+		put_one_call_open(frame);
+	return 0;
+}
+
+static int32_t crypt_open_finodelk_cbk(call_frame_t *frame,
+				       void *cookie,
+				       xlator_t *this,
+				       int32_t op_ret,
+				       int32_t op_errno,
+				       dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0) {
+		gf_log(this->name, GF_LOG_WARNING, "finodelk (LOCK) failed");
+		goto exit;
+	}
+	STACK_WIND(frame,
+		   load_mtd_open,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fgetxattr,
+		   local->fd,
+		   CRYPTO_FORMAT_PREFIX,
+		   NULL);
+	return 0;
+ exit:
+	put_one_call_open(frame);
+	return 0;
+}
+
+/*
+ * verify metadata against the specified pathname
+ */
+static int32_t crypt_open_cbk(call_frame_t *frame,
+			      void *cookie,
+			      xlator_t *this,
+			      int32_t op_ret,
+			      int32_t op_errno,
+			      fd_t *fd,
+			      dict_t *xdata)
+{
+	struct gf_flock lock = {0, };
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (local->fd->inode->ia_type == IA_IFLNK)
+		goto exit;
+	if (op_ret < 0)
+		goto exit;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+	else if (local->custom_mtd){
+		local->xdata = dict_new();
+		if (!local->xdata) {
+			local->op_ret = -1;
+			local->op_errno = ENOMEM;
+			gf_log ("crypt", GF_LOG_ERROR,
+				"Can not get new dict for mtd string");
+			goto exit;
+		}
+	}
+	lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = local->custom_mtd ? F_WRLCK : F_RDLCK;
+        lock.l_whence = SEEK_SET;
+
+	STACK_WIND(frame,
+		   crypt_open_finodelk_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ exit:
+	put_one_call_open(frame);
+	return 0;
+}
+
+static int32_t crypt_open(call_frame_t *frame,
+			  xlator_t *this,
+			  loc_t *loc,
+			  int32_t flags,
+			  fd_t *fd,
+			  dict_t *xdata)
+{
+	int32_t ret = ENOMEM;
+	crypt_local_t *local;
+
+	local = crypt_alloc_local(frame, this, GF_FOP_OPEN);
+	if (!local)
+		goto error;
+	local->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
+	if (!local->loc) {
+		ret = ENOMEM;
+		goto error;
+	}
+	memset(local->loc, 0, sizeof(*local->loc));
+	ret = loc_copy(local->loc, loc);
+	if (ret) {
+		GF_FREE(local->loc);
+		goto error;
+	}
+	local->fd = fd_ref(fd);
+
+	ret = is_custom_mtd(xdata);
+	if (ret < 0) {
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+		ret = EINVAL;
+		goto error;
+	}
+	local->custom_mtd = ret;
+
+	if ((flags & O_ACCMODE) == O_WRONLY)
+		/*
+		 * we can't open O_WRONLY, because
+		 * we need to do read-modify-write
+		 */
+		flags = (flags & ~O_ACCMODE) | O_RDWR;
+	/*
+	 * Make sure that out translated offsets
+	 * and counts won't be ignored
+	 */
+	flags &= ~O_APPEND;
+	get_one_call_nolock(frame);
+	STACK_WIND(frame,
+		   crypt_open_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->open,
+		   loc,
+		   flags,
+		   fd,
+		   xdata);
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(open,
+			    frame,
+			    -1,
+			    ret,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+static int32_t init_inode_info_tail(struct crypt_inode_info *info,
+				    struct master_cipher_info *master)
+{
+	int32_t ret;
+	struct object_cipher_info *object = &info->cinfo;
+
+#if DEBUG_CRYPT
+	gf_log("crypt", GF_LOG_DEBUG, "Init inode info for object %s",
+	       uuid_utoa(info->oid));
+#endif
+	ret = data_cipher_algs[object->o_alg][object->o_mode].set_private(info,
+									master);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Set private info failed");
+		return ret;
+	}
+	return 0;
+}
+
+/*
+ * Init inode info at ->create() time
+ */
+static void init_inode_info_create(struct crypt_inode_info *info,
+				   struct master_cipher_info *master,
+				   data_t *data)
+{
+	struct object_cipher_info *object;
+
+	info->nr_minor = CRYPT_XLATOR_ID;
+	memcpy(info->oid, data->data, data->len);
+
+	object = &info->cinfo;
+
+	object->o_alg        = master->m_alg;
+	object->o_mode       = master->m_mode;
+	object->o_block_bits = master->m_block_bits;
+	object->o_dkey_size  = master->m_dkey_size;
+}
+
+static void init_inode_info_head(struct crypt_inode_info *info, fd_t *fd)
+{
+	memcpy(info->oid, fd->inode->gfid, sizeof(uuid_t));
+}
+
+static int32_t crypt_create_done(call_frame_t *frame,
+				 void *cookie,
+				 xlator_t *this,
+				 int32_t op_ret,
+				 int32_t op_errno, dict_t *xdata)
+{
+	crypt_private_t *priv = this->private;
+	crypt_local_t *local = frame->local;
+	struct crypt_inode_info *info = local->info;
+	fd_t *local_fd = local->fd;
+	dict_t *local_xdata = local->xdata;
+	inode_t *local_inode = local->inode;
+
+	if (op_ret < 0) {
+		free_inode_info(info);
+		goto unwind;
+	}
+	op_errno = init_inode_info_tail(info, get_master_cinfo(priv));
+	if (op_errno) {
+		op_ret = -1;
+		free_inode_info(info);
+		goto unwind;
+	}
+	/*
+	 * FIXME: drop major subversion number
+	 */
+	op_ret = inode_ctx_put(local->fd->inode, this, (uint64_t)(long)info);
+	if (op_ret == -1) {
+		op_errno = EIO;
+		free_inode_info(info);
+		goto unwind;
+	}
+ unwind:
+	free_format(local);
+	STACK_UNWIND_STRICT(create,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    local_fd,
+			    local_inode,
+			    &local->buf,
+			    &local->prebuf,
+			    &local->postbuf,
+			    local_xdata);
+	fd_unref(local_fd);
+	inode_unref(local_inode);
+	if (local_xdata)
+		dict_unref(local_xdata);
+	return 0;
+}
+
+static int crypt_create_tail(call_frame_t *frame,
+			     void *cookie,
+			     xlator_t *this,
+			     int32_t op_ret,
+			     int32_t op_errno,
+			     dict_t *xdata)
+{
+	struct gf_flock  lock  = {0, };
+	crypt_local_t *local = frame->local;
+	fd_t *local_fd = local->fd;
+	dict_t *local_xdata = local->xdata;
+	inode_t *local_inode = local->inode;
+
+	dict_unref(local->xattr);
+
+	if (op_ret < 0)
+		goto error;
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+	STACK_WIND(frame,
+		   crypt_create_done,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   local->fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ error:
+	free_inode_info(local->info);
+	free_format(local);
+
+	STACK_UNWIND_STRICT(create,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    local_fd,
+			    local_inode,
+			    &local->buf,
+			    &local->prebuf,
+			    &local->postbuf,
+			    local_xdata);
+
+	fd_unref(local_fd);
+	inode_unref(local_inode);
+	if (local_xdata)
+		dict_unref(local_xdata);
+	return 0;
+}
+
+static int32_t crypt_create_finodelk_cbk(call_frame_t *frame,
+					 void *cookie,
+					 xlator_t *this,
+					 int32_t op_ret,
+					 int32_t op_errno,
+					 dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	struct crypt_inode_info *info = local->info;
+
+	if (op_ret < 0)
+		goto error;
+
+	STACK_WIND(frame,
+		   crypt_create_tail,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fsetxattr,
+		   local->fd,
+		   local->xattr, /* CRYPTO_FORMAT_PREFIX */
+		   0,
+		   NULL);
+	return 0;
+ error: 
+	free_inode_info(info);
+	free_format(local);
+	fd_unref(local->fd);
+	dict_unref(local->xattr);
+	if (local->xdata)
+		dict_unref(local->xdata);
+
+	STACK_UNWIND_STRICT(create,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    NULL,
+			    NULL,
+			    NULL,
+			    NULL,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+/*
+ * Create and store crypt-specific format on disk;
+ * Populate cache with private inode info
+ */
+static int32_t crypt_create_cbk(call_frame_t *frame,
+				void *cookie,
+				xlator_t *this,
+				int32_t op_ret,
+				int32_t op_errno,
+				fd_t *fd,
+				inode_t *inode,
+				struct iatt *buf,
+				struct iatt *preparent,
+				struct iatt *postparent,
+				dict_t *xdata)
+{
+	struct gf_flock lock = {0, };
+	crypt_local_t *local = frame->local;
+	struct crypt_inode_info *info = local->info;
+
+	if (op_ret < 0)
+		goto error;
+	if (xdata)
+		local->xdata = dict_ref(xdata);	
+	local->inode = inode_ref(inode);
+	local->buf = *buf;
+	local->prebuf = *preparent;
+	local->postbuf = *postparent;
+
+        lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+        STACK_WIND(frame,
+		   crypt_create_finodelk_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   local->fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);	
+	return 0;
+ error:
+	free_inode_info(info);
+	free_format(local);
+	fd_unref(local->fd);
+	dict_unref(local->xattr);
+
+	STACK_UNWIND_STRICT(create,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    NULL, NULL, NULL,
+			    NULL, NULL, NULL);
+	return 0;
+}
+
+static int32_t crypt_create(call_frame_t *frame,
+			    xlator_t *this,
+			    loc_t *loc,
+			    int32_t flags,
+			    mode_t mode,
+			    mode_t umask,
+			    fd_t *fd,
+			    dict_t *xdata)
+{
+	int ret;
+	data_t *data;
+	crypt_local_t *local;
+	crypt_private_t *priv;
+	struct master_cipher_info *master;
+	struct crypt_inode_info *info;
+
+	priv = this->private;
+	master = get_master_cinfo(priv);
+
+	if (master_alg_atomic(master)) {
+		/*
+		 * We can't open O_WRONLY, because we
+		 * need to do read-modify-write.
+		 */
+		if ((flags & O_ACCMODE) == O_WRONLY)
+			flags = (flags & ~O_ACCMODE) | O_RDWR;
+		/*
+		 * Make sure that out translated offsets
+		 * and counts won't be ignored
+		 */
+		flags &= ~O_APPEND;
+	}
+	local = crypt_alloc_local(frame, this, GF_FOP_CREATE);
+	if (!local) {
+		ret = ENOMEM;
+		goto error;
+	}
+	data = dict_get(xdata, "gfid-req");
+	if (!data) {
+		ret = EINVAL;
+		gf_log("crypt", GF_LOG_WARNING, "gfid not found");
+		goto error;
+	}
+	if (data->len != sizeof(uuid_t)) {
+		ret = EINVAL;
+		gf_log("crypt", GF_LOG_WARNING,
+		       "bad gfid size (%d), should be %d",
+		       (int)data->len, (int)sizeof(uuid_t));
+		goto error;
+	}
+	info = alloc_inode_info(local, loc);
+	if (!info){
+		ret = ENOMEM;
+		goto error;
+	}
+	/*
+	 * NOTE:
+	 * format has to be created BEFORE
+	 * proceeding to the untrusted server
+	 */
+	ret = alloc_format_create(local);
+	if (ret) {
+		free_inode_info(info);
+		goto error;
+	}
+	init_inode_info_create(info, master, data);
+
+	ret = create_format(local->format,
+			    loc,
+			    info,
+			    master);
+	if (ret) {
+		free_inode_info(info);
+		goto error;		
+	}
+	local->xattr = dict_new();
+	if (!local->xattr) {
+		free_inode_info(info);
+		free_format(local);	
+		goto error;
+	}
+	ret = dict_set_static_bin(local->xattr,
+				  CRYPTO_FORMAT_PREFIX,
+				  local->format,
+				  new_format_size());
+	if (ret) {
+		dict_unref(local->xattr);
+		free_inode_info(info);
+		free_format(local);
+		goto error;
+	}
+	ret = dict_set(local->xattr, FSIZE_XATTR_PREFIX, data_from_uint64(0));
+	if (ret) {
+		dict_unref(local->xattr);
+		free_inode_info(info);
+		free_format(local);
+		goto error;
+	}
+	local->fd = fd_ref(fd);
+
+	STACK_WIND(frame,
+		   crypt_create_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->create,
+		   loc,
+		   flags,
+		   mode,
+		   umask,
+		   fd,
+		   xdata);
+	return 0;
+ error:
+	gf_log("crypt", GF_LOG_WARNING, "can not create file");
+	STACK_UNWIND_STRICT(create,
+			    frame,
+			    -1,
+			    ret,
+			    NULL, NULL, NULL,
+			    NULL, NULL, NULL);
+	return 0;
+}
+
+/*
+ * FIXME: this should depends on the version of format string
+ */
+static int32_t filter_crypt_xattr(dict_t *dict,
+				  char *key, data_t *value, void *data)
+{
+        dict_del(dict, key);
+        return 0;
+}
+
+static int32_t crypt_fsetxattr(call_frame_t *frame,
+			       xlator_t *this,
+			       fd_t *fd,
+			       dict_t *dict,
+			       int32_t flags, dict_t *xdata)
+{
+	dict_foreach_fnmatch(dict, "trusted.glusterfs.crypt*",
+			     filter_crypt_xattr, NULL);
+	STACK_WIND(frame,
+		   default_fsetxattr_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fsetxattr,
+		   fd,
+		   dict,
+		   flags,
+		   xdata);
+	return 0;
+}
+
+/*
+ * TBD: verify file metadata before wind
+ */
+static int32_t crypt_setxattr(call_frame_t *frame,
+			      xlator_t *this,
+			      loc_t *loc,
+			      dict_t *dict,
+			      int32_t flags, dict_t *xdata)
+{
+	dict_foreach_fnmatch(dict, "trusted.glusterfs.crypt*",
+			     filter_crypt_xattr, NULL);
+	STACK_WIND(frame,
+		   default_setxattr_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->setxattr,
+		   loc,
+		   dict,
+		   flags,
+		   xdata);
+	return 0;
+}
+
+/*
+ * called as flush_cbk()
+ */
+static int32_t linkop_end(call_frame_t *frame,
+			  void *cookie,
+			  xlator_t *this,
+			  int32_t op_ret,
+			  int32_t op_errno,
+			  dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	linkop_unwind_handler_t unwind_fn;
+	unwind_fn = linkop_unwind_dispatch(local->fop);
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0 &&
+	    op_errno == ENOENT &&
+	    local->loc->inode->ia_type == IA_IFLNK) {
+		local->op_ret = 0;
+		local->op_errno = 0;
+	}
+	unwind_fn(frame);
+	return 0;
+}
+
+/*
+ * unpin inode on the server
+ */
+static int32_t link_flush(call_frame_t *frame,
+			  void *cookie,
+			  xlator_t *this,
+			  int32_t op_ret,
+			  int32_t op_errno,
+			  inode_t *inode,
+			  struct iatt *buf,
+			  struct iatt *preparent,
+			  struct iatt *postparent, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto error;
+	if (local->xdata) {
+		dict_unref(local->xdata);
+		local->xdata = NULL;
+	}
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+	local->inode = inode_ref(inode);
+	local->buf = *buf;
+	local->prebuf = *preparent;
+	local->postbuf = *postparent;
+
+	STACK_WIND(frame,
+		   linkop_end,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->flush,
+		   local->fd,
+		   NULL);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+	link_unwind(frame);
+	return 0;
+}
+
+void link_unwind(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	dict_t *xdata;
+	dict_t *xattr;
+	inode_t *inode;
+
+	if (!local) {
+		STACK_UNWIND_STRICT(link,
+				    frame,
+				    -1,
+				    ENOMEM,
+				    NULL,
+				    NULL,
+				    NULL,
+				    NULL,
+				    NULL);
+		return;
+	}
+	xdata = local->xdata;
+	xattr = local->xattr;
+	inode = local->inode;
+
+	if (local->loc){
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+	}
+	if (local->newloc) {
+		loc_wipe(local->newloc);
+		GF_FREE(local->newloc);
+	}
+	if (local->fd)
+		fd_unref(local->fd);
+	if (local->format)
+		GF_FREE(local->format);
+
+	STACK_UNWIND_STRICT(link,
+			    frame,
+			    local->op_ret,
+			    local->op_errno,
+			    inode,
+			    &local->buf,
+			    &local->prebuf,
+			    &local->postbuf,
+			    xdata);
+	if (xdata)
+		dict_unref(xdata);
+	if (xattr)
+		dict_unref(xattr);
+	if (inode)
+		inode_unref(inode);
+}
+
+void link_wind(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+
+	STACK_WIND(frame,
+		   link_flush,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->link,
+		   local->loc,
+		   local->newloc,
+		   local->xdata);
+}
+
+/*
+ * unlink()
+ */
+static int32_t unlink_flush(call_frame_t *frame,
+			    void *cookie,
+			    xlator_t *this,
+			    int32_t op_ret,
+			    int32_t op_errno,
+			    struct iatt *preparent,
+			    struct iatt *postparent, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto error;
+	local->prebuf = *preparent;
+	local->postbuf = *postparent;
+	if (local->xdata) {
+		dict_unref(local->xdata);
+		local->xdata = NULL;
+	}
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	STACK_WIND(frame,
+		   linkop_end,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->flush,
+		   local->fd,
+		   NULL);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+	unlink_unwind(frame);
+	return 0;
+}
+
+void unlink_unwind(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	dict_t *xdata;
+	dict_t *xattr;
+
+	if (!local) {
+		STACK_UNWIND_STRICT(unlink,
+				    frame,
+				    -1,
+				    ENOMEM,
+				    NULL,
+				    NULL,
+				    NULL);
+		return;
+	}
+	xdata = local->xdata;
+	xattr = local->xattr;
+	if (local->loc){
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+	}
+	if (local->fd)
+		fd_unref(local->fd);
+	if (local->format)
+		GF_FREE(local->format);
+
+	STACK_UNWIND_STRICT(unlink,
+			    frame,
+			    local->op_ret,
+			    local->op_errno,
+			    &local->prebuf,
+			    &local->postbuf,
+			    xdata);
+	if (xdata)
+		dict_unref(xdata);
+	if (xattr)
+		dict_unref(xattr);
+}
+
+void unlink_wind(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+
+	STACK_WIND(frame,
+		   unlink_flush,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->unlink,
+		   local->loc,
+		   local->flags,
+		   local->xdata);
+}
+
+void rename_unwind(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	dict_t *xdata;
+	dict_t *xattr;
+	struct iatt *prenewparent;
+	struct iatt *postnewparent;
+
+	if (!local) {
+		STACK_UNWIND_STRICT(rename,
+				    frame,
+				    -1,
+				    ENOMEM,
+				    NULL,
+				    NULL,
+				    NULL,
+				    NULL,
+				    NULL,
+				    NULL);	
+		return;
+	}
+	xdata = local->xdata;
+	xattr = local->xattr;
+	prenewparent = local->prenewparent;
+	postnewparent = local->postnewparent;	
+
+	if (local->loc){
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+	}
+	if (local->newloc){
+		loc_wipe(local->newloc);
+		GF_FREE(local->newloc);
+	}
+	if (local->fd)
+		fd_unref(local->fd);
+	if (local->format)
+		GF_FREE(local->format);
+
+	STACK_UNWIND_STRICT(rename,
+			    frame,
+			    local->op_ret,
+			    local->op_errno,
+			    &local->buf,
+			    &local->prebuf,
+			    &local->postbuf,
+			    prenewparent,
+			    postnewparent,
+			    xdata);
+	if (xdata)
+		dict_unref(xdata);
+	if (xattr)
+		dict_unref(xattr);
+	if (prenewparent)
+		GF_FREE(prenewparent);
+	if (postnewparent)
+		GF_FREE(postnewparent);
+}
+
+/*
+ * called as flush_cbk()
+ */
+static int32_t rename_end(call_frame_t *frame,
+			  void *cookie,
+			  xlator_t *this,
+			  int32_t op_ret,
+			  int32_t op_errno,
+			  dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	rename_unwind(frame);
+	return 0;
+}
+
+static int32_t rename_flush(call_frame_t *frame,
+			    void *cookie,
+			    xlator_t *this,
+			    int32_t op_ret,
+			    int32_t op_errno,
+			    struct iatt *buf,
+			    struct iatt *preoldparent,
+			    struct iatt *postoldparent,
+			    struct iatt *prenewparent,
+			    struct iatt *postnewparent,
+			    dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto error;
+	dict_unref(local->xdata);
+	local->xdata = NULL;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	local->buf = *buf;
+	local->prebuf = *preoldparent;
+	local->postbuf = *postoldparent;
+	if (prenewparent) {
+		local->prenewparent = GF_CALLOC(1, sizeof(*prenewparent),
+						gf_crypt_mt_iatt);
+		if (!local->prenewparent) {
+			op_errno = ENOMEM;
+			goto error;
+		}
+		*local->prenewparent = *prenewparent;
+	}
+	if (postnewparent) {
+		local->postnewparent = GF_CALLOC(1, sizeof(*postnewparent),
+						 gf_crypt_mt_iatt);
+		if (!local->postnewparent) {
+			op_errno = ENOMEM;
+			goto error;
+		}
+		*local->postnewparent = *postnewparent;
+	}
+	STACK_WIND(frame,
+		   rename_end,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->flush,
+		   local->fd,
+		   NULL);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+	rename_unwind(frame);
+	return 0;
+}
+
+void rename_wind(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+
+	STACK_WIND(frame,
+		   rename_flush,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->rename,
+		   local->loc,
+		   local->newloc,
+		   local->xdata);
+}
+
+static int32_t __do_linkop(call_frame_t *frame,
+			   void *cookie,
+			   xlator_t *this,
+			   int32_t op_ret,
+			   int32_t op_errno, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	linkop_wind_handler_t wind_fn;
+	linkop_unwind_handler_t unwind_fn;
+
+	wind_fn = linkop_wind_dispatch(local->fop);
+	unwind_fn = linkop_unwind_dispatch(local->fop);
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret >= 0)
+		wind_fn(frame, this);
+	else {
+		gf_log(this->name, GF_LOG_WARNING, "mtd unlock failed (%d)",
+		       op_errno);
+		unwind_fn(frame);
+	}
+	return 0;
+}
+
+static int32_t do_linkop(call_frame_t *frame,
+			 void *cookie,
+			 xlator_t *this,
+			 int32_t op_ret,
+			 int32_t op_errno,
+			 dict_t *xdata)
+{
+	struct gf_flock  lock  = {0, };
+	crypt_local_t *local = frame->local;
+	linkop_unwind_handler_t unwind_fn;
+
+	unwind_fn = linkop_unwind_dispatch(local->fop);
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if(op_ret < 0)
+		goto error;
+
+	lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+	STACK_WIND(frame,
+		   __do_linkop,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   local->fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ error:		
+	unwind_fn(frame);
+	return 0;
+}
+
+/*
+ * Update the metadata string (against the new pathname);
+ * submit the result
+ */
+static int32_t linkop_begin(call_frame_t *frame,
+			    void *cookie,
+			    xlator_t *this,
+			    int32_t op_ret,
+			    int32_t op_errno,
+			    fd_t *fd,
+			    dict_t *xdata)
+{
+	gf_boolean_t upload_info;
+	crypt_local_t *local = frame->local;
+	crypt_private_t *priv = this->private;
+	struct crypt_inode_info *info;
+	data_t *old_mtd;
+	uint32_t new_mtd_size;
+	uint64_t value = 0;
+	void (*unwind_fn)(call_frame_t *frame);
+	void (*wind_fn)(call_frame_t *frame, xlator_t *this);
+	mtd_op_t mop;
+
+	wind_fn = linkop_wind_dispatch(local->fop);
+	unwind_fn = linkop_unwind_dispatch(local->fop);
+	mop = linkop_mtdop_dispatch(local->fop);
+
+	if (local->fd->inode->ia_type == IA_IFLNK)
+		goto wind;
+	if (op_ret < 0)
+		/*
+		 * verification failed
+		 */
+		goto error;
+
+	old_mtd = dict_get(xdata, CRYPTO_FORMAT_PREFIX);
+	if (!old_mtd) {
+		op_errno = EIO;
+		gf_log (this->name, GF_LOG_DEBUG,
+			"Metadata string wasn't found");
+		goto error;
+	}
+	new_mtd_size = format_size(mop, old_mtd->len);
+	op_errno = alloc_format(local, new_mtd_size);
+	if (op_errno)
+		goto error;
+	/*
+	 * check for cached info 
+	 */
+	op_ret = inode_ctx_get(fd->inode, this, &value);
+	if (op_ret != -1) {
+		info = (struct crypt_inode_info *)(long)value;
+		if (info == NULL) {
+			gf_log (this->name, GF_LOG_WARNING,
+				"Inode info was not found");
+			op_errno = EINVAL;
+			goto error;
+		}
+		/*
+		 * info was found in the cache
+		 */
+		local->info = info;
+		upload_info = _gf_false;
+	}
+	else {
+		/*
+		 * info wasn't found in the cache;
+		 */
+		info = alloc_inode_info(local, local->loc);
+		if (!info)
+			goto error;
+		init_inode_info_head(info, fd);
+		local->info = info;
+		upload_info = _gf_true;
+	}
+	op_errno = open_format((unsigned char *)old_mtd->data,
+			       old_mtd->len,
+			       local->loc,
+			       info,
+			       get_master_cinfo(priv),
+			       local,
+			       upload_info);
+	if (op_errno)
+		goto error;
+	if (upload_info == _gf_true) {
+		op_errno = init_inode_info_tail(info,
+						get_master_cinfo(priv));
+		if (op_errno)
+			goto error;
+		op_errno = inode_ctx_put(fd->inode, this,
+					 (uint64_t)(long)(info));
+		if (op_errno == -1) {
+			op_errno = EIO;
+			goto error;
+		}
+	}
+	/*
+	 * update the format string (append/update/cup a MAC)
+	 */
+	op_errno = update_format(local->format,
+				 (unsigned char *)old_mtd->data,
+				 old_mtd->len,
+				 local->mac_idx,
+				 mop,
+				 local->newloc,
+				 info,
+				 get_master_cinfo(priv),
+				 local);
+	if (op_errno)
+		goto error;
+	/*
+	 * store the new format string on the server
+	 */
+	if (new_mtd_size) {
+		op_errno = dict_set_static_bin(local->xattr,
+					       CRYPTO_FORMAT_PREFIX,
+					       local->format,
+					       new_mtd_size);
+		if (op_errno)
+			goto error;
+	}
+	STACK_WIND(frame,
+		   do_linkop,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->setxattr,
+		   local->loc,
+		   local->xattr,
+		   0,
+		   NULL);
+	return 0;
+ wind:
+	wind_fn(frame, this);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+	unwind_fn(frame);
+	return 0;
+}
+
+static int32_t linkop_grab_local(call_frame_t *frame,
+				 xlator_t *this,
+				 loc_t *oldloc,
+				 loc_t *newloc,
+				 int flags, dict_t *xdata,
+				 glusterfs_fop_t op)
+{
+	int32_t ret = ENOMEM;
+	fd_t *fd;
+	crypt_local_t *local;
+
+	local = crypt_alloc_local(frame, this, op);
+	if (!local)
+		goto error;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	fd = fd_create(oldloc->inode, frame->root->pid);
+	if (!fd) {
+		gf_log(this->name, GF_LOG_ERROR, "Can not create fd");
+		goto error;
+	}
+	local->fd = fd;
+	local->flags = flags;
+	local->loc = GF_CALLOC(1, sizeof(*oldloc), gf_crypt_mt_loc);
+	if (!local->loc)
+		goto error;
+	memset(local->loc, 0, sizeof(*local->loc));
+	ret = loc_copy(local->loc, oldloc);
+	if (ret) {
+		GF_FREE(local->loc);
+		local->loc = NULL;
+		goto error;
+	}
+	if (newloc) {
+		local->newloc = GF_CALLOC(1, sizeof(*newloc), gf_crypt_mt_loc);
+		if (!local->newloc) {
+			GF_FREE(local->loc);
+			loc_wipe(local->loc);
+			goto error;
+		}
+		memset(local->newloc, 0, sizeof(*local->newloc));
+		ret = loc_copy(local->newloc, newloc);
+		if (ret) {
+			GF_FREE(local->loc);
+			loc_wipe(local->loc);
+			GF_FREE(local->newloc);
+			goto error;
+		}
+	}
+	local->xattr = dict_new();
+	if (!local->xattr) {
+		gf_log(this->name, GF_LOG_ERROR, "Can not create dict");
+		ret = ENOMEM;
+		goto error;
+	}
+	return 0;
+ error:
+	if (local->xdata)
+		dict_unref(local->xdata);
+	if (local->fd)
+		fd_unref(local->fd);
+	local->fd = 0;
+	local->loc = NULL;
+	local->newloc = NULL;
+
+	local->op_ret = -1;
+	local->op_errno = ret;
+
+	return ret;
+}
+
+/*
+ * read and verify locked metadata against the old pathname (via open);
+ * update the metadata string in accordance with the new pathname;
+ * submit modified metadata;
+ * wind;
+ */
+static int32_t linkop(call_frame_t *frame,
+		      xlator_t *this,
+		      loc_t *oldloc,
+		      loc_t *newloc,
+		      int flags,
+		      dict_t *xdata,
+		      glusterfs_fop_t op)
+{
+	int32_t ret;
+	dict_t *dict;
+	crypt_local_t *local;
+	void (*unwind_fn)(call_frame_t *frame);
+
+	unwind_fn = linkop_unwind_dispatch(op);
+
+	ret = linkop_grab_local(frame, this, oldloc, newloc, flags, xdata, op);
+	local = frame->local;
+	if (ret)
+		goto error;
+	dict = dict_new();
+	if (!dict) {
+		gf_log(this->name, GF_LOG_ERROR, "Can not create dict");
+		ret = ENOMEM;
+		goto error;
+	}
+	/*
+	 * Set a message to crypt_open() that we need
+	 * locked metadata string.
+	 * All link operations (link, unlink, rename)
+	 * need write lock
+	 */
+	msgflags_set_mtd_wlock(&local->msgflags);
+	ret = dict_set_static_bin(dict,
+				  MSGFLAGS_PREFIX,
+				  &local->msgflags,
+				  sizeof(local->msgflags));
+	if (ret) {
+		gf_log(this->name, GF_LOG_ERROR, "Can not set dict");
+		dict_unref(dict);
+		goto error;
+	}
+	/*
+	 * verify metadata against the old pathname
+	 * and retrieve locked metadata string
+	 */
+	STACK_WIND(frame,
+		   linkop_begin,
+		   this,
+		   this->fops->open, /* crypt_open() */
+		   oldloc,
+		   O_RDWR,
+		   local->fd,
+		   dict);
+	dict_unref(dict);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = ret;
+	unwind_fn(frame);
+	return 0;
+}
+
+static int32_t crypt_link(call_frame_t *frame, xlator_t *this,
+			  loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+	return linkop(frame, this, oldloc, newloc, 0, xdata, GF_FOP_LINK);
+}
+
+static int32_t crypt_unlink(call_frame_t *frame, xlator_t *this,
+			    loc_t *loc, int flags, dict_t *xdata)
+{
+	return linkop(frame, this, loc, NULL, flags, xdata, GF_FOP_UNLINK);
+}
+
+static int32_t crypt_rename(call_frame_t *frame, xlator_t *this,
+			    loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+	return linkop(frame, this, oldloc, newloc, 0, xdata, GF_FOP_RENAME);
+}
+
+static void put_one_call_open(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	if (put_one_call(local)) {
+		fd_t *fd = local->fd;
+		loc_t *loc = local->loc;
+		dict_t *xdata = local->xdata;
+
+		STACK_UNWIND_STRICT(open,
+				    frame,
+				    local->op_ret,
+				    local->op_errno,
+				    fd,
+				    xdata);
+		fd_unref(fd);
+		if (xdata)
+			dict_unref(xdata);
+		loc_wipe(loc);
+		GF_FREE(loc);
+	}
+}
+
+static int32_t __crypt_readv_done(call_frame_t *frame,
+				  void *cookie,
+				  xlator_t *this,
+				  int32_t op_ret,
+				  int32_t op_errno, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	fd_t *local_fd = local->fd;
+	dict_t *local_xdata = local->xdata;
+	/* read deals with data configs only */
+	struct iovec *avec = local->data_conf.avec;
+	char **pool = local->data_conf.pool;
+	int blocks_in_pool = local->data_conf.blocks_in_pool;
+	struct iobref *iobref = local->iobref;
+	struct iobref *iobref_data = local->iobref_data;
+
+	if (op_ret < 0) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "readv unlock failed (%d)", op_errno);
+		if (local->op_ret >= 0) {
+			local->op_ret = op_ret;
+			local->op_errno = op_errno;
+		}
+	}
+	dump_plain_text(local, avec);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "readv: ret_to_user: %d, iovec len: %d, ia_size: %llu",
+	       (int)(local->rw_count > 0 ? local->rw_count : local->op_ret),
+	       (int)(local->rw_count > 0 ? iovec_get_size(avec, local->data_conf.acount) : 0),
+	       (unsigned long long)local->buf.ia_size);
+
+	STACK_UNWIND_STRICT(readv,
+			    frame,
+			    local->rw_count > 0 ? local->rw_count : local->op_ret,
+			    local->op_errno,
+			    avec,
+			    avec ? local->data_conf.acount : 0,
+			    &local->buf,
+			    local->iobref,
+			    local_xdata);
+
+	free_avec(avec, pool, blocks_in_pool);
+	fd_unref(local_fd);
+	if (local_xdata)
+		dict_unref(local_xdata);
+	if (iobref)
+		iobref_unref(iobref);
+	if (iobref_data)
+		iobref_unref(iobref_data);
+	return 0;
+}
+
+static void crypt_readv_done(call_frame_t *frame, xlator_t *this)
+{
+	if (parent_is_crypt_xlator(frame, this))
+		/*
+		 *  don't unlock (it will be done by the parent)
+		 */
+		__crypt_readv_done(frame, NULL, this, 0, 0, NULL);
+	else {
+		crypt_local_t *local = frame->local;
+		struct gf_flock  lock  = {0, };
+
+		lock.l_type   = F_UNLCK;
+		lock.l_whence = SEEK_SET;
+		lock.l_start  = 0;
+		lock.l_len    = 0;
+		lock.l_pid    = 0;
+
+		STACK_WIND(frame,
+			   __crypt_readv_done,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->finodelk,
+			   this->name,
+			   local->fd,
+			   F_SETLKW,
+			   &lock,
+			   NULL);
+	}
+}
+
+static void put_one_call_readv(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	if (put_one_call(local))
+		crypt_readv_done(frame, this);
+}
+
+static int32_t __crypt_writev_done(call_frame_t *frame,
+				   void *cookie,
+				   xlator_t *this,
+				   int32_t op_ret,
+				   int32_t op_errno, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	fd_t *local_fd = local->fd;
+	dict_t *local_xdata = local->xdata;
+	int32_t ret_to_user;
+
+	if (local->xattr)
+		dict_unref(local->xattr);
+	/*
+	 * Calculate amout of butes to be returned
+	 * to user. We need to subtract paddings that
+	 * have been written as a part of atom.
+	 */
+	/*
+	 * subtract head padding
+	 */
+	if (local->rw_count == 0)
+		/*
+		 * Nothing has been written, it must be an error
+		 */
+		ret_to_user = local->op_ret;
+	else if (local->rw_count <= local->data_conf.off_in_head) {
+		gf_log("crypt", GF_LOG_WARNING, "Incomplete write");
+		ret_to_user = 0;
+	}
+	else
+		ret_to_user = local->rw_count -
+			local->data_conf.off_in_head;
+	/*
+	 * subtract tail padding
+	 */
+	if (ret_to_user > local->data_conf.orig_size)
+		ret_to_user = local->data_conf.orig_size;
+
+	if (local->iobref)
+		iobref_unref(local->iobref);
+	if (local->iobref_data)
+		iobref_unref(local->iobref_data);
+	free_avec_data(local);
+	free_avec_hole(local);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "writev: ret_to_user: %d", ret_to_user);
+
+	STACK_UNWIND_STRICT(writev,
+			    frame,
+			    ret_to_user,
+			    local->op_errno,
+			    &local->prebuf,
+			    &local->postbuf,
+			    local_xdata);
+	fd_unref(local_fd);
+	if (local_xdata)
+		dict_unref(local_xdata);
+	return 0;
+}
+
+static int32_t crypt_writev_done(call_frame_t *frame,
+				 void *cookie,
+				 xlator_t *this,
+				 int32_t op_ret,
+				 int32_t op_errno,
+				 dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		gf_log("crypt", GF_LOG_WARNING, "can not update file size");
+
+	if (parent_is_crypt_xlator(frame, this))
+		/*
+		 * don't unlock (it will be done by the parent)
+		 */
+		__crypt_writev_done(frame, NULL, this, 0, 0, NULL);
+	else {
+		struct gf_flock  lock  = {0, };
+
+		lock.l_type   = F_UNLCK;
+		lock.l_whence = SEEK_SET;
+		lock.l_start  = 0;
+		lock.l_len    = 0;
+		lock.l_pid    = 0;
+
+		STACK_WIND(frame,
+			   __crypt_writev_done,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->finodelk,
+			   this->name,
+			   local->fd,
+			   F_SETLKW,
+			   &lock,
+			   NULL);
+	}
+	return 0;
+}
+
+static void put_one_call_writev(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	if (put_one_call(local)) {
+		if (local->update_disk_file_size) {
+			int32_t ret;
+			/*
+			 * update file size, unlock the file and unwind
+			 */
+			ret = dict_set(local->xattr,
+				       FSIZE_XATTR_PREFIX,
+				       data_from_uint64(local->cur_file_size));
+			if (ret) {
+				gf_log("crypt", GF_LOG_WARNING,
+				       "can not set key to update file size");
+				crypt_writev_done(frame, NULL,
+						  this, 0, 0, NULL);
+				return;
+			}
+			gf_log("crypt", GF_LOG_DEBUG,
+			       "Updating disk file size to %llu",
+			       (unsigned long long)local->cur_file_size);
+			STACK_WIND(frame,
+				   crypt_writev_done,
+				   FIRST_CHILD(this),
+				   FIRST_CHILD(this)->fops->fsetxattr,
+				   local->fd,
+				   local->xattr, /* CRYPTO_FORMAT_PREFIX */
+				   0,
+				   NULL);
+		}
+		else
+			crypt_writev_done(frame, NULL, this, 0, 0, NULL);
+	}
+}
+
+static int32_t __crypt_ftruncate_done(call_frame_t *frame,
+				      void *cookie,
+				      xlator_t *this,
+				      int32_t op_ret,
+				      int32_t op_errno, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	fd_t *local_fd = local->fd;
+	dict_t *local_xdata = local->xdata;
+	char *iobase = local->vec.iov_base;
+
+	if (op_ret < 0) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "ftruncate unlock failed (%d)", op_errno);
+		if (local->op_ret >= 0) {
+			local->op_ret = op_ret;
+			local->op_errno = op_errno;
+		}
+	}
+	if (local->iobref_data)
+		iobref_unref(local->iobref_data);
+	free_avec_data(local);
+	free_avec_hole(local);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "ftruncate, return to user: presize=%llu, postsize=%llu",
+	       (unsigned long long)local->prebuf.ia_size,
+	       (unsigned long long)local->postbuf.ia_size);
+
+	STACK_UNWIND_STRICT(ftruncate,
+			    frame,
+			    local->op_ret < 0 ? -1 : 0,
+			    local->op_errno,
+			    &local->prebuf,
+			    &local->postbuf,
+			    local_xdata);
+	fd_unref(local_fd);
+	if (local_xdata)
+		dict_unref(local_xdata);
+	if (iobase)
+		GF_FREE(iobase);
+	return 0;
+}
+
+static int32_t crypt_ftruncate_done(call_frame_t *frame,
+				    void *cookie,
+				    xlator_t *this,
+				    int32_t op_ret,
+				    int32_t op_errno,
+				    dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	struct gf_flock  lock  = {0, };
+
+	dict_unref(local->xattr);
+	if (op_ret < 0)
+		gf_log("crypt", GF_LOG_WARNING, "can not update file size");
+
+	lock.l_type   = F_UNLCK;
+	lock.l_whence = SEEK_SET;
+	lock.l_start  = 0;
+	lock.l_len    = 0;
+	lock.l_pid    = 0;
+
+	STACK_WIND(frame,
+		   __crypt_ftruncate_done,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   local->fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+}
+
+static void put_one_call_ftruncate(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	if (put_one_call(local)) {
+		if (local->update_disk_file_size) {
+			int32_t ret;
+			/*
+			 * update file size, unlock the file and unwind
+			 */
+			ret = dict_set(local->xattr,
+				       FSIZE_XATTR_PREFIX,
+				       data_from_uint64(local->cur_file_size));
+			if (ret) {
+				gf_log("crypt", GF_LOG_WARNING,
+				       "can not set key to update file size");
+				crypt_ftruncate_done(frame, NULL,
+						     this, 0, 0, NULL);
+				return;
+			}
+			gf_log("crypt", GF_LOG_DEBUG,
+			       "Updating disk file size to %llu",
+			       (unsigned long long)local->cur_file_size);
+			STACK_WIND(frame,
+				   crypt_ftruncate_done,
+				   FIRST_CHILD(this),
+				   FIRST_CHILD(this)->fops->fsetxattr,
+				   local->fd,
+				   local->xattr, /* CRYPTO_FORMAT_PREFIX */
+				   0,
+				   NULL);
+		}
+		else
+			crypt_ftruncate_done(frame, NULL, this, 0, 0, NULL);
+	}
+}
+
+/*
+ * load regular file size for some FOPs
+ */
+static int32_t load_file_size(call_frame_t *frame,
+			      void *cookie,
+			      xlator_t *this,
+			      int32_t op_ret,
+			      int32_t op_errno,
+			      dict_t *dict,
+			      dict_t *xdata)
+{
+	data_t *data;
+	crypt_local_t *local = frame->local;
+
+	dict_t *local_xdata = local->xdata;
+	inode_t *local_inode = local->inode;
+
+	if (op_ret < 0)
+		goto unwind;
+	/*
+	 * load regular file size
+	 */
+	data = dict_get(dict, FSIZE_XATTR_PREFIX);
+	if (!data) {
+		if (local->xdata)
+			dict_unref(local->xdata);
+		gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
+		op_ret = -1;
+		op_errno = EIO;
+		goto unwind;
+	}
+	local->buf.ia_size = data_to_uint64(data);
+
+	gf_log(this->name, GF_LOG_DEBUG,
+	       "FOP %d: Translate regular file to %llu",
+	       local->fop,
+	       (unsigned long long)local->buf.ia_size);
+ unwind:	
+	if (local->fd)
+		fd_unref(local->fd);
+	if (local->loc) {
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+	}
+	switch (local->fop) {
+	case GF_FOP_FSTAT:
+		STACK_UNWIND_STRICT(fstat,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    op_ret >= 0 ? &local->buf : NULL,
+				    local->xdata);
+		break;
+	case GF_FOP_STAT:
+		STACK_UNWIND_STRICT(stat,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    op_ret >= 0 ? &local->buf : NULL,
+				    local->xdata);
+		break;
+	case GF_FOP_LOOKUP:
+		STACK_UNWIND_STRICT(lookup,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    op_ret >= 0 ? local->inode : NULL,
+				    op_ret >= 0 ? &local->buf : NULL,				    
+				    local->xdata,
+				    op_ret >= 0 ? &local->postbuf : NULL);
+		break;
+	case GF_FOP_READ:
+		STACK_UNWIND_STRICT(readv,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    NULL,
+				    0,
+				    op_ret >= 0 ? &local->buf : NULL,
+				    NULL,
+				    NULL);
+		break;
+	default:
+		gf_log(this->name, GF_LOG_WARNING,
+		       "Improper file operation %d", local->fop);
+	}
+	if (local_xdata)
+		dict_unref(local_xdata);
+	if (local_inode)
+		inode_unref(local_inode);
+	return 0;
+}
+
+static int32_t crypt_stat_common_cbk(call_frame_t *frame,
+				     void *cookie,
+				     xlator_t *this,
+				     int32_t op_ret,
+				     int32_t op_errno,
+				     struct iatt *buf, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto unwind;
+	if (!IA_ISREG(buf->ia_type))
+		goto unwind;
+
+	local->buf = *buf;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	switch (local->fop) {
+	case GF_FOP_FSTAT:
+		STACK_WIND(frame,
+			   load_file_size,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->fgetxattr,
+			   local->fd,
+			   FSIZE_XATTR_PREFIX,
+			   NULL);
+		break;
+	case GF_FOP_STAT:
+		STACK_WIND(frame,
+			   load_file_size,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->getxattr,
+			   local->loc,
+			   FSIZE_XATTR_PREFIX,
+			   NULL);
+		break;
+	default:
+		gf_log (this->name, GF_LOG_WARNING,
+			"Improper file operation %d", local->fop);
+	}
+	return 0;
+ unwind:
+	if (local->fd)
+		fd_unref(local->fd);
+	if (local->loc) {
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+	}
+	switch (local->fop) {
+	case GF_FOP_FSTAT:
+		STACK_UNWIND_STRICT(fstat,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    op_ret >= 0 ? buf : NULL,
+				    op_ret >= 0 ? xdata : NULL);
+		break;
+	case GF_FOP_STAT:
+		STACK_UNWIND_STRICT(stat,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    op_ret >= 0 ? buf : NULL,
+				    op_ret >= 0 ? xdata : NULL);
+		break;
+	default:
+		gf_log (this->name, GF_LOG_WARNING,
+			"Improper file operation %d", local->fop);
+	}
+	return 0;
+}
+
+static int32_t crypt_fstat(call_frame_t *frame,
+			   xlator_t *this,
+			   fd_t *fd, dict_t *xdata)
+{
+	crypt_local_t *local;
+
+	local = crypt_alloc_local(frame, this, GF_FOP_FSTAT);
+	if (!local)
+		goto error;
+	local->fd = fd_ref(fd);
+	STACK_WIND(frame,
+		   crypt_stat_common_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fstat,
+		   fd,
+		   xdata);	
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(fstat,
+			    frame,
+			    -1,
+			    ENOMEM,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+static int32_t crypt_stat(call_frame_t *frame,
+			  xlator_t *this,
+			  loc_t *loc, dict_t *xdata)
+{
+	int32_t ret;
+	crypt_local_t *local;
+
+	local = crypt_alloc_local(frame, this, GF_FOP_STAT);
+	if (!local)
+		goto error;
+	local->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
+	if (!local->loc)
+		goto error;
+	memset(local->loc, 0, sizeof(*local->loc));
+	ret = loc_copy(local->loc, loc);
+	if (ret) {
+		GF_FREE(local->loc);
+		goto error;
+	}
+	STACK_WIND(frame,
+		   crypt_stat_common_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->stat,
+		   loc,
+		   xdata);	
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(stat,
+			    frame,
+			    -1,
+			    ENOMEM,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+static int32_t crypt_lookup_cbk(call_frame_t *frame,
+				void *cookie,
+				xlator_t *this,
+				int32_t op_ret,
+				int32_t op_errno,
+				inode_t *inode,
+				struct iatt *buf, dict_t *xdata,
+				struct iatt *postparent)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto unwind;
+	if (!IA_ISREG(buf->ia_type))
+		goto unwind;
+
+	local->inode = inode_ref(inode);
+	local->buf = *buf;
+	local->postbuf = *postparent;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+	uuid_copy(local->loc->gfid, buf->ia_gfid);
+
+	STACK_WIND(frame,
+		   load_file_size,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->getxattr,
+		   local->loc,
+		   FSIZE_XATTR_PREFIX,
+		   NULL);
+	return 0;
+ unwind:
+	loc_wipe(local->loc);
+	GF_FREE(local->loc);
+	STACK_UNWIND_STRICT(lookup,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    inode,
+			    buf,
+			    xdata,
+			    postparent);
+	return 0;
+}
+
+static int32_t crypt_lookup(call_frame_t *frame,
+			    xlator_t *this,
+			    loc_t *loc, dict_t *xdata)
+{
+	int32_t ret;
+	crypt_local_t *local;
+
+	local = crypt_alloc_local(frame, this, GF_FOP_LOOKUP);
+	if (!local)
+		goto error;
+	local->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
+	if (!local->loc)
+		goto error;
+	memset(local->loc, 0, sizeof(*local->loc));
+	ret = loc_copy(local->loc, loc);
+	if (ret) {
+		GF_FREE(local->loc);
+		goto error;
+	}
+	gf_log(this->name, GF_LOG_DEBUG, "Lookup %s", loc->path);
+	STACK_WIND(frame,
+		   crypt_lookup_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->lookup,
+		   loc,
+		   xdata);
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(lookup,
+			    frame,
+			    -1,
+			    ENOMEM,
+			    NULL,
+			    NULL,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+/*
+ * for every regular directory entry find its real file size
+ * and update stat's buf properly
+ */
+static int32_t crypt_readdirp_cbk(call_frame_t *frame,
+				  void *cookie,
+				  xlator_t *this,
+				  int32_t op_ret,
+				  int32_t op_errno,
+				  gf_dirent_t *entries, dict_t *xdata)
+{
+	gf_dirent_t *entry = NULL;
+
+	if (op_ret < 0)
+		goto unwind;
+
+	list_for_each_entry (entry, (&entries->list), list) {
+		data_t *data;
+
+		if (!IA_ISREG(entry->d_stat.ia_type))
+			continue;
+		data = dict_get(entry->dict, FSIZE_XATTR_PREFIX);
+		if (!data){
+			gf_log("crypt", GF_LOG_WARNING,
+			       "Regular file size of direntry not found");
+			op_errno = EIO;
+			op_ret = -1;
+			break;
+		}
+		entry->d_stat.ia_size = data_to_uint64(data);
+	}
+ unwind:
+	STACK_UNWIND_STRICT(readdirp, frame, op_ret, op_errno, entries, xdata);
+	return 0;
+}
+
+/*
+ * ->readdirp() fills in-core inodes, so we need to set proper
+ * file sizes for all directory entries of the parent @fd.
+ * Actual updates take place in ->crypt_readdirp_cbk()
+ */
+static int32_t crypt_readdirp(call_frame_t *frame, xlator_t *this,
+			      fd_t *fd, size_t size, off_t offset,
+			      dict_t *xdata)
+{
+	int32_t ret = ENOMEM;
+
+	if (!xdata) {
+		xdata = dict_new();
+		if (!xdata)
+			goto error;
+	}
+	else
+		dict_ref(xdata);
+	/*
+	 * make sure that we'll have real file sizes at ->readdirp_cbk()
+	 */
+	ret = dict_set(xdata, FSIZE_XATTR_PREFIX, data_from_uint64(0));
+	if (ret) {
+		dict_unref(xdata);
+		goto error;
+	}
+	STACK_WIND(frame,
+		   crypt_readdirp_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->readdirp,
+		   fd,
+		   size,
+		   offset,
+		   xdata);
+	dict_unref(xdata);
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(readdirp, frame, -1, ret, NULL, NULL);
+	return 0;
+}
+
+static int32_t crypt_access(call_frame_t *frame,
+			    xlator_t *this,
+			    loc_t *loc,
+			    int32_t mask, dict_t *xdata)
+{
+	gf_log(this->name, GF_LOG_WARNING,
+	       "NFS mounts of encrypted volumes are unsupported");
+	STACK_UNWIND_STRICT(access, frame, -1, EPERM, NULL);
+	return 0;
+}
+
+int32_t master_set_block_size (xlator_t *this, crypt_private_t *priv,
+			       dict_t *options)
+{
+	uint64_t block_size = 0;
+	struct master_cipher_info *master = get_master_cinfo(priv);
+
+	if (options != NULL)
+		GF_OPTION_RECONF("block-size", block_size, options,
+				 size, error);
+	else
+		GF_OPTION_INIT("block-size", block_size, size, error);
+
+	switch (block_size) {
+	case 512:
+		master->m_block_bits = 9;
+		break;
+	case 1024:
+		master->m_block_bits = 10;
+		break;
+	case 2048:
+		master->m_block_bits = 11;
+		break;
+	case 4096:
+		master->m_block_bits = 12;
+		break;
+	default:
+		gf_log("crypt", GF_LOG_ERROR,
+		       "FATAL: unsupported block size %llu",
+		       (unsigned long long)block_size);
+		goto error;
+	}
+	return 0;
+ error:
+	return -1;
+}
+
+int32_t master_set_alg(xlator_t *this, crypt_private_t *priv)
+{
+	struct master_cipher_info *master = get_master_cinfo(priv);
+	master->m_alg = AES_CIPHER_ALG;
+	return 0;
+}
+
+int32_t master_set_mode(xlator_t *this, crypt_private_t *priv)
+{
+	struct master_cipher_info *master = get_master_cinfo(priv);
+	master->m_mode = XTS_CIPHER_MODE;
+	return 0;
+}
+
+/*
+ * set key size in bits to the master info
+ * Pre-conditions: cipher mode in the master info is uptodate.
+ */
+static int master_set_data_key_size (xlator_t *this, crypt_private_t *priv,
+				     dict_t *options)
+{
+	int32_t ret;
+	uint64_t key_size = 0;
+	struct master_cipher_info *master = get_master_cinfo(priv);
+
+	if (options != NULL)
+		GF_OPTION_RECONF("data-key-size", key_size, options,
+				 size, error);
+	else
+		GF_OPTION_INIT("data-key-size", key_size, size, error);
+
+	ret = data_cipher_algs[master->m_alg][master->m_mode].check_key(key_size);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "FATAL: wrong bin key size %llu for alg %d mode %d",
+		       (unsigned long long)key_size,
+		       (int)master->m_alg,
+		       (int)master->m_mode);
+		goto error;
+	}
+	master->m_dkey_size = key_size;
+	return 0;
+ error:
+	return -1;
+}
+
+static int is_hex(char *s) {
+	return  ('0' <= *s && *s <= '9') || ('a' <= *s && *s <= 'f');
+}
+
+static int parse_hex_buf(xlator_t *this, char *src, unsigned char *dst,
+			 int hex_size)
+{
+	int i;
+	int hex_byte = 0;
+
+	for (i = 0; i < (hex_size / 2); i++) {
+		if (!is_hex(src + i*2) || !is_hex(src + i*2 + 1)) {
+			gf_log("crypt", GF_LOG_ERROR,
+			       "FATAL: not hex symbol in key");
+			return -1;
+		}
+		if (sscanf(src + i*2, "%2x", &hex_byte) != 1) {
+			gf_log("crypt", GF_LOG_ERROR,
+			       "FATAL: can not parse hex key");
+			return -1;
+		}
+		dst[i] = hex_byte & 0xff;
+	}
+	return 0;
+}
+
+/*
+ * Parse options;
+ * install master volume key
+ */
+int32_t master_set_master_vol_key(xlator_t *this, crypt_private_t *priv)
+{
+	int32_t ret;
+	FILE *file = NULL;
+
+	int32_t key_size;
+	char *opt_key_file_pathname = NULL;
+
+	unsigned char bin_buf[MASTER_VOL_KEY_SIZE];
+	char hex_buf[2 * MASTER_VOL_KEY_SIZE];
+
+	struct master_cipher_info *master = get_master_cinfo(priv);
+	/*
+	 * extract master key passed via option
+	 */
+	GF_OPTION_INIT("master-key", opt_key_file_pathname, path, bad_key);
+
+	if (!opt_key_file_pathname) {
+		gf_log(this->name, GF_LOG_ERROR, "FATAL: missing master key");
+		return -1;
+	}
+	gf_log(this->name, GF_LOG_DEBUG, "handling file key %s",
+	       opt_key_file_pathname);
+
+	file = fopen(opt_key_file_pathname, "r");
+	if (file == NULL) {
+		gf_log(this->name, GF_LOG_ERROR,
+		       "FATAL: can not open file with master key");
+		return -1;
+	}
+	/*
+	 * extract hex key
+	 */
+	key_size = fread(hex_buf, 1, sizeof(hex_buf), file);
+	if (key_size < sizeof(hex_buf)) {
+		gf_log(this->name, GF_LOG_ERROR,
+		       "FATAL: master key is too short");
+		goto bad_key;
+	}
+	ret = parse_hex_buf(this, hex_buf, bin_buf, key_size);
+	if (ret)
+		goto bad_key;
+	memcpy(master->m_key, bin_buf, MASTER_VOL_KEY_SIZE);
+	memset(hex_buf, 0, sizeof(hex_buf));
+	fclose(file);
+
+ 	memset(bin_buf, 0, sizeof(bin_buf));
+	return 0;
+ bad_key:
+	gf_log(this->name, GF_LOG_ERROR, "FATAL: bad master key");
+	if (file)
+		fclose(file);
+	memset(bin_buf, 0, sizeof(bin_buf));
+	return -1;
+}
+
+/*
+ * Derive volume key for object-id authentication
+ */
+int32_t master_set_nmtd_vol_key(xlator_t *this, crypt_private_t *priv)
+{
+	return get_nmtd_vol_key(get_master_cinfo(priv));
+}
+
+int32_t crypt_init_xlator(xlator_t *this)
+{
+	int32_t ret;
+	crypt_private_t *priv = this->private; 
+
+	ret = master_set_alg(this, priv);
+	if (ret)
+		return ret;
+	ret = master_set_mode(this, priv);
+	if (ret)
+		return ret;
+	ret = master_set_block_size(this, priv, NULL);
+	if (ret)
+		return ret;
+	ret = master_set_data_key_size(this, priv, NULL);
+	if (ret)
+		return ret;
+	ret = master_set_master_vol_key(this, priv);
+	if (ret)
+		return ret;
+	return master_set_nmtd_vol_key(this, priv);
+}
+
+static int32_t crypt_alloc_private(xlator_t *this)
+{
+	this->private = GF_CALLOC(1, sizeof(crypt_private_t), gf_crypt_mt_priv);
+	if (!this->private) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "Can not allocate memory for private data");
+		return ENOMEM;
+	}
+	return 0;
+}
+
+static void crypt_free_private(xlator_t *this)
+{
+	crypt_private_t *priv = this->private;
+	if (priv) {
+		memset(priv, 0, sizeof(*priv));
+		GF_FREE(priv);
+	}
+}
+
+int32_t reconfigure (xlator_t *this, dict_t *options)
+{
+	int32_t ret = -1;
+	crypt_private_t *priv = NULL;
+
+	GF_VALIDATE_OR_GOTO ("crypt", this, error);
+	GF_VALIDATE_OR_GOTO (this->name, this->private, error);
+	GF_VALIDATE_OR_GOTO (this->name, options, error);
+
+	priv = this->private;
+
+	ret = master_set_block_size(this, priv, options);
+	if (ret) {
+		gf_log("this->name", GF_LOG_ERROR,
+		       "Failed to reconfure block size");
+		goto error;
+	}
+	ret = master_set_data_key_size(this, priv, options);
+	if (ret) {
+		gf_log("this->name", GF_LOG_ERROR,
+		       "Failed to reconfure data key size");
+		goto error;
+	}
+	return 0;
+ error:
+	return ret;
+}
+
+int32_t init(xlator_t *this)
+{
+	int32_t ret;
+
+	if (!this->children || this->children->next) {
+		gf_log ("crypt", GF_LOG_ERROR,
+			"FATAL: crypt should have exactly one child");
+		return EINVAL;
+	}
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+	ret = crypt_alloc_private(this);
+	if (ret)
+		return ret;
+	ret = crypt_init_xlator(this);
+	if (ret)
+		goto error;
+	this->local_pool = mem_pool_new(crypt_local_t, 64);
+        if (!this->local_pool) {
+		gf_log(this->name, GF_LOG_ERROR,
+		       "failed to create local_t's memory pool");
+		ret = ENOMEM;
+                goto error;
+        }
+	gf_log ("crypt", GF_LOG_INFO, "crypt xlator loaded");
+	return 0;
+ error:
+	crypt_free_private(this);
+	return ret;
+}
+
+void fini (xlator_t *this)
+{
+	crypt_free_private(this);
+}
+
+struct xlator_fops fops = {
+	.readv        = crypt_readv,
+	.writev       = crypt_writev,
+	.truncate     = crypt_truncate,
+	.ftruncate    = crypt_ftruncate,
+	.setxattr     = crypt_setxattr,
+	.fsetxattr    = crypt_fsetxattr,
+	.link         = crypt_link,
+	.unlink       = crypt_unlink,
+	.rename       = crypt_rename,
+	.open         = crypt_open,
+	.create       = crypt_create,
+	.stat         = crypt_stat,
+	.fstat        = crypt_fstat,
+	.lookup       = crypt_lookup,
+	.readdirp     = crypt_readdirp,
+	.access       = crypt_access   
+};
+
+struct xlator_cbks cbks = {
+	.forget       = crypt_forget
+};
+
+struct volume_options options[] = {
+	{ .key = {"master-key"},
+	  .type = GF_OPTION_TYPE_PATH,
+	  .description = "Pathname of regular file which contains master volume key"
+	},
+	{ .key = {"data-key-size"},
+	  .type = GF_OPTION_TYPE_SIZET,
+	  .description = "Data key size (bits)",
+	  .min = 256,
+	  .max = 512,
+	  .default_value = "256",
+	},
+	{ .key = {"block-size"},
+	  .type = GF_OPTION_TYPE_SIZET,
+	  .description = "Atom size (bits)",
+	  .min = 512,
+	  .max = 4096,
+	  .default_value = "4096"
+	},
+	{ .key  = {NULL} },
+};
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/crypt.h b/xlators/encryption/crypt/src/crypt.h
new file mode 100644
index 000000000..01a8542ab
--- /dev/null
+++ b/xlators/encryption/crypt/src/crypt.h
@@ -0,0 +1,899 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef __CRYPT_H__
+#define __CRYPT_H__
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+#include <openssl/aes.h>
+#include <openssl/evp.h>
+#include <openssl/sha.h>
+#include <openssl/hmac.h>
+#include <openssl/cmac.h>
+#include <openssl/modes.h>
+#include "crypt-mem-types.h"
+
+#define CRYPT_XLATOR_ID  (0)
+
+#define MAX_IOVEC_BITS (3)
+#define MAX_IOVEC (1 << MAX_IOVEC_BITS)
+#define KEY_FACTOR_BITS  (6)
+
+#define DEBUG_CRYPT (0)
+#define TRIVIAL_TFM (0)
+
+#define CRYPT_MIN_BLOCK_BITS (9)
+#define CRYPT_MAX_BLOCK_BITS (12)
+
+#define MASTER_VOL_KEY_SIZE (32)
+#define NMTD_VOL_KEY_SIZE (16)
+
+struct crypt_key {
+	uint32_t len;
+	const char *label;	
+};
+
+/*
+ * Add new key types to the end of this
+ * enumeration but before LAST_KEY_TYPE
+ */
+typedef enum {
+	MASTER_VOL_KEY,
+	NMTD_VOL_KEY,
+	NMTD_LINK_KEY,
+	EMTD_FILE_KEY,
+	DATA_FILE_KEY_256,
+	DATA_FILE_KEY_512,
+	LAST_KEY_TYPE
+}crypt_key_type;
+
+struct kderive_context {
+	const unsigned char *pkey;/* parent key */
+	uint32_t pkey_len;        /* parent key size, bits */
+	uint32_t ckey_len;        /* child key size, bits */
+	unsigned char *fid;       /* fixed input data, NIST 800-108, 5.1 */
+	uint32_t fid_len;         /* fid len, bytes */
+	unsigned char *out;       /* contains child keying material */
+	uint32_t out_len;         /* out len, bytes */
+};
+
+typedef enum {
+	DATA_ATOM,
+	HOLE_ATOM,
+	LAST_DATA_TYPE
+}atom_data_type;
+
+typedef enum {
+	HEAD_ATOM,
+	TAIL_ATOM,
+	FULL_ATOM,
+	LAST_LOCALITY_TYPE
+}atom_locality_type;
+
+typedef enum {
+	MTD_CREATE,
+	MTD_APPEND,
+	MTD_OVERWRITE,
+	MTD_CUT,
+	MTD_LAST_OP
+} mtd_op_t;
+
+struct xts128_context {
+	void      *key1, *key2;
+	block128_f block1,block2;
+};
+
+struct object_cipher_info {
+	cipher_alg_t  o_alg;
+        cipher_mode_t o_mode;
+	uint32_t      o_block_bits;
+        uint32_t      o_dkey_size; /* raw data key size in bits */
+	union {
+		struct {
+			unsigned char ivec[16];
+			AES_KEY dkey[2];
+			AES_KEY tkey;    /* key used for tweaking */
+			XTS128_CONTEXT xts;
+		} aes_xts;
+	} u;
+};
+
+struct master_cipher_info {
+	/*
+	 * attributes inherited by newly created regular files
+	 */
+	cipher_alg_t  m_alg;
+        cipher_mode_t m_mode;
+	uint32_t      m_block_bits;
+        uint32_t      m_dkey_size; /* raw key size in bits */
+	/*
+	 * master key
+	 */
+	unsigned char m_key[MASTER_VOL_KEY_SIZE];
+	/* 
+	 * volume key for oid authentication
+	 */
+	unsigned char m_nmtd_key[NMTD_VOL_KEY_SIZE];
+};
+
+/*
+* This info is not changed during file's life
+ */
+struct crypt_inode_info {
+#if DEBUG_CRYPT
+	loc_t *loc; /* pathname that the file has been
+		       opened, or created with */
+#endif
+	uint16_t nr_minor;
+	uuid_t oid;
+	struct object_cipher_info cinfo;
+};
+
+/*
+ * this should locate in secure memory
+ */
+typedef struct {
+	struct master_cipher_info master;
+} crypt_private_t;
+
+static inline struct master_cipher_info *get_master_cinfo(crypt_private_t *priv)
+{
+	return &priv->master;
+}
+
+static inline struct object_cipher_info *get_object_cinfo(struct crypt_inode_info
+							 *info)
+{
+	return &info->cinfo;
+}
+
+/*
+ * this describes layouts and properties
+ * of atoms in an aligned vector
+ */
+struct avec_config {
+	uint32_t atom_size;
+	atom_data_type type;
+	size_t orig_size;
+	off_t orig_offset;
+	size_t expanded_size;
+	off_t aligned_offset;
+
+	uint32_t off_in_head;
+	uint32_t off_in_tail;
+	uint32_t gap_in_tail;
+	uint32_t nr_full_blocks;
+
+	struct iovec *avec; /* aligned vector */
+	uint32_t acount; /* number of avec components. The same
+			  * as number of occupied logical blocks */
+	char **pool;
+	uint32_t blocks_in_pool;
+	uint32_t cursor; /* makes sense only for ordered writes,
+			  * so there is no races on this counter.
+			  *
+			  * Cursor is per-config object, we don't
+			  * reset cursor for atoms of different
+			  * localities (head, tail, full)
+			  */
+};
+
+
+typedef struct {
+	glusterfs_fop_t fop; /* code of FOP this local info built for */
+	fd_t *fd;
+	inode_t *inode;
+	loc_t *loc;
+	int32_t mac_idx;
+	loc_t *newloc;
+	int32_t flags;
+	int32_t wbflags;
+	struct crypt_inode_info *info;
+	struct iobref *iobref;
+	struct iobref *iobref_data;
+	off_t offset;
+
+	uint64_t old_file_size; /* per FOP, retrieved under lock held */
+	uint64_t cur_file_size; /* per iteration, before issuing IOs */
+	uint64_t new_file_size; /* per iteration, after issuing IOs */
+
+	uint64_t io_offset; /* offset of IOs issued per iteration */
+	uint64_t io_offset_nopad; /* offset of user's data in the atom */
+	uint32_t io_size; /* size of IOs issued per iteration */
+	uint32_t io_size_nopad; /* size of user's data in the IOs */
+	uint32_t eof_padding_size; /* size od EOF padding in the IOs */
+
+	gf_lock_t call_lock; /* protect nr_calls from many cbks */
+	int32_t nr_calls;
+
+	atom_data_type active_setup; /* which setup (hole or date)
+					is currently active */
+	/* data setup */
+	struct avec_config data_conf;
+
+	/* hole setup */
+	int hole_conv_in_proggress;
+	gf_lock_t hole_lock; /* protect hole config from many cbks */
+	int hole_handled;
+	struct avec_config hole_conf;
+	struct iatt buf;
+	struct iatt prebuf;
+	struct iatt postbuf;
+	struct iatt *prenewparent;
+	struct iatt *postnewparent;
+	int32_t op_ret;
+	int32_t op_errno;
+	int32_t rw_count; /* total read or written */
+	gf_lock_t rw_count_lock; /* protect the counter above */
+	unsigned char *format; /* for create, update format string */
+	uint32_t format_size;
+	uint32_t msgflags; /* messages for crypt_open() */
+	dict_t *xdata;
+	dict_t *xattr;
+	struct iovec vec; /* contains last file's atom for
+			     read-prune-write sequence */
+	gf_boolean_t custom_mtd;
+	/*
+	 * the next 3 fields are used by readdir and friends
+	 */
+	gf_dirent_t *de; /* directory entry */
+	char *de_path; /* pathname of directory entry */
+	uint32_t de_prefix_len; /* lenght of the parent's pathname */
+	gf_dirent_t *entries;
+
+	uint32_t update_disk_file_size:1;
+} crypt_local_t;
+
+/* This represents a (read)modify-write atom */
+struct rmw_atom {
+	atom_locality_type locality;
+	/*
+	 * read-modify-write sequence of the atom
+	 */
+	int32_t (*rmw)(call_frame_t *frame,
+		       void *cookie,
+		       xlator_t *this,
+		       int32_t op_ret,
+		       int32_t op_errno,
+		       struct iovec *vec,
+		       int32_t count,
+		       struct iatt *stbuf,
+		       struct iobref *iobref,
+		       dict_t *xdata);
+	/*
+	 * offset of the logical block in a file
+	 */
+	loff_t (*offset_at)(call_frame_t *frame,
+			    struct object_cipher_info *object);
+	/*
+	 * IO offset in an atom
+	 */
+	uint32_t (*offset_in)(call_frame_t *frame,
+			      struct object_cipher_info *object);
+	/*
+	 * number of bytes of plain text of this atom that user
+	 * wants to read/write.
+	 * It can be smaller than atom_size in the case of head
+	 * or tail atoms.
+	 */
+	uint32_t (*io_size_nopad)(call_frame_t *frame,
+				  struct object_cipher_info *object);
+	/*
+	 * which iovec represents the atom
+	 */
+	struct iovec *(*get_iovec)(call_frame_t *frame, uint32_t count);
+	/*
+	 * how many bytes of partial block should be uptodated by
+	 * reading from disk.
+	 * This is used to perform a read component of RMW (read-modify-write).
+	 */
+	uint32_t (*count_to_uptodate)(call_frame_t *frame, struct object_cipher_info *object);
+	struct avec_config *(*get_config)(call_frame_t *frame);
+};
+
+struct data_cipher_alg {
+	gf_boolean_t atomic; /* true means that algorithm requires
+				to pad data before cipher transform */
+	gf_boolean_t should_pad; /* true means that algorithm requires
+				    to pad the end of file with extra-data */
+	uint32_t blkbits; /* blksize = 1 << blkbits */
+	/*
+	 * any preliminary sanity checks goes here
+	 */
+	int32_t (*init)(void);
+	/*
+	 * set alg-mode specific inode info
+	 */
+	int32_t (*set_private)(struct crypt_inode_info *info,
+			       struct master_cipher_info *master);
+	/*
+	 * check alg-mode specific data key
+	 */
+	int32_t (*check_key)(uint32_t key_size);
+	void (*set_iv)(off_t offset, struct object_cipher_info *object);
+	int32_t (*encrypt)(const unsigned char *from, unsigned char *to,
+			   size_t length, off_t offset, const int enc,
+			   struct object_cipher_info *object);
+};
+
+/*
+ * version-dependent metadata loader
+ */
+struct crypt_mtd_loader {
+	/*
+	 * return core format size
+	 */
+	size_t (*format_size)(mtd_op_t op, size_t old_size);
+	/*
+	 * pack version-specific metadata of an object
+	 * at ->create()
+	 */
+	int32_t (*create_format)(unsigned char *wire,
+				 loc_t *loc,
+				 struct crypt_inode_info *info,
+				 struct master_cipher_info *master);
+	/*
+	 * extract version-specific metadata of an object
+	 * at ->open() time
+	 */
+	int32_t (*open_format)(unsigned char *wire,
+			       int32_t len,
+			       loc_t *loc,
+			       struct crypt_inode_info *info,
+			       struct master_cipher_info *master,
+			       crypt_local_t *local,
+			       gf_boolean_t load_info);
+	int32_t (*update_format)(unsigned char *new,
+				 unsigned char *old,
+				 size_t old_len,
+				 int32_t mac_idx,
+				 mtd_op_t op,
+				 loc_t *loc,
+				 struct crypt_inode_info *info,
+				 struct master_cipher_info *master,
+				 crypt_local_t *local);
+};
+
+typedef int32_t (*end_writeback_handler_t)(call_frame_t *frame,
+					   void *cookie,
+					   xlator_t *this,
+					   int32_t op_ret,
+					   int32_t op_errno,
+					   struct iatt *prebuf,
+					   struct iatt *postbuf,
+					   dict_t *xdata);
+typedef void (*linkop_wind_handler_t)(call_frame_t *frame, xlator_t *this);
+typedef void (*linkop_unwind_handler_t)(call_frame_t *frame);
+
+
+/* Declarations */
+
+/* keys.c */
+extern struct crypt_key crypt_keys[LAST_KEY_TYPE];
+int32_t get_nmtd_vol_key(struct master_cipher_info *master);
+int32_t get_nmtd_link_key(loc_t *loc,
+			  struct master_cipher_info *master,
+			  unsigned char *result);
+int32_t get_emtd_file_key(struct crypt_inode_info *info,
+			  struct master_cipher_info *master,
+			  unsigned char *result);
+int32_t get_data_file_key(struct crypt_inode_info *info,
+			  struct master_cipher_info *master,
+			  uint32_t keysize,
+			  unsigned char *key);
+/* data.c */
+extern struct data_cipher_alg data_cipher_algs[LAST_CIPHER_ALG][LAST_CIPHER_MODE];
+void encrypt_aligned_iov(struct object_cipher_info *object,
+			 struct iovec *vec,
+			 int count,
+			 off_t off);
+void decrypt_aligned_iov(struct object_cipher_info *object,
+			 struct iovec *vec,
+			 int count,
+			 off_t off);
+int32_t align_iov_by_atoms(xlator_t *this,
+			   crypt_local_t *local,
+			   struct object_cipher_info *object,
+			   struct iovec *vec /* input vector */,
+			   int32_t count /* number of vec components */,
+			   struct iovec *avec /* aligned vector */,
+			   char **blocks /* pool of blocks */,
+			   uint32_t *blocks_allocated,
+			   struct avec_config *conf);
+int32_t set_config_avec_data(xlator_t *this,
+			     crypt_local_t *local,
+			     struct avec_config *conf,
+			     struct object_cipher_info *object,
+			     struct iovec *vec,
+			     int32_t vec_count);
+int32_t set_config_avec_hole(xlator_t *this,
+			     crypt_local_t *local,
+			     struct avec_config *conf,
+			     struct object_cipher_info *object,
+			     glusterfs_fop_t fop);
+void set_gap_at_end(call_frame_t *frame, struct object_cipher_info *object,
+		    struct avec_config *conf, atom_data_type dtype);
+void set_config_offsets(call_frame_t *frame,
+			xlator_t *this,
+			uint64_t offset,
+			uint64_t count,
+			atom_data_type dtype,
+			int32_t setup_gap_in_tail);
+
+/* metadata.c */
+extern struct crypt_mtd_loader mtd_loaders [LAST_MTD_LOADER];
+
+int32_t alloc_format(crypt_local_t *local, size_t size);
+int32_t alloc_format_create(crypt_local_t *local);
+void free_format(crypt_local_t *local);
+size_t format_size(mtd_op_t op, size_t old_size);
+size_t new_format_size(void);
+int32_t open_format(unsigned char *str, int32_t len, loc_t *loc,
+		    struct crypt_inode_info *info,
+		    struct master_cipher_info *master, crypt_local_t *local,
+		    gf_boolean_t load_info);
+int32_t update_format(unsigned char *new, unsigned char *old,
+		      size_t old_len, int32_t mac_idx, mtd_op_t op, loc_t *loc,
+		      struct crypt_inode_info *info,
+		      struct master_cipher_info *master,
+		      crypt_local_t *local);
+int32_t create_format(unsigned char *wire,
+		      loc_t *loc,
+		      struct crypt_inode_info *info,
+		      struct master_cipher_info *master);
+
+/* atom.c */
+struct rmw_atom *atom_by_types(atom_data_type data,
+			       atom_locality_type locality);
+void submit_partial(call_frame_t *frame,
+		    xlator_t *this,
+		    fd_t *fd,
+		    atom_locality_type ltype);
+void submit_full(call_frame_t *frame, xlator_t *this);
+
+/* crypt.c */
+
+end_writeback_handler_t dispatch_end_writeback(glusterfs_fop_t fop);
+static size_t iovec_get_size(struct iovec *vec, uint32_t count);
+void set_local_io_params_writev(call_frame_t *frame,
+				struct object_cipher_info *object,
+				struct rmw_atom *atom, off_t io_offset,
+				uint32_t io_size);
+void link_wind(call_frame_t *frame, xlator_t *this);
+void unlink_wind(call_frame_t *frame, xlator_t *this);
+void link_unwind(call_frame_t *frame);
+void unlink_unwind(call_frame_t *frame);
+void rename_wind(call_frame_t *frame, xlator_t *this);
+void rename_unwind(call_frame_t *frame);
+
+/* Inline functions */
+
+static inline size_t iovec_get_size(struct iovec *vec, uint32_t count)
+{
+	int i;
+	size_t size = 0;
+	for (i = 0; i < count; i++)
+		size += vec[i].iov_len;
+	return size;
+}
+
+static inline int32_t crypt_xlator_id(void)
+{
+	return CRYPT_XLATOR_ID;
+}
+
+static inline mtd_loader_id current_mtd_loader(void)
+{
+	return MTD_LOADER_V1;
+}
+
+static inline uint32_t master_key_size (void)
+{
+	return crypt_keys[MASTER_VOL_KEY].len >> 3;
+}
+
+static inline uint32_t nmtd_vol_key_size (void)
+{
+	return crypt_keys[NMTD_VOL_KEY].len >> 3;
+}
+
+static inline uint32_t alg_mode_blkbits(cipher_alg_t alg,
+					cipher_mode_t mode)
+{
+	return data_cipher_algs[alg][mode].blkbits;
+}
+
+static inline uint32_t alg_mode_blksize(cipher_alg_t alg,
+					cipher_mode_t mode)
+{
+	return 1 << alg_mode_blkbits(alg, mode);
+}
+
+static inline gf_boolean_t alg_mode_atomic(cipher_alg_t alg,
+					   cipher_mode_t mode)
+{
+	return data_cipher_algs[alg][mode].atomic;
+}
+
+static inline gf_boolean_t alg_mode_should_pad(cipher_alg_t alg,
+					       cipher_mode_t mode)
+{
+	return data_cipher_algs[alg][mode].should_pad;
+}
+
+static inline uint32_t master_alg_blksize(struct master_cipher_info *mr)
+{
+	return alg_mode_blksize(mr->m_alg, mr->m_mode);
+}
+
+static inline uint32_t master_alg_blkbits(struct master_cipher_info *mr)
+{
+	return alg_mode_blkbits(mr->m_alg, mr->m_mode);
+}
+
+static inline gf_boolean_t master_alg_atomic(struct master_cipher_info *mr)
+{
+	return alg_mode_atomic(mr->m_alg, mr->m_mode);
+}
+
+static inline gf_boolean_t master_alg_should_pad(struct master_cipher_info *mr)
+{
+	return alg_mode_should_pad(mr->m_alg, mr->m_mode);
+}
+
+static inline uint32_t object_alg_blksize(struct object_cipher_info *ob)
+{
+	return alg_mode_blksize(ob->o_alg, ob->o_mode);
+}
+
+static inline uint32_t object_alg_blkbits(struct object_cipher_info *ob)
+{
+	return alg_mode_blkbits(ob->o_alg, ob->o_mode);
+}
+
+static inline gf_boolean_t object_alg_atomic(struct object_cipher_info *ob)
+{
+	return alg_mode_atomic(ob->o_alg, ob->o_mode);
+}
+
+static inline gf_boolean_t object_alg_should_pad(struct object_cipher_info *ob)
+{
+	return alg_mode_should_pad(ob->o_alg, ob->o_mode);
+}
+
+static inline uint32_t aes_raw_key_size(struct master_cipher_info *master)
+{
+	return master->m_dkey_size >> 3;
+}
+
+static inline struct avec_config *get_hole_conf(call_frame_t *frame)
+{
+	return &(((crypt_local_t *)frame->local)->hole_conf);
+}
+
+static inline struct avec_config *get_data_conf(call_frame_t *frame)
+{
+	return &(((crypt_local_t *)frame->local)->data_conf);
+}
+
+static inline int32_t get_atom_bits (struct object_cipher_info *object)
+{
+	return object->o_block_bits;
+}
+
+static inline int32_t get_atom_size (struct object_cipher_info *object)
+{
+	return 1 << get_atom_bits(object);
+}
+
+static inline int32_t has_head_block(struct avec_config *conf)
+{
+	return conf->off_in_head ||
+		(conf->acount == 1 && conf->off_in_tail);
+}
+
+static inline int32_t has_tail_block(struct avec_config *conf)
+{
+	return conf->off_in_tail && conf->acount > 1;
+}
+
+static inline int32_t has_full_blocks(struct avec_config *conf)
+{
+	return conf->nr_full_blocks;
+}
+
+static inline int32_t should_submit_head_block(struct avec_config *conf)
+{
+	return has_head_block(conf) && (conf->cursor == 0);
+}
+
+static inline int32_t should_submit_tail_block(struct avec_config *conf)
+{
+	return has_tail_block(conf) && (conf->cursor == conf->acount - 1);
+}
+
+static inline int32_t should_submit_full_block(struct avec_config *conf)
+{
+	uint32_t start = has_head_block(conf) ? 1 : 0;
+
+	return has_full_blocks(conf) &&
+		conf->cursor >= start &&
+		conf->cursor < start + conf->nr_full_blocks;
+}
+
+#if DEBUG_CRYPT
+static inline void crypt_check_input_len(size_t len,
+					 struct object_cipher_info *object)
+{
+	if (object_alg_should_pad(object) && (len & (object_alg_blksize(object) - 1)))
+		gf_log ("crypt", GF_LOG_DEBUG, "bad input len: %d", (int)len);
+}
+
+static inline void check_head_block(struct avec_config *conf)
+{
+	if (!has_head_block(conf))
+		gf_log("crypt", GF_LOG_DEBUG, "not a head atom");
+}
+
+static inline void check_tail_block(struct avec_config *conf)
+{
+	if (!has_tail_block(conf))
+		gf_log("crypt", GF_LOG_DEBUG, "not a tail atom");
+}
+
+static inline void check_full_block(struct avec_config *conf)
+{
+	if (!has_full_blocks(conf))
+		gf_log("crypt", GF_LOG_DEBUG, "not a full atom");
+}
+
+static inline void check_cursor_head(struct avec_config *conf)
+{
+	if (!has_head_block(conf))
+		gf_log("crypt",
+		       GF_LOG_DEBUG, "Illegal call of head atom method");
+	else if (conf->cursor != 0)
+		gf_log("crypt",
+		       GF_LOG_DEBUG, "Cursor (%d) is not at head atom",
+		       conf->cursor);
+}
+
+static inline void check_cursor_full(struct avec_config *conf)
+{
+	if (!has_full_blocks(conf))
+		gf_log("crypt",
+		       GF_LOG_DEBUG, "Illegal call of full atom method");
+	if (has_head_block(conf) && (conf->cursor == 0))
+		gf_log("crypt",
+		       GF_LOG_DEBUG, "Cursor is not at full atom");
+}
+
+/*
+ * FIXME: use avec->iov_len to check setup
+ */
+static inline int data_local_invariant(crypt_local_t *local)
+{
+	return 0;
+}
+
+#else
+#define crypt_check_input_len(len, object) noop
+#define check_head_block(conf) noop
+#define check_tail_block(conf) noop
+#define check_full_block(conf) noop
+#define check_cursor_head(conf) noop
+#define check_cursor_full(conf) noop
+
+#endif /* DEBUG_CRYPT */
+
+static inline struct avec_config *conf_by_type(call_frame_t *frame,
+					       atom_data_type dtype)
+{
+	struct avec_config *conf = NULL;
+
+	switch (dtype) {
+	case HOLE_ATOM:
+		conf = get_hole_conf(frame);
+		break;
+	case DATA_ATOM:
+		conf = get_data_conf(frame);
+		break;
+	default:
+		gf_log("crypt", GF_LOG_DEBUG, "bad atom type");
+	}
+	return conf;
+}
+
+static inline uint32_t nr_calls_head(struct avec_config *conf)
+{
+	return has_head_block(conf) ? 1 : 0;
+}
+
+static inline uint32_t nr_calls_tail(struct avec_config *conf)
+{
+	return has_tail_block(conf) ? 1 : 0;
+}
+
+static inline uint32_t nr_calls_full(struct avec_config *conf)
+{
+	switch(conf->type) {
+	case HOLE_ATOM:
+		return has_full_blocks(conf);
+	case DATA_ATOM:
+		return has_full_blocks(conf) ?
+			logical_blocks_occupied(0,
+						conf->nr_full_blocks,
+						MAX_IOVEC_BITS)	: 0;
+	default:
+		gf_log("crypt", GF_LOG_DEBUG, "bad atom data type");
+		return 0;
+	}
+}
+
+static inline uint32_t nr_calls(struct avec_config *conf)
+{
+	return nr_calls_head(conf) + nr_calls_tail(conf) + nr_calls_full(conf);
+}
+
+static inline uint32_t nr_calls_data(call_frame_t *frame)
+{
+	return nr_calls(get_data_conf(frame));
+}
+
+static inline uint32_t nr_calls_hole(call_frame_t *frame)
+{
+	return nr_calls(get_hole_conf(frame));
+}
+
+static inline void get_one_call_nolock(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+
+	++local->nr_calls;
+
+	//gf_log("crypt", GF_LOG_DEBUG, "get %d calls", 1);
+}
+
+static inline void get_one_call(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+
+	LOCK(&local->call_lock);
+	get_one_call_nolock(frame);
+	UNLOCK(&local->call_lock);
+}
+
+static inline void get_nr_calls_nolock(call_frame_t *frame, int32_t nr)
+{
+	crypt_local_t *local = frame->local;
+
+	local->nr_calls += nr;
+
+	//gf_log("crypt", GF_LOG_DEBUG, "get %d calls", nr);
+}
+
+static inline void get_nr_calls(call_frame_t *frame, int32_t nr)
+{
+	crypt_local_t *local = frame->local;
+
+	LOCK(&local->call_lock);
+	get_nr_calls_nolock(frame, nr);
+	UNLOCK(&local->call_lock);
+}
+
+static inline int put_one_call(crypt_local_t *local)
+{
+	uint32_t last = 0;
+
+	LOCK(&local->call_lock);
+	if (--local->nr_calls == 0)
+		last = 1;
+
+	//gf_log("crypt", GF_LOG_DEBUG, "put %d calls", 1);
+
+	UNLOCK(&local->call_lock);
+	return last;
+}
+
+static inline int is_appended_write(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = get_data_conf(frame);
+
+	return conf->orig_offset + conf->orig_size > local->old_file_size;
+}
+
+static inline int is_ordered_mode(call_frame_t *frame)
+{
+#if 0
+	crypt_local_t *local = frame->local;
+	return local->fop == GF_FOP_FTRUNCATE ||
+		(local->fop == GF_FOP_WRITE && is_appended_write(frame));
+#endif
+	return 1;
+}
+
+static inline int32_t hole_conv_completed(crypt_local_t *local)
+{
+	struct avec_config *conf = &local->hole_conf;
+	return conf->cursor == conf->acount;
+}
+
+static inline int32_t data_write_in_progress(crypt_local_t *local)
+{
+	return local->active_setup == DATA_ATOM;
+}
+
+static inline int32_t parent_is_crypt_xlator(call_frame_t *frame,
+					     xlator_t *this)
+{
+	return frame->parent->this == this;
+}
+
+static inline linkop_wind_handler_t linkop_wind_dispatch(glusterfs_fop_t fop)
+{
+	switch(fop){
+	case GF_FOP_LINK:
+		return link_wind;
+	case GF_FOP_UNLINK:
+		return unlink_wind;
+	case GF_FOP_RENAME:
+		return rename_wind;
+	default:
+		gf_log("crypt", GF_LOG_ERROR, "Bad link operation %d", fop);
+		return NULL;
+	}
+}
+
+static inline linkop_unwind_handler_t linkop_unwind_dispatch(glusterfs_fop_t fop)
+{
+	switch(fop){
+	case GF_FOP_LINK:
+		return link_unwind;
+	case GF_FOP_UNLINK:
+		return unlink_unwind;
+	case GF_FOP_RENAME:
+		return rename_unwind;
+	default:
+		gf_log("crypt", GF_LOG_ERROR, "Bad link operation %d", fop);
+		return NULL;		
+	}	
+}
+
+static inline mtd_op_t linkop_mtdop_dispatch(glusterfs_fop_t fop)
+{
+	switch (fop) {
+	case GF_FOP_LINK:
+		return MTD_APPEND;
+	case GF_FOP_UNLINK:
+		return MTD_CUT;
+	case GF_FOP_RENAME:
+		return MTD_OVERWRITE;
+	default:
+		gf_log("crypt", GF_LOG_WARNING, "Bad link operation %d", fop);
+		return MTD_LAST_OP;
+	}
+}
+
+#endif /* __CRYPT_H__ */
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/data.c b/xlators/encryption/crypt/src/data.c
new file mode 100644
index 000000000..762fa554a
--- /dev/null
+++ b/xlators/encryption/crypt/src/data.c
@@ -0,0 +1,769 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "defaults.h"
+#include "crypt-common.h"
+#include "crypt.h"
+
+static void set_iv_aes_xts(off_t offset, struct object_cipher_info *object)
+{
+	unsigned char *ivec;
+
+	ivec = object->u.aes_xts.ivec;
+
+	/* convert the tweak into a little-endian byte
+	 * array (IEEE P1619/D16, May 2007, section 5.1)
+	 */
+
+	*((uint64_t *)ivec) = htole64(offset);
+
+	/* ivec is padded with zeroes */
+}
+
+static int32_t aes_set_keys_common(unsigned char *raw_key, uint32_t key_size,
+				   AES_KEY *keys)
+{
+	int32_t ret;
+
+	ret = AES_set_encrypt_key(raw_key,
+				  key_size,
+				  &keys[AES_ENCRYPT]);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Set encrypt key failed");
+		return ret;
+	}
+	ret = AES_set_decrypt_key(raw_key,
+				  key_size,
+				  &keys[AES_DECRYPT]);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Set decrypt key failed");
+		return ret;
+	}
+	return 0;
+}
+
+/*
+ * set private cipher info for xts mode
+ */
+static int32_t set_private_aes_xts(struct crypt_inode_info *info,
+				   struct master_cipher_info *master)
+{
+	int ret;
+	struct object_cipher_info *object = get_object_cinfo(info);
+	unsigned char *data_key;
+	uint32_t subkey_size;
+
+	/* init tweak value */
+	memset(object->u.aes_xts.ivec, 0, 16);
+
+	data_key = GF_CALLOC(1, object->o_dkey_size, gf_crypt_mt_key);
+	if (!data_key)
+		return ENOMEM;
+
+	/*
+	 * retrieve data keying meterial
+	 */
+	ret = get_data_file_key(info, master, object->o_dkey_size, data_key);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Failed to retrieve data key");
+		GF_FREE(data_key);
+		return ret;
+	}
+	/*
+	 * parse compound xts key
+	 */
+	subkey_size = object->o_dkey_size >> 4; /* (xts-key-size-in-bytes / 2) */
+	/*
+	 * install key for data encryption
+	 */
+	ret = aes_set_keys_common(data_key,
+				  subkey_size << 3, object->u.aes_xts.dkey);
+	if (ret) {
+		GF_FREE(data_key);
+		return ret;
+	}
+	/*
+	 * set up key used to encrypt tweaks
+	 */
+	ret = AES_set_encrypt_key(data_key + subkey_size,
+				  object->o_dkey_size / 2,
+				  &object->u.aes_xts.tkey);
+	if (ret < 0)
+		gf_log("crypt", GF_LOG_ERROR, "Set tweak key failed");
+
+	GF_FREE(data_key);
+	return ret;
+}
+
+static int32_t aes_xts_init(void)
+{
+	cassert(AES_BLOCK_SIZE == (1 << AES_BLOCK_BITS));
+	return 0;
+}
+
+static int32_t check_key_aes_xts(uint32_t keysize)
+{
+	switch(keysize) {
+	case 256:
+	case 512:
+		return 0;
+	default:
+		break;
+	}
+	return -1;
+}
+
+static int32_t encrypt_aes_xts(const unsigned char *from,
+			       unsigned char *to, size_t length,
+			       off_t offset, const int enc,
+			       struct object_cipher_info *object)
+{
+	XTS128_CONTEXT ctx;
+	if (enc) {
+		ctx.key1 = &object->u.aes_xts.dkey[AES_ENCRYPT];
+		ctx.block1 = (block128_f)AES_encrypt;
+	}
+	else {
+		ctx.key1 = &object->u.aes_xts.dkey[AES_DECRYPT];
+		ctx.block1 = (block128_f)AES_decrypt;
+	}
+	ctx.key2 = &object->u.aes_xts.tkey;
+	ctx.block2 = (block128_f)AES_encrypt;
+
+	return CRYPTO_xts128_encrypt(&ctx,
+				     object->u.aes_xts.ivec,
+				     from,
+				     to,
+				     length, enc);
+}
+
+/*
+ * Cipher input chunk @from of length @len;
+ * @to: result of cipher transform;
+ * @off: offset in a file (must be cblock-aligned);
+ */
+static void cipher_data(struct object_cipher_info *object,
+			char *from,
+			char *to,
+			off_t off,
+			size_t len,
+			const int enc)
+{
+	crypt_check_input_len(len, object);
+
+#if TRIVIAL_TFM && DEBUG_CRYPT
+	return;
+#endif
+	data_cipher_algs[object->o_alg][object->o_mode].set_iv(off, object);
+	data_cipher_algs[object->o_alg][object->o_mode].encrypt
+		((const unsigned char *)from,
+		 (unsigned char *)to,
+		 len,
+		 off,
+		 enc,
+		 object);
+}
+
+#define MAX_CIPHER_CHUNK (1 << 30)
+
+/*
+ * Do cipher (encryption/decryption) transform of a
+ * continuous region of memory.
+ *
+ * @len: a number of bytes to transform;
+ * @buf: data to transform;
+ * @off: offset in a file, should be block-aligned
+ *       for atomic cipher modes and ksize-aligned
+ *       for other modes).
+ * @dir: direction of transform (encrypt/decrypt).
+ */
+static void cipher_region(struct object_cipher_info *object,
+			  char *from,
+			  char *to,
+			  off_t off,
+			  size_t len,
+			  int dir)
+{
+	while (len > 0) {
+		size_t to_cipher;
+
+		to_cipher = len;
+		if (to_cipher > MAX_CIPHER_CHUNK)
+			to_cipher = MAX_CIPHER_CHUNK;
+
+		/* this will reset IV */
+		cipher_data(object,
+			    from,
+			    to,
+			    off,
+			    to_cipher,
+			    dir);
+		from += to_cipher;
+		to   += to_cipher;
+		off  += to_cipher;
+		len  -= to_cipher;
+	}
+}
+
+/*
+ * Do cipher transform (encryption/decryption) of
+ * plaintext/ciphertext represented by @vec.
+ *
+ * Pre-conditions: @vec represents a continuous piece
+ * of data in a file at offset @off to be ciphered
+ * (encrypted/decrypted).
+ * @count is the number of vec's components. All the
+ * components must be block-aligned, the caller is
+ * responsible for this. @dir is "direction" of
+ * transform (encrypt/decrypt).
+ */
+static void cipher_aligned_iov(struct object_cipher_info *object,
+			       struct iovec *vec,
+			       int count,
+			       off_t off,
+			       int32_t dir)
+{
+	int i;
+	int len = 0;
+
+	for (i = 0; i < count; i++) {
+		cipher_region(object,
+			      vec[i].iov_base,
+			      vec[i].iov_base,
+			      off + len,
+			      vec[i].iov_len,
+			      dir);
+		len += vec[i].iov_len;
+	}
+}
+
+void encrypt_aligned_iov(struct object_cipher_info *object,
+			 struct iovec *vec,
+			 int count,
+			 off_t off)
+{
+	cipher_aligned_iov(object, vec, count, off, 1);
+}
+
+void decrypt_aligned_iov(struct object_cipher_info *object,
+			 struct iovec *vec,
+			 int count,
+			 off_t off)
+{
+	cipher_aligned_iov(object, vec, count, off, 0);
+}
+
+#if DEBUG_CRYPT
+static void compound_stream(struct iovec *vec, int count, char *buf, off_t skip)
+{
+	int i;
+	int off = 0;
+	for (i = 0; i < count; i++) {
+		memcpy(buf + off,
+		       vec[i].iov_base + skip,
+		       vec[i].iov_len - skip);
+
+		off += (vec[i].iov_len - skip);
+		skip = 0;
+	}
+}
+
+static void check_iovecs(struct iovec *vec, int cnt,
+			 struct iovec *avec, int acnt, uint32_t off_in_head)
+{
+	char *s1, *s2;
+	uint32_t size, asize;
+
+	size = iovec_get_size(vec, cnt);
+	asize = iovec_get_size(avec, acnt) - off_in_head;
+	if (size != asize) {
+		gf_log("crypt", GF_LOG_DEBUG, "size %d is not eq asize %d",
+		       size, asize);
+		return;
+	}
+	s1 = GF_CALLOC(1, size, gf_crypt_mt_data);
+	if (!s1) {
+		gf_log("crypt", GF_LOG_DEBUG, "Can not allocate stream ");
+		return;
+	}
+	s2 = GF_CALLOC(1, asize, gf_crypt_mt_data);
+	if (!s2) {
+		GF_FREE(s1);
+		gf_log("crypt", GF_LOG_DEBUG, "Can not allocate stream ");
+		return;
+	}
+	compound_stream(vec, cnt, s1, 0);
+	compound_stream(avec, acnt, s2, off_in_head);
+	if (memcmp(s1, s2, size))
+		gf_log("crypt", GF_LOG_DEBUG, "chunks of different data");
+	GF_FREE(s1);
+	GF_FREE(s2);
+}
+
+#else
+#define check_iovecs(vec, count, avec, avecn, off) noop
+#endif /* DEBUG_CRYPT */
+
+static char *data_alloc_block(xlator_t *this, crypt_local_t *local,
+			      int32_t block_size)
+{
+	struct iobuf *iobuf = NULL;
+
+	iobuf = iobuf_get2(this->ctx->iobuf_pool, block_size);
+	if (!iobuf) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "Failed to get iobuf");
+		return NULL;
+	}
+	if (!local->iobref_data) {
+		local->iobref_data = iobref_new();
+		if (!local->iobref_data) {
+			gf_log("crypt", GF_LOG_ERROR,
+			       "Failed to get iobref");
+			iobuf_unref(iobuf);
+			return NULL;
+		}
+	}
+	iobref_add(local->iobref_data, iobuf);
+	return iobuf->ptr;
+}
+
+/*
+ * Compound @avec, which represent the same data
+ * chunk as @vec, but has aligned components of
+ * specified block size. Alloc blocks, if needed.
+ * In particular, incomplete head and tail blocks
+ * must be allocated.
+ * Put number of allocated blocks to @num_blocks.
+ *
+ * Example:
+ *
+ * input: data chunk represented by 4 components
+ * [AB],[BC],[CD],[DE];
+ * output: 5 logical blocks (0, 1, 2, 3, 4).
+ *
+ *   A     B       C     D               E
+ *   *-----*+------*-+---*----+--------+-*
+ *   |     ||      | |   |    |        | |
+ * *-+-----+*------+-*---+----*--------*-+------*
+ * 0        1        2        3        4
+ *
+ * 0 - incomplete compound (head);
+ * 1, 2 - full compound;
+ * 3 - full non-compound (the case of reuse);
+ * 4 - incomplete non-compound (tail).
+ */
+int32_t align_iov_by_atoms(xlator_t *this,
+			   crypt_local_t *local,
+			   struct object_cipher_info *object,
+			   struct iovec *vec /* input vector */,
+			   int32_t count /* number of vec components */,
+			   struct iovec *avec /* aligned vector */,
+			   char **blocks /* pool of blocks */,
+			   uint32_t *blocks_allocated,
+			   struct avec_config *conf)
+{
+	int vecn = 0;      /* number of the current component in vec */
+	int avecn = 0;     /* number of the current component in avec */
+	off_t vec_off = 0; /* offset in the current vec component,
+			    * i.e. the number of bytes have already
+			    * been copied */
+	int32_t block_size = get_atom_size(object);
+	size_t to_process; /* number of vec's bytes to copy and(or) re-use */
+	int32_t off_in_head = conf->off_in_head;
+
+	to_process = iovec_get_size(vec, count);
+
+	while (to_process > 0) {
+		if (off_in_head ||
+		    vec[vecn].iov_len - vec_off < block_size) {
+			/*
+			 * less than block_size:
+			 * the case of incomplete (head or tail),
+			 * or compound block
+			 */
+			size_t copied = 0;
+			/*
+			 * populate the pool with a new block
+			 */
+			blocks[*blocks_allocated] = data_alloc_block(this,
+								    local,
+								    block_size);
+			if (!blocks[*blocks_allocated])
+				return -ENOMEM;
+			memset(blocks[*blocks_allocated], 0, off_in_head);
+			/*
+			 * fill the block with vec components
+			 */
+			do {
+				size_t to_copy;
+
+				to_copy = vec[vecn].iov_len - vec_off;
+				if (to_copy > block_size - off_in_head)
+					to_copy = block_size - off_in_head;
+
+				memcpy(blocks[*blocks_allocated] + off_in_head + copied,
+				       vec[vecn].iov_base + vec_off,
+				       to_copy);
+
+				copied += to_copy;
+				to_process -= to_copy;
+
+				vec_off += to_copy;
+				if (vec_off == vec[vecn].iov_len) {
+					/* finished with this vecn */
+					vec_off = 0;
+					vecn++;
+				}
+			} while (copied < (block_size - off_in_head) && to_process > 0);
+			/*
+			 * update avec
+			 */
+			avec[avecn].iov_len = off_in_head + copied;
+			avec[avecn].iov_base = blocks[*blocks_allocated];
+
+			(*blocks_allocated)++;
+			off_in_head = 0;
+		} else {
+			/*
+			 * the rest of the current vec component
+			 * is not less than block_size, so reuse
+			 * the memory buffer of the component.
+			 */
+			size_t to_reuse;
+			to_reuse = (to_process > block_size ?
+				    block_size :
+				    to_process);
+			avec[avecn].iov_len = to_reuse;
+			avec[avecn].iov_base = vec[vecn].iov_base + vec_off;
+
+			vec_off += to_reuse;
+			if (vec_off == vec[vecn].iov_len) {
+				/* finished with this vecn */
+				vec_off = 0;
+				vecn++;
+			}
+			to_process -= to_reuse;
+		}
+		avecn++;
+	}
+	check_iovecs(vec, count, avec, avecn, conf->off_in_head);
+	return 0;
+}
+
+/*
+ * allocate and setup aligned vector for data submission
+ * Pre-condition: @conf is set.
+ */
+int32_t set_config_avec_data(xlator_t *this,
+			     crypt_local_t *local,
+			     struct avec_config *conf,
+			     struct object_cipher_info *object,
+			     struct iovec *vec,
+			     int32_t vec_count)
+{
+	int32_t ret = ENOMEM;
+	struct iovec *avec;
+	char **pool;
+	uint32_t blocks_in_pool = 0;
+
+	conf->type = DATA_ATOM;
+
+	avec = GF_CALLOC(conf->acount, sizeof(*avec), gf_crypt_mt_iovec);
+	if (!avec)
+		return ret;
+	pool = GF_CALLOC(conf->acount, sizeof(pool), gf_crypt_mt_char);
+	if (!pool) {
+		GF_FREE(avec);
+		return ret;
+	}
+	if (!vec) {
+		/*
+		 * degenerated case: no data
+		 */
+		pool[0] = data_alloc_block(this, local, get_atom_size(object));
+		if (!pool[0])
+			goto free;
+		blocks_in_pool = 1;
+		avec->iov_base = pool[0];
+		avec->iov_len = conf->off_in_tail;
+	}
+	else {
+		ret = align_iov_by_atoms(this, local, object, vec, vec_count,
+					 avec, pool, &blocks_in_pool, conf);
+		if (ret)
+			goto free;
+	}
+	conf->avec = avec;
+	conf->pool = pool;
+	conf->blocks_in_pool = blocks_in_pool;
+	return 0;
+ free:
+	GF_FREE(avec);
+	GF_FREE(pool);
+	return ret;
+}
+
+/*
+ * allocate and setup aligned vector for hole submission
+ */
+int32_t set_config_avec_hole(xlator_t *this,
+			     crypt_local_t *local,
+			     struct avec_config *conf,
+			     struct object_cipher_info *object,
+			     glusterfs_fop_t fop)
+{
+	uint32_t i, idx;
+	struct iovec *avec;
+	char **pool;
+	uint32_t num_blocks;
+	uint32_t blocks_in_pool = 0;
+
+	conf->type = HOLE_ATOM;
+
+	num_blocks = conf->acount -
+		(conf->nr_full_blocks ? conf->nr_full_blocks - 1 : 0);
+
+	switch (fop) {
+	case GF_FOP_WRITE:
+		/*
+		 * hole goes before data
+		 */
+		if (num_blocks == 1 && conf->off_in_tail != 0)
+		/*
+		 * we won't submit a hole which fits into
+		 * a data atom: this part of hole will be
+		 * submitted with data write
+		 */
+			return 0;
+		break;
+	case GF_FOP_FTRUNCATE:
+		/*
+		 * expanding truncate, hole goes after data,
+		 * and will be submited in any case.
+ 		 */
+		break;
+	default:
+		gf_log("crypt", GF_LOG_WARNING,
+		       "bad file operation %d", fop);
+		return 0;
+	}
+	avec = GF_CALLOC(num_blocks, sizeof(*avec), gf_crypt_mt_iovec);
+	if (!avec)
+		return ENOMEM;
+	pool = GF_CALLOC(num_blocks, sizeof(pool), gf_crypt_mt_char);
+	if (!pool) {
+		GF_FREE(avec);
+		return ENOMEM;
+	}
+	for (i = 0; i < num_blocks; i++) {
+		pool[i] = data_alloc_block(this, local, get_atom_size(object));
+		if (pool[i] == NULL)
+			goto free;
+		blocks_in_pool++;
+	}
+	if (has_head_block(conf)) {
+		/* set head block */
+		idx = 0;
+		avec[idx].iov_base = pool[idx];
+		avec[idx].iov_len = get_atom_size(object);
+		memset(avec[idx].iov_base + conf->off_in_head,
+		       0,
+		       get_atom_size(object) - conf->off_in_head);
+	}
+	if (has_tail_block(conf)) {
+		/* set tail block */
+		idx = num_blocks - 1;
+		avec[idx].iov_base = pool[idx];
+		avec[idx].iov_len = get_atom_size(object);
+		memset(avec[idx].iov_base, 0, conf->off_in_tail);
+	}
+	if (has_full_blocks(conf)) {
+		/* set full block */
+		idx = conf->off_in_head ? 1 : 0;
+		avec[idx].iov_base = pool[idx];
+		avec[idx].iov_len = get_atom_size(object);
+		/*
+		 * since we re-use the buffer,
+		 * zeroes will be set every time
+		 * before encryption, see submit_full()
+		 */
+	}
+	conf->avec = avec;
+	conf->pool = pool;
+	conf->blocks_in_pool = blocks_in_pool;
+	return 0;
+ free:
+	GF_FREE(avec);
+	GF_FREE(pool);
+	return ENOMEM;
+}
+
+/* A helper for setting up config of partial atoms (which
+ * participate in read-modify-write sequence).
+ *
+ * Calculate and setup precise amount of "extra-bytes"
+ * that should be uptodated at the end of partial (not
+ * necessarily tail!) block.
+ *
+ * Pre-condition: local->old_file_size is valid!
+ * @conf contains setup, which is enough for correct calculation
+ * of has_tail_block(), ->get_offset().
+ */
+void set_gap_at_end(call_frame_t *frame, struct object_cipher_info *object,
+		    struct avec_config *conf, atom_data_type dtype)
+{
+	uint32_t to_block;
+	crypt_local_t *local = frame->local;
+	uint64_t old_file_size = local->old_file_size;
+	struct rmw_atom *partial = atom_by_types(dtype,
+						 has_tail_block(conf) ?
+						 TAIL_ATOM : HEAD_ATOM);
+
+	if (old_file_size <= partial->offset_at(frame, object))
+		to_block = 0;
+	else {
+		to_block = old_file_size - partial->offset_at(frame, object);
+		if (to_block > get_atom_size(object))
+			to_block = get_atom_size(object);
+	}
+	if (to_block > conf->off_in_tail)
+		conf->gap_in_tail = to_block - conf->off_in_tail;
+	else
+		/*
+		 * nothing to uptodate
+		 */
+		conf->gap_in_tail = 0;
+}
+
+/*
+ * fill struct avec_config with offsets layouts
+ */
+void set_config_offsets(call_frame_t *frame,
+			xlator_t *this,
+			uint64_t offset,
+			uint64_t count,
+			atom_data_type dtype,
+			int32_t set_gap)
+{
+	crypt_local_t *local;
+	struct object_cipher_info *object;
+	struct avec_config *conf;
+	uint32_t resid;
+
+	uint32_t atom_size;
+	uint32_t atom_bits;
+
+	size_t orig_size;
+	off_t orig_offset;
+	size_t expanded_size;
+	off_t aligned_offset;
+
+	uint32_t off_in_head = 0;
+	uint32_t off_in_tail = 0;
+	uint32_t nr_full_blocks;
+	int32_t size_full_blocks;
+
+	uint32_t acount; /* number of alifned components to write.
+			  * The same as number of occupied logical
+			  * blocks (atoms)
+			  */
+	local = frame->local;
+	object = &local->info->cinfo;
+	conf = (dtype == DATA_ATOM ?
+		get_data_conf(frame) : get_hole_conf(frame));
+
+	orig_offset = offset;
+	orig_size = count;
+
+	atom_size = get_atom_size(object);
+	atom_bits = get_atom_bits(object);
+
+	/*
+	 * Round-down the start,
+	 * round-up the end.
+	 */
+	resid = offset & (uint64_t)(atom_size - 1);
+
+	if (resid)
+		off_in_head = resid;
+	aligned_offset = offset - off_in_head;
+	expanded_size = orig_size + off_in_head;
+
+	/* calculate tail,
+	   expand size forward  */
+	resid = (offset + orig_size) & (uint64_t)(atom_size - 1);
+
+	if (resid) {
+		off_in_tail = resid;
+		expanded_size += (atom_size - off_in_tail);
+	}
+	/*
+	 * calculate number of occupied blocks
+	 */
+	acount = expanded_size >> atom_bits;
+	/*
+	 * calculate number of full blocks
+	 */
+	size_full_blocks = expanded_size;
+	if (off_in_head)
+		size_full_blocks -= atom_size;
+	if (off_in_tail && size_full_blocks > 0)
+		size_full_blocks -= atom_size;
+	nr_full_blocks = size_full_blocks >> atom_bits;
+
+	conf->atom_size = atom_size;
+	conf->orig_size = orig_size;
+	conf->orig_offset = orig_offset;
+	conf->expanded_size = expanded_size;
+	conf->aligned_offset = aligned_offset;
+
+	conf->off_in_head = off_in_head;
+	conf->off_in_tail = off_in_tail;
+	conf->nr_full_blocks = nr_full_blocks;
+	conf->acount = acount;
+	/*
+	 * Finally, calculate precise amount of
+	 * "extra-bytes" that should be uptodated
+	 * at the end.
+	 * Only if RMW is expected.
+	 */
+	if (off_in_tail && set_gap)
+		set_gap_at_end(frame, object, conf, dtype);
+}
+
+struct data_cipher_alg data_cipher_algs[LAST_CIPHER_ALG][LAST_CIPHER_MODE] = {
+	[AES_CIPHER_ALG][XTS_CIPHER_MODE] =
+	{ .atomic = _gf_true,
+	  .should_pad = _gf_true,
+	  .blkbits = AES_BLOCK_BITS,
+	  .init = aes_xts_init,
+	  .set_private = set_private_aes_xts,
+	  .check_key = check_key_aes_xts,
+	  .set_iv = set_iv_aes_xts,
+	  .encrypt = encrypt_aes_xts
+	}
+};
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/keys.c b/xlators/encryption/crypt/src/keys.c
new file mode 100644
index 000000000..4a1d3bb5a
--- /dev/null
+++ b/xlators/encryption/crypt/src/keys.c
@@ -0,0 +1,302 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "defaults.h"
+#include "crypt-common.h"
+#include "crypt.h"
+
+/* Key hierarchy
+                  
+                   +----------------+
+                   | MASTER_VOL_KEY |
+                   +-------+--------+
+                           |
+                           |
+          +----------------+----------------+
+          |                |                |
+          |                |                |
+  +-------+------+ +-------+-------+ +------+--------+
+  | NMTD_VOL_KEY | | EMTD_FILE_KEY | | DATA_FILE_KEY |
+  +-------+------+ +---------------+ +---------------+
+          |
+          |
+  +-------+-------+
+  | NMTD_LINK_KEY |
+  +---------------+
+
+ */
+
+#if DEBUG_CRYPT
+static void check_prf_iters(uint32_t num_iters)
+{
+	if (num_iters == 0)
+		gf_log ("crypt", GF_LOG_DEBUG,
+			"bad number of prf iterations : %d", num_iters);
+}
+#else
+#define check_prf_iters(num_iters) noop
+#endif /* DEBUG_CRYPT */
+
+unsigned char crypt_fake_oid[16] =
+	{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+
+/*
+ * derive key in the counter mode using
+ * sha256-based HMAC as PRF, see
+ * NIST Special Publication 800-108, 5.1)
+ */
+
+#define PRF_OUTPUT_SIZE  SHA256_DIGEST_LENGTH
+
+static int32_t kderive_init(struct kderive_context *ctx,
+			    const unsigned char *pkey, /* parent key */
+			    uint32_t pkey_size,  /* parent key size */
+			    const unsigned char *idctx, /* id-context */
+			    uint32_t idctx_size,
+			    crypt_key_type type  /* type of child key */)
+{
+	unsigned char *pos;
+	uint32_t llen = strlen(crypt_keys[type].label);	
+	/*
+	 * Compoud the fixed input data for KDF:
+	 * [i]_2 || Label || 0x00 || Id-Context || [L]_2),
+	 * NIST SP 800-108, 5.1
+	 */
+	ctx->fid_len =
+		sizeof(uint32_t) +
+		llen +
+		1 +
+		idctx_size +
+		sizeof(uint32_t);
+
+	ctx->fid = GF_CALLOC(ctx->fid_len, 1, gf_crypt_mt_key);
+	if (!ctx->fid)
+		return ENOMEM;
+	ctx->out_len = round_up(crypt_keys[type].len >> 3,
+				PRF_OUTPUT_SIZE);
+	ctx->out = GF_CALLOC(ctx->out_len, 1, gf_crypt_mt_key);
+	if (!ctx->out) {
+		GF_FREE(ctx->fid);
+		return ENOMEM;
+	}
+	ctx->pkey = pkey;
+	ctx->pkey_len = pkey_size;
+	ctx->ckey_len = crypt_keys[type].len;
+
+	pos = ctx->fid;
+
+	/* counter will be set up in kderive_rfn() */
+	pos += sizeof(uint32_t);
+
+	memcpy(pos, crypt_keys[type].label, llen);
+	pos += llen;
+
+	/* set up zero octet */
+	*pos = 0;
+	pos += 1;
+
+	memcpy(pos, idctx, idctx_size);
+	pos += idctx_size;
+
+	*((uint32_t *)pos) = htobe32(ctx->ckey_len);
+
+	return 0;
+}
+
+static void kderive_update(struct kderive_context *ctx)
+{	
+	uint32_t i;
+	HMAC_CTX hctx;
+	unsigned char *pos = ctx->out;
+	uint32_t *p_iter = (uint32_t *)ctx->fid;
+	uint32_t num_iters = ctx->out_len / PRF_OUTPUT_SIZE;
+
+	check_prf_iters(num_iters);
+
+	HMAC_CTX_init(&hctx);
+	for (i = 0; i < num_iters; i++) {
+		/*
+		 * update the iteration number in the fid
+		 */
+		*p_iter = htobe32(i);
+		HMAC_Init_ex(&hctx,
+			     ctx->pkey, ctx->pkey_len >> 3,
+			     EVP_sha256(),
+			     NULL);
+		HMAC_Update(&hctx, ctx->fid, ctx->fid_len);
+		HMAC_Final(&hctx, pos, NULL);
+
+		pos += PRF_OUTPUT_SIZE;
+	}
+	HMAC_CTX_cleanup(&hctx);
+}
+
+static void kderive_final(struct kderive_context *ctx, unsigned char *child)
+{
+	memcpy(child, ctx->out, ctx->ckey_len >> 3);
+	GF_FREE(ctx->fid);
+	GF_FREE(ctx->out);
+	memset(ctx, 0, sizeof(*ctx));
+}
+
+/*
+ * derive per-volume key for object ids aithentication
+ */
+int32_t get_nmtd_vol_key(struct master_cipher_info *master)
+{
+	int32_t ret;
+	struct kderive_context ctx;
+
+	ret = kderive_init(&ctx,
+			   master->m_key,
+			   master_key_size(),
+			   crypt_fake_oid, sizeof(uuid_t), NMTD_VOL_KEY);
+	if (ret)
+		return ret;
+	kderive_update(&ctx);
+	kderive_final(&ctx, master->m_nmtd_key);
+	return 0;
+}
+
+/*
+ * derive per-link key for aithentication of non-encrypted
+ * meta-data (nmtd)
+ */
+int32_t get_nmtd_link_key(loc_t *loc,
+			  struct master_cipher_info *master,
+			  unsigned char *result)
+{
+	int32_t ret;
+	struct kderive_context ctx;
+
+	ret = kderive_init(&ctx,
+			   master->m_nmtd_key,
+			   nmtd_vol_key_size(),
+			   (const unsigned char *)loc->path,
+			   strlen(loc->path), NMTD_LINK_KEY);
+	if (ret)
+		return ret;
+	kderive_update(&ctx);
+	kderive_final(&ctx, result);
+	return 0;
+}
+
+/*
+ * derive per-file key for encryption and authentication
+ * of encrypted part of metadata (emtd)
+ */
+int32_t get_emtd_file_key(struct crypt_inode_info *info,
+			  struct master_cipher_info *master,
+			  unsigned char *result)
+{
+	int32_t ret;
+	struct kderive_context ctx;
+
+	ret = kderive_init(&ctx,
+			   master->m_key,
+			   master_key_size(),
+			   info->oid, sizeof(uuid_t), EMTD_FILE_KEY);
+	if (ret)
+		return ret;
+	kderive_update(&ctx);
+	kderive_final(&ctx, result);
+	return 0;
+}
+
+static int32_t data_key_type_by_size(uint32_t keysize, crypt_key_type *type)
+{
+	int32_t ret = 0;
+	switch (keysize) {
+	case 256:
+		*type = DATA_FILE_KEY_256;
+		break;
+	case 512:
+		*type = DATA_FILE_KEY_512;
+		break;
+	default:
+		gf_log("crypt", GF_LOG_ERROR, "Unsupported data key size %d",
+		       keysize);
+		ret = ENOTSUP;
+		break;
+	}
+	return ret;
+}
+
+/*
+ * derive per-file key for data encryption
+ */
+int32_t get_data_file_key(struct crypt_inode_info *info,
+			  struct master_cipher_info *master,
+			  uint32_t keysize,
+			  unsigned char *key)
+{
+	int32_t ret;
+	struct kderive_context ctx;
+	crypt_key_type type;
+
+	ret = data_key_type_by_size(keysize, &type);
+	if (ret)
+		return ret;
+	ret = kderive_init(&ctx,
+			   master->m_key,
+			   master_key_size(),
+			   info->oid, sizeof(uuid_t), type);
+	if (ret)
+		return ret;
+	kderive_update(&ctx);
+	kderive_final(&ctx, key);
+	return 0;
+}
+
+/*
+ * NOTE: Don't change existing keys: it will break compatibility;
+ */
+struct crypt_key crypt_keys[LAST_KEY_TYPE] = {
+	[MASTER_VOL_KEY] =
+	{ .len = MASTER_VOL_KEY_SIZE << 3,
+	  .label = "volume-master",
+	},
+	[NMTD_VOL_KEY] =
+	{ .len = NMTD_VOL_KEY_SIZE << 3,
+	  .label = "volume-nmtd-key-generation"
+	},
+	[NMTD_LINK_KEY] =
+	{ .len = 128,
+	  .label = "link-nmtd-authentication"
+	},
+	[EMTD_FILE_KEY] =
+	{ .len = 128,
+	  .label = "file-emtd-encryption-and-auth"
+	},
+	[DATA_FILE_KEY_256] =
+	{ .len = 256,
+	  .label = "file-data-encryption-256"
+	},
+	[DATA_FILE_KEY_512] =
+	{ .len = 512,
+	  .label = "file-data-encryption-512"
+	}
+};
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/metadata.c b/xlators/encryption/crypt/src/metadata.c
new file mode 100644
index 000000000..36b14c055
--- /dev/null
+++ b/xlators/encryption/crypt/src/metadata.c
@@ -0,0 +1,605 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "defaults.h"
+#include "crypt-common.h"
+#include "crypt.h"
+#include "metadata.h"
+
+int32_t alloc_format(crypt_local_t *local, size_t size)
+{
+	if (size > 0) {
+		local->format = GF_CALLOC(1, size, gf_crypt_mt_mtd);
+		if (!local->format)
+			return ENOMEM;
+	}
+	local->format_size = size;
+	return 0;
+}
+
+int32_t alloc_format_create(crypt_local_t *local)
+{
+	return alloc_format(local, new_format_size());
+}
+
+void free_format(crypt_local_t *local)
+{
+	GF_FREE(local->format);
+}
+
+/*
+ * Check compatibility with extracted metadata
+ */
+static int32_t check_file_metadata(struct crypt_inode_info *info)
+{
+	struct object_cipher_info *object = &info->cinfo;
+
+	if (info->nr_minor != CRYPT_XLATOR_ID) {
+		gf_log("crypt", GF_LOG_WARNING,
+		       "unsupported minor subversion %d", info->nr_minor);
+		return EINVAL;
+	}
+	if (object->o_alg > LAST_CIPHER_ALG) {
+		gf_log("crypt", GF_LOG_WARNING,
+		       "unsupported cipher algorithm %d",
+		       object->o_alg);
+		return EINVAL;
+	}
+	if (object->o_mode > LAST_CIPHER_MODE) {
+		gf_log("crypt", GF_LOG_WARNING,
+		       "unsupported cipher mode %d",
+		       object->o_mode);
+		return EINVAL;
+	}
+	if (object->o_block_bits < CRYPT_MIN_BLOCK_BITS ||
+	    object->o_block_bits > CRYPT_MAX_BLOCK_BITS) {
+		gf_log("crypt", GF_LOG_WARNING, "unsupported block bits %d",
+		       object->o_block_bits);
+		return EINVAL;
+	}
+	/* TBD: check data key size */
+	return 0;
+}
+
+static size_t format_size_v1(mtd_op_t op, size_t old_size)
+{
+	
+	switch (op) {
+	case MTD_CREATE:
+		return sizeof(struct mtd_format_v1);
+	case MTD_OVERWRITE:
+		return old_size;
+	case MTD_APPEND:
+		return old_size + NMTD_8_MAC_SIZE;
+	case MTD_CUT:
+		if (old_size > sizeof(struct mtd_format_v1))
+			return old_size - NMTD_8_MAC_SIZE;
+		else
+			return 0;
+	default:
+		gf_log("crypt", GF_LOG_WARNING, "Bad mtd operation");
+		return 0;
+	}
+}
+
+/*
+ * Calculate size of the updated format string.
+ * Returned zero means that we don't need to update the format string.
+ */
+size_t format_size(mtd_op_t op, size_t old_size)
+{
+	size_t versioned;
+
+	versioned = mtd_loaders[current_mtd_loader()].format_size(op,
+				 old_size - sizeof(struct crypt_format));
+	if (versioned != 0)
+		return versioned + sizeof(struct crypt_format);
+	return 0;
+}
+
+/*
+ * size of the format string of newly created file (nr_links = 1)
+ */
+size_t new_format_size(void)
+{
+	return format_size(MTD_CREATE, 0);
+}
+
+/*
+ * Calculate per-link MAC by pathname
+ */
+static int32_t calc_link_mac_v1(struct mtd_format_v1 *fmt,
+				loc_t *loc,
+				unsigned char *result,
+				struct crypt_inode_info *info,
+				struct master_cipher_info *master)
+{
+	int32_t ret;	
+	unsigned char nmtd_link_key[16];
+	CMAC_CTX *cctx;
+	size_t len;
+
+	ret = get_nmtd_link_key(loc, master, nmtd_link_key);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Can not get nmtd link key");
+		return -1;
+	}
+	cctx = CMAC_CTX_new();
+	if (!cctx) {
+		gf_log("crypt", GF_LOG_ERROR, "CMAC_CTX_new failed");
+		return -1; 
+	}
+	ret = CMAC_Init(cctx, nmtd_link_key, sizeof(nmtd_link_key),
+			EVP_aes_128_cbc(), 0);
+	if (!ret) {
+		gf_log("crypt", GF_LOG_ERROR, "CMAC_Init failed");
+		CMAC_CTX_free(cctx);
+		return -1;
+	}
+	ret = CMAC_Update(cctx, get_NMTD_V1(info), SIZE_OF_NMTD_V1);
+	if (!ret) {
+		gf_log("crypt", GF_LOG_ERROR, "CMAC_Update failed");
+		CMAC_CTX_free(cctx);
+		return -1;
+	}
+	ret = CMAC_Final(cctx, result, &len);
+	CMAC_CTX_free(cctx);
+	if (!ret) {
+		gf_log("crypt", GF_LOG_ERROR, "CMAC_Final failed");
+		return -1;
+	}
+	return 0;
+}
+
+/*
+ * Create per-link MAC of index @idx by pathname
+ */
+static int32_t create_link_mac_v1(struct mtd_format_v1 *fmt,
+				  uint32_t idx,
+				  loc_t *loc,
+				  struct crypt_inode_info *info,
+				  struct master_cipher_info *master)
+{
+	int32_t ret;
+	unsigned char *mac;
+	unsigned char cmac[16];
+
+	mac = get_NMTD_V1_MAC(fmt) + idx * SIZE_OF_NMTD_V1_MAC;
+
+	ret = calc_link_mac_v1(fmt, loc, cmac, info, master);
+	if (ret)
+		return -1;
+	memcpy(mac, cmac, SIZE_OF_NMTD_V1_MAC);
+	return 0;
+}
+
+static int32_t create_format_v1(unsigned char *wire,
+				loc_t *loc,
+				struct crypt_inode_info *info,
+				struct master_cipher_info *master)
+{
+	int32_t ret;
+	struct mtd_format_v1 *fmt;
+	unsigned char mtd_key[16];
+	AES_KEY EMTD_KEY;
+	unsigned char nmtd_link_key[16];
+	uint32_t ad;
+	GCM128_CONTEXT *gctx;
+
+	fmt = (struct mtd_format_v1 *)wire;
+
+	fmt->minor_id = info->nr_minor;
+	fmt->alg_id = AES_CIPHER_ALG;
+	fmt->dkey_factor = master->m_dkey_size >> KEY_FACTOR_BITS;
+	fmt->block_bits = master->m_block_bits;
+	fmt->mode_id = master->m_mode;
+	/*
+	 * retrieve keys for the parts of metadata
+	 */
+	ret = get_emtd_file_key(info, master, mtd_key);
+	if (ret)
+		return ret;
+	ret = get_nmtd_link_key(loc, master, nmtd_link_key);
+	if (ret)
+		return ret;
+	
+	AES_set_encrypt_key(mtd_key, sizeof(mtd_key)*8, &EMTD_KEY);
+
+	gctx = CRYPTO_gcm128_new(&EMTD_KEY, (block128_f)AES_encrypt);
+
+	/* TBD: Check return values */
+
+	CRYPTO_gcm128_setiv(gctx, info->oid, sizeof(uuid_t));
+
+	ad = htole32(MTD_LOADER_V1);
+	ret = CRYPTO_gcm128_aad(gctx, (const unsigned char *)&ad, sizeof(ad));
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_aad failed");
+		CRYPTO_gcm128_release(gctx);
+		return ret;
+	}
+	ret = CRYPTO_gcm128_encrypt(gctx,
+				    get_EMTD_V1(fmt),
+				    get_EMTD_V1(fmt),
+				    SIZE_OF_EMTD_V1);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_encrypt failed");
+		CRYPTO_gcm128_release(gctx);
+		return ret;
+	}
+	/*
+	 * set MAC of encrypted part of metadata
+	 */
+	CRYPTO_gcm128_tag(gctx, get_EMTD_V1_MAC(fmt), SIZE_OF_EMTD_V1_MAC);
+	CRYPTO_gcm128_release(gctx);
+	/*
+	 * set the first MAC of non-encrypted part of metadata 
+	 */
+	return create_link_mac_v1(fmt, 0, loc, info, master);
+}
+
+/*
+ * Called by fops:
+ * ->create();
+ * ->link();
+ *
+ * Pack common and version-specific parts of file's metadata
+ * Pre-conditions: @info contains valid object-id.
+ */
+int32_t create_format(unsigned char *wire,
+		      loc_t *loc,
+		      struct crypt_inode_info *info,
+		      struct master_cipher_info *master)
+{
+	struct crypt_format *fmt = (struct crypt_format *)wire;
+
+	fmt->loader_id = current_mtd_loader();
+
+	wire += sizeof(struct crypt_format);
+	return mtd_loaders[current_mtd_loader()].create_format(wire, loc,
+							       info, master);
+}
+
+/*
+ * Append or overwrite per-link mac of @mac_idx index
+ * in accordance with the new pathname
+ */
+int32_t appov_link_mac_v1(unsigned char *new,
+			  unsigned char *old,
+			  uint32_t old_size,
+			  int32_t mac_idx,
+			  loc_t *loc,
+			  struct crypt_inode_info *info,
+			  struct master_cipher_info *master,
+			  crypt_local_t *local)
+{
+	memcpy(new, old, old_size);
+	return create_link_mac_v1((struct mtd_format_v1 *)new, mac_idx,
+				  loc, info, master);
+}
+
+/*
+ * Cut per-link mac of @mac_idx index
+ */
+static int32_t cut_link_mac_v1(unsigned char *new,
+			       unsigned char *old,			       
+			       uint32_t old_size,
+			       int32_t mac_idx,
+			       loc_t *loc,
+			       struct crypt_inode_info *info,
+			       struct master_cipher_info *master,
+			       crypt_local_t *local)
+{
+	memcpy(new,
+	       old,
+	       sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * (mac_idx - 1));
+
+	memcpy(new + sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * (mac_idx - 1),
+	       old + sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * mac_idx,
+	       old_size - (sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * mac_idx));
+	return 0;
+}
+
+int32_t update_format_v1(unsigned char *new,
+			 unsigned char *old,
+				size_t old_len,
+				int32_t mac_idx, /* of old name */
+				mtd_op_t op,
+				loc_t *loc,
+				struct crypt_inode_info *info,
+				struct master_cipher_info *master,
+				crypt_local_t *local)
+{
+	switch (op) {
+	case MTD_APPEND:
+		mac_idx = 1 + (old_len - sizeof(struct mtd_format_v1))/8;
+	case MTD_OVERWRITE:
+		return appov_link_mac_v1(new, old, old_len, mac_idx,
+					 loc, info, master, local);
+	case MTD_CUT:
+		return cut_link_mac_v1(new, old, old_len, mac_idx,
+				       loc, info, master, local);
+	default:
+		gf_log("crypt", GF_LOG_ERROR, "Bad  mtd operation %d", op);
+		return -1;
+	}
+}
+
+/*
+ * Called by fops:
+ *
+ * ->link()
+ * ->unlink()
+ * ->rename()
+ *
+ */
+int32_t update_format(unsigned char *new,
+		      unsigned char *old,
+		      size_t old_len,
+		      int32_t mac_idx,
+		      mtd_op_t op,
+		      loc_t *loc,
+		      struct crypt_inode_info *info,
+		      struct master_cipher_info *master,
+		      crypt_local_t *local)
+{
+	if (!new)
+		return 0;
+	memcpy(new, old, sizeof(struct crypt_format));
+
+	old += sizeof(struct crypt_format);
+	new += sizeof(struct crypt_format);
+	old_len -= sizeof(struct crypt_format);
+
+	return mtd_loaders[current_mtd_loader()].update_format(new, old,
+							       old_len,
+							       mac_idx, op,
+							       loc, info,
+							       master, local);
+}
+
+/*
+ * Perform preliminary checks of found metadata
+ * Return < 0 on errors;
+ * Return number of object-id MACs (>= 1) on success
+ */
+int32_t check_format_v1(uint32_t len, unsigned char *wire)
+{
+	uint32_t nr_links;
+
+	if (len < sizeof(struct mtd_format_v1)) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "v1-loader: bad metadata size %d", len);
+		goto error;
+	}
+	len -= sizeof(struct mtd_format_v1);
+	if (len % sizeof(nmtd_8_mac_t)) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "v1-loader: bad metadata format");
+		goto error;
+	}
+	nr_links = 1 + len / sizeof(nmtd_8_mac_t);
+	if (nr_links > _POSIX_LINK_MAX)
+		goto error;
+	return nr_links;
+ error:
+	return EIO;
+}
+
+/*
+ * Verify per-link MAC specified by index @idx
+ *
+ * return:
+ * -1 on errors;
+ *  0 on failed verification;
+ *  1 on sucessful verification
+ */
+static int32_t verify_link_mac_v1(struct mtd_format_v1 *fmt,
+				  uint32_t idx /* index of the mac to verify */,
+				  loc_t *loc,
+				  struct crypt_inode_info *info,
+				  struct master_cipher_info *master)
+{
+	int32_t ret;
+	unsigned char *mac;
+	unsigned char cmac[16];
+
+	mac = get_NMTD_V1_MAC(fmt) + idx * SIZE_OF_NMTD_V1_MAC;
+
+	ret = calc_link_mac_v1(fmt, loc, cmac, info, master);
+	if (ret)
+		return -1;
+	if (memcmp(cmac, mac, SIZE_OF_NMTD_V1_MAC))
+		return 0;
+	return 1;
+}
+
+/*
+ * Lookup per-link MAC by pathname.
+ *
+ * return index of the MAC, if it was found;
+ * return < 0 on errors, or if the MAC wasn't found
+ */
+static int32_t lookup_link_mac_v1(struct mtd_format_v1 *fmt,
+				  uint32_t nr_macs,
+				  loc_t *loc,
+				  struct crypt_inode_info *info,
+				  struct master_cipher_info *master)
+{
+	int32_t ret;
+	uint32_t idx;
+
+	for (idx = 0; idx < nr_macs; idx++) {
+		ret = verify_link_mac_v1(fmt, idx, loc, info, master);
+		if (ret < 0)
+			return ret;
+		if (ret > 0)
+			return idx;
+	}
+	return -ENOENT;
+}
+
+/*
+ * Extract version-specific part of metadata
+ */
+static int32_t open_format_v1(unsigned char *wire,
+			      int32_t len,
+			      loc_t *loc,
+			      struct crypt_inode_info *info,
+			      struct master_cipher_info *master,
+			      crypt_local_t *local,
+			      gf_boolean_t load_info)
+{
+	int32_t ret;
+	int32_t num_nmtd_macs;
+	struct mtd_format_v1 *fmt;
+	unsigned char mtd_key[16];
+	AES_KEY EMTD_KEY;
+	GCM128_CONTEXT *gctx;
+	uint32_t ad;
+	emtd_8_mac_t gmac;
+	struct object_cipher_info *object;
+
+	num_nmtd_macs = check_format_v1(len, wire);
+	if (num_nmtd_macs <= 0)
+		return EIO;
+	fmt = (struct mtd_format_v1 *)wire;
+
+	ret = lookup_link_mac_v1(fmt, num_nmtd_macs, loc, info, master);
+	if (ret < 0) {
+		gf_log("crypt", GF_LOG_ERROR, "NMTD verification failed");
+		return EINVAL;
+	}
+	local->mac_idx = ret;
+	if (load_info == _gf_false)
+		/* the case of partial open */
+		return 0;
+
+	object = &info->cinfo;
+
+	ret = get_emtd_file_key(info, master, mtd_key);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Can not retrieve metadata key");
+		return ret;
+	}
+	/*
+	 * decrypt encrypted meta-data
+	 */
+	ret = AES_set_encrypt_key(mtd_key, sizeof(mtd_key)*8, &EMTD_KEY);
+	if (ret < 0) {
+		gf_log("crypt", GF_LOG_ERROR, "Can not set encrypt key");
+		return ret;
+	}
+	gctx = CRYPTO_gcm128_new(&EMTD_KEY, (block128_f)AES_encrypt);
+	if (!gctx) {
+		gf_log("crypt", GF_LOG_ERROR, "Can not alloc gcm context");
+		return ENOMEM;
+	}
+	CRYPTO_gcm128_setiv(gctx, info->oid, sizeof(uuid_t));
+
+	ad = htole32(MTD_LOADER_V1);
+	ret = CRYPTO_gcm128_aad(gctx, (const unsigned char *)&ad, sizeof(ad));
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_aad failed");
+		CRYPTO_gcm128_release(gctx);
+		return ret;	
+	}
+	ret = CRYPTO_gcm128_decrypt(gctx,
+				    get_EMTD_V1(fmt),
+				    get_EMTD_V1(fmt),
+				    SIZE_OF_EMTD_V1);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_decrypt failed");
+		CRYPTO_gcm128_release(gctx);
+		return ret;
+	}
+	/*
+	 * verify metadata
+	 */
+	CRYPTO_gcm128_tag(gctx, gmac, sizeof(gmac));
+	CRYPTO_gcm128_release(gctx);
+	if (memcmp(gmac, get_EMTD_V1_MAC(fmt), SIZE_OF_EMTD_V1_MAC)) {
+		gf_log("crypt", GF_LOG_ERROR, "EMTD verification failed");
+		return EINVAL;
+	}
+	/*
+	 * load verified metadata to the private part of inode
+	 */
+	info->nr_minor = fmt->minor_id;
+
+	object->o_alg = fmt->alg_id;
+	object->o_dkey_size = fmt->dkey_factor << KEY_FACTOR_BITS;
+	object->o_block_bits = fmt->block_bits;
+	object->o_mode = fmt->mode_id;
+
+	return check_file_metadata(info);
+}
+
+/*
+ * perform metadata authentication against @loc->path;
+ * extract crypt-specific attribtes and populate @info
+ * with them (optional)
+ */
+int32_t open_format(unsigned char *str,
+		    int32_t len,
+		    loc_t *loc,
+		    struct crypt_inode_info *info,
+		    struct master_cipher_info *master,
+		    crypt_local_t *local,
+		    gf_boolean_t load_info)
+{
+	struct crypt_format *fmt;
+	if (len < sizeof(*fmt)) {
+		gf_log("crypt", GF_LOG_ERROR, "Bad core format");
+		return EIO;
+	}
+	fmt = (struct crypt_format *)str;
+
+	if (fmt->loader_id >= LAST_MTD_LOADER) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "Unsupported loader id %d", fmt->loader_id);
+		return EINVAL;
+	}
+	str += sizeof(*fmt);
+	len -= sizeof(*fmt);
+
+	return mtd_loaders[fmt->loader_id].open_format(str,
+						       len,
+						       loc,
+						       info,
+						       master,
+						       local,
+						       load_info);
+}
+
+struct crypt_mtd_loader mtd_loaders [LAST_MTD_LOADER] = {
+	[MTD_LOADER_V1] = 
+	{.format_size = format_size_v1,
+	 .create_format = create_format_v1,
+	 .open_format = open_format_v1,
+	 .update_format = update_format_v1
+	}
+};
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/metadata.h b/xlators/encryption/crypt/src/metadata.h
new file mode 100644
index 000000000..a92f149ef
--- /dev/null
+++ b/xlators/encryption/crypt/src/metadata.h
@@ -0,0 +1,74 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef __METADATA_H__
+#define __METADATA_H__
+
+#define NMTD_8_MAC_SIZE   (8)
+#define EMTD_8_MAC_SIZE   (8)
+
+typedef uint8_t nmtd_8_mac_t[NMTD_8_MAC_SIZE];
+typedef uint8_t emtd_8_mac_t[EMTD_8_MAC_SIZE] ;
+
+/*
+ * Version "v1" of file's metadata.
+ * Metadata of this version has 4 components:
+ * 
+ * 1) EMTD (Encrypted part of MeTaData);
+ * 2) NMTD (Non-encrypted part of MeTaData);
+ * 3) EMTD_MAC; (EMTD Message Authentication Code);
+ * 4) Array of per-link NMTD MACs (for every (hard)link it includes
+ *    exactly one MAC)
+ */
+struct mtd_format_v1 {
+	/* EMTD, encrypted part of meta-data */
+	uint8_t            alg_id;  /* cipher algorithm id (only AES for now) */
+        uint8_t            mode_id; /* cipher mode id; (only XTS for now) */
+	uint8_t            block_bits; /* encoded block size */
+	uint8_t            minor_id; /* client translator id */
+	uint8_t            dkey_factor; /* encoded size of the data key */
+	/* MACs */
+	emtd_8_mac_t       gmac; /* MAC of the encrypted meta-data, 8 bytes */
+	nmtd_8_mac_t       omac; /* per-link MACs of the non-encrypted
+				  * meta-data: at least one such MAC is always
+				  * present */
+} __attribute__((packed));
+
+/*
+ * NMTD, the non-encrypted part of metadata of version "v1"
+ * is file's gfid, which is generated on trusted machines.
+ */
+#define SIZE_OF_NMTD_V1 (sizeof(uuid_t))
+#define SIZE_OF_EMTD_V1 (offsetof(struct mtd_format_v1, gmac) -		\
+			 offsetof(struct mtd_format_v1, alg_id))
+#define SIZE_OF_NMTD_V1_MAC (NMTD_8_MAC_SIZE)
+#define SIZE_OF_EMTD_V1_MAC (EMTD_8_MAC_SIZE)
+
+static inline unsigned char *get_EMTD_V1(struct mtd_format_v1 *format)
+{
+        return &format->alg_id;
+}
+
+static inline unsigned char *get_NMTD_V1(struct crypt_inode_info *info)
+{
+        return info->oid;
+}
+
+static inline unsigned char *get_EMTD_V1_MAC(struct mtd_format_v1 *format)
+{
+        return format->gmac;
+}
+
+static inline unsigned char *get_NMTD_V1_MAC(struct mtd_format_v1 *format)
+{
+        return format->omac;
+}
+
+#endif /* __METADATA_H__ */
diff --git a/xlators/encryption/rot-13/src/Makefile.am b/xlators/encryption/rot-13/src/Makefile.am
index ba5e623d8..94e8d18e7 100644
--- a/xlators/encryption/rot-13/src/Makefile.am
+++ b/xlators/encryption/rot-13/src/Makefile.am
@@ -1,14 +1,15 @@
 xlator_LTLIBRARIES = rot-13.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/encryption
 
-rot_13_la_LDFLAGS = -module -avoidversion
+rot_13_la_LDFLAGS = -module -avoid-version
 
 rot_13_la_SOURCES = rot-13.c
 rot_13_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la 
 
 noinst_HEADERS = rot-13.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
diff --git a/xlators/encryption/rot-13/src/rot-13.c b/xlators/encryption/rot-13/src/rot-13.c
index 9d5dab033..b9ac29a72 100644
--- a/xlators/encryption/rot-13/src/rot-13.c
+++ b/xlators/encryption/rot-13/src/rot-13.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #include <ctype.h>
 #include <sys/uio.h>
 
@@ -32,13 +22,13 @@
 #include "rot-13.h"
 
 /*
- * This is a rot13 ``encryption'' xlator. It rot13's data when 
- * writing to disk and rot13's it back when reading it. 
+ * This is a rot13 ``encryption'' xlator. It rot13's data when
+ * writing to disk and rot13's it back when reading it.
  * This xlator is meant as an example, NOT FOR PRODUCTION
  * USE ;) (hence no error-checking)
  */
 
-void 
+void
 rot13 (char *buf, int len)
 {
 	int i;
@@ -68,14 +58,15 @@ rot13_readv_cbk (call_frame_t *frame,
                  struct iovec *vector,
                  int32_t count,
 		 struct iatt *stbuf,
-                 struct iobref *iobref)
+                 struct iobref *iobref, dict_t *xdata)
 {
 	rot_13_private_t *priv = (rot_13_private_t *)this->private;
-  
+
 	if (priv->decrypt_read)
 		rot13_iovec (vector, count);
 
-	STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector, count, stbuf, iobref);
+	STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector, count,
+                             stbuf, iobref, xdata);
 	return 0;
 }
 
@@ -84,13 +75,13 @@ rot13_readv (call_frame_t *frame,
              xlator_t *this,
              fd_t *fd,
              size_t size,
-             off_t offset)
+             off_t offset, uint32_t flags, dict_t *xdata)
 {
 	STACK_WIND (frame,
 		    rot13_readv_cbk,
 		    FIRST_CHILD (this),
 		    FIRST_CHILD (this)->fops->readv,
-		    fd, size, offset);
+		    fd, size, offset, flags, xdata);
 	return 0;
 }
 
@@ -101,9 +92,10 @@ rot13_writev_cbk (call_frame_t *frame,
                   int32_t op_ret,
                   int32_t op_errno,
                   struct iatt *prebuf,
-		  struct iatt *postbuf)
+		  struct iatt *postbuf, dict_t *xdata)
 {
-	STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf);
+	STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf,
+                             xdata);
 	return 0;
 }
 
@@ -112,20 +104,20 @@ rot13_writev (call_frame_t *frame,
               xlator_t *this,
               fd_t *fd,
               struct iovec *vector,
-              int32_t count, 
-              off_t offset,
-              struct iobref *iobref)
+              int32_t count,
+              off_t offset, uint32_t flags,
+              struct iobref *iobref, dict_t *xdata)
 {
 	rot_13_private_t *priv = (rot_13_private_t *)this->private;
 	if (priv->encrypt_write)
 		rot13_iovec (vector, count);
 
-	STACK_WIND (frame, 
+	STACK_WIND (frame,
 		    rot13_writev_cbk,
 		    FIRST_CHILD (this),
 		    FIRST_CHILD (this)->fops->writev,
-		    fd, vector, count, offset,
-                    iobref);
+		    fd, vector, count, offset, flags,
+                    iobref, xdata);
 	return 0;
 }
 
@@ -136,7 +128,7 @@ init (xlator_t *this)
 	rot_13_private_t *priv = NULL;
 
 	if (!this->children || this->children->next) {
-		gf_log ("rot13", GF_LOG_ERROR, 
+		gf_log ("rot13", GF_LOG_ERROR,
 			"FATAL: rot13 should have exactly one child");
 		return -1;
 	}
@@ -181,6 +173,9 @@ fini (xlator_t *this)
 {
 	rot_13_private_t *priv = this->private;
 
+        if (!priv)
+                return;
+        this->private = NULL;
 	GF_FREE (priv);
 
 	return;
@@ -191,15 +186,14 @@ struct xlator_fops fops = {
 	.writev       = rot13_writev
 };
 
-struct xlator_cbks cbks = {
-};
+struct xlator_cbks cbks;
 
 struct volume_options options[] = {
-	{ .key  = {"encrypt-write"}, 
+	{ .key  = {"encrypt-write"},
 	  .type = GF_OPTION_TYPE_BOOL
 	},
-	{ .key  = {"decrypt-read"}, 
-	  .type = GF_OPTION_TYPE_BOOL 
+	{ .key  = {"decrypt-read"},
+	  .type = GF_OPTION_TYPE_BOOL
 	},
 	{ .key  = {NULL} },
 };
diff --git a/xlators/encryption/rot-13/src/rot-13.h b/xlators/encryption/rot-13/src/rot-13.h
index 35a2df2da..3e9fc19c7 100644
--- a/xlators/encryption/rot-13/src/rot-13.h
+++ b/xlators/encryption/rot-13/src/rot-13.h
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __ROT_13_H__
 #define __ROT_13_H__
 
diff --git a/xlators/features/Makefile.am b/xlators/features/Makefile.am
index 6496f3ab0..d2f5ef192 100644
--- a/xlators/features/Makefile.am
+++ b/xlators/features/Makefile.am
@@ -1,3 +1,4 @@
-SUBDIRS = locks trash quota read-only access-control mac-compat #path-converter # filter
+SUBDIRS = locks quota read-only mac-compat quiesce marker index \
+	  protect compress changelog gfid-access $(GLUPY_SUBDIR) qemu-block # trash path-converter # filter
 
-CLEANFILES = 
+CLEANFILES =
diff --git a/xlators/features/access-control/src/Makefile.am b/xlators/features/access-control/src/Makefile.am
deleted file mode 100644
index 6ab8cc4ec..000000000
--- a/xlators/features/access-control/src/Makefile.am
+++ /dev/null
@@ -1,13 +0,0 @@
-xlator_LTLIBRARIES = access-control.la
-xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
-access_control_la_LDFLAGS = -module -avoidversion
-access_control_la_SOURCES = access-control.c
-access_control_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-
-noinst_HEADERS = access-control.h
-
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)\
-	-L$(xlatordir)/
-
-CLEANFILES =
diff --git a/xlators/features/access-control/src/access-control.c b/xlators/features/access-control/src/access-control.c
deleted file mode 100644
index b659a0133..000000000
--- a/xlators/features/access-control/src/access-control.c
+++ /dev/null
@@ -1,1846 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "access-control.h"
-#include "xlator.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "iatt.h"
-
-/* Careful, this function erases the stub from frame->local. Dont call this if
- * a subsequent callback requires retaining access to the stub. This should be
- * called at the end of all access-control related operations, i.e. once the
- * frame will be handed off to the actual fop and the next callback that will
- * be called is the default callback. IOW, the function where call_resume is
- * called.
- * NOTE: this is required because FRAME_DESTROY tries to free frame->local if
- * it finds it to be non-NULL.
- */
-call_stub_t *
-__get_frame_stub (call_frame_t *fr)
-{
-        call_stub_t     *st = NULL;
-
-        if (!fr)
-                return NULL;
-
-        st = fr->local;
-        fr->local = NULL;
-
-        return st;
-}
-
-
-int
-ac_test_owner_access (struct iatt *ia, uid_t uid, int accesstest)
-{
-        int     ret = -1;
-
-        if (!ia)
-                return -1;
-
-        /* First test permissions using the uid. */
-        if (ia->ia_uid != uid) {
-                ret = -1;
-                goto out;
-        }
-
-        /* At this point we know, the uid matches that of the stat structure, so
-         * if the caller does not care, we should return success.
-         */
-        if (ac_test_dontcare (accesstest)) {
-                ret = 0;
-                goto out;
-        }
-
-        if (ac_test_read (accesstest))
-                ret = IA_PROT_RUSR (ia->ia_prot);
-
-        if (ac_test_write (accesstest))
-                ret = IA_PROT_WUSR (ia->ia_prot);
-
-        if (ac_test_exec (accesstest))
-                ret = IA_PROT_XUSR (ia->ia_prot);
-
-        /* For failed access test for owner, we need to return EACCES */
-        if (!ret)
-                ret = -1;
-        else
-                ret = 0;
-out:
-        return ret;
-}
-
-
-int
-ac_test_group_access (struct iatt *ia, gid_t gid, gid_t *auxgids, int auxcount,
-                      int accesstest)
-{
-        int     ret = -1;
-        int     testgid = -1;
-        int     x = 0;
-
-        if (!ia)
-                return -1;
-        /* First, determine which gid to test against. This will be determined
-         * by first checking which of the gids given to us match the gid in the
-         * stat. If none match, then we go to checking with others as the user.
-         */
-
-        /* If we are only given the primary gid. Dont depend on @auxgids
-         * being NULL since I know users of this function can pass statically
-         * allocated arrays which cant be NULL and yet contain no valid gids.
-         */
-
-        if ((ia->ia_gid != gid) && (auxcount == 0)) {
-                ret = -1;
-                goto out;
-        }
-
-        if (ia->ia_gid == gid)
-                testgid = gid;
-        else {
-                for (; x < auxcount; ++x) {
-                        if (ia->ia_gid == auxgids[x]) {
-                                testgid = ia->ia_gid;
-                                break;
-                        }
-                }
-        }
-
-        /* None of the gids match with the gid in the stat. */
-        if (testgid == -1) {
-                ret = -1;
-                goto out;
-        }
-
-        /* At this point, at least one gid matches that in the stat, now we must
-         * check whether the caller is interested in the access check at all.
-         */
-        if (ac_test_dontcare (accesstest)) {
-                ret = 0;
-                goto out;
-        }
-
-        if (ac_test_read (accesstest))
-                ret = IA_PROT_RGRP (ia->ia_prot);
-
-        if (ac_test_write (accesstest))
-                ret = IA_PROT_WGRP (ia->ia_prot);
-
-        if (ac_test_exec (accesstest))
-                ret = IA_PROT_XGRP (ia->ia_prot);
-
-        if (!ret)
-                ret = -1;
-        else
-                ret = 0;
-
-out:
-        return ret;
-}
-
-
-int
-ac_test_other_access (struct iatt *ia, int accesstest)
-{
-        int     ret = 0;
-
-        if (!ia)
-                return -1;
-
-        if (ac_test_read (accesstest))
-                ret = IA_PROT_ROTH (ia->ia_prot);
-
-        if (ac_test_write (accesstest))
-                ret = IA_PROT_WOTH (ia->ia_prot);
-
-        if (ac_test_exec (accesstest))
-                ret = IA_PROT_XOTH (ia->ia_prot);
-
-        if (!ret)
-                ret = -1;
-        else
-                ret = 0;
-
-        return ret;
-}
-
-
-/* Returns -1 on a failed access test with @operrno set to the relevant error
- * number.
- */
-int
-ac_test_access (struct iatt *ia, uid_t uid, gid_t gid, gid_t *auxgids,
-                int auxcount, int accesstest, int testwho, int *operrno)
-{
-        int             ret = -1;
-
-        if ((!ia) || (!operrno))
-                return -1;
-
-        if ((uid == 0) && (gid == 0)) {
-                gf_log (ACTRL, GF_LOG_TRACE, "Root has access");
-                return 0;
-        }
-
-        if (ac_test_owner (testwho)) {
-                gf_log (ACTRL, GF_LOG_TRACE, "Testing owner access");
-                ret = ac_test_owner_access (ia, uid, accesstest);
-        }
-
-        if (ret == 0) {
-                gf_log (ACTRL, GF_LOG_TRACE, "Owner has access");
-                goto out;
-        }
-
-        if (ac_test_group (testwho)) {
-                gf_log (ACTRL, GF_LOG_TRACE, "Testing group access");
-                ret = ac_test_group_access (ia, gid, auxgids, auxcount,
-                                            accesstest);
-        }
-
-        if (ret == 0) {
-                gf_log (ACTRL, GF_LOG_TRACE, "Group has access");
-                goto out;
-        }
-
-        if (ac_test_other (testwho)) {
-                gf_log (ACTRL, GF_LOG_TRACE, "Testing other access");
-                ret = ac_test_other_access (ia, accesstest);
-        }
-
-        if (ret == 0)
-                gf_log (ACTRL, GF_LOG_TRACE, "Other has access");
-out:
-        if (ret == -1) {
-                gf_log (ACTRL, GF_LOG_TRACE, "No access allowed");
-                *operrno = EPERM;
-        }
-
-        return ret;
-}
-
-
-int
-ac_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path)
-{
-        int     ret = -EFAULT;
-
-        if (!loc)
-                return ret;
-
-        if (inode) {
-                loc->inode = inode_ref (inode);
-                loc->ino = inode->ino;
-        }
-
-        if (parent)
-                loc->parent = inode_ref (parent);
-
-        loc->path = gf_strdup (path);
-        if (!loc->path) {
-                gf_log (ACTRL, GF_LOG_ERROR, "strdup failed");
-                goto loc_wipe;
-        }
-
-        loc->name = strrchr (loc->path, '/');
-        if (loc->name)
-                loc->name++;
-        else
-                goto loc_wipe;
-
-        ret = 0;
-loc_wipe:
-        if (ret < 0)
-                loc_wipe (loc);
-
-        return ret;
-}
-
-
-int
-ac_inode_loc_fill (inode_t *inode, loc_t *loc)
-{
-        char            *resolvedpath = NULL;
-        inode_t         *parent = NULL;
-        int             ret = -EFAULT;
-
-        if ((!inode) || (!loc))
-                return ret;
-
-        if ((inode) && (inode->ino == 1))
-                goto ignore_parent;
-
-        parent = inode_parent (inode, 0, NULL);
-        if (!parent)
-                goto err;
-
-ignore_parent:
-        ret = inode_path (inode, NULL, &resolvedpath);
-        if (ret < 0)
-                goto err;
-
-        ret = ac_loc_fill (loc, inode, parent, resolvedpath);
-        if (ret < 0)
-                goto err;
-
-err:
-        if (parent)
-                inode_unref (parent);
-
-        if (resolvedpath)
-                GF_FREE (resolvedpath);
-
-        return ret;
-}
-
-
-int
-ac_parent_loc_fill (loc_t *parentloc, loc_t *childloc)
-{
-        if ((!parentloc) || (!childloc))
-                return -1;
-
-        return ac_inode_loc_fill (childloc->parent, parentloc);
-}
-
-
-int32_t
-ac_truncate_resume (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                    off_t offset)
-{
-        STACK_WIND (frame, default_truncate_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->truncate, loc, offset);
-        return 0;
-}
-
-
-int32_t
-ac_truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_WRITE, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1)
-                goto out;
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (truncate, frame, -1, op_errno, NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-
-        stub = fop_truncate_stub (frame, ac_truncate_resume, loc, offset);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        STACK_WIND (frame, ac_truncate_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, loc);
-
-        ret = 0;
-out:
-        if (ret < 0)
-                STACK_UNWIND_STRICT (truncate, frame, -1, -ret, NULL, NULL);
-
-        return 0;
-}
-
-
-int32_t
-ac_access_resume (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask)
-{
-        STACK_WIND (frame, default_access_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->access, loc, mask);
-        return 0;
-}
-
-
-int32_t
-ac_access_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-        int32_t         mask = 0;
-        int             acctest = 0;
-
-        stub = __get_frame_stub (frame);
-        mask = stub->args.access.mask;
-
-        /* If mask requests test for file existence then do not
-         * return a failure with ENOENT, instead return a failed
-         * access test.
-         */
-        if (op_ret == -1) {
-                if (mask & F_OK)
-                        op_errno = EACCES;
-                else
-                        op_errno = errno;
-
-                goto out;
-        }
-
-        if (R_OK & mask)
-                acctest |= ACCTEST_READ;
-        else if (W_OK & mask)
-                acctest |= ACCTEST_WRITE;
-        else if (X_OK & mask)
-                acctest |= ACCTEST_EXEC;
-        else
-                acctest = 0;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 acctest, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1)
-                goto out;
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (access, frame, -1, op_errno);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-
-        stub = fop_access_stub (frame, ac_access_resume, loc, mask);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        STACK_WIND (frame, ac_access_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, loc);
-        ret = 0;
-
-out:
-        if (ret < 0)
-                STACK_UNWIND_STRICT (access, frame, -1, -ret);
-
-        return 0;
-}
-
-
-int32_t
-ac_readlink_resume (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                    size_t size)
-{
-        STACK_WIND (frame, default_readlink_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->readlink, loc, size);
-        return 0;
-}
-
-
-int32_t
-ac_readlink_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_READ, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1)
-                goto out;
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (readlink, frame, -1, op_errno, NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-
-        stub = fop_readlink_stub (frame, ac_readlink_resume, loc, size);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        STACK_WIND (frame, ac_readlink_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, loc);
-        ret = 0;
-
-out:
-        if (ret < 0)
-                STACK_UNWIND_STRICT (readlink, frame, -1, -ret, NULL, NULL);
-
-        return 0;
-}
-
-
-int
-ac_mknod_resume (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-                 dev_t rdev, dict_t *params)
-{
-        STACK_WIND (frame, default_mknod_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, params);
-        return 0;
-}
-
-
-int32_t
-ac_mknod_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_WRITE, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1) {
-                op_errno = EACCES;
-                goto out;
-        }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (mknod, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int
-ac_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-          dev_t rdev, dict_t *params)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-        loc_t           parentloc = {0, };
-
-        stub = fop_mknod_stub (frame, ac_mknod_resume, loc, mode, rdev, params);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        ret = ac_parent_loc_fill (&parentloc, loc);
-        if (ret < 0)
-                goto out;
-
-        STACK_WIND (frame, ac_mknod_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, &parentloc);
-        loc_wipe (&parentloc);
-        ret = 0;
-
-out:
-        if (ret < 0) {
-                /* Erase any stored frame before unwinding. */
-                stub = __get_frame_stub (frame);
-                if (stub)
-                        call_stub_destroy (stub);
-                STACK_UNWIND_STRICT (mknod, frame, -1, -ret, NULL, NULL, NULL,
-                                     NULL);
-        }
-
-        return 0;
-}
-
-
-int
-ac_mkdir_resume (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-                 dict_t *params)
-{
-        STACK_WIND (frame, default_mkdir_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->mkdir, loc, mode, params);
-        return 0;
-}
-
-
-int32_t
-ac_mkdir_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_WRITE, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1) {
-                /* On a failed  write test on parent dir, we need to return
-                 * EACCES, not EPERM that is returned by default by
-                 * ac_test_access.
-                 */
-                op_errno = EACCES;
-                goto out;
-        }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (mkdir, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int
-ac_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-          dict_t *params)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-        loc_t           parentloc = {0, };
-
-        stub = fop_mkdir_stub (frame, ac_mkdir_resume, loc, mode, params);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        ret = ac_parent_loc_fill (&parentloc, loc);
-        if (ret < 0)
-                goto out;
-
-        STACK_WIND (frame, ac_mkdir_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, &parentloc);
-        loc_wipe (&parentloc);
-        ret = 0;
-
-out:
-        if (ret < 0) {
-                /* Erase the stored stub before unwinding. */
-                stub = __get_frame_stub (frame);
-                if (stub)
-                        call_stub_destroy (stub);
-                STACK_UNWIND_STRICT (mkdir, frame, -1, -ret, NULL, NULL, NULL,
-                                     NULL);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_unlink_resume (call_frame_t *frame, xlator_t *this, loc_t *loc)
-{
-        STACK_WIND (frame, default_unlink_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->unlink, loc);
-        return 0;
-}
-
-
-int32_t
-ac_unlink_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_WRITE, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1) {
-                op_errno = EACCES;
-                goto out;
-        }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-        loc_t           parentloc = {0, };
-
-        stub = fop_unlink_stub (frame, ac_unlink_resume, loc);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        ret = ac_parent_loc_fill (&parentloc, loc);
-        if (ret < 0)
-                goto out;
-
-        STACK_WIND (frame, ac_unlink_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, &parentloc);
-        loc_wipe (&parentloc);
-        ret = 0;
-
-out:
-        if (ret < 0) {
-                /* Erase the stored stub before unwinding. */
-                stub = __get_frame_stub (frame);
-                if (stub)
-                        call_stub_destroy (stub);
-                STACK_UNWIND_STRICT (unlink, frame, -1, -ret, NULL, NULL);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_rmdir_resume (call_frame_t *frame, xlator_t *this, loc_t *loc)
-{
-        STACK_WIND (frame, default_rmdir_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->rmdir, loc);
-        return 0;
-}
-
-
-int32_t
-ac_rmdir_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_WRITE, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1) {
-                op_errno = EACCES;
-                goto out;
-        }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (rmdir, frame, -1, op_errno, NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-        loc_t           parentloc = {0, };
-
-        stub = fop_rmdir_stub (frame, ac_rmdir_resume, loc);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        ret = ac_parent_loc_fill (&parentloc, loc);
-        if (ret < 0)
-                goto out;
-
-        STACK_WIND (frame, ac_rmdir_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, &parentloc);
-        loc_wipe (&parentloc);
-        ret = 0;
-
-out:
-        if (ret < 0) {
-                /* Erase the stored stub before unwinding. */
-                stub = __get_frame_stub (frame);
-                if (stub)
-                        call_stub_destroy (stub);
-                STACK_UNWIND_STRICT (rmdir, frame, -1, -ret, NULL, NULL);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_symlink_resume (call_frame_t *frame, xlator_t *this, const char *linkname,
-                   loc_t *loc, dict_t *params)
-{
-        STACK_WIND (frame, default_symlink_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->symlink, linkname, loc, params);
-        return 0;
-}
-
-
-int32_t
-ac_symlink_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_WRITE, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1) {
-                op_errno = EACCES;
-                goto out;
-        }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (symlink, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int
-ac_symlink (call_frame_t *frame, xlator_t *this, const char *linkname,
-            loc_t *loc, dict_t *params)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-        loc_t           parentloc = {0, };
-
-        stub = fop_symlink_stub (frame, ac_symlink_resume, linkname, loc,
-                                 params);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        ret = ac_parent_loc_fill (&parentloc, loc);
-        if (ret < 0)
-                goto out;
-
-        STACK_WIND (frame, ac_symlink_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, &parentloc);
-        loc_wipe (&parentloc);
-        ret = 0;
-
-out:
-        if (ret < 0) {
-                /* Erase the stored stub before unwinding. */
-                stub = __get_frame_stub (frame);
-                if (stub)
-                        call_stub_destroy (stub);
-                STACK_UNWIND_STRICT (symlink, frame, -1, -ret, NULL, NULL, NULL,
-                                     NULL);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_rename_resume (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
-                  loc_t *newloc)
-{
-        STACK_WIND (frame, default_rename_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->rename, oldloc, newloc);
-        return 0;
-}
-
-
-int32_t
-ac_rename_dst_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid,
-                                 frame->root->gid, frame->root->groups,
-                                 frame->root->ngrps, ACCTEST_WRITE,
-                                 ACCTEST_ANY, &op_errno);
-        if (op_ret == -1) {
-                op_errno = EACCES;
-               goto out;
-        }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_rename_src_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-        loc_t           parentloc = {0, };
-
-        stub = frame->local;
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid,
-                                 frame->root->gid, frame->root->groups,
-                                 frame->root->ngrps, ACCTEST_WRITE,
-                                 ACCTEST_ANY, &op_errno);
-        if (op_ret == -1) {
-                op_errno = EACCES;
-                goto out;
-        }
-
-        op_ret = ac_parent_loc_fill (&parentloc, &stub->args.rename.new);
-        if (op_ret == -1) {
-                op_errno = -EFAULT;
-                goto out;
-        }
-
-        STACK_WIND (frame, ac_rename_dst_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, &parentloc);
-        loc_wipe (&parentloc);
-
-out:
-        if (op_ret < 0) {
-                /* Erase the stored stub before unwinding. */
-                stub = __get_frame_stub (frame);
-                if (stub)
-                        call_stub_destroy (stub);
-                STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL, NULL);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-        loc_t           parentloc = {0, };
-
-        stub = fop_rename_stub (frame, ac_rename_resume, oldloc, newloc);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        ret = ac_parent_loc_fill (&parentloc, oldloc);
-        if (ret < 0)
-                goto out;
-
-        STACK_WIND (frame, ac_rename_src_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, &parentloc);
-        loc_wipe (&parentloc);
-        ret = 0;
-
-out:
-        if (ret < 0) {
-                /* Erase the stored stub before unwinding. */
-                stub = __get_frame_stub (frame);
-                if (stub)
-                        call_stub_destroy (stub);
-                STACK_UNWIND_STRICT (rename, frame, -1, -ret, NULL, NULL, NULL,
-                                     NULL, NULL);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_link_resume (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
-                loc_t *newloc)
-{
-        STACK_WIND (frame, default_link_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->link, oldloc, newloc);
-        return 0;
-}
-
-
-int32_t
-ac_link_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_WRITE, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1) {
-                /* By default ac_test_access sets the op_errno to EPERM
-                 * but in the case of link, we need to return EACCES to meet
-                 * posix requirements when a write permission is not available
-                 * for the new directory.
-                 */
-                op_errno = EACCES;
-                goto out;
-        }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-        loc_t           parentloc = {0, };
-
-        stub = fop_link_stub (frame, ac_link_resume, oldloc, newloc);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        ret = ac_parent_loc_fill (&parentloc, newloc);
-        if (ret < 0)
-                goto out;
-
-        STACK_WIND (frame, ac_link_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, &parentloc);
-        loc_wipe (&parentloc);
-        ret = 0;
-
-out:
-        if (ret < 0) {
-                /* Erase the stored stub before unwinding. */
-                stub = __get_frame_stub (frame);
-                if (stub)
-                        call_stub_destroy (stub);
-                STACK_UNWIND_STRICT (link, frame, -1, -ret, NULL, NULL, NULL,
-                                     NULL);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_create_resume (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                  int32_t flags, mode_t mode, fd_t *fd, dict_t *params)
-{
-        STACK_WIND (frame, default_create_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->create, loc, flags, mode,
-                    fd, params);
-        return 0;
-}
-
-
-int32_t
-ac_create_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_WRITE, ACCTEST_ANY, &op_errno);
-         if (op_ret == -1) {
-                op_errno = EACCES;
-                goto out;
-         }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (create, frame, -1, op_errno, NULL, NULL,
-                                     NULL, NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-           mode_t mode, fd_t *fd, dict_t *params)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-        loc_t           parentloc = {0, };
-
-        stub = fop_create_stub (frame, ac_create_resume, loc, flags, mode,
-                                fd, params);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        ret = ac_parent_loc_fill (&parentloc, loc);
-        if (ret < 0)
-                goto out;
-
-        STACK_WIND (frame, ac_create_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, &parentloc);
-        loc_wipe (&parentloc);
-        ret = 0;
-
-out:
-        if (ret < 0) {
-                /* Erase the stored stub before unwinding. */
-                stub = __get_frame_stub (frame);
-                if (stub)
-                        call_stub_destroy (stub);
-                STACK_UNWIND_STRICT (create, frame, -1, -ret, NULL, NULL, NULL,
-                                     NULL, NULL);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_open_resume (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-                fd_t *fd, int32_t wbflags)
-{
-        STACK_WIND (frame, default_open_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->open, loc, flags, fd, wbflags);
-        return 0;
-}
-
-
-int32_t
-ac_open_create_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                         int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_WRITE, ACCTEST_ANY, &op_errno);
-         if (op_ret == -1) {
-                op_errno = EACCES;
-                goto out;
-         }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (open, frame, -1, op_errno, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int
-ac_open_create (call_stub_t *stub)
-{
-        int             ret = -EFAULT;
-        loc_t           parentloc = {0, };
-        xlator_t        *this = NULL;
-
-        if (!stub)
-                return ret;
-
-        ret = ac_parent_loc_fill (&parentloc, &stub->args.open.loc);
-        if (ret < 0)
-                goto out;
-
-        this = stub->frame->this;
-        STACK_WIND (stub->frame, ac_open_create_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, &parentloc);
-        loc_wipe (&parentloc);
-        ret = 0;
-
-out:
-        return ret;
-}
-
-
-int32_t
-ac_open_only_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-        int             acctest = 0;
-        int32_t         flags = 0;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        flags = stub->args.open.flags;
-        /* The permissions we test for depend on how the open needs to be
-         * performed. */
-        if ((flags & O_ACCMODE) == O_RDONLY)
-                acctest = ACCTEST_READ;
-        else if (((flags & O_ACCMODE) == O_RDWR) ||
-                 ((flags & O_ACCMODE) == O_WRONLY))
-                acctest = ACCTEST_WRITE;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 acctest, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1)
-                goto out;
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (open, frame, -1, op_errno, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int
-ac_open_only (call_stub_t *stub)
-{
-        int             ret = -EFAULT;
-        xlator_t        *this = NULL;
-
-        if (!stub)
-                return ret;
-
-        this = stub->frame->this;
-        STACK_WIND (stub->frame, ac_open_only_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, &stub->args.open.loc);
-        return 0;
-}
-
-int32_t
-ac_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-         fd_t *fd, int32_t wbflags)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-
-        stub = fop_open_stub (frame, ac_open_resume, loc, flags, fd, wbflags);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        /* If we are not supposed to create the file then there is no need to
-         * check the parent dir permissions. */
-        if (flags & O_CREAT)
-                ret = ac_open_create (stub);
-        else
-                ret = ac_open_only (stub);
-
-out:
-        if (ret < 0) {
-                /* Erase the stored stub before unwinding. */
-                stub = __get_frame_stub (frame);
-                if (stub)
-                        call_stub_destroy (stub);
-                STACK_UNWIND_STRICT (open, frame, -1, -ret, NULL);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_readv_resume (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-                 off_t offset)
-{
-        STACK_WIND (frame, default_readv_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->readv, fd, size, offset);
-        return 0;
-}
-
-
-int32_t
-ac_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_READ, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1) {
-                op_errno = EACCES;
-                goto out;
-        }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (readv, frame, -1, op_errno, NULL, 0, NULL,
-                                     NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-          off_t offset)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-
-        stub = fop_readv_stub (frame, ac_readv_resume, fd, size, offset);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        STACK_WIND (frame, ac_readv_fstat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->fstat, fd);
-        ret = 0;
-
-out:
-        if (ret < 0)
-                STACK_UNWIND_STRICT (readv, frame, -1, -ret, NULL, 0, NULL,
-                                     NULL);
-
-        return 0;
-}
-
-
-int32_t
-ac_writev_resume (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                  struct iovec *vector, int32_t count, off_t offset,
-                  struct iobref *iobref)
-{
-        STACK_WIND (frame, default_writev_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->writev, fd, vector, count, offset,
-                    iobref);
-        return 0;
-}
-
-
-int32_t
-ac_writev_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_WRITE, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1) {
-                op_errno = EACCES;
-                goto out;
-        }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
-           int32_t count, off_t offset, struct iobref *iobref)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-
-        stub = fop_writev_stub (frame, ac_writev_resume, fd, vector, count,
-                                offset, iobref);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        STACK_WIND (frame, ac_writev_fstat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->fstat, fd);
-        ret = 0;
-
-out:
-        if (ret < 0)
-                STACK_UNWIND_STRICT (writev, frame, -1, -ret, NULL, NULL);
-
-        return 0;
-}
-
-
-int32_t
-ac_opendir_resume (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
-{
-        STACK_WIND (frame, default_opendir_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->opendir, loc, fd);
-        return 0;
-}
-
-
-int32_t
-ac_opendir_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_READ, ACCTEST_ANY, &op_errno);
-        if (op_ret == -1)
-                goto out;
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (opendir, frame, -1, op_errno, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-int32_t
-ac_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-
-        stub = fop_opendir_stub (frame, ac_opendir_resume, loc, fd);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        STACK_WIND (frame, ac_opendir_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, loc);
-        ret = 0;
-
-out:
-        if (ret < 0)
-                STACK_UNWIND_STRICT (opendir, frame, -1, -ret, NULL);
-
-        return 0;
-}
-
-
-int32_t
-ac_setattr_resume (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                   struct iatt *buf, int32_t valid)
-{
-        STACK_WIND (frame, default_setattr_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->setattr, loc, buf, valid);
-        return 0;
-}
-
-
-int32_t
-ac_setattr_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-        int32_t         valid = 0;
-        struct iatt     *setbuf = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_DONTCARE, ACCTEST_OWNER,
-                                 &op_errno);
-        if (op_ret == -1)
-                goto out;
-
-        valid = stub->args.setattr.valid;
-        setbuf = &stub->args.setattr.stbuf;
-        if (gf_attr_uid_set (valid) || gf_attr_gid_set (valid)) {
-                /* chown returns EPERM if the operation would change the
-                 * ownership, but the effective user ID is not the
-                 * super-user and the process is not an owner of the file.
-                 * Ref: posix-testsuite/chown/07.t
-                 */
-                if ((frame->root->uid != 0) && (gf_attr_uid_set (valid))) {
-                        if (buf->ia_uid != setbuf->ia_uid) {
-                                op_ret = -1;
-                                op_errno = EPERM;
-                                goto out;
-                        }
-                }
-
-                /* non-super-user can modify file group if he is owner of a
-                 * file and gid he is setting is in his groups list.
-                 * Ref: posix-testsuite/chown/00.t
-                 */
-                if ((frame->root->uid != 0) && (gf_attr_gid_set (valid))) {
-                        if (frame->root->uid != buf->ia_uid) {
-                                op_ret = -1;
-                                op_errno = EPERM;
-                                goto out;
-                        }
-
-                        op_ret = ac_test_access (setbuf, 0, frame->root->gid,
-                                                 frame->root->groups,
-                                                 frame->root->ngrps,
-                                                 ACCTEST_DONTCARE,
-                                                 ACCTEST_GROUP, &op_errno);
-                        if (op_ret == -1)
-                                goto out;
-                }
-        }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (setattr, frame, -1, op_errno, NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-int32_t
-ac_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *buf,
-            int32_t valid)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-
-        stub = fop_setattr_stub (frame, ac_setattr_resume, loc, buf, valid);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        STACK_WIND (frame, ac_setattr_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, loc);
-        ret = 0;
-
-out:
-        if (ret < 0)
-                STACK_UNWIND_STRICT (setattr, frame, -1, -ret, NULL, NULL);
-
-        return 0;
-}
-
-
-int32_t
-ac_fsetattr_resume (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                    struct iatt *buf, int32_t valid)
-{
-        STACK_WIND (frame, default_fsetattr_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->fsetattr, fd, buf, valid);
-        return 0;
-}
-
-
-int32_t
-ac_fsetattr_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno, struct iatt *buf)
-{
-        call_stub_t     *stub = NULL;
-        int32_t         valid = 0;
-        struct iatt     *setbuf = NULL;
-
-        stub = __get_frame_stub (frame);
-        if (op_ret == -1)
-                goto out;
-
-        op_ret = ac_test_access (buf, frame->root->uid, frame->root->gid,
-                                 frame->root->groups, frame->root->ngrps,
-                                 ACCTEST_DONTCARE, ACCTEST_OWNER,
-                                 &op_errno);
-        if (op_ret == -1)
-                goto out;
-
-        valid = stub->args.fsetattr.valid;
-        setbuf = &stub->args.fsetattr.stbuf;
-        if (gf_attr_uid_set (valid) && gf_attr_gid_set (valid)) {
-                /* chown returns EPERM if the operation would change the
-                 * ownership, but the effective user ID is not the
-                 * super-user and the process is not an owner of the file.
-                 * Ref: posix-testsuite/chown/07.t
-                 */
-                if ((frame->root->uid != 0) && (gf_attr_uid_set (valid))) {
-                        if (buf->ia_uid != setbuf->ia_uid) {
-                                op_ret = -1;
-                                op_errno = EPERM;
-                                goto out;
-                        }
-                }
-
-                /* non-super-user can modify file group if he is owner of a
-                 * file and gid he is setting is in his groups list.
-                 * Ref: posix-testsuite/chown/00.t
-                 */
-                if ((frame->root->uid != 0) && (gf_attr_gid_set (valid))) {
-                        if (frame->root->uid != buf->ia_uid) {
-                                op_ret = -1;
-                                op_errno = EPERM;
-                                goto out;
-                        }
-
-                        op_ret = ac_test_access (buf, 0, frame->root->gid,
-                                                 frame->root->groups,
-                                                 frame->root->ngrps,
-                                                 ACCTEST_DONTCARE,
-                                                 ACCTEST_GROUP, &op_errno);
-                        if (op_ret == -1)
-                                goto out;
-                }
-        }
-
-        call_resume (stub);
-out:
-        if (op_ret < 0) {
-                STACK_UNWIND_STRICT (fsetattr, frame, -1, op_errno, NULL, NULL);
-                if (stub)
-                        call_stub_destroy (stub);
-        }
-
-        return 0;
-}
-
-
-int32_t
-ac_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *buf,
-             int32_t valid)
-{
-        call_stub_t     *stub = NULL;
-        int             ret = -EFAULT;
-
-        stub = fop_fsetattr_stub (frame, ac_fsetattr_resume, fd, buf, valid);
-	if (!stub) {
-		gf_log (this->name, GF_LOG_ERROR, "cannot create call stub: "
-                        "(out of memory)");
-                ret = -ENOMEM;
-                goto out;
-	}
-
-        frame->local = stub;
-        STACK_WIND (frame, ac_fsetattr_fstat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->fstat, fd);
-        ret = 0;
-
-out:
-        if (ret < 0)
-                STACK_UNWIND_STRICT (fsetattr, frame, -1, -ret, NULL, NULL);
-
-        return 0;
-}
-
-
-struct xlator_fops fops = {
-        .truncate       = ac_truncate,
-        .access         = ac_access,
-        .readlink       = ac_readlink,
-        .mknod          = ac_mknod,
-        .mkdir          = ac_mkdir,
-        .unlink         = ac_unlink,
-        .rmdir          = ac_rmdir,
-        .symlink        = ac_symlink,
-        .rename         = ac_rename,
-        .link           = ac_link,
-        .create         = ac_create,
-        .open           = ac_open,
-        .readv          = ac_readv,
-        .writev         = ac_writev,
-        .opendir        = ac_opendir,
-        .setattr        = ac_setattr,
-        .fsetattr       = ac_fsetattr,
-};
-
-int
-init (xlator_t *this)
-{
-        return 0;
-}
-
-void
-fini (xlator_t *this)
-{
-        return;
-}
-
-struct xlator_cbks cbks = {
-};
diff --git a/xlators/features/access-control/src/access-control.h b/xlators/features/access-control/src/access-control.h
deleted file mode 100644
index bfc0d7752..000000000
--- a/xlators/features/access-control/src/access-control.h
+++ /dev/null
@@ -1,55 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-
-#ifndef __ACCESS_CONTROL_H_
-#define __ACCESS_CONTROL_H_
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#define ACTRL                   "access-control"
-#define ACCTEST_READ            0x1
-#define ACCTEST_WRITE           0x2
-#define ACCTEST_EXEC            0x4
-#define ACCTEST_DONTCARE        0x8
-
-/* Note if the caller is only interested in ownership test i.e. one of the below
-+ * in combination with GF_ACCTEST_DONTCARE, then only one type of user's owner
-+ * ship can be tested with one call to gf_test_access, i.e. we can only
-+ * check of either owner and group, if both need to be tested for a specific
-+ * (uid, gid) pair then two calls will be needed.
-+ */
-#define ACCTEST_OWNER           0x1
-#define ACCTEST_GROUP           0x2
-#define ACCTEST_OTHER           0x4
-
-/* Signifies any user, as long as we get access. */
-#define ACCTEST_ANY             (ACCTEST_OWNER | ACCTEST_GROUP | ACCTEST_OTHER)
-
-#define ac_test_owner(acc)      ((acc) & ACCTEST_OWNER)
-#define ac_test_group(acc)      ((acc) & ACCTEST_GROUP)
-#define ac_test_other(acc)      ((acc) & ACCTEST_OTHER)
-#define ac_test_dontcare(acc)   ((acc) & ACCTEST_DONTCARE)
-#define ac_test_read(acc)       ((acc) & ACCTEST_READ)
-#define ac_test_write(acc)      ((acc) & ACCTEST_WRITE)
-#define ac_test_exec(acc)       ((acc) & ACCTEST_EXEC)
-#endif
diff --git a/xlators/features/changelog/Makefile.am b/xlators/features/changelog/Makefile.am
new file mode 100644
index 000000000..153bb6850
--- /dev/null
+++ b/xlators/features/changelog/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src lib
+
+CLEANFILES =
diff --git a/xlators/features/access-control/Makefile.am b/xlators/features/changelog/lib/Makefile.am
index a985f42a8..a985f42a8 100644
--- a/xlators/features/access-control/Makefile.am
+++ b/xlators/features/changelog/lib/Makefile.am
diff --git a/xlators/features/changelog/lib/examples/c/get-changes.c b/xlators/features/changelog/lib/examples/c/get-changes.c
new file mode 100644
index 000000000..14562585a
--- /dev/null
+++ b/xlators/features/changelog/lib/examples/c/get-changes.c
@@ -0,0 +1,87 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+/**
+ * get set of new changes every 10 seconds (just print the file names)
+ *
+ * Compile it using:
+ *  gcc -o getchanges `pkg-config --cflags libgfchangelog` get-changes.c \
+ *  `pkg-config --libs libgfchangelog`
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <sys/un.h>
+#include <limits.h>
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <errno.h>
+
+#include "changelog.h"
+
+#define handle_error(fn)                                \
+        printf ("%s (reason: %s)\n", fn, strerror (errno))
+
+int
+main (int argc, char ** argv)
+{
+        int     i           = 0;
+        int     ret         = 0;
+        ssize_t nr_changes  = 0;
+        ssize_t changes     = 0;
+        char fbuf[PATH_MAX] = {0,};
+
+        /* get changes for brick "/home/vshankar/export/yow/yow-1" */
+        ret = gf_changelog_register ("/home/vshankar/export/yow/yow-1",
+                                     "/tmp/scratch", "/tmp/change.log", 9, 5);
+        if (ret) {
+                handle_error ("register failed");
+                goto out;
+        }
+
+        while (1) {
+                i = 0;
+                nr_changes = gf_changelog_scan ();
+                if (nr_changes < 0) {
+                        handle_error ("scan(): ");
+                        break;
+                }
+
+                if (nr_changes == 0)
+                        goto next;
+
+                printf ("Got %ld changelog files\n", nr_changes);
+
+                while ( (changes =
+                         gf_changelog_next_change (fbuf, PATH_MAX)) > 0) {
+                        printf ("changelog file [%d]: %s\n", ++i, fbuf);
+
+                        /* process changelog */
+                        /* ... */
+                        /* ... */
+                        /* ... */
+                        /* done processing */
+
+                        ret = gf_changelog_done (fbuf);
+                        if (ret)
+                                handle_error ("gf_changelog_done");
+                }
+
+                if (changes == -1)
+                        handle_error ("gf_changelog_next_change");
+
+        next:
+                sleep (10);
+        }
+
+ out:
+        return ret;
+}
diff --git a/xlators/features/changelog/lib/examples/python/changes.py b/xlators/features/changelog/lib/examples/python/changes.py
new file mode 100644
index 000000000..d21db8eab
--- /dev/null
+++ b/xlators/features/changelog/lib/examples/python/changes.py
@@ -0,0 +1,32 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import libgfchangelog
+
+cl = libgfchangelog.Changes()
+
+def get_changes(brick, scratch_dir, log_file, log_level, interval):
+    change_list = []
+    try:
+        cl.cl_register(brick, scratch_dir, log_file, log_level)
+        while True:
+            cl.cl_scan()
+            change_list = cl.cl_getchanges()
+            if change_list:
+                print change_list
+            for change in change_list:
+                print('done with %s' % (change))
+                cl.cl_done(change)
+            time.sleep(interval)
+    except OSError:
+        ex = sys.exc_info()[1]
+        print ex
+
+if __name__ == '__main__':
+    if len(sys.argv) != 5:
+        print("usage: %s <brick> <scratch-dir> <log-file> <fetch-interval>"
+              % (sys.argv[0]))
+        sys.exit(1)
+    get_changes(sys.argv[1], sys.argv[2], sys.argv[3], 9, int(sys.argv[4]))
diff --git a/xlators/features/changelog/lib/examples/python/libgfchangelog.py b/xlators/features/changelog/lib/examples/python/libgfchangelog.py
new file mode 100644
index 000000000..68ec3baf1
--- /dev/null
+++ b/xlators/features/changelog/lib/examples/python/libgfchangelog.py
@@ -0,0 +1,64 @@
+import os
+from ctypes import *
+from ctypes.util import find_library
+
+class Changes(object):
+    libgfc = CDLL(find_library("gfchangelog"), use_errno=True)
+
+    @classmethod
+    def geterrno(cls):
+        return get_errno()
+
+    @classmethod
+    def raise_oserr(cls):
+        errn = cls.geterrno()
+        raise OSError(errn, os.strerror(errn))
+
+    @classmethod
+    def _get_api(cls, call):
+        return getattr(cls.libgfc, call)
+
+    @classmethod
+    def cl_register(cls, brick, path, log_file, log_level, retries = 0):
+        ret = cls._get_api('gf_changelog_register')(brick, path,
+                                                    log_file, log_level, retries)
+        if ret == -1:
+            cls.raise_oserr()
+
+    @classmethod
+    def cl_scan(cls):
+        ret = cls._get_api('gf_changelog_scan')()
+        if ret == -1:
+            cls.raise_oserr()
+
+    @classmethod
+    def cl_startfresh(cls):
+        ret = cls._get_api('gf_changelog_start_fresh')()
+        if ret == -1:
+            cls.raise_oserr()
+
+    @classmethod
+    def cl_getchanges(cls):
+        """ remove hardcoding for path name length """
+        def clsort(f):
+            return f.split('.')[-1]
+        changes = []
+        buf = create_string_buffer('\0', 4096)
+        call = cls._get_api('gf_changelog_next_change')
+
+        while True:
+            ret = call(buf, 4096)
+            if ret in (0, -1):
+                break;
+            changes.append(buf.raw[:ret-1])
+        if ret == -1:
+            cls.raise_oserr()
+        # cleanup tracker
+        cls.cl_startfresh()
+        return sorted(changes, key=clsort)
+
+    @classmethod
+    def cl_done(cls, clfile):
+        ret = cls._get_api('gf_changelog_done')(clfile)
+        if ret == -1:
+            cls.raise_oserr()
diff --git a/xlators/features/changelog/lib/src/Makefile.am b/xlators/features/changelog/lib/src/Makefile.am
new file mode 100644
index 000000000..fbaaea628
--- /dev/null
+++ b/xlators/features/changelog/lib/src/Makefile.am
@@ -0,0 +1,37 @@
+libgfchangelog_la_CFLAGS = -Wall $(GF_CFLAGS) $(GF_DARWIN_LIBGLUSTERFS_CFLAGS) \
+			-DDATADIR=\"$(localstatedir)\"
+
+libgfchangelog_la_CPPFLAGS = $(GF_CPPFLAGS) -D__USE_FILE_OFFSET64 -fpic \
+				-I../../../src/ -I$(top_srcdir)/libglusterfs/src \
+				-I$(top_srcdir)/xlators/features/changelog/src \
+				-DDATADIR=\"$(localstatedir)\"
+
+libgfchangelog_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \
+			$(GF_GLUSTERFS_LIBS)
+
+libgfchangelog_la_LDFLAGS = $(GF_LDFLAGS)
+
+libgfchangelogdir = $(includedir)/glusterfs/gfchangelog
+lib_LTLIBRARIES = libgfchangelog.la
+
+CONTRIB_BUILDDIR = $(top_builddir)/contrib
+
+libgfchangelog_la_SOURCES = gf-changelog.c gf-changelog-process.c \
+		gf-changelog-helpers.c $(CONTRIBDIR)/uuid/clear.c \
+		$(CONTRIBDIR)/uuid/copy.c $(CONTRIBDIR)/uuid/gen_uuid.c \
+		$(CONTRIBDIR)/uuid/pack.c $(CONTRIBDIR)/uuid/parse.c \
+		$(CONTRIBDIR)/uuid/unparse.c $(CONTRIBDIR)/uuid/uuid_time.c \
+		$(CONTRIBDIR)/uuid/compare.c $(CONTRIBDIR)/uuid/isnull.c \
+		$(CONTRIBDIR)/uuid/unpack.c
+
+noinst_HEADERS = gf-changelog-helpers.h $(CONTRIBDIR)/uuid/uuidd.h \
+		$(CONTRIBDIR)/uuid/uuid.h $(CONTRIBDIR)/uuid/uuidP.h \
+		$(CONTRIB_BUILDDIR)/uuid/uuid_types.h
+
+libgfchangelog_HEADERS = changelog.h
+
+CLEANFILES =
+CONFIG_CLEAN_FILES = $(CONTRIB_BUILDDIR)/uuid/uuid_types.h
+
+$(top_builddir)/libglusterfs/src/libglusterfs.la:
+	$(MAKE) -C $(top_builddir)/libglusterfs/src/ all
diff --git a/xlators/features/changelog/lib/src/changelog.h b/xlators/features/changelog/lib/src/changelog.h
new file mode 100644
index 000000000..5cddfb583
--- /dev/null
+++ b/xlators/features/changelog/lib/src/changelog.h
@@ -0,0 +1,31 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _GF_CHANGELOG_H
+#define _GF_CHANGELOG_H
+
+/* API set */
+
+int
+gf_changelog_register (char *brick_path, char *scratch_dir,
+                       char *log_file, int log_levl, int max_reconnects);
+ssize_t
+gf_changelog_scan ();
+
+int
+gf_changelog_start_fresh ();
+
+ssize_t
+gf_changelog_next_change (char *bufptr, size_t maxlen);
+
+int
+gf_changelog_done (char *file);
+
+#endif
diff --git a/xlators/features/changelog/lib/src/gf-changelog-helpers.c b/xlators/features/changelog/lib/src/gf-changelog-helpers.c
new file mode 100644
index 000000000..1eef8bf04
--- /dev/null
+++ b/xlators/features/changelog/lib/src/gf-changelog-helpers.c
@@ -0,0 +1,180 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#include "changelog-mem-types.h"
+#include "gf-changelog-helpers.h"
+
+ssize_t gf_changelog_read_path (int fd, char *buffer, size_t bufsize)
+{
+        return read (fd, buffer, bufsize);
+}
+
+size_t
+gf_changelog_write (int fd, char *buffer, size_t len)
+{
+        ssize_t size = 0;
+        size_t writen = 0;
+
+        while (writen < len) {
+                size = write (fd,
+                              buffer + writen, len - writen);
+                if (size <= 0)
+                        break;
+
+                writen += size;
+        }
+
+        return writen;
+}
+
+void
+gf_rfc3986_encode (unsigned char *s, char *enc, char *estr)
+{
+        for (; *s; s++) {
+                if (estr[*s])
+                        sprintf(enc, "%c", estr[*s]);
+                else
+                        sprintf(enc, "%%%02X", *s);
+                while (*++enc);
+        }
+}
+
+/**
+ * thread safe version of readline with buffering
+ * (taken from Unix Network Programming Volume I, W.R. Stevens)
+ *
+ * This is favoured over fgets() as we'd need to ftruncate()
+ * (see gf_changelog_scan() API) to record new changelog files.
+ * stream open functions does have a truncate like api (although
+ * that can be done via @fflush(fp), @ftruncate(fd) and @fseek(fp),
+ * but this involves mixing POSIX file descriptors and stream FILE *).
+ *
+ * NOTE: This implmentation still does work with more than one fd's
+ *       used to perform gf_readline(). For this very reason it's not
+ *       made a part of libglusterfs.
+ */
+
+static pthread_key_t rl_key;
+static pthread_once_t rl_once = PTHREAD_ONCE_INIT;
+
+static void
+readline_destructor (void *ptr)
+{
+        GF_FREE (ptr);
+}
+
+static void
+readline_once (void)
+{
+        pthread_key_create (&rl_key, readline_destructor);
+}
+
+static ssize_t
+my_read (read_line_t *tsd, int fd, char *ptr)
+{
+        if (tsd->rl_cnt <= 0) {
+                if ( (tsd->rl_cnt = read (fd, tsd->rl_buf, MAXLINE)) < 0 )
+                        return -1;
+                else if (tsd->rl_cnt == 0)
+                        return 0;
+                tsd->rl_bufptr = tsd->rl_buf;
+        }
+
+        tsd->rl_cnt--;
+        *ptr = *tsd->rl_bufptr++;
+        return 1;
+}
+
+static int
+gf_readline_init_once (read_line_t **tsd)
+{
+        if (pthread_once (&rl_once, readline_once) != 0)
+                return -1;
+
+        *tsd = pthread_getspecific (rl_key);
+        if (*tsd)
+                goto out;
+
+        *tsd = GF_CALLOC (1, sizeof (**tsd),
+                          gf_changelog_mt_libgfchangelog_rl_t);
+        if (!*tsd)
+                return -1;
+
+        if (pthread_setspecific (rl_key, *tsd) != 0)
+                return -1;
+
+ out:
+        return 0;
+}
+
+ssize_t
+gf_readline (int fd, void *vptr, size_t maxlen)
+{
+        size_t       n   = 0;
+        size_t       rc  = 0;
+        char         c   = ' ';
+        char        *ptr = NULL;
+        read_line_t *tsd = NULL;
+
+        if (gf_readline_init_once (&tsd))
+                return -1;
+
+        ptr = vptr;
+        for (n = 1; n < maxlen; n++) {
+                if ( (rc = my_read (tsd, fd, &c)) == 1 ) {
+                        *ptr++ = c;
+                        if (c == '\n')
+                                break;
+                } else if (rc == 0) {
+                        *ptr = '\0';
+                        return (n - 1);
+                } else
+                        return -1;
+        }
+
+        *ptr = '\0';
+        return n;
+
+}
+
+off_t
+gf_lseek (int fd, off_t offset, int whence)
+{
+        off_t        off = 0;
+        read_line_t *tsd = NULL;
+
+        if (gf_readline_init_once (&tsd))
+                return -1;
+
+        if ( (off = lseek (fd, offset, whence)) == -1)
+                return -1;
+
+        tsd->rl_cnt = 0;
+        tsd->rl_bufptr = tsd->rl_buf;
+
+        return off;
+}
+
+int
+gf_ftruncate (int fd, off_t length)
+{
+        read_line_t *tsd = NULL;
+
+        if (gf_readline_init_once (&tsd))
+                return -1;
+
+        if (ftruncate (fd, 0))
+                return -1;
+
+        tsd->rl_cnt = 0;
+        tsd->rl_bufptr = tsd->rl_buf;
+
+        return 0;
+}
diff --git a/xlators/features/changelog/lib/src/gf-changelog-helpers.h b/xlators/features/changelog/lib/src/gf-changelog-helpers.h
new file mode 100644
index 000000000..3aa6ed7b8
--- /dev/null
+++ b/xlators/features/changelog/lib/src/gf-changelog-helpers.h
@@ -0,0 +1,97 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _GF_CHANGELOG_HELPERS_H
+#define _GF_CHANGELOG_HELPERS_H
+
+#include <unistd.h>
+#include <dirent.h>
+#include <limits.h>
+#include <pthread.h>
+
+#include <xlator.h>
+
+#define GF_CHANGELOG_TRACKER  "tracker"
+
+#define GF_CHANGELOG_CURRENT_DIR    ".current"
+#define GF_CHANGELOG_PROCESSED_DIR  ".processed"
+#define GF_CHANGELOG_PROCESSING_DIR ".processing"
+
+#ifndef MAXLINE
+#define MAXLINE 4096
+#endif
+
+#define GF_CHANGELOG_FILL_BUFFER(ptr, ascii, off, len) do {     \
+                memcpy (ascii + off, ptr, len);                 \
+                off += len;                                     \
+        } while (0)
+
+typedef struct read_line {
+        int rl_cnt;
+        char *rl_bufptr;
+        char rl_buf[MAXLINE];
+} read_line_t;
+
+typedef struct gf_changelog {
+        xlator_t *this;
+
+        /* 'processing' directory stream */
+        DIR *gfc_dir;
+
+        /* fd to the tracker file */
+        int gfc_fd;
+
+        /* connection retries */
+        int gfc_connretries;
+
+        char gfc_sockpath[PATH_MAX];
+
+        char gfc_brickpath[PATH_MAX];
+
+        /* socket for recieving notifications */
+        int gfc_sockfd;
+
+        char *gfc_working_dir;
+
+        /* RFC 3986 string encoding */
+        char rfc3986[256];
+
+        char gfc_current_dir[PATH_MAX];
+        char gfc_processed_dir[PATH_MAX];
+        char gfc_processing_dir[PATH_MAX];
+
+        pthread_t gfc_changelog_processor;
+} gf_changelog_t;
+
+int
+gf_changelog_notification_init (xlator_t *this, gf_changelog_t *gfc);
+
+void *
+gf_changelog_process (void *data);
+
+ssize_t
+gf_changelog_read_path (int fd, char *buffer, size_t bufsize);
+
+void
+gf_rfc3986_encode (unsigned char *s, char *enc, char *estr);
+
+size_t
+gf_changelog_write (int fd, char *buffer, size_t len);
+
+ssize_t
+gf_readline (int fd, void *vptr, size_t maxlen);
+
+int
+gf_ftruncate (int fd, off_t length);
+
+off_t
+gf_lseek (int fd, off_t offset, int whence);
+
+#endif
diff --git a/xlators/features/changelog/lib/src/gf-changelog-process.c b/xlators/features/changelog/lib/src/gf-changelog-process.c
new file mode 100644
index 000000000..df7204931
--- /dev/null
+++ b/xlators/features/changelog/lib/src/gf-changelog-process.c
@@ -0,0 +1,571 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#include <unistd.h>
+#include <pthread.h>
+
+#include "uuid.h"
+#include "globals.h"
+#include "glusterfs.h"
+
+#include "gf-changelog-helpers.h"
+
+/* from the changelog translator */
+#include "changelog-misc.h"
+
+extern int byebye;
+
+/**
+ * number of gfid records after fop number
+ */
+int nr_gfids[] = {
+        [GF_FOP_MKNOD]   = 1,
+        [GF_FOP_MKDIR]   = 1,
+        [GF_FOP_UNLINK]  = 1,
+        [GF_FOP_RMDIR]   = 1,
+        [GF_FOP_SYMLINK] = 1,
+        [GF_FOP_RENAME]  = 2,
+        [GF_FOP_LINK]    = 1,
+        [GF_FOP_CREATE]  = 1,
+};
+
+static char *
+binary_to_ascii (uuid_t uuid)
+{
+        return uuid_utoa (uuid);
+}
+
+static char *
+conv_noop (char *ptr) { return ptr; }
+
+#define VERIFY_SEPARATOR(ptr, plen, perr)       \
+        {                                       \
+                if (*(ptr + plen) != '\0') {    \
+                        perr = 1;               \
+                        break;                  \
+                }                               \
+        }
+
+#define MOVER_MOVE(mover, nleft, bytes)         \
+        {                                       \
+                mover += bytes;                 \
+                nleft -= bytes;                 \
+        }                                       \
+
+#define PARSE_GFID(mov, ptr, le, fn, perr)                      \
+        {                                                       \
+                VERIFY_SEPARATOR (mov, le, perr);               \
+                ptr = fn (mov);                                 \
+                if (!ptr) {                                     \
+                        perr = 1;                               \
+                        break;                                  \
+                }                                               \
+        }
+
+#define FILL_AND_MOVE(pt, buf, of, mo, nl, le)                          \
+        {                                                               \
+                GF_CHANGELOG_FILL_BUFFER (pt, buf, of, strlen (pt));    \
+                MOVER_MOVE (mo, nl, le);                                \
+        }
+
+
+#define PARSE_GFID_MOVE(ptr, uuid, mover, nleft, perr)          \
+        {                                                       \
+                memcpy (uuid, mover, sizeof (uuid_t));          \
+                ptr = binary_to_ascii (uuid);                   \
+                if (!ptr) {                                     \
+                        perr = 1;                               \
+                        break;                                  \
+                }                                               \
+                MOVER_MOVE (mover, nleft, sizeof (uuid_t));     \
+        }                                                       \
+
+#define LINE_BUFSIZE  3*PATH_MAX /* enough buffer for extra chars too */
+
+/**
+ * using mmap() makes parsing easy. fgets() cannot be used here as
+ * the binary gfid could contain a line-feed (0x0A), in that case fgets()
+ * would read an incomplete line and parsing would fail. using POSIX fds
+ * would result is additional code to maintain state in case of partial
+ * reads of data (where multiple entries do not fit extirely in the buffer).
+ *
+ * mmap() gives the flexibility of pointing to an offset in the file
+ * without us worrying about reading it in memory (VM does that for us for
+ * free).
+ */
+
+static int
+gf_changelog_parse_binary (xlator_t *this,
+                           gf_changelog_t *gfc, int from_fd, int to_fd,
+                           size_t start_offset, struct stat *stbuf)
+
+{
+        int     ret              = -1;
+        off_t   off              = 0;
+        off_t   nleft            = 0;
+        uuid_t  uuid             = {0,};
+        char   *ptr              = NULL;
+        char   *bname_start      = NULL;
+        char   *bname_end        = NULL;
+        char   *mover            = NULL;
+        char   *start            = NULL;
+        char    current_mover    = ' ';
+        size_t  blen             = 0;
+        int     parse_err        = 0;
+        char ascii[LINE_BUFSIZE] = {0,};
+
+        nleft = stbuf->st_size;
+
+        start = (char *) mmap (NULL, nleft,
+                               PROT_READ, MAP_PRIVATE, from_fd, 0);
+        if (!start) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "mmap() error (reason: %s)", strerror (errno));
+                goto out;
+        }
+
+        mover = start;
+
+        MOVER_MOVE (mover, nleft, start_offset);
+
+        while (nleft > 0) {
+
+                off = blen = 0;
+                ptr = bname_start = bname_end = NULL;
+
+                current_mover = *mover;
+
+                switch (current_mover) {
+                case 'D':
+                case 'M':
+                        MOVER_MOVE (mover, nleft, 1);
+                        PARSE_GFID_MOVE (ptr, uuid, mover, nleft, parse_err);
+
+                        break;
+
+                case 'E':
+                        MOVER_MOVE (mover, nleft, 1);
+                        PARSE_GFID_MOVE (ptr, uuid, mover, nleft, parse_err);
+
+                        bname_start = mover;
+                        if ( (bname_end = strchr (mover, '\n')) == NULL ) {
+                                parse_err = 1;
+                                break;
+                        }
+
+                        blen = bname_end - bname_start;
+                        MOVER_MOVE (mover, nleft, blen);
+
+                        break;
+
+                default:
+                        parse_err = 1;
+                }
+
+                if (parse_err)
+                        break;
+
+                GF_CHANGELOG_FILL_BUFFER (&current_mover, ascii, off, 1);
+                GF_CHANGELOG_FILL_BUFFER (" ", ascii, off, 1);
+                GF_CHANGELOG_FILL_BUFFER (ptr, ascii, off, strlen (ptr));
+                if (blen)
+                        GF_CHANGELOG_FILL_BUFFER (bname_start,
+                                                  ascii, off, blen);
+                GF_CHANGELOG_FILL_BUFFER ("\n", ascii, off, 1);
+
+                if (gf_changelog_write (to_fd, ascii, off) != off) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "processing binary changelog failed due to "
+                                " error in writing ascii change (reason: %s)",
+                                strerror (errno));
+                        break;
+                }
+
+                MOVER_MOVE (mover, nleft, 1);
+        }
+
+        if ( (nleft == 0) && (!parse_err))
+                ret = 0;
+
+        if (munmap (start, stbuf->st_size))
+                gf_log (this->name, GF_LOG_ERROR,
+                        "munmap() error (reason: %s)", strerror (errno));
+ out:
+        return ret;
+}
+
+/**
+ * ascii decoder:
+ *  - separate out one entry from another
+ *  - use fop name rather than fop number
+ */
+static int
+gf_changelog_parse_ascii (xlator_t *this,
+                          gf_changelog_t *gfc, int from_fd, int to_fd,
+                          size_t start_offset, struct stat *stbuf)
+{
+        int         ng            = 0;
+        int         ret           = -1;
+        int         fop           = 0;
+        int         len           = 0;
+        off_t       off           = 0;
+        off_t       nleft         = 0;
+        char       *ptr           = NULL;
+        char       *eptr          = NULL;
+        char       *start         = NULL;
+        char       *mover         = NULL;
+        int         parse_err     = 0;
+        char        current_mover = ' ';
+        char ascii[LINE_BUFSIZE]  = {0,};
+        const char *fopname       = NULL;
+
+        nleft = stbuf->st_size;
+
+        start = (char *) mmap (NULL, nleft,
+                               PROT_READ, MAP_PRIVATE, from_fd, 0);
+        if (!start) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "mmap() error (reason: %s)", strerror (errno));
+                goto out;
+        }
+
+        mover = start;
+
+        MOVER_MOVE (mover, nleft, start_offset);
+
+        while (nleft > 0) {
+                off = 0;
+                current_mover = *mover;
+
+                GF_CHANGELOG_FILL_BUFFER (&current_mover, ascii, off, 1);
+                GF_CHANGELOG_FILL_BUFFER (" ", ascii, off, 1);
+
+                switch (current_mover) {
+                case 'D':
+                case 'M':
+                        MOVER_MOVE (mover, nleft, 1);
+
+                        /* target gfid */
+                        PARSE_GFID (mover, ptr, UUID_CANONICAL_FORM_LEN,
+                                    conv_noop, parse_err);
+                        FILL_AND_MOVE(ptr, ascii, off,
+                                      mover, nleft, UUID_CANONICAL_FORM_LEN);
+                        break;
+
+                case 'E':
+                        MOVER_MOVE (mover, nleft, 1);
+
+                        /* target gfid */
+                        PARSE_GFID (mover, ptr, UUID_CANONICAL_FORM_LEN,
+                                    conv_noop, parse_err);
+                        FILL_AND_MOVE (ptr, ascii, off,
+                                       mover, nleft, UUID_CANONICAL_FORM_LEN);
+                        FILL_AND_MOVE (" ", ascii, off,
+                                       mover, nleft, 1);
+
+                        /* fop */
+                        len = strlen (mover);
+                        VERIFY_SEPARATOR (mover, len, parse_err);
+
+                        fop = atoi (mover);
+                        if ( (fopname = gf_fop_list[fop]) == NULL) {
+                                parse_err = 1;
+                                break;
+                        }
+
+                        MOVER_MOVE (mover, nleft, len);
+
+                        len = strlen (fopname);
+                        GF_CHANGELOG_FILL_BUFFER (fopname, ascii, off, len);
+
+                        /* pargfid + bname */
+                        ng = nr_gfids[fop];
+                        while (ng-- > 0) {
+                                MOVER_MOVE (mover, nleft, 1);
+                                len = strlen (mover);
+                                GF_CHANGELOG_FILL_BUFFER (" ", ascii, off, 1);
+
+                                PARSE_GFID (mover, ptr, len,
+                                            conv_noop, parse_err);
+                                eptr = calloc (3, strlen (ptr));
+                                if (!eptr) {
+                                        parse_err = 1;
+                                        break;
+                                }
+
+                                gf_rfc3986_encode ((unsigned char *) ptr,
+                                                   eptr, gfc->rfc3986);
+                                FILL_AND_MOVE (eptr, ascii, off,
+                                               mover, nleft, len);
+                                free (eptr);
+                        }
+
+                        break;
+                default:
+                        parse_err = 1;
+                }
+
+                if (parse_err)
+                        break;
+
+                GF_CHANGELOG_FILL_BUFFER ("\n", ascii, off, 1);
+
+                if (gf_changelog_write (to_fd, ascii, off) != off) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "processing ascii changelog failed due to "
+                                " wrror in writing change (reason: %s)",
+                                strerror (errno));
+                        break;
+                }
+
+                MOVER_MOVE (mover, nleft, 1);
+
+        }
+
+        if ( (nleft == 0) && (!parse_err))
+                ret = 0;
+
+        if (munmap (start, stbuf->st_size))
+                gf_log (this->name, GF_LOG_ERROR,
+                        "munmap() error (reason: %s)", strerror (errno));
+
+ out:
+        return ret;
+}
+
+#define COPY_BUFSIZE  8192
+static int
+gf_changelog_copy (xlator_t *this, int from_fd, int to_fd)
+{
+        ssize_t size                  = 0;
+        char   buffer[COPY_BUFSIZE+1] = {0,};
+
+        while (1) {
+                size = read (from_fd, buffer, COPY_BUFSIZE);
+                if (size <= 0)
+                        break;
+
+                if (gf_changelog_write (to_fd,
+                                        buffer, size) != size) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "error processing ascii changlog");
+                        size = -1;
+                        break;
+                }
+        }
+
+        return (size < 0 ? -1 : 0);
+}
+
+static int
+gf_changelog_decode (xlator_t *this, gf_changelog_t *gfc, int from_fd,
+                     int to_fd, struct stat *stbuf, int *zerob)
+{
+        int    ret        = -1;
+        int    encoding   = -1;
+        size_t elen       = 0;
+        char buffer[1024] = {0,};
+
+        CHANGELOG_GET_ENCODING (from_fd, buffer, 1024, encoding, elen);
+        if (encoding == -1) /* unknown encoding */
+                goto out;
+
+        if (!CHANGELOG_VALID_ENCODING (encoding))
+                goto out;
+
+        if (elen == stbuf->st_size) {
+                *zerob = 1;
+                goto out;
+        }
+
+        /**
+         * start processing after the header
+         */
+        lseek (from_fd, elen, SEEK_SET);
+
+        switch (encoding) {
+        case CHANGELOG_ENCODE_BINARY:
+                /**
+                 * this ideally should have been a part of changelog-encoders.c
+                 * (ie. part of the changelog translator).
+                 */
+                ret = gf_changelog_parse_binary (this, gfc, from_fd,
+                                                 to_fd, elen, stbuf);
+                break;
+
+        case CHANGELOG_ENCODE_ASCII:
+                ret = gf_changelog_parse_ascii (this, gfc, from_fd,
+                                                to_fd, elen, stbuf);
+                break;
+        default:
+                ret = gf_changelog_copy (this, from_fd, to_fd);
+        }
+
+ out:
+        return ret;
+}
+
+static int
+gf_changelog_consume (xlator_t *this, gf_changelog_t *gfc, char *from_path)
+{
+        int         ret        = -1;
+        int         fd1        = 0;
+        int         fd2        = 0;
+        int         zerob      = 0;
+        struct stat stbuf      = {0,};
+        char dest[PATH_MAX]    = {0,};
+        char to_path[PATH_MAX] = {0,};
+
+        ret = stat (from_path, &stbuf);
+        if (ret || !S_ISREG(stbuf.st_mode)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "stat failed on changelog file: %s", from_path);
+                goto out;
+        }
+
+        fd1 = open (from_path, O_RDONLY);
+        if (fd1 < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "cannot open changelog file: %s (reason: %s)",
+                        from_path, strerror (errno));
+                goto out;
+        }
+
+        (void) snprintf (to_path, PATH_MAX, "%s%s",
+                         gfc->gfc_current_dir, basename (from_path));
+        (void) snprintf (dest, PATH_MAX, "%s%s",
+                         gfc->gfc_processing_dir, basename (from_path));
+
+        fd2 = open (to_path, O_CREAT | O_TRUNC | O_RDWR,
+                    S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+        if (fd2 < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "cannot create ascii changelog file %s (reason %s)",
+                        to_path, strerror (errno));
+                goto close_fd;
+        } else {
+                ret = gf_changelog_decode (this, gfc, fd1,
+                                           fd2, &stbuf, &zerob);
+
+                close (fd2);
+
+                if (!ret) {
+                        /* move it to processing on a successfull
+                           decode */
+                        ret = rename (to_path, dest);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "error moving %s to processing dir"
+                                        " (reason: %s)", to_path,
+                                        strerror (errno));
+                }
+
+                /* remove it from .current if it's an empty file */
+                if (zerob) {
+                        ret = unlink (to_path);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "could not unlink %s (reason: %s",
+                                        to_path, strerror (errno));
+                }
+        }
+
+ close_fd:
+        close (fd1);
+
+ out:
+        return ret;
+}
+
+static char *
+gf_changelog_ext_change (xlator_t *this,
+                         gf_changelog_t *gfc, char *path, size_t readlen)
+{
+        int     alo = 0;
+        int     ret = 0;
+        size_t  len = 0;
+        char   *buf = NULL;
+
+        buf = path;
+        while (len < readlen) {
+                if (*buf == '\0') {
+                        alo = 1;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "processing changelog: %s", path);
+                        ret = gf_changelog_consume (this, gfc, path);
+                }
+
+                if (ret)
+                        break;
+
+                len++; buf++;
+                if (alo) {
+                        alo = 0;
+                        path = buf;
+                }
+        }
+
+        return (ret) ? NULL : path;
+}
+
+void *
+gf_changelog_process (void *data)
+{
+        ssize_t         len      = 0;
+        ssize_t         offlen   = 0;
+        xlator_t       *this     = NULL;
+        char           *sbuf     = NULL;
+        gf_changelog_t *gfc      = NULL;
+        char from_path[PATH_MAX] = {0,};
+
+        gfc = (gf_changelog_t *) data;
+        this = gfc->this;
+
+        pthread_detach (pthread_self());
+
+        for (;;) {
+                len = gf_changelog_read_path (gfc->gfc_sockfd,
+                                              from_path + offlen,
+                                              PATH_MAX - offlen);
+                if (len < 0)
+                        continue; /* ignore it for now */
+
+                if (len == 0) { /* close() from the changelog translator */
+                        gf_log (this->name, GF_LOG_INFO, "close from changelog"
+                                " notification translator.");
+
+                        if (gfc->gfc_connretries != 1) {
+                                if (!gf_changelog_notification_init(this, gfc))
+                                        continue;
+                        }
+
+                        byebye = 1;
+                        break;
+                }
+
+                len += offlen;
+                sbuf = gf_changelog_ext_change (this, gfc, from_path, len);
+                if (!sbuf) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "could not extract changelog filename");
+                        continue;
+                }
+
+                offlen = 0;
+                if (sbuf != (from_path + len)) {
+                        offlen = from_path + len - sbuf;
+                        memmove (from_path, sbuf, offlen);
+                }
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "byebye (%d) from processing thread...", byebye);
+        return NULL;
+}
diff --git a/xlators/features/changelog/lib/src/gf-changelog.c b/xlators/features/changelog/lib/src/gf-changelog.c
new file mode 100644
index 000000000..ca8e373e7
--- /dev/null
+++ b/xlators/features/changelog/lib/src/gf-changelog.c
@@ -0,0 +1,515 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#include <errno.h>
+#include <dirent.h>
+#include <stddef.h>
+#include <sys/types.h>
+#include <sys/socket.h>
+#include <sys/un.h>
+
+#ifndef _GNU_SOURCE
+#define _GNU_SOURCE
+#endif
+#include <string.h>
+
+#include "globals.h"
+#include "glusterfs.h"
+#include "logging.h"
+
+#include "gf-changelog-helpers.h"
+
+/* from the changelog translator */
+#include "changelog-misc.h"
+#include "changelog-mem-types.h"
+
+int byebye = 0;
+
+static void
+gf_changelog_cleanup (gf_changelog_t *gfc)
+{
+        /* socket */
+        if (gfc->gfc_sockfd != -1)
+                close (gfc->gfc_sockfd);
+        /* tracker fd */
+        if (gfc->gfc_fd != -1)
+                close (gfc->gfc_fd);
+        /* processing dir */
+        if (gfc->gfc_dir)
+                closedir (gfc->gfc_dir);
+
+        if (gfc->gfc_working_dir)
+                free (gfc->gfc_working_dir); /* allocated by realpath */
+}
+
+void
+__attribute__ ((constructor)) gf_changelog_ctor (void)
+{
+        glusterfs_ctx_t *ctx = NULL;
+
+        ctx = glusterfs_ctx_new ();
+        if (!ctx)
+                return;
+
+        if (glusterfs_globals_init (ctx)) {
+                free (ctx);
+                ctx = NULL;
+                return;
+        }
+
+        THIS->ctx = ctx;
+}
+
+void
+__attribute__ ((destructor)) gf_changelog_dtor (void)
+{
+        xlator_t        *this = NULL;
+        glusterfs_ctx_t *ctx  = NULL;
+        gf_changelog_t  *gfc  = NULL;
+
+        this = THIS;
+        if (!this)
+                return;
+
+        ctx = this->ctx;
+        gfc = this->private;
+
+        if (gfc) {
+                gf_changelog_cleanup (gfc);
+                GF_FREE (gfc);
+        }
+
+        if (ctx) {
+                pthread_mutex_destroy (&ctx->lock);
+                free (ctx);
+                ctx = NULL;
+        }
+}
+
+
+static int
+gf_changelog_open_dirs (gf_changelog_t *gfc)
+{
+        int  ret                    = -1;
+        DIR *dir                    = NULL;
+        int  tracker_fd             = 0;
+        char tracker_path[PATH_MAX] = {0,};
+
+        (void) snprintf (gfc->gfc_current_dir, PATH_MAX,
+                         "%s/"GF_CHANGELOG_CURRENT_DIR"/",
+                         gfc->gfc_working_dir);
+        ret = mkdir_p (gfc->gfc_current_dir, 0600, _gf_false);
+        if (ret)
+                goto out;
+
+        (void) snprintf (gfc->gfc_processed_dir, PATH_MAX,
+                         "%s/"GF_CHANGELOG_PROCESSED_DIR"/",
+                         gfc->gfc_working_dir);
+        ret = mkdir_p (gfc->gfc_processed_dir, 0600, _gf_false);
+        if (ret)
+                goto out;
+
+        (void) snprintf (gfc->gfc_processing_dir, PATH_MAX,
+                         "%s/"GF_CHANGELOG_PROCESSING_DIR"/",
+                         gfc->gfc_working_dir);
+        ret = mkdir_p (gfc->gfc_processing_dir, 0600, _gf_false);
+        if (ret)
+                goto out;
+
+        dir = opendir (gfc->gfc_processing_dir);
+        if (!dir) {
+                gf_log ("", GF_LOG_ERROR,
+                        "opendir() error [reason: %s]", strerror (errno));
+                goto out;
+        }
+
+        gfc->gfc_dir = dir;
+
+        (void) snprintf (tracker_path, PATH_MAX,
+                         "%s/"GF_CHANGELOG_TRACKER, gfc->gfc_working_dir);
+
+        tracker_fd = open (tracker_path, O_CREAT | O_APPEND | O_RDWR,
+                           S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+        if (tracker_fd < 0) {
+                closedir (gfc->gfc_dir);
+                ret = -1;
+                goto out;
+        }
+
+        gfc->gfc_fd = tracker_fd;
+        ret = 0;
+ out:
+        return ret;
+}
+
+int
+gf_changelog_notification_init (xlator_t *this, gf_changelog_t *gfc)
+{
+        int                ret    = 0;
+        int                len    = 0;
+        int                tries  = 0;
+        int                sockfd = 0;
+        struct sockaddr_un remote;
+
+        this = gfc->this;
+
+        if (gfc->gfc_sockfd != -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Reconnecting...");
+                close (gfc->gfc_sockfd);
+        }
+
+        sockfd = socket (AF_UNIX, SOCK_STREAM, 0);
+        if (sockfd < 0) {
+                ret = -1;
+                goto out;
+        }
+
+        CHANGELOG_MAKE_SOCKET_PATH (gfc->gfc_brickpath,
+                                    gfc->gfc_sockpath, PATH_MAX);
+        gf_log (this->name, GF_LOG_INFO,
+                "connecting to changelog socket: %s (brick: %s)",
+                gfc->gfc_sockpath, gfc->gfc_brickpath);
+
+        remote.sun_family = AF_UNIX;
+        strcpy (remote.sun_path, gfc->gfc_sockpath);
+
+        len = strlen (remote.sun_path) + sizeof (remote.sun_family);
+
+        while (tries < gfc->gfc_connretries) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "connection attempt %d/%d...",
+                        tries + 1, gfc->gfc_connretries);
+
+                /* initiate a connect */
+                if (connect (sockfd, (struct sockaddr *) &remote, len) == 0) {
+                        gfc->gfc_sockfd = sockfd;
+                        break;
+                }
+
+                tries++;
+                sleep (2);
+        }
+
+        if (tries == gfc->gfc_connretries) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "could not connect to changelog socket!"
+                        " bailing out...");
+                ret = -1;
+        } else
+                gf_log (this->name, GF_LOG_INFO,
+                        "connection successful");
+
+ out:
+        return ret;
+}
+
+int
+gf_changelog_done (char *file)
+{
+        int             ret    = -1;
+        char           *buffer = NULL;
+        xlator_t       *this   = NULL;
+        gf_changelog_t *gfc    = NULL;
+        char to_path[PATH_MAX] = {0,};
+
+        errno = EINVAL;
+
+        this = THIS;
+        if (!this)
+                goto out;
+
+        gfc = (gf_changelog_t *) this->private;
+        if (!gfc)
+                goto out;
+
+        if (!file || !strlen (file))
+                goto out;
+
+        /* make sure 'file' is inside ->gfc_working_dir */
+        buffer = realpath (file, NULL);
+        if (!buffer)
+                goto out;
+
+        if (strncmp (gfc->gfc_working_dir,
+                     buffer, strlen (gfc->gfc_working_dir)))
+                goto out;
+
+        (void) snprintf (to_path, PATH_MAX, "%s%s",
+                         gfc->gfc_processed_dir, basename (buffer));
+        gf_log (this->name, GF_LOG_DEBUG,
+                "moving %s to processed directory", file);
+        ret = rename (buffer, to_path);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "cannot move %s to %s (reason: %s)",
+                        file, to_path, strerror (errno));
+                goto out;
+        }
+
+        ret = 0;
+
+ out:
+        if (buffer)
+                free (buffer); /* allocated by realpath() */
+        return ret;
+}
+
+/**
+ * @API
+ *  for a set of changelogs, start from the begining
+ */
+int
+gf_changelog_start_fresh ()
+{
+        xlator_t *this = NULL;
+        gf_changelog_t *gfc = NULL;
+
+        this = THIS;
+        if (!this)
+                goto out;
+
+        errno = EINVAL;
+
+        gfc = (gf_changelog_t *) this->private;
+        if (!gfc)
+                goto out;
+
+        if (gf_ftruncate (gfc->gfc_fd, 0))
+                goto out;
+
+        return 0;
+
+ out:
+        return -1;
+}
+
+/**
+ * @API
+ * return the next changelog file entry. zero means all chanelogs
+ * consumed.
+ */
+ssize_t
+gf_changelog_next_change (char *bufptr, size_t maxlen)
+{
+        ssize_t         size       = 0;
+        int             tracker_fd = 0;
+        xlator_t       *this       = NULL;
+        gf_changelog_t *gfc        = NULL;
+        char buffer[PATH_MAX]      = {0,};
+
+        errno = EINVAL;
+
+        this = THIS;
+        if (!this)
+                goto out;
+
+        gfc = (gf_changelog_t *) this->private;
+        if (!gfc)
+                goto out;
+
+        tracker_fd = gfc->gfc_fd;
+
+        size = gf_readline (tracker_fd, buffer, maxlen);
+        if (size < 0)
+                goto out;
+        if (size == 0)
+                return 0;
+
+        memcpy (bufptr, buffer, size - 1);
+        *(buffer + size) = '\0';
+
+        return size;
+
+ out:
+        return -1;
+}
+
+/**
+ * @API
+ *  gf_changelog_scan() - scan and generate a list of change entries
+ *
+ * calling this api multiple times (without calling gf_changlog_done())
+ * would result new changelogs(s) being refreshed in the tracker file.
+ * This call also acts as a cancellation point for the consumer.
+ */
+ssize_t
+gf_changelog_scan ()
+{
+        int             ret        = 0;
+        int             tracker_fd = 0;
+        size_t          len        = 0;
+        size_t          off        = 0;
+        xlator_t       *this       = NULL;
+        size_t          nr_entries = 0;
+        gf_changelog_t *gfc        = NULL;
+        struct dirent  *entryp     = NULL;
+        struct dirent  *result     = NULL;
+        char buffer[PATH_MAX]      = {0,};
+
+        this = THIS;
+        if (!this)
+                goto out;
+
+        gfc = (gf_changelog_t *) this->private;
+        if (!gfc)
+                goto out;
+
+        /**
+         * do we need to protect 'byebye' with locks? worst, the
+         * consumer would get notified during next scan().
+         */
+        if (byebye) {
+                errno = ECONNREFUSED;
+                goto out;
+        }
+
+        errno = EINVAL;
+
+        tracker_fd = gfc->gfc_fd;
+
+        if (gf_ftruncate (tracker_fd, 0))
+                goto out;
+
+        len = offsetof(struct dirent, d_name)
+                + pathconf(gfc->gfc_processing_dir, _PC_NAME_MAX) + 1;
+        entryp = GF_CALLOC (1, len,
+                            gf_changelog_mt_libgfchangelog_dirent_t);
+        if (!entryp)
+                goto out;
+
+        rewinddir (gfc->gfc_dir);
+        while (1) {
+                ret = readdir_r (gfc->gfc_dir, entryp, &result);
+                if (ret || !result)
+                        break;
+
+                if ( !strcmp (basename (entryp->d_name), ".")
+                     || !strcmp (basename (entryp->d_name), "..") )
+                        continue;
+
+                nr_entries++;
+
+                GF_CHANGELOG_FILL_BUFFER (gfc->gfc_processing_dir,
+                                          buffer, off,
+                                          strlen (gfc->gfc_processing_dir));
+                GF_CHANGELOG_FILL_BUFFER (entryp->d_name, buffer,
+                                          off, strlen (entryp->d_name));
+                GF_CHANGELOG_FILL_BUFFER ("\n", buffer, off, 1);
+
+                if (gf_changelog_write (tracker_fd, buffer, off) != off) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "error writing changelog filename"
+                                " to tracker file");
+                        break;
+                }
+                off = 0;
+        }
+
+        GF_FREE (entryp);
+
+        if (!result) {
+                if (gf_lseek (tracker_fd, 0, SEEK_SET) != -1)
+                        return nr_entries;
+        }
+ out:
+        return -1;
+}
+
+/**
+ * @API
+ *  gf_changelog_register() - register a client for updates.
+ */
+int
+gf_changelog_register (char *brick_path, char *scratch_dir,
+                       char *log_file, int log_level, int max_reconnects)
+{
+        int             i    = 0;
+        int             ret  = -1;
+        int             errn = 0;
+        xlator_t       *this = NULL;
+        gf_changelog_t *gfc  = NULL;
+
+        this = THIS;
+        if (!this->ctx)
+                goto out;
+
+        errno = ENOMEM;
+
+        gfc = GF_CALLOC (1, sizeof (*gfc),
+                         gf_changelog_mt_libgfchangelog_t);
+        if (!gfc)
+                goto out;
+
+        gfc->this = this;
+
+        gfc->gfc_dir = NULL;
+        gfc->gfc_fd = gfc->gfc_sockfd = -1;
+
+        gfc->gfc_working_dir = realpath (scratch_dir, NULL);
+        if (!gfc->gfc_working_dir) {
+                errn = errno;
+                goto cleanup;
+        }
+
+        ret = gf_changelog_open_dirs (gfc);
+        if (ret) {
+                errn = errno;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "could not create entries in scratch dir");
+                goto cleanup;
+        }
+
+        /* passing ident as NULL means to use default ident for syslog */
+        if (gf_log_init (this->ctx, log_file, NULL))
+                goto cleanup;
+
+        gf_log_set_loglevel ((log_level == -1) ? GF_LOG_INFO :
+                             log_level);
+
+        gfc->gfc_connretries = (max_reconnects <= 0) ? 1 : max_reconnects;
+        (void) strncpy (gfc->gfc_brickpath, brick_path, PATH_MAX);
+
+        ret = gf_changelog_notification_init (this, gfc);
+        if (ret) {
+                errn = errno;
+                goto cleanup;
+        }
+
+        ret = gf_thread_create (&gfc->gfc_changelog_processor,
+				NULL, gf_changelog_process, gfc);
+        if (ret) {
+                errn = errno;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "error creating changelog processor thread"
+                        " new changes won't be recorded!!!");
+                goto cleanup;
+        }
+
+        for (; i < 256; i++) {
+                gfc->rfc3986[i] =
+                        (isalnum(i) || i == '~' ||
+                        i == '-' || i == '.' || i == '_') ? i : 0;
+        }
+
+        ret = 0;
+        this->private = gfc;
+
+        goto out;
+
+ cleanup:
+        gf_changelog_cleanup (gfc);
+        GF_FREE (gfc);
+        this->private = NULL;
+        errno = errn;
+
+ out:
+        return ret;
+}
diff --git a/xlators/features/changelog/src/Makefile.am b/xlators/features/changelog/src/Makefile.am
new file mode 100644
index 000000000..e85031ad4
--- /dev/null
+++ b/xlators/features/changelog/src/Makefile.am
@@ -0,0 +1,19 @@
+xlator_LTLIBRARIES = changelog.la
+
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+noinst_HEADERS = changelog-helpers.h changelog-mem-types.h changelog-rt.h \
+	changelog-misc.h changelog-encoders.h changelog-notifier.h
+
+changelog_la_LDFLAGS = -module -avoidversion
+
+changelog_la_SOURCES = changelog.c changelog-rt.c changelog-helpers.c \
+	changelog-encoders.c changelog-notifier.c
+changelog_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src -fPIC -D_FILE_OFFSET_BITS=64 \
+	-D_GNU_SOURCE -D$(GF_HOST_OS) -shared -nostartfiles -DDATADIR=\"$(localstatedir)\"
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
diff --git a/xlators/features/changelog/src/changelog-encoders.c b/xlators/features/changelog/src/changelog-encoders.c
new file mode 100644
index 000000000..553eec85c
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-encoders.c
@@ -0,0 +1,176 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "changelog-encoders.h"
+
+size_t
+entry_fn (void *data, char *buffer, gf_boolean_t encode)
+{
+        char    *tmpbuf = NULL;
+        size_t  bufsz  = 0;
+        struct changelog_entry_fields *ce = NULL;
+
+        ce = (struct changelog_entry_fields *) data;
+
+        if (encode) {
+                tmpbuf = uuid_utoa (ce->cef_uuid);
+                CHANGELOG_FILL_BUFFER (buffer, bufsz, tmpbuf, strlen (tmpbuf));
+        } else {
+                CHANGELOG_FILL_BUFFER (buffer, bufsz,
+                                       ce->cef_uuid, sizeof (uuid_t));
+        }
+
+        CHANGELOG_FILL_BUFFER (buffer, bufsz, "/", 1);
+        CHANGELOG_FILL_BUFFER (buffer, bufsz,
+                               ce->cef_bname, strlen (ce->cef_bname));
+        return bufsz;
+}
+
+size_t
+fop_fn (void *data, char *buffer, gf_boolean_t encode)
+{
+        char buf[10]          = {0,};
+        size_t         bufsz = 0;
+        glusterfs_fop_t fop   = 0;
+
+        fop = *(glusterfs_fop_t *) data;
+
+        if (encode) {
+                (void) snprintf (buf, sizeof (buf), "%d", fop);
+                CHANGELOG_FILL_BUFFER (buffer, bufsz, buf, strlen (buf));
+        } else
+                CHANGELOG_FILL_BUFFER (buffer, bufsz, &fop, sizeof (fop));
+
+        return bufsz;
+}
+
+void
+entry_free_fn (void *data)
+{
+        changelog_opt_t *co = data;
+
+        if (!co)
+                return;
+
+        GF_FREE (co->co_entry.cef_bname);
+}
+
+/**
+ * try to write all data in one shot
+ */
+
+static inline void
+changelog_encode_write_xtra (changelog_log_data_t *cld,
+                             char *buffer, size_t *off, gf_boolean_t encode)
+{
+        int              i      = 0;
+        size_t           offset = 0;
+        void            *data   = NULL;
+        changelog_opt_t *co     = NULL;
+
+        offset = *off;
+
+        co = (changelog_opt_t *) cld->cld_ptr;
+
+        for (; i < cld->cld_xtra_records; i++, co++) {
+                CHANGELOG_FILL_BUFFER (buffer, offset, "\0", 1);
+
+                switch (co->co_type) {
+                case CHANGELOG_OPT_REC_FOP:
+                        data = &co->co_fop;
+                        break;
+                case CHANGELOG_OPT_REC_ENTRY:
+                        data = &co->co_entry;
+                        break;
+                }
+
+                if (co->co_convert)
+                        offset += co->co_convert (data,
+                                                  buffer + offset, encode);
+                else /* no coversion: write it out as it is */
+                        CHANGELOG_FILL_BUFFER (buffer, offset,
+                                               data, co->co_len);
+        }
+
+        *off = offset;
+}
+
+int
+changelog_encode_ascii (xlator_t *this, changelog_log_data_t *cld)
+{
+        size_t            off      = 0;
+        size_t            gfid_len = 0;
+        char             *gfid_str = NULL;
+        char             *buffer   = NULL;
+        changelog_priv_t *priv     = NULL;
+
+        priv = this->private;
+
+        gfid_str = uuid_utoa (cld->cld_gfid);
+        gfid_len = strlen (gfid_str);
+
+        /* extra bytes for decorations */
+        buffer = alloca (gfid_len + cld->cld_ptr_len + 10);
+        CHANGELOG_STORE_ASCII (priv, buffer,
+                               off, gfid_str, gfid_len, cld);
+
+        if (cld->cld_xtra_records)
+                changelog_encode_write_xtra (cld, buffer, &off, _gf_true);
+
+        CHANGELOG_FILL_BUFFER (buffer, off, "\0", 1);
+
+        return changelog_write_change (priv, buffer, off);
+}
+
+int
+changelog_encode_binary (xlator_t *this, changelog_log_data_t *cld)
+{
+        size_t            off    = 0;
+        char             *buffer = NULL;
+        changelog_priv_t *priv   = NULL;
+
+        priv = this->private;
+
+        /* extra bytes for decorations */
+        buffer = alloca (sizeof (uuid_t) + cld->cld_ptr_len + 10);
+        CHANGELOG_STORE_BINARY (priv, buffer, off, cld->cld_gfid, cld);
+
+        if (cld->cld_xtra_records)
+                changelog_encode_write_xtra (cld, buffer, &off, _gf_false);
+
+        CHANGELOG_FILL_BUFFER (buffer, off, "\0", 1);
+
+        return changelog_write_change (priv, buffer, off);
+}
+
+static struct changelog_encoder
+cb_encoder[] = {
+        [CHANGELOG_ENCODE_BINARY] =
+        {
+                .encoder = CHANGELOG_ENCODE_BINARY,
+                .encode = changelog_encode_binary,
+        },
+        [CHANGELOG_ENCODE_ASCII] =
+        {
+                .encoder = CHANGELOG_ENCODE_ASCII,
+                .encode = changelog_encode_ascii,
+        },
+};
+
+void
+changelog_encode_change( changelog_priv_t * priv)
+{
+        priv->ce = &cb_encoder[priv->encode_mode];
+}
diff --git a/xlators/features/changelog/src/changelog-encoders.h b/xlators/features/changelog/src/changelog-encoders.h
new file mode 100644
index 000000000..a3efbee05
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-encoders.h
@@ -0,0 +1,46 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_ENCODERS_H
+#define _CHANGELOG_ENCODERS_H
+
+#include "xlator.h"
+#include "defaults.h"
+
+#include "changelog-helpers.h"
+
+#define CHANGELOG_STORE_ASCII(priv, buf, off, gfid, gfid_len, cld) do { \
+                CHANGELOG_FILL_BUFFER (buffer, off,                     \
+                                       priv->maps[cld->cld_type], 1);   \
+                CHANGELOG_FILL_BUFFER (buffer,                          \
+                                       off, gfid, gfid_len);            \
+        } while (0)
+
+#define CHANGELOG_STORE_BINARY(priv, buf, off, gfid, cld) do {          \
+                CHANGELOG_FILL_BUFFER (buffer, off,                     \
+                                       priv->maps[cld->cld_type], 1);   \
+                CHANGELOG_FILL_BUFFER (buffer,                          \
+                                       off, gfid, sizeof (uuid_t));     \
+        } while (0)
+
+size_t
+entry_fn (void *data, char *buffer, gf_boolean_t encode);
+size_t
+fop_fn (void *data, char *buffer, gf_boolean_t encode);
+void
+entry_free_fn (void *data);
+int
+changelog_encode_binary (xlator_t *, changelog_log_data_t *);
+int
+changelog_encode_ascii (xlator_t *, changelog_log_data_t *);
+void
+changelog_encode_change(changelog_priv_t *);
+
+#endif /* _CHANGELOG_ENCODERS_H */
diff --git a/xlators/features/changelog/src/changelog-helpers.c b/xlators/features/changelog/src/changelog-helpers.c
new file mode 100644
index 000000000..7ab0091b5
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-helpers.c
@@ -0,0 +1,693 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "logging.h"
+#include "iobuf.h"
+
+#include "changelog-helpers.h"
+#include "changelog-mem-types.h"
+
+#include "changelog-encoders.h"
+#include <pthread.h>
+
+void
+changelog_thread_cleanup (xlator_t *this, pthread_t thr_id)
+{
+        int   ret    = 0;
+        void *retval = NULL;
+
+        /* send a cancel request to the thread */
+        ret = pthread_cancel (thr_id);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "could not cancel thread (reason: %s)",
+                        strerror (errno));
+                goto out;
+        }
+
+        ret = pthread_join (thr_id, &retval);
+        if (ret || (retval != PTHREAD_CANCELED)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "cancel request not adhered as expected"
+                        " (reason: %s)", strerror (errno));
+        }
+
+ out:
+        return;
+}
+
+inline void *
+changelog_get_usable_buffer (changelog_local_t *local)
+{
+        changelog_log_data_t *cld = NULL;
+
+        cld = &local->cld;
+        if (!cld->cld_iobuf)
+                return NULL;
+
+        return cld->cld_iobuf->ptr;
+}
+
+inline void
+changelog_set_usable_record_and_length (changelog_local_t *local,
+                                        size_t len, int xr)
+{
+        changelog_log_data_t *cld = NULL;
+
+        cld = &local->cld;
+
+        cld->cld_ptr_len = len;
+        cld->cld_xtra_records = xr;
+}
+
+void
+changelog_local_cleanup (xlator_t *xl, changelog_local_t *local)
+{
+        int                   i   = 0;
+        changelog_opt_t      *co  = NULL;
+        changelog_log_data_t *cld = NULL;
+
+        if (!local)
+                return;
+
+        cld = &local->cld;
+
+        /* cleanup dynamic allocation for extra records */
+        if (cld->cld_xtra_records) {
+                co = (changelog_opt_t *) cld->cld_ptr;
+                for (; i < cld->cld_xtra_records; i++, co++)
+                        if (co->co_free)
+                                co->co_free (co);
+        }
+
+        CHANGELOG_IOBUF_UNREF (cld->cld_iobuf);
+
+        if (local->inode)
+                inode_unref (local->inode);
+
+        mem_put (local);
+}
+
+inline int
+changelog_write (int fd, char *buffer, size_t len)
+{
+        ssize_t size = 0;
+        size_t writen = 0;
+
+        while (writen < len) {
+                size = write (fd,
+                              buffer + writen, len - writen);
+                if (size <= 0)
+                        break;
+
+                writen += size;
+        }
+
+        return (writen != len);
+}
+
+static int
+changelog_rollover_changelog (xlator_t *this,
+                              changelog_priv_t *priv, unsigned long ts)
+{
+        int   ret            = -1;
+        int   notify         = 0;
+        char *bname          = NULL;
+        char ofile[PATH_MAX] = {0,};
+        char nfile[PATH_MAX] = {0,};
+
+        if (priv->changelog_fd != -1) {
+                close (priv->changelog_fd);
+                priv->changelog_fd = -1;
+        }
+
+        (void) snprintf (ofile, PATH_MAX,
+                         "%s/"CHANGELOG_FILE_NAME, priv->changelog_dir);
+        (void) snprintf (nfile, PATH_MAX,
+                         "%s/"CHANGELOG_FILE_NAME".%lu",
+                         priv->changelog_dir, ts);
+
+        ret = rename (ofile, nfile);
+        if (!ret)
+                notify = 1;
+
+        if (ret && (errno == ENOENT)) {
+                ret = 0;
+        }
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "error renaming %s -> %s (reason %s)",
+                        ofile, nfile, strerror (errno));
+        }
+
+        if (notify) {
+                bname = basename (nfile);
+                gf_log (this->name, GF_LOG_DEBUG, "notifying: %s", bname);
+                ret = changelog_write (priv->wfd, bname, strlen (bname) + 1);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to send file name to notify thread"
+                                " (reason: %s)", strerror (errno));
+                }
+        }
+
+        return ret;
+}
+
+int
+changelog_open (xlator_t *this,
+                changelog_priv_t *priv)
+{
+        int fd                        = 0;
+        int ret                       = -1;
+        int flags                     = 0;
+        char buffer[1024]             = {0,};
+        char changelog_path[PATH_MAX] = {0,};
+
+        (void) snprintf (changelog_path, PATH_MAX,
+                         "%s/"CHANGELOG_FILE_NAME,
+                         priv->changelog_dir);
+
+        flags |= (O_CREAT | O_RDWR);
+        if (priv->fsync_interval == 0)
+                flags |= O_SYNC;
+
+        fd = open (changelog_path, flags,
+                   S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+        if (fd < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "unable to open/create changelog file %s"
+                        " (reason: %s). change-logging will be"
+                        " inactive", changelog_path, strerror (errno));
+                goto out;
+        }
+
+        priv->changelog_fd = fd;
+
+        (void) snprintf (buffer, 1024, CHANGELOG_HEADER,
+                         CHANGELOG_VERSION_MAJOR,
+                         CHANGELOG_VERSION_MINOR,
+                         priv->ce->encoder);
+        ret = changelog_write_change (priv, buffer, strlen (buffer));
+        if (ret) {
+                close (priv->changelog_fd);
+                priv->changelog_fd = -1;
+                goto out;
+        }
+
+        ret = 0;
+
+ out:
+        return ret;
+}
+
+int
+changelog_start_next_change (xlator_t *this,
+                             changelog_priv_t *priv,
+                             unsigned long ts, gf_boolean_t finale)
+{
+        int ret = -1;
+
+        ret = changelog_rollover_changelog (this, priv, ts);
+
+        if (!ret && !finale)
+                ret = changelog_open (this, priv);
+
+        return ret;
+}
+
+/**
+ * return the length of entry
+ */
+inline size_t
+changelog_entry_length ()
+{
+        return sizeof (changelog_log_data_t);
+}
+
+int
+changelog_fill_rollover_data (changelog_log_data_t *cld, gf_boolean_t is_last)
+{
+        struct timeval tv = {0,};
+
+        cld->cld_type = CHANGELOG_TYPE_ROLLOVER;
+
+        if (gettimeofday (&tv, NULL))
+                return -1;
+
+        cld->cld_roll_time = (unsigned long) tv.tv_sec;
+        cld->cld_finale = is_last;
+        return 0;
+}
+
+int
+changelog_write_change (changelog_priv_t *priv, char *buffer, size_t len)
+{
+        return changelog_write (priv->changelog_fd, buffer, len);
+}
+
+inline int
+changelog_handle_change (xlator_t *this,
+                         changelog_priv_t *priv, changelog_log_data_t *cld)
+{
+        int ret = 0;
+
+        if (CHANGELOG_TYPE_IS_ROLLOVER (cld->cld_type)) {
+                changelog_encode_change(priv);
+                ret = changelog_start_next_change (this, priv,
+                                                   cld->cld_roll_time,
+                                                   cld->cld_finale);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Problem rolling over changelog(s)");
+                goto out;
+        }
+
+        /**
+         * case when there is reconfigure done (disabling changelog) and there
+         * are still fops that have updates in prgress.
+         */
+        if (priv->changelog_fd == -1)
+                return 0;
+
+        if (CHANGELOG_TYPE_IS_FSYNC (cld->cld_type)) {
+                ret = fsync (priv->changelog_fd);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "fsync failed (reason: %s)",
+                                strerror (errno));
+                }
+                goto out;
+        }
+
+        ret = priv->ce->encode (this, cld);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "error writing changelog to disk");
+        }
+
+ out:
+        return ret;
+}
+
+changelog_local_t *
+changelog_local_init (xlator_t *this, inode_t *inode,
+                      uuid_t gfid, int xtra_records,
+                      gf_boolean_t update_flag)
+{
+        changelog_local_t *local = NULL;
+        struct iobuf      *iobuf = NULL;
+
+        /**
+         * We relax the presence of inode if @update_flag is true.
+         * The caller (implmentation of the fop) needs to be careful to
+         * not blindly use local->inode.
+         */
+        if (!update_flag && !inode) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                                  "inode needed for version checking !!!");
+                goto out;
+        }
+
+        if (xtra_records) {
+                iobuf = iobuf_get2 (this->ctx->iobuf_pool,
+                                    xtra_records * CHANGELOG_OPT_RECORD_LEN);
+                if (!iobuf)
+                        goto out;
+        }
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                CHANGELOG_IOBUF_UNREF (iobuf);
+                goto out;
+        }
+
+        local->update_no_check = update_flag;
+
+        uuid_copy (local->cld.cld_gfid, gfid);
+
+        local->cld.cld_iobuf = iobuf;
+        local->cld.cld_xtra_records = 0; /* set by the caller */
+
+        if (inode)
+                local->inode = inode_ref (inode);
+
+ out:
+        return local;
+}
+
+int
+changelog_forget (xlator_t *this, inode_t *inode)
+{
+        uint64_t ctx_addr = 0;
+        changelog_inode_ctx_t *ctx = NULL;
+
+        inode_ctx_del (inode, this, &ctx_addr);
+        if (!ctx_addr)
+                return 0;
+
+        ctx = (changelog_inode_ctx_t *) (long) ctx_addr;
+        GF_FREE (ctx);
+
+        return 0;
+}
+
+int
+changelog_inject_single_event (xlator_t *this,
+                               changelog_priv_t *priv,
+                               changelog_log_data_t *cld)
+{
+        return priv->cd.dispatchfn (this, priv, priv->cd.cd_data, cld, NULL);
+}
+
+/**
+ * TODO: these threads have many thing in common (wake up after
+ * a certain time etc..). move them into separate routine.
+ */
+void *
+changelog_rollover (void *data)
+{
+        int                     ret   = 0;
+        xlator_t               *this  = NULL;
+        struct timeval          tv    = {0,};
+        changelog_log_data_t    cld   = {0,};
+        changelog_time_slice_t *slice = NULL;
+        changelog_priv_t       *priv  = data;
+
+        this = priv->cr.this;
+        slice = &priv->slice;
+
+        while (1) {
+                tv.tv_sec  = priv->rollover_time;
+                tv.tv_usec = 0;
+
+                ret = select (0, NULL, NULL, NULL, &tv);
+                if (ret)
+                        continue;
+
+                ret = changelog_fill_rollover_data (&cld, _gf_false);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to fill rollover data");
+                        continue;
+                }
+
+                LOCK (&priv->lock);
+                {
+                        ret = changelog_inject_single_event (this, priv, &cld);
+                        if (!ret)
+                                SLICE_VERSION_UPDATE (slice);
+                }
+                UNLOCK (&priv->lock);
+        }
+
+        return NULL;
+}
+
+void *
+changelog_fsync_thread (void *data)
+{
+        int                   ret  = 0;
+        xlator_t             *this = NULL;
+        struct timeval        tv   = {0,};
+        changelog_log_data_t  cld  = {0,};
+        changelog_priv_t     *priv = data;
+
+        this = priv->cf.this;
+        cld.cld_type = CHANGELOG_TYPE_FSYNC;
+
+        while (1) {
+                tv.tv_sec  = priv->fsync_interval;
+                tv.tv_usec = 0;
+
+                ret = select (0, NULL, NULL, NULL, &tv);
+                if (ret)
+                        continue;
+
+                ret = changelog_inject_single_event (this, priv, &cld);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to inject fsync event");
+        }
+
+        return NULL;
+}
+
+/* macros for inode/changelog version checks */
+
+#define INODE_VERSION_UPDATE(priv, inode, iver, slice, type) do {       \
+                LOCK (&inode->lock);                                    \
+                {                                                       \
+                        LOCK (&priv->lock);                             \
+                        {                                               \
+                                *iver = slice->changelog_version[type]; \
+                        }                                               \
+                        UNLOCK (&priv->lock);                           \
+                }                                                       \
+                UNLOCK (&inode->lock);                                  \
+        } while (0)
+
+#define INODE_VERSION_EQUALS_SLICE(priv, ver, slice, type, upd) do {    \
+                LOCK (&priv->lock);                                     \
+                {                                                       \
+                        upd = (ver == slice->changelog_version[type])   \
+                                ? _gf_false : _gf_true;                 \
+                }                                                       \
+                UNLOCK (&priv->lock);                                   \
+        } while (0)
+
+static int
+__changelog_inode_ctx_set (xlator_t *this,
+                           inode_t *inode, changelog_inode_ctx_t *ctx)
+{
+        uint64_t ctx_addr = (uint64_t) ctx;
+        return __inode_ctx_set (inode, this, &ctx_addr);
+}
+
+/**
+ * one shot routine to get the address and the value of a inode version
+ * for a particular type.
+ */
+static changelog_inode_ctx_t *
+__changelog_inode_ctx_get (xlator_t *this,
+                           inode_t *inode, unsigned long **iver,
+                           unsigned long *version, changelog_log_type type)
+{
+        int                    ret      = 0;
+        uint64_t               ctx_addr = 0;
+        changelog_inode_ctx_t *ctx      = NULL;
+
+        ret = __inode_ctx_get (inode, this, &ctx_addr);
+        if (ret < 0)
+                ctx_addr = 0;
+        if (ctx_addr != 0) {
+                ctx = (changelog_inode_ctx_t *) (long)ctx_addr;
+                goto out;
+        }
+
+        ctx = GF_CALLOC (1, sizeof (*ctx), gf_changelog_mt_inode_ctx_t);
+        if (!ctx)
+                goto out;
+
+        ret = __changelog_inode_ctx_set (this, inode, ctx);
+        if (ret) {
+                GF_FREE (ctx);
+                ctx = NULL;
+        }
+
+ out:
+        if (ctx && iver && version) {
+                *iver = CHANGELOG_INODE_VERSION_TYPE (ctx, type);
+                *version = **iver;
+        }
+
+        return ctx;
+}
+
+static changelog_inode_ctx_t *
+changelog_inode_ctx_get (xlator_t *this,
+                         inode_t *inode, unsigned long **iver,
+                         unsigned long *version, changelog_log_type type)
+{
+        changelog_inode_ctx_t *ctx = NULL;
+
+        LOCK (&inode->lock);
+        {
+                ctx = __changelog_inode_ctx_get (this,
+                                                 inode, iver, version, type);
+        }
+        UNLOCK (&inode->lock);
+
+        return ctx;
+}
+
+/**
+ * This is the main update routine. Locking has been made granular so as to
+ * maximize parallelism of fops - I'll try to explain it below using execution
+ * timelines.
+ *
+ * Basically, the contention is between multiple execution threads of this
+ * routine and the roll-over thread. So, instead of having a big lock, we hold
+ * granular locks: inode->lock and priv->lock. Now I'll explain what happens
+ * when there is an update and a roll-over at just about the same time.
+ * NOTE:
+ *  - the dispatcher itself synchronizes updates via it's own lock
+ *  - the slice version in incremented by the roll-over thread
+ *
+ * Case 1: When the rollover thread wins before the inode version can be
+ * compared with the slice version.
+ *
+ *          [updater]                 |             [rollover]
+ *                                    |
+ *                                    |           <SLICE: 1, 1, 1>
+ * <changelog_update>                 |
+ *   <changelog_inode_ctx_get>        |
+ *      <CTX: 1, 1, 1>                |
+ *                                    |         <dispatch-rollover-event>
+ *                                    |         LOCK (&priv->lock)
+ *                                    |            <SLICE_VERSION_UPDATE>
+ *                                    |              <SLICE: 2, 2, 2>
+ *                                    |         UNLOCK (&priv->lock)
+ *                                    |
+ * LOCK (&priv->lock)                 |
+ *   <INODE_VERSION_EQUALS_SLICE>     |
+ *    I: 1 <-> S: 2                   |
+ *    update: true                    |
+ * UNLOCK (&priv->lock)               |
+ *                                    |
+ * <if update == true>                |
+ *  <dispath-update-event>            |
+ *  <INODE_VERSION_UPDATE>            |
+ *   LOCK (&inode->lock)              |
+ *    LOCK (&priv->lock)              |
+ *     <CTX: 2, 1, 1>                 |
+ *    UNLOCK (&priv->lock)            |
+ *   UNLOCK (&inode->lock)            |
+ *
+ * Therefore, the change gets recorded in the next change (no lost change). If
+ * the slice version was ahead of the inode version (say I:1, S: 2), then
+ * anyway the comparison would result in a update (I: 1, S: 3).
+ *
+ * If the rollover time is too less, then there is another contention when the
+ * updater tries to bring up inode version to the slice version (this is also
+ * the case when the roll-over thread wakes up during INODE_VERSION_UPDATE.
+ *
+ *   <CTX: 1, 1, 1>                   |       <SLICE: 2, 2, 2>
+ *                                    |
+ *                                    |
+ * <dispath-update-event>             |
+ * <INODE_VERSION_UPDATE>             |
+ *  LOCK (&inode->lock)               |
+ *   LOCK (&priv->lock)               |
+ *    <CTX: 2, 1, 1>                  |
+ *   UNLOCK (&priv->lock)             |
+ *  UNLOCK (&inode->lock)             |
+ *                                    |         <dispatch-rollover-event>
+ *                                    |         LOCK (&priv->lock)
+ *                                    |            <SLICE_VERSION_UPDATE>
+ *                                    |              <SLICE: 3, 3, 3>
+ *                                    |         UNLOCK (&priv->lock)
+ *
+ *
+ * Case 2: When the fop thread wins
+ *
+ *          [updater]                 |             [rollover]
+ *                                    |
+ *                                    |           <SLICE: 1, 1, 1>
+ * <changelog_update>                 |
+ *   <changelog_inode_ctx_get>        |
+ *      <CTX: 0, 0, 0>                |
+ *                                    |
+ * LOCK (&priv->lock)                 |
+ *   <INODE_VERSION_EQUALS_SLICE>     |
+ *    I: 0 <-> S: 1                   |
+ *    update: true                    |
+ * UNLOCK (&priv->lock)               |
+ *                                    |         <dispatch-rollover-event>
+ *                                    |         LOCK (&priv->lock)
+ *                                    |            <SLICE_VERSION_UPDATE>
+ *                                    |              <SLICE: 2, 2, 2>
+ *                                    |         UNLOCK (&priv->lock)
+ * <if update == true>                |
+ *  <dispath-update-event>            |
+ *  <INODE_VERSION_UPDATE>            |
+ *   LOCK (&inode->lock)              |
+ *    LOCK (&priv->lock)              |
+ *     <CTX: 2, 0, 0>                 |
+ *    UNLOCK (&priv->lock)            |
+ *   UNLOCK (&inode->lock)            |
+ *
+ * Here again, if the inode version was equal to the slice version (I: 1, S: 1)
+ * then there is no need to record an update (as the equality of the two version
+ * signifies an update was recorded in the current time slice).
+ */
+inline void
+changelog_update (xlator_t *this, changelog_priv_t *priv,
+                  changelog_local_t *local, changelog_log_type type)
+{
+        int                     ret        = 0;
+        unsigned long          *iver       = NULL;
+        unsigned long           version    = 0;
+        inode_t                *inode      = NULL;
+        changelog_time_slice_t *slice      = NULL;
+        changelog_inode_ctx_t  *ctx        = NULL;
+        changelog_log_data_t   *cld_0      = NULL;
+        changelog_log_data_t   *cld_1      = NULL;
+        changelog_local_t      *next_local = NULL;
+        gf_boolean_t            need_upd   = _gf_true;
+
+        slice = &priv->slice;
+
+        /**
+         * for fops that do not require inode version checking
+         */
+        if (local->update_no_check)
+                goto update;
+
+        inode = local->inode;
+
+        ctx = changelog_inode_ctx_get (this,
+                                       inode, &iver, &version, type);
+        if (!ctx)
+                goto update;
+
+        INODE_VERSION_EQUALS_SLICE (priv, version, slice, type, need_upd);
+
+ update:
+        if (need_upd) {
+                cld_0 = &local->cld;
+                cld_0->cld_type = type;
+
+                if ( (next_local = local->prev_entry) != NULL ) {
+                        cld_1 = &next_local->cld;
+                        cld_1->cld_type = type;
+                }
+
+                ret = priv->cd.dispatchfn (this, priv,
+                                           priv->cd.cd_data, cld_0, cld_1);
+
+                /**
+                 * update after the dispatcher has successfully done
+                 * it's job.
+                 */
+                if (!local->update_no_check && iver && !ret)
+                        INODE_VERSION_UPDATE (priv, inode, iver, slice, type);
+        }
+
+        return;
+}
diff --git a/xlators/features/changelog/src/changelog-helpers.h b/xlators/features/changelog/src/changelog-helpers.h
new file mode 100644
index 000000000..ad79636b0
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-helpers.h
@@ -0,0 +1,395 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_HELPERS_H
+#define _CHANGELOG_HELPERS_H
+
+#include "locking.h"
+#include "timer.h"
+#include "pthread.h"
+#include "iobuf.h"
+
+#include "changelog-misc.h"
+
+/**
+ * the changelog entry
+ */
+typedef struct changelog_log_data {
+        /* rollover related */
+        unsigned long cld_roll_time;
+
+        /* reopen changelog? */
+        gf_boolean_t cld_finale;
+
+        changelog_log_type cld_type;
+
+        /**
+         * sincd gfid is _always_ a necessity, it's not a part
+         * of the iobuf. by doing this we do not add any overhead
+         * for data and metadata related fops.
+         */
+        uuid_t        cld_gfid;
+
+        /**
+         * iobufs are used for optionals records: pargfid, path,
+         * write offsets etc.. It's the fop implementers job
+         * to allocate (iobuf_get() in the fop) and get unref'ed
+         * in the callback (CHANGELOG_STACK_UNWIND).
+         */
+        struct iobuf *cld_iobuf;
+
+#define cld_ptr cld_iobuf->ptr
+
+        /**
+         * after allocation you can point this to the length of
+         * usable data, but make sure it does not exceed the
+         * the size of the requested iobuf.
+         */
+        size_t        cld_iobuf_len;
+
+#define cld_ptr_len cld_iobuf_len
+
+        /**
+         * number of optional records
+         */
+        int cld_xtra_records;
+} changelog_log_data_t;
+
+/**
+ * holder for dispatch function and private data
+ */
+
+typedef struct changelog_priv changelog_priv_t;
+
+typedef struct changelog_dispatcher {
+        void *cd_data;
+        int (*dispatchfn) (xlator_t *, changelog_priv_t *, void *,
+                           changelog_log_data_t *, changelog_log_data_t *);
+} changelog_dispatcher_t;
+
+struct changelog_bootstrap {
+        changelog_mode_t mode;
+        int (*ctor) (xlator_t *, changelog_dispatcher_t *);
+        int (*dtor) (xlator_t *, changelog_dispatcher_t *);
+};
+
+struct changelog_encoder {
+        changelog_encoder_t encoder;
+        int (*encode) (xlator_t *, changelog_log_data_t *);
+};
+
+
+/* xlator private */
+
+typedef struct changelog_time_slice {
+        /**
+         * just in case we need nanosecond granularity some day.
+         * field is unused as of now (maybe we'd need it later).
+         */
+        struct timeval tv_start;
+
+        /**
+         * version of changelog file, incremented each time changes
+         * rollover.
+         */
+        unsigned long changelog_version[CHANGELOG_MAX_TYPE];
+} changelog_time_slice_t;
+
+typedef struct changelog_rollover {
+        /* rollover thread */
+        pthread_t rollover_th;
+
+        xlator_t *this;
+} changelog_rollover_t;
+
+typedef struct changelog_fsync {
+        /* fsync() thread */
+        pthread_t fsync_th;
+
+        xlator_t *this;
+} changelog_fsync_t;
+
+# define CHANGELOG_MAX_CLIENTS  5
+typedef struct changelog_notify {
+        /* reader end of the pipe */
+        int rfd;
+
+        /* notifier thread */
+        pthread_t notify_th;
+
+        /* unique socket path */
+        char sockpath[PATH_MAX];
+
+        int socket_fd;
+
+        /**
+         * simple array of accept()'ed fds. Not scalable at all
+         * for large number of clients, but it's okay as we have
+         * a ahrd limit in this version (@CHANGELOG_MAX_CLIENTS).
+         */
+        int client_fd[CHANGELOG_MAX_CLIENTS];
+
+        xlator_t *this;
+} changelog_notify_t;
+
+struct changelog_priv {
+        gf_boolean_t active;
+
+        /* to generate unique socket file per brick */
+        char *changelog_brick;
+
+        /* logging directory */
+        char *changelog_dir;
+
+        /* one file for all changelog types */
+        int changelog_fd;
+
+        gf_lock_t lock;
+
+        /* writen end of the pipe */
+        int wfd;
+
+        /* rollover time */
+        int32_t rollover_time;
+
+        /* fsync() interval */
+        int32_t fsync_interval;
+
+        /* changelog type maps */
+        const char *maps[CHANGELOG_MAX_TYPE];
+
+        /* time slicer */
+        changelog_time_slice_t slice;
+
+        /* context of the updater */
+        changelog_dispatcher_t cd;
+
+        /* context of the rollover thread */
+        changelog_rollover_t cr;
+
+        /* context of fsync thread */
+        changelog_fsync_t cf;
+
+        /* context of the notifier thread */
+        changelog_notify_t cn;
+
+        /* operation mode */
+        changelog_mode_t op_mode;
+
+        /* bootstrap routine for 'current' logger */
+        struct changelog_bootstrap *cb;
+
+        /* encoder mode */
+        changelog_encoder_t encode_mode;
+
+        /* encoder */
+        struct changelog_encoder *ce;
+};
+
+struct changelog_local {
+        inode_t              *inode;
+        gf_boolean_t          update_no_check;
+
+        changelog_log_data_t  cld;
+
+        /**
+         * ->prev_entry is used in cases when there needs to be
+         * additional changelog entry for the parent (eg. rename)
+         * It's analogous to ->next in single linked list world,
+         * but we call it as ->prev_entry... ha ha ha
+         */
+        struct changelog_local *prev_entry;
+};
+
+typedef struct changelog_local changelog_local_t;
+
+/* inode version is stored in inode ctx */
+typedef struct changelog_inode_ctx {
+        unsigned long iversion[CHANGELOG_MAX_TYPE];
+} changelog_inode_ctx_t;
+
+#define CHANGELOG_INODE_VERSION_TYPE(ctx, type)  &(ctx->iversion[type])
+
+/**
+ * Optional Records:
+ *  fops that need to save additional information request a array of
+ *  @changelog_opt_t struct. The array is allocated via @iobufs.
+ */
+typedef enum {
+        CHANGELOG_OPT_REC_FOP,
+        CHANGELOG_OPT_REC_ENTRY,
+} changelog_optional_rec_type_t;
+
+struct changelog_entry_fields {
+        uuid_t  cef_uuid;
+        char   *cef_bname;
+};
+
+typedef struct {
+        /**
+         * @co_covert can be used to do post-processing of the record before
+         * it's persisted to the CHANGELOG. If this is NULL, then the record
+         * is persisted as per it's in memory format.
+         */
+        size_t (*co_convert) (void *data, char *buffer, gf_boolean_t encode);
+
+        /* release routines */
+        void (*co_free) (void *data);
+
+        /* type of the field */
+        changelog_optional_rec_type_t co_type;
+
+        /**
+         * sizeof of the 'valid' field in the union. This field is not used if
+         * @co_convert is specified.
+         */
+        size_t co_len;
+
+        union {
+                glusterfs_fop_t co_fop;
+                struct changelog_entry_fields co_entry;
+        };
+} changelog_opt_t;
+
+#define CHANGELOG_OPT_RECORD_LEN  sizeof (changelog_opt_t)
+
+/**
+ * helpers routines
+ */
+
+void
+changelog_thread_cleanup (xlator_t *this, pthread_t thr_id);
+inline void *
+changelog_get_usable_buffer (changelog_local_t *local);
+inline void
+changelog_set_usable_record_and_length (changelog_local_t *local,
+                                        size_t len, int xr);
+void
+changelog_local_cleanup (xlator_t *xl, changelog_local_t *local);
+changelog_local_t *
+changelog_local_init (xlator_t *this, inode_t *inode, uuid_t gfid,
+                      int xtra_records, gf_boolean_t update_flag);
+int
+changelog_start_next_change (xlator_t *this,
+                             changelog_priv_t *priv,
+                             unsigned long ts, gf_boolean_t finale);
+int
+changelog_open (xlator_t *this, changelog_priv_t *priv);
+int
+changelog_fill_rollover_data (changelog_log_data_t *cld, gf_boolean_t is_last);
+int
+changelog_inject_single_event (xlator_t *this,
+                               changelog_priv_t *priv,
+                               changelog_log_data_t *cld);
+inline size_t
+changelog_entry_length ();
+inline int
+changelog_write (int fd, char *buffer, size_t len);
+int
+changelog_write_change (changelog_priv_t *priv, char *buffer, size_t len);
+inline int
+changelog_handle_change (xlator_t *this,
+                         changelog_priv_t *priv, changelog_log_data_t *cld);
+inline void
+changelog_update (xlator_t *this, changelog_priv_t *priv,
+                  changelog_local_t *local, changelog_log_type type);
+void *
+changelog_rollover (void *data);
+void *
+changelog_fsync_thread (void *data);
+int
+changelog_forget (xlator_t *this, inode_t *inode);
+
+/* macros */
+
+#define CHANGELOG_STACK_UNWIND(fop, frame, params ...) do {             \
+                changelog_local_t *__local = NULL;                      \
+                xlator_t          *__xl    = NULL;                      \
+                if (frame) {                                            \
+                        __local      = frame->local;                    \
+                        __xl         = frame->this;                     \
+                        frame->local = NULL;                            \
+                }                                                       \
+                STACK_UNWIND_STRICT (fop, frame, params);               \
+                changelog_local_cleanup (__xl, __local);                \
+                if (__local && __local->prev_entry)                     \
+                        changelog_local_cleanup (__xl,                  \
+                                                 __local->prev_entry);  \
+        } while (0)
+
+#define CHANGELOG_IOBUF_REF(iobuf) do {         \
+                if (iobuf)                      \
+                        iobuf_ref (iobuf);      \
+        } while (0)
+
+#define CHANGELOG_IOBUF_UNREF(iobuf) do {       \
+                if (iobuf)                      \
+                        iobuf_unref (iobuf);    \
+        } while (0)
+
+#define CHANGELOG_FILL_BUFFER(buffer, off, val, len) do {       \
+                memcpy (buffer + off, val, len);                \
+                off += len;                                     \
+        } while (0)
+
+#define SLICE_VERSION_UPDATE(slice) do {                \
+                int i = 0;                              \
+                for (; i < CHANGELOG_MAX_TYPE; i++) {   \
+                        slice->changelog_version[i]++;  \
+                }                                       \
+        } while (0)
+
+#define CHANGLOG_FILL_FOP_NUMBER(co, fop, converter, xlen) do { \
+                co->co_convert = converter;                     \
+                co->co_free = NULL;                             \
+                co->co_type = CHANGELOG_OPT_REC_FOP;            \
+                co->co_fop = fop;                               \
+                xlen += sizeof (fop);                           \
+        } while (0)
+
+#define CHANGELOG_FILL_ENTRY(co, pargfid, bname,                        \
+                             converter, freefn, xlen, label)            \
+        do {                                                            \
+                co->co_convert = converter;                             \
+                co->co_free = freefn;                                   \
+                co->co_type = CHANGELOG_OPT_REC_ENTRY;                  \
+                uuid_copy (co->co_entry.cef_uuid, pargfid);             \
+                co->co_entry.cef_bname = gf_strdup(bname);              \
+                if (!co->co_entry.cef_bname)                            \
+                        goto label;                                     \
+                xlen += (UUID_CANONICAL_FORM_LEN + strlen (bname));     \
+        } while (0)
+
+#define CHANGELOG_INIT(this, local, inode, gfid, xrec)                  \
+        local = changelog_local_init (this, inode, gfid, xrec, _gf_false)
+
+#define CHANGELOG_INIT_NOCHECK(this, local, inode, gfid, xrec)          \
+        local = changelog_local_init (this, inode, gfid, xrec, _gf_true)
+
+#define CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, label) do { \
+                if (!priv->active)                              \
+                        goto label;                             \
+                /* ignore rebalance process's activity. */      \
+                if (frame->root->pid == GF_CLIENT_PID_DEFRAG)   \
+                        goto label;                             \
+        } while (0)
+
+/* ignore internal fops */
+#define CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO(dict, label) do {           \
+                if (dict && dict_get (dict, GLUSTERFS_INTERNAL_FOP_KEY)) \
+                        goto label;                                     \
+        } while (0)
+
+#define CHANGELOG_COND_GOTO(priv, cond, label) do {                    \
+                if (!priv->active || cond)                             \
+                        goto label;                                    \
+        } while (0)
+
+#endif /* _CHANGELOG_HELPERS_H */
diff --git a/xlators/features/changelog/src/changelog-mem-types.h b/xlators/features/changelog/src/changelog-mem-types.h
new file mode 100644
index 000000000..d72464eab
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-mem-types.h
@@ -0,0 +1,29 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_MEM_TYPES_H
+#define _CHANGELOG_MEM_TYPES_H
+
+#include "mem-types.h"
+
+enum gf_changelog_mem_types {
+        gf_changelog_mt_priv_t                  = gf_common_mt_end + 1,
+        gf_changelog_mt_str_t                   = gf_common_mt_end + 2,
+        gf_changelog_mt_batch_t                 = gf_common_mt_end + 3,
+        gf_changelog_mt_rt_t                    = gf_common_mt_end + 4,
+        gf_changelog_mt_inode_ctx_t             = gf_common_mt_end + 5,
+        gf_changelog_mt_libgfchangelog_t        = gf_common_mt_end + 6,
+        gf_changelog_mt_libgfchangelog_rl_t     = gf_common_mt_end + 7,
+        gf_changelog_mt_libgfchangelog_dirent_t = gf_common_mt_end + 8,
+        gf_changelog_mt_changelog_buffer_t      = gf_common_mt_end + 9,
+        gf_changelog_mt_end
+};
+
+#endif
diff --git a/xlators/features/changelog/src/changelog-misc.h b/xlators/features/changelog/src/changelog-misc.h
new file mode 100644
index 000000000..0712a3771
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-misc.h
@@ -0,0 +1,101 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_MISC_H
+#define _CHANGELOG_MISC_H
+
+#include "glusterfs.h"
+#include "common-utils.h"
+
+#define CHANGELOG_MAX_TYPE  3
+#define CHANGELOG_FILE_NAME "CHANGELOG"
+
+#define CHANGELOG_VERSION_MAJOR  1
+#define CHANGELOG_VERSION_MINOR  0
+
+#define CHANGELOG_UNIX_SOCK  DEFAULT_VAR_RUN_DIRECTORY"/changelog-%s.sock"
+
+/**
+ * header starts with the version and the format of the changelog.
+ * 'version' not much of a use now.
+ */
+#define CHANGELOG_HEADER                                                \
+        "GlusterFS Changelog | version: v%d.%d | encoding : %d\n"
+
+#define CHANGELOG_MAKE_SOCKET_PATH(brick_path, sockpath, len) do {      \
+                char md5_sum[MD5_DIGEST_LENGTH*2+1] = {0,};             \
+                md5_wrapper((unsigned char *) brick_path,               \
+                            strlen(brick_path),                         \
+                            md5_sum);                                   \
+                (void) snprintf (sockpath, len,                         \
+                                 CHANGELOG_UNIX_SOCK, md5_sum);         \
+        } while (0)
+
+/**
+ * ... used by libgfchangelog.
+ */
+#define CHANGELOG_GET_ENCODING(fd, buffer, len, enc, enc_len) do {      \
+                FILE *fp;                                               \
+                int fd_dup, maj, min;                                   \
+                                                                        \
+                enc = -1;                                               \
+                fd_dup = dup (fd);                                      \
+                                                                        \
+                if (fd_dup != -1) {                                     \
+                        fp = fdopen (fd_dup, "r");                      \
+                        if (fp) {                                       \
+                                if (fgets (buffer, len, fp)) {          \
+                                        elen = strlen (buffer);         \
+                                        sscanf (buffer,                 \
+                                                CHANGELOG_HEADER,       \
+                                                &maj, &min, &enc);      \
+                                }                                       \
+                                fclose (fp);                            \
+                        } else {                                        \
+                                close (fd_dup);                         \
+                        }                                               \
+                }                                                       \
+        } while (0)
+
+/**
+ * everything after 'CHANGELOG_TYPE_ENTRY' are internal types
+ * (ie. none of the fops trigger this type of event), hence
+ * CHANGELOG_MAX_TYPE = 3
+ */
+typedef enum {
+        CHANGELOG_TYPE_DATA = 0,
+        CHANGELOG_TYPE_METADATA,
+        CHANGELOG_TYPE_ENTRY,
+        CHANGELOG_TYPE_ROLLOVER,
+        CHANGELOG_TYPE_FSYNC,
+} changelog_log_type;
+
+/* operation modes - RT for now */
+typedef enum {
+        CHANGELOG_MODE_RT = 0,
+} changelog_mode_t;
+
+/* encoder types */
+
+typedef enum {
+        CHANGELOG_ENCODE_MIN = 0,
+        CHANGELOG_ENCODE_BINARY,
+        CHANGELOG_ENCODE_ASCII,
+        CHANGELOG_ENCODE_MAX,
+} changelog_encoder_t;
+
+#define CHANGELOG_VALID_ENCODING(enc)                                   \
+        (enc > CHANGELOG_ENCODE_MIN && enc < CHANGELOG_ENCODE_MAX)
+
+#define CHANGELOG_TYPE_IS_ENTRY(type)  (type == CHANGELOG_TYPE_ENTRY)
+#define CHANGELOG_TYPE_IS_ROLLOVER(type)  (type == CHANGELOG_TYPE_ROLLOVER)
+#define CHANGELOG_TYPE_IS_FSYNC(type)  (type == CHANGELOG_TYPE_FSYNC)
+
+#endif /* _CHANGELOG_MISC_H */
diff --git a/xlators/features/changelog/src/changelog-notifier.c b/xlators/features/changelog/src/changelog-notifier.c
new file mode 100644
index 000000000..1f8b31253
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-notifier.c
@@ -0,0 +1,314 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#include "changelog-notifier.h"
+
+#include <pthread.h>
+
+inline static void
+changelog_notify_clear_fd (changelog_notify_t *cn, int i)
+{
+        cn->client_fd[i] = -1;
+}
+
+inline static void
+changelog_notify_save_fd (changelog_notify_t *cn, int i, int fd)
+{
+        cn->client_fd[i] = fd;
+}
+
+static int
+changelog_notify_insert_fd (xlator_t *this, changelog_notify_t *cn, int fd)
+{
+        int i   = 0;
+        int ret = 0;
+
+        for (; i < CHANGELOG_MAX_CLIENTS; i++) {
+                if (cn->client_fd[i] == -1)
+                        break;
+        }
+
+        if (i == CHANGELOG_MAX_CLIENTS) {
+                /**
+                 * this case should not be hit as listen() would limit
+                 * the number of completely established connections.
+                 */
+                gf_log (this->name, GF_LOG_WARNING,
+                        "hit max client limit (%d)", CHANGELOG_MAX_CLIENTS);
+                ret = -1;
+        }
+        else
+                changelog_notify_save_fd (cn, i, fd);
+
+        return ret;
+}
+
+static void
+changelog_notify_fill_rset (changelog_notify_t *cn, fd_set *rset, int *maxfd)
+{
+        int i = 0;
+
+        FD_ZERO (rset);
+
+        FD_SET (cn->socket_fd, rset);
+        *maxfd = cn->socket_fd;
+
+        FD_SET (cn->rfd, rset);
+        *maxfd = max (*maxfd, cn->rfd);
+
+        for (; i < CHANGELOG_MAX_CLIENTS; i++) {
+                if (cn->client_fd[i] != -1) {
+                        FD_SET (cn->client_fd[i], rset);
+                        *maxfd = max (*maxfd, cn->client_fd[i]);
+                }
+        }
+
+        *maxfd = *maxfd + 1;
+}
+
+static int
+changelog_notify_client (changelog_notify_t *cn, char *path, ssize_t len)
+{
+        int i = 0;
+        int ret = 0;
+
+        for (; i < CHANGELOG_MAX_CLIENTS; i++) {
+                if (cn->client_fd[i] == -1)
+                        continue;
+
+                if (changelog_write (cn->client_fd[i],
+                                     path, len)) {
+                        ret = -1;
+
+                        close (cn->client_fd[i]);
+                        changelog_notify_clear_fd (cn, i);
+                }
+        }
+
+        return ret;
+}
+
+static void
+changelog_notifier_init (changelog_notify_t *cn)
+{
+        int i = 0;
+
+        cn->socket_fd = -1;
+
+        for (; i < CHANGELOG_MAX_CLIENTS; i++) {
+                changelog_notify_clear_fd (cn, i);
+        }
+}
+
+static void
+changelog_close_client_conn (changelog_notify_t *cn)
+{
+        int i = 0;
+
+        for (; i < CHANGELOG_MAX_CLIENTS; i++) {
+                if (cn->client_fd[i] == -1)
+                        continue;
+
+                close (cn->client_fd[i]);
+                changelog_notify_clear_fd (cn, i);
+        }
+}
+
+static void
+changelog_notifier_cleanup (void *arg)
+{
+        changelog_notify_t *cn = NULL;
+
+        cn = (changelog_notify_t *) arg;
+
+        changelog_close_client_conn (cn);
+
+        if (cn->socket_fd != -1)
+                close (cn->socket_fd);
+
+        if (cn->rfd)
+                close (cn->rfd);
+
+        if (unlink (cn->sockpath))
+                gf_log ("", GF_LOG_WARNING,
+                        "could not unlink changelog socket file"
+                        " %s (reason: %s", cn->sockpath, strerror (errno));
+}
+
+void *
+changelog_notifier (void *data)
+{
+        int                 i         = 0;
+        int                 fd        = 0;
+        int                 max_fd    = 0;
+        int                 len       = 0;
+        ssize_t             readlen   = 0;
+        xlator_t           *this      = NULL;
+        changelog_priv_t   *priv      = NULL;
+        changelog_notify_t *cn        = NULL;
+        struct sockaddr_un  local     = {0,};
+        char path[PATH_MAX]           = {0,};
+        char abspath[PATH_MAX]        = {0,};
+
+        char buffer;
+        fd_set rset;
+
+        priv = (changelog_priv_t *) data;
+
+        cn = &priv->cn;
+        this = cn->this;
+
+        pthread_cleanup_push (changelog_notifier_cleanup, cn);
+
+        changelog_notifier_init (cn);
+
+        cn->socket_fd = socket (AF_UNIX, SOCK_STREAM, 0);
+        if (cn->socket_fd < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "changelog socket error (reason: %s)",
+                        strerror (errno));
+                goto out;
+        }
+
+        CHANGELOG_MAKE_SOCKET_PATH (priv->changelog_brick,
+                                    cn->sockpath, PATH_MAX);
+        if (unlink (cn->sockpath) < 0) {
+                if (errno != ENOENT) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Could not unlink changelog socket file (%s)"
+                                " (reason: %s)",
+                                CHANGELOG_UNIX_SOCK, strerror (errno));
+                        goto cleanup;
+                }
+        }
+
+        local.sun_family = AF_UNIX;
+        strcpy (local.sun_path, cn->sockpath);
+
+        len = strlen (local.sun_path) + sizeof (local.sun_family);
+
+        /* bind to the unix domain socket */
+        if (bind (cn->socket_fd, (struct sockaddr *) &local, len) < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not bind to changelog socket (reason: %s)",
+                        strerror (errno));
+                goto cleanup;
+        }
+
+        /* listen for incoming connections */
+        if (listen (cn->socket_fd, CHANGELOG_MAX_CLIENTS) < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "listen() error on changelog socket (reason: %s)",
+                        strerror (errno));
+                goto cleanup;
+        }
+
+        /**
+         * simple select() on all to-be-read file descriptors. This method
+         * though old school works pretty well when you have a handfull of
+         * fd's to be watched (clients).
+         *
+         * Future TODO: move this to epoll based notification facility if
+         *              number of clients increase.
+         */
+        for (;;) {
+                changelog_notify_fill_rset (cn, &rset, &max_fd);
+
+                if (select (max_fd, &rset, NULL, NULL, NULL) < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "select() returned -1 (reason: %s)",
+                                strerror (errno));
+                        sleep (2);
+                        continue;
+                }
+
+                if (FD_ISSET (cn->socket_fd, &rset)) {
+                        fd = accept (cn->socket_fd, NULL, NULL);
+                        if (fd < 0) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "accept error on changelog socket"
+                                        " (reason: %s)", strerror (errno));
+                        } else if (changelog_notify_insert_fd (this, cn, fd)) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "hit max client limit");
+                        }
+                }
+
+                if (FD_ISSET (cn->rfd, &rset)) {
+                        /**
+                         * read changelog filename and notify all connected
+                         * clients.
+                         */
+                        readlen = 0;
+                        while (readlen < PATH_MAX) {
+                                len = read (cn->rfd, &path[readlen++], 1);
+                                if (len == -1) {
+                                        break;
+                                }
+
+                                if (len == 0) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "rollover thread sent EOF"
+                                                " on pipe - possibly a crash.");
+                                        /* be blunt and close all connections */
+                                        pthread_exit(NULL);
+                                }
+
+                                if (path[readlen - 1] == '\0')
+                                        break;
+                        }
+
+                        /* should we close all client connections here too? */
+                        if (len < 0 || readlen == PATH_MAX) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Could not get pathname from rollover"
+                                        " thread or pathname too long");
+                                goto process_rest;
+                        }
+
+                        (void) snprintf (abspath, PATH_MAX,
+                                         "%s/%s", priv->changelog_dir, path);
+                        if (changelog_notify_client (cn, abspath,
+                                                     strlen (abspath) + 1))
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "could not notify some clients with new"
+                                        " changelogs");
+                }
+
+        process_rest:
+                for (i = 0; i < CHANGELOG_MAX_CLIENTS; i++) {
+                        if ( (fd = cn->client_fd[i]) == -1 )
+                                continue;
+
+                        if (FD_ISSET (fd, &rset)) {
+                                /**
+                                 * the only data we accept from the client is a
+                                 * disconnect. Anything else is treated as bogus
+                                 * and is silently discarded (also warned!!!).
+                                 */
+                                if ( (readlen = read (fd, &buffer, 1)) <= 0 ) {
+                                        close (fd);
+                                        changelog_notify_clear_fd (cn, i);
+                                } else {
+                                        /* silently discard data and log */
+                                        gf_log (this->name, GF_LOG_WARNING,
+                                                "misbehaving changelog client");
+                                }
+                        }
+                }
+
+        }
+
+ cleanup:;
+        pthread_cleanup_pop (1);
+
+ out:
+        return NULL;
+}
diff --git a/xlators/features/changelog/src/changelog-notifier.h b/xlators/features/changelog/src/changelog-notifier.h
new file mode 100644
index 000000000..55e728356
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-notifier.h
@@ -0,0 +1,19 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_NOTIFIER_H
+#define _CHANGELOG_NOTIFIER_H
+
+#include "changelog-helpers.h"
+
+void *
+changelog_notifier (void *data);
+
+#endif
diff --git a/xlators/features/changelog/src/changelog-rt.c b/xlators/features/changelog/src/changelog-rt.c
new file mode 100644
index 000000000..c147f68ca
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-rt.c
@@ -0,0 +1,72 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "logging.h"
+
+#include "changelog-rt.h"
+#include "changelog-mem-types.h"
+
+int
+changelog_rt_init (xlator_t *this, changelog_dispatcher_t *cd)
+{
+        changelog_rt_t *crt = NULL;
+
+        crt = GF_CALLOC (1, sizeof (*crt),
+                         gf_changelog_mt_rt_t);
+        if (!crt)
+                return -1;
+
+        LOCK_INIT (&crt->lock);
+
+        cd->cd_data = crt;
+        cd->dispatchfn = &changelog_rt_enqueue;
+
+        return 0;
+}
+
+int
+changelog_rt_fini (xlator_t *this, changelog_dispatcher_t *cd)
+{
+        changelog_rt_t *crt = NULL;
+
+        crt = cd->cd_data;
+
+        LOCK_DESTROY (&crt->lock);
+        GF_FREE (crt);
+
+        return 0;
+}
+
+int
+changelog_rt_enqueue (xlator_t *this, changelog_priv_t *priv, void *cbatch,
+                      changelog_log_data_t *cld_0, changelog_log_data_t *cld_1)
+{
+        int             ret = 0;
+        changelog_rt_t *crt = NULL;
+
+        crt = (changelog_rt_t *) cbatch;
+
+        LOCK (&crt->lock);
+        {
+                ret = changelog_handle_change (this, priv, cld_0);
+                if (!ret && cld_1)
+                        ret = changelog_handle_change (this, priv, cld_1);
+        }
+        UNLOCK (&crt->lock);
+
+        return ret;
+}
diff --git a/xlators/features/changelog/src/changelog-rt.h b/xlators/features/changelog/src/changelog-rt.h
new file mode 100644
index 000000000..1fc2bbc5b
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-rt.h
@@ -0,0 +1,33 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_RT_H
+#define _CHANGELOG_RT_H
+
+#include "locking.h"
+#include "timer.h"
+#include "pthread.h"
+
+#include "changelog-helpers.h"
+
+/* unused as of now - may be you would need it later */
+typedef struct changelog_rt {
+        gf_lock_t lock;
+} changelog_rt_t;
+
+int
+changelog_rt_init (xlator_t *this, changelog_dispatcher_t *cd);
+int
+changelog_rt_fini (xlator_t *this, changelog_dispatcher_t *cd);
+int
+changelog_rt_enqueue (xlator_t *this, changelog_priv_t *priv, void *cbatch,
+                      changelog_log_data_t *cld_0, changelog_log_data_t *cld_1);
+
+#endif /* _CHANGELOG_RT_H */
diff --git a/xlators/features/changelog/src/changelog.c b/xlators/features/changelog/src/changelog.c
new file mode 100644
index 000000000..cea0e8c70
--- /dev/null
+++ b/xlators/features/changelog/src/changelog.c
@@ -0,0 +1,1477 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "logging.h"
+#include "iobuf.h"
+
+#include "changelog-rt.h"
+
+#include "changelog-encoders.h"
+#include "changelog-mem-types.h"
+
+#include <pthread.h>
+
+#include "changelog-notifier.h"
+
+static struct changelog_bootstrap
+cb_bootstrap[] = {
+        {
+                .mode = CHANGELOG_MODE_RT,
+                .ctor = changelog_rt_init,
+                .dtor = changelog_rt_fini,
+        },
+};
+
+/* Entry operations - TYPE III */
+
+/**
+ * entry operations do not undergo inode version checking.
+ */
+
+/* {{{ */
+
+/* rmdir */
+
+int32_t
+changelog_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (rmdir, frame, op_ret, op_errno,
+                                preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_rmdir (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, int xflags, dict_t *xdata)
+{
+        size_t            xtra_len = 0;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local,
+                                NULL, loc->inode->gfid, 2);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
+
+ wind:
+        STACK_WIND (frame, changelog_rmdir_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->rmdir,
+                    loc, xflags, xdata);
+        return 0;
+}
+
+/* unlink */
+
+int32_t
+changelog_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                      struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (unlink, frame, op_ret, op_errno,
+                                preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_unlink (call_frame_t *frame, xlator_t *this,
+                  loc_t *loc, int xflags, dict_t *xdata)
+{
+        size_t            xtra_len = 0;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+        CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO (xdata, wind);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, loc->inode->gfid, 2);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
+
+ wind:
+        STACK_WIND (frame, changelog_unlink_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->unlink,
+                    loc, xflags, xdata);
+        return 0;
+}
+
+/* rename */
+
+int32_t
+changelog_rename_cbk (call_frame_t *frame,
+                      void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno,
+                      struct iatt *buf, struct iatt *preoldparent,
+                      struct iatt *postoldparent, struct iatt *prenewparent,
+                      struct iatt *postnewparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (rename, frame, op_ret, op_errno,
+                                buf, preoldparent, postoldparent,
+                                prenewparent, postnewparent, xdata);
+        return 0;
+}
+
+
+int32_t
+changelog_rename (call_frame_t *frame, xlator_t *this,
+                  loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+        size_t            xtra_len  = 0;
+        changelog_priv_t *priv      = NULL;
+        changelog_opt_t  *co        = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        /* 3 == fop + oldloc + newloc */
+        CHANGELOG_INIT_NOCHECK (this, frame->local,
+                                NULL, oldloc->inode->gfid, 3);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, oldloc->pargfid, oldloc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, newloc->pargfid, newloc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 3);
+
+ wind:
+        STACK_WIND (frame, changelog_rename_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->rename,
+                    oldloc, newloc, xdata);
+        return 0;
+}
+
+/* link */
+
+int32_t
+changelog_link_cbk (call_frame_t *frame,
+                    void *cookie, xlator_t *this, int32_t op_ret,
+                    int32_t op_errno, inode_t *inode,
+                    struct iatt *buf, struct iatt *preparent,
+                    struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (link, frame, op_ret, op_errno,
+                                inode, buf, preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_link (call_frame_t *frame,
+                xlator_t *this, loc_t *oldloc,
+                loc_t *newloc, dict_t *xdata)
+{
+        size_t            xtra_len = 0;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+        CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO (xdata, wind);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, oldloc->gfid, 2);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, newloc->pargfid, newloc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
+
+ wind:
+        STACK_WIND (frame, changelog_link_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->link,
+                    oldloc, newloc, xdata);
+        return 0;
+}
+
+/* mkdir */
+
+int32_t
+changelog_mkdir_cbk (call_frame_t *frame,
+                     void *cookie, xlator_t *this, int32_t op_ret,
+                     int32_t op_errno, inode_t *inode,
+                     struct iatt *buf, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
+                                inode, buf, preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_mkdir (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
+{
+        int               ret      = -1;
+        uuid_t            gfid     = {0,};
+        void             *uuid_req = NULL;
+        size_t            xtra_len = 0;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        ret = dict_get_ptr (xdata, "gfid-req", &uuid_req);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "failed to get gfid from dict");
+                goto wind;
+        }
+        uuid_copy (gfid, uuid_req);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 2);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
+
+ wind:
+        STACK_WIND (frame, changelog_mkdir_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->mkdir,
+                    loc, mode, umask, xdata);
+        return 0;
+}
+
+/* symlink */
+
+int32_t
+changelog_symlink_cbk (call_frame_t *frame,
+                       void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno,
+                       inode_t *inode, struct iatt *buf, struct iatt *preparent,
+                       struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (symlink, frame, op_ret, op_errno,
+                                inode, buf, preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_symlink (call_frame_t *frame, xlator_t *this,
+                   const char *linkname, loc_t *loc,
+                   mode_t umask, dict_t *xdata)
+{
+        int               ret      = -1;
+        size_t            xtra_len = 0;
+        uuid_t            gfid     = {0,};
+        void             *uuid_req = NULL;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        ret = dict_get_ptr (xdata, "gfid-req", &uuid_req);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "failed to get gfid from dict");
+                goto wind;
+        }
+        uuid_copy (gfid, uuid_req);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 2);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
+
+ wind:
+        STACK_WIND (frame, changelog_symlink_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->symlink,
+                    linkname, loc, umask, xdata);
+        return 0;
+}
+
+/* mknod */
+
+int32_t
+changelog_mknod_cbk (call_frame_t *frame,
+                     void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, inode_t *inode,
+                     struct iatt *buf, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (mknod, frame, op_ret, op_errno,
+                                inode, buf, preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_mknod (call_frame_t *frame,
+                 xlator_t *this, loc_t *loc,
+                 mode_t mode, dev_t dev, mode_t umask, dict_t *xdata)
+{
+        int               ret      = -1;
+        uuid_t            gfid     = {0,};
+        void             *uuid_req = NULL;
+        size_t            xtra_len = 0;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        ret = dict_get_ptr (xdata, "gfid-req", &uuid_req);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "failed to get gfid from dict");
+                goto wind;
+        }
+        uuid_copy (gfid, uuid_req);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 2);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
+
+ wind:
+        STACK_WIND (frame, changelog_mknod_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->mknod,
+                    loc, mode, dev, umask, xdata);
+        return 0;
+}
+
+/* creat */
+
+int32_t
+changelog_create_cbk (call_frame_t *frame,
+                      void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno,
+                      fd_t *fd, inode_t *inode, struct iatt *buf,
+                      struct iatt *preparent,
+                      struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (create, frame,
+                                op_ret, op_errno, fd, inode,
+                                buf, preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_create (call_frame_t *frame, xlator_t *this,
+                  loc_t *loc, int32_t flags, mode_t mode,
+                  mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        int               ret      = -1;
+        uuid_t            gfid     = {0,};
+        void             *uuid_req = NULL;
+        changelog_opt_t  *co       = NULL;
+        changelog_priv_t *priv     = NULL;
+        size_t            xtra_len = 0;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        ret = dict_get_ptr (xdata, "gfid-req", &uuid_req);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "failed to get gfid from dict");
+                goto wind;
+        }
+        uuid_copy (gfid, uuid_req);
+
+        /* init with two extra records */
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 2);
+        if (!frame->local)
+                goto wind;
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
+
+ wind:
+        STACK_WIND (frame, changelog_create_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->create,
+                    loc, flags, mode, umask, fd, xdata);
+        return 0;
+}
+
+/* }}} */
+
+
+/* Metadata modification fops - TYPE II */
+
+/* {{{ */
+
+/* {f}setattr */
+
+int32_t
+changelog_fsetattr_cbk (call_frame_t *frame,
+                        void *cookie, xlator_t *this, int32_t op_ret,
+                        int32_t op_errno, struct iatt *preop_stbuf,
+                        struct iatt *postop_stbuf, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (fsetattr, frame, op_ret, op_errno,
+                                preop_stbuf, postop_stbuf, xdata);
+
+        return 0;
+
+
+}
+
+int32_t
+changelog_fsetattr (call_frame_t *frame,
+                    xlator_t *this, fd_t *fd,
+                    struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        fd->inode, fd->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_fsetattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsetattr,
+                    fd, stbuf, valid, xdata);
+        return 0;
+
+
+}
+
+int32_t
+changelog_setattr_cbk (call_frame_t *frame,
+                       void *cookie, xlator_t *this, int32_t op_ret,
+                       int32_t op_errno, struct iatt *preop_stbuf,
+                       struct iatt *postop_stbuf, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (setattr, frame, op_ret, op_errno,
+                                preop_stbuf, postop_stbuf, xdata);
+
+        return 0;
+}
+
+int32_t
+changelog_setattr (call_frame_t *frame,
+                   xlator_t *this, loc_t *loc,
+                   struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        loc->inode, loc->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_setattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->setattr,
+                    loc, stbuf, valid, xdata);
+        return 0;
+}
+
+/* {f}removexattr */
+
+int32_t
+changelog_fremovexattr_cbk (call_frame_t *frame,
+                            void *cookie, xlator_t *this,
+                            int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+int32_t
+changelog_fremovexattr (call_frame_t *frame, xlator_t *this,
+                        fd_t *fd, const char *name, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        fd->inode, fd->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_fremovexattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->fremovexattr,
+                    fd, name, xdata);
+        return 0;
+}
+
+int32_t
+changelog_removexattr_cbk (call_frame_t *frame,
+                           void *cookie, xlator_t *this,
+                           int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+int32_t
+changelog_removexattr (call_frame_t *frame, xlator_t *this,
+                       loc_t *loc, const char *name, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        loc->inode, loc->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_removexattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->removexattr,
+                    loc, name, xdata);
+        return 0;
+}
+
+/* {f}setxattr */
+
+int32_t
+changelog_setxattr_cbk (call_frame_t *frame,
+                        void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+int32_t
+changelog_setxattr (call_frame_t *frame,
+                    xlator_t *this, loc_t *loc,
+                    dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        loc->inode, loc->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_setxattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->setxattr,
+                    loc, dict, flags, xdata);
+        return 0;
+}
+
+int32_t
+changelog_fsetxattr_cbk (call_frame_t *frame,
+                         void *cookie, xlator_t *this, int32_t op_ret,
+                         int32_t op_errno, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+int32_t
+changelog_fsetxattr (call_frame_t *frame,
+                     xlator_t *this, fd_t *fd, dict_t *dict,
+                     int32_t flags, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        fd->inode, fd->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_fsetxattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsetxattr,
+                    fd, dict, flags, xdata);
+        return 0;
+}
+
+/* }}} */
+
+
+/* Data modification fops - TYPE I */
+
+/* {{{ */
+
+/* {f}truncate() */
+
+int32_t
+changelog_truncate_cbk (call_frame_t *frame,
+                        void *cookie, xlator_t *this, int32_t op_ret,
+                        int32_t op_errno, struct iatt *prebuf,
+                        struct iatt *postbuf, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_DATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (truncate, frame,
+                                op_ret, op_errno, prebuf, postbuf, xdata);
+        return 0;
+}
+
+int32_t
+changelog_truncate (call_frame_t *frame,
+                    xlator_t *this, loc_t *loc, off_t offset, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        loc->inode, loc->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_truncate_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->truncate,
+                    loc, offset, xdata);
+        return 0;
+}
+
+int32_t
+changelog_ftruncate_cbk (call_frame_t *frame,
+                         void *cookie, xlator_t *this, int32_t op_ret,
+                         int32_t op_errno, struct iatt *prebuf,
+                         struct iatt *postbuf, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_DATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (ftruncate, frame,
+                                op_ret, op_errno, prebuf, postbuf, xdata);
+        return 0;
+}
+
+int32_t
+changelog_ftruncate (call_frame_t *frame,
+                     xlator_t *this, fd_t *fd, off_t offset, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        fd->inode, fd->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_ftruncate_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->ftruncate,
+                    fd, offset, xdata);
+        return 0;
+}
+
+/* writev() */
+
+int32_t
+changelog_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                    struct iatt *postbuf,
+                    dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret <= 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_DATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (writev, frame,
+                                op_ret, op_errno, prebuf, postbuf, xdata);
+        return 0;
+}
+
+int32_t
+changelog_writev (call_frame_t *frame,
+                  xlator_t *this, fd_t *fd, struct iovec *vector,
+                  int32_t count, off_t offset, uint32_t flags,
+                  struct iobref *iobref, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        fd->inode, fd->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_writev_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->writev, fd, vector,
+                    count, offset, flags, iobref, xdata);
+        return 0;
+}
+
+/* }}} */
+
+/**
+ * The
+ *   - @init ()
+ *   - @fini ()
+ *   - @reconfigure ()
+ *   ... and helper routines
+ */
+
+/**
+ * needed if there are more operation modes in the future.
+ */
+static void
+changelog_assign_opmode (changelog_priv_t *priv, char *mode)
+{
+        if ( strncmp (mode, "realtime", 8) == 0 ) {
+                priv->op_mode = CHANGELOG_MODE_RT;
+        }
+}
+
+static void
+changelog_assign_encoding (changelog_priv_t *priv, char *enc)
+{
+        if ( strncmp (enc, "binary", 6) == 0 ) {
+                priv->encode_mode = CHANGELOG_ENCODE_BINARY;
+        } else if ( strncmp (enc, "ascii", 5) == 0 ) {
+                priv->encode_mode = CHANGELOG_ENCODE_ASCII;
+        }
+}
+
+/* cleanup any helper threads that are running */
+static void
+changelog_cleanup_helper_threads (xlator_t *this, changelog_priv_t *priv)
+{
+        if (priv->cr.rollover_th) {
+                changelog_thread_cleanup (this, priv->cr.rollover_th);
+                priv->cr.rollover_th = 0;
+        }
+
+        if (priv->cf.fsync_th) {
+                changelog_thread_cleanup (this, priv->cf.fsync_th);
+                priv->cf.fsync_th = 0;
+        }
+}
+
+/* spawn helper thread; cleaning up in case of errors */
+static int
+changelog_spawn_helper_threads (xlator_t *this, changelog_priv_t *priv)
+{
+        int ret = 0;
+
+        priv->cr.this = this;
+        ret = gf_thread_create (&priv->cr.rollover_th,
+				NULL, changelog_rollover, priv);
+        if (ret)
+                goto out;
+
+        if (priv->fsync_interval) {
+                priv->cf.this = this;
+                ret = gf_thread_create (&priv->cf.fsync_th,
+					NULL, changelog_fsync_thread, priv);
+        }
+
+        if (ret)
+                changelog_cleanup_helper_threads (this, priv);
+
+ out:
+        return ret;
+}
+
+/* cleanup the notifier thread */
+static int
+changelog_cleanup_notifier (xlator_t *this, changelog_priv_t *priv)
+{
+        int ret = 0;
+
+        if (priv->cn.notify_th) {
+                changelog_thread_cleanup (this, priv->cn.notify_th);
+                priv->cn.notify_th = 0;
+
+                ret = close (priv->wfd);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "error closing writer end of notifier pipe"
+                                " (reason: %s)", strerror (errno));
+        }
+
+        return ret;
+}
+
+/* spawn the notifier thread - nop if already running */
+static int
+changelog_spawn_notifier (xlator_t *this, changelog_priv_t *priv)
+{
+        int ret        = 0;
+        int flags      = 0;
+        int pipe_fd[2] = {0, 0};
+
+        if (priv->cn.notify_th)
+                goto out; /* notifier thread already running */
+
+        ret = pipe (pipe_fd);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Cannot create pipe (reason: %s)", strerror (errno));
+                goto out;
+        }
+
+        /* writer is non-blocking */
+        flags = fcntl (pipe_fd[1], F_GETFL);
+        flags |= O_NONBLOCK;
+
+        ret = fcntl (pipe_fd[1], F_SETFL, flags);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to set O_NONBLOCK flag");
+                goto out;
+        }
+
+        priv->wfd = pipe_fd[1];
+
+        priv->cn.this = this;
+        priv->cn.rfd  = pipe_fd[0];
+
+        ret = gf_thread_create (&priv->cn.notify_th,
+				NULL, changelog_notifier, priv);
+
+ out:
+        return ret;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_changelog_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_WARNING, "Memory accounting"
+                        " init failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+static int
+changelog_init (xlator_t *this, changelog_priv_t *priv)
+{
+        int                  i   = 0;
+        int                  ret = -1;
+        struct timeval       tv  = {0,};
+        changelog_log_data_t cld = {0,};
+
+        ret = gettimeofday (&tv, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "gettimeofday() failure");
+                goto out;
+        }
+
+        priv->slice.tv_start = tv;
+
+        priv->maps[CHANGELOG_TYPE_DATA]     = "D ";
+        priv->maps[CHANGELOG_TYPE_METADATA] = "M ";
+        priv->maps[CHANGELOG_TYPE_ENTRY]    = "E ";
+
+        for (; i < CHANGELOG_MAX_TYPE; i++) {
+                /* start with version 1 */
+                priv->slice.changelog_version[i] = 1;
+        }
+
+        if (!priv->active)
+                return ret;
+
+        /* spawn the notifier thread */
+        ret = changelog_spawn_notifier (this, priv);
+        if (ret)
+                goto out;
+
+        /**
+         * start with a fresh changelog file every time. this is done
+         * in case there was an encoding change. so... things are kept
+         * simple here.
+         */
+        ret = changelog_fill_rollover_data (&cld, _gf_false);
+        if (ret)
+                goto out;
+
+        LOCK (&priv->lock);
+        {
+                ret = changelog_inject_single_event (this, priv, &cld);
+        }
+        UNLOCK (&priv->lock);
+
+        /* ... and finally spawn the helpers threads */
+        ret = changelog_spawn_helper_threads (this, priv);
+
+ out:
+        return ret;
+}
+
+int
+reconfigure (xlator_t *this, dict_t *options)
+{
+        int                     ret            = 0;
+        char                   *tmp            = NULL;
+        changelog_priv_t       *priv           = NULL;
+        gf_boolean_t            active_earlier = _gf_true;
+        gf_boolean_t            active_now     = _gf_true;
+        changelog_time_slice_t *slice          = NULL;
+        changelog_log_data_t    cld            = {0,};
+
+        priv = this->private;
+        if (!priv)
+                goto out;
+
+        ret = -1;
+        active_earlier = priv->active;
+
+        /* first stop the rollover and the fsync thread */
+        changelog_cleanup_helper_threads (this, priv);
+
+        GF_OPTION_RECONF ("changelog-dir", tmp, options, str, out);
+        if (!tmp) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "\"changelog-dir\" option is not set");
+                goto out;
+        }
+
+        GF_FREE (priv->changelog_dir);
+        priv->changelog_dir = gf_strdup (tmp);
+        if (!priv->changelog_dir)
+                goto out;
+
+        ret = mkdir_p (priv->changelog_dir, 0600, _gf_true);
+        if (ret)
+                goto out;
+
+        GF_OPTION_RECONF ("changelog", active_now, options, bool, out);
+
+        /**
+         * changelog_handle_change() handles changes that could possibly
+         * have been submit changes before changelog deactivation.
+         */
+        if (!active_now)
+                priv->active = _gf_false;
+
+        GF_OPTION_RECONF ("op-mode", tmp, options, str, out);
+        changelog_assign_opmode (priv, tmp);
+
+        tmp = NULL;
+
+        GF_OPTION_RECONF ("encoding", tmp, options, str, out);
+        changelog_assign_encoding (priv, tmp);
+
+        GF_OPTION_RECONF ("rollover-time",
+                          priv->rollover_time, options, int32, out);
+        GF_OPTION_RECONF ("fsync-interval",
+                          priv->fsync_interval, options, int32, out);
+
+        if (active_now || active_earlier) {
+                ret = changelog_fill_rollover_data (&cld, !active_now);
+                if (ret)
+                        goto out;
+
+                slice = &priv->slice;
+
+                LOCK (&priv->lock);
+                {
+                        ret = changelog_inject_single_event (this, priv, &cld);
+                        if (!ret && active_now)
+                                SLICE_VERSION_UPDATE (slice);
+                }
+                UNLOCK (&priv->lock);
+
+                if (ret)
+                        goto out;
+
+                if (active_now) {
+                        ret = changelog_spawn_notifier (this, priv);
+                        if (!ret)
+                                ret = changelog_spawn_helper_threads (this,
+                                                                      priv);
+                } else
+                        ret = changelog_cleanup_notifier (this, priv);
+        }
+
+ out:
+        if (ret) {
+                ret = changelog_cleanup_notifier (this, priv);
+        } else {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "changelog reconfigured");
+                if (active_now)
+                        priv->active = _gf_true;
+        }
+
+        return ret;
+}
+
+int32_t
+init (xlator_t *this)
+{
+        int               ret  = -1;
+        char             *tmp  = NULL;
+        changelog_priv_t *priv = NULL;
+
+        GF_VALIDATE_OR_GOTO ("changelog", this, out);
+
+        if (!this->children || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "translator needs a single subvolume");
+                goto out;
+        }
+
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "dangling volume. please check volfile");
+                goto out;
+        }
+
+        priv = GF_CALLOC (1, sizeof (*priv), gf_changelog_mt_priv_t);
+        if (!priv)
+                goto out;
+
+        this->local_pool = mem_pool_new (changelog_local_t, 64);
+        if (!this->local_pool) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local memory pool");
+                goto out;
+        }
+
+        LOCK_INIT (&priv->lock);
+
+        GF_OPTION_INIT ("changelog-brick", tmp, str, out);
+        if (!tmp) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "\"changelog-brick\" option is not set");
+                goto out;
+        }
+
+        priv->changelog_brick = gf_strdup (tmp);
+        if (!priv->changelog_brick)
+                goto out;
+        tmp = NULL;
+
+        GF_OPTION_INIT ("changelog-dir", tmp, str, out);
+        if (!tmp) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "\"changelog-dir\" option is not set");
+                goto out;
+        }
+
+        priv->changelog_dir = gf_strdup (tmp);
+        if (!priv->changelog_dir)
+                goto out;
+        tmp = NULL;
+
+        /**
+         * create the directory even if change-logging would be inactive
+         * so that consumers can _look_ into it (finding nothing...)
+         */
+        ret = mkdir_p (priv->changelog_dir, 0600, _gf_true);
+        if (ret)
+                goto out;
+
+        GF_OPTION_INIT ("changelog", priv->active, bool, out);
+
+        GF_OPTION_INIT ("op-mode", tmp, str, out);
+        changelog_assign_opmode (priv, tmp);
+
+        tmp = NULL;
+
+        GF_OPTION_INIT ("encoding", tmp, str, out);
+        changelog_assign_encoding (priv, tmp);
+
+        GF_OPTION_INIT ("rollover-time", priv->rollover_time, int32, out);
+
+        GF_OPTION_INIT ("fsync-interval", priv->fsync_interval, int32, out);
+
+        changelog_encode_change(priv);
+
+        GF_ASSERT (cb_bootstrap[priv->op_mode].mode == priv->op_mode);
+        priv->cb = &cb_bootstrap[priv->op_mode];
+
+        /* ... now bootstrap the logger */
+        ret = priv->cb->ctor (this, &priv->cd);
+        if (ret)
+                goto out;
+
+        priv->changelog_fd = -1;
+        ret = changelog_init (this, priv);
+        if (ret)
+                goto out;
+
+        gf_log (this->name, GF_LOG_DEBUG, "changelog translator loaded");
+
+ out:
+        if (ret) {
+                if (this->local_pool)
+                        mem_pool_destroy (this->local_pool);
+                if (priv->cb) {
+                        ret = priv->cb->dtor (this, &priv->cd);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "error in cleanup during init()");
+                }
+                GF_FREE (priv->changelog_brick);
+                GF_FREE (priv->changelog_dir);
+                GF_FREE (priv);
+                this->private = NULL;
+        } else
+                this->private = priv;
+
+        return ret;
+}
+
+void
+fini (xlator_t *this)
+{
+        int               ret  = -1;
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+
+        if (priv) {
+                ret = priv->cb->dtor (this, &priv->cd);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "error in fini");
+                mem_pool_destroy (this->local_pool);
+                GF_FREE (priv->changelog_brick);
+                GF_FREE (priv->changelog_dir);
+                GF_FREE (priv);
+        }
+
+        this->private = NULL;
+
+        return;
+}
+
+struct xlator_fops fops = {
+        .mknod        = changelog_mknod,
+        .mkdir        = changelog_mkdir,
+        .create       = changelog_create,
+        .symlink      = changelog_symlink,
+        .writev       = changelog_writev,
+        .truncate     = changelog_truncate,
+        .ftruncate    = changelog_ftruncate,
+        .link         = changelog_link,
+        .rename       = changelog_rename,
+        .unlink       = changelog_unlink,
+        .rmdir        = changelog_rmdir,
+        .setattr      = changelog_setattr,
+        .fsetattr     = changelog_fsetattr,
+        .setxattr     = changelog_setxattr,
+        .fsetxattr    = changelog_fsetxattr,
+        .removexattr  = changelog_removexattr,
+        .fremovexattr = changelog_fremovexattr,
+};
+
+struct xlator_cbks cbks = {
+        .forget = changelog_forget,
+};
+
+struct volume_options options[] = {
+        {.key = {"changelog"},
+         .type = GF_OPTION_TYPE_BOOL,
+         .default_value = "off",
+         .description = "enable/disable change-logging"
+        },
+        {.key = {"changelog-brick"},
+         .type = GF_OPTION_TYPE_PATH,
+         .description = "brick path to generate unique socket file name."
+                       " should be the export directory of the volume strictly."
+        },
+        {.key = {"changelog-dir"},
+         .type = GF_OPTION_TYPE_PATH,
+         .description = "directory for the changelog files"
+        },
+        {.key = {"op-mode"},
+         .type = GF_OPTION_TYPE_STR,
+         .default_value = "realtime",
+         .value = {"realtime"},
+         .description = "operation mode - futuristic operation modes"
+        },
+        {.key = {"encoding"},
+         .type = GF_OPTION_TYPE_STR,
+         .default_value = "ascii",
+         .value = {"binary", "ascii"},
+         .description = "encoding type for changelogs"
+        },
+        {.key = {"rollover-time"},
+         .default_value = "60",
+         .type = GF_OPTION_TYPE_TIME,
+         .description = "time to switch to a new changelog file (in seconds)"
+        },
+        {.key = {"fsync-interval"},
+         .type = GF_OPTION_TYPE_TIME,
+         .default_value = "0",
+         .description = "do not open CHANGELOG file with O_SYNC mode."
+                        " instead perform fsync() at specified intervals"
+        },
+        {.key = {NULL}
+        },
+};
diff --git a/xlators/protocol/legacy/lib/Makefile.am b/xlators/features/compress/Makefile.am
index d471a3f92..a985f42a8 100644
--- a/xlators/protocol/legacy/lib/Makefile.am
+++ b/xlators/features/compress/Makefile.am
@@ -1,3 +1,3 @@
 SUBDIRS = src
 
-CLEANFILES = 
+CLEANFILES =
diff --git a/xlators/features/compress/src/Makefile.am b/xlators/features/compress/src/Makefile.am
new file mode 100644
index 000000000..4a64b52a9
--- /dev/null
+++ b/xlators/features/compress/src/Makefile.am
@@ -0,0 +1,17 @@
+xlator_LTLIBRARIES = cdc.la
+
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+noinst_HEADERS = cdc.h cdc-mem-types.h
+
+cdc_la_LDFLAGS = -module -avoidversion $(LIBZ_LIBS)
+
+cdc_la_SOURCES = cdc.c cdc-helper.c
+cdc_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -D$(GF_HOST_OS) \
+-shared -nostartfiles  $(LIBZ_CFLAGS)
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
diff --git a/xlators/features/compress/src/cdc-helper.c b/xlators/features/compress/src/cdc-helper.c
new file mode 100644
index 000000000..54432ff45
--- /dev/null
+++ b/xlators/features/compress/src/cdc-helper.c
@@ -0,0 +1,547 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+
+#include "cdc.h"
+#include "cdc-mem-types.h"
+
+#ifdef HAVE_LIB_Z
+#include "zlib.h"
+#endif
+
+#ifdef HAVE_LIB_Z
+/* gzip header looks something like this
+ * (RFC 1950)
+ *
+ * +---+---+---+---+---+---+---+---+---+---+
+ * |ID1|ID2|CM |FLG|     MTIME     |XFL|OS |
+ * +---+---+---+---+---+---+---+---+---+---+
+ *
+ * Data is usually sent without this header i.e
+ * Data sent = <compressed-data> + trailer(8)
+ * The trailer contains the checksum.
+ *
+ * gzip_header is added only during debugging.
+ * Refer to the function cdc_dump_iovec_to_disk
+ */
+static const char gzip_header[10] =
+        {
+                '\037', '\213', Z_DEFLATED, 0,
+                0,      0,      0,          0,
+                0,      GF_CDC_OS_ID
+        };
+
+static int32_t
+cdc_next_iovec (xlator_t *this, cdc_info_t *ci)
+{
+        int ret = -1;
+
+        ci->ncount++;
+        /* check for iovec overflow -- should not happen */
+        if (ci->ncount == MAX_IOVEC) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Zlib output buffer overflow"
+                        " ->ncount (%d) | ->MAX_IOVEC (%d)",
+                        ci->ncount, MAX_IOVEC);
+                goto out;
+        }
+
+        ret = 0;
+
+ out:
+        return ret;
+}
+
+static void
+cdc_put_long (unsigned char *string, unsigned long x)
+{
+        string[0] = (unsigned char) (x & 0xff);
+        string[1] = (unsigned char) ((x & 0xff00) >> 8);
+        string[2] = (unsigned char) ((x & 0xff0000) >> 16);
+        string[3] = (unsigned char) ((x & 0xff000000) >> 24);
+}
+
+static unsigned long
+cdc_get_long (unsigned char *buf)
+{
+        return ((unsigned long) buf[0])
+                | (((unsigned long) buf[1]) << 8)
+                | (((unsigned long) buf[2]) << 16)
+                | (((unsigned long) buf[3]) << 24);
+}
+
+static int32_t
+cdc_init_gzip_trailer (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci)
+{
+        int   ret = -1;
+        char *buf = NULL;
+
+        ret = cdc_next_iovec (this, ci);
+        if (ret)
+                goto out;
+
+        buf = CURR_VEC(ci).iov_base =
+                (char *) GF_CALLOC (1, GF_CDC_VALIDATION_SIZE,
+                                    gf_cdc_mt_gzip_trailer_t);
+
+        if (!CURR_VEC(ci).iov_base)
+                goto out;
+
+        CURR_VEC(ci).iov_len = GF_CDC_VALIDATION_SIZE;
+
+        cdc_put_long ((unsigned char *)&buf[0], ci->crc);
+        cdc_put_long ((unsigned char *)&buf[4], ci->stream.total_in);
+
+        ret = 0;
+
+ out:
+        return ret;
+}
+
+static int32_t
+cdc_alloc_iobuf_and_init_vec (xlator_t *this,
+                              cdc_priv_t *priv, cdc_info_t *ci,
+                              int size)
+{
+        int           ret       = -1;
+        int           alloc_len = 0;
+        struct iobuf *iobuf     = NULL;
+
+        ret = cdc_next_iovec (this, ci);
+        if (ret)
+                goto out;
+
+        alloc_len = size ? size : ci->buffer_size;
+
+        iobuf = iobuf_get2 (this->ctx->iobuf_pool, alloc_len);
+        if (!iobuf)
+                goto out;
+
+        ret = iobref_add (ci->iobref, iobuf);
+        if (ret)
+                goto out;
+
+        /* Initialize this iovec */
+        CURR_VEC(ci).iov_base = iobuf->ptr;
+        CURR_VEC(ci).iov_len  = alloc_len;
+
+        ret = 0;
+
+ out:
+        return ret;
+}
+
+static void
+cdc_init_zlib_output_stream (cdc_priv_t *priv, cdc_info_t *ci, int size)
+{
+        ci->stream.next_out  = (unsigned char *) CURR_VEC(ci).iov_base;
+        ci->stream.avail_out = size ? size : ci->buffer_size;
+}
+
+/* This routine is for testing and debugging only.
+ * Data written = header(10) + <compressed-data> + trailer(8)
+ * So each gzip dump file is at least 18 bytes in size.
+ */
+void
+cdc_dump_iovec_to_disk (xlator_t *this, cdc_info_t *ci, const char *file)
+{
+        int    i      = 0;
+        int    fd     = 0;
+        size_t writen = 0;
+        size_t total_writen = 0;
+
+        fd = open (file, O_WRONLY|O_CREAT|O_TRUNC, 0777 );
+        if (fd < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Cannot open file: %s", file);
+                return;
+        }
+
+        writen = write (fd, (char *) gzip_header, 10);
+        total_writen += writen;
+        for (i = 0; i < ci->ncount; i++) {
+                writen = write (fd, (char *) ci->vec[i].iov_base, ci->vec[i].iov_len);
+                total_writen += writen;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                        "dump'd %zu bytes to %s", total_writen, GF_CDC_DEBUG_DUMP_FILE );
+
+        close (fd);
+}
+
+static int32_t
+cdc_flush_libz_buffer (cdc_priv_t *priv, xlator_t *this, cdc_info_t *ci,
+                       int (*libz_func)(z_streamp, int),
+                       int flush)
+{
+        int32_t ret = Z_OK;
+        int done = 0;
+        unsigned int deflate_len = 0;
+
+        for (;;) {
+                deflate_len = ci->buffer_size - ci->stream.avail_out;
+
+                if (deflate_len != 0) {
+                        CURR_VEC(ci).iov_len = deflate_len;
+
+                        ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
+                        if (ret) {
+                                ret = Z_MEM_ERROR;
+                                break;
+                        }
+
+                        /* Re-position Zlib output buffer */
+                        cdc_init_zlib_output_stream (priv, ci, 0);
+                }
+
+                if (done) {
+                        ci->ncount--;
+                        break;
+                }
+
+                ret = libz_func (&ci->stream, flush);
+
+                if (ret == Z_BUF_ERROR) {
+                        ret = Z_OK;
+                        ci->ncount--;
+                        break;
+                }
+
+                done = (ci->stream.avail_out != 0 || ret == Z_STREAM_END);
+
+                if (ret != Z_OK && ret != Z_STREAM_END)
+                        break;
+        }
+
+        return ret;
+}
+
+static int32_t
+do_cdc_compress (struct iovec *vec, xlator_t *this, cdc_priv_t *priv,
+                 cdc_info_t *ci)
+{
+        int ret = -1;
+
+        /* Initialize defalte */
+        ret = deflateInit2 (&ci->stream, priv->cdc_level, Z_DEFLATED,
+                            priv->window_size, priv->mem_level,
+                            Z_DEFAULT_STRATEGY);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "unable to init Zlib (retval: %d)", ret);
+                goto out;
+        }
+
+        ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
+        if (ret)
+                goto out;
+
+        /* setup output buffer */
+        cdc_init_zlib_output_stream (priv, ci, 0);
+
+        /* setup input buffer */
+        ci->stream.next_in  = (unsigned char *) vec->iov_base;
+        ci->stream.avail_in = vec->iov_len;
+
+        ci->crc = crc32 (ci->crc, (const Bytef *) vec->iov_base, vec->iov_len);
+
+        gf_log (this->name, GF_LOG_DEBUG, "crc=%lu len=%d buffer_size=%d",
+                ci->crc, ci->stream.avail_in, ci->buffer_size);
+
+        /* compress !! */
+        while (ci->stream.avail_in != 0) {
+                if (ci->stream.avail_out == 0) {
+
+                        CURR_VEC(ci).iov_len = ci->buffer_size;
+
+                        ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
+                        if (ret)
+                                break;
+
+                        /* Re-position Zlib output buffer */
+                        cdc_init_zlib_output_stream (priv, ci, 0);
+                }
+
+                ret = deflate (&ci->stream, Z_NO_FLUSH);
+                if (ret != Z_OK)
+                        break;
+        }
+
+ out:
+        return ret;
+}
+
+int32_t
+cdc_compress (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci,
+              dict_t **xdata)
+{
+        int ret = -1;
+        int i   = 0;
+
+        ci->iobref = iobref_new ();
+        if (!ci->iobref)
+                goto out;
+
+        if (!*xdata) {
+                *xdata = dict_new ();
+                if (!*xdata) {
+                        gf_log (this->name, GF_LOG_ERROR, "Cannot allocate xdata"
+                                " dict");
+                        goto out;
+                }
+        }
+
+        /* data */
+        for (i = 0; i < ci->count; i++) {
+                ret = do_cdc_compress (&ci->vector[i], this, priv, ci);
+                if (ret != Z_OK)
+                        goto deflate_cleanup_out;
+        }
+
+        /* flush zlib buffer */
+        ret = cdc_flush_libz_buffer (priv, this, ci, deflate, Z_FINISH);
+        if (!(ret == Z_OK || ret == Z_STREAM_END)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Compression Error: ret (%d)", ret);
+                ret = -1;
+                goto deflate_cleanup_out;
+        }
+
+        /* trailer */
+        ret = cdc_init_gzip_trailer (this, priv, ci);
+        if (ret)
+                goto deflate_cleanup_out;
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Compressed %ld to %ld bytes",
+                ci->stream.total_in, ci->stream.total_out);
+
+        ci->nbytes = ci->stream.total_out + GF_CDC_VALIDATION_SIZE;
+
+        /* set deflated canary value for identification */
+        ret = dict_set_int32 (*xdata, GF_CDC_DEFLATE_CANARY_VAL, 1);
+        if (ret) {
+                /* Send uncompressed data if we can't _tell_ the client
+                 * that deflated data is on it's way. So, we just log
+                 * the faliure and continue as usual.
+                 */
+                 gf_log (this->name, GF_LOG_ERROR,
+                 "Data deflated, but could not set canary"
+                 " value in dict for identification");
+        }
+
+        /* This is to be used in testing */
+        if ( priv->debug ) {
+                cdc_dump_iovec_to_disk (this, ci, GF_CDC_DEBUG_DUMP_FILE );
+        }
+
+ deflate_cleanup_out:
+        (void) deflateEnd(&ci->stream);
+
+ out:
+        return ret;
+}
+
+
+/* deflate content is checked by the presence of a canary
+ * value in the dict as the key
+ */
+static int32_t
+cdc_check_content_for_deflate (dict_t *xdata)
+{
+        return dict_get (xdata, GF_CDC_DEFLATE_CANARY_VAL) ? -1 : 0;
+}
+
+static unsigned long
+cdc_extract_crc (char *trailer)
+{
+        return cdc_get_long ((unsigned char *) &trailer[0]);
+}
+
+static unsigned long
+cdc_extract_size (char *trailer)
+{
+        return cdc_get_long ((unsigned char *) &trailer[4]);
+}
+
+static int32_t
+cdc_validate_inflate (cdc_info_t *ci, unsigned long crc,
+                      unsigned long len)
+{
+        return !((crc == ci->crc)
+                 /* inflated length is hidden inside
+                  * Zlib stream struct */
+                 && (len == ci->stream.total_out));
+}
+
+static int32_t
+do_cdc_decompress (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci)
+{
+        int            ret          = -1;
+        int            i            = 0;
+        int            len          = 0;
+        char          *inflte       = NULL;
+        char          *trailer      = NULL;
+        struct iovec   vec          = {0,};
+        unsigned long  computed_crc = 0;
+        unsigned long  computed_len = 0;
+
+        ret = inflateInit2 (&ci->stream, priv->window_size);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Zlib: Unable to initialize inflate");
+                goto out;
+        }
+
+        vec = THIS_VEC(ci, 0);
+
+        trailer = (char *) (((char *) vec.iov_base) + vec.iov_len
+                - GF_CDC_VALIDATION_SIZE);
+
+        /* CRC of uncompressed data */
+        computed_crc = cdc_extract_crc (trailer);
+
+        /* size of uncomrpessed data */
+        computed_len = cdc_extract_size (trailer);
+
+        gf_log (this->name, GF_LOG_DEBUG, "crc=%lu len=%lu buffer_size=%d",
+                computed_crc, computed_len, ci->buffer_size);
+
+        inflte = vec.iov_base ;
+        len = vec.iov_len - GF_CDC_VALIDATION_SIZE;
+
+        /* allocate buffer of the original length of the data */
+        ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
+        if (ret)
+                goto out;
+
+        /* setup output buffer */
+        cdc_init_zlib_output_stream (priv, ci, 0);
+
+        /* setup input buffer */
+        ci->stream.next_in = (unsigned char *) inflte;
+        ci->stream.avail_in = len;
+
+        while (ci->stream.avail_in != 0) {
+                if (ci->stream.avail_out == 0) {
+                        CURR_VEC(ci).iov_len = ci->buffer_size;
+
+                        ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
+                        if (ret)
+                                break;
+
+                        /* Re-position Zlib output buffer */
+                        cdc_init_zlib_output_stream (priv, ci, 0);
+                }
+
+                ret = inflate (&ci->stream, Z_NO_FLUSH);
+                if (ret == Z_STREAM_ERROR)
+                        break;
+        }
+
+        /* flush zlib buffer */
+        ret = cdc_flush_libz_buffer (priv, this, ci, inflate, Z_SYNC_FLUSH);
+        if (!(ret == Z_OK || ret == Z_STREAM_END)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Decompression Error: ret (%d)", ret);
+                ret = -1;
+                goto out;
+        }
+
+        /* compute CRC of the uncompresses data to check for
+         * correctness */
+
+        for (i = 0; i < ci->ncount; i++) {
+                ci->crc = crc32 (ci->crc,
+                                 (const Bytef *) ci->vec[i].iov_base,
+                                 ci->vec[i].iov_len);
+        }
+
+        /* validate inflated data */
+        ret = cdc_validate_inflate (ci, computed_crc, computed_len);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Checksum or length mismatched in inflated data");
+        }
+
+ out:
+        return ret;
+}
+
+int32_t
+cdc_decompress (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci,
+                dict_t *xdata)
+{
+        int32_t ret = -1;
+
+        /* check for deflate content */
+        if (!cdc_check_content_for_deflate (xdata)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Content not deflated, passing through ...");
+                goto passthrough_out;
+        }
+
+        ci->iobref = iobref_new ();
+        if (!ci->iobref)
+                goto passthrough_out;
+
+        /* do we need to do this? can we assume that one iovec
+         * will hold per request data everytime?
+         *
+         * server/client protocol seems to deal with a single
+         * iovec even if op_ret > 1M. So, it looks ok to
+         * assume that a single iovec will contain all the
+         * data (This saves us a lot from finding the trailer
+         * and the data since it could have been split-up onto
+         * two adjacent iovec's.
+         *
+         * But, in case this translator is loaded above quick-read
+         * for some reason, then it's entirely possible that we get
+         * multiple iovec's...
+         *
+         * This case (handled below) is not tested. (by loading the
+         * xlator below quick-read)
+         */
+
+        /* @@ I_HOPE_THIS_IS_NEVER_HIT */
+        if (ci->count > 1) {
+                gf_log (this->name, GF_LOG_WARNING, "unable to handle"
+                        " multiple iovecs (%d in number)", ci->count);
+                goto inflate_cleanup_out;
+                /* TODO: coallate all iovecs in one */
+        }
+
+        ret = do_cdc_decompress (this, priv, ci);
+        if (ret)
+                goto inflate_cleanup_out;
+
+        ci->nbytes = ci->stream.total_out;
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Inflated %ld to %ld bytes",
+                ci->stream.total_in, ci->stream.total_out);
+
+ inflate_cleanup_out:
+        (void) inflateEnd (&ci->stream);
+
+ passthrough_out:
+        return ret;
+}
+
+#endif
diff --git a/xlators/features/compress/src/cdc-mem-types.h b/xlators/features/compress/src/cdc-mem-types.h
new file mode 100644
index 000000000..efa008059
--- /dev/null
+++ b/xlators/features/compress/src/cdc-mem-types.h
@@ -0,0 +1,22 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __CDC_MEM_TYPES_H
+#define __CDC_MEM_TYPES_H
+
+#include "mem-types.h"
+
+enum gf_cdc_mem_types {
+        gf_cdc_mt_priv_t         = gf_common_mt_end + 1,
+        gf_cdc_mt_vec_t          = gf_common_mt_end + 2,
+        gf_cdc_mt_gzip_trailer_t = gf_common_mt_end + 3,
+};
+
+#endif
diff --git a/xlators/features/compress/src/cdc.c b/xlators/features/compress/src/cdc.c
new file mode 100644
index 000000000..eb7d87c56
--- /dev/null
+++ b/xlators/features/compress/src/cdc.c
@@ -0,0 +1,342 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#include <sys/uio.h>
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "logging.h"
+
+#include "cdc.h"
+#include "cdc-mem-types.h"
+
+static void
+cdc_cleanup_iobref (cdc_info_t *ci)
+{
+        assert(ci->iobref != NULL);
+        iobref_clear (ci->iobref);
+}
+
+int32_t
+cdc_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno,
+               struct iovec *vector, int32_t count,
+               struct iatt *stbuf, struct iobref *iobref,
+               dict_t *xdata)
+{
+        int         ret   = -1;
+        cdc_priv_t *priv  = NULL;
+        cdc_info_t  ci    = {0,};
+
+        GF_VALIDATE_OR_GOTO ("cdc", this, default_out);
+        GF_VALIDATE_OR_GOTO (this->name, frame, default_out);
+
+        priv = this->private;
+
+        if (op_ret <= 0)
+                goto default_out;
+
+        if ( (priv->min_file_size != 0)
+             && (op_ret < priv->min_file_size) )
+                goto default_out;
+
+        ci.count       = count;
+        ci.ibytes      = op_ret;
+        ci.vector      = vector;
+        ci.buf         = NULL;
+        ci.iobref      = NULL;
+        ci.ncount      = 0;
+        ci.crc         = 0;
+        ci.buffer_size = GF_CDC_DEF_BUFFERSIZE;
+
+/* A readv compresses on the server side and decompresses on the client side
+ */
+        if (priv->op_mode == GF_CDC_MODE_SERVER) {
+                ret = cdc_compress (this, priv, &ci, &xdata);
+        } else if (priv->op_mode == GF_CDC_MODE_CLIENT) {
+                ret = cdc_decompress (this, priv, &ci, xdata);
+        } else {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Invalid operation mode (%d)", priv->op_mode);
+        }
+
+        if (ret)
+                goto default_out;
+
+        STACK_UNWIND_STRICT (readv, frame, ci.nbytes, op_errno,
+                             ci.vec, ci.ncount, stbuf, iobref,
+                             xdata);
+        cdc_cleanup_iobref (&ci);
+        return 0;
+
+ default_out:
+        STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno,
+                             vector, count, stbuf, iobref, xdata);
+        return 0;
+}
+
+int32_t
+cdc_readv (call_frame_t *frame, xlator_t *this,
+           fd_t *fd, size_t size, off_t offset, uint32_t flags,
+           dict_t *xdata)
+{
+        fop_readv_cbk_t cbk = NULL;
+
+#ifdef HAVE_LIB_Z
+        cbk = cdc_readv_cbk;
+#else
+        cbk = default_readv_cbk;
+#endif
+        STACK_WIND (frame, cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readv,
+                    fd, size, offset, flags, xdata);
+        return 0;
+}
+
+int32_t
+cdc_writev_cbk (call_frame_t *frame,
+                void *cookie,
+                xlator_t *this,
+                int32_t op_ret,
+                int32_t op_errno,
+                struct iatt *prebuf,
+                struct iatt *postbuf, dict_t *xdata)
+{
+
+	STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+	return 0;
+}
+
+int32_t
+cdc_writev (call_frame_t *frame,
+            xlator_t *this,
+            fd_t *fd,
+            struct iovec *vector,
+            int32_t count,
+            off_t offset,
+            uint32_t flags,
+            struct iobref *iobref, dict_t *xdata)
+{
+	int	     ret   = -1;
+	cdc_priv_t  *priv  = NULL;
+	cdc_info_t   ci    = {0,};
+	size_t       isize = 0;
+
+	GF_VALIDATE_OR_GOTO ("cdc", this, default_out);
+	GF_VALIDATE_OR_GOTO (this->name, frame, default_out);
+
+	priv = this->private;
+
+	isize =  iov_length(vector, count);
+
+	if (isize <= 0)
+                goto default_out;
+
+        if ( (priv->min_file_size != 0)
+             && (isize < priv->min_file_size) )
+                goto default_out;
+
+        ci.count       = count;
+        ci.ibytes      = isize;
+        ci.vector      = vector;
+        ci.buf         = NULL;
+        ci.iobref      = NULL;
+        ci.ncount      = 0;
+        ci.crc         = 0;
+        ci.buffer_size = GF_CDC_DEF_BUFFERSIZE;
+
+/* A writev compresses on the client side and decompresses on the server side
+ */
+	if (priv->op_mode == GF_CDC_MODE_CLIENT) {
+		ret = cdc_compress (this, priv, &ci, &xdata);
+	} else if (priv->op_mode == GF_CDC_MODE_SERVER) {
+		ret = cdc_decompress (this, priv, &ci, xdata);
+	} else {
+		gf_log (this->name, GF_LOG_ERROR, "Invalid operation mode (%d) ", priv->op_mode);
+	}
+
+	if (ret)
+		goto default_out;
+
+	STACK_WIND (frame,
+                    cdc_writev_cbk,
+                    FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->writev,
+                    fd, ci.vec, ci.ncount, offset, flags,
+                    iobref, xdata);
+
+        cdc_cleanup_iobref (&ci);
+        return 0;
+
+ default_out:
+	STACK_WIND (frame,
+                    cdc_writev_cbk,
+                    FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->writev,
+                    fd, vector, count, offset, flags,
+                    iobref, xdata);
+	return 0;
+}
+
+int32_t
+init (xlator_t *this)
+{
+        int         ret      = -1;
+        char       *temp_str = NULL;
+        cdc_priv_t *priv     = NULL;
+
+        GF_VALIDATE_OR_GOTO ("cdc", this, err);
+
+        if (!this->children || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Need subvolume == 1");
+                goto err;
+        }
+
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Dangling volume. Check volfile");
+        }
+
+        priv = GF_CALLOC (1, sizeof (*priv), gf_cdc_mt_priv_t);
+        if (!priv) {
+                goto err;
+        }
+
+        /* Check if debug mode is turned on */
+        GF_OPTION_INIT ("debug", priv->debug, bool, err);
+        if( priv->debug ) {
+                gf_log (this->name, GF_LOG_DEBUG, "CDC debug option turned on");
+        }
+
+        /* Set Gzip Window Size */
+        GF_OPTION_INIT ("window-size", priv->window_size, int32, err);
+        if ( (priv->window_size > GF_CDC_MAX_WINDOWSIZE)
+             || (priv->window_size < GF_CDC_DEF_WINDOWSIZE) ) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Invalid gzip window size (%d), using default",
+                        priv->window_size);
+                priv->window_size = GF_CDC_DEF_WINDOWSIZE;
+        }
+
+        /* Set Gzip (De)Compression Level */
+        GF_OPTION_INIT ("compression-level", priv->cdc_level, int32, err);
+        if ( ((priv->cdc_level < 1) || (priv->cdc_level > 9))
+             && (priv->cdc_level != GF_CDC_DEF_COMPRESSION) ) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Invalid gzip (de)compression level (%d),"
+                        " using default", priv->cdc_level);
+                priv->cdc_level = GF_CDC_DEF_COMPRESSION;
+        }
+
+        /* Set Gzip Memory Level */
+        GF_OPTION_INIT ("mem-level", priv->mem_level, int32, err);
+        if ( (priv->mem_level < 1) || (priv->mem_level > 9) ) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Invalid gzip memory level, using the default");
+                priv->mem_level = GF_CDC_DEF_MEMLEVEL;
+        }
+
+        /* Set min file size to enable compression */
+        GF_OPTION_INIT ("min-size", priv->min_file_size, int32, err);
+
+        /* Mode of operation - Server/Client */
+        ret = dict_get_str (this->options, "mode", &temp_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "Operation mode not specified !!");
+                goto err;
+        }
+
+        if (GF_CDC_MODE_IS_CLIENT (temp_str)) {
+                priv->op_mode = GF_CDC_MODE_CLIENT;
+        } else if (GF_CDC_MODE_IS_SERVER (temp_str)) {
+                priv->op_mode = GF_CDC_MODE_SERVER;
+        } else {
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "Bogus operation mode (%s) specified", temp_str);
+                goto err;
+        }
+
+        this->private = priv;
+        gf_log (this->name, GF_LOG_DEBUG, "CDC xlator loaded in (%s) mode",temp_str);
+        return 0;
+
+ err:
+        if (priv)
+                GF_FREE (priv);
+
+        return -1;
+}
+
+void
+fini (xlator_t *this)
+{
+        cdc_priv_t *priv = this->private;
+
+        if (priv)
+                GF_FREE (priv);
+        this->private = NULL;
+        return;
+}
+
+struct xlator_fops fops = {
+        .readv  = cdc_readv,
+        .writev = cdc_writev,
+};
+
+struct xlator_cbks cbks = {
+};
+
+struct volume_options options[] = {
+        { .key  = {"window-size"},
+          .default_value = "-15",
+          .type = GF_OPTION_TYPE_INT,
+          .description = "Size of the zlib history buffer."
+        },
+        { .key  = {"mem-level"},
+          .default_value = "8",
+          .type = GF_OPTION_TYPE_INT,
+          .description = "Memory allocated for internal compression state.\
+                          1 uses minimum memory but is slow and reduces \
+                          compression ratio; memLevel=9 uses maximum memory \
+                          for optimal speed. The default value is 8."
+        },
+        { .key  = {"compression-level"},
+          .default_value = "-1",
+          .type = GF_OPTION_TYPE_INT,
+          .description = "Compression levels \
+                          0 : no compression, 1 : best speed, \
+                          9 : best compression, -1 : default compression "
+        },
+        { .key  = {"min-size"},
+          .default_value = "0",
+          .type = GF_OPTION_TYPE_INT,
+          .description = "Data is compressed only when its size exceeds this."
+        },
+        { .key  = {"mode"},
+          .value = {"server", "client"},
+          .type = GF_OPTION_TYPE_STR,
+          .description = "Set on the basis of where the xlator is loaded."
+        },
+        { .key = {"debug"},
+          .default_value = "false",
+          .type = GF_OPTION_TYPE_BOOL,
+          .description = "This is used in testing. Will dump compressed data \
+                          to disk as a gzip file."
+        },
+        { .key  = {NULL}
+        },
+};
diff --git a/xlators/features/compress/src/cdc.h b/xlators/features/compress/src/cdc.h
new file mode 100644
index 000000000..71f4d2317
--- /dev/null
+++ b/xlators/features/compress/src/cdc.h
@@ -0,0 +1,107 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __CDC_H
+#define __CDC_H
+
+#ifdef HAVE_LIB_Z
+#include "zlib.h"
+#endif
+
+#include "xlator.h"
+
+#ifndef MAX_IOVEC
+#define MAX_IOVEC 16
+#endif
+
+typedef struct cdc_priv {
+        int window_size;
+        int mem_level;
+        int cdc_level;
+        int min_file_size;
+        int op_mode;
+        gf_boolean_t debug;
+        gf_lock_t lock;
+} cdc_priv_t;
+
+typedef struct cdc_info {
+        /* input bits */
+        int            count;
+        int32_t        ibytes;
+        struct iovec  *vector;
+        struct iatt   *buf;
+
+        /* output bits */
+        int            ncount;
+        int            nbytes;
+        int            buffer_size;
+        struct iovec   vec[MAX_IOVEC];
+        struct iobref *iobref;
+
+        /* zlib bits */
+#ifdef HAVE_LIB_Z
+        z_stream      stream;
+#endif
+        unsigned long crc;
+} cdc_info_t;
+
+#define NVEC(ci) (ci->ncount - 1)
+#define CURR_VEC(ci) ci->vec[ci->ncount - 1]
+#define THIS_VEC(ci, i) ci->vector[i]
+
+/* Gzip defaults */
+#define GF_CDC_DEF_WINDOWSIZE  -15 /* default value */
+#define GF_CDC_MAX_WINDOWSIZE  -8  /* max value     */
+
+#ifdef HAVE_LIB_Z
+#define GF_CDC_DEF_COMPRESSION Z_DEFAULT_COMPRESSION
+#else
+#define GF_CDC_DEF_COMPRESSION -1
+#endif
+
+#define GF_CDC_DEF_MEMLEVEL    8
+#define GF_CDC_DEF_BUFFERSIZE  262144 // 256K - default compression buffer size
+
+/* Operation mode
+ * If xlator is loaded on client, readv decompresses and writev compresses
+ * If xlator is loaded on server, readv compresses and writev decompresses
+ */
+#define GF_CDC_MODE_CLIENT   0
+#define GF_CDC_MODE_SERVER   1
+
+/* min size of data to do cmpression
+ * 0 == compress even 1byte
+ */
+#define GF_CDC_MIN_CHUNK_SIZE 0
+
+#define GF_CDC_VALIDATION_SIZE 8
+
+#define GF_CDC_OS_ID 0xFF
+#define GF_CDC_DEFLATE_CANARY_VAL "deflate"
+#define GF_CDC_DEBUG_DUMP_FILE "/tmp/cdcdump.gz"
+
+#define GF_CDC_MODE_IS_CLIENT(m) \
+        (strcmp (m, "client") == 0)
+
+#define GF_CDC_MODE_IS_SERVER(m) \
+        (strcmp (m, "server") == 0)
+
+int32_t
+cdc_compress (xlator_t *this,
+              cdc_priv_t *priv,
+              cdc_info_t *ci,
+              dict_t **xdata);
+int32_t
+cdc_decompress (xlator_t *this,
+                cdc_priv_t *priv,
+                cdc_info_t *ci,
+                dict_t *xdata);
+
+#endif
diff --git a/xlators/features/filter/src/Makefile.am b/xlators/features/filter/src/Makefile.am
index d473b9ea1..d1fda8b0a 100644
--- a/xlators/features/filter/src/Makefile.am
+++ b/xlators/features/filter/src/Makefile.am
@@ -1,15 +1,16 @@
 xlator_LTLIBRARIES = filter.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/testing/features
 
-filter_la_LDFLAGS = -module -avoidversion
+filter_la_LDFLAGS = -module -avoid-version
 
 filter_la_SOURCES = filter.c
 filter_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la 
 
 noinst_HEADERS = filter-mem-types.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/features/filter/src/filter-mem-types.h b/xlators/features/filter/src/filter-mem-types.h
index cca354438..47a17249b 100644
--- a/xlators/features/filter/src/filter-mem-types.h
+++ b/xlators/features/filter/src/filter-mem-types.h
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __FILTER_MEM_TYPES_H__
 #define __FILTER_MEM_TYPES_H__
 
diff --git a/xlators/features/filter/src/filter.c b/xlators/features/filter/src/filter.c
index beede16e8..1d4887b71 100644
--- a/xlators/features/filter/src/filter.c
+++ b/xlators/features/filter/src/filter.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
diff --git a/xlators/performance/stat-prefetch/Makefile.am b/xlators/features/gfid-access/Makefile.am
index af437a64d..af437a64d 100644
--- a/xlators/performance/stat-prefetch/Makefile.am
+++ b/xlators/features/gfid-access/Makefile.am
diff --git a/xlators/features/gfid-access/src/Makefile.am b/xlators/features/gfid-access/src/Makefile.am
new file mode 100644
index 000000000..db53affaa
--- /dev/null
+++ b/xlators/features/gfid-access/src/Makefile.am
@@ -0,0 +1,15 @@
+xlator_LTLIBRARIES = gfid-access.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+gfid_access_la_LDFLAGS = -module -avoid-version
+
+gfid_access_la_SOURCES = gfid-access.c
+gfid_access_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+noinst_HEADERS = gfid-access.h gfid-access-mem-types.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
diff --git a/xlators/features/gfid-access/src/gfid-access-mem-types.h b/xlators/features/gfid-access/src/gfid-access-mem-types.h
new file mode 100644
index 000000000..168d67b43
--- /dev/null
+++ b/xlators/features/gfid-access/src/gfid-access-mem-types.h
@@ -0,0 +1,23 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _GFID_ACCESS_MEM_TYPES_H
+#define _GFID_ACCESS_MEM_TYPES_H
+
+#include "mem-types.h"
+
+enum gf_changelog_mem_types {
+        gf_gfid_access_mt_priv_t = gf_common_mt_end + 1,
+        gf_gfid_access_mt_gfid_t,
+        gf_gfid_access_mt_end
+};
+
+#endif
+
diff --git a/xlators/features/gfid-access/src/gfid-access.c b/xlators/features/gfid-access/src/gfid-access.c
new file mode 100644
index 000000000..da0ba7e50
--- /dev/null
+++ b/xlators/features/gfid-access/src/gfid-access.c
@@ -0,0 +1,1172 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "gfid-access.h"
+#include "inode.h"
+#include "byte-order.h"
+
+
+
+void
+ga_newfile_args_free (ga_newfile_args_t *args)
+{
+        if (!args)
+                goto out;
+
+        GF_FREE (args->bname);
+
+        if (S_ISLNK (args->st_mode) && args->args.symlink.linkpath) {
+                GF_FREE (args->args.symlink.linkpath);
+                args->args.symlink.linkpath = NULL;
+        }
+
+        mem_put (args);
+out:
+        return;
+}
+
+
+void
+ga_heal_args_free (ga_heal_args_t *args)
+{
+        if (!args)
+                goto out;
+
+        GF_FREE (args->bname);
+
+        mem_put (args);
+out:
+        return;
+}
+
+
+ga_newfile_args_t *
+ga_newfile_parse_args (xlator_t *this, data_t *data)
+{
+        ga_newfile_args_t *args     = NULL;
+        ga_private_t      *priv     = NULL;
+        int                len      = 0;
+        int                blob_len = 0;
+        int                min_len  = 0;
+        void              *blob     = NULL;
+
+        priv = this->private;
+
+        blob = data->data;
+        blob_len = data->len;
+
+        min_len = sizeof (args->uid) + sizeof (args->gid) + sizeof (args->gfid)
+                + sizeof (args->st_mode) + 2 + 2;
+        if (blob_len < min_len) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Invalid length: Total length is less "
+                        "than minimum length.");
+                goto err;
+        }
+
+        args = mem_get0 (priv->newfile_args_pool);
+        if (args == NULL)
+                goto err;
+
+        args->uid = ntoh32 (*(uint32_t *)blob);
+        blob += sizeof (uint32_t);
+        blob_len -= sizeof (uint32_t);
+
+        args->gid = ntoh32 (*(uint32_t *)blob);
+        blob += sizeof (uint32_t);
+        blob_len -= sizeof (uint32_t);
+
+        memcpy (args->gfid, blob, sizeof (args->gfid));
+        blob += sizeof (args->gfid);
+        blob_len -= sizeof (args->gfid);
+
+        args->st_mode = ntoh32 (*(uint32_t *)blob);
+        blob += sizeof (uint32_t);
+        blob_len -= sizeof (uint32_t);
+
+        len = strnlen (blob, blob_len);
+        if (len == blob_len)
+        if (len == blob_len) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "gfid: %s. No null byte present.",
+                        args->gfid);
+                goto err;
+        }
+
+        args->bname = GF_CALLOC (1, (len + 1), gf_common_mt_char);
+        if (args->bname == NULL)
+                goto err;
+
+        memcpy (args->bname, blob, (len + 1));
+        blob += (len + 1);
+        blob_len -= (len + 1);
+
+        if (S_ISDIR (args->st_mode)) {
+                if (blob_len < sizeof (uint32_t)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.mkdir.mode = ntoh32 (*(uint32_t *)blob);
+                blob += sizeof (uint32_t);
+                blob_len -= sizeof (uint32_t);
+
+                if (blob_len < sizeof (uint32_t)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.mkdir.umask = ntoh32 (*(uint32_t *)blob);
+                blob += sizeof (uint32_t);
+                blob_len -= sizeof (uint32_t);
+                if (blob_len < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+        } else if (S_ISLNK (args->st_mode)) {
+                len = strnlen (blob, blob_len);
+                if (len == blob_len) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.symlink.linkpath = GF_CALLOC (1, len + 1,
+                                                         gf_common_mt_char);
+                if (args->args.symlink.linkpath == NULL)
+                        goto err;
+
+                memcpy (args->args.symlink.linkpath, blob, (len + 1));
+                blob += (len + 1);
+                blob_len -= (len + 1);
+        } else {
+                if (blob_len < sizeof (uint32_t)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.mknod.mode = ntoh32 (*(uint32_t *)blob);
+                blob += sizeof (uint32_t);
+                blob_len -= sizeof (uint32_t);
+
+                if (blob_len < sizeof (uint32_t)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.mknod.rdev = ntoh32 (*(uint32_t *)blob);
+                blob += sizeof (uint32_t);
+                blob_len -= sizeof (uint32_t);
+
+                if (blob_len < sizeof (uint32_t)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.mknod.umask = ntoh32 (*(uint32_t *)blob);
+                blob += sizeof (uint32_t);
+                blob_len -= sizeof (uint32_t);
+        }
+
+        if (blob_len) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "gfid: %s. Invalid length",
+                        args->gfid);
+                goto err;
+        }
+
+        return args;
+
+err:
+        if (args)
+                ga_newfile_args_free (args);
+
+        return NULL;
+}
+
+ga_heal_args_t *
+ga_heal_parse_args (xlator_t *this, data_t *data)
+{
+        ga_heal_args_t *args     = NULL;
+        ga_private_t   *priv     = NULL;
+        void           *blob     = NULL;
+        int             len      = 0;
+        int             blob_len = 0;
+
+        blob = data->data;
+        blob_len = data->len;
+
+        priv = this->private;
+
+        /* bname should at least contain a character */
+        if (blob_len < (sizeof (args->gfid) + 2))
+                goto err;
+
+        args = mem_get0 (priv->heal_args_pool);
+        if (!args)
+                goto err;
+
+        memcpy (args->gfid, blob, sizeof (args->gfid));
+        blob += sizeof (args->gfid);
+        blob_len -= sizeof (args->gfid);
+
+        len = strnlen (blob, blob_len);
+        if (len == blob_len)
+                goto err;
+
+        args->bname = GF_CALLOC (1, len + 1, gf_common_mt_char);
+        if (!args->bname)
+                goto err;
+
+        memcpy (args->bname, blob, len);
+        blob_len -= (len + 1);
+
+        if (blob_len)
+                goto err;
+
+        return args;
+
+err:
+        if (args)
+                ga_heal_args_free (args);
+
+        return NULL;
+}
+
+static int32_t
+ga_fill_tmp_loc (loc_t *loc, xlator_t *this, char *gfid,
+                 char *bname, dict_t *xdata, loc_t *new_loc)
+{
+        int       ret    = -1;
+        uint64_t  value  = 0;
+        inode_t  *parent = NULL;
+
+        parent = loc->inode;
+        ret = inode_ctx_get (loc->inode, this, &value);
+        if (!ret) {
+                parent = (void *)value;
+        }
+
+        /* parent itself should be looked up */
+        uuid_copy (new_loc->pargfid, parent->gfid);
+        new_loc->parent = inode_ref (parent);
+
+        new_loc->inode = inode_grep (parent->table, parent, bname);
+        if (!new_loc->inode)
+                new_loc->inode = inode_new (parent->table);
+
+        loc_path (new_loc, bname);
+        new_loc->name = basename (new_loc->path);
+
+        /* As GFID would not be set on the entry yet, lets not send entry
+           gfid in the request */
+        /*uuid_copy (new_loc->gfid, (const unsigned char *)gfid); */
+
+        ret = dict_set_static_bin (xdata, "gfid-req", gfid, 16);
+        if (ret < 0)
+                goto out;
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+
+
+static gf_boolean_t
+__is_gfid_access_dir (uuid_t gfid)
+{
+        uuid_t  aux_gfid;
+
+        memset (aux_gfid, 0, 16);
+        aux_gfid[15] = GF_AUX_GFID;
+
+        if (uuid_compare (gfid, aux_gfid) == 0)
+                return _gf_true;
+
+        return _gf_false;
+}
+
+int32_t
+ga_forget (xlator_t *this, inode_t *inode)
+{
+        int       ret = -1;
+        uint64_t  value = 0;
+        inode_t  *tmp_inode = NULL;
+
+        ret = inode_ctx_del (inode, this, &value);
+        if (ret)
+                goto out;
+
+        tmp_inode = (void *)value;
+        inode_unref (tmp_inode);
+
+out:
+        return 0;
+}
+
+
+static int
+ga_heal_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+             int32_t op_ret, int32_t op_errno,
+             inode_t *inode, struct iatt *stat, dict_t *dict,
+             struct iatt *postparent)
+{
+        call_frame_t *orig_frame = NULL;
+
+        orig_frame = frame->local;
+        frame->local = NULL;
+
+        /* don't worry about inode linking and other stuff. They'll happen on
+         * the next lookup.
+         */
+        STACK_DESTROY (frame->root);
+
+        STACK_UNWIND_STRICT (setxattr, orig_frame, op_ret, op_errno, dict);
+
+        return 0;
+}
+
+static int
+ga_newentry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno,
+                 inode_t *inode, struct iatt *buf,
+                 struct iatt *preparent, struct iatt *postparent,
+                 dict_t *xdata)
+{
+        call_frame_t *orig_frame = NULL;
+
+        orig_frame = frame->local;
+        frame->local = NULL;
+
+        /* don't worry about inode linking and other stuff. They'll happen on
+         * the next lookup.
+         */
+        STACK_DESTROY (frame->root);
+
+        STACK_UNWIND_STRICT (setxattr, orig_frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+int32_t
+ga_new_entry (call_frame_t *frame, xlator_t *this, loc_t *loc, data_t *data,
+              dict_t *xdata)
+{
+        int                ret       = -1;
+        ga_newfile_args_t *args      = NULL;
+        loc_t              tmp_loc   = {0,};
+        call_frame_t      *new_frame = NULL;
+        mode_t             mode      = 0;
+
+        args = ga_newfile_parse_args (this, data);
+        if (!args)
+                goto out;
+
+        if (!xdata)
+                xdata = dict_new ();
+
+        ret = ga_fill_tmp_loc (loc, this, args->gfid,
+                               args->bname, xdata, &tmp_loc);
+        if (ret)
+                goto out;
+
+        new_frame = copy_frame (frame);
+        if (!new_frame)
+                goto out;
+        new_frame->local = (void *)frame;
+
+        new_frame->root->uid = args->uid;
+        new_frame->root->gid = args->gid;
+
+        if (S_ISDIR (args->st_mode)) {
+                STACK_WIND (new_frame, ga_newentry_cbk,
+                            FIRST_CHILD(this), FIRST_CHILD(this)->fops->mkdir,
+                            &tmp_loc, args->args.mkdir.mode,
+                            args->args.mkdir.umask, xdata);
+        } else if (S_ISLNK (args->st_mode)) {
+                STACK_WIND (new_frame, ga_newentry_cbk,
+                            FIRST_CHILD(this), FIRST_CHILD(this)->fops->symlink,
+                            args->args.symlink.linkpath,
+                            &tmp_loc, 0, xdata);
+        } else {
+                /* use 07777 (4 7s) for considering the Sticky bits etc) */
+                mode = (S_IFMT & args->st_mode) |
+                        (07777 | args->args.mknod.mode);;
+
+                STACK_WIND (new_frame, ga_newentry_cbk,
+                            FIRST_CHILD(this), FIRST_CHILD(this)->fops->mknod,
+                            &tmp_loc, mode,
+                            args->args.mknod.rdev, args->args.mknod.umask,
+                            xdata);
+        }
+
+        ret = 0;
+out:
+        ga_newfile_args_free (args);
+
+        return ret;
+}
+
+int32_t
+ga_heal_entry (call_frame_t *frame, xlator_t *this, loc_t *loc, data_t *data,
+               dict_t *xdata)
+{
+        int             ret       = -1;
+        ga_heal_args_t *args      = NULL;
+        loc_t           tmp_loc   = {0,};
+        call_frame_t   *new_frame = NULL;
+
+        args = ga_heal_parse_args (this, data);
+        if (!args)
+                goto out;
+
+        if (!xdata)
+                xdata = dict_new ();
+
+        ret = ga_fill_tmp_loc (loc, this, args->gfid, args->bname,
+                               xdata, &tmp_loc);
+        if (ret)
+                goto out;
+
+        new_frame = copy_frame (frame);
+        if (!new_frame)
+                goto out;
+        new_frame->local = (void *)frame;
+
+        STACK_WIND (new_frame, ga_heal_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD(this)->fops->lookup,
+                    &tmp_loc, xdata);
+
+        ret = 0;
+out:
+        if (args)
+                ga_heal_args_free (args);
+
+        return ret;
+}
+
+int32_t
+ga_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno,
+                 dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+ga_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+             int32_t flags, dict_t *xdata)
+{
+
+        data_t  *data     = NULL;
+        int      op_errno = ENOMEM;
+        int      ret      = 0;
+        inode_t *unref    = NULL;
+
+        if ((loc->name && !strcmp (GF_GFID_DIR, loc->name)) &&
+            ((loc->parent &&
+              __is_root_gfid (loc->parent->gfid)) ||
+             __is_root_gfid (loc->pargfid))) {
+                op_errno = EPERM;
+                goto err;
+        }
+
+        data = dict_get (dict, GF_FUSE_AUX_GFID_NEWFILE);
+        if (data) {
+                ret = ga_new_entry (frame, this, loc, data, xdata);
+                if (ret)
+                        goto err;
+                return 0;
+        }
+
+        data = dict_get (dict, GF_FUSE_AUX_GFID_HEAL);
+        if (data) {
+                ret = ga_heal_entry (frame, this, loc, data, xdata);
+                if (ret)
+                        goto err;
+                return 0;
+        }
+
+        //If the inode is a virtual inode change the inode otherwise perform
+        //the operation on same inode
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, ga_setxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr, loc, dict, flags,
+                    xdata);
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (setxattr, frame, -1, op_errno, xdata);
+        return 0;
+}
+
+
+int32_t
+ga_virtual_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, inode_t *inode,
+                       struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+        int       j           = 0;
+        int       i           = 0;
+        int       ret         = 0;
+        uint64_t  temp_ino    = 0;
+        inode_t  *cbk_inode   = NULL;
+        inode_t  *true_inode  = NULL;
+        uuid_t    random_gfid = {0,};
+
+        if (frame->local)
+                cbk_inode = frame->local;
+        else
+                cbk_inode = inode;
+
+        frame->local = NULL;
+        if (op_ret)
+                goto unwind;
+
+        if (!IA_ISDIR (buf->ia_type))
+                goto unwind;
+
+        /* need to send back a different inode for linking in itable */
+        if (cbk_inode == inode) {
+                /* check if the inode is in the 'itable' or
+                   if its just previously discover()'d inode */
+                true_inode = inode_find (inode->table, buf->ia_gfid);
+                if (!true_inode) {
+                        cbk_inode = inode_new (inode->table);
+
+                        if (!cbk_inode) {
+                                op_ret = -1;
+                                op_errno = ENOMEM;
+                                goto unwind;
+                        }
+                        /* the inode is not present in itable, ie, the actual
+                           path is not yet looked up. Use the current inode
+                           itself for now */
+                        inode_ref (inode);
+                } else {
+                        /* 'inode_ref()' has been done in inode_find() */
+                        inode = true_inode;
+                }
+
+                ret = inode_ctx_put (cbk_inode, this, (uint64_t)inode);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "failed to set the inode ctx with"
+                                "the actual inode");
+                        if (inode)
+                                inode_unref (inode);
+                }
+                inode = NULL;
+        }
+
+        if (!uuid_is_null (cbk_inode->gfid)) {
+                /* if the previous linked inode is used, use the
+                   same gfid */
+                uuid_copy (random_gfid, cbk_inode->gfid);
+        } else {
+                /* replace the buf->ia_gfid to a random gfid
+                   for directory, for files, what we received is fine */
+                uuid_generate (random_gfid);
+        }
+
+        uuid_copy (buf->ia_gfid, random_gfid);
+
+        for (i = 15; i > (15 - 8); i--) {
+                temp_ino += (uint64_t)(buf->ia_gfid[i]) << j;
+                j += 8;
+        }
+        buf->ia_ino = temp_ino;
+
+unwind:
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, cbk_inode, buf,
+                             xdata, postparent);
+
+        return 0;
+}
+
+int32_t
+ga_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, inode_t *inode,
+               struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+        ga_private_t *priv = NULL;
+
+        /* if the entry in question is not 'root',
+           then follow the normal path */
+        if (op_ret || !__is_root_gfid(buf->ia_gfid))
+                goto unwind;
+
+        priv = this->private;
+
+        /* do we need to copy root stbuf everytime? */
+        /* mostly yes, as we want to have the 'stat' info show latest
+           in every _cbk() */
+
+        /* keep the reference for root stat buf */
+        priv->root_stbuf = *buf;
+        priv->gfiddir_stbuf = priv->root_stbuf;
+        priv->gfiddir_stbuf.ia_gfid[15] = GF_AUX_GFID;
+        priv->gfiddir_stbuf.ia_ino = GF_AUX_GFID;
+
+unwind:
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
+                             xdata, postparent);
+        return 0;
+}
+
+int32_t
+ga_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+        ga_private_t *priv     = NULL;
+        int           ret      = -1;
+        uuid_t        tmp_gfid = {0,};
+        loc_t         tmp_loc  = {0,};
+        uint64_t      value    = 0;
+        inode_t      *inode    = NULL;
+        inode_t      *true_inode    = NULL;
+        int32_t       op_errno = ENOENT;
+
+        /* if its discover(), no need for any action here */
+        if (!loc->name)
+                goto wind;
+
+        /* if its revalidate, and inode is not of type directory,
+           proceed with 'wind' */
+        if (loc->inode && loc->inode->ia_type &&
+            !IA_ISDIR (loc->inode->ia_type))
+                goto wind;
+
+        priv = this->private;
+
+        /* need to check if the lookup is on virtual dir */
+        if ((loc->name && !strcmp (GF_GFID_DIR, loc->name)) &&
+            ((loc->parent && __is_root_gfid (loc->parent->gfid)) ||
+             __is_root_gfid (loc->pargfid))) {
+                /* this means, the query is on '/.gfid', return the fake stat,
+                   and say success */
+
+                STACK_UNWIND_STRICT (lookup, frame, 0, 0, loc->inode,
+                                     &priv->gfiddir_stbuf, xdata,
+                                     &priv->root_stbuf);
+                return 0;
+        }
+
+        /* now, check if the lookup() is on an existing entry,
+           but on gfid-path */
+        if (!((loc->parent && __is_gfid_access_dir (loc->parent->gfid)) ||
+              __is_gfid_access_dir (loc->pargfid)))
+                goto wind;
+
+        /* make sure the 'basename' is actually a 'canonical-gfid',
+           otherwise, return error */
+        ret = uuid_parse (loc->name, tmp_gfid);
+        if (ret)
+                goto err;
+
+        /* if its fresh lookup, go ahead and send it down, if not,
+           for directory, we need indirection to actual dir inode */
+        if (!(loc->inode && loc->inode->ia_type))
+                goto discover;
+
+        /* revalidate on directory */
+        ret = inode_ctx_get (loc->inode, this, &value);
+        if (ret)
+                goto err;
+
+        inode = (void *)value;
+
+        /* valid inode, already looked up, work on that */
+        if (inode->ia_type)
+                goto discover;
+
+        /* check if the inode is in the 'itable' or
+           if its just previously discover()'d inode */
+        true_inode = inode_find (loc->inode->table, tmp_gfid);
+        if (true_inode) {
+                /* time do another lookup and update the context
+                   with proper inode */
+                op_errno = ESTALE;
+                goto err;
+        }
+
+discover:
+        /* for the virtual entries, we don't need to send 'gfid-req' key, as
+           for these entries, we don't want to 'set' a new gfid */
+        if (xdata)
+                dict_del (xdata, "gfid-req");
+
+        uuid_copy (tmp_loc.gfid, tmp_gfid);
+
+        /* if revalidate, then we need to have the proper reference */
+        if (inode) {
+                tmp_loc.inode = inode_ref (inode);
+                frame->local = loc->inode;
+        } else {
+                tmp_loc.inode = inode_ref (loc->inode);
+        }
+
+        STACK_WIND (frame, ga_virtual_lookup_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, &tmp_loc, xdata);
+
+        inode_unref (tmp_loc.inode);
+
+        return 0;
+
+wind:
+        /* used for all the normal lookup path */
+        STACK_WIND (frame, ga_lookup_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, loc, xdata);
+
+        return 0;
+
+err:
+        STACK_UNWIND_STRICT (lookup, frame, -1, op_errno, loc->inode,
+                             &priv->gfiddir_stbuf, xdata,
+                             &priv->root_stbuf);
+        return 0;
+}
+
+int
+ga_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+          mode_t umask, dict_t *xdata)
+{
+        int op_errno = 0;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        STACK_WIND (frame, default_mkdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mkdir, loc, mode, umask,
+                    xdata);
+
+        return 0;
+
+err:
+        STACK_UNWIND_STRICT (mkdir, frame, -1, op_errno, loc->inode,
+                             NULL, NULL, NULL, xdata);
+        return 0;
+}
+
+
+int
+ga_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+           mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        int op_errno = 0;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        STACK_WIND (frame, default_create_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->create,
+                    loc, flags, mode, umask, fd, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (create, frame, -1, op_errno, NULL,
+                             NULL, NULL, NULL, NULL, xdata);
+
+        return 0;
+
+}
+
+int
+ga_symlink (call_frame_t *frame, xlator_t *this, const char *linkname,
+            loc_t *loc, mode_t umask, dict_t *xdata)
+{
+        int op_errno = 0;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        STACK_WIND (frame, default_symlink_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->symlink,
+                    linkname, loc, umask, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (symlink, frame, -1, op_errno, NULL,
+                             NULL, NULL, NULL, xdata);
+
+        return 0;
+}
+
+int
+ga_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+          dev_t rdev, mode_t umask, dict_t *xdata)
+{
+        int op_errno = 0;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        STACK_WIND (frame, default_mknod_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mknod, loc, mode, rdev,
+                    umask, xdata);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (mknod, frame, -1, op_errno, NULL,
+                             NULL, NULL, NULL, xdata);
+
+        return 0;
+}
+
+int
+ga_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
+          dict_t *xdata)
+{
+        int op_errno = 0;
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_rmdir_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->rmdir,
+                    loc, flag, xdata);
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (rmdir, frame, -1, op_errno, NULL,
+                             NULL, xdata);
+
+        return 0;
+}
+
+int
+ga_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t xflag,
+           dict_t *xdata)
+{
+        int op_errno = 0;
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_unlink_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->unlink,
+                    loc, xflag, xdata);
+
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL,
+                             NULL, xdata);
+
+        return 0;
+}
+
+int
+ga_rename (call_frame_t *frame, xlator_t *this,
+           loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+        int op_errno = 0;
+        inode_t *oldloc_unref = NULL;
+        inode_t *newloc_unref = NULL;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (oldloc, op_errno, err);
+        GFID_ACCESS_ENTRY_OP_CHECK (newloc, op_errno, err);
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, oldloc, oldloc_unref,
+                                         handle_newloc);
+
+handle_newloc:
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, newloc, newloc_unref, wind);
+
+wind:
+        STACK_WIND (frame, default_rename_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->rename,
+                    oldloc, newloc, xdata);
+
+        if (oldloc_unref)
+                inode_unref (oldloc_unref);
+
+        if (newloc_unref)
+                inode_unref (newloc_unref);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL,
+                             NULL, NULL, NULL, NULL, xdata);
+
+        return 0;
+}
+
+
+int
+ga_link (call_frame_t *frame, xlator_t *this,
+         loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+        int op_errno = 0;
+        inode_t *oldloc_unref = NULL;
+        inode_t *newloc_unref = NULL;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (oldloc, op_errno, err);
+        GFID_ACCESS_ENTRY_OP_CHECK (newloc, op_errno, err);
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, oldloc, oldloc_unref,
+                                         handle_newloc);
+
+handle_newloc:
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, newloc, newloc_unref, wind);
+
+wind:
+        STACK_WIND (frame, default_link_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->link,
+                    oldloc, newloc, xdata);
+
+        if (oldloc_unref)
+                inode_unref (oldloc_unref);
+
+        if (newloc_unref)
+                inode_unref (newloc_unref);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL,
+                             NULL, NULL, NULL, xdata);
+
+        return 0;
+}
+
+int32_t
+ga_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            fd_t *fd, dict_t *xdata)
+{
+        int op_errno = 0;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        /* also check if the loc->inode itself is virtual
+           inode, if yes, return with failure, mainly because we
+           can't handle all the readdirp and other things on it. */
+        if (inode_ctx_get (loc->inode, this, NULL) == 0) {
+                op_errno = ENOTSUP;
+                goto err;
+        }
+
+        STACK_WIND (frame, default_opendir_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->opendir,
+                    loc, fd, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (opendir, frame, -1, op_errno, NULL, xdata);
+
+        return 0;
+}
+
+int32_t
+ga_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             const char *name, dict_t *xdata)
+{
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+}
+
+int32_t
+ga_stat (call_frame_t *frame, xlator_t *this, loc_t *loc,
+         dict_t *xdata)
+{
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_stat_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->stat, loc, xdata);
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+}
+
+int32_t
+ga_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            struct iatt *stbuf, int32_t valid,
+            dict_t *xdata)
+{
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_setattr_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->setattr, loc, stbuf, valid,
+                    xdata);
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+}
+
+int32_t
+ga_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                const char *name, dict_t *xdata)
+{
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_removexattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->removexattr, loc, name,
+                    xdata);
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+}
+
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_gfid_access_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_WARNING, "Memory accounting"
+                        " init failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+int32_t
+init (xlator_t *this)
+{
+        ga_private_t *priv = NULL;
+        int ret = -1;
+
+        if (!this->children || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "not configured with exactly one child. exiting");
+                goto out;
+        }
+
+        /* This can be the top of graph in certain cases */
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "dangling volume. check volfile ");
+        }
+
+        /* TODO: define a mem-type structure */
+        priv = GF_CALLOC (1, sizeof (*priv), gf_gfid_access_mt_priv_t);
+        if (!priv)
+                goto out;
+
+        priv->newfile_args_pool = mem_pool_new (ga_newfile_args_t, 512);
+        if (!priv->newfile_args_pool)
+                goto out;
+
+        priv->heal_args_pool = mem_pool_new (ga_heal_args_t, 512);
+        if (!priv->heal_args_pool)
+                goto out;
+
+        this->private = priv;
+
+        ret = 0;
+out:
+        if (ret && priv) {
+                if (priv->newfile_args_pool)
+                        mem_pool_destroy (priv->newfile_args_pool);
+                GF_FREE (priv);
+        }
+
+        return ret;
+}
+
+void
+fini (xlator_t *this)
+{
+        ga_private_t *priv = NULL;
+        priv = this->private;
+        this->private = NULL;
+
+        if (priv) {
+                if (priv->newfile_args_pool)
+                        mem_pool_destroy (priv->newfile_args_pool);
+                if (priv->heal_args_pool)
+                        mem_pool_destroy (priv->heal_args_pool);
+                GF_FREE (priv);
+        }
+
+        return;
+}
+
+
+struct xlator_fops fops = {
+        .lookup = ga_lookup,
+
+        /* entry fops */
+        .mkdir   = ga_mkdir,
+        .mknod   = ga_mknod,
+        .create  = ga_create,
+        .symlink = ga_symlink,
+        .link    = ga_link,
+        .unlink  = ga_unlink,
+        .rmdir   = ga_rmdir,
+        .rename  = ga_rename,
+
+        /* handle any other directory operations here */
+        .opendir  = ga_opendir,
+        .stat     = ga_stat,
+        .setattr  = ga_setattr,
+        .getxattr = ga_getxattr,
+        .removexattr = ga_removexattr,
+
+        /* special fop to handle more entry creations */
+        .setxattr = ga_setxattr,
+};
+
+struct xlator_cbks cbks = {
+        .forget = ga_forget,
+};
+
+struct volume_options options[] = {
+        /* This translator doesn't take any options, or provide any options */
+        { .key  = {NULL} },
+};
diff --git a/xlators/features/gfid-access/src/gfid-access.h b/xlators/features/gfid-access/src/gfid-access.h
new file mode 100644
index 000000000..e13c9b724
--- /dev/null
+++ b/xlators/features/gfid-access/src/gfid-access.h
@@ -0,0 +1,128 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef __GFID_ACCESS_H__
+#define __GFID_ACCESS_H__
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "defaults.h"
+#include "gfid-access-mem-types.h"
+
+#define UUID_CANONICAL_FORM_LEN 36
+
+#define GF_FUSE_AUX_GFID_NEWFILE "glusterfs.gfid.newfile"
+#define GF_FUSE_AUX_GFID_HEAL    "glusterfs.gfid.heal"
+
+#define GF_GFID_KEY "GLUSTERFS_GFID"
+#define GF_GFID_DIR ".gfid"
+#define GF_AUX_GFID 0xd
+
+#define GFID_ACCESS_GET_VALID_DIR_INODE(x,l,unref,lbl) do {             \
+                int       ret       = 0;                                \
+                uint64_t  value     = 0;                                \
+                inode_t  *tmp_inode = NULL;                             \
+                                                                        \
+                /* if its an entry operation, on the virtual */         \
+                /* directory inode as parent, we need to handle */      \
+                /* it properly */                                       \
+                if (l->parent) {                                        \
+                        ret = inode_ctx_get (l->parent, x, &value);     \
+                        if (ret)                                        \
+                                goto lbl;                               \
+                        tmp_inode = (inode_t *)value;                   \
+                        unref = inode_ref (tmp_inode);                  \
+                        l->parent = tmp_inode;                          \
+                        /* if parent is virtual, no need to handle */   \
+                        /* loc->inode */                                \
+                        break;                                          \
+                }                                                       \
+                                                                        \
+                /* if its an inode operation, on the virtual */         \
+                /* directory inode itself, we need to handle */         \
+                /* it properly */                                       \
+                if (l->inode) {                                         \
+                        ret = inode_ctx_get (l->inode, x, &value);      \
+                        if (ret)                                        \
+                                goto lbl;                               \
+                        tmp_inode = (inode_t *)value;                   \
+                        unref = inode_ref (tmp_inode);                  \
+                        l->inode = tmp_inode;                           \
+                }                                                       \
+                                                                        \
+        } while (0)
+
+#define GFID_ACCESS_ENTRY_OP_CHECK(loc,err,lbl)    do {                 \
+                /* need to check if the lookup is on virtual dir */     \
+                if ((loc->name && !strcmp (GF_GFID_DIR, loc->name)) &&  \
+                    ((loc->parent &&                                    \
+                      __is_root_gfid (loc->parent->gfid)) ||            \
+                      __is_root_gfid (loc->pargfid))) {                 \
+                        err = EEXIST;                                   \
+                        goto lbl;                                       \
+                }                                                       \
+                                                                        \
+                /* now, check if the lookup() is on an existing */      \
+                /* entry, but on gfid-path */                           \
+                if ((loc->parent &&                                     \
+                     __is_gfid_access_dir (loc->parent->gfid)) ||       \
+                    __is_gfid_access_dir (loc->pargfid)) {              \
+                        err = EPERM;                                    \
+                        goto lbl;                                       \
+                }                                                       \
+        } while (0)
+
+
+typedef struct {
+        unsigned int  uid;
+        unsigned int  gid;
+        char          gfid[UUID_CANONICAL_FORM_LEN + 1];
+        unsigned int  st_mode;
+        char         *bname;
+
+        union {
+                struct _symlink_in {
+                        char     *linkpath;
+                } __attribute__ ((__packed__)) symlink;
+
+                struct _mknod_in {
+                        unsigned int mode;
+                        unsigned int rdev;
+                        unsigned int umask;
+                } __attribute__ ((__packed__)) mknod;
+
+                struct _mkdir_in {
+                        unsigned int mode;
+                        unsigned int umask;
+                } __attribute__ ((__packed__)) mkdir;
+        } __attribute__ ((__packed__)) args;
+} __attribute__((__packed__)) ga_newfile_args_t;
+
+typedef struct {
+        char      gfid[UUID_CANONICAL_FORM_LEN + 1];
+        char     *bname; /* a null terminated basename */
+} __attribute__((__packed__)) ga_heal_args_t;
+
+struct ga_private {
+        /* root inode's stbuf */
+        struct iatt root_stbuf;
+        struct iatt gfiddir_stbuf;
+        struct mem_pool *newfile_args_pool;
+        struct mem_pool *heal_args_pool;
+};
+typedef struct ga_private ga_private_t;
+
+#endif /* __GFID_ACCESS_H__ */
diff --git a/xlators/protocol/legacy/server/Makefile.am b/xlators/features/glupy/Makefile.am
index d471a3f92..a985f42a8 100644
--- a/xlators/protocol/legacy/server/Makefile.am
+++ b/xlators/features/glupy/Makefile.am
@@ -1,3 +1,3 @@
 SUBDIRS = src
 
-CLEANFILES = 
+CLEANFILES =
diff --git a/xlators/features/glupy/doc/README.md b/xlators/features/glupy/doc/README.md
new file mode 100644
index 000000000..2d7b30ef6
--- /dev/null
+++ b/xlators/features/glupy/doc/README.md
@@ -0,0 +1,44 @@
+This is just the very start for a GlusterFS[1] meta-translator that will
+allow translator code to be written in Python.  It's based on the standard
+Python embedding (not extending) techniques, plus a dash of the ctypes module.
+The interface is a pretty minimal adaptation of the dispatches and callbacks
+from the C API[2] to Python, as follows:
+
+* Dispatch functions and callbacks must be defined on an "xlator" class
+  derived from gluster.Translator so that they'll be auto-registered with
+  the C translator during initialization.
+
+* For each dispatch or callback function you want to intercept, you define a
+  Python function using the xxx\_fop\_t or xxx\_cbk\_t decorator.
+
+* The arguments for each operation are different, so you'll need to refer to
+  the C API.  GlusterFS-specific types are used (though only loc\_t is fully
+  defined so far) and type correctness is enforced by ctypes.
+
+* If you do intercept a dispatch function, it is your responsibility to call
+  xxx\_wind (like STACK\_WIND in the C API but operation-specific) to pass
+  the request to the next translator.  If you do not intercept a function, it
+  will default the same way as for C (pass through to the same operation with
+  the same arguments on the first child translator).
+
+* If you intercept a callback function, it is your responsibility to call
+  xxx\_unwind (like STACK\_UNWIND\_STRICT in the C API) to pass the request back
+  to the caller.
+
+So far only the lookup and create operations are handled this way, to support
+the "negative lookup" example.  Now that the basic infrastructure is in place,
+adding more functions should be very quick, though with that much boilerplate I
+might pause to write a code generator.  I also plan to add structure
+definitions and interfaces for some of the utility functions in libglusterfs
+(especially those having to do with inode and fd context) in the fairly near
+future.  Note that you can also use ctypes to get at anything not explicitly
+exposed to Python already.
+
+_If you're coming here because of the Linux Journal article, please note that
+the code has evolved since that was written. The version that matches the
+article is here:_
+
+https://github.com/jdarcy/glupy/tree/4bbae91ba459ea46ef32f2966562492e4ca9187a
+
+[1] http://www.gluster.org
+[2] http://hekafs.org/dist/xlator_api_2.html
diff --git a/xlators/features/glupy/doc/TESTING b/xlators/features/glupy/doc/TESTING
new file mode 100644
index 000000000..e05f17f49
--- /dev/null
+++ b/xlators/features/glupy/doc/TESTING
@@ -0,0 +1,9 @@
+Loading a translator written in Python using the glupy meta translator
+-------------------------------------------------------------------------------
+'test.vol' is a simple volfile with the debug-trace Python translator on top
+of a brick. The volfile can be mounted using the following command.
+
+$ glusterfs --debug -f test.vol /path/to/mntpt
+
+If then file operations are performed on the newly mounted file system, log
+output would be printed by the Python translator on the standard output.
diff --git a/xlators/features/glupy/doc/test.vol b/xlators/features/glupy/doc/test.vol
new file mode 100644
index 000000000..0751a488c
--- /dev/null
+++ b/xlators/features/glupy/doc/test.vol
@@ -0,0 +1,10 @@
+volume vol-posix
+    type storage/posix
+    option directory /path/to/brick
+end-volume
+
+volume vol-glupy
+    type features/glupy
+    option module-name debug-trace
+    subvolumes vol-posix
+end-volume
diff --git a/xlators/features/glupy/src/Makefile.am b/xlators/features/glupy/src/Makefile.am
new file mode 100644
index 000000000..960862839
--- /dev/null
+++ b/xlators/features/glupy/src/Makefile.am
@@ -0,0 +1,20 @@
+xlator_LTLIBRARIES = glupy.la
+
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+glupydir = $(xlatordir)/glupy
+
+glupy_PYTHON = gluster.py negative.py helloworld.py debug-trace.py
+
+glupy_la_LDFLAGS = -module -avoid-version -shared -nostartfiles
+glupy_la_SOURCES = glupy.c
+glupy_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \
+	 -lpthread -l$(BUILD_PYTHON_LIB)
+
+noinst_HEADERS = glupy.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src -isystem $(BUILD_PYTHON_INC)
+
+AM_CFLAGS = -Wall -fno-strict-aliasing -DGLUSTER_PYTHON_PATH=\"$(glupydir)\" $(GF_CFLAGS)
+
+CLEANFILES =
diff --git a/xlators/features/glupy/src/debug-trace.py b/xlators/features/glupy/src/debug-trace.py
new file mode 100644
index 000000000..53e76546b
--- /dev/null
+++ b/xlators/features/glupy/src/debug-trace.py
@@ -0,0 +1,774 @@
+import sys
+import stat
+from uuid import UUID
+from time import strftime, localtime
+from gluster import *
+# This translator was written primarily to test the fop entry point definitions
+# and structure definitions in 'gluster.py'.
+# It is similar to the debug-trace translator, one of the already available
+# translator types written in C, that logs the arguments passed to the fops and
+# their corresponding cbk functions.
+
+dl.get_id.restype = c_long
+dl.get_id.argtypes = [ POINTER(call_frame_t) ]
+
+dl.get_rootunique.restype = c_uint64
+dl.get_rootunique.argtypes = [ POINTER(call_frame_t) ]
+
+def uuid2str (gfid):
+        return str(UUID(''.join(map("{0:02x}".format, gfid))))
+
+
+def st_mode_from_ia (prot, filetype):
+        st_mode = 0
+        type_bit = 0
+        prot_bit = 0
+
+        if filetype == IA_IFREG:
+                type_bit = stat.S_IFREG
+        elif filetype == IA_IFDIR:
+                type_bit = stat.S_IFDIR
+        elif filetype == IA_IFLNK:
+                type_bit = stat.S_IFLNK
+        elif filetype == IA_IFBLK:
+                type_bit = stat.S_IFBLK
+        elif filetype == IA_IFCHR:
+                type_bit = stat.S_IFCHR
+        elif filetype == IA_IFIFO:
+                type_bit = stat.S_IFIFO
+        elif filetype == IA_IFSOCK:
+                type_bit = stat.S_IFSOCK
+        elif filetype == IA_INVAL:
+                pass
+
+
+        if prot.suid:
+                prot_bit |= stat.S_ISUID
+        if prot.sgid:
+                prot_bit |= stat.S_ISGID
+        if prot.sticky:
+                prot_bit |= stat.S_ISVTX
+
+        if prot.owner.read:
+                prot_bit |= stat.S_IRUSR
+        if prot.owner.write:
+                prot_bit |= stat.S_IWUSR
+        if prot.owner.execn:
+                prot_bit |= stat.S_IXUSR
+
+        if prot.group.read:
+                prot_bit |= stat.S_IRGRP
+        if prot.group.write:
+                prot_bit |= stat.S_IWGRP
+        if prot.group.execn:
+                prot_bit |= stat.S_IXGRP
+
+        if prot.other.read:
+                prot_bit |= stat.S_IROTH
+        if prot.other.write:
+                prot_bit |= stat.S_IWOTH
+        if prot.other.execn:
+                prot_bit |= stat.S_IXOTH
+
+        st_mode = (type_bit | prot_bit)
+
+        return st_mode
+
+
+def trace_stat2str (buf):
+        gfid = uuid2str(buf.contents.ia_gfid)
+        mode = st_mode_from_ia(buf.contents.ia_prot, buf.contents.ia_type)
+        atime_buf = strftime("[%b %d %H:%M:%S]",
+                             localtime(buf.contents.ia_atime))
+        mtime_buf = strftime("[%b %d %H:%M:%S]",
+                             localtime(buf.contents.ia_mtime))
+        ctime_buf = strftime("[%b %d %H:%M:%S]",
+                             localtime(buf.contents.ia_ctime))
+        return ("(gfid={0:s}, ino={1:d}, mode={2:o}, nlink={3:d}, uid ={4:d}, "+
+                "gid ={5:d}, size={6:d}, blocks={7:d}, atime={8:s}, mtime={9:s}, "+
+                "ctime={10:s})").format(gfid, buf.contents.ia_no, mode,
+                                        buf.contents.ia_nlink,
+                                        buf.contents.ia_uid,
+                                        buf.contents.ia_gid,
+                                        buf.contents.ia_size,
+                                        buf.contents.ia_blocks,
+                                        atime_buf, mtime_buf,
+                                        ctime_buf)
+
+class xlator(Translator):
+
+        def __init__(self, c_this):
+                Translator.__init__(self, c_this)
+                self.gfids = {}
+
+        def lookup_fop(self, frame, this, loc, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.gfid)
+                print("GLUPY TRACE LOOKUP FOP- {0:d}: gfid={1:s}; " +
+                      "path={2:s}").format(unique, gfid, loc.contents.path)
+                self.gfids[key] = gfid
+                dl.wind_lookup(frame, POINTER(xlator_t)(), loc, xdata)
+                return 0
+
+        def lookup_cbk(self, frame, cookie, this, op_ret, op_errno,
+                       inode, buf, xdata, postparent):
+                unique =dl.get_rootunique(frame)
+                key =dl.get_id(frame)
+                if op_ret == 0:
+                        gfid = uuid2str(buf.contents.ia_gfid)
+                        statstr = trace_stat2str(buf)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE LOOKUP CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; *buf={3:s}; " +
+                              "*postparent={4:s}").format(unique, gfid,
+                                                          op_ret, statstr,
+                                                          postparentstr)
+                else:
+                        gfid = self.gfids[key]
+                        print("GLUPY TRACE LOOKUP CBK - {0:d}: gfid={1:s};" +
+                              " op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                      gfid,
+                                                                      op_ret,
+                                                                      op_errno)
+                del self.gfids[key]
+                dl.unwind_lookup(frame, cookie, this, op_ret, op_errno,
+                                 inode, buf, xdata, postparent)
+                return 0
+
+        def create_fop(self, frame, this, loc, flags, mode, umask, fd,
+                       xdata):
+                unique = dl.get_rootunique(frame)
+                gfid = uuid2str(loc.contents.gfid)
+                print("GLUPY TRACE CREATE FOP- {0:d}: gfid={1:s}; path={2:s}; " +
+                      "fd={3:s}; flags=0{4:o}; mode=0{5:o}; " +
+                      "umask=0{6:o}").format(unique, gfid, loc.contents.path,
+                                             fd, flags, mode, umask)
+                dl.wind_create(frame, POINTER(xlator_t)(), loc, flags,mode,
+                               umask, fd, xdata)
+                return 0
+
+        def create_cbk(self, frame, cookie, this, op_ret, op_errno, fd,
+                       inode, buf, preparent, postparent, xdata):
+                unique = dl.get_rootunique(frame)
+                if op_ret >= 0:
+                        gfid = uuid2str(inode.contents.gfid)
+                        statstr = trace_stat2str(buf)
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE CREATE CBK- {0:d}: gfid={1:s};" +
+                              " op_ret={2:d}; fd={3:s}; *stbuf={4:s}; " +
+                              "*preparent={5:s};" +
+                              " *postparent={6:s}").format(unique, gfid, op_ret,
+                                                           fd, statstr,
+                                                           preparentstr,
+                                                           postparentstr)
+                else:
+                        print ("GLUPY TRACE CREATE CBK- {0:d}: op_ret={1:d}; " +
+                              "op_errno={2:d}").format(unique, op_ret, op_errno)
+                dl.unwind_create(frame, cookie, this, op_ret, op_errno, fd,
+                                 inode, buf, preparent, postparent, xdata)
+                return 0
+
+        def open_fop(self, frame, this, loc, flags, fd, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE OPEN FOP- {0:d}: gfid={1:s}; path={2:s}; "+
+                      "flags={3:d}; fd={4:s}").format(unique, gfid,
+                                                      loc.contents.path, flags,
+                                                      fd)
+                self.gfids[key] = gfid
+                dl.wind_open(frame, POINTER(xlator_t)(), loc, flags, fd, xdata)
+                return 0
+
+        def open_cbk(self, frame, cookie, this, op_ret, op_errno, fd, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE OPEN CBK- {0:d}: gfid={1:s}; op_ret={2:d}; "
+                      "op_errno={3:d}; *fd={4:s}").format(unique, gfid,
+                                                          op_ret, op_errno, fd)
+                del self.gfids[key]
+                dl.unwind_open(frame, cookie, this, op_ret, op_errno, fd,
+                               xdata)
+                return 0
+
+        def readv_fop(self, frame, this, fd, size, offset, flags, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE READV FOP- {0:d}: gfid={1:s}; "+
+                      "fd={2:s}; size ={3:d}; offset={4:d}; " +
+                      "flags=0{5:x}").format(unique, gfid, fd, size, offset,
+                                             flags)
+                self.gfids[key] = gfid
+                dl.wind_readv (frame, POINTER(xlator_t)(), fd, size, offset,
+                               flags, xdata)
+                return 0
+
+        def readv_cbk(self, frame, cookie, this, op_ret, op_errno, vector,
+                      count, buf, iobref, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret >= 0:
+                        statstr = trace_stat2str(buf)
+                        print("GLUPY TRACE READV CBK- {0:d}: gfid={1:s}, "+
+                              "op_ret={2:d}; *buf={3:s};").format(unique, gfid,
+                                                                  op_ret,
+                                                                  statstr)
+
+                else:
+                        print("GLUPY TRACE READV CBK- {0:d}: gfid={1:s}, "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_readv (frame, cookie, this, op_ret, op_errno,
+                                 vector, count, buf, iobref, xdata)
+                return 0
+
+        def writev_fop(self, frame, this, fd, vector, count, offset, flags,
+                       iobref, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE  WRITEV FOP- {0:d}: gfid={1:s}; " +
+                      "fd={2:s}; count={3:d}; offset={4:d}; " +
+                      "flags=0{5:x}").format(unique, gfid, fd, count, offset,
+                                             flags)
+                self.gfids[key] = gfid
+                dl.wind_writev(frame, POINTER(xlator_t)(), fd, vector, count,
+                               offset, flags, iobref, xdata)
+                return 0
+
+        def writev_cbk(self, frame, cookie, this, op_ret, op_errno, prebuf,
+                       postbuf, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                if op_ret >= 0:
+                        preopstr = trace_stat2str(prebuf)
+                        postopstr = trace_stat2str(postbuf)
+                        print("GLUPY TRACE WRITEV CBK- {0:d}: op_ret={1:d}; " +
+                              "*prebuf={2:s}; " +
+                              "*postbuf={3:s}").format(unique, op_ret, preopstr,
+                                                       postopstr)
+                else:
+                        gfid = self.gfids[key]
+                        print("GLUPY TRACE WRITEV CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_writev (frame, cookie, this, op_ret, op_errno,
+                                  prebuf, postbuf, xdata)
+                return 0
+
+        def opendir_fop(self, frame, this, loc, fd, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE OPENDIR FOP- {0:d}: gfid={1:s}; path={2:s}; "+
+                      "fd={3:s}").format(unique, gfid, loc.contents.path, fd)
+                self.gfids[key] = gfid
+                dl.wind_opendir(frame, POINTER(xlator_t)(), loc, fd, xdata)
+                return 0
+
+        def opendir_cbk(self, frame, cookie, this, op_ret, op_errno, fd,
+                        xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE OPENDIR CBK- {0:d}: gfid={1:s}; op_ret={2:d};"+
+                      " op_errno={3:d}; fd={4:s}").format(unique, gfid, op_ret,
+                                                          op_errno, fd)
+                del self.gfids[key]
+                dl.unwind_opendir(frame, cookie, this, op_ret, op_errno,
+                                  fd, xdata)
+                return 0
+
+        def readdir_fop(self, frame, this, fd, size, offset, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE READDIR FOP- {0:d}:  gfid={1:s}; fd={2:s}; " +
+                      "size={3:d}; offset={4:d}").format(unique, gfid, fd, size,
+                                                         offset)
+                self.gfids[key] = gfid
+                dl.wind_readdir(frame, POINTER(xlator_t)(), fd, size, offset,
+                                xdata)
+                return 0
+
+        def readdir_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
+                        xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE READDIR CBK- {0:d}: gfid={1:s}; op_ret={2:d};"+
+                      " op_errno={3:d}").format(unique, gfid, op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_readdir(frame, cookie, this, op_ret, op_errno, buf,
+                                  xdata)
+                return 0
+
+        def readdirp_fop(self, frame, this, fd, size, offset, dictionary):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE READDIRP FOP- {0:d}: gfid={1:s}; fd={2:s}; "+
+                      " size={3:d}; offset={4:d}").format(unique, gfid, fd, size,
+                                                          offset)
+                self.gfids[key] = gfid
+                dl.wind_readdirp(frame, POINTER(xlator_t)(), fd, size, offset,
+                                 dictionary)
+                return 0
+
+        def readdirp_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
+                         xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE READDIRP CBK- {0:d}: gfid={1:s}; "+
+                      "op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
+                                                             op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_readdirp(frame, cookie, this, op_ret, op_errno, buf,
+                                  xdata)
+                return 0
+
+        def mkdir_fop(self, frame, this, loc, mode, umask, xdata):
+                unique = dl.get_rootunique(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE MKDIR FOP- {0:d}: gfid={1:s}; path={2:s}; " +
+                      "mode={3:d}; umask=0{4:o}").format(unique, gfid,
+                                                         loc.contents.path, mode,
+                                                         umask)
+                dl.wind_mkdir(frame, POINTER(xlator_t)(), loc, mode, umask,
+                              xdata)
+                return 0
+
+        def mkdir_cbk(self, frame, cookie, this, op_ret, op_errno, inode, buf,
+                      preparent, postparent,  xdata):
+                unique = dl.get_rootunique(frame)
+                if op_ret == 0:
+                        gfid = uuid2str(inode.contents.gfid)
+                        statstr = trace_stat2str(buf)
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE MKDIR CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; *stbuf={3:s}; *prebuf={4:s}; "+
+                              "*postbuf={5:s} ").format(unique, gfid, op_ret,
+                                                        statstr,
+                                                        preparentstr,
+                                                        postparentstr)
+                else:
+                        print("GLUPY TRACE MKDIR CBK- {0:d}:  op_ret={1:d}; "+
+                              "op_errno={2:d}").format(unique, op_ret, op_errno)
+                dl.unwind_mkdir(frame, cookie, this, op_ret, op_errno, inode,
+                                buf, preparent, postparent, xdata)
+                return 0
+
+        def rmdir_fop(self, frame, this, loc, flags, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE RMDIR FOP- {0:d}: gfid={1:s}; path={2:s}; "+
+                      "flags={3:d}").format(unique, gfid, loc.contents.path,
+                                            flags)
+                self.gfids[key] = gfid
+                dl.wind_rmdir(frame, POINTER(xlator_t)(), loc, flags, xdata)
+                return 0
+
+        def rmdir_cbk(self, frame, cookie, this, op_ret, op_errno, preparent,
+                      postparent, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE RMDIR CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; *prebuf={3:s}; "+
+                              "*postbuf={4:s}").format(unique, gfid, op_ret,
+                                                       preparentstr,
+                                                       postparentstr)
+                else:
+                        print("GLUPY TRACE RMDIR CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_rmdir(frame, cookie, this, op_ret, op_errno,
+                                preparent, postparent, xdata)
+                return 0
+
+        def stat_fop(self, frame, this, loc, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE STAT FOP- {0:d}: gfid={1:s}; " +
+                      " path={2:s}").format(unique, gfid, loc.contents.path)
+                self.gfids[key] = gfid
+                dl.wind_stat(frame, POINTER(xlator_t)(), loc, xdata)
+                return 0
+
+        def stat_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
+                     xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        statstr = trace_stat2str(buf)
+                        print("GLUPY TRACE STAT CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d};  *buf={3:s};").format(unique,
+                                                                   gfid,
+                                                                   op_ret,
+                                                                   statstr)
+                else:
+                        print("GLUPY TRACE STAT CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_stat(frame, cookie, this, op_ret, op_errno,
+                               buf, xdata)
+                return 0
+
+        def fstat_fop(self, frame, this, fd, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE FSTAT FOP- {0:d}:  gfid={1:s}; " +
+                      "fd={2:s}").format(unique, gfid, fd)
+                self.gfids[key] = gfid
+                dl.wind_fstat(frame, POINTER(xlator_t)(), fd, xdata)
+                return 0
+
+        def fstat_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
+                      xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        statstr = trace_stat2str(buf)
+                        print("GLUPY TRACE FSTAT CBK- {0:d}: gfid={1:s} "+
+                              " op_ret={2:d}; *buf={3:s}").format(unique,
+                                                                  gfid,
+                                                                  op_ret,
+                                                                  statstr)
+                else:
+                        print("GLUPY TRACE FSTAT CBK- {0:d}: gfid={1:s} "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique.
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_fstat(frame, cookie, this, op_ret, op_errno,
+                                buf, xdata)
+                return 0
+
+        def statfs_fop(self, frame, this, loc, xdata):
+                unique = dl.get_rootunique(frame)
+                if loc.contents.inode:
+                        gfid = uuid2str(loc.contents.inode.contents.gfid)
+                else:
+                        gfid = "0"
+                print("GLUPY TRACE STATFS FOP- {0:d}: gfid={1:s}; "+
+                      "path={2:s}").format(unique, gfid, loc.contents.path)
+                dl.wind_statfs(frame, POINTER(xlator_t)(), loc, xdata)
+                return 0
+
+        def statfs_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
+                       xdata):
+                unique = dl.get_rootunique(frame)
+                if op_ret == 0:
+                        #TBD: print buf (pointer to an iovec type object)
+                        print("GLUPY TRACE STATFS CBK {0:d}: "+
+                              "op_ret={1:d}").format(unique, op_ret)
+                else:
+                        print("GLUPY TRACE STATFS CBK-  {0:d}"+
+                              "op_ret={1:d}; op_errno={2:d}").format(unique,
+                                                                     op_ret,
+                                                                     op_errno)
+                dl.unwind_statfs(frame, cookie, this, op_ret, op_errno,
+                                 buf, xdata)
+                return 0
+
+        def getxattr_fop(self, frame, this, loc, name, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE GETXATTR FOP- {0:d}: gfid={1:s}; path={2:s};"+
+                      " name={3:s}").format(unique, gfid, loc.contents.path,
+                                            name)
+                self.gfids[key]=gfid
+                dl.wind_getxattr(frame, POINTER(xlator_t)(), loc, name, xdata)
+                return 0
+
+        def getxattr_cbk(self, frame, cookie, this, op_ret, op_errno,
+                         dictionary, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE GETXATTR CBK- {0:d}: gfid={1:s}; "+
+                      "op_ret={2:d}; op_errno={3:d}; "+
+                      " dictionary={4:s}").format(unique, gfid, op_ret, op_errno,
+                                                  dictionary)
+                del self.gfids[key]
+                dl.unwind_getxattr(frame, cookie, this, op_ret, op_errno,
+                                   dictionary, xdata)
+                return 0
+
+        def fgetxattr_fop(self, frame, this, fd, name, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE FGETXATTR FOP- {0:d}: gfid={1:s}; fd={2:s}; "+
+                      "name={3:s}").format(unique, gfid, fd, name)
+                self.gfids[key] = gfid
+                dl.wind_fgetxattr(frame, POINTER(xlator_t)(), fd, name, xdata)
+                return 0
+
+        def fgetxattr_cbk(self, frame, cookie, this, op_ret, op_errno,
+                          dictionary, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE FGETXATTR CBK- {0:d}: gfid={1:s}; "+
+                      "op_ret={2:d}; op_errno={3:d};"+
+                      " dictionary={4:s}").format(unique, gfid, op_ret,
+                                                  op_errno, dictionary)
+                del self.gfids[key]
+                dl.unwind_fgetxattr(frame, cookie, this, op_ret, op_errno,
+                                    dictionary, xdata)
+                return 0
+
+        def setxattr_fop(self, frame, this, loc, dictionary, flags, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE SETXATTR FOP- {0:d}:  gfid={1:s}; path={2:s};"+
+                      " flags={3:d}").format(unique, gfid, loc.contents.path,
+                                             flags)
+                self.gfids[key] = gfid
+                dl.wind_setxattr(frame, POINTER(xlator_t)(), loc, dictionary,
+                                 flags, xdata)
+                return 0
+
+        def setxattr_cbk(self, frame, cookie, this, op_ret, op_errno, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE SETXATTR CBK- {0:d}: gfid={1:s}; "+
+                      "op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
+                                                             op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_setxattr(frame, cookie, this, op_ret, op_errno,
+                                   xdata)
+                return 0
+
+        def fsetxattr_fop(self, frame, this, fd, dictionary, flags, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE FSETXATTR FOP- {0:d}: gfid={1:s}; fd={2:p}; "+
+                      "flags={3:d}").format(unique, gfid, fd, flags)
+                self.gfids[key] = gfid
+                dl.wind_fsetxattr(frame, POINTER(xlator_t)(), fd, dictionary,
+                                  flags, xdata)
+                return 0
+
+        def fsetxattr_cbk(self, frame, cookie, this, op_ret, op_errno, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE FSETXATTR CBK- {0:d}: gfid={1:s};  "+
+                      "op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
+                                                             op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_fsetxattr(frame, cookie, this, op_ret, op_errno,
+                                   xdata)
+                return 0
+
+        def removexattr_fop(self, frame, this, loc, name, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE REMOVEXATTR FOP- {0:d}:  gfid={1:s}; "+
+                      "path={2:s}; name={3:s}").format(unique, gfid,
+                                                       loc.contents.path,
+                                                       name)
+                self.gfids[key] = gfid
+                dl.wind_removexattr(frame, POINTER(xlator_t)(), loc, name,
+                                    xdata)
+                return 0
+
+        def removexattr_cbk(self, frame, cookie, this, op_ret, op_errno,
+                            xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE REMOVEXATTR CBK- {0:d}: gfid={1:s} "+
+                      " op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
+                                                              op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_removexattr(frame, cookie, this, op_ret, op_errno,
+                                      xdata)
+                return 0
+
+        def link_fop(self, frame, this, oldloc, newloc, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                if (newloc.contents.inode):
+                        newgfid = uuid2str(newloc.contents.inode.contents.gfid)
+                else:
+                        newgfid = "0"
+                oldgfid = uuid2str(oldloc.contents.inode.contents.gfid)
+                print("GLUPY TRACE LINK FOP-{0:d}: oldgfid={1:s}; oldpath={2:s};"+
+                      "newgfid={3:s};"+
+                      "newpath={4:s}").format(unique, oldgfid,
+                                              oldloc.contents.path,
+                                              newgfid,
+                                              newloc.contents.path)
+                self.gfids[key] =  oldgfid
+                dl.wind_link(frame, POINTER(xlator_t)(), oldloc, newloc,
+                             xdata)
+                return 0
+
+        def link_cbk(self, frame, cookie, this, op_ret, op_errno, inode, buf,
+                     preparent, postparent, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        statstr = trace_stat2str(buf)
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE LINK CBK- {0:d}: op_ret={1:d} "+
+                              "*stbuf={2:s}; *prebuf={3:s}; "+
+                              "*postbuf={4:s} ").format(unique, op_ret, statstr,
+                                                        preparentstr,
+                                                        postparentstr)
+                else:
+                        print("GLUPY TRACE LINK CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; "+
+                              "op_errno={3:d}").format(unique, gfid,
+                                                       op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_link(frame, cookie, this, op_ret, op_errno, inode,
+                               buf, preparent, postparent, xdata)
+                return 0
+
+        def unlink_fop(self, frame, this, loc, xflag, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE UNLINK FOP- {0:d}; gfid={1:s}; path={2:s}; "+
+                      "flag={3:d}").format(unique, gfid, loc.contents.path,
+                                           xflag)
+                self.gfids[key] = gfid
+                dl.wind_unlink(frame, POINTER(xlator_t)(), loc, xflag,
+                               xdata)
+                return 0
+
+        def unlink_cbk(self, frame, cookie, this, op_ret, op_errno,
+                       preparent, postparent, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE UNLINK CBK- {0:d}: gfid ={1:s}; "+
+                              "op_ret={2:d}; *prebuf={3:s}; "+
+                              "*postbuf={4:s} ").format(unique, gfid, op_ret,
+                                                        preparentstr,
+                                                        postparentstr)
+                else:
+                        print("GLUPY TRACE UNLINK CBK: {0:d}: gfid ={1:s}; "+
+                              "op_ret={2:d}; "+
+                              "op_errno={3:d}").format(unique, gfid, op_ret,
+                                                       op_errno)
+                del self.gfids[key]
+                dl.unwind_unlink(frame, cookie, this, op_ret, op_errno,
+                                 preparent, postparent, xdata)
+                return 0
+
+        def readlink_fop(self, frame, this, loc, size, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE READLINK FOP- {0:d}:  gfid={1:s}; path={2:s};"+
+                      " size={3:d}").format(unique, gfid, loc.contents.path,
+                                            size)
+                self.gfids[key] = gfid
+                dl.wind_readlink(frame, POINTER(xlator_t)(), loc, size,
+                               xdata)
+                return 0
+
+        def readlink_cbk(self, frame, cookie, this, op_ret, op_errno,
+                         buf, stbuf, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid  = self.gfids[key]
+                if op_ret == 0:
+                        statstr = trace_stat2str(stbuf)
+                        print("GLUPY TRACE READLINK CBK- {0:d}: gfid={1:s} "+
+                              " op_ret={2:d}; op_errno={3:d}; *prebuf={4:s}; "+
+                              "*postbuf={5:s} ").format(unique, gfid,
+                                                        op_ret, op_errno,
+                                                        buf, statstr)
+                else:
+                        print("GLUPY TRACE READLINK CBK- {0:d}: gfid={1:s} "+
+                              " op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                      gfid,
+                                                                      op_ret,
+                                                                      op_errno)
+                del self.gfids[key]
+                dl.unwind_readlink(frame, cookie, this, op_ret, op_errno, buf,
+                                   stbuf, xdata)
+                return 0
+
+        def symlink_fop(self, frame, this, linkpath, loc, umask, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE SYMLINK FOP- {0:d}: gfid={1:s}; "+
+                      "linkpath={2:s}; path={3:s};"+
+                      "umask=0{4:o}").format(unique, gfid, linkpath,
+                                             loc.contents.path, umask)
+                self.gfids[key] = gfid
+                dl.wind_symlink(frame, POINTER(xlator_t)(), linkpath, loc,
+                                umask, xdata)
+                return 0
+
+        def symlink_cbk(self, frame, cookie, this, op_ret, op_errno,
+                        inode, buf, preparent, postparent, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        statstr = trace_stat2str(buf)
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE SYMLINK CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; *stbuf={3:s}; *preparent={4:s}; "+
+                              "*postparent={5:s}").format(unique, gfid,
+                                                          op_ret, statstr,
+                                                          preparentstr,
+                                                          postparentstr)
+                else:
+                        print("GLUPY TRACE SYMLINK CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_symlink(frame, cookie, this, op_ret, op_errno,
+                                  inode, buf, preparent, postparent, xdata)
+                return 0
diff --git a/xlators/features/glupy/src/glupy.c b/xlators/features/glupy/src/glupy.c
new file mode 100644
index 000000000..dc86c0071
--- /dev/null
+++ b/xlators/features/glupy/src/glupy.c
@@ -0,0 +1,2470 @@
+/*
+  Copyright (c) 2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  GlusterFS is free software; you can redistribute it and/or modify
+  it under the terms of the GNU General Public License as published
+  by the Free Software Foundation; either version 3 of the License,
+  or (at your option) any later version.
+
+  GlusterFS is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+  General Public License for more details.
+
+  You should have received a copy of the GNU General Public License
+  along with this program.  If not, see
+  <http://www.gnu.org/licenses/>.
+*/
+
+#include <ctype.h>
+#include <sys/uio.h>
+#include <Python.h>
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "xlator.h"
+#include "logging.h"
+#include "defaults.h"
+
+#include "glupy.h"
+
+/* UTILITY FUNCTIONS FOR FOP-SPECIFIC CODE */
+
+pthread_key_t gil_init_key;
+
+PyGILState_STATE
+glupy_enter (void)
+{
+#if 0
+        if (!pthread_getspecific(gil_init_key)) {
+                PyEval_ReleaseLock();
+                (void)pthread_setspecific(gil_init_key,(void *)1);
+        }
+#endif
+
+        return PyGILState_Ensure();
+}
+
+void
+glupy_leave (PyGILState_STATE gstate)
+{
+        PyGILState_Release(gstate);
+}
+
+/* FOP: LOOKUP */
+
+int32_t
+glupy_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, inode_t *inode,
+                  struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_LOOKUP]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_lookup_cbk_t)(priv->cbks[GLUPY_LOOKUP]))(
+                frame, cookie, this, op_ret, op_errno,
+                inode, buf, xdata, postparent);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
+                             xdata, postparent);
+        return 0;
+}
+
+int32_t
+glupy_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
+              dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_LOOKUP]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_lookup_t)(priv->fops[GLUPY_LOOKUP]))(
+                frame, this, loc, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_lookup_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, loc, xdata);
+        return 0;
+}
+
+void
+wind_lookup (call_frame_t *frame, xlator_t *xl, loc_t *loc, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_lookup_cbk,xl,xl->fops->lookup,loc,xdata);
+}
+
+void
+unwind_lookup (call_frame_t *frame, long cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, inode_t *inode,
+               struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(lookup,frame,op_ret,op_errno,
+                            inode,buf,xdata,postparent);
+}
+
+void
+set_lookup_fop (long py_this, fop_lookup_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_LOOKUP] = (long)fop;
+}
+
+void
+set_lookup_cbk (long py_this, fop_lookup_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_LOOKUP] = (long)cbk;
+}
+
+/* FOP: CREATE */
+
+int32_t
+glupy_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+                  struct iatt *buf, struct iatt *preparent,
+                  struct iatt *postparent, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_CREATE]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_create_cbk_t)(priv->cbks[GLUPY_CREATE]))(
+                frame, cookie, this, op_ret, op_errno,
+                fd, inode, buf, preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+              mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_CREATE]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_create_t)(priv->fops[GLUPY_CREATE]))(
+                frame, this, loc, flags, mode, umask, fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_create_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->create, loc, flags, mode, umask,
+                    fd, xdata);
+        return 0;
+}
+
+void
+wind_create (call_frame_t *frame, xlator_t *xl, loc_t *loc, int32_t flags,
+             mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_create_cbk,xl, xl->fops->create,
+                    loc, flags, mode, umask, fd, xdata);
+}
+
+void
+unwind_create (call_frame_t *frame, long cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+               struct iatt *buf, struct iatt *preparent,
+               struct iatt *postparent, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
+                             preparent, postparent, xdata);
+}
+
+void
+set_create_fop (long py_this, fop_create_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_CREATE] = (long)fop;
+}
+
+void
+set_create_cbk (long py_this, fop_create_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_CREATE] = (long)cbk;
+}
+
+/* FOP: OPEN */
+
+int32_t
+glupy_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_OPEN]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_open_cbk_t)(priv->cbks[GLUPY_OPEN]))(
+                frame, cookie, this, op_ret, op_errno,
+                fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+        return 0;
+}
+
+int32_t
+glupy_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            int32_t flags, fd_t *fd, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_OPEN]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_open_t)(priv->fops[GLUPY_OPEN]))(
+                frame, this, loc, flags, fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_open_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+        return 0;
+}
+
+void
+wind_open (call_frame_t *frame, xlator_t *xl, loc_t *loc, int32_t flags,
+           fd_t *fd, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_open_cbk, xl, xl->fops->open, loc, flags,
+                    fd, xdata);
+}
+
+void
+unwind_open (call_frame_t *frame, long cookie, xlator_t *this,
+             int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+}
+
+void
+set_open_fop (long py_this, fop_open_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->fops[GLUPY_OPEN] = (long)fop;
+}
+
+void
+set_open_cbk (long py_this, fop_open_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->cbks[GLUPY_OPEN] = (long)cbk;
+}
+
+/* FOP: READV */
+
+int32_t
+glupy_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, struct iovec *vector,
+                 int32_t count, struct iatt *stbuf, struct iobref *iobref,
+                 dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_READV]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_readv_cbk_t)(priv->cbks[GLUPY_READV]))(
+                frame, cookie, this, op_ret, op_errno,
+                vector, count, stbuf, iobref, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector,
+                             count, stbuf, iobref, xdata);
+        return 0;
+}
+
+int32_t
+glupy_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
+             size_t size, off_t offset, uint32_t flags, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_READV]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_readv_t)(priv->fops[GLUPY_READV]))(
+                frame, this, fd, size, offset, flags, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_readv_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readv, fd, size, offset,
+                    flags, xdata);
+        return 0;
+}
+
+void
+wind_readv (call_frame_t *frame, xlator_t *xl, fd_t *fd, size_t size,
+            off_t offset, uint32_t flags, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_readv_cbk, xl, xl->fops->readv, fd, size,
+                    offset, flags, xdata);
+}
+
+void
+unwind_readv (call_frame_t *frame, long cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, struct iovec *vector,
+              int32_t count, struct iatt *stbuf, struct iobref *iobref,
+              dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector,
+                             count, stbuf, iobref, xdata);
+}
+
+void
+set_readv_fop (long py_this, fop_readv_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->fops[GLUPY_READV] = (long)fop;
+}
+
+void
+set_readv_cbk (long py_this, fop_readv_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->cbks[GLUPY_READV] = (long)cbk;
+}
+
+/* FOP: WRITEV */
+
+int32_t
+glupy_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                  struct iatt *postbuf, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_WRITEV]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_writev_cbk_t)(priv->cbks[GLUPY_WRITEV]))(
+                frame, cookie, this, op_ret, op_errno,
+                prebuf, postbuf, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+        return 0;
+}
+
+int32_t
+glupy_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+              struct iovec *vector, int32_t count, off_t offset,
+              uint32_t flags, struct iobref *iobref, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_WRITEV]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_writev_t)(priv->fops[GLUPY_WRITEV]))(
+                frame, this, fd, vector, count, offset, flags,
+                iobref, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_writev_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->writev, fd, vector, count,
+                    offset, flags, iobref, xdata);
+        return 0;
+}
+
+void
+wind_writev (call_frame_t *frame, xlator_t *xl, fd_t *fd, struct iovec *vector,
+             int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
+             dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_writev_cbk, xl, xl->fops->writev, fd, vector,
+                    count, offset, flags, iobref, xdata);
+}
+
+void
+unwind_writev (call_frame_t *frame, long cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+               struct iatt *postbuf, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+}
+
+void
+set_writev_fop (long py_this, fop_writev_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->fops[GLUPY_WRITEV] = (long)fop;
+}
+
+void
+set_writev_cbk (long py_this, fop_writev_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->cbks[GLUPY_WRITEV] = (long)cbk;
+}
+
+
+/* FOP: OPENDIR */
+
+int32_t
+glupy_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, fd_t *fd,
+                   dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_OPENDIR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_opendir_cbk_t)(priv->cbks[GLUPY_OPENDIR]))(
+                frame, cookie, this, op_ret, op_errno,
+                fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
+        return 0;
+}
+
+int32_t
+glupy_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc,
+               fd_t *fd, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_OPENDIR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_opendir_t)(priv->fops[GLUPY_OPENDIR]))(
+                frame, this, loc, fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_opendir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
+        return 0;
+}
+
+void
+wind_opendir (call_frame_t *frame, xlator_t *xl, loc_t *loc, fd_t *fd, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_opendir_cbk,xl,xl->fops->opendir,loc,fd,xdata);
+}
+
+void
+unwind_opendir (call_frame_t *frame, long cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(opendir,frame,op_ret,op_errno,
+                            fd,xdata);
+}
+
+void
+set_opendir_fop (long py_this, fop_opendir_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_OPENDIR] = (long)fop;
+}
+
+void
+set_opendir_cbk (long py_this, fop_opendir_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_OPENDIR] = (long)cbk;
+}
+
+/* FOP: READDIR */
+
+int32_t
+glupy_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                   dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_READDIR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_readdir_cbk_t)(priv->cbks[GLUPY_READDIR]))(
+                frame, cookie, this, op_ret, op_errno,
+                entries,  xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, entries,
+                             xdata);
+        return 0;
+}
+
+int32_t
+glupy_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
+               size_t size, off_t offset, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_READDIR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_readdir_t)(priv->fops[GLUPY_READDIR]))(
+                frame, this, fd, size, offset, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_readdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdir,fd, size, offset, xdata);
+        return 0;
+}
+
+void
+wind_readdir(call_frame_t *frame, xlator_t *xl, fd_t *fd, size_t size,
+             off_t offset, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_readdir_cbk,xl,xl->fops->readdir,fd,size,offset,xdata);
+}
+
+void
+unwind_readdir (call_frame_t *frame, long cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(readdir,frame,op_ret,op_errno,
+                            entries, xdata);
+}
+
+void
+set_readdir_fop (long py_this, fop_readdir_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_READDIR] = (long)fop;
+}
+
+void
+set_readdir_cbk (long py_this, fop_readdir_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_READDIR] = (long)cbk;
+}
+
+
+/* FOP: READDIRP */
+
+int32_t
+glupy_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                    dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_READDIRP]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_readdirp_cbk_t)(priv->cbks[GLUPY_READDIRP]))(
+                frame, cookie, this, op_ret, op_errno,
+                entries,  xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries,
+                             xdata);
+        return 0;
+}
+
+int32_t
+glupy_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                size_t size, off_t offset, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_READDIRP]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_readdirp_t)(priv->fops[GLUPY_READDIRP]))(
+                frame, this, fd, size, offset, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_readdirp_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdirp,fd, size, offset, xdata);
+        return 0;
+}
+
+void
+wind_readdirp (call_frame_t *frame, xlator_t *xl, fd_t *fd, size_t size,
+               off_t offset, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_readdirp_cbk,xl,xl->fops->readdirp,fd,size,offset,xdata);
+}
+
+void
+unwind_readdirp (call_frame_t *frame, long cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                 dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(readdirp,frame,op_ret,op_errno,
+                            entries, xdata);
+}
+
+void
+set_readdirp_fop (long py_this, fop_readdirp_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_READDIRP] = (long)fop;
+}
+
+void
+set_readdirp_cbk (long py_this, fop_readdirp_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_READDIRP] = (long)cbk;
+}
+
+
+/* FOP:STAT */
+
+int32_t
+glupy_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_STAT]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_stat_cbk_t)(priv->cbks[GLUPY_STAT]))(
+                frame, cookie, this, op_ret, op_errno,
+                buf, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+int32_t
+glupy_stat (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_STAT]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_stat_t)(priv->fops[GLUPY_STAT]))(
+                frame, this, loc, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_stat_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->stat, loc, xdata);
+        return 0;
+}
+
+void
+wind_stat (call_frame_t *frame, xlator_t *xl, loc_t *loc, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_stat_cbk,xl,xl->fops->stat,loc,xdata);
+}
+
+void
+unwind_stat (call_frame_t *frame, long cookie, xlator_t *this,
+             int32_t op_ret, int32_t op_errno, struct iatt *buf,
+             dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(stat,frame,op_ret,op_errno,
+                            buf,xdata);
+}
+
+void
+set_stat_fop (long py_this, fop_stat_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_STAT] = (long)fop;
+}
+
+void
+set_stat_cbk (long py_this, fop_stat_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_STAT] = (long)cbk;
+}
+
+
+/* FOP: FSTAT */
+
+int32_t
+glupy_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_FSTAT]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_fstat_cbk_t)(priv->cbks[GLUPY_FSTAT]))(
+                frame, cookie, this, op_ret, op_errno,
+                buf, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+int32_t
+glupy_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd,
+             dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_FSTAT]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_fstat_t)(priv->fops[GLUPY_FSTAT]))(
+                frame, this, fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_fstat_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fstat, fd, xdata);
+        return 0;
+}
+
+void
+wind_fstat (call_frame_t *frame, xlator_t *xl, fd_t *fd, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_fstat_cbk,xl,xl->fops->fstat,fd,xdata);
+}
+
+void
+unwind_fstat (call_frame_t *frame, long cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, struct iatt *buf,
+              dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(fstat,frame,op_ret,op_errno,
+                            buf,xdata);
+}
+
+void
+set_fstat_fop (long py_this, fop_fstat_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_FSTAT] = (long)fop;
+}
+
+void
+set_fstat_cbk (long py_this, fop_fstat_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_FSTAT] = (long)cbk;
+}
+
+/* FOP:STATFS */
+
+int32_t
+glupy_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct statvfs *buf, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_STATFS]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_statfs_cbk_t)(priv->cbks[GLUPY_STATFS]))(
+                frame, cookie, this, op_ret, op_errno,
+                buf, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+int32_t
+glupy_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc,
+              dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_STATFS]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_statfs_t)(priv->fops[GLUPY_STATFS]))(
+                frame, this, loc, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_statfs_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->statfs, loc, xdata);
+        return 0;
+}
+
+void
+wind_statfs (call_frame_t *frame, xlator_t *xl, loc_t *loc, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_statfs_cbk,xl,xl->fops->statfs,loc,xdata);
+}
+
+void
+unwind_statfs (call_frame_t *frame, long cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, struct statvfs *buf,
+               dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(statfs,frame,op_ret,op_errno,
+                            buf,xdata);
+}
+
+void
+set_statfs_fop (long py_this, fop_statfs_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_STATFS] = (long)fop;
+}
+
+void
+set_statfs_cbk (long py_this, fop_statfs_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_STATFS] = (long)cbk;
+}
+
+
+/* FOP: SETXATTR */
+
+int32_t
+glupy_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_SETXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_setxattr_cbk_t)(priv->cbks[GLUPY_SETXATTR]))(
+                frame, cookie, this, op_ret, op_errno,
+                xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+glupy_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_SETXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_setxattr_t)(priv->fops[GLUPY_SETXATTR]))(
+                frame, this, loc, dict, flags, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_setxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr, loc, dict,
+                    flags, xdata);
+        return 0;
+}
+
+void
+wind_setxattr (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+               dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_setxattr_cbk, xl, xl->fops->setxattr,
+                    loc, dict, flags, xdata);
+}
+
+
+void
+unwind_setxattr (call_frame_t *frame, long cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+
+}
+
+void
+set_setxattr_fop (long py_this, fop_setxattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_SETXATTR] = (long)fop;
+}
+
+void
+set_setxattr_cbk (long py_this, fop_setxattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_SETXATTR] = (long)cbk;
+}
+
+/* FOP: GETXATTR */
+
+int32_t
+glupy_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *dict,
+                    dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_GETXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_getxattr_cbk_t)(priv->cbks[GLUPY_GETXATTR]))(
+                frame, cookie, this, op_ret, op_errno, dict,
+                xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict,
+                             xdata);
+        return 0;
+}
+
+int32_t
+glupy_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                const char *name, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_GETXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_getxattr_t)(priv->fops[GLUPY_GETXATTR]))(
+                frame, this, loc, name, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, loc, name,
+                    xdata);
+        return 0;
+}
+
+void
+wind_getxattr (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+               const char *name, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_getxattr_cbk, xl, xl->fops->getxattr,
+                    loc, name, xdata);
+}
+
+
+void
+unwind_getxattr (call_frame_t *frame, long cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *dict,
+                 dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict,
+                             xdata);
+
+}
+
+
+void
+set_getxattr_fop (long py_this, fop_getxattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_GETXATTR] = (long)fop;
+}
+
+
+void
+set_getxattr_cbk (long py_this, fop_getxattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_GETXATTR] = (long)cbk;
+}
+
+/* FOP: FSETXATTR */
+
+int32_t
+glupy_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_FSETXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_fsetxattr_cbk_t)(priv->cbks[GLUPY_FSETXATTR]))(
+                frame, cookie, this, op_ret, op_errno,
+                xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+glupy_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_FSETXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_fsetxattr_t)(priv->fops[GLUPY_FSETXATTR]))(
+                frame, this, fd, dict, flags, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_fsetxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr, fd, dict,
+                    flags, xdata);
+        return 0;
+}
+
+void
+wind_fsetxattr (call_frame_t *frame, xlator_t *xl, fd_t *fd,
+                dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_fsetxattr_cbk, xl, xl->fops->fsetxattr,
+                    fd, dict, flags, xdata);
+}
+
+
+void
+unwind_fsetxattr (call_frame_t *frame, long cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+
+}
+
+void
+set_fsetxattr_fop (long py_this, fop_fsetxattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_FSETXATTR] = (long)fop;
+}
+
+void
+set_fsetxattr_cbk (long py_this, fop_fsetxattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_FSETXATTR] = (long)cbk;
+}
+
+/* FOP: FGETXATTR */
+
+int32_t
+glupy_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *dict,
+                     dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_FGETXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_fgetxattr_cbk_t)(priv->cbks[GLUPY_FGETXATTR]))(
+                frame, cookie, this, op_ret, op_errno, dict,
+                xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict,
+                             xdata);
+        return 0;
+}
+
+int32_t
+glupy_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 const char *name, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_FGETXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_fgetxattr_t)(priv->fops[GLUPY_FGETXATTR]))(
+                frame, this, fd, name, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_fgetxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fgetxattr, fd, name,
+                    xdata);
+        return 0;
+}
+
+void
+wind_fgetxattr (call_frame_t *frame, xlator_t *xl, fd_t *fd,
+                const char *name, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_fgetxattr_cbk, xl, xl->fops->fgetxattr,
+                    fd, name, xdata);
+}
+
+
+void
+unwind_fgetxattr (call_frame_t *frame, long cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *dict,
+                  dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict,
+                             xdata);
+
+}
+
+
+void
+set_fgetxattr_fop (long py_this, fop_fgetxattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_FGETXATTR] = (long)fop;
+}
+
+
+void
+set_fgetxattr_cbk (long py_this, fop_fgetxattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_FGETXATTR] = (long)cbk;
+}
+
+/* FOP:REMOVEXATTR */
+
+int32_t
+glupy_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_REMOVEXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_removexattr_cbk_t)(priv->cbks[GLUPY_REMOVEXATTR]))(
+                frame, cookie, this, op_ret, op_errno, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+glupy_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                   const char *name, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_REMOVEXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_removexattr_t)(priv->fops[GLUPY_REMOVEXATTR]))(
+                frame, this, loc, name, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_removexattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->removexattr, loc, name,
+                    xdata);
+        return 0;
+}
+
+void
+wind_removexattr (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+                  const char *name, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_removexattr_cbk, xl, xl->fops->removexattr,
+                    loc, name, xdata);
+}
+
+
+void
+unwind_removexattr (call_frame_t *frame, long cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
+
+}
+
+void
+set_removexattr_fop (long py_this, fop_removexattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_REMOVEXATTR] = (long)fop;
+}
+
+void
+set_removexattr_cbk (long py_this, fop_removexattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_REMOVEXATTR] = (long)cbk;
+}
+
+
+/* FOP:FREMOVEXATTR */
+
+int32_t
+glupy_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_FREMOVEXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_fremovexattr_cbk_t)(priv->cbks[GLUPY_FREMOVEXATTR]))(
+                frame, cookie, this, op_ret, op_errno, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+glupy_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                   const char *name, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_FREMOVEXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_fremovexattr_t)(priv->fops[GLUPY_FREMOVEXATTR]))(
+                frame, this, fd, name, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_fremovexattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fremovexattr, fd, name,
+                    xdata);
+        return 0;
+}
+
+void
+wind_fremovexattr (call_frame_t *frame, xlator_t *xl, fd_t *fd,
+                   const char *name, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_fremovexattr_cbk, xl, xl->fops->fremovexattr,
+                    fd, name, xdata);
+}
+
+
+void
+unwind_fremovexattr (call_frame_t *frame, long cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
+
+}
+
+void
+set_fremovexattr_fop (long py_this, fop_fremovexattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_FREMOVEXATTR] = (long)fop;
+}
+
+void
+set_fremovexattr_cbk (long py_this, fop_fremovexattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_FREMOVEXATTR] = (long)cbk;
+}
+
+
+/* FOP: LINK*/
+int32_t
+glupy_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, inode_t *inode,
+                struct iatt *buf, struct iatt *preparent,
+                struct iatt *postparent, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_LINK]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_link_cbk_t)(priv->cbks[GLUPY_LINK]))(
+                frame, cookie, this, op_ret, op_errno,
+                inode, buf, preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+           dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_LINK]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_link_t)(priv->fops[GLUPY_LINK]))(
+                frame, this, oldloc, newloc, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_link_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->link, oldloc, newloc,
+                    xdata);
+        return 0;
+}
+
+void
+wind_link (call_frame_t *frame, xlator_t *xl, loc_t *oldloc, loc_t *newloc,
+           dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_link_cbk, xl, xl->fops->link,
+                    oldloc, newloc, xdata);
+}
+
+void
+unwind_link (call_frame_t *frame, long cookie, xlator_t *this,
+             int32_t op_ret, int32_t op_errno, inode_t *inode,
+             struct iatt *buf, struct iatt *preparent,
+             struct iatt *postparent, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+}
+
+void
+set_link_fop (long py_this, fop_link_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_LINK] = (long)fop;
+}
+
+void
+set_link_cbk (long py_this, fop_link_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_LINK] = (long)cbk;
+}
+
+/* FOP: SYMLINK*/
+int32_t
+glupy_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, inode_t *inode,
+                   struct iatt *buf, struct iatt *preparent,
+                   struct iatt *postparent, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_SYMLINK]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_symlink_cbk_t)(priv->cbks[GLUPY_SYMLINK]))(
+                frame, cookie, this, op_ret, op_errno,
+                inode, buf, preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_symlink(call_frame_t *frame, xlator_t *this, const char *linkname,
+              loc_t *loc, mode_t umask, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_SYMLINK]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_symlink_t)(priv->fops[GLUPY_SYMLINK]))(
+                frame, this, linkname, loc, umask, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_symlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->symlink, linkname, loc,
+                    umask, xdata);
+        return 0;
+}
+
+void
+wind_symlink (call_frame_t *frame, xlator_t *xl, const char *linkname,
+              loc_t *loc, mode_t umask, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_symlink_cbk, xl, xl->fops->symlink,
+                    linkname, loc, umask, xdata);
+}
+
+void
+unwind_symlink (call_frame_t *frame, long cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, inode_t *inode,
+                struct iatt *buf, struct iatt *preparent,
+                struct iatt *postparent, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+}
+
+void
+set_symlink_fop (long py_this, fop_symlink_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_SYMLINK] = (long)fop;
+}
+
+void
+set_symlink_cbk (long py_this, fop_symlink_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_SYMLINK] = (long)cbk;
+}
+
+
+/* FOP: READLINK */
+int32_t
+glupy_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, const char *path,
+                    struct iatt *buf, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_READLINK]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_readlink_cbk_t)(priv->cbks[GLUPY_READLINK]))(
+                frame, cookie, this, op_ret, op_errno,
+                path, buf, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path,
+                             buf, xdata);
+        return 0;
+}
+
+int32_t
+glupy_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                size_t size, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_READLINK]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_readlink_t)(priv->fops[GLUPY_READLINK]))(
+                frame, this, loc, size, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_readlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readlink, loc,
+                    size, xdata);
+        return 0;
+}
+
+void
+wind_readlink (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+               size_t size, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_readlink_cbk, xl, xl->fops->readlink,
+                    loc, size, xdata);
+}
+
+void
+unwind_readlink (call_frame_t *frame, long cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, const char *path,
+                 struct iatt *buf, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path, buf,
+                             xdata);
+}
+
+void
+set_readlink_fop (long py_this, fop_readlink_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_READLINK] = (long)fop;
+}
+
+void
+set_readlink_cbk (long py_this, fop_readlink_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_READLINK] = (long)cbk;
+}
+
+
+/* FOP: UNLINK */
+
+int32_t
+glupy_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno,
+                  struct iatt *preparent, struct iatt *postparent,
+                  dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_UNLINK]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_unlink_cbk_t)(priv->cbks[GLUPY_UNLINK]))(
+                frame, cookie, this, op_ret, op_errno,
+                preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno, preparent,
+                             postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
+              int xflags, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_UNLINK]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_unlink_t)(priv->fops[GLUPY_UNLINK]))(
+                frame, this, loc, xflags, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_unlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->unlink, loc,
+                    xflags, xdata);
+        return 0;
+}
+
+void
+wind_unlink (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+             int xflags, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_unlink_cbk, xl, xl->fops->unlink,
+                    loc, xflags, xdata);
+}
+
+void
+unwind_unlink (call_frame_t *frame, long cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno,
+               struct iatt *preparent, struct iatt *postparent,
+               dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
+                             preparent, postparent, xdata);
+}
+
+void
+set_unlink_fop (long py_this, fop_unlink_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_UNLINK] = (long)fop;
+}
+
+void
+set_unlink_cbk (long py_this, fop_unlink_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_UNLINK] = (long)cbk;
+}
+
+
+/* FOP: MKDIR */
+
+int32_t
+glupy_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, inode_t *inode,
+                 struct iatt *buf, struct iatt *preparent,
+                 struct iatt *postparent, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_MKDIR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_mkdir_cbk_t)(priv->cbks[GLUPY_MKDIR]))(
+                frame, cookie, this, op_ret, op_errno,
+                inode, buf, preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+             mode_t umask, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_MKDIR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_mkdir_t)(priv->fops[GLUPY_MKDIR]))(
+                frame, this, loc, mode, umask, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_mkdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mkdir, loc, mode, umask,
+                    xdata);
+        return 0;
+}
+
+void
+wind_mkdir (call_frame_t *frame, xlator_t *xl, loc_t *loc, mode_t mode,
+            mode_t umask, dict_t *xdata)
+{
+
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_mkdir_cbk, xl, xl->fops->mkdir,
+                    loc, mode, umask, xdata);
+}
+
+void
+unwind_mkdir (call_frame_t *frame, long cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, inode_t *inode,
+              struct iatt *buf, struct iatt *preparent,
+              struct iatt *postparent, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+}
+
+void
+set_mkdir_fop (long py_this, fop_mkdir_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_MKDIR] = (long)fop;
+}
+
+void
+set_mkdir_cbk (long py_this, fop_mkdir_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_MKDIR] = (long)cbk;
+}
+
+
+/* FOP: RMDIR */
+
+int32_t
+glupy_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno,
+                 struct iatt *preparent, struct iatt *postparent,
+                 dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_RMDIR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_rmdir_cbk_t)(priv->cbks[GLUPY_RMDIR]))(
+                frame, cookie, this, op_ret, op_errno,
+                preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno, preparent,
+                             postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             int xflags, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_RMDIR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_rmdir_t)(priv->fops[GLUPY_RMDIR]))(
+                frame, this, loc, xflags, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_rmdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rmdir, loc,
+                    xflags, xdata);
+        return 0;
+}
+
+void
+wind_rmdir (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+            int xflags, dict_t *xdata)
+{
+
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_rmdir_cbk, xl, xl->fops->rmdir,
+                    loc, xflags, xdata);
+}
+
+void
+unwind_rmdir (call_frame_t *frame, long cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno,
+              struct iatt *preparent, struct iatt *postparent,
+              dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno,
+                             preparent, postparent, xdata);
+}
+
+void
+set_rmdir_fop (long py_this, fop_rmdir_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_RMDIR] = (long)fop;
+}
+
+void
+set_rmdir_cbk (long py_this, fop_rmdir_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_RMDIR] = (long)cbk;
+}
+
+
+/* NON-FOP-SPECIFIC CODE */
+
+
+long
+get_id (call_frame_t *frame)
+{
+        return (long)(frame->local);
+}
+
+uint64_t
+get_rootunique (call_frame_t *frame)
+{
+        return frame->root->unique;
+}
+
+int32_t
+init (xlator_t *this)
+{
+        glupy_private_t         *priv           = NULL;
+        char                    *module_name    = NULL;
+        PyObject                *py_mod_name    = NULL;
+        PyObject                *py_init_func   = NULL;
+        PyObject                *py_args        = NULL;
+        PyObject                *syspath        = NULL;
+	PyObject                *path           = NULL;
+	static gf_boolean_t      py_inited      = _gf_false;
+        void *                   err_cleanup    = &&err_return;
+
+        if (dict_get_str(this->options,"module-name",&module_name) != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "missing module-name");
+                return -1;
+        }
+
+	priv = GF_CALLOC (1, sizeof (glupy_private_t), gf_glupy_mt_priv);
+        if (!priv) {
+                goto *err_cleanup;
+        }
+        this->private = priv;
+        err_cleanup = &&err_free_priv;
+
+	if (!py_inited) {
+                Py_Initialize();
+                PyEval_InitThreads();
+#if 0
+                (void)pthread_key_create(&gil_init_key,NULL);
+                (void)pthread_setspecific(gil_init_key,(void *)1);
+#endif
+                /* PyEval_InitThreads takes this "for" us.  No thanks. */
+                PyEval_ReleaseLock();
+                py_inited = _gf_true;
+        }
+
+	/* Adjust python's path */
+	syspath = PySys_GetObject("path");
+	path = PyString_FromString(GLUSTER_PYTHON_PATH);
+	PyList_Append(syspath, path);
+	Py_DECREF(path);
+
+	py_mod_name = PyString_FromString(module_name);
+        if (!py_mod_name) {
+                gf_log (this->name, GF_LOG_ERROR, "could not create name");
+                if (PyErr_Occurred()) {
+                        PyErr_Print();
+                }
+                goto *err_cleanup;
+        }
+
+        gf_log (this->name, GF_LOG_ERROR, "py_mod_name = %s", module_name);
+        priv->py_module = PyImport_Import(py_mod_name);
+        Py_DECREF(py_mod_name);
+        if (!priv->py_module) {
+                gf_log (this->name, GF_LOG_ERROR, "Python import failed");
+                if (PyErr_Occurred()) {
+                        PyErr_Print();
+                }
+                goto *err_cleanup;
+        }
+        err_cleanup = &&err_deref_module;
+
+        py_init_func = PyObject_GetAttrString(priv->py_module, "xlator");
+        if (!py_init_func || !PyCallable_Check(py_init_func)) {
+                gf_log (this->name, GF_LOG_ERROR, "missing init func");
+                if (PyErr_Occurred()) {
+                        PyErr_Print();
+                }
+                goto *err_cleanup;
+        }
+        err_cleanup = &&err_deref_init;
+
+        py_args = PyTuple_New(1);
+        if (!py_args) {
+                gf_log (this->name, GF_LOG_ERROR, "could not create args");
+                if (PyErr_Occurred()) {
+                        PyErr_Print();
+                }
+                goto *err_cleanup;
+        }
+        PyTuple_SetItem(py_args,0,PyLong_FromLong((long)this));
+
+        /* TBD: pass in list of children */
+        priv->py_xlator = PyObject_CallObject(py_init_func, py_args);
+        Py_DECREF(py_args);
+        if (!priv->py_xlator) {
+                gf_log (this->name, GF_LOG_ERROR, "Python init failed");
+                if (PyErr_Occurred()) {
+                        PyErr_Print();
+                }
+                goto *err_cleanup;
+        }
+        gf_log (this->name, GF_LOG_INFO, "init returned %p", priv->py_xlator);
+
+        return 0;
+
+err_deref_init:
+        Py_DECREF(py_init_func);
+err_deref_module:
+        Py_DECREF(priv->py_module);
+err_free_priv:
+        GF_FREE(priv);
+err_return:
+        return -1;
+}
+
+void
+fini (xlator_t *this)
+{
+        glupy_private_t *priv = this->private;
+
+        if (!priv)
+                return;
+        Py_DECREF(priv->py_xlator);
+        Py_DECREF(priv->py_module);
+        this->private = NULL;
+        GF_FREE (priv);
+
+        return;
+}
+
+struct xlator_fops fops = {
+        .lookup       = glupy_lookup,
+        .create       = glupy_create,
+        .open         = glupy_open,
+        .readv        = glupy_readv,
+        .writev       = glupy_writev,
+        .opendir      = glupy_opendir,
+        .readdir      = glupy_readdir,
+        .stat         = glupy_stat,
+        .fstat        = glupy_fstat,
+        .setxattr     = glupy_setxattr,
+        .getxattr     = glupy_getxattr,
+        .fsetxattr    = glupy_fsetxattr,
+        .fgetxattr    = glupy_fgetxattr,
+        .removexattr  = glupy_removexattr,
+        .fremovexattr = glupy_fremovexattr,
+        .link         = glupy_link,
+        .unlink       = glupy_unlink,
+        .readlink     = glupy_readlink,
+        .symlink      = glupy_symlink,
+        .mkdir        = glupy_mkdir,
+        .rmdir        = glupy_rmdir,
+        .statfs       = glupy_statfs,
+        .readdirp     = glupy_readdirp
+};
+
+struct xlator_cbks cbks = {
+};
+
+struct volume_options options[] = {
+        { .key  = {NULL} },
+};
diff --git a/xlators/features/glupy/src/glupy.h b/xlators/features/glupy/src/glupy.h
new file mode 100644
index 000000000..8661fce88
--- /dev/null
+++ b/xlators/features/glupy/src/glupy.h
@@ -0,0 +1,69 @@
+/*
+   Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com>
+   This file is part of GlusterFS.
+
+   GlusterFS is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published
+   by the Free Software Foundation; either version 3 of the License,
+   or (at your option) any later version.
+
+   GlusterFS is distributed in the hope that it will be useful, but
+   WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see
+   <http://www.gnu.org/licenses/>.
+*/
+
+#ifndef __GLUPY_H__
+#define __GLUPY_H__
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+#include "mem-types.h"
+
+enum {
+        GLUPY_LOOKUP = 0,
+        GLUPY_CREATE,
+        GLUPY_OPEN,
+        GLUPY_READV,
+        GLUPY_WRITEV,
+        GLUPY_OPENDIR,
+        GLUPY_READDIR,
+        GLUPY_READDIRP,
+        GLUPY_STAT,
+        GLUPY_FSTAT,
+        GLUPY_STATFS,
+        GLUPY_SETXATTR,
+        GLUPY_GETXATTR,
+        GLUPY_FSETXATTR,
+        GLUPY_FGETXATTR,
+        GLUPY_REMOVEXATTR,
+        GLUPY_FREMOVEXATTR,
+        GLUPY_LINK,
+        GLUPY_UNLINK,
+        GLUPY_READLINK,
+        GLUPY_SYMLINK,
+        GLUPY_MKNOD,
+        GLUPY_MKDIR,
+        GLUPY_RMDIR,
+        GLUPY_N_FUNCS
+};
+
+typedef struct {
+        PyObject        *py_module;
+        PyObject        *py_xlator;
+        long            fops[GLUPY_N_FUNCS];
+        long            cbks[GLUPY_N_FUNCS];
+} glupy_private_t;
+
+enum gf_glupy_mem_types_ {
+        gf_glupy_mt_priv = gf_common_mt_end + 1,
+        gf_glupy_mt_end
+};
+
+#endif /* __GLUPY_H__ */
diff --git a/xlators/features/glupy/src/gluster.py b/xlators/features/glupy/src/gluster.py
new file mode 100644
index 000000000..a5daa77d3
--- /dev/null
+++ b/xlators/features/glupy/src/gluster.py
@@ -0,0 +1,841 @@
+import sys
+from ctypes import *
+
+dl = CDLL("",RTLD_GLOBAL)
+
+
+class call_frame_t (Structure):
+        pass
+
+class dev_t (Structure):
+        pass
+
+
+class dict_t (Structure):
+        pass
+
+
+class gf_dirent_t (Structure):
+        pass
+
+
+class iobref_t (Structure):
+	pass
+
+
+class iovec_t (Structure):
+        pass
+
+
+class list_head (Structure):
+        pass
+
+list_head._fields_ = [
+                ("next", POINTER(list_head)),
+                ("prev", POINTER(list_head))
+        ]
+
+
+class rwxperm_t (Structure):
+        _fields_ = [
+                ("read", c_uint8, 1),
+                ("write", c_uint8, 1),
+                ("execn", c_uint8, 1)
+        ]
+
+
+class statvfs_t (Structure):
+        pass
+
+
+class xlator_t (Structure):
+        pass
+
+
+class ia_prot_t (Structure):
+        _fields_ = [
+                ("suid", c_uint8, 1),
+                ("sgid", c_uint8, 1),
+                ("sticky", c_uint8, 1),
+                ("owner", rwxperm_t),
+                ("group", rwxperm_t),
+                ("other", rwxperm_t)
+        ]
+
+# For checking file type.
+(IA_INVAL, IA_IFREG, IA_IFDIR, IA_IFLNK, IA_IFBLK, IA_IFCHR, IA_IFIFO,
+ IA_IFSOCK) = xrange(8)
+
+
+class iatt_t (Structure):
+        _fields_ = [
+                ("ia_no", c_uint64),
+                ("ia_gfid", c_ubyte * 16),
+                ("ia_dev", c_uint64),
+                ("ia_type", c_uint),
+                ("ia_prot", ia_prot_t),
+                ("ia_nlink", c_uint32),
+                ("ia_uid", c_uint32),
+                ("ia_gid", c_uint32),
+                ("ia_rdev", c_uint64),
+                ("ia_size", c_uint64),
+                ("ia_blksize", c_uint32),
+                ("ia_blocks", c_uint64),
+                ("ia_atime", c_uint32 ),
+                ("ia_atime_nsec", c_uint32),
+                ("ia_mtime", c_uint32),
+                ("ia_mtime_nsec", c_uint32),
+                ("ia_ctime", c_uint32),
+                ("ia_ctime_nsec", c_uint32)
+        ]
+
+
+class mem_pool (Structure):
+        _fields_ = [
+                ("list", list_head),
+                ("hot_count", c_int),
+                ("cold_count", c_int),
+                ("lock", c_void_p),
+                ("padded_sizeof_type", c_ulong),
+                ("pool", c_void_p),
+                ("pool_end", c_void_p),
+                ("real_sizeof_type", c_int),
+                ("alloc_count", c_uint64),
+                ("pool_misses", c_uint64),
+                ("max_alloc", c_int),
+                ("curr_stdalloc", c_int),
+                ("max_stdalloc", c_int),
+                ("name", c_char_p),
+                ("global_list", list_head)
+        ]
+
+
+class U_ctx_key_inode (Union):
+        _fields_ = [
+                ("key", c_uint64),
+                ("xl_key", POINTER(xlator_t))
+        ]
+
+
+class U_ctx_value1 (Union):
+        _fields_ = [
+                ("value1", c_uint64),
+                ("ptr1", c_void_p)
+        ]
+
+
+class U_ctx_value2 (Union):
+        _fields_ = [
+                ("value2", c_uint64),
+                ("ptr2", c_void_p)
+        ]
+
+class inode_ctx (Structure):
+        _anonymous_ = ("u_key","u_value1","u_value2",)
+        _fields_ = [
+                ("u_key", U_ctx_key_inode),
+                ("u_value1", U_ctx_value1),
+                ("u_value2", U_ctx_value2)
+        ]
+
+class inode_t (Structure):
+        pass
+
+class inode_table_t (Structure):
+        _fields_ = [
+                ("lock", c_void_p),
+                ("hashsize", c_size_t),
+                ("name", c_char_p),
+                ("root", POINTER(inode_t)),
+                ("xl", POINTER(xlator_t)),
+                ("lru_limit", c_uint32),
+                ("inode_hash", POINTER(list_head)),
+                ("name_hash", POINTER(list_head)),
+                ("active", list_head),
+                ("active_size", c_uint32),
+                ("lru", list_head),
+                ("lru_size", c_uint32),
+                ("purge", list_head),
+                ("purge_size", c_uint32),
+                ("inode_pool", POINTER(mem_pool)),
+                ("dentry_pool", POINTER(mem_pool)),
+                ("fd_mem_pool", POINTER(mem_pool))
+        ]
+
+inode_t._fields_ = [
+                ("table", POINTER(inode_table_t)),
+                ("gfid", c_ubyte * 16),
+                ("lock", c_void_p),
+                ("nlookup", c_uint64),
+                ("fd_count", c_uint32),
+                ("ref", c_uint32),
+                ("ia_type", c_uint),
+                ("fd_list", list_head),
+                ("dentry_list", list_head),
+                ("hashv", list_head),
+                ("listv", list_head),
+                ("ctx", POINTER(inode_ctx))
+        ]
+
+
+
+class U_ctx_key_fd (Union):
+        _fields_ = [
+                ("key", c_uint64),
+                ("xl_key", c_void_p)
+        ]
+
+class fd_lk_ctx (Structure):
+        _fields_ = [
+                ("lk_list", list_head),
+                ("ref", c_int),
+                ("lock", c_void_p)
+        ]
+
+class fd_ctx (Structure):
+        _anonymous_ = ("u_key","u_value1")
+        _fields_ = [
+                ("u_key", U_ctx_key_fd),
+                ("u_value1", U_ctx_value1)
+        ]
+
+class fd_t (Structure):
+        _fields_ = [
+                ("pid", c_uint64),
+                ("flags", c_int32),
+                ("refcount", c_int32),
+                ("inode_list", list_head),
+                ("inode", POINTER(inode_t)),
+                ("lock", c_void_p),
+                ("ctx", POINTER(fd_ctx)),
+                ("xl_count", c_int),
+                ("lk_ctx", POINTER(fd_lk_ctx)),
+                ("anonymous", c_uint)
+        ]
+
+class loc_t (Structure):
+        _fields_ = [
+                ("path", c_char_p),
+                ("name", c_char_p),
+                ("inode", POINTER(inode_t)),
+                ("parent", POINTER(inode_t)),
+                ("gfid", c_ubyte * 16),
+                ("pargfid", c_ubyte * 16),
+        ]
+
+
+
+def _init_op (a_class, fop, cbk, wind, unwind):
+        # Decorators, used by translators. We could pass the signatures as
+        # parameters, but it's actually kind of nice to keep them around for
+        # inspection.
+        a_class.fop_type = apply(CFUNCTYPE,a_class.fop_sig)
+        a_class.cbk_type = apply(CFUNCTYPE,a_class.cbk_sig)
+        # Dispatch-function registration.
+        fop.restype = None
+        fop.argtypes = [ c_long, a_class.fop_type ]
+        # Callback-function registration.
+        cbk.restype = None
+        cbk.argtypes = [ c_long, a_class.cbk_type ]
+        # STACK_WIND function.
+        wind.restype = None
+        wind.argtypes = list(a_class.fop_sig[1:])
+        # STACK_UNWIND function.
+        unwind.restype = None
+        unwind.argtypes = list(a_class.cbk_sig[1:])
+
+class OpLookup:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
+                   POINTER(dict_t), POINTER(iatt_t))
+_init_op (OpLookup, dl.set_lookup_fop, dl.set_lookup_cbk,
+                    dl.wind_lookup,    dl.unwind_lookup)
+
+class OpCreate:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_int, c_uint, c_uint, POINTER(fd_t),
+                   POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(fd_t), POINTER(inode_t),
+                   POINTER(iatt_t), POINTER(iatt_t), POINTER(iatt_t),
+                   POINTER(dict_t))
+_init_op (OpCreate, dl.set_create_fop, dl.set_create_cbk,
+                    dl.wind_create,    dl.unwind_create)
+
+class OpOpen:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_int, POINTER(fd_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(fd_t), POINTER(dict_t))
+_init_op (OpOpen, dl.set_open_fop, dl.set_open_cbk,
+                  dl.wind_open,    dl.unwind_open)
+
+class OpReadv:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), c_size_t, c_long, c_uint32, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iovec_t), c_int, POINTER(iatt_t),
+                   POINTER(iobref_t), POINTER(dict_t))
+_init_op (OpReadv, dl.set_readv_fop, dl.set_readv_cbk,
+                   dl.wind_readv,    dl.unwind_readv)
+class OpWritev:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), POINTER(iovec_t), c_int, c_long, c_uint32,
+                   POINTER(iobref_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iatt_t), POINTER(iatt_t),
+                   POINTER(dict_t))
+_init_op (OpWritev, dl.set_writev_fop, dl.set_writev_cbk,
+                    dl.wind_writev,    dl.unwind_writev)
+
+class OpOpendir:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(fd_t) ,POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(fd_t), POINTER(dict_t))
+_init_op (OpOpendir, dl.set_opendir_fop, dl.set_opendir_cbk,
+                     dl.wind_opendir,    dl.unwind_opendir)
+
+class OpReaddir:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), c_size_t, c_long, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(gf_dirent_t), POINTER(dict_t))
+_init_op (OpReaddir, dl.set_readdir_fop, dl.set_readdir_cbk,
+                     dl.wind_readdir,    dl.unwind_readdir)
+
+class OpReaddirp:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), c_size_t, c_long, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(gf_dirent_t), POINTER(dict_t))
+_init_op (OpReaddirp, dl.set_readdirp_fop, dl.set_readdirp_cbk,
+                      dl.wind_readdirp,    dl.unwind_readdirp)
+
+class OpStat:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpStat, dl.set_stat_fop, dl.set_stat_cbk,
+                  dl.wind_stat,    dl.unwind_stat)
+
+class OpFstat:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpFstat, dl.set_fstat_fop, dl.set_fstat_cbk,
+                   dl.wind_fstat,    dl.unwind_fstat)
+
+class OpStatfs:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(statvfs_t), POINTER(dict_t))
+_init_op (OpStatfs, dl.set_statfs_fop, dl.set_statfs_cbk,
+                    dl.wind_statfs,    dl.unwind_statfs)
+
+
+class OpSetxattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(dict_t), c_int32,
+                   POINTER (dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t))
+_init_op (OpSetxattr, dl.set_setxattr_fop, dl.set_setxattr_cbk,
+                      dl.wind_setxattr,    dl.unwind_setxattr)
+
+class OpGetxattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_char_p, POINTER (dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t), POINTER(dict_t))
+_init_op (OpGetxattr, dl.set_getxattr_fop, dl.set_getxattr_cbk,
+                      dl.wind_getxattr,    dl.unwind_getxattr)
+
+class OpFsetxattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), POINTER(dict_t), c_int32,
+                   POINTER (dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t))
+_init_op (OpFsetxattr, dl.set_fsetxattr_fop, dl.set_fsetxattr_cbk,
+                       dl.wind_fsetxattr,    dl.unwind_fsetxattr)
+
+class OpFgetxattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), c_char_p, POINTER (dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t), POINTER(dict_t))
+_init_op (OpFgetxattr, dl.set_fgetxattr_fop, dl.set_fgetxattr_cbk,
+                       dl.wind_fgetxattr,    dl.unwind_fgetxattr)
+
+class OpRemovexattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_char_p, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t))
+_init_op (OpRemovexattr, dl.set_removexattr_fop, dl.set_removexattr_cbk,
+                         dl.wind_removexattr,    dl.unwind_removexattr)
+
+
+class OpFremovexattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), c_char_p, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t))
+_init_op (OpFremovexattr, dl.set_fremovexattr_fop, dl.set_fremovexattr_cbk,
+                          dl.wind_fremovexattr,    dl.unwind_fremovexattr)
+
+class OpLink:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(loc_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
+                   POINTER(iatt_t), POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpLink, dl.set_link_fop, dl.set_link_cbk,
+                  dl.wind_link,    dl.unwind_link)
+
+class OpSymlink:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   c_char_p, POINTER(loc_t), c_uint, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
+                   POINTER(iatt_t), POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpSymlink, dl.set_symlink_fop, dl.set_symlink_cbk,
+                     dl.wind_symlink,    dl.unwind_symlink)
+
+class OpUnlink:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_int, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iatt_t), POINTER(iatt_t),
+                   POINTER(dict_t))
+_init_op (OpUnlink, dl.set_unlink_fop, dl.set_unlink_cbk,
+                    dl.wind_unlink,    dl.unwind_unlink)
+
+class OpReadlink:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_size_t, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, c_char_p, POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpReadlink, dl.set_readlink_fop, dl.set_readlink_cbk,
+                      dl.wind_readlink,    dl.unwind_readlink)
+
+class OpMkdir:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_uint, c_uint, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
+                   POINTER(iatt_t), POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpMkdir, dl.set_mkdir_fop, dl.set_mkdir_cbk,
+                   dl.wind_mkdir,    dl.unwind_mkdir)
+
+class OpRmdir:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_int, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iatt_t), POINTER(iatt_t),
+                   POINTER(dict_t))
+_init_op (OpRmdir, dl.set_rmdir_fop, dl.set_rmdir_cbk,
+                   dl.wind_rmdir,    dl.unwind_rmdir)
+
+
+class Translator:
+        def __init__ (self, c_this):
+                # This is only here to keep references to the stubs we create,
+                # because ctypes doesn't and glupy.so can't because it doesn't
+                # get a pointer to the actual Python object. It's a dictionary
+                # instead of a list in case we ever allow changing fops/cbks
+                # after initialization and need to look them up.
+                self.stub_refs = {}
+                funcs = dir(self.__class__)
+                if "lookup_fop" in funcs:
+                        @OpLookup.fop_type
+                        def stub (frame, this, loc, xdata, s=self):
+                                return s.lookup_fop (frame, this, loc, xdata)
+                        self.stub_refs["lookup_fop"] = stub
+                        dl.set_lookup_fop(c_this,stub)
+                if "lookup_cbk" in funcs:
+                        @OpLookup.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, inode,
+                                  buf, xdata, postparent, s=self):
+                                return s.lookup_cbk(frame, cookie, this, op_ret,
+                                                    op_errno, inode, buf, xdata,
+                                                    postparent)
+                        self.stub_refs["lookup_cbk"] = stub
+                        dl.set_lookup_cbk(c_this,stub)
+                if "create_fop" in funcs:
+                        @OpCreate.fop_type
+                        def stub (frame, this, loc, flags, mode, umask, fd,
+                                  xdata, s=self):
+                                return s.create_fop (frame, this, loc, flags,
+                                                     mode, umask, fd, xdata)
+                        self.stub_refs["create_fop"] = stub
+                        dl.set_create_fop(c_this,stub)
+                if "create_cbk" in funcs:
+                        @OpCreate.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, fd,
+                                  inode, buf, preparent, postparent, xdata,
+                                  s=self):
+                                return s.create_cbk (frame, cookie, this,
+                                                     op_ret, op_errno, fd,
+                                                     inode, buf, preparent,
+                                                     postparent, xdata)
+                        self.stub_refs["create_cbk"] = stub
+                        dl.set_create_cbk(c_this,stub)
+                if "open_fop" in funcs:
+                        @OpOpen.fop_type
+                        def stub (frame, this, loc, flags, fd,
+                                  xdata, s=self):
+                                return s.open_fop (frame, this, loc, flags,
+                                                   fd, xdata)
+                        self.stub_refs["open_fop"] = stub
+                        dl.set_open_fop(c_this,stub)
+                if "open_cbk" in funcs:
+                        @OpOpen.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, fd,
+                                  xdata, s=self):
+                                return s.open_cbk (frame, cookie, this,
+                                                   op_ret, op_errno, fd,
+                                                   xdata)
+                        self.stub_refs["open_cbk"] = stub
+                        dl.set_open_cbk(c_this,stub)
+                if "readv_fop" in funcs:
+                        @OpReadv.fop_type
+                        def stub (frame, this, fd, size, offset, flags,
+                                  xdata, s=self):
+                                return s.readv_fop (frame, this, fd, size,
+                                                    offset, flags, xdata)
+                        self.stub_refs["readv_fop"] = stub
+                        dl.set_readv_fop(c_this,stub)
+                if "readv_cbk" in funcs:
+                        @OpReadv.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  vector, count, stbuf, iobref, xdata,
+                                  s=self):
+                                return s.readv_cbk (frame, cookie, this,
+                                                    op_ret, op_errno, vector,
+                                                    count, stbuf, iobref,
+                                                    xdata)
+                        self.stub_refs["readv_cbk"] = stub
+                        dl.set_readv_cbk(c_this,stub)
+                if "writev_fop" in funcs:
+                        @OpWritev.fop_type
+                        def stub (frame, this, fd, vector, count,
+                                  offset, flags, iobref, xdata, s=self):
+                                return s.writev_fop (frame, this, fd, vector,
+                                                     count, offset, flags,
+                                                     iobref, xdata)
+                        self.stub_refs["writev_fop"] = stub
+                        dl.set_writev_fop(c_this,stub)
+                if "writev_cbk" in funcs:
+                        @OpWritev.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  prebuf, postbuf, xdata, s=self):
+                                return s.writev_cbk (frame, cookie, this,
+                                                     op_ret, op_errno, prebuf,
+                                                     postbuf, xdata)
+                        self.stub_refs["writev_cbk"] = stub
+                        dl.set_writev_cbk(c_this,stub)
+                if "opendir_fop" in funcs:
+                        @OpOpendir.fop_type
+                        def stub (frame, this, loc, fd, xdata, s=self):
+                                return s.opendir_fop (frame, this, loc, fd,
+                                                      xdata)
+                        self.stub_refs["opendir_fop"] = stub
+                        dl.set_opendir_fop(c_this,stub)
+                if "opendir_cbk" in funcs:
+                        @OpOpendir.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, fd,
+                                  xdata, s=self):
+                                return s.opendir_cbk(frame, cookie, this,
+                                                     op_ret, op_errno, fd,
+                                                     xdata)
+                        self.stub_refs["opendir_cbk"] = stub
+                        dl.set_opendir_cbk(c_this,stub)
+                if "readdir_fop" in funcs:
+                        @OpReaddir.fop_type
+                        def stub (frame, this, fd, size, offset, xdata, s=self):
+                                return s.readdir_fop (frame, this, fd, size,
+                                                      offset, xdata)
+                        self.stub_refs["readdir_fop"] = stub
+                        dl.set_readdir_fop(c_this,stub)
+                if "readdir_cbk" in funcs:
+                        @OpReaddir.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  entries, xdata, s=self):
+                                return s.readdir_cbk(frame, cookie, this,
+                                                     op_ret, op_errno, entries,
+                                                     xdata)
+                        self.stub_refs["readdir_cbk"] = stub
+                        dl.set_readdir_cbk(c_this,stub)
+                if "readdirp_fop" in funcs:
+                        @OpReaddirp.fop_type
+                        def stub (frame, this, fd, size, offset, xdata, s=self):
+                                return s.readdirp_fop (frame, this, fd, size,
+                                                       offset, xdata)
+                        self.stub_refs["readdirp_fop"] = stub
+                        dl.set_readdirp_fop(c_this,stub)
+                if "readdirp_cbk" in funcs:
+                        @OpReaddirp.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  entries, xdata, s=self):
+                                return s.readdirp_cbk (frame, cookie, this,
+                                                       op_ret, op_errno,
+                                                       entries, xdata)
+                        self.stub_refs["readdirp_cbk"] = stub
+                        dl.set_readdirp_cbk(c_this,stub)
+                if "stat_fop" in funcs:
+                        @OpStat.fop_type
+                        def stub (frame, this, loc, xdata, s=self):
+                                return s.stat_fop (frame, this, loc, xdata)
+                        self.stub_refs["stat_fop"] = stub
+                        dl.set_stat_fop(c_this,stub)
+                if "stat_cbk" in funcs:
+                        @OpStat.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, buf,
+                                  xdata, s=self):
+                                return s.stat_cbk(frame, cookie, this, op_ret,
+                                                  op_errno, buf, xdata)
+                        self.stub_refs["stat_cbk"] = stub
+                        dl.set_stat_cbk(c_this,stub)
+                if "fstat_fop" in funcs:
+                        @OpFstat.fop_type
+                        def stub (frame, this, fd, xdata, s=self):
+                                return s.fstat_fop (frame, this, fd, xdata)
+                        self.stub_refs["fstat_fop"] = stub
+                        dl.set_fstat_fop(c_this,stub)
+                if "fstat_cbk" in funcs:
+                        @OpFstat.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, buf,
+                                  xdata, s=self):
+                                return s.fstat_cbk(frame, cookie, this, op_ret,
+                                                   op_errno, buf, xdata)
+                        self.stub_refs["fstat_cbk"] = stub
+                        dl.set_fstat_cbk(c_this,stub)
+                if "statfs_fop" in funcs:
+                        @OpStatfs.fop_type
+                        def stub (frame, this, loc, xdata, s=self):
+                                return s.statfs_fop (frame, this, loc, xdata)
+                        self.stub_refs["statfs_fop"] = stub
+                        dl.set_statfs_fop(c_this,stub)
+                if "statfs_cbk" in funcs:
+                        @OpStatfs.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, buf,
+                                  xdata, s=self):
+                                return s.statfs_cbk (frame, cookie, this,
+                                                     op_ret, op_errno, buf,
+                                                     xdata)
+                        self.stub_refs["statfs_cbk"] = stub
+                        dl.set_statfs_cbk(c_this,stub)
+                if "setxattr_fop" in funcs:
+                        @OpSetxattr.fop_type
+                        def stub (frame, this, loc, dictionary, flags, xdata,
+                                  s=self):
+                                return s.setxattr_fop (frame, this, loc,
+                                                       dictionary, flags,
+                                                       xdata)
+                        self.stub_refs["setxattr_fop"] = stub
+                        dl.set_setxattr_fop(c_this,stub)
+                if "setxattr_cbk" in funcs:
+                        @OpSetxattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, xdata,
+                                  s=self):
+                                return s.setxattr_cbk(frame, cookie, this,
+                                                      op_ret, op_errno, xdata)
+                        self.stub_refs["setxattr_cbk"] = stub
+                        dl.set_setxattr_cbk(c_this,stub)
+                if "getxattr_fop" in funcs:
+                        @OpGetxattr.fop_type
+                        def stub (frame, this, loc, name, xdata, s=self):
+                                return s.getxattr_fop (frame, this, loc, name,
+                                                       xdata)
+                        self.stub_refs["getxattr_fop"] = stub
+                        dl.set_getxattr_fop(c_this,stub)
+                if "getxattr_cbk" in funcs:
+                        @OpGetxattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  dictionary, xdata, s=self):
+                                return s.getxattr_cbk(frame, cookie, this,
+                                                      op_ret, op_errno,
+                                                      dictionary, xdata)
+                        self.stub_refs["getxattr_cbk"] = stub
+                        dl.set_getxattr_cbk(c_this,stub)
+                if "fsetxattr_fop" in funcs:
+                        @OpFsetxattr.fop_type
+                        def stub (frame, this, fd, dictionary, flags, xdata,
+                                  s=self):
+                                return s.fsetxattr_fop (frame, this, fd,
+                                                        dictionary, flags,
+                                                        xdata)
+                        self.stub_refs["fsetxattr_fop"] = stub
+                        dl.set_fsetxattr_fop(c_this,stub)
+                if "fsetxattr_cbk" in funcs:
+                        @OpFsetxattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, xdata,
+                                  s=self):
+                                return s.fsetxattr_cbk(frame, cookie, this,
+                                                       op_ret, op_errno, xdata)
+                        self.stub_refs["fsetxattr_cbk"] = stub
+                        dl.set_fsetxattr_cbk(c_this,stub)
+                if "fgetxattr_fop" in funcs:
+                        @OpFgetxattr.fop_type
+                        def stub (frame, this, fd, name, xdata, s=self):
+                                return s.fgetxattr_fop (frame, this, fd, name,
+                                                        xdata)
+                        self.stub_refs["fgetxattr_fop"] = stub
+                        dl.set_fgetxattr_fop(c_this,stub)
+                if "fgetxattr_cbk" in funcs:
+                        @OpFgetxattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  dictionary, xdata, s=self):
+                                return s.fgetxattr_cbk(frame, cookie, this,
+                                                       op_ret, op_errno,
+                                                       dictionary, xdata)
+                        self.stub_refs["fgetxattr_cbk"] = stub
+                        dl.set_fgetxattr_cbk(c_this,stub)
+                if "removexattr_fop" in funcs:
+                        @OpRemovexattr.fop_type
+                        def stub (frame, this, loc, name, xdata, s=self):
+                                return s.removexattr_fop (frame, this, loc,
+                                                          name, xdata)
+                        self.stub_refs["removexattr_fop"] = stub
+                        dl.set_removexattr_fop(c_this,stub)
+                if "removexattr_cbk" in funcs:
+                        @OpRemovexattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  xdata, s=self):
+                                return s.removexattr_cbk(frame, cookie, this,
+                                                         op_ret, op_errno,
+                                                         xdata)
+                        self.stub_refs["removexattr_cbk"] = stub
+                        dl.set_removexattr_cbk(c_this,stub)
+                if "fremovexattr_fop" in funcs:
+                        @OpFremovexattr.fop_type
+                        def stub (frame, this, fd, name, xdata, s=self):
+                                return s.fremovexattr_fop (frame, this, fd,
+                                                           name, xdata)
+                        self.stub_refs["fremovexattr_fop"] = stub
+                        dl.set_fremovexattr_fop(c_this,stub)
+                if "fremovexattr_cbk" in funcs:
+                        @OpFremovexattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  xdata, s=self):
+                                return s.fremovexattr_cbk(frame, cookie, this,
+                                                          op_ret, op_errno,
+                                                          xdata)
+                        self.stub_refs["fremovexattr_cbk"] = stub
+                        dl.set_fremovexattr_cbk(c_this,stub)
+                if "link_fop" in funcs:
+                        @OpLink.fop_type
+                        def stub (frame, this, oldloc, newloc,
+                                  xdata, s=self):
+                                return s.link_fop (frame, this, oldloc,
+                                                   newloc, xdata)
+                        self.stub_refs["link_fop"] = stub
+                        dl.set_link_fop(c_this,stub)
+                if "link_cbk" in funcs:
+                        @OpLink.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  inode, buf, preparent, postparent, xdata,
+                                  s=self):
+                                return s.link_cbk (frame, cookie, this,
+                                                   op_ret, op_errno, inode,
+                                                   buf, preparent,
+                                                   postparent, xdata)
+                        self.stub_refs["link_cbk"] = stub
+                        dl.set_link_cbk(c_this,stub)
+                if "symlink_fop" in funcs:
+                        @OpSymlink.fop_type
+                        def stub (frame, this, linkname, loc,
+                                  umask, xdata, s=self):
+                                return s.symlink_fop (frame, this, linkname,
+                                                      loc, umask, xdata)
+                        self.stub_refs["symlink_fop"] = stub
+                        dl.set_symlink_fop(c_this,stub)
+                if "symlink_cbk" in funcs:
+                        @OpSymlink.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  inode, buf, preparent, postparent, xdata,
+                                  s=self):
+                                return s.symlink_cbk (frame, cookie, this,
+                                                      op_ret, op_errno, inode,
+                                                      buf, preparent,
+                                                      postparent, xdata)
+                        self.stub_refs["symlink_cbk"] = stub
+                        dl.set_symlink_cbk(c_this,stub)
+                if "unlink_fop" in funcs:
+                        @OpUnlink.fop_type
+                        def stub (frame, this, loc, xflags,
+                                  xdata, s=self):
+                                return s.unlink_fop (frame, this, loc,
+                                                     xflags, xdata)
+                        self.stub_refs["unlink_fop"] = stub
+                        dl.set_unlink_fop(c_this,stub)
+                if "unlink_cbk" in funcs:
+                        @OpUnlink.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  preparent, postparent, xdata, s=self):
+                                return s.unlink_cbk (frame, cookie, this,
+                                                     op_ret, op_errno,
+                                                     preparent, postparent,
+                                                     xdata)
+                        self.stub_refs["unlink_cbk"] = stub
+                        dl.set_unlink_cbk(c_this,stub)
+                if "readlink_fop" in funcs:
+                        @OpReadlink.fop_type
+                        def stub (frame, this, loc, size,
+                                  xdata, s=self):
+                                return s.readlink_fop (frame, this, loc,
+                                                       size, xdata)
+                        self.stub_refs["readlink_fop"] = stub
+                        dl.set_readlink_fop(c_this,stub)
+                if "readlink_cbk" in funcs:
+                        @OpReadlink.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  path, buf, xdata, s=self):
+                                return s.readlink_cbk (frame, cookie, this,
+                                                       op_ret, op_errno,
+                                                       path, buf, xdata)
+                        self.stub_refs["readlink_cbk"] = stub
+                        dl.set_readlink_cbk(c_this,stub)
+                if "mkdir_fop" in funcs:
+                        @OpMkdir.fop_type
+                        def stub (frame, this, loc, mode, umask, xdata,
+                                  s=self):
+                                return s.mkdir_fop (frame, this, loc, mode,
+                                                    umask, xdata)
+                        self.stub_refs["mkdir_fop"] = stub
+                        dl.set_mkdir_fop(c_this,stub)
+                if "mkdir_cbk" in funcs:
+                        @OpMkdir.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, inode,
+                                  buf, preparent, postparent, xdata, s=self):
+                                return s.mkdir_cbk (frame, cookie, this,
+                                                    op_ret, op_errno, inode,
+                                                    buf, preparent,
+                                                    postparent, xdata)
+                        self.stub_refs["mkdir_cbk"] = stub
+                        dl.set_mkdir_cbk(c_this,stub)
+                if "rmdir_fop" in funcs:
+                        @OpRmdir.fop_type
+                        def stub (frame, this, loc, xflags,
+                                  xdata, s=self):
+                                return s.rmdir_fop (frame, this, loc,
+                                                    xflags, xdata)
+                        self.stub_refs["rmdir_fop"] = stub
+                        dl.set_rmdir_fop(c_this,stub)
+                if "rmdir_cbk" in funcs:
+                        @OpRmdir.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  preparent, postparent, xdata, s=self):
+                                return s.rmdir_cbk (frame, cookie, this,
+                                                    op_ret, op_errno,
+                                                    preparent, postparent,
+                                                    xdata)
+                        self.stub_refs["rmdir_cbk"] = stub
+                        dl.set_rmdir_cbk(c_this,stub)
diff --git a/xlators/features/glupy/src/helloworld.py b/xlators/features/glupy/src/helloworld.py
new file mode 100644
index 000000000..8fe403711
--- /dev/null
+++ b/xlators/features/glupy/src/helloworld.py
@@ -0,0 +1,19 @@
+import sys
+from gluster import *
+
+class xlator (Translator):
+
+    def __init__(self, c_this):
+        Translator.__init__(self, c_this)
+
+    def lookup_fop(self, frame, this, loc, xdata):
+        print "Python xlator: Hello!"
+        dl.wind_lookup(frame, POINTER(xlator_t)(), loc, xdata)
+        return 0
+
+    def lookup_cbk(self, frame, cookie, this, op_ret, op_errno, inode, buf,
+                   xdata, postparent):
+        print "Python xlator: Hello again!"
+        dl.unwind_lookup(frame, cookie, this, op_ret, op_errno, inode, buf,
+                         xdata, postparent)
+        return 0
diff --git a/xlators/features/glupy/src/negative.py b/xlators/features/glupy/src/negative.py
new file mode 100644
index 000000000..1023602b9
--- /dev/null
+++ b/xlators/features/glupy/src/negative.py
@@ -0,0 +1,92 @@
+import sys
+from uuid import UUID
+from gluster import *
+
+# Negative-lookup-caching example.  If a file wasn't there the last time we
+# looked, it's probably still not there.  This translator keeps track of
+# those failed lookups for us, and returns ENOENT without needing to pass the
+# call any further for repeated requests.
+
+# If we were doing this for real, we'd need separate caches for each xlator
+# instance.  The easiest way to do this would be to have xlator.__init__
+# "register" each instance in a module-global dict, with the key as the C
+# translator address and the value as the xlator object itself.  For testing
+# and teaching, it's sufficient just to have one cache.  The keys are parent
+# GFIDs, and the entries are lists of names within that parent that we know
+# don't exist.
+cache = {}
+
+# TBD: we need a better way of handling per-request data (frame->local in C).
+dl.get_id.restype = c_long
+dl.get_id.argtypes = [ POINTER(call_frame_t) ]
+
+def uuid2str (gfid):
+    return str(UUID(''.join(map("{0:02x}".format, gfid))))
+
+class xlator (Translator):
+
+    def __init__ (self, c_this):
+        self.requests = {}
+        Translator.__init__(self,c_this)
+
+    def lookup_fop (self, frame, this, loc, xdata):
+        pargfid = uuid2str(loc.contents.pargfid)
+        print "lookup FOP: %s:%s" % (pargfid, loc.contents.name)
+        # Check the cache.
+        if cache.has_key(pargfid):
+            if loc.contents.name in cache[pargfid]:
+                print "short-circuiting for %s:%s" % (pargfid,
+                    loc.contents.name)
+                dl.unwind_lookup(frame,0,this,-1,2,None,None,None,None)
+                return 0
+        key = dl.get_id(frame)
+        self.requests[key] = (pargfid, loc.contents.name[:])
+        # TBD: get real child xl from init, pass it here
+        dl.wind_lookup(frame,POINTER(xlator_t)(),loc,xdata)
+        return 0
+
+    def lookup_cbk (self, frame, cookie, this, op_ret, op_errno, inode, buf,
+                    xdata, postparent):
+        print "lookup CBK: %d (%d)" % (op_ret, op_errno)
+        key = dl.get_id(frame)
+        pargfid, name = self.requests[key]
+        # Update the cache.
+        if op_ret == 0:
+            print "found %s, removing from cache" % name
+            if cache.has_key(pargfid):
+                cache[pargfid].discard(name)
+        elif op_errno == 2:    # ENOENT
+            print "failed to find %s, adding to cache" % name
+            if cache.has_key(pargfid):
+                cache[pargfid].add(name)
+            else:
+                cache[pargfid] = set([name])
+        del self.requests[key]
+        dl.unwind_lookup(frame,cookie,this,op_ret,op_errno,
+                         inode,buf,xdata,postparent)
+        return 0
+
+    def create_fop (self, frame, this, loc, flags, mode, umask, fd, xdata):
+        pargfid = uuid2str(loc.contents.pargfid)
+        print "create FOP: %s:%s" % (pargfid, loc.contents.name)
+        key = dl.get_id(frame)
+        self.requests[key] = (pargfid, loc.contents.name[:])
+        # TBD: get real child xl from init, pass it here
+        dl.wind_create(frame,POINTER(xlator_t)(),loc,flags,mode,umask,fd,xdata)
+        return 0
+
+    def create_cbk (self, frame, cookie, this, op_ret, op_errno, fd, inode,
+                    buf, preparent, postparent, xdata):
+        print "create CBK: %d (%d)" % (op_ret, op_errno)
+        key = dl.get_id(frame)
+        pargfid, name = self.requests[key]
+        # Update the cache.
+        if op_ret == 0:
+            print "created %s, removing from cache" % name
+            if cache.has_key(pargfid):
+                cache[pargfid].discard(name)
+        del self.requests[key]
+        dl.unwind_create(frame,cookie,this,op_ret,op_errno,fd,inode,buf,
+                         preparent,postparent,xdata)
+        return 0
+
diff --git a/xlators/storage/bdb/Makefile.am b/xlators/features/index/Makefile.am
index d471a3f92..a985f42a8 100644
--- a/xlators/storage/bdb/Makefile.am
+++ b/xlators/features/index/Makefile.am
@@ -1,3 +1,3 @@
 SUBDIRS = src
 
-CLEANFILES = 
+CLEANFILES =
diff --git a/xlators/features/index/src/Makefile.am b/xlators/features/index/src/Makefile.am
new file mode 100644
index 000000000..73bb8972e
--- /dev/null
+++ b/xlators/features/index/src/Makefile.am
@@ -0,0 +1,17 @@
+xlator_LTLIBRARIES = index.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+index_la_LDFLAGS = -module -avoid-version
+
+index_la_SOURCES = index.c
+index_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+noinst_HEADERS = index.h index-mem-types.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) \
+	-I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/rpc/xdr/src \
+	-I$(top_srcdir)/rpc/rpc-lib/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
diff --git a/xlators/features/index/src/index-mem-types.h b/xlators/features/index/src/index-mem-types.h
new file mode 100644
index 000000000..553d492df
--- /dev/null
+++ b/xlators/features/index/src/index-mem-types.h
@@ -0,0 +1,22 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __QUIESCE_MEM_TYPES_H__
+#define __QUIESCE_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_index_mem_types_ {
+        gf_index_mt_priv_t = gf_common_mt_end + 1,
+        gf_index_inode_ctx_t = gf_common_mt_end + 2,
+        gf_index_fd_ctx_t = gf_common_mt_end + 3,
+        gf_index_mt_end
+};
+#endif
diff --git a/xlators/features/index/src/index.c b/xlators/features/index/src/index.c
new file mode 100644
index 000000000..9253120f3
--- /dev/null
+++ b/xlators/features/index/src/index.c
@@ -0,0 +1,1489 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "index.h"
+#include "options.h"
+#include "glusterfs3-xdr.h"
+#include "syncop.h"
+
+#define XATTROP_SUBDIR "xattrop"
+#define BASE_INDICES_HOLDER_SUBDIR "base_indices_holder"
+
+call_stub_t *
+__index_dequeue (struct list_head *callstubs)
+{
+        call_stub_t *stub = NULL;
+
+        if (!list_empty (callstubs)) {
+                stub = list_entry (callstubs->next, call_stub_t, list);
+                list_del_init (&stub->list);
+        }
+
+        return stub;
+}
+
+inline static void
+__index_enqueue (struct list_head *callstubs, call_stub_t *stub)
+{
+        list_add_tail (&stub->list, callstubs);
+}
+
+static void
+worker_enqueue (xlator_t *this, call_stub_t *stub)
+{
+        index_priv_t    *priv = NULL;
+
+        priv = this->private;
+        pthread_mutex_lock (&priv->mutex);
+        {
+                __index_enqueue (&priv->callstubs, stub);
+                pthread_cond_signal (&priv->cond);
+        }
+        pthread_mutex_unlock (&priv->mutex);
+}
+
+void *
+index_worker (void *data)
+{
+        index_priv_t     *priv = NULL;
+        xlator_t         *this = NULL;
+        call_stub_t      *stub = NULL;
+        int               ret = 0;
+
+        THIS = data;
+        this = data;
+        priv = this->private;
+
+        for (;;) {
+                pthread_mutex_lock (&priv->mutex);
+                {
+                        while (list_empty (&priv->callstubs)) {
+                                ret = pthread_cond_wait (&priv->cond,
+                                                         &priv->mutex);
+                        }
+
+                        stub = __index_dequeue (&priv->callstubs);
+                }
+                pthread_mutex_unlock (&priv->mutex);
+
+                if (stub) /* guard against spurious wakeups */
+                        call_resume (stub);
+        }
+
+        return NULL;
+}
+int
+__index_inode_ctx_get (inode_t *inode, xlator_t *this, index_inode_ctx_t **ctx)
+{
+        int               ret = 0;
+        index_inode_ctx_t *ictx = NULL;
+        uint64_t          tmpctx = 0;
+
+        ret = __inode_ctx_get (inode, this, &tmpctx);
+        if (!ret) {
+                ictx = (index_inode_ctx_t*) (long) tmpctx;
+                goto out;
+        }
+        ictx = GF_CALLOC (1, sizeof (*ictx), gf_index_inode_ctx_t);
+        if (!ictx) {
+                ret = -1;
+                goto out;
+        }
+
+        INIT_LIST_HEAD (&ictx->callstubs);
+        ret = __inode_ctx_put (inode, this, (uint64_t)ictx);
+        if (ret) {
+                GF_FREE (ictx);
+                ictx = NULL;
+                goto out;
+        }
+out:
+        if (ictx)
+                *ctx = ictx;
+        return ret;
+}
+
+int
+index_inode_ctx_get (inode_t *inode, xlator_t *this, index_inode_ctx_t **ctx)
+{
+        int               ret = 0;
+
+        LOCK (&inode->lock);
+        {
+                ret = __index_inode_ctx_get (inode, this, ctx);
+        }
+        UNLOCK (&inode->lock);
+
+        return ret;
+}
+
+static void
+make_index_dir_path (char *base, const char *subdir,
+                     char *index_dir, size_t len)
+{
+        snprintf (index_dir, len, "%s/%s", base, subdir);
+}
+
+int
+index_dir_create (xlator_t *this, const char *subdir)
+{
+        int          ret = 0;
+        struct       stat st = {0};
+        char         fullpath[PATH_MAX] = {0};
+        char         path[PATH_MAX] = {0};
+        char         *dir = NULL;
+        index_priv_t *priv = NULL;
+        size_t       len = 0;
+        size_t       pathlen = 0;
+
+        priv = this->private;
+        make_index_dir_path (priv->index_basepath, subdir, fullpath,
+                             sizeof (fullpath));
+        ret = stat (fullpath, &st);
+        if (!ret) {
+                if (!S_ISDIR (st.st_mode))
+                        ret = -2;
+                goto out;
+        }
+
+        pathlen = strlen (fullpath);
+        if ((pathlen > 1) && fullpath[pathlen - 1] == '/')
+                fullpath[pathlen - 1] = '\0';
+        dir = strchr (fullpath, '/');
+        while (dir) {
+                dir = strchr (dir + 1, '/');
+                if (dir)
+                        len = pathlen - strlen (dir);
+                else
+                        len = pathlen;
+                strncpy (path, fullpath, len);
+                path[len] = '\0';
+                ret = mkdir (path, 0600);
+                if (ret && (errno != EEXIST))
+                        goto out;
+        }
+        ret = 0;
+out:
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "%s/%s: Failed to "
+                        "create (%s)", priv->index_basepath, subdir,
+                        strerror (errno));
+        } else if (ret == -2) {
+                gf_log (this->name, GF_LOG_ERROR, "%s/%s: Failed to create, "
+                        "path exists, not a directory ", priv->index_basepath,
+                        subdir);
+        }
+        return ret;
+}
+
+void
+index_get_index (index_priv_t *priv, uuid_t index)
+{
+        LOCK (&priv->lock);
+        {
+                uuid_copy (index, priv->index);
+        }
+        UNLOCK (&priv->lock);
+}
+
+void
+index_generate_index (index_priv_t *priv, uuid_t index)
+{
+        LOCK (&priv->lock);
+        {
+                //To prevent duplicate generates.
+                //This method fails if number of contending threads is greater
+                //than MAX_LINK count of the fs
+                if (!uuid_compare (priv->index, index))
+                        uuid_generate (priv->index);
+                uuid_copy (index, priv->index);
+        }
+        UNLOCK (&priv->lock);
+}
+
+static void
+make_index_path (char *base, const char *subdir, uuid_t index,
+                 char *index_path, size_t len)
+{
+        make_index_dir_path (base, subdir, index_path, len);
+        snprintf (index_path + strlen (index_path), len - strlen (index_path),
+                  "/%s-%s", subdir, uuid_utoa (index));
+}
+
+static void
+make_gfid_path (char *base, const char *subdir, uuid_t gfid,
+                char *gfid_path, size_t len)
+{
+        make_index_dir_path (base, subdir, gfid_path, len);
+        snprintf (gfid_path + strlen (gfid_path), len - strlen (gfid_path),
+                  "/%s", uuid_utoa (gfid));
+}
+
+static void
+make_file_path (char *base, const char *subdir, const char *filename,
+                char *file_path, size_t len)
+{
+        make_index_dir_path (base, subdir, file_path, len);
+        snprintf (file_path + strlen (file_path), len - strlen (file_path),
+                  "/%s", filename);
+}
+
+static void
+check_delete_stale_index_file (xlator_t *this, char *filename)
+{
+        int             ret = 0;
+        struct stat     st = {0};
+        struct stat     base_index_st = {0};
+        char            filepath[PATH_MAX] = {0};
+        char            filepath_under_base_indices_holder[PATH_MAX] = {0};
+        index_priv_t    *priv = NULL;
+
+        priv = this->private;
+        if (priv->to_be_healed_states != synced_state)
+                return;
+
+        make_file_path (priv->index_basepath, XATTROP_SUBDIR,
+                        filename, filepath, sizeof (filepath));
+
+        make_file_path (priv->index_basepath, BASE_INDICES_HOLDER_SUBDIR,
+                        filename, filepath_under_base_indices_holder,
+                        sizeof (filepath_under_base_indices_holder));
+
+
+        ret = stat (filepath_under_base_indices_holder, &base_index_st);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Base index is not created"
+                        "under index/base_indices_holder");
+                return;
+        }
+
+        ret = stat (filepath, &st);
+        if (!ret && st.st_nlink == 2) {
+                unlink (filepath);
+                unlink (filepath_under_base_indices_holder);
+        }
+}
+
+static int
+index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
+                    size_t size, gf_dirent_t *entries, readdir_directory type)
+{
+        off_t     in_case = -1;
+        size_t    filled = 0;
+        int       count = 0;
+        char      entrybuf[sizeof(struct dirent) + 256 + 8];
+        struct dirent  *entry          = NULL;
+        int32_t              this_size      = -1;
+        gf_dirent_t          *this_entry     = NULL;
+        xlator_t             *this = NULL;
+
+        this = THIS;
+        if (!off) {
+                rewinddir (dir);
+        } else {
+                seekdir (dir, off);
+        }
+
+        while (filled <= size) {
+                in_case = telldir (dir);
+
+                if (in_case == -1) {
+                        gf_log (THIS->name, GF_LOG_ERROR,
+                                "telldir failed on dir=%p: %s",
+                                dir, strerror (errno));
+                        goto out;
+                }
+
+                errno = 0;
+                entry = NULL;
+                readdir_r (dir, (struct dirent *)entrybuf, &entry);
+
+                if (!entry) {
+                        if (errno == EBADF) {
+                                gf_log (THIS->name, GF_LOG_WARNING,
+                                        "readdir failed on dir=%p: %s",
+                                        dir, strerror (errno));
+                                goto out;
+                        }
+                        break;
+                }
+
+                if (!strncmp (entry->d_name, XATTROP_SUBDIR"-",
+                              strlen (XATTROP_SUBDIR"-")) &&
+                              (type == INDEX_XATTROP)) {
+                        check_delete_stale_index_file (this, entry->d_name);
+                        continue;
+                }
+
+                this_size = max (sizeof (gf_dirent_t),
+                                 sizeof (gfs3_dirplist))
+                        + strlen (entry->d_name) + 1;
+
+                if (this_size + filled > size) {
+                        seekdir (dir, in_case);
+                        break;
+                }
+
+                this_entry = gf_dirent_for_name (entry->d_name);
+
+                if (!this_entry) {
+                        gf_log (THIS->name, GF_LOG_ERROR,
+                                "could not create gf_dirent for entry %s: (%s)",
+                                entry->d_name, strerror (errno));
+                        goto out;
+                }
+                this_entry->d_off = telldir (dir);
+                this_entry->d_ino = entry->d_ino;
+
+                list_add_tail (&this_entry->list, &entries->list);
+
+                filled += this_size;
+                count ++;
+        }
+
+        if ((!readdir (dir) && (errno == 0)))
+                /* Indicate EOF */
+                errno = ENOENT;
+out:
+        return count;
+}
+
+int
+sync_base_indices (void *index_priv)
+{
+        index_priv_t    *priv = NULL;
+        DIR             *dir_base_holder  = NULL;
+        DIR             *xattrop_dir = NULL;
+        struct dirent   *entry = NULL;
+        char            base_indices_holder[PATH_MAX] = {0};
+        char            xattrop_directory[PATH_MAX] = {0};
+        char            base_index_path[PATH_MAX] = {0};
+        char            xattrop_index_path[PATH_MAX] = {0};
+        int             ret = 0;
+
+        priv = index_priv;
+
+        snprintf (base_indices_holder, PATH_MAX, "%s/%s", priv->index_basepath,
+                  BASE_INDICES_HOLDER_SUBDIR);
+        snprintf (xattrop_directory, PATH_MAX, "%s/%s", priv->index_basepath,
+                  XATTROP_SUBDIR);
+
+        if ((dir_base_holder = opendir(base_indices_holder)) == NULL) {
+                ret = -1;
+                goto out;
+        }
+        if ((xattrop_dir = opendir (xattrop_directory)) == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        priv->to_be_healed_states = sync_started;
+        while ((entry = readdir(xattrop_dir)) != NULL) {
+                if (!strcmp (entry->d_name, ".") ||
+                    !strcmp (entry->d_name, "..")) {
+                        continue;
+                }
+                if (strncmp (entry->d_name, XATTROP_SUBDIR"-",
+                    strlen (XATTROP_SUBDIR"-"))) {
+                        continue;
+                }
+                if (!strncmp (entry->d_name, XATTROP_SUBDIR"-",
+                    strlen (XATTROP_SUBDIR"-"))) {
+
+                    snprintf (xattrop_index_path, PATH_MAX, "%s/%s",
+                              xattrop_directory, entry->d_name);
+
+                    snprintf (base_index_path, PATH_MAX, "%s/%s",
+                              base_indices_holder, entry->d_name);
+
+                    ret = link (xattrop_index_path, base_index_path);
+                    if (ret && errno != EEXIST)
+                        goto out;
+
+                }
+        }
+        ret = closedir (xattrop_dir);
+        if (ret)
+                goto out;
+        ret = closedir (dir_base_holder);
+        if (ret)
+                goto out;
+
+        ret = 0;
+out:
+        return ret;
+
+}
+
+int
+base_indices_syncing_done (int ret, call_frame_t *frame, void *data)
+{
+        index_priv_t         *priv = NULL;
+        priv = data;
+
+        if (!priv)
+                goto out;
+
+        if (ret) {
+                priv->to_be_healed_states = sync_not_started;
+        } else {
+                priv->to_be_healed_states = synced_state;
+        }
+
+        STACK_DESTROY (frame->root);
+
+out:
+        return 0;
+}
+
+int
+sync_base_indices_from_xattrop (xlator_t *this)
+{
+
+        index_priv_t          *priv = NULL;
+        char                  base_indices_holder[PATH_MAX] =  {0};
+        int                   ret = 0;
+        struct stat           st = {0};
+        DIR                   *dir = NULL;
+        struct dirent         *entry = NULL;
+        call_frame_t          *frame = NULL;
+
+        priv = this->private;
+
+        if (priv->to_be_healed_states != sync_not_started) {
+                ret = -1;
+                goto out;
+        }
+
+        snprintf (base_indices_holder, PATH_MAX, "%s/%s", priv->index_basepath,
+                  BASE_INDICES_HOLDER_SUBDIR);
+
+        ret = stat (base_indices_holder, &st);
+
+        if (ret && (errno != ENOENT)) {
+                goto out;
+        } else if (errno == ENOENT) {
+                ret = index_dir_create (this, BASE_INDICES_HOLDER_SUBDIR);
+                if (ret)
+                        goto out;
+        } else {
+                if ((dir = opendir (base_indices_holder)) == NULL) {
+                        ret = -1;
+                        goto out;
+                }
+                while ((entry = readdir (dir)) != NULL) {
+                        if (!strcmp (entry->d_name, ".") ||
+                            !strcmp (entry->d_name,"..")) {
+                                continue;
+                        }
+                        ret = unlink (entry->d_name);
+                        if (ret)
+                                goto out;
+                }
+                closedir (dir);
+        }
+
+        /*At this point of time we have index/base_indicies_holder directory
+         *is with no entries*/
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame) {
+                ret = -1;
+                goto out;
+        }
+        set_lk_owner_from_ptr (&frame->root->lk_owner, frame->root);
+
+        frame->root->pid = LOW_PRIO_PROC_PID;
+
+        ret = synctask_new (this->ctx->env, sync_base_indices,
+                            base_indices_syncing_done,frame, priv);
+
+
+
+out:
+        return ret;
+
+}
+
+int
+index_add (xlator_t *this, uuid_t gfid, const char *subdir)
+{
+        int32_t           op_errno = 0;
+        char              gfid_path[PATH_MAX] = {0};
+        char              index_path[PATH_MAX] = {0};
+        char              base_path[PATH_MAX] = {0};
+        int               ret = 0;
+        uuid_t            index = {0};
+        index_priv_t      *priv = NULL;
+        struct stat       st = {0};
+        int               fd = 0;
+        int               index_created = 0;
+
+        priv = this->private;
+        GF_ASSERT_AND_GOTO_WITH_ERROR (this->name, !uuid_is_null (gfid),
+                                       out, op_errno, EINVAL);
+
+        make_gfid_path (priv->index_basepath, subdir, gfid,
+                        gfid_path, sizeof (gfid_path));
+
+        ret = stat (gfid_path, &st);
+        if (!ret)
+                goto out;
+        index_get_index (priv, index);
+        make_index_path (priv->index_basepath, subdir,
+                         index, index_path, sizeof (index_path));
+        ret = link (index_path, gfid_path);
+        if (!ret || (errno == EEXIST))  {
+                ret = 0;
+                index_created = 1;
+                goto out;
+        }
+
+
+        op_errno = errno;
+        if (op_errno == ENOENT) {
+                ret = index_dir_create (this, subdir);
+                if (ret)
+                        goto out;
+        } else if (op_errno == EMLINK) {
+                index_generate_index (priv, index);
+                make_index_path (priv->index_basepath, subdir,
+                                 index, index_path, sizeof (index_path));
+        } else {
+                goto out;
+        }
+
+        fd = creat (index_path, 0);
+        if ((fd < 0) && (errno != EEXIST)) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR, "%s: Not able to "
+                        "create index (%s)", uuid_utoa (gfid),
+                        strerror (errno));
+                goto out;
+        }
+
+        if (fd >= 0)
+                close (fd);
+
+        ret = link (index_path, gfid_path);
+        if (ret && (errno != EEXIST)) {
+                gf_log (this->name, GF_LOG_ERROR, "%s: Not able to "
+                        "add to index (%s)", uuid_utoa (gfid),
+                        strerror (errno));
+                goto out;
+        } else {
+                index_created = 1;
+        }
+
+        if (priv->to_be_healed_states != sync_not_started) {
+                 make_index_path (priv->index_basepath,
+                                  GF_BASE_INDICES_HOLDER_GFID,
+                                  index, base_path, sizeof (base_path));
+                 ret = link (index_path, base_path);
+                 if (ret)
+                         goto out;
+        }
+
+        ret = 0;
+out:
+        /*If base_indices_holder is not created: create and sync
+         *If directory is present: delete contents and start syncing
+         *If syncing is in progress :No need to do any thing
+         *If syncing is done: No need to do anything*/
+        if (!ret) {
+                switch (priv->to_be_healed_states) {
+                        case sync_not_started:
+                                ret = sync_base_indices_from_xattrop (this);
+                                break;
+                        case sync_started:
+                        case synced_state:
+                                /*No need to do anything*/
+                                break;
+                }
+        }
+        return ret;
+}
+
+int
+index_del (xlator_t *this, uuid_t gfid, const char *subdir)
+{
+        int32_t      op_errno __attribute__((unused)) = 0;
+        index_priv_t *priv = NULL;
+        int          ret = 0;
+        char         gfid_path[PATH_MAX] = {0};
+
+        priv = this->private;
+        GF_ASSERT_AND_GOTO_WITH_ERROR (this->name, !uuid_is_null (gfid),
+                                       out, op_errno, EINVAL);
+        make_gfid_path (priv->index_basepath, subdir, gfid,
+                        gfid_path, sizeof (gfid_path));
+        ret = unlink (gfid_path);
+        if (ret && (errno != ENOENT)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to delete from index (%s)",
+                        gfid_path, strerror (errno));
+                ret = -errno;
+                goto out;
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+static int
+_check_key_is_zero_filled (dict_t *d, char *k, data_t *v,
+                           void *tmp)
+{
+        if (mem_0filled ((const char*)v->data, v->len)) {
+                /* -1 means, no more iterations, treat as 'break' */
+                return -1;
+        }
+        return 0;
+}
+
+
+void
+_xattrop_index_action (xlator_t *this, inode_t *inode,  dict_t *xattr)
+{
+        gf_boolean_t      zero_xattr = _gf_true;
+        index_inode_ctx_t *ctx = NULL;
+        int               ret = 0;
+
+        ret = dict_foreach (xattr, _check_key_is_zero_filled, NULL);
+        if (ret == -1)
+                zero_xattr = _gf_false;
+
+        ret = index_inode_ctx_get (inode, this, &ctx);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Not able to %s %s -> index",
+                        zero_xattr?"add":"del", uuid_utoa (inode->gfid));
+                goto out;
+        }
+        if (zero_xattr) {
+                if (ctx->state == NOTIN)
+                        goto out;
+                ret = index_del (this, inode->gfid, XATTROP_SUBDIR);
+                if (!ret)
+                        ctx->state = NOTIN;
+        } else {
+                if (ctx->state == IN)
+                        goto out;
+                ret = index_add (this, inode->gfid, XATTROP_SUBDIR);
+                if (!ret)
+                        ctx->state = IN;
+        }
+out:
+        return;
+}
+
+void
+fop_xattrop_index_action (xlator_t *this, inode_t *inode, dict_t *xattr)
+{
+        _xattrop_index_action (this, inode, xattr);
+}
+
+void
+fop_fxattrop_index_action (xlator_t *this, inode_t *inode, dict_t *xattr)
+{
+        _xattrop_index_action (this, inode, xattr);
+}
+
+inline gf_boolean_t
+index_xattrop_track (loc_t *loc, gf_xattrop_flags_t flags, dict_t *dict)
+{
+        return (flags == GF_XATTROP_ADD_ARRAY);
+}
+
+inline gf_boolean_t
+index_fxattrop_track (fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict)
+{
+        return (flags == GF_XATTROP_ADD_ARRAY);
+}
+
+int
+__index_fd_ctx_get (fd_t *fd, xlator_t *this, index_fd_ctx_t **ctx)
+{
+        int               ret = 0;
+        index_fd_ctx_t    *fctx = NULL;
+        uint64_t          tmpctx = 0;
+        char              index_dir[PATH_MAX] = {0};
+        index_priv_t      *priv = NULL;
+
+        priv = this->private;
+        if (uuid_compare (fd->inode->gfid, priv->xattrop_vgfid)) {
+                ret = -EINVAL;
+                goto out;
+        }
+
+        ret = __fd_ctx_get (fd, this, &tmpctx);
+        if (!ret) {
+                fctx = (index_fd_ctx_t*) (long) tmpctx;
+                goto out;
+        }
+
+        fctx = GF_CALLOC (1, sizeof (*fctx), gf_index_fd_ctx_t);
+        if (!fctx) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
+        make_index_dir_path (priv->index_basepath, XATTROP_SUBDIR,
+                             index_dir, sizeof (index_dir));
+        fctx->dir = opendir (index_dir);
+        if (!fctx->dir) {
+                ret = -errno;
+                GF_FREE (fctx);
+                fctx = NULL;
+                goto out;
+        }
+
+        ret = __fd_ctx_set (fd, this, (uint64_t)(long)fctx);
+        if (ret) {
+                GF_FREE (fctx);
+                fctx = NULL;
+                ret = -EINVAL;
+                goto out;
+        }
+out:
+        if (fctx)
+                *ctx = fctx;
+        return ret;
+}
+
+int
+index_fd_ctx_get (fd_t *fd, xlator_t *this, index_fd_ctx_t **ctx)
+{
+        int     ret = 0;
+        LOCK (&fd->lock);
+        {
+                ret = __index_fd_ctx_get (fd, this, ctx);
+        }
+        UNLOCK (&fd->lock);
+        return ret;
+}
+
+//new - Not NULL means start a fop
+//new - NULL means done processing the fop
+void
+index_queue_process (xlator_t *this, inode_t *inode, call_stub_t *new)
+{
+        call_stub_t       *stub = NULL;
+        index_inode_ctx_t *ctx = NULL;
+        int               ret = 0;
+        call_frame_t      *frame = NULL;
+
+        LOCK (&inode->lock);
+        {
+                ret = __index_inode_ctx_get (inode, this, &ctx);
+                if (ret)
+                        goto unlock;
+
+                if (new) {
+                        __index_enqueue (&ctx->callstubs, new);
+                        new = NULL;
+                } else {
+                        ctx->processing = _gf_false;
+                }
+
+                if (!ctx->processing) {
+                        stub = __index_dequeue (&ctx->callstubs);
+                        if (stub)
+                                ctx->processing = _gf_true;
+                        else
+                                ctx->processing = _gf_false;
+                }
+        }
+unlock:
+        UNLOCK (&inode->lock);
+
+        if (ret && new) {
+                frame = new->frame;
+                if (new->fop == GF_FOP_XATTROP) {
+                        INDEX_STACK_UNWIND (xattrop, frame, -1, ENOMEM,
+                                            NULL, NULL);
+                } else if (new->fop == GF_FOP_FXATTROP) {
+                        INDEX_STACK_UNWIND (fxattrop, frame, -1, ENOMEM,
+                                            NULL, NULL);
+                }
+                call_stub_destroy (new);
+        } else if (stub) {
+                call_resume (stub);
+        }
+        return;
+}
+
+int32_t
+index_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, dict_t *xattr, dict_t *xdata)
+{
+        inode_t *inode = NULL;
+
+        inode = inode_ref (frame->local);
+        if (op_ret < 0)
+                goto out;
+        fop_xattrop_index_action (this, frame->local, xattr);
+out:
+        INDEX_STACK_UNWIND (xattrop, frame, op_ret, op_errno, xattr, xdata);
+        index_queue_process (this, inode, NULL);
+        inode_unref (inode);
+
+        return 0;
+}
+
+int32_t
+index_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xattr,
+                    dict_t *xdata)
+{
+        inode_t *inode = NULL;
+
+        inode = inode_ref (frame->local);
+        if (op_ret < 0)
+                goto out;
+
+        fop_fxattrop_index_action (this, frame->local, xattr);
+out:
+        INDEX_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, xattr, xdata);
+        index_queue_process (this, inode, NULL);
+        inode_unref (inode);
+
+        return 0;
+}
+
+int
+index_xattrop_wrapper (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                       gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+{
+        STACK_WIND (frame, index_xattrop_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->xattrop, loc, optype, xattr,
+                    xdata);
+        return 0;
+}
+
+int
+index_fxattrop_wrapper (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                        gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+{
+        STACK_WIND (frame, index_fxattrop_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->fxattrop, fd, optype, xattr,
+                    xdata);
+        return 0;
+}
+
+int32_t
+index_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
+	       gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+        call_stub_t     *stub = NULL;
+
+        if (!index_xattrop_track (loc, flags, dict))
+                goto out;
+
+        frame->local = inode_ref (loc->inode);
+        stub = fop_xattrop_stub (frame, index_xattrop_wrapper,
+                                 loc, flags, dict, xdata);
+        if (!stub) {
+                INDEX_STACK_UNWIND (xattrop, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        index_queue_process (this, loc->inode, stub);
+        return 0;
+out:
+        STACK_WIND (frame, default_xattrop_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop, loc, flags, dict, xdata);
+        return 0;
+}
+
+int32_t
+index_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+
+        if (!index_fxattrop_track (fd, flags, dict))
+                goto out;
+
+        frame->local = inode_ref (fd->inode);
+        stub = fop_fxattrop_stub (frame, index_fxattrop_wrapper,
+                                  fd, flags, dict, xdata);
+        if (!stub) {
+                INDEX_STACK_UNWIND (fxattrop, frame, -1, ENOMEM, NULL, xdata);
+                return 0;
+        }
+
+        index_queue_process (this, fd->inode, stub);
+        return 0;
+out:
+        STACK_WIND (frame, default_fxattrop_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fxattrop, fd, flags, dict, xdata);
+        return 0;
+}
+
+int32_t
+index_getxattr_wrapper (call_frame_t *frame, xlator_t *this,
+                        loc_t *loc, const char *name, dict_t *xdata)
+{
+        index_priv_t    *priv = NULL;
+        dict_t          *xattr = NULL;
+        int             ret = 0;
+
+        priv = this->private;
+
+        xattr = dict_new ();
+        if (!xattr) {
+                ret = -ENOMEM;
+                goto done;
+        }
+
+        if (!strcmp (name, GF_XATTROP_INDEX_GFID)) {
+
+                ret = dict_set_static_bin (xattr, (char*)name,
+                                           priv->xattrop_vgfid,
+                                           sizeof (priv->xattrop_vgfid));
+
+        } else if (!strcmp (name, GF_BASE_INDICES_HOLDER_GFID)) {
+
+                ret = dict_set_static_bin (xattr, (char*)name,
+                                           priv->base_indices_holder_vgfid,
+                                      sizeof (priv->base_indices_holder_vgfid));
+        }
+        if (ret) {
+                ret = -ENOMEM;
+                gf_log (THIS->name, GF_LOG_ERROR, "xattrop index "
+                        "gfid set failed");
+                goto done;
+        }
+done:
+        if (ret)
+                STACK_UNWIND_STRICT (getxattr, frame, -1, -ret, xattr, xdata);
+        else
+                STACK_UNWIND_STRICT (getxattr, frame, 0, 0, xattr, xdata);
+
+        if (xattr)
+                dict_unref (xattr);
+
+        return 0;
+}
+
+int32_t
+index_lookup_wrapper (call_frame_t *frame, xlator_t *this,
+                      loc_t *loc, dict_t *xattr_req)
+{
+        index_priv_t    *priv = NULL;
+        struct stat     lstatbuf = {0};
+        int             ret = 0;
+        int32_t         op_errno = EINVAL;
+        int32_t         op_ret = -1;
+        char            path[PATH_MAX] = {0};
+        struct iatt     stbuf        = {0, };
+        struct iatt     postparent = {0,};
+        dict_t          *xattr = NULL;
+        gf_boolean_t    is_dir = _gf_false;
+
+        priv = this->private;
+
+        VALIDATE_OR_GOTO (loc, done);
+        if (!uuid_compare (loc->gfid, priv->xattrop_vgfid)) {
+                make_index_dir_path (priv->index_basepath, XATTROP_SUBDIR,
+                                     path, sizeof (path));
+                is_dir = _gf_true;
+        } else if (!uuid_compare (loc->pargfid, priv->xattrop_vgfid)) {
+                make_file_path (priv->index_basepath, XATTROP_SUBDIR,
+                                loc->name, path, sizeof (path));
+        } else if (!uuid_compare (loc->gfid,priv->base_indices_holder_vgfid)){
+                make_index_dir_path (priv->index_basepath,
+                                     BASE_INDICES_HOLDER_SUBDIR, path,
+                                    sizeof (path));
+                is_dir = _gf_true;
+        } else if (!uuid_compare (loc->pargfid, priv->base_indices_holder_vgfid)) {
+                make_file_path (priv->index_basepath,
+                                BASE_INDICES_HOLDER_SUBDIR,loc->name, path,
+                                sizeof (path));
+        }
+
+        ret = lstat (path, &lstatbuf);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "Stat failed on index dir "
+                        "(%s)", strerror (errno));
+                op_errno = errno;
+                goto done;
+        } else if (!S_ISDIR (lstatbuf.st_mode) && is_dir) {
+                gf_log (this->name, GF_LOG_DEBUG, "Stat failed on index dir, "
+                        "not a directory");
+                op_errno = ENOENT;
+                goto done;
+        }
+        xattr = dict_new ();
+        if (!xattr) {
+                op_errno = ENOMEM;
+                goto done;
+        }
+
+        iatt_from_stat (&stbuf, &lstatbuf);
+        if (is_dir && !uuid_compare (loc->gfid, priv->xattrop_vgfid)) {
+                uuid_copy (stbuf.ia_gfid, priv->xattrop_vgfid);
+        } else if (is_dir &&
+                !uuid_compare (loc->gfid, priv->base_indices_holder_vgfid)) {
+                uuid_copy (stbuf.ia_gfid, priv->base_indices_holder_vgfid);
+        } else {
+                uuid_generate (stbuf.ia_gfid);
+        }
+        stbuf.ia_ino = -1;
+        op_ret = 0;
+done:
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno,
+                             loc->inode, &stbuf, xattr, &postparent);
+        if (xattr)
+                dict_unref (xattr);
+        return 0;
+}
+
+int32_t
+base_indices_readdir_wrapper (call_frame_t *frame, xlator_t *this,
+                              fd_t *fd, size_t size, off_t off, dict_t *xdata)
+{
+        index_priv_t    *priv = NULL;
+        char            base_indices_holder[PATH_MAX] = {0};
+        DIR             *dir = NULL;
+        int32_t         op_ret = -1;
+        int32_t         op_errno = 0;
+        int             count = 0;
+        gf_dirent_t     entries;
+
+        priv = this->private;
+
+        make_index_dir_path (priv->index_basepath, BASE_INDICES_HOLDER_SUBDIR,
+                             base_indices_holder, sizeof (base_indices_holder));
+
+        dir = opendir (base_indices_holder);
+        if (!dir) {
+                op_errno = EINVAL;
+                goto done;
+        }
+
+
+        INIT_LIST_HEAD (&entries.list);
+
+        count = index_fill_readdir (fd, dir, off, size, &entries,
+                                    BASE_INDICES_HOLDER);
+        /* pick ENOENT to indicate EOF */
+        op_errno = errno;
+        op_ret = count;
+        closedir (dir);
+done:
+        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, &entries, xdata);
+        gf_dirent_free (&entries);
+        return 0;
+}
+
+int32_t
+index_readdir_wrapper (call_frame_t *frame, xlator_t *this,
+                       fd_t *fd, size_t size, off_t off, dict_t *xdata)
+{
+        index_fd_ctx_t       *fctx           = NULL;
+        DIR                  *dir            = NULL;
+        int                   ret            = -1;
+        int32_t               op_ret         = -1;
+        int32_t               op_errno       = 0;
+        int                   count          = 0;
+        gf_dirent_t           entries;
+
+        INIT_LIST_HEAD (&entries.list);
+
+        ret = index_fd_ctx_get (fd, this, &fctx);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "pfd is NULL, fd=%p", fd);
+                op_errno = -ret;
+                goto done;
+        }
+
+        dir = fctx->dir;
+
+        if (!dir) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dir is NULL for fd=%p", fd);
+                op_errno = EINVAL;
+                goto done;
+        }
+
+        count = index_fill_readdir (fd, dir, off, size, &entries,
+                                    INDEX_XATTROP);
+
+        /* pick ENOENT to indicate EOF */
+        op_errno = errno;
+        op_ret = count;
+done:
+        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, &entries, xdata);
+        gf_dirent_free (&entries);
+        return 0;
+}
+
+int
+index_unlink_wrapper (call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
+                      dict_t *xdata)
+{
+        index_priv_t    *priv = NULL;
+        int32_t         op_ret = 0;
+        int32_t         op_errno = 0;
+        int             ret = 0;
+        struct  iatt    preparent = {0};
+        struct  iatt    postparent = {0};
+        char            index_dir[PATH_MAX] = {0};
+        struct  stat    lstatbuf = {0};
+        uuid_t          gfid = {0};
+
+        priv = this->private;
+        make_index_dir_path (priv->index_basepath, XATTROP_SUBDIR,
+                             index_dir, sizeof (index_dir));
+        ret = lstat (index_dir, &lstatbuf);
+        if (ret < 0) {
+                op_ret = -1;
+                op_errno = errno;
+                goto done;
+        }
+
+        iatt_from_stat (&preparent, &lstatbuf);
+        uuid_copy (preparent.ia_gfid, priv->xattrop_vgfid);
+        preparent.ia_ino = -1;
+        uuid_parse (loc->name, gfid);
+        ret = index_del (this, gfid, XATTROP_SUBDIR);
+        if (ret < 0) {
+                op_ret = -1;
+                op_errno = -ret;
+                goto done;
+        }
+        memset (&lstatbuf, 0, sizeof (lstatbuf));
+        ret = lstat (index_dir, &lstatbuf);
+        if (ret < 0) {
+                op_ret = -1;
+                op_errno = errno;
+                goto done;
+        }
+        iatt_from_stat (&postparent, &lstatbuf);
+        uuid_copy (postparent.ia_gfid, priv->xattrop_vgfid);
+        postparent.ia_ino = -1;
+done:
+        INDEX_STACK_UNWIND (unlink, frame, op_ret, op_errno, &preparent,
+                            &postparent, xdata);
+        return 0;
+}
+
+int32_t
+index_getxattr (call_frame_t *frame, xlator_t *this,
+                loc_t *loc, const char *name, dict_t *xdata)
+{
+        call_stub_t     *stub = NULL;
+
+        if (!name)
+                goto out;
+        if (strcmp (GF_XATTROP_INDEX_GFID, name) &&
+            strcmp (GF_BASE_INDICES_HOLDER_GFID, name))
+                goto out;
+
+        stub = fop_getxattr_stub (frame, index_getxattr_wrapper, loc, name,
+                                  xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+        worker_enqueue (this, stub);
+        return 0;
+out:
+        STACK_WIND (frame, default_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+        return 0;
+}
+
+int32_t
+index_lookup (call_frame_t *frame, xlator_t *this,
+              loc_t *loc, dict_t *xattr_req)
+{
+        call_stub_t     *stub = NULL;
+        index_priv_t    *priv = NULL;
+
+        priv = this->private;
+
+        if (uuid_compare (loc->gfid, priv->xattrop_vgfid) &&
+            uuid_compare (loc->pargfid, priv->xattrop_vgfid) &&
+            uuid_compare (loc->gfid, priv->base_indices_holder_vgfid) &&
+            uuid_compare (loc->pargfid, priv->base_indices_holder_vgfid))
+                goto normal;
+
+        stub = fop_lookup_stub (frame, index_lookup_wrapper, loc, xattr_req);
+        if (!stub) {
+                STACK_UNWIND_STRICT (lookup, frame, -1, ENOMEM, loc->inode,
+                                     NULL, NULL, NULL);
+                return 0;
+        }
+        worker_enqueue (this, stub);
+        return 0;
+normal:
+        STACK_WIND (frame, default_lookup_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+
+        return 0;
+}
+
+int32_t
+index_readdir (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, size_t size, off_t off, dict_t *xdata)
+{
+        call_stub_t     *stub = NULL;
+        index_priv_t    *priv = NULL;
+
+        priv = this->private;
+        if (uuid_compare (fd->inode->gfid, priv->xattrop_vgfid) &&
+            uuid_compare (fd->inode->gfid, priv->base_indices_holder_vgfid))
+                goto out;
+
+        if (!uuid_compare (fd->inode->gfid, priv->xattrop_vgfid)) {
+                stub = fop_readdir_stub (frame, index_readdir_wrapper, fd, size, 
+                                         off, xdata);
+        } else if (!uuid_compare (fd->inode->gfid,
+                                  priv->base_indices_holder_vgfid)) {
+                stub = fop_readdir_stub (frame, base_indices_readdir_wrapper,
+                                         fd, size, off, xdata);
+        }
+
+        if (!stub) {
+                STACK_UNWIND_STRICT (readdir, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+        worker_enqueue (this, stub);
+        return 0;
+out:
+        STACK_WIND (frame, default_readdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdir, fd, size, off, xdata);
+        return 0;
+}
+
+int
+index_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+              dict_t *xdata)
+{
+        call_stub_t     *stub = NULL;
+        index_priv_t    *priv = NULL;
+
+        priv = this->private;
+        if (uuid_compare (loc->pargfid, priv->xattrop_vgfid))
+                goto out;
+
+        stub = fop_unlink_stub (frame, index_unlink_wrapper, loc, xflag, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (unlink, frame, -1, ENOMEM, NULL, NULL,
+                                     NULL);
+                return 0;
+        }
+        worker_enqueue (this, stub);
+        return 0;
+out:
+        STACK_WIND (frame, default_unlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+        return 0;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        ret = xlator_mem_acct_init (this, gf_index_mt_end + 1);
+
+        return ret;
+}
+
+int
+init (xlator_t *this)
+{
+        int ret = -1;
+        index_priv_t *priv = NULL;
+        pthread_t thread;
+        pthread_attr_t  w_attr;
+        gf_boolean_t    mutex_inited = _gf_false;
+        gf_boolean_t    cond_inited  = _gf_false;
+        gf_boolean_t    attr_inited  = _gf_false;
+
+	if (!this->children || this->children->next) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"'index' not configured with exactly one child");
+                goto out;
+	}
+
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+
+        priv = GF_CALLOC (1, sizeof (*priv), gf_index_mt_priv_t);
+        if (!priv)
+                goto out;
+
+        LOCK_INIT (&priv->lock);
+        if ((ret = pthread_cond_init(&priv->cond, NULL)) != 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "pthread_cond_init failed (%d)", ret);
+                goto out;
+        }
+        cond_inited = _gf_true;
+
+        if ((ret = pthread_mutex_init(&priv->mutex, NULL)) != 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "pthread_mutex_init failed (%d)", ret);
+                goto out;
+        }
+        mutex_inited = _gf_true;
+
+        if ((ret = pthread_attr_init (&w_attr)) != 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "pthread_attr_init failed (%d)", ret);
+                goto out;
+        }
+        attr_inited = _gf_true;
+
+        ret = pthread_attr_setstacksize (&w_attr, INDEX_THREAD_STACK_SIZE);
+        if (ret == EINVAL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Using default thread stack size");
+        }
+        GF_OPTION_INIT ("index-base", priv->index_basepath, path, out);
+        uuid_generate (priv->index);
+        uuid_generate (priv->xattrop_vgfid);
+        /*base_indices_holder is a directory which contains hard links to
+         * all base indices inside indices/xattrop directory*/
+        uuid_generate (priv->base_indices_holder_vgfid);
+        INIT_LIST_HEAD (&priv->callstubs);
+
+        this->private = priv;
+        ret = gf_thread_create (&thread, &w_attr, index_worker, this);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "Failed to create "
+                        "worker thread, aborting");
+                goto out;
+        }
+        ret = 0;
+out:
+        if (ret) {
+                if (cond_inited)
+                        pthread_cond_destroy (&priv->cond);
+                if (mutex_inited)
+                        pthread_mutex_destroy (&priv->mutex);
+                if (priv)
+                        GF_FREE (priv);
+                this->private = NULL;
+        }
+        if (attr_inited)
+                pthread_attr_destroy (&w_attr);
+        return ret;
+}
+
+void
+fini (xlator_t *this)
+{
+        index_priv_t *priv = NULL;
+
+        priv = this->private;
+        if (!priv)
+                goto out;
+        this->private = NULL;
+        LOCK_DESTROY (&priv->lock);
+        pthread_cond_destroy (&priv->cond);
+        pthread_mutex_destroy (&priv->mutex);
+        GF_FREE (priv);
+out:
+        return;
+}
+
+int
+index_forget (xlator_t *this, inode_t *inode)
+{
+        uint64_t tmp_cache = 0;
+        if (!inode_ctx_del (inode, this, &tmp_cache))
+                GF_FREE ((index_inode_ctx_t*) (long)tmp_cache);
+
+        return 0;
+}
+
+int32_t
+index_releasedir (xlator_t *this, fd_t *fd)
+{
+        index_fd_ctx_t *fctx = NULL;
+        uint64_t        ctx = 0;
+        int             ret = 0;
+
+        ret = fd_ctx_del (fd, this, &ctx);
+        if (ret < 0)
+                goto out;
+
+        fctx = (index_fd_ctx_t*) (long) ctx;
+        if (fctx->dir)
+                closedir (fctx->dir);
+
+        GF_FREE (fctx);
+out:
+        return 0;
+}
+
+int32_t
+index_release (xlator_t *this, fd_t *fd)
+{
+        index_fd_ctx_t *fctx = NULL;
+        uint64_t        ctx = 0;
+        int             ret = 0;
+
+        ret = fd_ctx_del (fd, this, &ctx);
+        if (ret < 0)
+                goto out;
+
+        fctx = (index_fd_ctx_t*) (long) ctx;
+        GF_FREE (fctx);
+out:
+        return 0;
+}
+
+int
+notify (xlator_t *this, int event, void *data, ...)
+{
+        int     ret = 0;
+        ret = default_notify (this, event, data);
+        return ret;
+}
+
+struct xlator_fops fops = {
+	.xattrop     = index_xattrop,
+	.fxattrop    = index_fxattrop,
+
+        //interface functions follow
+        .getxattr    = index_getxattr,
+        .lookup      = index_lookup,
+        .readdir     = index_readdir,
+        .unlink      = index_unlink
+};
+
+struct xlator_dumpops dumpops;
+
+struct xlator_cbks cbks = {
+        .forget         = index_forget,
+        .release        = index_release,
+        .releasedir     = index_releasedir
+};
+
+struct volume_options options[] = {
+        { .key  = {"index-base" },
+          .type = GF_OPTION_TYPE_PATH,
+          .description = "path where the index files need to be stored",
+        },
+        { .key  = {NULL} },
+};
diff --git a/xlators/features/index/src/index.h b/xlators/features/index/src/index.h
new file mode 100644
index 000000000..d6dcb1c23
--- /dev/null
+++ b/xlators/features/index/src/index.h
@@ -0,0 +1,73 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __INDEX_H__
+#define __INDEX_H__
+
+#include "xlator.h"
+#include "call-stub.h"
+#include "defaults.h"
+#include "byte-order.h"
+#include "common-utils.h"
+#include "index-mem-types.h"
+
+#define INDEX_THREAD_STACK_SIZE   ((size_t)(1024*1024))
+
+typedef enum {
+        UNKNOWN,
+        IN,
+        NOTIN
+} index_state_t;
+
+typedef struct index_inode_ctx {
+        gf_boolean_t processing;
+        struct list_head callstubs;
+        index_state_t state;
+} index_inode_ctx_t;
+
+typedef struct index_fd_ctx {
+        DIR *dir;
+} index_fd_ctx_t;
+
+typedef enum {
+        sync_not_started,
+        sync_started,
+        synced_state,
+} to_be_healed_states_t;
+
+typedef enum {
+        INDEX_XATTROP,
+        BASE_INDICES_HOLDER,
+} readdir_directory;
+
+typedef struct index_priv {
+        char *index_basepath;
+        uuid_t index;
+        gf_lock_t lock;
+        uuid_t xattrop_vgfid;//virtual gfid of the xattrop index dir
+        uuid_t base_indices_holder_vgfid; //virtual gfid of the
+                                          //to_be_healed_xattrop directory
+        struct list_head callstubs;
+        pthread_mutex_t mutex;
+        pthread_cond_t  cond;
+        to_be_healed_states_t to_be_healed_states;
+} index_priv_t;
+
+#define INDEX_STACK_UNWIND(fop, frame, params ...)      \
+do {                                                    \
+        if (frame) {                                    \
+                inode_t *_inode = frame->local;         \
+                frame->local = NULL;                    \
+                inode_unref (_inode);                   \
+        }                                               \
+        STACK_UNWIND_STRICT (fop, frame, params);       \
+} while (0)
+
+#endif
diff --git a/xlators/features/locks/src/Makefile.am b/xlators/features/locks/src/Makefile.am
index d10b874be..0f79731b4 100644
--- a/xlators/features/locks/src/Makefile.am
+++ b/xlators/features/locks/src/Makefile.am
@@ -1,15 +1,18 @@
 xlator_LTLIBRARIES = locks.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
 
-locks_la_LDFLAGS = -module -avoidversion
+locks_la_LDFLAGS = -module -avoid-version
 
-locks_la_SOURCES = common.c posix.c entrylk.c inodelk.c
-locks_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la 
+locks_la_SOURCES = common.c posix.c entrylk.c inodelk.c reservelk.c \
+		   clear.c
+locks_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = locks.h common.h locks-mem-types.h
+noinst_HEADERS = locks.h common.h locks-mem-types.h clear.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -fno-strict-aliasing -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src $(GF_CFLAGS) -shared -nostartfiles
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+
+AM_CFLAGS = -Wall -fno-strict-aliasing $(GF_CFLAGS)
 
 CLEANFILES = 
 
@@ -17,4 +20,4 @@ uninstall-local:
 	rm -f $(DESTDIR)$(xlatordir)/posix-locks.so
 
 install-data-hook:
-	ln -sf locks.so $(DESTDIR)$(xlatordir)/posix-locks.so
-\ No newline at end of file
+	ln -sf locks.so $(DESTDIR)$(xlatordir)/posix-locks.so
diff --git a/xlators/features/locks/src/clear.c b/xlators/features/locks/src/clear.c
new file mode 100644
index 000000000..124b9ad0f
--- /dev/null
+++ b/xlators/features/locks/src/clear.c
@@ -0,0 +1,424 @@
+/*
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#include <unistd.h>
+#include <fcntl.h>
+#include <limits.h>
+#include <pthread.h>
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "compat.h"
+#include "xlator.h"
+#include "inode.h"
+#include "logging.h"
+#include "common-utils.h"
+
+#include "locks.h"
+#include "common.h"
+#include "statedump.h"
+#include "clear.h"
+
+int
+clrlk_get_kind (char *kind)
+{
+        char  *clrlk_kinds[CLRLK_KIND_MAX] = {"dummy", "blocked", "granted",
+                                              "all"};
+        int   ret_kind                     = CLRLK_KIND_MAX;
+        int   i                            = 0;
+
+        for (i = CLRLK_BLOCKED; i < CLRLK_KIND_MAX; i++) {
+                if (!strcmp (clrlk_kinds[i], kind)) {
+                        ret_kind = i;
+                        break;
+                }
+        }
+
+        return ret_kind;
+}
+
+int
+clrlk_get_type (char *type)
+{
+        char    *clrlk_types[CLRLK_TYPE_MAX]    = {"inode", "entry", "posix"};
+        int     ret_type                        = CLRLK_TYPE_MAX;
+        int     i                               = 0;
+
+        for (i = CLRLK_INODE; i < CLRLK_TYPE_MAX; i++) {
+                if (!strcmp (clrlk_types[i], type)) {
+                        ret_type = i;
+                        break;
+                }
+        }
+
+        return ret_type;
+}
+
+int
+clrlk_get_lock_range (char *range_str, struct gf_flock *ulock,
+                      gf_boolean_t *chk_range)
+{
+        int     ret     =       -1;
+
+        if (!chk_range)
+                goto out;
+
+        if (!range_str) {
+                ret = 0;
+                *chk_range = _gf_false;
+                goto out;
+        }
+
+        if (sscanf (range_str, "%hd,%"PRId64"-""%"PRId64, &ulock->l_whence,
+                    &ulock->l_start, &ulock->l_len) != 3) {
+                goto out;
+        }
+
+        ret = 0;
+        *chk_range = _gf_true;
+out:
+        return ret;
+}
+
+int
+clrlk_parse_args (const char* cmd, clrlk_args *args)
+{
+        char            *opts           = NULL;
+        char            *cur            = NULL;
+        char            *tok            = NULL;
+        char            *sptr           = NULL;
+        char            *free_ptr       = NULL;
+        char            kw[KW_MAX]     = {[KW_TYPE]     = 't',
+                                          [KW_KIND]     = 'k',
+                                          };
+        int             ret             = -1;
+        int             i               = 0;
+
+        GF_ASSERT (cmd);
+        free_ptr = opts = GF_CALLOC (1, strlen (cmd), gf_common_mt_char);
+        if (!opts)
+                goto out;
+
+        if (sscanf (cmd, GF_XATTR_CLRLK_CMD".%s", opts) < 1) {
+                ret = -1;
+                goto out;
+        }
+
+        /*clr_lk_prefix.ttype.kkind.args, args - type specific*/
+        cur = opts;
+        for (i = 0; i < KW_MAX && (tok = strtok_r (cur, ".", &sptr));
+             cur = NULL, i++) {
+                if (tok[0] != kw[i]) {
+                        ret = -1;
+                        goto out;
+                }
+                if (i == KW_TYPE)
+                        args->type = clrlk_get_type (tok+1);
+                if (i == KW_KIND)
+                        args->kind = clrlk_get_kind (tok+1);
+        }
+
+        if ((args->type == CLRLK_TYPE_MAX) || (args->kind == CLRLK_KIND_MAX))
+                goto out;
+
+        /*optional args, neither range nor basename can 'legally' contain
+         * "/" in them*/
+        tok = strtok_r (NULL, "/", &sptr);
+        if (tok)
+                args->opts = gf_strdup (tok);
+
+        ret = 0;
+out:
+        GF_FREE (free_ptr);
+        return ret;
+}
+
+int
+clrlk_clear_posixlk (xlator_t *this, pl_inode_t *pl_inode, clrlk_args *args,
+                     int *blkd, int *granted, int *op_errno)
+{
+        posix_lock_t            *plock          = NULL;
+        posix_lock_t            *tmp            = NULL;
+        struct gf_flock         ulock           = {0, };
+        int                     ret             = -1;
+        int                     bcount          = 0;
+        int                     gcount          = 0;
+        gf_boolean_t            chk_range       = _gf_false;
+
+        if (clrlk_get_lock_range (args->opts, &ulock, &chk_range)) {
+                *op_errno = EINVAL;
+                goto out;
+        }
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (plock, tmp, &pl_inode->ext_list,
+                                          list) {
+                        if ((plock->blocked &&
+                             !(args->kind & CLRLK_BLOCKED)) ||
+                            (!plock->blocked &&
+                             !(args->kind & CLRLK_GRANTED)))
+                                continue;
+
+                        if (chk_range &&
+                            (plock->user_flock.l_whence != ulock.l_whence
+                            || plock->user_flock.l_start != ulock.l_start
+                            || plock->user_flock.l_len != ulock.l_len))
+                                continue;
+
+                        list_del_init (&plock->list);
+                        if (plock->blocked) {
+                                bcount++;
+                                pl_trace_out (this, plock->frame, NULL, NULL,
+                                              F_SETLKW, &plock->user_flock,
+                                              -1, EAGAIN, NULL);
+
+                                STACK_UNWIND_STRICT (lk, plock->frame, -1, EAGAIN,
+                                                     &plock->user_flock, NULL);
+
+                        } else {
+                                gcount++;
+                        }
+                        GF_FREE (plock);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+        grant_blocked_locks (this, pl_inode);
+        ret = 0;
+out:
+        *blkd    = bcount;
+        *granted = gcount;
+        return ret;
+}
+
+/* Returns 0 on success and -1 on failure */
+int
+clrlk_clear_inodelk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+                     clrlk_args *args, int *blkd, int *granted, int *op_errno)
+{
+        pl_inode_lock_t         *ilock          = NULL;
+        pl_inode_lock_t         *tmp            = NULL;
+        struct gf_flock         ulock           = {0, };
+        int                     ret             = -1;
+        int                     bcount          = 0;
+        int                     gcount          = 0;
+        gf_boolean_t            chk_range       = _gf_false;
+        struct list_head        released;
+
+        INIT_LIST_HEAD (&released);
+        if (clrlk_get_lock_range (args->opts, &ulock, &chk_range)) {
+                *op_errno = EINVAL;
+                goto out;
+        }
+
+        if (args->kind & CLRLK_BLOCKED)
+                goto blkd;
+
+        if (args->kind & CLRLK_GRANTED)
+                goto granted;
+
+blkd:
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (ilock, tmp, &dom->blocked_inodelks,
+                                          blocked_locks) {
+                        if (chk_range &&
+                            (ilock->user_flock.l_whence != ulock.l_whence
+                            || ilock->user_flock.l_start != ulock.l_start
+                            || ilock->user_flock.l_len != ulock.l_len))
+                                continue;
+
+                        bcount++;
+                        list_del_init (&ilock->blocked_locks);
+                        list_add (&ilock->blocked_locks, &released);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (ilock, tmp, &released, blocked_locks) {
+                list_del_init (&ilock->blocked_locks);
+                pl_trace_out (this, ilock->frame, NULL, NULL, F_SETLKW,
+                              &ilock->user_flock, -1, EAGAIN,
+                              ilock->volume);
+                STACK_UNWIND_STRICT (inodelk, ilock->frame, -1,
+                                     EAGAIN, NULL);
+                //No need to take lock as the locks are only in one list
+                __pl_inodelk_unref (ilock);
+        }
+
+        if (!(args->kind & CLRLK_GRANTED)) {
+                ret = 0;
+                goto out;
+        }
+
+granted:
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (ilock, tmp, &dom->inodelk_list,
+                                          list) {
+                        if (chk_range &&
+                            (ilock->user_flock.l_whence != ulock.l_whence
+                            || ilock->user_flock.l_start != ulock.l_start
+                            || ilock->user_flock.l_len != ulock.l_len))
+                                continue;
+
+                        gcount++;
+                        list_del_init (&ilock->list);
+                        list_add (&ilock->list, &released);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (ilock, tmp, &released, list) {
+                list_del_init (&ilock->list);
+                //No need to take lock as the locks are only in one list
+                __pl_inodelk_unref (ilock);
+        }
+
+        ret = 0;
+out:
+        grant_blocked_inode_locks (this, pl_inode, dom);
+        *blkd    = bcount;
+        *granted = gcount;
+        return ret;
+}
+
+/* Returns 0 on success and -1 on failure */
+int
+clrlk_clear_entrylk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+                     clrlk_args *args, int *blkd, int *granted, int *op_errno)
+{
+        pl_entry_lock_t         *elock          = NULL;
+        pl_entry_lock_t         *tmp            = NULL;
+        int                     bcount          = 0;
+        int                     gcount          = 0;
+        int                     ret             = -1;
+        struct list_head        removed;
+        struct list_head        released;
+
+        INIT_LIST_HEAD (&released);
+        if (args->kind & CLRLK_BLOCKED)
+                goto blkd;
+
+        if (args->kind & CLRLK_GRANTED)
+                goto granted;
+
+blkd:
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (elock, tmp, &dom->blocked_entrylks,
+                                          blocked_locks) {
+                        if (args->opts) {
+                                if (!elock->basename ||
+                                    strcmp (elock->basename, args->opts))
+                                        continue;
+                        }
+
+                        bcount++;
+
+                        list_del_init (&elock->blocked_locks);
+                        list_add_tail (&elock->blocked_locks, &released);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (elock, tmp, &released, blocked_locks) {
+                list_del_init (&elock->blocked_locks);
+                entrylk_trace_out (this, elock->frame, elock->volume, NULL, NULL,
+                                   elock->basename, ENTRYLK_LOCK, elock->type,
+                                   -1, EAGAIN);
+                STACK_UNWIND_STRICT (entrylk, elock->frame, -1, EAGAIN, NULL);
+                GF_FREE ((char *) elock->basename);
+                GF_FREE (elock->connection_id);
+                GF_FREE (elock);
+        }
+
+        if (!(args->kind & CLRLK_GRANTED)) {
+                ret = 0;
+                goto out;
+        }
+
+granted:
+        INIT_LIST_HEAD (&removed);
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (elock, tmp, &dom->entrylk_list,
+                                          domain_list) {
+                        if (args->opts) {
+                                if (!elock->basename ||
+                                    strcmp (elock->basename, args->opts))
+                                        continue;
+                        }
+
+                        gcount++;
+                        list_del_init (&elock->domain_list);
+                        list_add_tail (&elock->domain_list, &removed);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (elock, tmp, &removed, domain_list) {
+                grant_blocked_entry_locks (this, pl_inode, elock, dom);
+        }
+
+        ret = 0;
+out:
+        *blkd    = bcount;
+        *granted = gcount;
+        return ret;
+}
+
+int
+clrlk_clear_lks_in_all_domains (xlator_t *this, pl_inode_t *pl_inode,
+                                clrlk_args *args, int *blkd, int *granted,
+                                int *op_errno)
+{
+        pl_dom_list_t   *dom            = NULL;
+        int             ret             = -1;
+        int             tmp_bcount      = 0;
+        int             tmp_gcount      = 0;
+
+        if (list_empty (&pl_inode->dom_list)) {
+                ret = 0;
+                goto out;
+        }
+
+        list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
+                tmp_bcount = tmp_gcount = 0;
+
+                switch (args->type)
+                {
+                case CLRLK_INODE:
+                        ret = clrlk_clear_inodelk (this, pl_inode, dom, args,
+                                             &tmp_bcount, &tmp_gcount,
+                                             op_errno);
+                        if (ret)
+                                goto out;
+                        break;
+                case CLRLK_ENTRY:
+                        ret = clrlk_clear_entrylk (this, pl_inode, dom, args,
+                                             &tmp_bcount, &tmp_gcount,
+                                             op_errno);
+                        if (ret)
+                                goto out;
+                        break;
+                }
+
+                *blkd    += tmp_bcount;
+                *granted += tmp_gcount;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
diff --git a/xlators/features/locks/src/clear.h b/xlators/features/locks/src/clear.h
new file mode 100644
index 000000000..511f3f74a
--- /dev/null
+++ b/xlators/features/locks/src/clear.h
@@ -0,0 +1,76 @@
+/*
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef __CLEAR_H__
+#define __CLEAR_H__
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "compat-errno.h"
+#include "stack.h"
+#include "call-stub.h"
+#include "locks.h"
+
+typedef enum {
+        CLRLK_INODE,
+        CLRLK_ENTRY,
+        CLRLK_POSIX,
+        CLRLK_TYPE_MAX
+} clrlk_type;
+
+typedef enum {
+        CLRLK_BLOCKED = 1,
+        CLRLK_GRANTED,
+        CLRLK_ALL,
+        CLRLK_KIND_MAX
+} clrlk_kind;
+
+typedef enum {
+        KW_TYPE,
+        KW_KIND,
+        /*add new keywords here*/
+        KW_MAX
+} clrlk_opts;
+
+struct _clrlk_args;
+typedef struct _clrlk_args clrlk_args;
+
+struct _clrlk_args {
+        int type;
+        int kind;
+        char *opts;
+};
+
+int
+clrlk_get__kind (char *kind);
+int
+clrlk_get_type (char *type);
+int
+clrlk_get_lock_range (char *range_str, struct gf_flock *ulock,
+                      gf_boolean_t *chk_range);
+int
+clrlk_parse_args (const char* cmd, clrlk_args *args);
+
+int
+clrlk_clear_posixlk (xlator_t *this, pl_inode_t *pl_inode, clrlk_args *args,
+                     int *blkd, int *granted, int *op_errno);
+int
+clrlk_clear_inodelk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+                     clrlk_args *args, int *blkd, int *granted, int *op_errno);
+int
+clrlk_clear_entrylk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+                     clrlk_args *args, int *blkd, int *granted, int *op_errno);
+int
+clrlk_clear_lks_in_all_domains (xlator_t *this, pl_inode_t *pl_inode,
+                                clrlk_args *args, int *blkd, int *granted,
+                                int *op_errno);
+#endif /* __CLEAR_H__ */
diff --git a/xlators/features/locks/src/common.c b/xlators/features/locks/src/common.c
index b34cd9781..b3309580d 100644
--- a/xlators/features/locks/src/common.c
+++ b/xlators/features/locks/src/common.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #include <unistd.h>
 #include <fcntl.h>
 #include <limits.h>
@@ -42,24 +32,23 @@ static int
 __is_lock_grantable (pl_inode_t *pl_inode, posix_lock_t *lock);
 static void
 __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock);
+static int
+pl_send_prelock_unlock (xlator_t *this, pl_inode_t *pl_inode,
+                        posix_lock_t *old_lock);
 
 static pl_dom_list_t *
-allocate_domain (const char *volume)
+__allocate_domain (const char *volume)
 {
         pl_dom_list_t *dom = NULL;
 
         dom = GF_CALLOC (1, sizeof (*dom),
                          gf_locks_mt_pl_dom_list_t);
         if (!dom)
-                return NULL;
-
+                goto out;
 
         dom->domain = gf_strdup(volume);
-	if (!dom->domain) {
-		gf_log ("posix-locks", GF_LOG_TRACE,
-			"Out of Memory");
-		return NULL;
-	}
+        if (!dom->domain)
+                goto out;
 
         gf_log ("posix-locks", GF_LOG_TRACE,
                 "New domain allocated: %s", dom->domain);
@@ -70,6 +59,12 @@ allocate_domain (const char *volume)
         INIT_LIST_HEAD (&dom->inodelk_list);
         INIT_LIST_HEAD (&dom->blocked_inodelks);
 
+out:
+        if (dom && (NULL == dom->domain)) {
+                GF_FREE (dom);
+                dom = NULL;
+        }
+
         return dom;
 }
 
@@ -81,19 +76,28 @@ get_domain (pl_inode_t *pl_inode, const char *volume)
 {
         pl_dom_list_t *dom = NULL;
 
-        list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
-                if (strcmp (dom->domain, volume) == 0)
-                        goto found;
+        GF_VALIDATE_OR_GOTO ("posix-locks", pl_inode, out);
+        GF_VALIDATE_OR_GOTO ("posix-locks", volume, out);
 
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
+                        if (strcmp (dom->domain, volume) == 0)
+                                goto unlock;
+                }
 
+                dom = __allocate_domain (volume);
+                if (dom)
+                        list_add (&dom->inode_list, &pl_inode->dom_list);
         }
-
-        dom = allocate_domain(volume);
-
-        if (dom)
-                list_add (&dom->inode_list, &pl_inode->dom_list);
-found:
-
+unlock:
+        pthread_mutex_unlock (&pl_inode->mutex);
+        if (dom) {
+                gf_log ("posix-locks", GF_LOG_TRACE, "Domain %s found", volume);
+        } else {
+                gf_log ("posix-locks", GF_LOG_TRACE, "Domain %s not found", volume);
+        }
+out:
         return dom;
 }
 
@@ -103,6 +107,12 @@ fd_to_fdnum (fd_t *fd)
         return ((unsigned long) fd);
 }
 
+fd_t *
+fd_from_fdnum (posix_lock_t *lock)
+{
+        return ((fd_t *) lock->fd_num);
+}
+
 int
 __pl_inode_is_empty (pl_inode_t *pl_inode)
 {
@@ -126,10 +136,10 @@ __pl_inode_is_empty (pl_inode_t *pl_inode)
 void
 pl_print_locker (char *str, int size, xlator_t *this, call_frame_t *frame)
 {
-        snprintf (str, size, "Pid=%llu, lk-owner=%llu, Transport=%p, Frame=%llu",
+        snprintf (str, size, "Pid=%llu, lk-owner=%s, Client=%p, Frame=%llu",
                   (unsigned long long) frame->root->pid,
-                  (unsigned long long) frame->root->lk_owner,
-                  (void *)frame->root->trans,
+                  lkowner_utoa (&frame->root->lk_owner),
+                  frame->root->client,
                   (unsigned long long) frame->root->unique);
 }
 
@@ -159,18 +169,17 @@ pl_print_lockee (char *str, int size, fd_t *fd, loc_t *loc)
                         ipath = NULL;
         }
 
-        snprintf (str, size, "ino=%llu, fd=%p, path=%s",
-                  (unsigned long long) inode->ino, fd,
+        snprintf (str, size, "gfid=%s, fd=%p, path=%s",
+                  uuid_utoa (inode->gfid), fd,
                   ipath ? ipath : "<nul>");
 
-        if (ipath)
-                GF_FREE (ipath);
+        GF_FREE (ipath);
 }
 
 
 void
 pl_print_lock (char *str, int size, int cmd,
-               struct flock *flock, uint64_t owner)
+               struct gf_flock *flock, gf_lkowner_t *owner)
 {
         char *cmd_str = NULL;
         char *type_str = NULL;
@@ -218,17 +227,17 @@ pl_print_lock (char *str, int size, int cmd,
         }
 
         snprintf (str, size, "lock=FCNTL, cmd=%s, type=%s, "
-                  "start=%llu, len=%llu, pid=%llu, lk-owner=%llu",
+                  "start=%llu, len=%llu, pid=%llu, lk-owner=%s",
                   cmd_str, type_str, (unsigned long long) flock->l_start,
                   (unsigned long long) flock->l_len,
                   (unsigned long long) flock->l_pid,
-                  (unsigned long long) owner);
+                  lkowner_utoa (owner));
 }
 
 
 void
 pl_trace_in (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-             int cmd, struct flock *flock, const char *domain)
+             int cmd, struct gf_flock *flock, const char *domain)
 {
         posix_locks_private_t  *priv = NULL;
         char                    pl_locker[256];
@@ -245,9 +254,9 @@ pl_trace_in (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
         if (domain)
                 pl_print_inodelk (pl_lock, 256, cmd, flock, domain);
         else
-                pl_print_lock (pl_lock, 256, cmd, flock, frame->root->lk_owner);
+                pl_print_lock (pl_lock, 256, cmd, flock, &frame->root->lk_owner);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[REQUEST] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 pl_locker, pl_lockee, pl_lock);
 }
@@ -276,7 +285,7 @@ pl_print_verdict (char *str, int size, int op_ret, int op_errno)
 
 void
 pl_trace_out (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-              int cmd, struct flock *flock, int op_ret, int op_errno, const char *domain)
+              int cmd, struct gf_flock *flock, int op_ret, int op_errno, const char *domain)
 
 {
         posix_locks_private_t  *priv = NULL;
@@ -295,11 +304,11 @@ pl_trace_out (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
         if (domain)
                 pl_print_inodelk (pl_lock, 256, cmd, flock, domain);
         else
-                pl_print_lock (pl_lock, 256, cmd, flock, frame->root->lk_owner);
+                pl_print_lock (pl_lock, 256, cmd, flock, &frame->root->lk_owner);
 
         pl_print_verdict (verdict, 32, op_ret, op_errno);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[%s] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 verdict, pl_locker, pl_lockee, pl_lock);
 }
@@ -307,7 +316,7 @@ pl_trace_out (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
 
 void
 pl_trace_block (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-                int cmd, struct flock *flock, const char *domain)
+                int cmd, struct gf_flock *flock, const char *domain)
 
 {
         posix_locks_private_t  *priv = NULL;
@@ -325,9 +334,9 @@ pl_trace_block (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
         if (domain)
                 pl_print_inodelk (pl_lock, 256, cmd, flock, domain);
         else
-                pl_print_lock (pl_lock, 256, cmd, flock, frame->root->lk_owner);
+                pl_print_lock (pl_lock, 256, cmd, flock, &frame->root->lk_owner);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[BLOCKED] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 pl_locker, pl_lockee, pl_lock);
 }
@@ -354,7 +363,7 @@ pl_trace_flush (xlator_t *this, call_frame_t *frame, fd_t *fd)
         pl_print_locker (pl_locker, 256, this, frame);
         pl_print_lockee (pl_lockee, 256, fd, NULL);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[FLUSH] Locker = {%s} Lockee = {%s}",
                 pl_locker, pl_lockee);
 }
@@ -372,7 +381,7 @@ pl_trace_release (xlator_t *this, fd_t *fd)
 
         pl_print_lockee (pl_lockee, 256, fd, NULL);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[RELEASE] Lockee = {%s}", pl_lockee);
 }
 
@@ -383,6 +392,7 @@ pl_update_refkeeper (xlator_t *this, inode_t *inode)
         pl_inode_t *pl_inode  = NULL;
         int         is_empty  = 0;
         int         need_unref = 0;
+        int         need_ref = 0;
 
         pl_inode = pl_inode_get (this, inode);
 
@@ -396,13 +406,17 @@ pl_update_refkeeper (xlator_t *this, inode_t *inode)
                 }
 
                 if (!is_empty && !pl_inode->refkeeper) {
-                        pl_inode->refkeeper = inode_ref (inode);
+                        need_ref = 1;
+                        pl_inode->refkeeper = inode;
                 }
         }
         pthread_mutex_unlock (&pl_inode->mutex);
 
         if (need_unref)
                 inode_unref (inode);
+
+        if (need_ref)
+                inode_ref (inode);
 }
 
 
@@ -410,74 +424,78 @@ pl_inode_t *
 pl_inode_get (xlator_t *this, inode_t *inode)
 {
         uint64_t    tmp_pl_inode = 0;
-	pl_inode_t *pl_inode = NULL;
-//	mode_t      st_mode = 0;
-	int         ret = 0;
-
-	ret = inode_ctx_get (inode, this,&tmp_pl_inode);
-	if (ret == 0) {
-                pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
-		goto out;
-        }
-	pl_inode = GF_CALLOC (1, sizeof (*pl_inode),
-                              gf_locks_mt_pl_inode_t);
-	if (!pl_inode) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "Allocating new pl inode");
+        pl_inode_t *pl_inode = NULL;
+        int         ret = 0;
 
-/*
-	st_mode  = inode->st_mode;
-	if ((st_mode & S_ISGID) && !(st_mode & S_IXGRP))
-		pl_inode->mandatory = 1;
-*/
+        LOCK (&inode->lock);
+        {
+                ret = __inode_ctx_get (inode, this, &tmp_pl_inode);
+                if (ret == 0) {
+                        pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
+                        goto unlock;
+                }
+                pl_inode = GF_CALLOC (1, sizeof (*pl_inode),
+                                      gf_locks_mt_pl_inode_t);
+                if (!pl_inode) {
+                        goto unlock;
+                }
 
-	pthread_mutex_init (&pl_inode->mutex, NULL);
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Allocating new pl inode");
 
-	INIT_LIST_HEAD (&pl_inode->dom_list);
-	INIT_LIST_HEAD (&pl_inode->ext_list);
-	INIT_LIST_HEAD (&pl_inode->rw_list);
+                pthread_mutex_init (&pl_inode->mutex, NULL);
 
-	inode_ctx_put (inode, this, (uint64_t)(long)(pl_inode));
+                INIT_LIST_HEAD (&pl_inode->dom_list);
+                INIT_LIST_HEAD (&pl_inode->ext_list);
+                INIT_LIST_HEAD (&pl_inode->rw_list);
+                INIT_LIST_HEAD (&pl_inode->reservelk_list);
+                INIT_LIST_HEAD (&pl_inode->blocked_reservelks);
+                INIT_LIST_HEAD (&pl_inode->blocked_calls);
 
-out:
-	return pl_inode;
+                __inode_ctx_put (inode, this, (uint64_t)(long)(pl_inode));
+        }
+unlock:
+        UNLOCK (&inode->lock);
+
+        return pl_inode;
 }
 
 
 /* Create a new posix_lock_t */
 posix_lock_t *
-new_posix_lock (struct flock *flock, void *transport, pid_t client_pid,
-                uint64_t owner, fd_t *fd)
+new_posix_lock (struct gf_flock *flock, client_t *client, pid_t client_pid,
+                gf_lkowner_t *owner, fd_t *fd)
 {
-	posix_lock_t *lock = NULL;
+        posix_lock_t *lock = NULL;
+
+        GF_VALIDATE_OR_GOTO ("posix-locks", flock, out);
+        GF_VALIDATE_OR_GOTO ("posix-locks", client, out);
+        GF_VALIDATE_OR_GOTO ("posix-locks", fd, out);
 
-	lock = GF_CALLOC (1, sizeof (posix_lock_t),
+        lock = GF_CALLOC (1, sizeof (posix_lock_t),
                           gf_locks_mt_posix_lock_t);
-	if (!lock) {
-		return NULL;
-	}
+        if (!lock) {
+                goto out;
+        }
 
-	lock->fl_start = flock->l_start;
-	lock->fl_type  = flock->l_type;
+        lock->fl_start = flock->l_start;
+        lock->fl_type  = flock->l_type;
 
-	if (flock->l_len == 0)
-		lock->fl_end = LLONG_MAX;
-	else
-		lock->fl_end = flock->l_start + flock->l_len - 1;
+        if (flock->l_len == 0)
+                lock->fl_end = LLONG_MAX;
+        else
+                lock->fl_end = flock->l_start + flock->l_len - 1;
 
-	lock->transport  = transport;
+        lock->client     = client;
         lock->fd_num     = fd_to_fdnum (fd);
-	lock->client_pid = client_pid;
-        lock->owner      = owner;
+        lock->fd         = fd;
+        lock->client_pid = client_pid;
+        lock->owner      = *owner;
 
-	INIT_LIST_HEAD (&lock->list);
+        INIT_LIST_HEAD (&lock->list);
 
-	return lock;
+out:
+        return lock;
 }
 
 
@@ -485,7 +503,7 @@ new_posix_lock (struct flock *flock, void *transport, pid_t client_pid,
 void
 __delete_lock (pl_inode_t *pl_inode, posix_lock_t *lock)
 {
-	list_del_init (&lock->list);
+        list_del_init (&lock->list);
 }
 
 
@@ -493,32 +511,37 @@ __delete_lock (pl_inode_t *pl_inode, posix_lock_t *lock)
 void
 __destroy_lock (posix_lock_t *lock)
 {
-	GF_FREE (lock);
+        GF_FREE (lock);
 }
 
 
-/* Convert a posix_lock to a struct flock */
+/* Convert a posix_lock to a struct gf_flock */
 void
-posix_lock_to_flock (posix_lock_t *lock, struct flock *flock)
+posix_lock_to_flock (posix_lock_t *lock, struct gf_flock *flock)
 {
-	flock->l_pid   = lock->client_pid;
-	flock->l_type  = lock->fl_type;
-	flock->l_start = lock->fl_start;
+        flock->l_pid   = lock->client_pid;
+        flock->l_type  = lock->fl_type;
+        flock->l_start = lock->fl_start;
+        flock->l_owner = lock->owner;
 
-	if (lock->fl_end == LLONG_MAX)
-		flock->l_len = 0;
-	else
-		flock->l_len = lock->fl_end - lock->fl_start + 1;
+        if (lock->fl_end == LLONG_MAX)
+                flock->l_len = 0;
+        else
+                flock->l_len = lock->fl_end - lock->fl_start + 1;
 }
 
-
 /* Insert the lock into the inode's lock list */
 static void
 __insert_lock (pl_inode_t *pl_inode, posix_lock_t *lock)
 {
-	list_add_tail (&lock->list, &pl_inode->ext_list);
+        if (lock->blocked)
+                gettimeofday (&lock->blkd_time, NULL);
+        else
+                gettimeofday (&lock->granted_time, NULL);
 
-	return;
+        list_add_tail (&lock->list, &pl_inode->ext_list);
+
+        return;
 }
 
 
@@ -526,14 +549,14 @@ __insert_lock (pl_inode_t *pl_inode, posix_lock_t *lock)
 int
 locks_overlap (posix_lock_t *l1, posix_lock_t *l2)
 {
-	/* 
-	   Note:
-	   FUSE always gives us absolute offsets, so no need to worry 
-	   about SEEK_CUR or SEEK_END
-	*/
-
-	return ((l1->fl_end >= l2->fl_start) &&
-		(l2->fl_end >= l1->fl_start));
+        /*
+           Note:
+           FUSE always gives us absolute offsets, so no need to worry
+           about SEEK_CUR or SEEK_END
+        */
+
+        return ((l1->fl_end >= l2->fl_start) &&
+                (l2->fl_end >= l1->fl_start));
 }
 
 
@@ -542,8 +565,8 @@ int
 same_owner (posix_lock_t *l1, posix_lock_t *l2)
 {
 
-                return ((l1->owner == l2->owner) &&
-                        (l1->transport  == l2->transport));
+        return (is_same_lkowner (&l1->owner, &l2->owner) &&
+                (l1->client == l2->client));
 
 }
 
@@ -552,15 +575,15 @@ same_owner (posix_lock_t *l1, posix_lock_t *l2)
 void
 __delete_unlck_locks (pl_inode_t *pl_inode)
 {
-	posix_lock_t *l = NULL;
-	posix_lock_t *tmp = NULL;
+        posix_lock_t *l = NULL;
+        posix_lock_t *tmp = NULL;
 
-	list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
-		if (l->fl_type == F_UNLCK) {
-			__delete_lock (pl_inode, l);
-			__destroy_lock (l);
-		}
-	}
+        list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
+                if (l->fl_type == F_UNLCK) {
+                        __delete_lock (pl_inode, l);
+                        __destroy_lock (l);
+                }
+        }
 }
 
 
@@ -568,22 +591,22 @@ __delete_unlck_locks (pl_inode_t *pl_inode)
 static posix_lock_t *
 add_locks (posix_lock_t *l1, posix_lock_t *l2)
 {
-	posix_lock_t *sum = NULL;
+        posix_lock_t *sum = NULL;
 
-	sum = GF_CALLOC (1, sizeof (posix_lock_t),
+        sum = GF_CALLOC (1, sizeof (posix_lock_t),
                          gf_locks_mt_posix_lock_t);
-	if (!sum)
-		return NULL;
+        if (!sum)
+                return NULL;
 
-	sum->fl_start = min (l1->fl_start, l2->fl_start);
-	sum->fl_end   = max (l1->fl_end, l2->fl_end);
+        sum->fl_start = min (l1->fl_start, l2->fl_start);
+        sum->fl_end   = max (l1->fl_end, l2->fl_end);
 
-	return sum;
+        return sum;
 }
 
 /* Subtract two locks */
 struct _values {
-	posix_lock_t *locks[3];
+        posix_lock_t *locks[3];
 };
 
 /* {big} must always be contained inside {small} */
@@ -591,89 +614,88 @@ static struct _values
 subtract_locks (posix_lock_t *big, posix_lock_t *small)
 {
 
-	struct _values v = { .locks = {0, 0, 0} };
+        struct _values v = { .locks = {0, 0, 0} };
 
-	if ((big->fl_start == small->fl_start) &&
-	    (big->fl_end   == small->fl_end)) {
-		/* both edges coincide with big */
-		v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
+        if ((big->fl_start == small->fl_start) &&
+            (big->fl_end   == small->fl_end)) {
+                /* both edges coincide with big */
+                v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
                                         gf_locks_mt_posix_lock_t);
                 if (!v.locks[0])
                         goto out;
-		memcpy (v.locks[0], big, sizeof (posix_lock_t));
-		v.locks[0]->fl_type = small->fl_type;
+                memcpy (v.locks[0], big, sizeof (posix_lock_t));
+                v.locks[0]->fl_type = small->fl_type;
                 goto done;
-	}
+        }
 
-	if ((small->fl_start > big->fl_start) &&
+        if ((small->fl_start > big->fl_start) &&
             (small->fl_end   < big->fl_end)) {
-		/* both edges lie inside big */
-		v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
+                /* both edges lie inside big */
+                v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
                                         gf_locks_mt_posix_lock_t);
                 if (!v.locks[0])
                         goto out;
 
-		v.locks[1] = GF_CALLOC (1, sizeof (posix_lock_t),
+                v.locks[1] = GF_CALLOC (1, sizeof (posix_lock_t),
                                         gf_locks_mt_posix_lock_t);
                 if (!v.locks[1])
                         goto out;
 
-		v.locks[2] = GF_CALLOC (1, sizeof (posix_lock_t),
+                v.locks[2] = GF_CALLOC (1, sizeof (posix_lock_t),
                                         gf_locks_mt_posix_lock_t);
                 if (!v.locks[1])
                         goto out;
 
-		memcpy (v.locks[0], big, sizeof (posix_lock_t));
-		v.locks[0]->fl_end = small->fl_start - 1;
+                memcpy (v.locks[0], big, sizeof (posix_lock_t));
+                v.locks[0]->fl_end = small->fl_start - 1;
 
-		memcpy (v.locks[1], small, sizeof (posix_lock_t));
+                memcpy (v.locks[1], small, sizeof (posix_lock_t));
 
-		memcpy (v.locks[2], big, sizeof (posix_lock_t));
-		v.locks[2]->fl_start = small->fl_end + 1;
+                memcpy (v.locks[2], big, sizeof (posix_lock_t));
+                v.locks[2]->fl_start = small->fl_end + 1;
                 goto done;
 
-	}
+        }
 
-	/* one edge coincides with big */
+        /* one edge coincides with big */
         if (small->fl_start == big->fl_start) {
-		v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
+                v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
                                         gf_locks_mt_posix_lock_t);
                 if (!v.locks[0])
                         goto out;
 
-		v.locks[1] = GF_CALLOC (1, sizeof (posix_lock_t),
+                v.locks[1] = GF_CALLOC (1, sizeof (posix_lock_t),
                                         gf_locks_mt_posix_lock_t);
                 if (!v.locks[1])
                         goto out;
 
-		memcpy (v.locks[0], big, sizeof (posix_lock_t));
-		v.locks[0]->fl_start = small->fl_end + 1;
+                memcpy (v.locks[0], big, sizeof (posix_lock_t));
+                v.locks[0]->fl_start = small->fl_end + 1;
 
-		memcpy (v.locks[1], small, sizeof (posix_lock_t));
+                memcpy (v.locks[1], small, sizeof (posix_lock_t));
                 goto done;
-	}
+        }
 
-	if (small->fl_end  == big->fl_end) {
-		v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
+        if (small->fl_end  == big->fl_end) {
+                v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
                                         gf_locks_mt_posix_lock_t);
                 if (!v.locks[0])
                         goto out;
 
-		v.locks[1] = GF_CALLOC (1, sizeof (posix_lock_t),
+                v.locks[1] = GF_CALLOC (1, sizeof (posix_lock_t),
                                         gf_locks_mt_posix_lock_t);
                 if (!v.locks[1])
                         goto out;
 
-		memcpy (v.locks[0], big, sizeof (posix_lock_t));
-		v.locks[0]->fl_end = small->fl_start - 1;
+                memcpy (v.locks[0], big, sizeof (posix_lock_t));
+                v.locks[0]->fl_end = small->fl_start - 1;
 
-		memcpy (v.locks[1], small, sizeof (posix_lock_t));
+                memcpy (v.locks[1], small, sizeof (posix_lock_t));
                 goto done;
-	}
+        }
 
-        gf_log ("posix-locks", GF_LOG_ERROR,
-                "Unexpected case in subtract_locks. Please send "
-                "a bug report to gluster-devel@nongnu.org");
+        GF_ASSERT (0);
+        gf_log ("posix-locks", GF_LOG_ERROR, "Unexpected case in subtract_locks");
 
 out:
         if (v.locks[0]) {
@@ -693,6 +715,36 @@ done:
         return v;
 }
 
+static posix_lock_t *
+first_conflicting_overlap (pl_inode_t *pl_inode, posix_lock_t *lock)
+{
+        posix_lock_t *l = NULL;
+        posix_lock_t *conf = NULL;
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry (l, &pl_inode->ext_list, list) {
+                        if (l->blocked)
+                                continue;
+
+                        if (locks_overlap (l, lock)) {
+                                if (same_owner (l, lock))
+                                        continue;
+
+                                if ((l->fl_type == F_WRLCK) ||
+                                    (lock->fl_type == F_WRLCK)) {
+                                        conf = l;
+                                        goto unlock;
+                                }
+                        }
+                }
+        }
+unlock:
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        return conf;
+}
+
 /*
   Start searching from {begin}, and return the first lock that
   conflicts, NULL if no conflict
@@ -751,6 +803,8 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock)
         struct _values v = { .locks = {0, 0, 0} };
 
         list_for_each_entry_safe (conf, t, &pl_inode->ext_list, list) {
+                if (conf->blocked)
+                        continue;
                 if (!locks_overlap (conf, lock))
                         continue;
 
@@ -759,7 +813,7 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock)
                                 sum = add_locks (lock, conf);
 
                                 sum->fl_type    = lock->fl_type;
-                                sum->transport  = lock->transport;
+                                sum->client     = lock->client;
                                 sum->fd_num     = lock->fd_num;
                                 sum->client_pid = lock->client_pid;
                                 sum->owner      = lock->owner;
@@ -768,6 +822,8 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock)
                                 __destroy_lock (conf);
 
                                 __destroy_lock (lock);
+                                INIT_LIST_HEAD (&sum->list);
+                                posix_lock_to_flock (sum, &sum->user_flock);
                                 __insert_and_merge (pl_inode, sum);
 
                                 return;
@@ -775,7 +831,7 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock)
                                 sum = add_locks (lock, conf);
 
                                 sum->fl_type    = conf->fl_type;
-                                sum->transport  = conf->transport;
+                                sum->client     = conf->client;
                                 sum->fd_num     = conf->fd_num;
                                 sum->client_pid = conf->client_pid;
                                 sum->owner      = conf->owner;
@@ -795,6 +851,8 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock)
                                                 continue;
 
                                         INIT_LIST_HEAD (&v.locks[i]->list);
+                                        posix_lock_to_flock (v.locks[i],
+                                                       &v.locks[i]->user_flock);
                                         __insert_and_merge (pl_inode,
                                                             v.locks[i]);
                                 }
@@ -848,7 +906,7 @@ __grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode, struct list_head *g
                 list_del_init (&l->list);
 
                 if (__is_lock_grantable (pl_inode, l)) {
-			conf = GF_CALLOC (1, sizeof (*conf),
+                        conf = GF_CALLOC (1, sizeof (*conf),
                                           gf_locks_mt_posix_lock_t);
 
                         if (!conf) {
@@ -863,10 +921,9 @@ __grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode, struct list_head *g
                         posix_lock_to_flock (l, &conf->user_flock);
 
                         gf_log (this->name, GF_LOG_TRACE,
-                                "%s (pid=%d) lk-owner:%"PRIu64" %"PRId64" - %"PRId64" => Granted",
+                                "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => Granted",
                                 l->fl_type == F_UNLCK ? "Unlock" : "Lock",
-                                l->client_pid,
-                                l->owner,
+                                l->client_pid, lkowner_utoa (&l->owner),
                                 l->user_flock.l_start,
                                 l->user_flock.l_len);
 
@@ -902,7 +959,8 @@ grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode)
                 pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW,
                               &lock->user_flock, 0, 0, NULL);
 
-                STACK_UNWIND (lock->frame, 0, 0, &lock->user_flock);
+                STACK_UNWIND_STRICT (lk, lock->frame, 0, 0,
+                                     &lock->user_flock, NULL);
 
                 GF_FREE (lock);
         }
@@ -910,6 +968,52 @@ grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode)
         return;
 }
 
+static int
+pl_send_prelock_unlock (xlator_t *this, pl_inode_t *pl_inode,
+                        posix_lock_t *old_lock)
+{
+        struct gf_flock  flock       = {0,};
+        posix_lock_t *unlock_lock = NULL;
+
+        struct list_head granted_list;
+        posix_lock_t     *tmp = NULL;
+        posix_lock_t     *lock = NULL;
+
+        int ret = -1;
+
+        INIT_LIST_HEAD (&granted_list);
+
+        flock.l_type   = F_UNLCK;
+        flock.l_whence = old_lock->user_flock.l_whence;
+        flock.l_start  = old_lock->user_flock.l_start;
+        flock.l_len    = old_lock->user_flock.l_len;
+
+
+        unlock_lock = new_posix_lock (&flock, old_lock->client,
+                                      old_lock->client_pid, &old_lock->owner,
+                                      old_lock->fd);
+        GF_VALIDATE_OR_GOTO (this->name, unlock_lock, out);
+        ret = 0;
+
+        __insert_and_merge (pl_inode, unlock_lock);
+
+        __grant_blocked_locks (this, pl_inode, &granted_list);
+
+        list_for_each_entry_safe (lock, tmp, &granted_list, list) {
+                list_del_init (&lock->list);
+
+                pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW,
+                              &lock->user_flock, 0, 0, NULL);
+
+                STACK_UNWIND_STRICT (lk, lock->frame, 0, 0,
+                                     &lock->user_flock, NULL);
+
+                GF_FREE (lock);
+        }
+
+out:
+        return ret;
+}
 
 int
 pl_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
@@ -921,21 +1025,38 @@ pl_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
 
         pthread_mutex_lock (&pl_inode->mutex);
         {
+                /* Send unlock before the actual lock to
+                   prevent lock upgrade / downgrade
+                   problems only if:
+                   - it is a blocking call
+                   - it has other conflicting locks
+                */
+
+                if (can_block &&
+                    !(__is_lock_grantable (pl_inode, lock))) {
+                        ret = pl_send_prelock_unlock (this, pl_inode,
+                                                      lock);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                "Could not send pre-lock "
+                                        "unlock");
+                }
+
                 if (__is_lock_grantable (pl_inode, lock)) {
                         gf_log (this->name, GF_LOG_TRACE,
-                                "%s (pid=%d) lk-owner:%"PRIu64" %"PRId64" - %"PRId64" => OK",
+                                "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => OK",
                                 lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                                 lock->client_pid,
-                                lock->owner,
+                                lkowner_utoa (&lock->owner),
                                 lock->user_flock.l_start,
                                 lock->user_flock.l_len);
                         __insert_and_merge (pl_inode, lock);
                 } else if (can_block) {
                         gf_log (this->name, GF_LOG_TRACE,
-                                "%s (pid=%d) lk-owner:%"PRIu64" %"PRId64" - %"PRId64" => Blocked",
+                                "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => Blocked",
                                 lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                                 lock->client_pid,
-                                lock->owner,
+                                lkowner_utoa (&lock->owner),
                                 lock->user_flock.l_start,
                                 lock->user_flock.l_len);
                         lock->blocked = 1;
@@ -943,10 +1064,10 @@ pl_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
                         ret = -1;
                 } else {
                         gf_log (this->name, GF_LOG_TRACE,
-                                "%s (pid=%d) lk-owner:%"PRIu64" %"PRId64" - %"PRId64" => NOK",
+                                "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => NOK",
                                 lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                                 lock->client_pid,
-                                lock->owner,
+                                lkowner_utoa (&lock->owner),
                                 lock->user_flock.l_start,
                                 lock->user_flock.l_len);
                         errno = EAGAIN;
@@ -968,7 +1089,7 @@ pl_getlk (pl_inode_t *pl_inode, posix_lock_t *lock)
 {
         posix_lock_t *conf = NULL;
 
-        conf = first_overlap (pl_inode, lock);
+        conf = first_conflicting_overlap (pl_inode, lock);
 
         if (conf == NULL) {
                 lock->fl_type = F_UNLCK;
@@ -977,3 +1098,124 @@ pl_getlk (pl_inode_t *pl_inode, posix_lock_t *lock)
 
         return conf;
 }
+
+
+struct _lock_table *
+pl_lock_table_new (void)
+{
+        struct _lock_table *new = NULL;
+
+        new = GF_CALLOC (1, sizeof (struct _lock_table), gf_common_mt_lock_table);
+        if (new == NULL) {
+                goto out;
+        }
+        INIT_LIST_HEAD (&new->entrylk_lockers);
+        INIT_LIST_HEAD (&new->inodelk_lockers);
+        LOCK_INIT (&new->lock);
+out:
+        return new;
+}
+
+
+int
+pl_add_locker (struct _lock_table *table, const char *volume,
+               loc_t *loc, fd_t *fd, pid_t pid, gf_lkowner_t *owner,
+               glusterfs_fop_t type)
+{
+        int32_t         ret = -1;
+        struct _locker *new = NULL;
+
+        GF_VALIDATE_OR_GOTO ("lock-table", table, out);
+        GF_VALIDATE_OR_GOTO ("lock-table", volume, out);
+
+        new = GF_CALLOC (1, sizeof (struct _locker), gf_common_mt_locker);
+        if (new == NULL) {
+                goto out;
+        }
+        INIT_LIST_HEAD (&new->lockers);
+
+        new->volume = gf_strdup (volume);
+
+        if (fd == NULL) {
+                loc_copy (&new->loc, loc);
+        } else {
+                new->fd = fd_ref (fd);
+        }
+
+        new->pid   = pid;
+        new->owner = *owner;
+
+        LOCK (&table->lock);
+        {
+                if (type == GF_FOP_ENTRYLK)
+                        list_add_tail (&new->lockers, &table->entrylk_lockers);
+                else
+                        list_add_tail (&new->lockers, &table->inodelk_lockers);
+        }
+        UNLOCK (&table->lock);
+out:
+        return ret;
+}
+
+int
+pl_del_locker (struct _lock_table *table, const char *volume,
+               loc_t *loc, fd_t *fd, gf_lkowner_t *owner, glusterfs_fop_t type)
+{
+        struct _locker    *locker = NULL;
+        struct _locker    *tmp = NULL;
+        int32_t            ret = -1;
+        struct list_head  *head = NULL;
+        struct list_head   del;
+
+        GF_VALIDATE_OR_GOTO ("lock-table", table, out);
+        GF_VALIDATE_OR_GOTO ("lock-table", volume, out);
+
+        INIT_LIST_HEAD (&del);
+
+        LOCK (&table->lock);
+        {
+                if (type == GF_FOP_ENTRYLK) {
+                        head = &table->entrylk_lockers;
+                } else {
+                        head = &table->inodelk_lockers;
+                }
+
+                list_for_each_entry_safe (locker, tmp, head, lockers) {
+                        if (!is_same_lkowner (&locker->owner, owner) ||
+                            strcmp (locker->volume, volume))
+                                continue;
+
+                        /*
+                         * It is possible for inodelk lock to come on anon-fd
+                         * and inodelk unlock to come on normal fd in case of
+                         * client re-opens. So don't check for fds to be equal.
+                         */
+                        if (locker->fd && fd)
+                                list_move_tail (&locker->lockers, &del);
+                        else if (locker->loc.inode && loc &&
+                                 (locker->loc.inode == loc->inode))
+                                list_move_tail (&locker->lockers, &del);
+                }
+        }
+        UNLOCK (&table->lock);
+
+        tmp = NULL;
+        locker = NULL;
+
+        list_for_each_entry_safe (locker, tmp, &del, lockers) {
+                list_del_init (&locker->lockers);
+                if (locker->fd)
+                        fd_unref (locker->fd);
+                else
+                        loc_wipe (&locker->loc);
+
+                GF_FREE (locker->volume);
+                GF_FREE (locker);
+        }
+
+        ret = 0;
+out:
+        return ret;
+
+}
+
diff --git a/xlators/features/locks/src/common.h b/xlators/features/locks/src/common.h
index d70729447..db19ec978 100644
--- a/xlators/features/locks/src/common.h
+++ b/xlators/features/locks/src/common.h
@@ -1,28 +1,55 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef __COMMON_H__
 #define __COMMON_H__
 
+#include "lkowner.h"
+/*dump locks format strings */
+#define RANGE_FMT               "type=%s, whence=%hd, start=%llu, len=%llu"
+#define ENTRY_FMT               "type=%s on basename=%s"
+#define DUMP_GEN_FMT            "pid = %llu, owner=%s, client=%p"
+#define GRNTD_AT                "granted at %s"
+#define BLKD_AT                 "blocked at %s"
+#define CONN_ID                 "connection-id=%s"
+#define DUMP_BLKD_FMT           DUMP_GEN_FMT", "CONN_ID", "BLKD_AT
+#define DUMP_GRNTD_FMT          DUMP_GEN_FMT", "CONN_ID", "GRNTD_AT
+#define DUMP_BLKD_GRNTD_FMT     DUMP_GEN_FMT", "CONN_ID", "BLKD_AT", "GRNTD_AT
+
+#define ENTRY_BLKD_FMT          ENTRY_FMT", "DUMP_BLKD_FMT
+#define ENTRY_GRNTD_FMT         ENTRY_FMT", "DUMP_GRNTD_FMT
+#define ENTRY_BLKD_GRNTD_FMT    ENTRY_FMT", "DUMP_BLKD_GRNTD_FMT
+
+#define RANGE_BLKD_FMT          RANGE_FMT", "DUMP_BLKD_FMT
+#define RANGE_GRNTD_FMT         RANGE_FMT", "DUMP_GRNTD_FMT
+#define RANGE_BLKD_GRNTD_FMT    RANGE_FMT", "DUMP_BLKD_GRNTD_FMT
+
+#define SET_FLOCK_PID(flock, lock) ((flock)->l_pid = lock->client_pid)
+
+struct _locker {
+        struct list_head  lockers;
+        char             *volume;
+        loc_t             loc;
+        fd_t             *fd;
+        gf_lkowner_t      owner;
+        pid_t             pid;
+};
+
+struct _lock_table {
+        struct list_head  inodelk_lockers;
+        struct list_head  entrylk_lockers;
+        gf_lock_t         lock;
+};
+
 posix_lock_t *
-new_posix_lock (struct flock *flock, void *transport, pid_t client_pid,
-                uint64_t owner, fd_t *fd);
+new_posix_lock (struct gf_flock *flock, client_t *client, pid_t client_pid,
+                gf_lkowner_t *owner, fd_t *fd);
 
 pl_inode_t *
 pl_inode_get (xlator_t *this, inode_t *inode);
@@ -38,7 +65,7 @@ void
 grant_blocked_locks (xlator_t *this, pl_inode_t *inode);
 
 void
-posix_lock_to_flock (posix_lock_t *lock, struct flock *flock);
+posix_lock_to_flock (posix_lock_t *lock, struct gf_flock *flock);
 
 int
 locks_overlap (posix_lock_t *l1, posix_lock_t *l2);
@@ -54,13 +81,14 @@ pl_dom_list_t *
 get_domain (pl_inode_t *pl_inode, const char *volume);
 
 void
-grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom);
+grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode,
+                           pl_dom_list_t *dom);
 
 void
 __delete_inode_lock (pl_inode_lock_t *lock);
 
 void
-__destroy_inode_lock (pl_inode_lock_t *lock);
+__pl_inodelk_unref (pl_inode_lock_t *lock);
 
 void
 grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
@@ -69,19 +97,23 @@ grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
 void pl_update_refkeeper (xlator_t *this, inode_t *inode);
 
 int32_t
-get_inodelk_count (xlator_t *this, inode_t *inode);
+__get_inodelk_count (xlator_t *this, pl_inode_t *pl_inode, char *domname);
+int32_t
+get_inodelk_count (xlator_t *this, inode_t *inode, char *domname);
 
 int32_t
+__get_entrylk_count (xlator_t *this, pl_inode_t *pl_inode);
+int32_t
 get_entrylk_count (xlator_t *this, inode_t *inode);
 
 void pl_trace_in (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-                  int cmd, struct flock *flock, const char *domain);
+                  int cmd, struct gf_flock *flock, const char *domain);
 
 void pl_trace_out (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-                   int cmd, struct flock *flock, int op_ret, int op_errno, const char *domain);
+                   int cmd, struct gf_flock *flock, int op_ret, int op_errno, const char *domain);
 
 void pl_trace_block (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-                     int cmd, struct flock *flock, const char *domain);
+                     int cmd, struct gf_flock *flock, const char *domain);
 
 void pl_trace_flush (xlator_t *this, call_frame_t *frame, fd_t *fd);
 
@@ -108,7 +140,7 @@ void
 pl_print_locker (char *str, int size, xlator_t *this, call_frame_t *frame);
 
 void
-pl_print_inodelk (char *str, int size, int cmd, struct flock *flock, const char *domain);
+pl_print_inodelk (char *str, int size, int cmd, struct gf_flock *flock, const char *domain);
 
 void
 pl_trace_release (xlator_t *this, fd_t *fd);
@@ -116,4 +148,40 @@ pl_trace_release (xlator_t *this, fd_t *fd);
 unsigned long
 fd_to_fdnum (fd_t *fd);
 
+fd_t *
+fd_from_fdnum (posix_lock_t *lock);
+
+int
+pl_reserve_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+                  int can_block);
+int
+reservelks_equal (posix_lock_t *l1, posix_lock_t *l2);
+
+int
+pl_verify_reservelk (xlator_t *this, pl_inode_t *pl_inode,
+                     posix_lock_t *lock, int can_block);
+int
+pl_reserve_unlock (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *reqlock);
+
+uint32_t
+check_entrylk_on_basename (xlator_t *this, inode_t *parent, char *basename);
+
+int32_t
+pl_add_locker (struct _lock_table *table, const char *volume,
+               loc_t *loc,
+               fd_t *fd,
+               pid_t pid,
+               gf_lkowner_t *owner,
+               glusterfs_fop_t type);
+
+int32_t
+pl_del_locker (struct _lock_table *table, const char *volume,
+               loc_t *loc,
+               fd_t *fd,
+               gf_lkowner_t *owner,
+               glusterfs_fop_t type);
+
+struct _lock_table *
+pl_lock_table_new (void);
+
 #endif /* __COMMON_H__ */
diff --git a/xlators/features/locks/src/entrylk.c b/xlators/features/locks/src/entrylk.c
index 1e941e600..0785dc547 100644
--- a/xlators/features/locks/src/entrylk.c
+++ b/xlators/features/locks/src/entrylk.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -35,29 +25,30 @@
 
 static pl_entry_lock_t *
 new_entrylk_lock (pl_inode_t *pinode, const char *basename, entrylk_type type,
-		  void *trans, pid_t client_pid, uint64_t owner, const char *volume)
+                  client_t *client, pid_t client_pid, gf_lkowner_t *owner,
+                  const char *volume)
 
 {
-	pl_entry_lock_t *newlock = NULL;
+        pl_entry_lock_t *newlock = NULL;
 
-	newlock = GF_CALLOC (1, sizeof (pl_entry_lock_t),
+        newlock = GF_CALLOC (1, sizeof (pl_entry_lock_t),
                              gf_locks_mt_pl_entry_lock_t);
-	if (!newlock) {
-		goto out;
-	}
+        if (!newlock) {
+                goto out;
+        }
 
-	newlock->basename       = basename ? gf_strdup (basename) : NULL;
-	newlock->type           = type;
-	newlock->trans          = trans;
-	newlock->volume         = volume;
-        newlock->client_pid     = client_pid;
-        newlock->owner          = owner;
+        newlock->basename   = basename ? gf_strdup (basename) : NULL;
+        newlock->type       = type;
+        newlock->trans      = client;
+        newlock->volume     = volume;
+        newlock->client_pid = client_pid;
+        newlock->owner      = *owner;
 
-	INIT_LIST_HEAD (&newlock->domain_list);
-	INIT_LIST_HEAD (&newlock->blocked_locks);
+        INIT_LIST_HEAD (&newlock->domain_list);
+        INIT_LIST_HEAD (&newlock->blocked_locks);
 
 out:
-	return newlock;
+        return newlock;
 }
 
 
@@ -77,16 +68,16 @@ out:
 static int
 names_conflict (const char *n1, const char *n2)
 {
-	return all_names (n1) || all_names (n2) || !strcmp (n1, n2);
+        return all_names (n1) || all_names (n2) || !strcmp (n1, n2);
 }
 
 
-static int
+static inline int
 __same_entrylk_owner (pl_entry_lock_t *l1, pl_entry_lock_t *l2)
 {
 
-        return ((l1->owner == l2->owner) &&
-		(l1->trans  == l2->trans));
+        return (is_same_lkowner (&l1->owner, &l2->owner) &&
+                (l1->trans  == l2->trans));
 }
 
 
@@ -99,33 +90,33 @@ __same_entrylk_owner (pl_entry_lock_t *l1, pl_entry_lock_t *l2)
 static pl_entry_lock_t *
 __lock_grantable (pl_dom_list_t *dom, const char *basename, entrylk_type type)
 {
-	pl_entry_lock_t *lock = NULL;
+        pl_entry_lock_t *lock = NULL;
 
-	if (list_empty (&dom->entrylk_list))
-		return NULL;
+        if (list_empty (&dom->entrylk_list))
+                return NULL;
 
-	list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
-		if (names_conflict (lock->basename, basename))
-			return lock;
-	}
+        list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
+                if (names_conflict (lock->basename, basename))
+                        return lock;
+        }
 
-	return NULL;
+        return NULL;
 }
 
 static pl_entry_lock_t *
 __blocked_lock_conflict (pl_dom_list_t *dom, const char *basename, entrylk_type type)
 {
-	pl_entry_lock_t *lock = NULL;
+        pl_entry_lock_t *lock = NULL;
 
-	if (list_empty (&dom->blocked_entrylks))
-		return NULL;
+        if (list_empty (&dom->blocked_entrylks))
+                return NULL;
 
-	list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
-		if (names_conflict (lock->basename, basename))
-			return lock;
-	}
+        list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
+                if (names_conflict (lock->basename, basename))
+                        return lock;
+        }
 
-	return NULL;
+        return NULL;
 }
 
 static int
@@ -133,23 +124,23 @@ __owner_has_lock (pl_dom_list_t *dom, pl_entry_lock_t *newlock)
 {
         pl_entry_lock_t *lock = NULL;
 
-	list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
-		if (__same_entrylk_owner (lock, newlock))
-			return 1;
-	}
+        list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
+                if (__same_entrylk_owner (lock, newlock))
+                        return 1;
+        }
 
-	list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
-		if (__same_entrylk_owner (lock, newlock))
-			return 1;
-	}
+        list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
+                if (__same_entrylk_owner (lock, newlock))
+                        return 1;
+        }
 
-	return 0;
+        return 0;
 }
 
 static int
 names_equal (const char *n1, const char *n2)
 {
-	return (n1 == NULL && n2 == NULL) || (n1 && n2 && !strcmp (n1, n2));
+        return (n1 == NULL && n2 == NULL) || (n1 && n2 && !strcmp (n1, n2));
 }
 
 void
@@ -213,7 +204,7 @@ entrylk_trace_in (xlator_t *this, call_frame_t *frame, const char *domain,
         pl_print_lockee (pl_lockee, 256, fd, loc);
         pl_print_entrylk (pl_entrylk, 256, cmd, type, basename, domain);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[REQUEST] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 pl_locker, pl_lockee, pl_entrylk);
 }
@@ -240,7 +231,7 @@ entrylk_trace_out (xlator_t *this, call_frame_t *frame, const char *domain,
         pl_print_entrylk (pl_entrylk, 256, cmd, type, basename, domain);
         pl_print_verdict (verdict, 32, op_ret, op_errno);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[%s] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 verdict, pl_locker, pl_lockee, pl_entrylk);
 }
@@ -266,7 +257,7 @@ entrylk_trace_block (xlator_t *this, call_frame_t *frame, const char *volume,
         pl_print_lockee (pl_lockee, 256, fd, loc);
         pl_print_entrylk (pl_entrylk, 256, cmd, type, basename, volume);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[BLOCKED] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 pl_locker, pl_lockee, pl_entrylk);
 }
@@ -284,21 +275,21 @@ entrylk_trace_block (xlator_t *this, call_frame_t *frame, const char *volume,
 static pl_entry_lock_t *
 __find_most_matching_lock (pl_dom_list_t *dom, const char *basename)
 {
-	pl_entry_lock_t *lock;
-	pl_entry_lock_t *all = NULL;
-	pl_entry_lock_t *exact = NULL;
-
-	if (list_empty (&dom->entrylk_list))
-		return NULL;
-
-	list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
-		if (all_names (lock->basename))
-			all = lock;
-		else if (names_equal (lock->basename, basename))
-			exact = lock;
-	}
+        pl_entry_lock_t *lock;
+        pl_entry_lock_t *all = NULL;
+        pl_entry_lock_t *exact = NULL;
+
+        if (list_empty (&dom->entrylk_list))
+                return NULL;
+
+        list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
+                if (all_names (lock->basename))
+                        all = lock;
+                else if (names_equal (lock->basename, basename))
+                        exact = lock;
+        }
 
-	return (exact ? exact : all);
+        return (exact ? exact : all);
 }
 
 /**
@@ -314,55 +305,55 @@ __find_most_matching_lock (pl_dom_list_t *dom, const char *basename)
 
 int
 __lock_name (pl_inode_t *pinode, const char *basename, entrylk_type type,
-	     call_frame_t *frame, pl_dom_list_t *dom, xlator_t *this, int nonblock)
+             call_frame_t *frame, pl_dom_list_t *dom, xlator_t *this,
+             int nonblock, char *conn_id)
 {
-	pl_entry_lock_t *lock       = NULL;
-	pl_entry_lock_t *conf       = NULL;
-	void            *trans      = NULL;
-        pid_t            client_pid = 0;
-        uint64_t         owner      = 0;
-
-	int ret = -EINVAL;
-
-	trans = frame->root->trans;
-        client_pid = frame->root->pid;
-        owner      = frame->root->lk_owner;
-
-	lock = new_entrylk_lock (pinode, basename, type, trans, client_pid, owner, dom->domain);
-	if (!lock) {
-		ret = -ENOMEM;
-		goto out;
-	}
+        pl_entry_lock_t *lock       = NULL;
+        pl_entry_lock_t *conf       = NULL;
+        int              ret        = -EINVAL;
+
+        lock = new_entrylk_lock (pinode, basename, type,
+                                 frame->root->client, frame->root->pid,
+                                 &frame->root->lk_owner, dom->domain);
+        if (!lock) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
         lock->frame   = frame;
         lock->this    = this;
-        lock->trans   = trans;
-
-	conf = __lock_grantable (dom, basename, type);
-	if (conf) {
-		ret = -EAGAIN;
-		if (nonblock){
-			if (lock->basename)
-				GF_FREE ((char *)lock->basename);
+        lock->trans   = frame->root->client;
+
+        if (conn_id) {
+                lock->connection_id = gf_strdup (conn_id);
+        }
+
+        conf = __lock_grantable (dom, basename, type);
+        if (conf) {
+                ret = -EAGAIN;
+                if (nonblock){
+                        GF_FREE (lock->connection_id);
+                        GF_FREE ((char *)lock->basename);
                         GF_FREE (lock);
                         goto out;
 
                 }
 
-		list_add_tail (&lock->blocked_locks, &dom->blocked_entrylks);
+                gettimeofday (&lock->blkd_time, NULL);
+                list_add_tail (&lock->blocked_locks, &dom->blocked_entrylks);
 
-		gf_log (this->name, GF_LOG_TRACE,
-			"Blocking lock: {pinode=%p, basename=%s}",
-			pinode, basename);
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Blocking lock: {pinode=%p, basename=%s}",
+                        pinode, basename);
 
-		goto out;
-	}
+                goto out;
+        }
 
         if ( __blocked_lock_conflict (dom, basename, type) && !(__owner_has_lock (dom, lock))) {
                 ret = -EAGAIN;
                 if (nonblock) {
-			if (lock->basename)
-				GF_FREE ((char *) lock->basename);
+                        GF_FREE (lock->connection_id);
+                        GF_FREE ((char *) lock->basename);
                         GF_FREE (lock);
                         goto out;
 
@@ -370,22 +361,24 @@ __lock_name (pl_inode_t *pinode, const char *basename, entrylk_type type,
                 lock->frame     = frame;
                 lock->this      = this;
 
+                gettimeofday (&lock->blkd_time, NULL);
                 list_add_tail (&lock->blocked_locks, &dom->blocked_entrylks);
 
                 gf_log (this->name, GF_LOG_TRACE,
                         "Lock is grantable, but blocking to prevent starvation");
-		gf_log (this->name, GF_LOG_TRACE,
-			"Blocking lock: {pinode=%p, basename=%s}",
-			pinode, basename);
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Blocking lock: {pinode=%p, basename=%s}",
+                        pinode, basename);
 
                 ret = -EAGAIN;
-		goto out;
+                goto out;
         }
         switch (type) {
 
-	case ENTRYLK_WRLCK:
-		list_add_tail (&lock->domain_list, &dom->entrylk_list);
-		break;
+        case ENTRYLK_WRLCK:
+                gettimeofday (&lock->granted_time, NULL);
+                list_add_tail (&lock->domain_list, &dom->entrylk_list);
+                break;
 
         default:
 
@@ -393,11 +386,11 @@ __lock_name (pl_inode_t *pinode, const char *basename, entrylk_type type,
                         "Invalid type for entrylk specified: %d", type);
                 ret = -EINVAL;
                 goto out;
-	}
+        }
 
-	ret = 0;
+        ret = 0;
 out:
-	return ret;
+        return ret;
 }
 
 /**
@@ -410,319 +403,366 @@ out:
 pl_entry_lock_t *
 __unlock_name (pl_dom_list_t *dom, const char *basename, entrylk_type type)
 {
-	pl_entry_lock_t *lock = NULL;
-	pl_entry_lock_t *ret_lock = NULL;
-
-	lock = __find_most_matching_lock (dom, basename);
-
-	if (!lock) {
-		gf_log ("locks", GF_LOG_DEBUG,
-			"unlock on %s (type=ENTRYLK_WRLCK) attempted but no matching lock found",
-			basename);
-		goto out;
-	}
-
-	if (names_equal (lock->basename, basename)
-	    && lock->type == type) {
-
-		if (type == ENTRYLK_WRLCK) {
-			list_del_init (&lock->domain_list);
-			ret_lock = lock;
-		}
-	} else {
-		gf_log ("locks", GF_LOG_DEBUG,
-			"Unlock for a non-existing lock!");
-		goto out;
-	}
+        pl_entry_lock_t *lock = NULL;
+        pl_entry_lock_t *ret_lock = NULL;
+
+        lock = __find_most_matching_lock (dom, basename);
+
+        if (!lock) {
+                gf_log ("locks", GF_LOG_DEBUG,
+                        "unlock on %s (type=ENTRYLK_WRLCK) attempted but no matching lock found",
+                        basename);
+                goto out;
+        }
+
+        if (names_equal (lock->basename, basename)
+            && lock->type == type) {
+
+                if (type == ENTRYLK_WRLCK) {
+                        list_del_init (&lock->domain_list);
+                        ret_lock = lock;
+                }
+        } else {
+                gf_log ("locks", GF_LOG_DEBUG,
+                        "Unlock for a non-existing lock!");
+                goto out;
+        }
 
 out:
-	return ret_lock;
+        return ret_lock;
 }
 
+uint32_t
+check_entrylk_on_basename (xlator_t *this, inode_t *parent, char *basename)
+{
+        uint32_t        entrylk = 0;
+        pl_inode_t      *pinode = 0;
+        pl_dom_list_t   *dom = NULL;
+        pl_entry_lock_t *conf       = NULL;
+
+        pinode = pl_inode_get (this, parent);
+        if (!pinode)
+                goto out;
+        pthread_mutex_lock (&pinode->mutex);
+        {
+                list_for_each_entry (dom, &pinode->dom_list, inode_list) {
+                        conf = __lock_grantable (dom, basename, ENTRYLK_WRLCK);
+                        if (conf && conf->basename) {
+                                entrylk = 1;
+                                break;
+                        }
+                }
+        }
+        pthread_mutex_unlock (&pinode->mutex);
+
+out:
+        return entrylk;
+}
 
 void
 __grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
-			     pl_dom_list_t *dom, struct list_head *granted)
+                             pl_dom_list_t *dom, struct list_head *granted)
 {
-	int              bl_ret = 0;
-	pl_entry_lock_t *bl   = NULL;
-	pl_entry_lock_t *tmp  = NULL;
+        int              bl_ret = 0;
+        pl_entry_lock_t *bl   = NULL;
+        pl_entry_lock_t *tmp  = NULL;
 
         struct list_head blocked_list;
 
         INIT_LIST_HEAD (&blocked_list);
         list_splice_init (&dom->blocked_entrylks, &blocked_list);
-        
-	list_for_each_entry_safe (bl, tmp, &blocked_list,
-				  blocked_locks) {
 
-		list_del_init (&bl->blocked_locks);
+        list_for_each_entry_safe (bl, tmp, &blocked_list,
+                                  blocked_locks) {
+
+                list_del_init (&bl->blocked_locks);
 
 
-		gf_log ("locks", GF_LOG_TRACE,
-			"Trying to unblock: {pinode=%p, basename=%s}",
-			pl_inode, bl->basename);
+                gf_log ("locks", GF_LOG_TRACE,
+                        "Trying to unblock: {pinode=%p, basename=%s}",
+                        pl_inode, bl->basename);
 
-		bl_ret = __lock_name (pl_inode, bl->basename, bl->type,
-				      bl->frame, dom, bl->this, 0);
+                bl_ret = __lock_name (pl_inode, bl->basename, bl->type,
+                                      bl->frame, dom, bl->this, 0,
+                                      bl->connection_id);
 
-		if (bl_ret == 0) {
-			list_add (&bl->blocked_locks, granted);
-		} else {
+                if (bl_ret == 0) {
+                        list_add (&bl->blocked_locks, granted);
+                } else {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "should never happen");
-			if (bl->basename)
-				GF_FREE ((char *)bl->basename);
-			GF_FREE (bl);
-		}
-	}
-	return;
+                        GF_FREE (bl->connection_id);
+                        GF_FREE ((char *)bl->basename);
+                        GF_FREE (bl);
+                }
+        }
+        return;
 }
 
 /* Grants locks if possible which are blocked on a lock */
 void
 grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
-			   pl_entry_lock_t *unlocked, pl_dom_list_t *dom)
+                           pl_entry_lock_t *unlocked, pl_dom_list_t *dom)
 {
-	struct list_head  granted_list;
-	pl_entry_lock_t  *tmp = NULL;
-	pl_entry_lock_t  *lock = NULL;
+        struct list_head  granted_list;
+        pl_entry_lock_t  *tmp = NULL;
+        pl_entry_lock_t  *lock = NULL;
 
-	INIT_LIST_HEAD (&granted_list);
+        INIT_LIST_HEAD (&granted_list);
 
-	pthread_mutex_lock (&pl_inode->mutex);
-	{
-		__grant_blocked_entry_locks (this, pl_inode, dom, &granted_list);
-	}
-	pthread_mutex_unlock (&pl_inode->mutex);
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                __grant_blocked_entry_locks (this, pl_inode, dom,
+                                             &granted_list);
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
 
-	list_for_each_entry_safe (lock, tmp, &granted_list, blocked_locks) {
-		list_del_init (&lock->blocked_locks);
+        list_for_each_entry_safe (lock, tmp, &granted_list, blocked_locks) {
+                list_del_init (&lock->blocked_locks);
 
                 entrylk_trace_out (this, lock->frame, NULL, NULL, NULL,
                                    lock->basename, ENTRYLK_LOCK, lock->type,
                                    0, 0);
 
-		STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0);
+                STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0, NULL);
 
-	}
+                GF_FREE (lock->connection_id);
+                GF_FREE ((char *)lock->basename);
+                GF_FREE (lock);
+        }
 
-	GF_FREE ((char *)unlocked->basename);
-	GF_FREE (unlocked);
+        GF_FREE ((char *)unlocked->basename);
+        GF_FREE (unlocked->connection_id);
+        GF_FREE (unlocked);
 
-	return;
+        return;
 }
 
 /**
- * release_entry_locks_for_transport: release all entry locks from this
- * transport for this loc_t
+ * release_entry_locks_for_client: release all entry locks from this
+ * client for this loc_t
  */
 
 static int
-release_entry_locks_for_transport (xlator_t *this, pl_inode_t *pinode,
-				   pl_dom_list_t *dom, void *trans)
+release_entry_locks_for_client (xlator_t *this, pl_inode_t *pinode,
+                                pl_dom_list_t *dom, client_t *client)
 {
-	pl_entry_lock_t  *lock = NULL;
-	pl_entry_lock_t  *tmp = NULL;
-	struct list_head  granted;
+        pl_entry_lock_t  *lock = NULL;
+        pl_entry_lock_t  *tmp = NULL;
+        struct list_head  granted;
         struct list_head  released;
 
-	INIT_LIST_HEAD (&granted);
+        INIT_LIST_HEAD (&granted);
         INIT_LIST_HEAD (&released);
 
-	pthread_mutex_lock (&pinode->mutex);
-	{
+        pthread_mutex_lock (&pinode->mutex);
+        {
                 list_for_each_entry_safe (lock, tmp, &dom->blocked_entrylks,
                                           blocked_locks) {
-                        if (lock->trans != trans)
+                        if (lock->trans != client)
                                 continue;
 
                         list_del_init (&lock->blocked_locks);
 
-			gf_log (this->name, GF_LOG_TRACE,
+                        gf_log (this->name, GF_LOG_TRACE,
                                 "releasing lock on  held by "
-                                "{transport=%p}",trans);
+                                "{client=%p}", client);
 
                         list_add (&lock->blocked_locks, &released);
 
                 }
 
-		list_for_each_entry_safe (lock, tmp, &dom->entrylk_list,
-					  domain_list) {
-			if (lock->trans != trans)
-				continue;
+                list_for_each_entry_safe (lock, tmp, &dom->entrylk_list,
+                                          domain_list) {
+                        if (lock->trans != client)
+                                continue;
 
-			list_del_init (&lock->domain_list);
+                        list_del_init (&lock->domain_list);
 
-			gf_log (this->name, GF_LOG_TRACE,
+                        gf_log (this->name, GF_LOG_TRACE,
                                 "releasing lock on  held by "
-                                "{transport=%p}",trans);
+                                "{client=%p}", client);
 
-			GF_FREE ((char *)lock->basename);
-			GF_FREE (lock);
-		}
+                        GF_FREE ((char *)lock->basename);
+                        GF_FREE (lock->connection_id);
+                        GF_FREE (lock);
+                }
 
-		__grant_blocked_entry_locks (this, pinode, dom, &granted);
+                __grant_blocked_entry_locks (this, pinode, dom, &granted);
 
-	}
+        }
 
-	pthread_mutex_unlock (&pinode->mutex);
+        pthread_mutex_unlock (&pinode->mutex);
 
         list_for_each_entry_safe (lock, tmp, &released, blocked_locks) {
                 list_del_init (&lock->blocked_locks);
 
-                STACK_UNWIND_STRICT (entrylk, lock->frame, -1, EAGAIN);
+                STACK_UNWIND_STRICT (entrylk, lock->frame, -1, EAGAIN, NULL);
 
-                if (lock->basename)
-                        GF_FREE ((char *)lock->basename);
+                GF_FREE ((char *)lock->basename);
+                GF_FREE (lock->connection_id);
                 GF_FREE (lock);
 
         }
 
-	list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) {
-		list_del_init (&lock->blocked_locks);
+        list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) {
+                list_del_init (&lock->blocked_locks);
 
-		STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0);
+                STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0, NULL);
 
-		if (lock->basename)
-			GF_FREE ((char *)lock->basename);
-		GF_FREE (lock);
-	}
+                GF_FREE ((char *)lock->basename);
+                GF_FREE (lock->connection_id);
+                GF_FREE (lock);
+        }
 
-	return 0;
+        return 0;
 }
 
 /* Common entrylk code called by pl_entrylk and pl_fentrylk */
 int
 pl_common_entrylk (call_frame_t *frame, xlator_t *this,
                    const char *volume, inode_t *inode, const char *basename,
-                   entrylk_cmd cmd, entrylk_type type, loc_t *loc, fd_t *fd)
+                   entrylk_cmd cmd, entrylk_type type, loc_t *loc, fd_t *fd,
+                   dict_t *xdata)
+
 {
-        uint64_t owner    = 0;
-	int32_t  op_ret   = -1;
-	int32_t  op_errno = 0;
-
-	void *        transport = NULL;
-	pid_t         pid       = -1;
-
-	pl_inode_t *     pinode   = NULL;
-	int              ret      = -1;
-	pl_entry_lock_t *unlocked = NULL;
-	char             unwind   = 1;
-
-	pl_dom_list_t	  *dom = NULL;
-
-	pinode = pl_inode_get (this, inode);
-	if (!pinode) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		op_errno = ENOMEM;
-		goto out;
-	}
-
-	dom = get_domain (pinode, volume);
-	if (!dom){
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto out;
-	}
+        int32_t          op_ret   = -1;
+        int32_t          op_errno = 0;
+        int              ret      = -1;
+        char             unwind   = 1;
+        GF_UNUSED int    dict_ret = -1;
+        pl_inode_t      *pinode   = NULL;
+        pl_entry_lock_t *unlocked = NULL;
+        pl_dom_list_t   *dom      = NULL;
+        char            *conn_id  = NULL;
+        pl_ctx_t        *ctx      = NULL;
+
+        if (xdata)
+                dict_ret = dict_get_str (xdata, "connection-id", &conn_id);
+
+        pinode = pl_inode_get (this, inode);
+        if (!pinode) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-        entrylk_trace_in (this, frame, volume, fd, loc, basename, cmd, type);
+        dom = get_domain (pinode, volume);
+        if (!dom){
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	pid       = frame->root->pid;
-        owner     = frame->root->lk_owner;
-	transport = frame->root->trans;
+        entrylk_trace_in (this, frame, volume, fd, loc, basename, cmd, type);
 
-	if (owner == 0) {
-		/*
+        if (frame->root->lk_owner.len == 0) {
+                /*
                   this is a special case that means release
-                  all locks from this transport
-		*/
+                  all locks from this client
+                */
 
-		gf_log (this->name, GF_LOG_TRACE,
-			"Releasing locks for transport %p", transport);
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Releasing locks for client %p", frame->root->client);
 
-		release_entry_locks_for_transport (this, pinode, dom, transport);
-		op_ret = 0;
+                release_entry_locks_for_client (this, pinode, dom,
+                                                frame->root->client);
+                op_ret = 0;
 
-		goto out;
-	}
+                goto out;
+        }
 
-	switch (cmd) {
-	case ENTRYLK_LOCK:
-		pthread_mutex_lock (&pinode->mutex);
-		{
-			ret = __lock_name (pinode, basename, type,
-					   frame, dom, this, 0);
-		}
-		pthread_mutex_unlock (&pinode->mutex);
+        switch (cmd) {
+        case ENTRYLK_LOCK:
+                pthread_mutex_lock (&pinode->mutex);
+                {
+                        ret = __lock_name (pinode, basename, type,
+                                           frame, dom, this, 0, conn_id);
+                }
+                pthread_mutex_unlock (&pinode->mutex);
 
                 op_errno = -ret;
-		if (ret < 0) {
+                if (ret < 0) {
                         if (ret == -EAGAIN)
                                 unwind = 0;
                         else
                                 unwind = 1;
-			goto out;
-		} else {
+                        goto out;
+                } else {
                         op_ret = 0;
                         op_errno = 0;
                         unwind = 1;
                         goto out;
                 }
 
-		break;
+                break;
 
-	case ENTRYLK_LOCK_NB:
-		pthread_mutex_lock (&pinode->mutex);
-		{
-			ret = __lock_name (pinode, basename, type,
-					   frame, dom, this, 1);
-		}
-		pthread_mutex_unlock (&pinode->mutex);
+        case ENTRYLK_LOCK_NB:
+                unwind = 1;
+                pthread_mutex_lock (&pinode->mutex);
+                {
+                        ret = __lock_name (pinode, basename, type,
+                                           frame, dom, this, 1, conn_id);
+                }
+                pthread_mutex_unlock (&pinode->mutex);
 
-		if (ret < 0)
-			op_errno = -ret;
+                if (ret < 0) {
+                        op_errno = -ret;
+                        goto out;
+                }
 
-                unwind = 1;
-                goto out;
-		break;
+                break;
 
-	case ENTRYLK_UNLOCK:
-		pthread_mutex_lock (&pinode->mutex);
-		{
+        case ENTRYLK_UNLOCK:
+                pthread_mutex_lock (&pinode->mutex);
+                {
                         unlocked = __unlock_name (dom, basename, type);
-		}
-		pthread_mutex_unlock (&pinode->mutex);
+                }
+                pthread_mutex_unlock (&pinode->mutex);
 
-		if (unlocked)
-			grant_blocked_entry_locks (this, pinode, unlocked, dom);
+                if (unlocked)
+                        grant_blocked_entry_locks (this, pinode, unlocked, dom);
 
-		break;
+                break;
 
-	default:
-		gf_log (this->name, GF_LOG_ERROR,
-			"Unexpected case in entrylk (cmd=%d). Please file"
+        default:
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unexpected case in entrylk (cmd=%d). Please file"
                         "a bug report at http://bugs.gluster.com", cmd);
-		goto out;
-	}
+                goto out;
+        }
 
-	op_ret = 0;
+        op_ret = 0;
 out:
         pl_update_refkeeper (this, inode);
-	if (unwind) {
+        if (unwind) {
                 entrylk_trace_out (this, frame, volume, fd, loc, basename,
                                    cmd, type, op_ret, op_errno);
 
-		STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno);
-	} else {
+                ctx = pl_ctx_get (frame->root->client, this);
+
+                if (ctx == NULL) {
+                        gf_log (this->name, GF_LOG_INFO, "pl_ctx_get() failed");
+                        goto unwind;
+                }
+
+                if (cmd == ENTRYLK_UNLOCK)
+                        pl_del_locker (ctx->ltable, volume, loc, fd,
+                                       &frame->root->lk_owner,
+                                       GF_FOP_ENTRYLK);
+                else
+                        pl_add_locker (ctx->ltable, volume, loc, fd,
+                                       frame->root->pid,
+                                       &frame->root->lk_owner,
+                                       GF_FOP_ENTRYLK);
+
+unwind:
+                STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, NULL);
+        } else {
                 entrylk_trace_block (this, frame, volume, fd, loc, basename,
                                      cmd, type);
         }
 
 
-	return 0;
+        return 0;
 }
 
 /**
@@ -733,11 +773,11 @@ out:
 
 int
 pl_entrylk (call_frame_t *frame, xlator_t *this,
-	    const char *volume, loc_t *loc, const char *basename,
-	    entrylk_cmd cmd, entrylk_type type)
+            const char *volume, loc_t *loc, const char *basename,
+            entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-
-        pl_common_entrylk (frame, this, volume, loc->inode, basename, cmd, type, loc, NULL);
+        pl_common_entrylk (frame, this, volume, loc->inode, basename, cmd,
+                           type, loc, NULL, xdata);
 
         return 0;
 }
@@ -752,16 +792,16 @@ pl_entrylk (call_frame_t *frame, xlator_t *this,
 int
 pl_fentrylk (call_frame_t *frame, xlator_t *this,
              const char *volume, fd_t *fd, const char *basename,
-             entrylk_cmd cmd, entrylk_type type)
+             entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-
-        pl_common_entrylk (frame, this, volume, fd->inode, basename, cmd, type, NULL, fd);
+        pl_common_entrylk (frame, this, volume, fd->inode, basename, cmd,
+                           type, NULL, fd, xdata);
 
         return 0;
 }
 
 
-static int32_t
+int32_t
 __get_entrylk_count (xlator_t *this, pl_inode_t *pl_inode)
 {
         int32_t            count = 0;
@@ -770,24 +810,10 @@ __get_entrylk_count (xlator_t *this, pl_inode_t *pl_inode)
 
         list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
                 list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
-
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                " XATTR DEBUG"
-                                " domain: %s  %s on %s state = Active",
-                                dom->domain,
-                                lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : 
-                                "ENTRYLK_WRLCK", lock->basename);
                         count++;
                 }
 
                 list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
-
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                " XATTR DEBUG"
-                                " domain: %s  %s on %s state = Blocked",
-                                dom->domain,
-                                lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : 
-                                "ENTRYLK_WRLCK", lock->basename);
                         count++;
                 }
 
diff --git a/xlators/features/locks/src/inodelk.c b/xlators/features/locks/src/inodelk.c
index eaae8d4de..508523e11 100644
--- a/xlators/features/locks/src/inodelk.c
+++ b/xlators/features/locks/src/inodelk.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -33,30 +23,40 @@
 #include "locks.h"
 #include "common.h"
 
-void
+inline void
 __delete_inode_lock (pl_inode_lock_t *lock)
 {
-	list_del (&lock->list);
+        list_del (&lock->list);
 }
 
-void
-__destroy_inode_lock (pl_inode_lock_t *lock)
+static inline void
+__pl_inodelk_ref (pl_inode_lock_t *lock)
+{
+        lock->ref++;
+}
+
+inline void
+__pl_inodelk_unref (pl_inode_lock_t *lock)
 {
-	GF_FREE (lock);
+        lock->ref--;
+        if (!lock->ref) {
+                GF_FREE (lock->connection_id);
+                GF_FREE (lock);
+        }
 }
 
 /* Check if 2 inodelks are conflicting on type. Only 2 shared locks don't conflict */
-static int
+static inline int
 inodelk_type_conflict (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
 {
-	if (l2->fl_type == F_WRLCK || l1->fl_type == F_WRLCK)
-		return 1;
+        if (l2->fl_type == F_WRLCK || l1->fl_type == F_WRLCK)
+                return 1;
 
-	return 0;
+        return 0;
 }
 
 void
-pl_print_inodelk (char *str, int size, int cmd, struct flock *flock, const char *domain)
+pl_print_inodelk (char *str, int size, int cmd, struct gf_flock *flock, const char *domain)
 {
         char *cmd_str = NULL;
         char *type_str = NULL;
@@ -115,15 +115,16 @@ pl_print_inodelk (char *str, int size, int cmd, struct flock *flock, const char
 static int
 inodelk_overlap (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
 {
-	return ((l1->fl_end >= l2->fl_start) &&
-		(l2->fl_end >= l1->fl_start));
+        return ((l1->fl_end >= l2->fl_start) &&
+                (l2->fl_end >= l1->fl_start));
 }
 
 /* Returns true if the 2 inodelks have the same owner */
-static int same_inodelk_owner (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
+static inline int
+same_inodelk_owner (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
 {
-	return ((l1->owner == l2->owner) &&
-                (l1->transport  == l2->transport));
+        return (is_same_lkowner (&l1->owner, &l2->owner) &&
+                (l1->client == l2->client));
 }
 
 /* Returns true if the 2 inodelks conflict with each other */
@@ -138,18 +139,19 @@ inodelk_conflict (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
 static pl_inode_lock_t *
 __inodelk_grantable (pl_dom_list_t *dom, pl_inode_lock_t *lock)
 {
-	pl_inode_lock_t *l = NULL;
-	pl_inode_lock_t *ret = NULL;
-	if (list_empty (&dom->inodelk_list))
-		goto out;
-	list_for_each_entry (l, &dom->inodelk_list, list){
-		if (inodelk_conflict (lock, l)) {
-			ret = l;
-			goto out;
-		}
-	}
+        pl_inode_lock_t *l = NULL;
+        pl_inode_lock_t *ret = NULL;
+        if (list_empty (&dom->inodelk_list))
+                goto out;
+        list_for_each_entry (l, &dom->inodelk_list, list){
+                if (inodelk_conflict (lock, l) &&
+                    !same_inodelk_owner (lock, l)) {
+                        ret = l;
+                        goto out;
+                }
+        }
 out:
-	return ret;
+        return ret;
 }
 
 static pl_inode_lock_t *
@@ -158,18 +160,18 @@ __blocked_lock_conflict (pl_dom_list_t *dom, pl_inode_lock_t *lock)
         pl_inode_lock_t *l   = NULL;
         pl_inode_lock_t *ret = NULL;
 
-	if (list_empty (&dom->blocked_entrylks))
-		return NULL;
+        if (list_empty (&dom->blocked_inodelks))
+                return NULL;
 
-	list_for_each_entry (l, &dom->blocked_inodelks, blocked_locks) {
-		if (inodelk_conflict (lock, l)) {
-			ret = l;
-			goto out;
+        list_for_each_entry (l, &dom->blocked_inodelks, blocked_locks) {
+                if (inodelk_conflict (lock, l)) {
+                        ret = l;
+                        goto out;
                 }
-	}
+        }
 
 out:
-	return ret;
+        return ret;
 }
 
 static int
@@ -177,17 +179,17 @@ __owner_has_lock (pl_dom_list_t *dom, pl_inode_lock_t *newlock)
 {
         pl_inode_lock_t *lock = NULL;
 
-	list_for_each_entry (lock, &dom->entrylk_list, list) {
-		if (same_inodelk_owner (lock, newlock))
-			return 1;
-	}
+        list_for_each_entry (lock, &dom->inodelk_list, list) {
+                if (same_inodelk_owner (lock, newlock))
+                        return 1;
+        }
 
-	list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
-		if (same_inodelk_owner (lock, newlock))
-			return 1;
-	}
+        list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) {
+                if (same_inodelk_owner (lock, newlock))
+                        return 1;
+        }
 
-	return 0;
+        return 0;
 }
 
 
@@ -196,81 +198,85 @@ __owner_has_lock (pl_dom_list_t *dom, pl_inode_lock_t *newlock)
  */
 static int
 __lock_inodelk (xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock,
-		int can_block,  pl_dom_list_t *dom)
+                int can_block,  pl_dom_list_t *dom)
 {
-	pl_inode_lock_t *conf = NULL;
-	int ret = -EINVAL;
+        pl_inode_lock_t *conf = NULL;
+        int ret = -EINVAL;
 
-	conf = __inodelk_grantable (dom, lock);
-	if (conf){
-		ret = -EAGAIN;
-		if (can_block == 0)
-			goto out;
+        conf = __inodelk_grantable (dom, lock);
+        if (conf){
+                ret = -EAGAIN;
+                if (can_block == 0)
+                        goto out;
 
-		list_add_tail (&lock->blocked_locks, &dom->blocked_inodelks);
+                gettimeofday (&lock->blkd_time, NULL);
+                list_add_tail (&lock->blocked_locks, &dom->blocked_inodelks);
 
                 gf_log (this->name, GF_LOG_TRACE,
-                        "%s (pid=%d) lk-owner:%"PRIu64" %"PRId64" - %"PRId64" => Blocked",
+                        "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => Blocked",
                         lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                         lock->client_pid,
-                        lock->owner,
+                        lkowner_utoa (&lock->owner),
                         lock->user_flock.l_start,
                         lock->user_flock.l_len);
 
 
-		goto out;
-	}
+                goto out;
+        }
 
         if (__blocked_lock_conflict (dom, lock) && !(__owner_has_lock (dom, lock))) {
                 ret = -EAGAIN;
                 if (can_block == 0)
                         goto out;
 
+                gettimeofday (&lock->blkd_time, NULL);
                 list_add_tail (&lock->blocked_locks, &dom->blocked_inodelks);
 
                 gf_log (this->name, GF_LOG_TRACE,
                         "Lock is grantable, but blocking to prevent starvation");
-		gf_log (this->name, GF_LOG_TRACE,
-                        "%s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" => Blocked",
+                gf_log (this->name, GF_LOG_TRACE,
+                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => Blocked",
                         lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                         lock->client_pid,
-                        lock->owner,
+                        lkowner_utoa (&lock->owner),
                         lock->user_flock.l_start,
                         lock->user_flock.l_len);
 
 
-		goto out;
+                goto out;
         }
-	list_add (&lock->list, &dom->inodelk_list);
+        __pl_inodelk_ref (lock);
+        gettimeofday (&lock->granted_time, NULL);
+        list_add (&lock->list, &dom->inodelk_list);
 
-	ret = 0;
+        ret = 0;
 
 out:
-	return ret;
+        return ret;
 }
 
 /* Return true if the two inodelks have exactly same lock boundaries */
 static int
 inodelks_equal (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
 {
-	if ((l1->fl_start == l2->fl_start) &&
-	    (l1->fl_end == l2->fl_end))
-		return 1;
+        if ((l1->fl_start == l2->fl_start) &&
+            (l1->fl_end == l2->fl_end))
+                return 1;
 
-	return 0;
+        return 0;
 }
 
 
 static pl_inode_lock_t *
 find_matching_inodelk (pl_inode_lock_t *lock, pl_dom_list_t *dom)
 {
-	pl_inode_lock_t *l = NULL;
-	list_for_each_entry (l, &dom->inodelk_list, list) {
-		if (inodelks_equal (l, lock) &&
+        pl_inode_lock_t *l = NULL;
+        list_for_each_entry (l, &dom->inodelk_list, list) {
+                if (inodelks_equal (l, lock) &&
                     same_inodelk_owner (l, lock))
-			return l;
-	}
-	return NULL;
+                        return l;
+        }
+        return NULL;
 }
 
 /* Set F_UNLCK removes a lock which has the exact same lock boundaries
@@ -280,105 +286,110 @@ static pl_inode_lock_t *
 __inode_unlock_lock (xlator_t *this, pl_inode_lock_t *lock, pl_dom_list_t *dom)
 {
 
-	pl_inode_lock_t *conf = NULL;
+        pl_inode_lock_t *conf = NULL;
 
-	conf = find_matching_inodelk (lock, dom);
-	if (!conf) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        " Matching lock not found for unlock");
-		goto out;
+        conf = find_matching_inodelk (lock, dom);
+        if (!conf) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        " Matching lock not found for unlock %llu-%llu, by %s "
+                        "on %p", (unsigned long long)lock->fl_start,
+                        (unsigned long long)lock->fl_end,
+                        lkowner_utoa (&lock->owner), lock->client);
+                goto out;
         }
-	__delete_inode_lock (conf);
+        __delete_inode_lock (conf);
         gf_log (this->name, GF_LOG_DEBUG,
-                " Matching lock found for unlock");
-        __destroy_inode_lock (lock);
-
+                " Matching lock found for unlock %llu-%llu, by %s on %p",
+                (unsigned long long)lock->fl_start,
+                (unsigned long long)lock->fl_end, lkowner_utoa (&lock->owner),
+                lock->client);
 
 out:
-	return conf;
-
-
+        return conf;
 }
 static void
-__grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, 
-			     struct list_head *granted, pl_dom_list_t *dom)
+__grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode,
+                             struct list_head *granted, pl_dom_list_t *dom)
 {
-	int	      bl_ret = 0;
-	pl_inode_lock_t *bl = NULL;
-	pl_inode_lock_t *tmp = NULL;
+        int              bl_ret = 0;
+        pl_inode_lock_t *bl = NULL;
+        pl_inode_lock_t *tmp = NULL;
 
         struct list_head blocked_list;
 
         INIT_LIST_HEAD (&blocked_list);
         list_splice_init (&dom->blocked_inodelks, &blocked_list);
 
-	list_for_each_entry_safe (bl, tmp, &blocked_list, blocked_locks) {
+        list_for_each_entry_safe (bl, tmp, &blocked_list, blocked_locks) {
 
-		list_del_init (&bl->blocked_locks);
+                list_del_init (&bl->blocked_locks);
 
-		bl_ret = __lock_inodelk (this, pl_inode, bl, 1, dom);
+                bl_ret = __lock_inodelk (this, pl_inode, bl, 1, dom);
 
-		if (bl_ret == 0) {
-			list_add (&bl->blocked_locks, granted);
+                if (bl_ret == 0) {
+                        list_add (&bl->blocked_locks, granted);
                 }
         }
-	return;
+        return;
 }
 
 /* Grant all inodelks blocked on a lock */
 void
-grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom)
+grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode,
+                           pl_dom_list_t *dom)
 {
-	struct list_head granted;
-	pl_inode_lock_t *lock;
-	pl_inode_lock_t *tmp;
-
-	INIT_LIST_HEAD (&granted);
+        struct list_head granted;
+        pl_inode_lock_t *lock;
+        pl_inode_lock_t *tmp;
 
-        if (list_empty (&dom->blocked_inodelks)) {
-                gf_log (this->name, GF_LOG_TRACE,
-                        "No blocked locks to be granted for domain: %s", dom->domain);
-        }
+        INIT_LIST_HEAD (&granted);
 
         pthread_mutex_lock (&pl_inode->mutex);
-	{
-		__grant_blocked_inode_locks (this, pl_inode, &granted, dom);
-	}
+        {
+                __grant_blocked_inode_locks (this, pl_inode, &granted, dom);
+        }
         pthread_mutex_unlock (&pl_inode->mutex);
 
-	list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) {
+        list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) {
                 gf_log (this->name, GF_LOG_TRACE,
-			"%s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" => Granted",
-			lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
-			lock->client_pid,
-			lock->owner,
-			lock->user_flock.l_start,
-			lock->user_flock.l_len);
+                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => Granted",
+                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                        lock->client_pid,
+                        lkowner_utoa (&lock->owner),
+                        lock->user_flock.l_start,
+                        lock->user_flock.l_len);
 
-		pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW,
-			      &lock->user_flock, 0, 0, lock->volume);
+                pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW,
+                              &lock->user_flock, 0, 0, lock->volume);
 
-		STACK_UNWIND_STRICT (inodelk, lock->frame, 0, 0);
-	}
+                STACK_UNWIND_STRICT (inodelk, lock->frame, 0, 0, NULL);
+        }
 
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) {
+                        list_del_init (&lock->blocked_locks);
+                        __pl_inodelk_unref (lock);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
 }
 
-/* Release all inodelks from this transport */
+/* Release all inodelks from this client */
 static int
-release_inode_locks_of_transport (xlator_t *this, pl_dom_list_t *dom,
-                                  inode_t *inode, void *trans)
+release_inode_locks_of_client (xlator_t *this, pl_dom_list_t *dom,
+                                  inode_t *inode, client_t *client)
 {
-	pl_inode_lock_t *tmp = NULL;
-	pl_inode_lock_t *l = NULL;
+        pl_inode_lock_t *tmp = NULL;
+        pl_inode_lock_t *l = NULL;
 
         pl_inode_t * pinode = NULL;
 
-        struct list_head granted;
         struct list_head released;
 
         char *path = NULL;
+        char *file = NULL;
 
-        INIT_LIST_HEAD (&granted);
         INIT_LIST_HEAD (&released);
 
         pinode = pl_inode_get (this, inode);
@@ -387,251 +398,304 @@ release_inode_locks_of_transport (xlator_t *this, pl_dom_list_t *dom,
         {
 
                 list_for_each_entry_safe (l, tmp, &dom->blocked_inodelks, blocked_locks) {
-                        if (l->transport != trans)
+                        if (l->client != client)
                                 continue;
 
                         list_del_init (&l->blocked_locks);
 
-                        if (inode_path (inode, NULL, &path) < 0) {
-                                gf_log (this->name, GF_LOG_TRACE,
-                                        "inode_path failed");
-                                goto unlock;
-                        }
+                        inode_path (inode, NULL, &path);
+                        if (path)
+                                file = path;
+                        else
+                                file = uuid_utoa (inode->gfid);
 
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "releasing lock on %s held by "
-                                "{transport=%p, pid=%"PRId64" lk-owner=%"PRIu64"}",
-                                path, trans,
-                                (uint64_t) l->client_pid,
-                                l->owner);
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "releasing blocking lock on %s held by "
+                                "{client=%p, pid=%"PRId64" lk-owner=%s}",
+                                file, client, (uint64_t) l->client_pid,
+                                lkowner_utoa (&l->owner));
 
                         list_add (&l->blocked_locks, &released);
-
+                        if (path) {
+                                GF_FREE (path);
+                                path = NULL;
+                        }
                 }
 
                 list_for_each_entry_safe (l, tmp, &dom->inodelk_list, list) {
-                        if (l->transport != trans)
+                        if (l->client != client)
                                 continue;
 
-                        __delete_inode_lock (l);
-			__destroy_inode_lock (l);
-
-
-                        if (inode_path (inode, NULL, &path) < 0) {
-                                gf_log (this->name, GF_LOG_TRACE,
-                                        "inode_path failed");
-                                goto unlock;
+                        inode_path (inode, NULL, &path);
+                        if (path)
+                                file = path;
+                        else
+                                file = uuid_utoa (inode->gfid);
+
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "releasing granted lock on %s held by "
+                                "{client=%p, pid=%"PRId64" lk-owner=%s}",
+                                file, client, (uint64_t) l->client_pid,
+                                lkowner_utoa (&l->owner));
+
+                        if (path) {
+                                GF_FREE (path);
+                                path = NULL;
                         }
 
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "releasing lock on %s held by "
-                                "{transport=%p, pid=%"PRId64" lk-owner=%"PRIu64"}",
-                                path, trans,
-                                (uint64_t) l->client_pid,
-                                l->owner);
-
-
+                        __delete_inode_lock (l);
+                        __pl_inodelk_unref (l);
                 }
         }
-unlock:
-        if (path)
-                GF_FREE (path);
+        GF_FREE (path);
 
         pthread_mutex_unlock (&pinode->mutex);
 
         list_for_each_entry_safe (l, tmp, &released, blocked_locks) {
                 list_del_init (&l->blocked_locks);
 
-                STACK_UNWIND_STRICT (inodelk, l->frame, -1, EAGAIN);
-                GF_FREE (l);
+                STACK_UNWIND_STRICT (inodelk, l->frame, -1, EAGAIN, NULL);
+                //No need to take lock as the locks are only in one list
+                __pl_inodelk_unref (l);
         }
 
-	grant_blocked_inode_locks (this, pinode, dom);
-	return 0;
+        grant_blocked_inode_locks (this, pinode, dom);
+        return 0;
 }
 
 
 static int
 pl_inode_setlk (xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock,
-		int can_block,  pl_dom_list_t *dom)
+                int can_block,  pl_dom_list_t *dom)
 {
-	int ret = -EINVAL;
+        int ret = -EINVAL;
         pl_inode_lock_t *retlock = NULL;
+        gf_boolean_t    unref = _gf_true;
 
-	pthread_mutex_lock (&pl_inode->mutex);
-	{
-		if (lock->fl_type != F_UNLCK) {
-			ret = __lock_inodelk (this, pl_inode, lock, can_block, dom);
-			if (ret == 0)
-				gf_log (this->name, GF_LOG_TRACE,
-                                        "%s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" => OK",
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                if (lock->fl_type != F_UNLCK) {
+                        ret = __lock_inodelk (this, pl_inode, lock, can_block, dom);
+                        if (ret == 0) {
+                                gf_log (this->name, GF_LOG_TRACE,
+                                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => OK",
                                         lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                                         lock->client_pid,
-                                        lock->owner,
+                                        lkowner_utoa (&lock->owner),
                                         lock->fl_start,
                                         lock->fl_end);
+                        } else if (ret == -EAGAIN) {
+                                gf_log (this->name, GF_LOG_TRACE,
+                                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => NOK",
+                                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                                        lock->client_pid,
+                                        lkowner_utoa (&lock->owner),
+                                        lock->user_flock.l_start,
+                                        lock->user_flock.l_len);
+                                if (can_block)
+                                        unref = _gf_false;
+                        }
+                } else {
+                        retlock = __inode_unlock_lock (this, lock, dom);
+                        if (!retlock) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "Bad Unlock issued on Inode lock");
+                                ret = -EINVAL;
+                                goto out;
+                        }
+                        __pl_inodelk_unref (retlock);
 
-			if (ret == -EAGAIN)
-				gf_log (this->name, GF_LOG_TRACE,
-                                        "%s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" => NOK",
-					lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
-					lock->client_pid,
-                                        lock->owner,
-					lock->user_flock.l_start,
-					lock->user_flock.l_len);
-
-			goto out;
-		}
-
-
-		retlock = __inode_unlock_lock (this, lock, dom);
-		if (!retlock) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"Bad Unlock issued on Inode lock");
-                        ret = -EINVAL;
-                        goto out;
+                        ret = 0;
                 }
-		__destroy_inode_lock (retlock);
-
-                ret = 0;
-
-
-	}
+        }
 out:
-	pthread_mutex_unlock (&pl_inode->mutex);
-	grant_blocked_inode_locks (this, pl_inode, dom);
+        if (unref)
+                __pl_inodelk_unref (lock);
+        pthread_mutex_unlock (&pl_inode->mutex);
+        grant_blocked_inode_locks (this, pl_inode, dom);
         return ret;
 }
 
 /* Create a new inode_lock_t */
 pl_inode_lock_t *
-new_inode_lock (struct flock *flock, void *transport, pid_t client_pid,
-                uint64_t owner, const char *volume)
+new_inode_lock (struct gf_flock *flock, client_t *client, pid_t client_pid,
+                call_frame_t *frame, xlator_t *this, const char *volume,
+                char *conn_id)
 
 {
-	pl_inode_lock_t *lock = NULL;
+        pl_inode_lock_t *lock = NULL;
 
-	lock = GF_CALLOC (1, sizeof (*lock),
+        lock = GF_CALLOC (1, sizeof (*lock),
                           gf_locks_mt_pl_inode_lock_t);
-	if (!lock) {
-		return NULL;
-	}
+        if (!lock) {
+                return NULL;
+        }
 
-	lock->fl_start = flock->l_start;
-	lock->fl_type  = flock->l_type;
+        lock->fl_start = flock->l_start;
+        lock->fl_type  = flock->l_type;
 
-	if (flock->l_len == 0)
-		lock->fl_end = LLONG_MAX;
-	else
-		lock->fl_end = flock->l_start + flock->l_len - 1;
+        if (flock->l_len == 0)
+                lock->fl_end = LLONG_MAX;
+        else
+                lock->fl_end = flock->l_start + flock->l_len - 1;
 
-	lock->transport  = transport;
-	lock->client_pid = client_pid;
-        lock->owner      = owner;
-	lock->volume     = volume;
+        lock->client     = client;
+        lock->client_pid = client_pid;
+        lock->volume     = volume;
+        lock->owner      = frame->root->lk_owner;
+        lock->frame      = frame;
+        lock->this       = this;
 
-	INIT_LIST_HEAD (&lock->list);
-	INIT_LIST_HEAD (&lock->blocked_locks);
+        if (conn_id) {
+                lock->connection_id = gf_strdup (conn_id);
+        }
+
+        INIT_LIST_HEAD (&lock->list);
+        INIT_LIST_HEAD (&lock->blocked_locks);
+        __pl_inodelk_ref (lock);
+
+        return lock;
+}
+
+int32_t
+_pl_convert_volume (const char *volume, char **res)
+{
+        char    *mdata_vol = NULL;
+        int     ret = 0;
+
+        mdata_vol = strrchr (volume, ':');
+        //if the volume already ends with :metadata don't bother
+        if (mdata_vol && (strcmp (mdata_vol, ":metadata") == 0))
+                return 0;
 
-	return lock;
+        ret = gf_asprintf (res, "%s:metadata", volume);
+        if (ret <= 0)
+                return ENOMEM;
+        return 0;
+}
+
+int32_t
+_pl_convert_volume_for_special_range (struct gf_flock *flock,
+                                      const char *volume, char **res)
+{
+        int32_t     ret = 0;
+
+        if ((flock->l_start == LLONG_MAX -1) &&
+            (flock->l_len == 0)) {
+                ret = _pl_convert_volume (volume, res);
+        }
+
+        return ret;
 }
 
 /* Common inodelk code called from pl_inodelk and pl_finodelk */
 int
 pl_common_inodelk (call_frame_t *frame, xlator_t *this,
                    const char *volume, inode_t *inode, int32_t cmd,
-                   struct flock *flock, loc_t *loc, fd_t *fd)
+                   struct gf_flock *flock, loc_t *loc, fd_t *fd, dict_t *xdata)
 {
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-	int     ret      = -1;
-	int     can_block = 0;
-	void *                  transport  = NULL;
-	pid_t                   client_pid = -1;
-        uint64_t                owner      = -1;
-	pl_inode_t *            pinode     = NULL;
-	pl_inode_lock_t *       reqlock    = NULL;
-	pl_dom_list_t *		dom	   = NULL;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (inode, unwind);
-	VALIDATE_OR_GOTO (flock, unwind);
-
-	if ((flock->l_start < 0) || (flock->l_len < 0)) {
-		op_errno = EINVAL;
-		goto unwind;
-	}
+        int32_t           op_ret     = -1;
+        int32_t           op_errno   = 0;
+        int               ret        = -1;
+        GF_UNUSED int     dict_ret   = -1;
+        int               can_block  = 0;
+        pl_inode_t *      pinode     = NULL;
+        pl_inode_lock_t * reqlock    = NULL;
+        pl_dom_list_t *   dom        = NULL;
+        char             *res        = NULL;
+        char             *res1       = NULL;
+        char             *conn_id    = NULL;
+        pl_ctx_t         *ctx        = NULL;
+
+        if (xdata)
+                dict_ret = dict_get_str (xdata, "connection-id", &conn_id);
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (inode, unwind);
+        VALIDATE_OR_GOTO (flock, unwind);
+
+        if ((flock->l_start < 0) || (flock->l_len < 0)) {
+                op_errno = EINVAL;
+                goto unwind;
+        }
 
-        pl_trace_in (this, frame, fd, loc, cmd, flock, volume);
+        op_errno = _pl_convert_volume_for_special_range (flock, volume, &res);
+        if (op_errno)
+                goto unwind;
+        if (res)
+                volume = res;
 
-	transport  = frame->root->trans;
-	client_pid = frame->root->pid;
-        owner      = frame->root->lk_owner;
+        pl_trace_in (this, frame, fd, loc, cmd, flock, volume);
 
-	pinode = pl_inode_get (this, inode);
-	if (!pinode) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		op_errno = ENOMEM;
-		goto unwind;
-	}
+        pinode = pl_inode_get (this, inode);
+        if (!pinode) {
+                op_errno = ENOMEM;
+                goto unwind;
+        }
 
-	dom = get_domain (pinode, volume);
+        dom = get_domain (pinode, volume);
+        if (!dom) {
+                op_errno = ENOMEM;
+                goto unwind;
+        }
 
-	if (owner == 0) {
-		/*
+        if (frame->root->lk_owner.len == 0) {
+                /*
                   special case: this means release all locks
-                  from this transport
-		*/
-		gf_log (this->name, GF_LOG_TRACE,
-			"Releasing all locks from transport %p", transport);
-
-		release_inode_locks_of_transport (this, dom, inode, transport);
-		goto unwind;
-	}
-
-	reqlock = new_inode_lock (flock, transport, client_pid, owner, volume);
-
-	if (!reqlock) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		op_ret = -1;
-		op_errno = ENOMEM;
-		goto unwind;
-	}
-
-        reqlock->frame = frame;
-        reqlock->this  = this;
-
-	switch (cmd) {
-	case F_SETLKW:
-		can_block = 1;
-		reqlock->frame = frame;
-		reqlock->this  = this;
-
-		/* fall through */
-
-	case F_SETLK:
-		memcpy (&reqlock->user_flock, flock, sizeof (struct flock));
-		ret = pl_inode_setlk (this, pinode, reqlock,
+                  from this client
+                */
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Releasing all locks from client %p", frame->root->client);
+
+                release_inode_locks_of_client (this, dom, inode, frame->root->client);
+                _pl_convert_volume (volume, &res1);
+                if (res1) {
+                        dom = get_domain (pinode, res1);
+                        if (dom)
+                                release_inode_locks_of_client (this, dom,
+                                                        inode, frame->root->client);
+                }
+
+                op_ret = 0;
+                goto unwind;
+        }
+
+        reqlock = new_inode_lock (flock, frame->root->client, frame->root->pid,
+                                  frame, this, volume, conn_id);
+
+        if (!reqlock) {
+                op_ret = -1;
+                op_errno = ENOMEM;
+                goto unwind;
+        }
+
+
+        switch (cmd) {
+        case F_SETLKW:
+                can_block = 1;
+
+                /* fall through */
+
+        case F_SETLK:
+                memcpy (&reqlock->user_flock, flock, sizeof (struct gf_flock));
+                ret = pl_inode_setlk (this, pinode, reqlock,
                                       can_block, dom);
 
-		if (ret < 0) {
-                        if (can_block) {
+                if (ret < 0) {
+                        if ((can_block) && (F_UNLCK != flock->l_type)) {
                                 pl_trace_block (this, frame, fd, loc,
                                                 cmd, flock, volume);
-				goto out;
+                                goto out;
                         }
-			gf_log (this->name, GF_LOG_TRACE, "returning EAGAIN");
-			op_errno = -ret;
-			__destroy_inode_lock (reqlock);
-			goto unwind;
-		}
-		break;
-
-	default:
-		op_errno = ENOTSUP;
-		gf_log (this->name, GF_LOG_DEBUG,
+                        gf_log (this->name, GF_LOG_TRACE, "returning EAGAIN");
+                        op_errno = -ret;
+                        goto unwind;
+                }
+                break;
+
+        default:
+                op_errno = ENOTSUP;
+                gf_log (this->name, GF_LOG_DEBUG,
                         "Lock command F_GETLK not supported for [f]inodelk "
                         "(cmd=%d)",
                         cmd);
@@ -640,86 +704,103 @@ pl_common_inodelk (call_frame_t *frame, xlator_t *this,
 
         op_ret = 0;
 
+        ctx = pl_ctx_get (frame->root->client, this);
+
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_INFO, "pl_ctx_get() failed");
+                goto unwind;
+        }
+
+        if (flock->l_type == F_UNLCK)
+                pl_del_locker (ctx->ltable, volume, loc, fd,
+                               &frame->root->lk_owner,
+                               GF_FOP_INODELK);
+        else
+                pl_add_locker (ctx->ltable, volume, loc, fd,
+                               frame->root->pid,
+                               &frame->root->lk_owner,
+                               GF_FOP_INODELK);
+
 unwind:
-	if ((inode != NULL) && (flock !=NULL)) {
-		pl_update_refkeeper (this, inode);
-		pl_trace_out (this, frame, fd, loc, cmd, flock, op_ret, op_errno, volume);
-	}
+        if ((inode != NULL) && (flock !=NULL)) {
+                pl_update_refkeeper (this, inode);
+                pl_trace_out (this, frame, fd, loc, cmd, flock, op_ret, op_errno, volume);
+        }
 
-        STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno);
+        STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, NULL);
 out:
+        GF_FREE (res);
+        GF_FREE (res1);
         return 0;
 }
 
 int
 pl_inodelk (call_frame_t *frame, xlator_t *this,
-            const char *volume, loc_t *loc, int32_t cmd, struct flock *flock)
+            const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock,
+            dict_t *xdata)
 {
-
-        pl_common_inodelk (frame, this, volume, loc->inode, cmd, flock, loc, NULL);
+        pl_common_inodelk (frame, this, volume, loc->inode, cmd, flock,
+                           loc, NULL, xdata);
 
         return 0;
 }
 
 int
 pl_finodelk (call_frame_t *frame, xlator_t *this,
-             const char *volume, fd_t *fd, int32_t cmd, struct flock *flock)
+             const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock,
+             dict_t *xdata)
 {
-
-        pl_common_inodelk (frame, this, volume, fd->inode, cmd, flock, NULL, fd);
+        pl_common_inodelk (frame, this, volume, fd->inode, cmd, flock,
+                           NULL, fd, xdata);
 
         return 0;
 
 }
 
+static inline int32_t
+__get_inodelk_dom_count (pl_dom_list_t *dom)
+{
+        pl_inode_lock_t     *lock   = NULL;
+        int32_t             count   = 0;
 
-static int32_t
-__get_inodelk_count (xlator_t *this, pl_inode_t *pl_inode)
+        list_for_each_entry (lock, &dom->inodelk_list, list) {
+                count++;
+        }
+        list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) {
+                count++;
+        }
+        return count;
+}
+
+/* Returns the no. of locks (blocked/granted) held on a given domain name
+ * If @domname is NULL, returns the no. of locks in all the domains present.
+ * If @domname is non-NULL and non-existent, returns 0 */
+int32_t
+__get_inodelk_count (xlator_t *this, pl_inode_t *pl_inode, char *domname)
 {
         int32_t            count  = 0;
-        pl_inode_lock_t   *lock   = NULL;
         pl_dom_list_t     *dom    = NULL;
 
         list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
-                list_for_each_entry (lock, &dom->inodelk_list, list) {
-
-			gf_log (this->name, GF_LOG_DEBUG,
-                                " XATTR DEBUG"
-				" domain: %s %s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" "
-                                "state = Active",
-                                dom->domain,
-				lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
-				lock->client_pid,
-                                lock->owner,
-				lock->user_flock.l_start,
-				lock->user_flock.l_len);
-
-                        count++;
-                }
-
-                list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) {
+                if (domname) {
+                        if (strcmp (domname, dom->domain) == 0) {
+                                count = __get_inodelk_dom_count (dom);
+                                goto out;
+                        }
 
-			gf_log (this->name, GF_LOG_DEBUG,
-                                " XATTR DEBUG"
-				" domain: %s %s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" "
-                                "state = Blocked",
-                                dom->domain,
-				lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
-				lock->client_pid,
-                                lock->owner,
-				lock->user_flock.l_start,
-				lock->user_flock.l_len);
+                } else {
+                    /* Counting locks from all domains */
+                        count += __get_inodelk_dom_count (dom);
 
-                        count++;
                 }
-
         }
 
+out:
         return count;
 }
 
 int32_t
-get_inodelk_count (xlator_t *this, inode_t *inode)
+get_inodelk_count (xlator_t *this, inode_t *inode, char *domname)
 {
         pl_inode_t   *pl_inode = NULL;
         uint64_t      tmp_pl_inode = 0;
@@ -735,7 +816,7 @@ get_inodelk_count (xlator_t *this, inode_t *inode)
 
         pthread_mutex_lock (&pl_inode->mutex);
         {
-                count = __get_inodelk_count (this, pl_inode);
+                count = __get_inodelk_count (this, pl_inode, domname);
         }
         pthread_mutex_unlock (&pl_inode->mutex);
 
diff --git a/xlators/features/locks/src/locks-mem-types.h b/xlators/features/locks/src/locks-mem-types.h
index cf5024086..08aeb0a79 100644
--- a/xlators/features/locks/src/locks-mem-types.h
+++ b/xlators/features/locks/src/locks-mem-types.h
@@ -1,23 +1,13 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
 
-
 #ifndef __LOCKS_MEM_TYPES_H__
 #define __LOCKS_MEM_TYPES_H__
 
@@ -32,7 +22,7 @@ enum gf_locks_mem_types_ {
         gf_locks_mt_truncate_ops,
         gf_locks_mt_pl_rw_req_t,
         gf_locks_mt_posix_locks_private_t,
-        gf_locks_mt_pl_local_t,
+        gf_locks_mt_pl_fdctx_t,
         gf_locks_mt_end
 };
 #endif
diff --git a/xlators/features/locks/src/locks.h b/xlators/features/locks/src/locks.h
index 60474615e..76fc941d7 100644
--- a/xlators/features/locks/src/locks.h
+++ b/xlators/features/locks/src/locks.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef __POSIX_LOCKS_H__
 #define __POSIX_LOCKS_H__
 
@@ -29,6 +19,9 @@
 #include "stack.h"
 #include "call-stub.h"
 #include "locks-mem-types.h"
+#include "client_t.h"
+
+#include "lkowner.h"
 
 struct __pl_fd;
 
@@ -40,24 +33,29 @@ struct __posix_lock {
         off_t              fl_end;
 
         short              blocked;    /* waiting to acquire */
-        struct flock       user_flock; /* the flock supplied by the user */
+        struct gf_flock    user_flock; /* the flock supplied by the user */
         xlator_t          *this;       /* required for blocked locks */
         unsigned long      fd_num;
 
+        fd_t              *fd;
         call_frame_t      *frame;
 
+        struct timeval     blkd_time;   /*time at which lock was queued into blkd list*/
+        struct timeval     granted_time; /*time at which lock was queued into active list*/
+
         /* These two together serve to uniquely identify each process
            across nodes */
 
-        void              *transport;     /* to identify client node */
+        void              *client;     /* to identify client node */
+        gf_lkowner_t       owner;
         pid_t              client_pid;    /* pid of client process */
-        uint64_t           owner;         /* lock owner from fuse */
 };
 typedef struct __posix_lock posix_lock_t;
 
 struct __pl_inode_lock {
         struct list_head   list;
         struct list_head   blocked_locks; /* list_head pointing to blocked_inodelks */
+        int                ref;
 
         short              fl_type;
         off_t              fl_start;
@@ -65,18 +63,23 @@ struct __pl_inode_lock {
 
         const char        *volume;
 
-        struct flock       user_flock; /* the flock supplied by the user */
+        struct gf_flock    user_flock; /* the flock supplied by the user */
         xlator_t          *this;       /* required for blocked locks */
         fd_t              *fd;
 
         call_frame_t      *frame;
 
+        struct timeval     blkd_time;   /*time at which lock was queued into blkd list*/
+        struct timeval     granted_time; /*time at which lock was queued into active list*/
+
         /* These two together serve to uniquely identify each process
            across nodes */
 
-        void              *transport;     /* to identify client node */
+        void              *client;     /* to identify client node */
+        gf_lkowner_t       owner;
         pid_t              client_pid;    /* pid of client process */
-        uint64_t           owner;
+
+        char              *connection_id; /* stores the client connection id */
 };
 typedef struct __pl_inode_lock pl_inode_lock_t;
 
@@ -109,9 +112,14 @@ struct __entry_lock {
         const char       *basename;
         entrylk_type      type;
 
-        void      *trans;
-        pid_t             client_pid;    /* pid of client process */
-        uint64_t          owner;
+        struct timeval     blkd_time;   /*time at which lock was queued into blkd list*/
+        struct timeval     granted_time; /*time at which lock was queued into active list*/
+
+        void             *trans;
+        gf_lkowner_t      owner;
+	pid_t             client_pid;    /* pid of client process */
+
+        char             *connection_id; /* stores the client connection id */
 };
 typedef struct __entry_lock pl_entry_lock_t;
 
@@ -125,6 +133,9 @@ struct __pl_inode {
         struct list_head dom_list;       /* list of domains */
         struct list_head ext_list;       /* list of fcntl locks */
         struct list_head rw_list;        /* list of waiting r/w requests */
+        struct list_head reservelk_list;        /* list of reservelks */
+        struct list_head blocked_reservelks;        /* list of blocked reservelks */
+        struct list_head blocked_calls;  /* List of blocked lock calls while a reserve is held*/
         int              mandatory;      /* if mandatory locking is enabled */
 
         inode_t          *refkeeper;     /* hold refs on an inode while locks are
@@ -142,12 +153,40 @@ typedef struct __pl_fd pl_fd_t;
 typedef struct {
         gf_boolean_t    mandatory;      /* if mandatory locking is enabled */
         gf_boolean_t    trace;          /* trace lock requests in and out */
+        char           *brickname;
 } posix_locks_private_t;
 
+
 typedef struct {
         gf_boolean_t   entrylk_count_req;
         gf_boolean_t   inodelk_count_req;
+        gf_boolean_t   inodelk_dom_count_req;
         gf_boolean_t   posixlk_count_req;
+        gf_boolean_t   parent_entrylk_req;
+
+        /* used by {f,}truncate */
+        loc_t  loc;
+        fd_t  *fd;
+        off_t  offset;
+        dict_t *xdata;
+        enum {TRUNCATE, FTRUNCATE} op;
 } pl_local_t;
 
+
+typedef struct {
+        struct list_head locks_list;
+} pl_fdctx_t;
+
+
+typedef struct _locks_ctx {
+        gf_lock_t            ltable_lock; /* only for replace, 
+                                             ltable has its own internal 
+                                             lock for operations */
+        struct _lock_table  *ltable;
+} pl_ctx_t;
+
+
+pl_ctx_t *
+pl_ctx_get (client_t *client, xlator_t *xlator);
+
 #endif /* __POSIX_LOCKS_H__ */
diff --git a/xlators/features/locks/src/posix.c b/xlators/features/locks/src/posix.c
index 165280265..7bfb38a51 100644
--- a/xlators/features/locks/src/posix.c
+++ b/xlators/features/locks/src/posix.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #include <unistd.h>
 #include <fcntl.h>
 #include <limits.h>
@@ -37,6 +27,9 @@
 #include "locks.h"
 #include "common.h"
 #include "statedump.h"
+#include "clear.h"
+#include "defaults.h"
+#include "syncop.h"
 
 #ifndef LLONG_MAX
 #define LLONG_MAX LONG_LONG_MAX /* compat with old gcc */
@@ -47,37 +40,87 @@
 
 void do_blocked_rw (pl_inode_t *);
 static int __rw_allowable (pl_inode_t *, posix_lock_t *, glusterfs_fop_t);
+static int format_brickname(char *);
+int pl_lockinfo_get_brickname (xlator_t *, inode_t *, int32_t *);
+static int fetch_pathinfo(xlator_t *, inode_t *, int32_t *, char **);
 
-struct _truncate_ops {
-        loc_t  loc;
-        fd_t  *fd;
-        off_t  offset;
-        enum {TRUNCATE, FTRUNCATE} op;
-};
+static pl_fdctx_t *
+pl_new_fdctx ()
+{
+        pl_fdctx_t *fdctx = NULL;
+
+        fdctx = GF_CALLOC (1, sizeof (*fdctx),
+                           gf_locks_mt_pl_fdctx_t);
+        GF_VALIDATE_OR_GOTO ("posix-locks", fdctx, out);
+
+        INIT_LIST_HEAD (&fdctx->locks_list);
+
+out:
+        return fdctx;
+}
+
+static pl_fdctx_t *
+pl_check_n_create_fdctx (xlator_t *this, fd_t *fd)
+{
+        int         ret   = 0;
+        uint64_t    tmp   = 0;
+        pl_fdctx_t *fdctx = NULL;
 
+        GF_VALIDATE_OR_GOTO ("posix-locks", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, fd, out);
+
+        LOCK (&fd->lock);
+        {
+                ret = __fd_ctx_get (fd, this, &tmp);
+                if ((ret != 0) || (tmp == 0)) {
+                        fdctx = pl_new_fdctx ();
+                        if (fdctx == NULL) {
+                                goto unlock;
+                        }
+                }
+
+                ret = __fd_ctx_set (fd, this, (uint64_t)(long)fdctx);
+                if (ret != 0) {
+                        GF_FREE (fdctx);
+                        fdctx = NULL;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "failed to set fd ctx");
+                }
+        }
+unlock:
+        UNLOCK (&fd->lock);
+
+out:
+        return fdctx;
+}
 
 int
 pl_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                 struct iatt *postbuf)
+                 struct iatt *postbuf, dict_t *xdata)
 {
-        struct _truncate_ops *local = NULL;
+        pl_local_t *local = NULL;
 
         local = frame->local;
 
         if (local->op == TRUNCATE)
                 loc_wipe (&local->loc);
 
+        if (local->xdata)
+                dict_unref (local->xdata);
+        if (local->fd)
+                fd_unref (local->fd);
+
         STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno,
-                             prebuf, postbuf);
+                             prebuf, postbuf, xdata);
         return 0;
 }
 
 
 static int
 truncate_allowed (pl_inode_t *pl_inode,
-                  void *transport, pid_t client_pid,
-                  uint64_t owner, off_t offset)
+                  client_t *client, pid_t client_pid,
+                  gf_lkowner_t *owner, off_t offset)
 {
         posix_lock_t *l = NULL;
         posix_lock_t  region = {.list = {0, }, };
@@ -85,9 +128,9 @@ truncate_allowed (pl_inode_t *pl_inode,
 
         region.fl_start   = offset;
         region.fl_end     = LLONG_MAX;
-        region.transport  = transport;
+        region.client     = client;
         region.client_pid = client_pid;
-        region.owner      = owner;
+        region.owner      = *owner;
 
         pthread_mutex_lock (&pl_inode->mutex);
         {
@@ -96,6 +139,8 @@ truncate_allowed (pl_inode_t *pl_inode,
                             && locks_overlap (&region, l)
                             && !same_owner (&region, l)) {
                                 ret = 0;
+                                gf_log ("posix-locks", GF_LOG_TRACE, "Truncate "
+                                        "allowed");
                                 break;
                         }
                 }
@@ -108,10 +153,11 @@ truncate_allowed (pl_inode_t *pl_inode,
 
 static int
 truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                   int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                   dict_t *xdata)
 {
         posix_locks_private_t *priv = NULL;
-        struct _truncate_ops  *local = NULL;
+        pl_local_t            *local = NULL;
         inode_t               *inode = NULL;
         pl_inode_t            *pl_inode = NULL;
 
@@ -133,8 +179,6 @@ truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         pl_inode = pl_inode_get (this, inode);
         if (!pl_inode) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
                 op_ret   = -1;
                 op_errno = ENOMEM;
                 goto unwind;
@@ -142,8 +186,8 @@ truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         if (priv->mandatory
             && pl_inode->mandatory
-            && !truncate_allowed (pl_inode, frame->root->trans,
-                                  frame->root->pid, frame->root->lk_owner,
+            && !truncate_allowed (pl_inode, frame->root->client,
+                                  frame->root->pid, &frame->root->lk_owner,
                                   local->offset)) {
                 op_ret   = -1;
                 op_errno = EAGAIN;
@@ -154,53 +198,58 @@ truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         case TRUNCATE:
                 STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this),
                             FIRST_CHILD (this)->fops->truncate,
-                            &local->loc, local->offset);
+                            &local->loc, local->offset, local->xdata);
                 break;
         case FTRUNCATE:
                 STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this),
                             FIRST_CHILD (this)->fops->ftruncate,
-                            local->fd, local->offset);
+                            local->fd, local->offset, local->xdata);
                 break;
         }
 
         return 0;
 
 unwind:
+        gf_log (this->name, GF_LOG_ERROR, "truncate failed with ret: %d, "
+                "error: %s", op_ret, strerror (op_errno));
         if (local->op == TRUNCATE)
                 loc_wipe (&local->loc);
+        if (local->xdata)
+                dict_unref (local->xdata);
+        if (local->fd)
+                fd_unref (local->fd);
 
-        STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, buf, NULL);
+        STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, buf, NULL, xdata);
         return 0;
 }
 
 
 int
 pl_truncate (call_frame_t *frame, xlator_t *this,
-             loc_t *loc, off_t offset)
+             loc_t *loc, off_t offset, dict_t *xdata)
 {
-        struct _truncate_ops *local = NULL;
+        pl_local_t *local = NULL;
 
-	local = GF_CALLOC (1, sizeof (struct _truncate_ops),
-                           gf_locks_mt_truncate_ops);
-        if (!local) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
-                goto unwind;
-        }
+        local = mem_get0 (this->local_pool);
+        GF_VALIDATE_OR_GOTO (this->name, local, unwind);
 
         local->op         = TRUNCATE;
         local->offset     = offset;
         loc_copy (&local->loc, loc);
+        if (xdata)
+                local->xdata = dict_ref (xdata);
 
         frame->local = local;
 
         STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, loc);
+                    FIRST_CHILD (this)->fops->stat, loc, NULL);
 
         return 0;
 
 unwind:
-        STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM, NULL, NULL);
+        gf_log (this->name, GF_LOG_ERROR, "truncate for %s failed with ret: %d, "
+                "error: %s", loc->path, -1, strerror (ENOMEM));
+        STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
 
         return 0;
 }
@@ -208,34 +257,54 @@ unwind:
 
 int
 pl_ftruncate (call_frame_t *frame, xlator_t *this,
-              fd_t *fd, off_t offset)
+              fd_t *fd, off_t offset, dict_t *xdata)
 {
-        struct _truncate_ops *local = NULL;
+        pl_local_t *local = NULL;
 
-	local = GF_CALLOC (1, sizeof (struct _truncate_ops),
-                           gf_locks_mt_truncate_ops);
-        if (!local) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
-                goto unwind;
-        }
+        local = mem_get0 (this->local_pool);
+        GF_VALIDATE_OR_GOTO (this->name, local, unwind);
 
         local->op         = FTRUNCATE;
         local->offset     = offset;
-        local->fd         = fd;
+        local->fd         = fd_ref (fd);
+        if (xdata)
+                local->xdata = dict_ref (xdata);
 
         frame->local = local;
 
         STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->fstat, fd);
+                    FIRST_CHILD(this)->fops->fstat, fd, xdata);
         return 0;
 
 unwind:
-        STACK_UNWIND_STRICT (ftruncate, frame, -1, ENOMEM, NULL, NULL);
+        gf_log (this->name, GF_LOG_ERROR, "ftruncate failed with ret: %d, "
+                "error: %s", -1, strerror (ENOMEM));
+        STACK_UNWIND_STRICT (ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
 
         return 0;
 }
 
+int
+pl_locks_by_fd (pl_inode_t *pl_inode, fd_t *fd)
+{
+       posix_lock_t *l = NULL;
+       int found = 0;
+
+       pthread_mutex_lock (&pl_inode->mutex);
+       {
+
+               list_for_each_entry (l, &pl_inode->ext_list, list) {
+                       if ((l->fd_num == fd_to_fdnum(fd))) {
+                               found = 1;
+                               break;
+                       }
+               }
+
+       }
+       pthread_mutex_unlock (&pl_inode->mutex);
+       return found;
+}
+
 static void
 delete_locks_of_fd (xlator_t *this, pl_inode_t *pl_inode, fd_t *fd)
 {
@@ -265,7 +334,8 @@ delete_locks_of_fd (xlator_t *this, pl_inode_t *pl_inode, fd_t *fd)
 
        list_for_each_entry_safe (l, tmp, &blocked_list, list) {
                list_del_init(&l->list);
-               STACK_UNWIND_STRICT (lk, l->frame, -1, EAGAIN, &l->user_flock);
+               STACK_UNWIND_STRICT (lk, l->frame, -1, EAGAIN, &l->user_flock,
+                                    NULL);
                __destroy_lock (l);
        }
 
@@ -277,7 +347,7 @@ delete_locks_of_fd (xlator_t *this, pl_inode_t *pl_inode, fd_t *fd)
 
 static void
 __delete_locks_of_owner (pl_inode_t *pl_inode,
-                         void *transport, uint64_t owner)
+                         client_t *client, gf_lkowner_t *owner)
 {
         posix_lock_t *tmp = NULL;
         posix_lock_t *l = NULL;
@@ -285,16 +355,18 @@ __delete_locks_of_owner (pl_inode_t *pl_inode,
         /* TODO: what if it is a blocked lock with pending l->frame */
 
         list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
-                if ((l->transport == transport)
-                    && (l->owner == owner)) {
-			gf_log ("posix-locks", GF_LOG_TRACE,
+                if (l->blocked)
+                        continue;
+                if ((l->client == client) &&
+                    is_same_lkowner (&l->owner, owner)) {
+                        gf_log ("posix-locks", GF_LOG_TRACE,
                                 " Flushing lock"
-				"%s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" state: %s",
-				l->fl_type == F_UNLCK ? "Unlock" : "Lock",
-				l->client_pid,
-                                l->owner,
-				l->user_flock.l_start,
-				l->user_flock.l_len,
+                                "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" state: %s",
+                                l->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                                l->client_pid,
+                                lkowner_utoa (&l->owner),
+                                l->user_flock.l_start,
+                                l->user_flock.l_len,
                                 l->blocked == 1 ? "Blocked" : "Active");
 
                         __delete_lock (pl_inode, l);
@@ -305,52 +377,578 @@ __delete_locks_of_owner (pl_inode_t *pl_inode,
         return;
 }
 
+
+int32_t
+pl_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+
+}
+
+int32_t
+pl_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             const char *name, dict_t *xdata)
+{
+        int32_t                 op_errno        = EINVAL;
+        int                     op_ret          = -1;
+        int32_t                 bcount          = 0;
+        int32_t                 gcount          = 0;
+        char                    key[PATH_MAX]   = {0, };
+        char                    *lk_summary     = NULL;
+        pl_inode_t              *pl_inode       = NULL;
+        dict_t                  *dict           = NULL;
+        clrlk_args              args            = {0,};
+        char                    *brickname      = NULL;
+
+        if (!name)
+                goto usual;
+
+        if (strncmp (name, GF_XATTR_CLRLK_CMD, strlen (GF_XATTR_CLRLK_CMD)))
+                goto usual;
+
+        if (clrlk_parse_args (name, &args)) {
+                op_errno = EINVAL;
+                goto out;
+        }
+
+        dict = dict_new ();
+        if (!dict) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        pl_inode = pl_inode_get (this, loc->inode);
+        if (!pl_inode) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        switch (args.type) {
+                case CLRLK_INODE:
+                case CLRLK_ENTRY:
+                        op_ret = clrlk_clear_lks_in_all_domains (this, pl_inode,
+                                                                 &args, &bcount,
+                                                                 &gcount,
+                                                                 &op_errno);
+                        if (op_ret)
+                                goto out;
+                        break;
+                case CLRLK_POSIX:
+                        op_ret = clrlk_clear_posixlk (this, pl_inode, &args,
+                                                      &bcount, &gcount,
+                                                      &op_errno);
+                        if (op_ret)
+                                goto out;
+                        break;
+                case CLRLK_TYPE_MAX:
+                        op_errno = EINVAL;
+                        goto out;
+        }
+
+        op_ret = fetch_pathinfo (this, loc->inode, &op_errno, &brickname);
+        if (op_ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Couldn't get brickname");
+        } else {
+                op_ret = format_brickname(brickname);
+                if (op_ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Couldn't format brickname");
+                        GF_FREE(brickname);
+                        brickname = NULL;
+                }
+        }
+
+        if (!gcount && !bcount) {
+                if (gf_asprintf (&lk_summary, "No locks cleared.") == -1) {
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        goto out;
+                }
+        } else if (gf_asprintf (&lk_summary, "%s: %s blocked locks=%d "
+                                "granted locks=%d",
+                                (brickname == NULL)? this->name : brickname,
+                                (args.type == CLRLK_INODE)? "inode":
+                                (args.type == CLRLK_ENTRY)? "entry":
+                                (args.type == CLRLK_POSIX)? "posix": " ",
+                                bcount, gcount) == -1) {
+                op_ret = -1;
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        strncpy (key, name, strlen (name));
+        if (dict_set_dynstr (dict, key, lk_summary)) {
+                op_ret = -1;
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        op_ret = 0;
+out:
+        GF_FREE(brickname);
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+
+        GF_FREE (args.opts);
+        if (op_ret && lk_summary)
+                GF_FREE (lk_summary);
+        if (dict)
+                dict_unref (dict);
+        return 0;
+
+usual:
+        STACK_WIND (frame, pl_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+        return 0;
+}
+
+static int
+format_brickname(char *brickname)
+{
+       int   ret             = -1;
+       char *hostname        = NULL;
+       char *volume          = NULL;
+       char *saveptr         = NULL;
+
+       if (!brickname)
+               goto out;
+
+       strtok_r(brickname, ":", &saveptr);
+       hostname = gf_strdup(strtok_r(NULL, ":", &saveptr));
+       if (hostname == NULL)
+               goto out;
+       volume = gf_strdup(strtok_r(NULL, ".", &saveptr));
+       if (volume == NULL)
+               goto out;
+
+       sprintf(brickname, "%s:%s", hostname, volume);
+
+       ret = 0;
+out:
+       GF_FREE(hostname);
+       GF_FREE(volume);
+       return ret;
+}
+
+static int
+fetch_pathinfo (xlator_t *this, inode_t *inode, int32_t *op_errno,
+                char **brickname)
+{
+        int                    ret       = -1;
+        loc_t                  loc       = {0, };
+        dict_t                *dict      = NULL;
+
+        if (!brickname)
+                goto out;
+
+        if (!op_errno)
+                goto out;
+
+        uuid_copy (loc.gfid, inode->gfid);
+        loc.inode = inode_ref (inode);
+
+        ret = syncop_getxattr (FIRST_CHILD(this), &loc, &dict,
+                               GF_XATTR_PATHINFO_KEY);
+        if (ret < 0) {
+                *op_errno = errno;
+                goto out;
+        }
+
+        ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, brickname);
+        if (ret)
+                goto out;
+
+        *brickname = gf_strdup(*brickname);
+        if (*brickname == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if (dict != NULL) {
+                dict_unref (dict);
+        }
+        loc_wipe(&loc);
+
+        return ret;
+}
+
+
+int
+pl_lockinfo_get_brickname (xlator_t *this, inode_t *inode, int32_t *op_errno)
+{
+        int                    ret       = -1;
+        posix_locks_private_t *priv      = NULL;
+        char                  *brickname = NULL;
+        char                  *end       = NULL;
+        char                  *tmp       = NULL;
+
+        priv = this->private;
+
+        ret = fetch_pathinfo (this, inode, op_errno, &brickname);
+        if (ret)
+                goto out;
+
+        end = strrchr (brickname, ':');
+        if (!end) {
+                GF_FREE(brickname);
+                ret = -1;
+                goto out;
+        }
+
+        tmp = brickname;
+        brickname = gf_strndup (brickname, (end - brickname));
+        if (brickname == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        priv->brickname = brickname;
+        ret = 0;
+out:
+        GF_FREE(tmp);
+        return ret;
+}
+
+char *
+pl_lockinfo_key (xlator_t *this, inode_t *inode, int32_t *op_errno)
+{
+        posix_locks_private_t *priv = NULL;
+        char                  *key  = NULL;
+        int                    ret = 0;
+
+        priv = this->private;
+
+        if (priv->brickname == NULL) {
+                ret = pl_lockinfo_get_brickname (this, inode, op_errno);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "cannot get brickname");
+                        goto out;
+                }
+        }
+
+        key = priv->brickname;
+out:
+        return key;
+}
+
+int32_t
+pl_fgetxattr_handle_lockinfo (xlator_t *this, fd_t *fd,
+                              dict_t *dict, int32_t *op_errno)
+{
+        pl_inode_t    *pl_inode = NULL;
+        char          *key      = NULL, *buf = NULL;
+        int32_t        op_ret   = 0;
+        unsigned long  fdnum    = 0, len = 0;
+        dict_t        *tmp      = NULL;
+
+        pl_inode = pl_inode_get (this, fd->inode);
+
+        if (!pl_inode) {
+                gf_log (this->name, GF_LOG_DEBUG, "Could not get inode.");
+                *op_errno = EBADFD;
+                op_ret = -1;
+                goto out;
+        }
+
+        if (!pl_locks_by_fd (pl_inode, fd)) {
+                op_ret = 0;
+                goto out;
+        }
+
+        fdnum = fd_to_fdnum (fd);
+
+        key = pl_lockinfo_key (this, fd->inode, op_errno);
+        if (key == NULL) {
+                op_ret = -1;
+                goto out;
+        }
+
+        tmp = dict_new ();
+        if (tmp == NULL) {
+                op_ret = -1;
+                *op_errno = ENOMEM;
+                goto out;
+        }
+
+        op_ret = dict_set_uint64 (tmp, key, fdnum);
+        if (op_ret < 0) {
+                *op_errno = -op_ret;
+                op_ret = -1;
+                gf_log (this->name, GF_LOG_WARNING, "setting lockinfo value "
+                        "(%lu) for fd (ptr:%p inode-gfid:%s) failed (%s)",
+                        fdnum, fd, uuid_utoa (fd->inode->gfid),
+                        strerror (*op_errno));
+                goto out;
+        }
+
+        len = dict_serialized_length (tmp);
+        if (len < 0) {
+                *op_errno = -op_ret;
+                op_ret = -1;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dict_serialized_length failed (%s) while handling "
+                        "lockinfo for fd (ptr:%p inode-gfid:%s)",
+                        strerror (*op_errno), fd, uuid_utoa (fd->inode->gfid));
+                goto out;
+        }
+
+        buf = GF_CALLOC (1, len, gf_common_mt_char);
+        if (buf == NULL) {
+                op_ret = -1;
+                *op_errno = ENOMEM;
+                goto out;
+        }
+
+        op_ret = dict_serialize (tmp, buf);
+        if (op_ret < 0) {
+                *op_errno = -op_ret;
+                op_ret = -1;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dict_serialize failed (%s) while handling lockinfo "
+                        "for fd (ptr: %p inode-gfid:%s)", strerror (*op_errno),
+                        fd, uuid_utoa (fd->inode->gfid));
+                goto out;
+        }
+
+        op_ret = dict_set_dynptr (dict, GF_XATTR_LOCKINFO_KEY, buf, len);
+        if (op_ret < 0) {
+                *op_errno = -op_ret;
+                op_ret = -1;
+                gf_log (this->name, GF_LOG_WARNING, "setting lockinfo value "
+                        "(%lu) for fd (ptr:%p inode-gfid:%s) failed (%s)",
+                        fdnum, fd, uuid_utoa (fd->inode->gfid),
+                        strerror (*op_errno));
+                goto out;
+        }
+
+        buf = NULL;
+out:
+        if (tmp != NULL) {
+                dict_unref (tmp);
+        }
+
+        if (buf != NULL) {
+                GF_FREE (buf);
+        }
+
+        return op_ret;
+}
+
+
+int32_t
+pl_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+              const char *name, dict_t *xdata)
+{
+        int32_t  op_ret = 0, op_errno = 0;
+        dict_t  *dict   = NULL;
+
+        if (!name) {
+                goto usual;
+        }
+
+        if (strcmp (name, GF_XATTR_LOCKINFO_KEY) == 0) {
+                dict = dict_new ();
+                if (dict == NULL) {
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        goto unwind;
+                }
+
+                op_ret = pl_fgetxattr_handle_lockinfo (this, fd, dict,
+                                                       &op_errno);
+                if (op_ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "getting lockinfo on fd (ptr:%p inode-gfid:%s) "
+                                "failed (%s)", fd, uuid_utoa (fd->inode->gfid),
+                                strerror (op_errno));
+                }
+
+                goto unwind;
+        } else {
+                goto usual;
+        }
+
+unwind:
+        STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, NULL);
+        if (dict != NULL) {
+                dict_unref (dict);
+        }
+
+        return 0;
+
+usual:
+        STACK_WIND (frame, default_fgetxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+        return 0;
+}
+
+int32_t
+pl_migrate_locks (call_frame_t *frame, fd_t *newfd, uint64_t oldfd_num,
+                  int32_t *op_errno)
+{
+        pl_inode_t   *pl_inode  = NULL;
+        uint64_t      newfd_num = 0;
+        posix_lock_t *l         = NULL;
+        int32_t       op_ret    = 0;
+
+        newfd_num = fd_to_fdnum (newfd);
+
+        pl_inode = pl_inode_get (frame->this, newfd->inode);
+        if (pl_inode == NULL) {
+                op_ret = -1;
+                *op_errno = EBADFD;
+                goto out;
+        }
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry (l, &pl_inode->ext_list, list) {
+                        if (l->fd_num == oldfd_num) {
+                                l->fd_num = newfd_num;
+                                l->client = frame->root->client;
+                        }
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        op_ret = 0;
+out:
+        return op_ret;
+}
+
+int32_t
+pl_fsetxattr_handle_lockinfo (call_frame_t *frame, fd_t *fd, char *lockinfo_buf,
+                              int len, int32_t *op_errno)
+{
+        int32_t   op_ret    = -1;
+        dict_t   *lockinfo  = NULL;
+        uint64_t  oldfd_num = 0;
+        char     *key       = NULL;
+
+        lockinfo = dict_new ();
+        if (lockinfo == NULL) {
+                op_ret = -1;
+                *op_errno = ENOMEM;
+                goto out;
+        }
+
+        op_ret = dict_unserialize (lockinfo_buf, len, &lockinfo);
+        if (op_ret < 0) {
+                *op_errno = -op_ret;
+                op_ret = -1;
+                goto out;
+        }
+
+        key = pl_lockinfo_key (frame->this, fd->inode, op_errno);
+        if (key == NULL) {
+                op_ret = -1;
+                goto out;
+        }
+
+        op_ret = dict_get_uint64 (lockinfo, key, &oldfd_num);
+
+        if (oldfd_num == 0) {
+                op_ret = 0;
+                goto out;
+        }
+
+        op_ret = pl_migrate_locks (frame, fd, oldfd_num, op_errno);
+        if (op_ret < 0) {
+                gf_log (frame->this->name, GF_LOG_WARNING,
+                        "migration of locks from oldfd (ptr:%p) to newfd "
+                        "(ptr:%p) (inode-gfid:%s)", (void *)oldfd_num, fd,
+                        uuid_utoa (fd->inode->gfid));
+                goto out;
+        }
+
+out:
+        dict_unref (lockinfo);
+
+        return op_ret;
+}
+
+int32_t
+pl_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+              int32_t flags, dict_t *xdata)
+{
+        int32_t  op_ret       = 0, op_errno = 0;
+        void    *lockinfo_buf = NULL;
+        int      len          = 0;
+
+        op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
+                                       &lockinfo_buf, &len);
+        if (lockinfo_buf == NULL) {
+                goto usual;
+        }
+
+        op_ret = pl_fsetxattr_handle_lockinfo (frame, fd, lockinfo_buf, len,
+                                               &op_errno);
+        if (op_ret < 0) {
+                goto unwind;
+        }
+
+usual:
+        STACK_WIND (frame, default_fsetxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+        return 0;
+
+unwind:
+        STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, NULL);
+        return 0;
+}
+
 int32_t
 pl_opendir_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-		     fd_t *fd)
+                void *cookie,
+                xlator_t *this,
+                int32_t op_ret,
+                int32_t op_errno,
+                fd_t *fd, dict_t *xdata)
 {
-        int dummy = 1;
-        int ret = -1;
+        pl_fdctx_t *fdctx = NULL;
 
         if (op_ret < 0)
                 goto unwind;
 
-        ret = fd_ctx_set (fd, this, dummy);
-        if (ret != 0)
-                gf_log (this->name, GF_LOG_ERROR,
-                        "setting context for fd=%p in locks failed.", fd);
+        fdctx = pl_check_n_create_fdctx (this, fd);
+        if (!fdctx) {
+                op_errno = ENOMEM;
+                op_ret   = -1;
+                goto unwind;
+        }
 
 unwind:
-	STACK_UNWIND_STRICT (opendir,
+        STACK_UNWIND_STRICT (opendir,
                              frame,
                              op_ret,
                              op_errno,
-                             fd);
-	return 0;
+                             fd, xdata);
+        return 0;
 }
 
-int32_t 
+int32_t
 pl_opendir (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, fd_t *fd)
+             loc_t *loc, fd_t *fd, dict_t *xdata)
 {
-	STACK_WIND (frame,
-		    pl_opendir_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->opendir,
-		    loc, fd);
-	return 0;
+        STACK_WIND (frame,
+                    pl_opendir_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->opendir,
+                    loc, fd, xdata);
+        return 0;
 
 }
 
 int
 pl_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-              int32_t op_ret, int32_t op_errno)
+              int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno);
+        STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
 
         return 0;
 }
@@ -358,38 +956,35 @@ pl_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 int
 pl_flush (call_frame_t *frame, xlator_t *this,
-          fd_t *fd)
+          fd_t *fd, dict_t *xdata)
 {
-        pl_inode_t            *pl_inode = NULL;
-        uint64_t              owner     = -1;
-
-        owner = frame->root->lk_owner;
+        pl_inode_t *pl_inode = NULL;
 
         pl_inode = pl_inode_get (this, fd->inode);
 
         if (!pl_inode) {
                 gf_log (this->name, GF_LOG_DEBUG, "Could not get inode.");
-                STACK_UNWIND_STRICT (flush, frame, -1, EBADFD);
+                STACK_UNWIND_STRICT (flush, frame, -1, EBADFD, NULL);
                 return 0;
         }
 
         pl_trace_flush (this, frame, fd);
 
-        if (owner == 0) {
+        if (frame->root->lk_owner.len == 0) {
                 /* Handle special case when protocol/server sets lk-owner to zero.
                  * This usually happens due to a client disconnection. Hence, free
                  * all locks opened with this fd.
                  */
                 gf_log (this->name, GF_LOG_TRACE,
-			"Releasing all locks with fd %p", fd);
+                        "Releasing all locks with fd %p", fd);
                 delete_locks_of_fd (this, pl_inode, fd);
                 goto wind;
 
         }
         pthread_mutex_lock (&pl_inode->mutex);
         {
-                __delete_locks_of_owner (pl_inode, frame->root->trans,
-                                         owner);
+                __delete_locks_of_owner (pl_inode, frame->root->client,
+                                         &frame->root->lk_owner);
         }
         pthread_mutex_unlock (&pl_inode->mutex);
 
@@ -399,28 +994,29 @@ pl_flush (call_frame_t *frame, xlator_t *this,
 
 wind:
         STACK_WIND (frame, pl_flush_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->flush, fd);
+                    FIRST_CHILD(this)->fops->flush, fd, xdata);
         return 0;
 }
 
 
 int
 pl_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-             int32_t op_ret, int32_t op_errno, fd_t *fd)
+             int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
-        int dummy = 1;
-        int ret = -1;
+        pl_fdctx_t *fdctx = NULL;
 
         if (op_ret < 0)
                 goto unwind;
 
-        ret = fd_ctx_set (fd, this, dummy);
-        if (ret != 0)
-                gf_log (this->name, GF_LOG_ERROR,
-                        "setting context for fd=%p in locks failed.", fd);
+        fdctx = pl_check_n_create_fdctx (this, fd);
+        if (!fdctx) {
+                op_errno = ENOMEM;
+                op_ret   = -1;
+                goto unwind;
+        }
 
 unwind:
-        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd);
+        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
 
         return 0;
 }
@@ -428,12 +1024,11 @@ unwind:
 
 int
 pl_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-         fd_t *fd, int32_t wbflags)
+         fd_t *fd, dict_t *xdata)
 {
-        /* why isn't O_TRUNC being handled ? */
         STACK_WIND (frame, pl_open_cbk,
                     FIRST_CHILD(this), FIRST_CHILD(this)->fops->open,
-                    loc, flags & ~O_TRUNC, fd, wbflags);
+                    loc, flags, fd, xdata);
 
         return 0;
 }
@@ -443,22 +1038,23 @@ int
 pl_create_cbk (call_frame_t *frame, void *cookie,
                xlator_t *this, int32_t op_ret, int32_t op_errno,
                fd_t *fd, inode_t *inode, struct iatt *buf,
-               struct iatt *preparent, struct iatt *postparent)
+               struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-        int dummy = 1;
-        int ret = -1;
+        pl_fdctx_t *fdctx = NULL;
 
         if (op_ret < 0)
                 goto unwind;
 
-        ret = fd_ctx_set (fd, this, dummy);
-        if (ret != 0)
-                gf_log (this->name, GF_LOG_ERROR,
-                        "setting context for fd=%p in locks failed.", fd);
+        fdctx = pl_check_n_create_fdctx (this, fd);
+        if (!fdctx) {
+                op_errno = ENOMEM;
+                op_ret   = -1;
+                goto unwind;
+        }
 
 unwind:
         STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
-                             preparent, postparent);
+                             preparent, postparent, xdata);
 
         return 0;
 }
@@ -466,11 +1062,12 @@ unwind:
 
 int
 pl_create (call_frame_t *frame, xlator_t *this,
-           loc_t *loc, int32_t flags, mode_t mode, fd_t *fd, dict_t *params)
+           loc_t *loc, int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
+           dict_t *xdata)
 {
         STACK_WIND (frame, pl_create_cbk,
                     FIRST_CHILD (this), FIRST_CHILD (this)->fops->create,
-                    loc, flags, mode, fd, params);
+                    loc, flags, mode, umask, fd, xdata);
         return 0;
 }
 
@@ -479,10 +1076,10 @@ int
 pl_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
               int32_t op_ret, int32_t op_errno,
               struct iovec *vector, int32_t count, struct iatt *stbuf,
-              struct iobref *iobref)
+              struct iobref *iobref, dict_t *xdata)
 {
         STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno,
-                             vector, count, stbuf, iobref);
+                             vector, count, stbuf, iobref, xdata);
 
         return 0;
 }
@@ -490,9 +1087,10 @@ pl_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 int
 pl_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-               struct iatt *postbuf)
+               struct iatt *postbuf, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf);
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf,
+                             xdata);
 
         return 0;
 }
@@ -550,12 +1148,12 @@ __rw_allowable (pl_inode_t *pl_inode, posix_lock_t *region,
 
 
 int
-pl_readv_cont (call_frame_t *frame, xlator_t *this,
-               fd_t *fd, size_t size, off_t offset)
+pl_readv_cont (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+               off_t offset, uint32_t flags, dict_t *xdata)
 {
         STACK_WIND (frame, pl_readv_cbk,
                     FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv,
-                    fd, size, offset);
+                    fd, size, offset, flags, xdata);
 
         return 0;
 }
@@ -563,7 +1161,7 @@ pl_readv_cont (call_frame_t *frame, xlator_t *this,
 
 int
 pl_readv (call_frame_t *frame, xlator_t *this,
-          fd_t *fd, size_t size, off_t offset)
+          fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
 {
         posix_locks_private_t *priv = NULL;
         pl_inode_t            *pl_inode = NULL;
@@ -580,7 +1178,7 @@ pl_readv (call_frame_t *frame, xlator_t *this,
         if (priv->mandatory && pl_inode->mandatory) {
                 region.fl_start   = offset;
                 region.fl_end     = offset + size - 1;
-                region.transport  = frame->root->trans;
+                region.client     = frame->root->client;
                 region.fd_num     = fd_to_fdnum(fd);
                 region.client_pid = frame->root->pid;
                 region.owner      = frame->root->lk_owner;
@@ -601,21 +1199,18 @@ pl_readv (call_frame_t *frame, xlator_t *this,
                                 goto unlock;
                         }
 
-			rw = GF_CALLOC (1, sizeof (*rw),
+                        rw = GF_CALLOC (1, sizeof (*rw),
                                         gf_locks_mt_pl_rw_req_t);
                         if (!rw) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory.");
                                 op_errno = ENOMEM;
                                 op_ret = -1;
                                 goto unlock;
                         }
 
                         rw->stub = fop_readv_stub (frame, pl_readv_cont,
-                                                   fd, size, offset);
+                                                   fd, size, offset, flags,
+                                                   xdata);
                         if (!rw->stub) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory.");
                                 op_errno = ENOMEM;
                                 op_ret = -1;
                                 GF_FREE (rw);
@@ -634,12 +1229,12 @@ pl_readv (call_frame_t *frame, xlator_t *this,
         if (wind_needed) {
                 STACK_WIND (frame, pl_readv_cbk,
                             FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv,
-                            fd, size, offset);
+                            fd, size, offset, flags, xdata);
         }
 
         if (op_ret == -1)
                 STACK_UNWIND_STRICT (readv, frame, -1, op_errno,
-                                     NULL, 0, NULL, NULL);
+                                     NULL, 0, NULL, NULL, NULL);
 
         return 0;
 }
@@ -648,11 +1243,11 @@ pl_readv (call_frame_t *frame, xlator_t *this,
 int
 pl_writev_cont (call_frame_t *frame, xlator_t *this, fd_t *fd,
                 struct iovec *vector, int count, off_t offset,
-                struct iobref *iobref)
+                uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
         STACK_WIND (frame, pl_writev_cbk,
                     FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev,
-                    fd, vector, count, offset, iobref);
+                    fd, vector, count, offset, flags, iobref, xdata);
 
         return 0;
 }
@@ -661,7 +1256,7 @@ pl_writev_cont (call_frame_t *frame, xlator_t *this, fd_t *fd,
 int
 pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
            struct iovec *vector, int32_t count, off_t offset,
-           struct iobref *iobref)
+           uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
         posix_locks_private_t *priv = NULL;
         pl_inode_t            *pl_inode = NULL;
@@ -671,14 +1266,13 @@ pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
         int                    op_errno = 0;
         char                   wind_needed = 1;
 
-
         priv = this->private;
         pl_inode = pl_inode_get (this, fd->inode);
 
         if (priv->mandatory && pl_inode->mandatory) {
                 region.fl_start   = offset;
                 region.fl_end     = offset + iov_length (vector, count) - 1;
-                region.transport  = frame->root->trans;
+                region.client     = frame->root->client;
                 region.fd_num     = fd_to_fdnum(fd);
                 region.client_pid = frame->root->pid;
                 region.owner      = frame->root->lk_owner;
@@ -699,11 +1293,9 @@ pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
                                 goto unlock;
                         }
 
-			rw = GF_CALLOC (1, sizeof (*rw),
+                        rw = GF_CALLOC (1, sizeof (*rw),
                                         gf_locks_mt_pl_rw_req_t);
                         if (!rw) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory.");
                                 op_errno = ENOMEM;
                                 op_ret = -1;
                                 goto unlock;
@@ -711,10 +1303,8 @@ pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
                         rw->stub = fop_writev_stub (frame, pl_writev_cont,
                                                     fd, vector, count, offset,
-                                                    iobref);
+                                                    flags, iobref, xdata);
                         if (!rw->stub) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory.");
                                 op_errno = ENOMEM;
                                 op_ret = -1;
                                 GF_FREE (rw);
@@ -733,33 +1323,203 @@ pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
         if (wind_needed)
                 STACK_WIND (frame, pl_writev_cbk,
                             FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev,
-                            fd, vector, count, offset, iobref);
+                            fd, vector, count, offset, flags, iobref, xdata);
 
         if (op_ret == -1)
-                STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL);
+                STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL,
+                                     NULL);
 
         return 0;
 }
 
+static int
+__fd_has_locks (pl_inode_t *pl_inode, fd_t *fd)
+{
+        int           found = 0;
+        posix_lock_t *l     = NULL;
+
+        list_for_each_entry (l, &pl_inode->ext_list, list) {
+                if ((l->fd_num == fd_to_fdnum(fd))) {
+                        found = 1;
+                        break;
+                }
+        }
+
+        return found;
+}
+
+static posix_lock_t *
+lock_dup (posix_lock_t *lock)
+{
+        posix_lock_t *new_lock = NULL;
+
+        new_lock = new_posix_lock (&lock->user_flock, lock->client,
+                                   lock->client_pid, &lock->owner,
+                                   (fd_t *)lock->fd_num);
+        return new_lock;
+}
+
+static int
+__dup_locks_to_fdctx (pl_inode_t *pl_inode, fd_t *fd,
+                      pl_fdctx_t *fdctx)
+{
+        posix_lock_t *l        = NULL;
+        posix_lock_t *duplock = NULL;
+        int ret = 0;
+
+        list_for_each_entry (l, &pl_inode->ext_list, list) {
+                if ((l->fd_num == fd_to_fdnum(fd))) {
+                        duplock = lock_dup (l);
+                        if (!duplock) {
+                                ret = -1;
+                                break;
+                        }
+
+                        list_add_tail (&duplock->list, &fdctx->locks_list);
+                }
+        }
+
+        return ret;
+}
+
+static int
+__copy_locks_to_fdctx (pl_inode_t *pl_inode, fd_t *fd,
+                      pl_fdctx_t *fdctx)
+{
+        int ret = 0;
+
+        ret = __dup_locks_to_fdctx (pl_inode, fd, fdctx);
+        if (ret)
+                goto out;
+
+out:
+        return ret;
+
+}
+
+static void
+pl_mark_eol_lock (posix_lock_t *lock)
+{
+        lock->user_flock.l_type = GF_LK_EOL;
+        return;
+}
+
+static posix_lock_t *
+__get_next_fdctx_lock (pl_fdctx_t *fdctx)
+{
+        posix_lock_t *lock = NULL;
+
+        GF_ASSERT (fdctx);
+
+        if (list_empty (&fdctx->locks_list)) {
+                gf_log (THIS->name, GF_LOG_DEBUG,
+                        "fdctx lock list empty");
+                goto out;
+        }
+
+        lock = list_entry (fdctx->locks_list.next, typeof (*lock),
+                           list);
+
+        GF_ASSERT (lock);
+
+        list_del_init (&lock->list);
+
+out:
+        return lock;
+}
+
+static int
+__set_next_lock_fd (pl_fdctx_t *fdctx, posix_lock_t *reqlock)
+{
+        posix_lock_t *lock  = NULL;
+        int           ret   = 0;
+
+        GF_ASSERT (fdctx);
+
+        lock = __get_next_fdctx_lock (fdctx);
+        if (!lock) {
+                gf_log (THIS->name, GF_LOG_DEBUG,
+                        "marking EOL in reqlock");
+                pl_mark_eol_lock (reqlock);
+                goto out;
+        }
+
+        reqlock->user_flock  = lock->user_flock;
+        reqlock->fl_start    = lock->fl_start;
+        reqlock->fl_type     = lock->fl_type;
+        reqlock->fl_end      = lock->fl_end;
+        reqlock->owner       = lock->owner;
+
+out:
+        if (lock)
+                __destroy_lock (lock);
+
+        return ret;
+}
+
+static int
+pl_getlk_fd (xlator_t *this, pl_inode_t *pl_inode,
+             fd_t *fd, posix_lock_t *reqlock)
+{
+        uint64_t    tmp   = 0;
+        pl_fdctx_t *fdctx = NULL;
+        int         ret   = 0;
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                if (!__fd_has_locks (pl_inode, fd)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "fd=%p has no active locks", fd);
+                        ret = 0;
+                        goto unlock;
+                }
+
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "There are active locks on fd");
+
+                ret = fd_ctx_get (fd, this, &tmp);
+                fdctx = (pl_fdctx_t *)(long) tmp;
+
+                if (list_empty (&fdctx->locks_list)) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "no fdctx -> copying all locks on fd");
+
+                        ret = __copy_locks_to_fdctx (pl_inode, fd, fdctx);
+                        if (ret) {
+                                goto unlock;
+                        }
+
+                        ret = __set_next_lock_fd (fdctx, reqlock);
+
+                } else {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "fdctx present -> returning the next lock");
+                        ret = __set_next_lock_fd (fdctx, reqlock);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "could not get next lock of fd");
+                                goto unlock;
+                        }
+                }
+        }
+
+unlock:
+        pthread_mutex_unlock (&pl_inode->mutex);
+        return ret;
+
+}
 
 int
 pl_lk (call_frame_t *frame, xlator_t *this,
-       fd_t *fd, int32_t cmd, struct flock *flock)
+       fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
-        void                  *transport = NULL;
-        pid_t                  client_pid = 0;
-        uint64_t               owner      = 0;
-        pl_inode_t            *pl_inode = NULL;
-        int                    op_ret = 0;
-        int                    op_errno = 0;
-        int                    can_block = 0;
-        posix_lock_t          *reqlock = NULL;
-        posix_lock_t          *conf = NULL;
-        int                    ret = 0;
-
-        transport  = frame->root->trans;
-        client_pid = frame->root->pid;
-        owner      = frame->root->lk_owner;
+        pl_inode_t   *pl_inode   = NULL;
+        int           op_ret     = 0;
+        int           op_errno   = 0;
+        int           can_block  = 0;
+        posix_lock_t *reqlock    = NULL;
+        posix_lock_t *conf       = NULL;
+        int           ret        = 0;
 
         if ((flock->l_start < 0) || (flock->l_len < 0)) {
                 op_ret = -1;
@@ -769,19 +1529,15 @@ pl_lk (call_frame_t *frame, xlator_t *this,
 
         pl_inode = pl_inode_get (this, fd->inode);
         if (!pl_inode) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
                 op_ret = -1;
                 op_errno = ENOMEM;
                 goto unwind;
         }
 
-        reqlock = new_posix_lock (flock, transport, client_pid,
-                                  owner, fd);
+        reqlock = new_posix_lock (flock, frame->root->client, frame->root->pid,
+                                  &frame->root->lk_owner, fd);
 
         if (!reqlock) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
                 op_ret = -1;
                 op_errno = ENOMEM;
                 goto unwind;
@@ -791,6 +1547,68 @@ pl_lk (call_frame_t *frame, xlator_t *this,
 
         switch (cmd) {
 
+        case F_RESLK_LCKW:
+                can_block = 1;
+
+                /* fall through */
+        case F_RESLK_LCK:
+                memcpy (&reqlock->user_flock, flock, sizeof (struct gf_flock));
+                reqlock->frame = frame;
+                reqlock->this = this;
+
+                ret = pl_reserve_setlk (this, pl_inode, reqlock,
+                                        can_block);
+                if (ret < 0) {
+                        if (can_block)
+                                goto out;
+
+                        op_ret = -1;
+                        op_errno = -ret;
+                        __destroy_lock (reqlock);
+                        goto unwind;
+                }
+                /* Finally a getlk and return the call */
+                conf = pl_getlk (pl_inode, reqlock);
+                if (conf)
+                        posix_lock_to_flock (conf, flock);
+                break;
+
+        case F_RESLK_UNLCK:
+                reqlock->frame = frame;
+                reqlock->this = this;
+                ret = pl_reserve_unlock (this, pl_inode, reqlock);
+                if (ret < 0) {
+                        op_ret = -1;
+                        op_errno = -ret;
+                }
+                __destroy_lock (reqlock);
+                goto unwind;
+
+                break;
+
+        case F_GETLK_FD:
+                reqlock->frame = frame;
+                reqlock->this = this;
+                ret = pl_verify_reservelk (this, pl_inode, reqlock, can_block);
+                GF_ASSERT (ret >= 0);
+
+                ret = pl_getlk_fd (this, pl_inode, fd, reqlock);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "getting locks on fd failed");
+                        op_ret = -1;
+                        op_errno = ENOLCK;
+                        goto unwind;
+                }
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Replying with a lock on fd for healing");
+
+                posix_lock_to_flock (reqlock, flock);
+                __destroy_lock (reqlock);
+
+                break;
+
 #if F_GETLK != F_GETLK64
         case F_GETLK64:
 #endif
@@ -815,12 +1633,18 @@ pl_lk (call_frame_t *frame, xlator_t *this,
         case F_SETLK64:
 #endif
         case F_SETLK:
-                memcpy (&reqlock->user_flock, flock, sizeof (struct flock));
+                memcpy (&reqlock->user_flock, flock, sizeof (struct gf_flock));
+                ret = pl_verify_reservelk (this, pl_inode, reqlock, can_block);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "Lock blocked due to conflicting reserve lock");
+                        goto out;
+                }
                 ret = pl_setlk (this, pl_inode, reqlock,
                                 can_block);
 
                 if (ret == -1) {
-                        if (can_block) {
+                        if ((can_block) && (F_UNLCK != flock->l_type)) {
                                 pl_trace_block (this, frame, fd, NULL, cmd, flock, NULL);
                                 goto out;
                         }
@@ -828,13 +1652,22 @@ pl_lk (call_frame_t *frame, xlator_t *this,
                         op_ret = -1;
                         op_errno = EAGAIN;
                         __destroy_lock (reqlock);
+
+                } else if ((0 == ret) && (F_UNLCK == flock->l_type)) {
+                        /* For NLM's last "unlock on fd" detection */
+                        if (pl_locks_by_fd (pl_inode, fd))
+                                flock->l_type = F_RDLCK;
+                        else
+                                flock->l_type = F_UNLCK;
                 }
         }
 
 unwind:
         pl_trace_out (this, frame, fd, NULL, cmd, flock, op_ret, op_errno, NULL);
         pl_update_refkeeper (this, fd->inode);
-        STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, flock);
+
+
+        STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, flock, xdata);
 out:
         return 0;
 }
@@ -849,120 +1682,121 @@ pl_forget (xlator_t *this,
 
         posix_lock_t *ext_tmp = NULL;
         posix_lock_t *ext_l   = NULL;
-	struct list_head posixlks_released;
+        struct list_head posixlks_released;
 
         pl_inode_lock_t *ino_tmp = NULL;
         pl_inode_lock_t *ino_l   = NULL;
-	struct list_head inodelks_released;
+        struct list_head inodelks_released;
 
         pl_rw_req_t *rw_tmp = NULL;
         pl_rw_req_t *rw_req = NULL;
 
         pl_entry_lock_t *entry_tmp = NULL;
         pl_entry_lock_t *entry_l   = NULL;
-	struct list_head entrylks_released;
+        struct list_head entrylks_released;
 
         pl_dom_list_t *dom = NULL;
         pl_dom_list_t *dom_tmp = NULL;
 
-	INIT_LIST_HEAD (&posixlks_released);
-	INIT_LIST_HEAD (&inodelks_released);
-	INIT_LIST_HEAD (&entrylks_released);
+        INIT_LIST_HEAD (&posixlks_released);
+        INIT_LIST_HEAD (&inodelks_released);
+        INIT_LIST_HEAD (&entrylks_released);
 
         pl_inode = pl_inode_get (this, inode);
 
-	pthread_mutex_lock (&pl_inode->mutex);
-	{
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
 
-		if (!list_empty (&pl_inode->rw_list)) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Pending R/W requests found, releasing.");
+                if (!list_empty (&pl_inode->rw_list)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Pending R/W requests found, releasing.");
 
-			list_for_each_entry_safe (rw_req, rw_tmp, &pl_inode->rw_list,
-						  list) {
+                        list_for_each_entry_safe (rw_req, rw_tmp, &pl_inode->rw_list,
+                                                  list) {
 
-				list_del (&rw_req->list);
-				GF_FREE (rw_req);
-			}
-		}
+                                list_del (&rw_req->list);
+                                GF_FREE (rw_req);
+                        }
+                }
 
-		if (!list_empty (&pl_inode->ext_list)) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Pending fcntl locks found, releasing.");
+                if (!list_empty (&pl_inode->ext_list)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Pending fcntl locks found, releasing.");
 
-			list_for_each_entry_safe (ext_l, ext_tmp, &pl_inode->ext_list,
-						  list) {
+                        list_for_each_entry_safe (ext_l, ext_tmp, &pl_inode->ext_list,
+                                                  list) {
 
-				__delete_lock (pl_inode, ext_l);
-				if (ext_l->blocked) {
-					list_add_tail (&ext_l->list, &posixlks_released);
-					continue;
-				}
-				__destroy_lock (ext_l);
-			}
-		}
+                                __delete_lock (pl_inode, ext_l);
+                                if (ext_l->blocked) {
+                                        list_add_tail (&ext_l->list, &posixlks_released);
+                                        continue;
+                                }
+                                __destroy_lock (ext_l);
+                        }
+                }
 
 
-		list_for_each_entry_safe (dom, dom_tmp, &pl_inode->dom_list, inode_list) {
+                list_for_each_entry_safe (dom, dom_tmp, &pl_inode->dom_list, inode_list) {
 
-			if (!list_empty (&dom->inodelk_list)) {
-				gf_log (this->name, GF_LOG_WARNING,
-					"Pending inode locks found, releasing.");
+                        if (!list_empty (&dom->inodelk_list)) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "Pending inode locks found, releasing.");
 
-				list_for_each_entry_safe (ino_l, ino_tmp, &dom->inodelk_list, list) {
-					__delete_inode_lock (ino_l);
-					__destroy_inode_lock (ino_l);
-				}
+                                list_for_each_entry_safe (ino_l, ino_tmp, &dom->inodelk_list, list) {
+                                        __delete_inode_lock (ino_l);
+                                        __pl_inodelk_unref (ino_l);
+                                }
 
-				list_splice_init (&dom->blocked_inodelks, &inodelks_released);
-				
+                                list_splice_init (&dom->blocked_inodelks, &inodelks_released);
 
-			}
-			if (!list_empty (&dom->entrylk_list)) {
-				gf_log (this->name, GF_LOG_WARNING,
-					"Pending entry locks found, releasing.");
 
-				list_for_each_entry_safe (entry_l, entry_tmp, &dom->entrylk_list, domain_list) {
-					list_del_init (&entry_l->domain_list);
+                        }
+                        if (!list_empty (&dom->entrylk_list)) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "Pending entry locks found, releasing.");
 
-					if (entry_l->basename)
-						GF_FREE ((char *)entry_l->basename);
-					GF_FREE (entry_l);
-				}
+                                list_for_each_entry_safe (entry_l, entry_tmp, &dom->entrylk_list, domain_list) {
+                                        list_del_init (&entry_l->domain_list);
 
-				list_splice_init (&dom->blocked_entrylks, &entrylks_released);
-			}
+                                        GF_FREE ((char *)entry_l->basename);
+                                        GF_FREE (entry_l->connection_id);
+                                        GF_FREE (entry_l);
+                                }
 
-			list_del (&dom->inode_list);
-			gf_log ("posix-locks", GF_LOG_TRACE,
-				" Cleaning up domain: %s", dom->domain);
-			GF_FREE ((char *)(dom->domain));
-			GF_FREE (dom);
-		}
+                                list_splice_init (&dom->blocked_entrylks, &entrylks_released);
+                        }
 
-	}
-	pthread_mutex_unlock (&pl_inode->mutex);
+                        list_del (&dom->inode_list);
+                        gf_log ("posix-locks", GF_LOG_TRACE,
+                                " Cleaning up domain: %s", dom->domain);
+                        GF_FREE ((char *)(dom->domain));
+                        GF_FREE (dom);
+                }
 
-	list_for_each_entry_safe (ext_l, ext_tmp, &posixlks_released, list) {
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
 
-		STACK_UNWIND_STRICT (lk, ext_l->frame, -1, 0, &ext_l->user_flock);
-		__destroy_lock (ext_l);
-	}
+        list_for_each_entry_safe (ext_l, ext_tmp, &posixlks_released, list) {
+
+                STACK_UNWIND_STRICT (lk, ext_l->frame, -1, 0,
+                                     &ext_l->user_flock, NULL);
+                __destroy_lock (ext_l);
+        }
 
-	list_for_each_entry_safe (ino_l, ino_tmp, &inodelks_released, blocked_locks) {
+        list_for_each_entry_safe (ino_l, ino_tmp, &inodelks_released, blocked_locks) {
 
-		STACK_UNWIND_STRICT (inodelk, ino_l->frame, -1, 0);
-		__destroy_inode_lock (ino_l);
-	}
+                STACK_UNWIND_STRICT (inodelk, ino_l->frame, -1, 0, NULL);
+                __pl_inodelk_unref (ino_l);
+        }
 
-	list_for_each_entry_safe (entry_l, entry_tmp, &entrylks_released, blocked_locks) {
+        list_for_each_entry_safe (entry_l, entry_tmp, &entrylks_released, blocked_locks) {
 
-		STACK_UNWIND_STRICT (entrylk, entry_l->frame, -1, 0);
-		if (entry_l->basename)
-			GF_FREE ((char *)entry_l->basename);
-		GF_FREE (entry_l);
+                STACK_UNWIND_STRICT (entrylk, entry_l->frame, -1, 0, NULL);
+                GF_FREE ((char *)entry_l->basename);
+                GF_FREE (entry_l->connection_id);
+                GF_FREE (entry_l);
 
-	}
+        }
 
         GF_FREE (pl_inode);
 
@@ -975,8 +1809,14 @@ pl_release (xlator_t *this, fd_t *fd)
         pl_inode_t *pl_inode     = NULL;
         uint64_t    tmp_pl_inode = 0;
         int         ret          = -1;
+        uint64_t    tmp          = 0;
+        pl_fdctx_t *fdctx        = NULL;
+
+        if (fd == NULL) {
+                goto out;
+        }
 
-	ret = inode_ctx_get (fd->inode, this, &tmp_pl_inode);
+        ret = inode_ctx_get (fd->inode, this, &tmp_pl_inode);
         if (ret != 0)
                 goto out;
 
@@ -988,11 +1828,48 @@ pl_release (xlator_t *this, fd_t *fd)
                 "Releasing all locks with fd %p", fd);
 
         delete_locks_of_fd (this, pl_inode, fd);
+        pl_update_refkeeper (this, fd->inode);
+
+        ret = fd_ctx_del (fd, this, &tmp);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Could not get fdctx");
+                goto out;
+        }
+
+        fdctx = (pl_fdctx_t *)(long)tmp;
 
+        GF_FREE (fdctx);
 out:
         return ret;
 }
-static int32_t
+
+int
+pl_releasedir (xlator_t *this, fd_t *fd)
+{
+        int         ret          = -1;
+        uint64_t    tmp          = 0;
+        pl_fdctx_t *fdctx        = NULL;
+
+        if (fd == NULL) {
+                goto out;
+        }
+
+        ret = fd_ctx_del (fd, this, &tmp);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Could not get fdctx");
+                goto out;
+        }
+
+        fdctx = (pl_fdctx_t *)(long)tmp;
+
+        GF_FREE (fdctx);
+out:
+        return ret;
+}
+
+int32_t
 __get_posixlk_count (xlator_t *this, pl_inode_t *pl_inode)
 {
         posix_lock_t *lock   = NULL;
@@ -1000,16 +1877,6 @@ __get_posixlk_count (xlator_t *this, pl_inode_t *pl_inode)
 
         list_for_each_entry (lock, &pl_inode->ext_list, list) {
 
-			gf_log (this->name, GF_LOG_DEBUG,
-                                " XATTR DEBUG"
-				"%s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" state: %s",
-				lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
-				lock->client_pid,
-                                lock->owner,
-				lock->user_flock.l_start,
-				lock->user_flock.l_len,
-                                lock->blocked == 1 ? "Blocked" : "Active");
-
                 count++;
         }
 
@@ -1042,6 +1909,24 @@ out:
 }
 
 void
+pl_parent_entrylk_xattr_fill (xlator_t *this, inode_t *parent,
+                              char *basename, dict_t *dict)
+{
+        uint32_t         entrylk = 0;
+        int             ret     = -1;
+
+        if (!parent || !basename || !strlen (basename))
+                goto out;
+        entrylk = check_entrylk_on_basename (this, parent, basename);
+out:
+        ret = dict_set_uint32 (dict, GLUSTERFS_PARENT_ENTRYLK, entrylk);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        " dict_set failed on key %s", GLUSTERFS_PARENT_ENTRYLK);
+        }
+}
+
+void
 pl_entrylk_xattr_fill (xlator_t *this, inode_t *inode,
                        dict_t *dict)
 {
@@ -1058,19 +1943,34 @@ pl_entrylk_xattr_fill (xlator_t *this, inode_t *inode,
 }
 
 void
-pl_inodelk_xattr_fill (xlator_t *this, inode_t *inode,
-                       dict_t *dict)
+pl_inodelk_xattr_fill (xlator_t *this, inode_t *inode, dict_t *dict,
+                       gf_boolean_t per_dom)
 {
         int32_t     count = 0;
         int         ret   = -1;
+        char        *domname = NULL;
+
+
+        if (per_dom){
+                ret = dict_get_str (dict, GLUSTERFS_INODELK_DOM_COUNT,
+                                    &domname);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                "value for key %s",GLUSTERFS_INODELK_DOM_COUNT);
+                        goto out;
+                }
+        }
+
+        count = get_inodelk_count (this, inode, domname);
 
-        count = get_inodelk_count (this, inode);
         ret = dict_set_int32 (dict, GLUSTERFS_INODELK_COUNT, count);
         if (ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        " dict_set failed on key %s", GLUSTERFS_INODELK_COUNT);
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to set count for "
+                        "key %s", GLUSTERFS_INODELK_COUNT);
         }
 
+out:
+        return;
 }
 
 void
@@ -1091,58 +1991,63 @@ pl_posixlk_xattr_fill (xlator_t *this, inode_t *inode,
 
 int32_t
 pl_lookup_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    inode_t *inode,
-		    struct iatt *buf,
-                    dict_t *dict,
+                    void *cookie,
+                    xlator_t *this,
+                    int32_t op_ret,
+                    int32_t op_errno,
+                    inode_t *inode,
+                    struct iatt *buf,
+                    dict_t *xdata,
                     struct iatt *postparent)
 {
         pl_local_t *local = NULL;
 
-        if (!frame->local) {
-                goto out;
-        }
+        GF_VALIDATE_OR_GOTO (this->name, frame->local, out);
 
-        if (op_ret) {
+        if (op_ret)
                 goto out;
-        }
 
         local = frame->local;
 
+        if (local->parent_entrylk_req)
+                pl_parent_entrylk_xattr_fill (this, local->loc.parent,
+                                              (char*)local->loc.name, xdata);
         if (local->entrylk_count_req)
-                pl_entrylk_xattr_fill (this, inode, dict);
+                pl_entrylk_xattr_fill (this, inode, xdata);
         if (local->inodelk_count_req)
-                pl_inodelk_xattr_fill (this, inode, dict);
+                pl_inodelk_xattr_fill (this, inode, xdata, _gf_false);
+        if (local->inodelk_dom_count_req)
+                pl_inodelk_xattr_fill (this, inode, xdata, _gf_true);
         if (local->posixlk_count_req)
-                pl_posixlk_xattr_fill (this, inode, dict);
+                pl_posixlk_xattr_fill (this, inode, xdata);
 
 
+out:
+        local = frame->local;
         frame->local = NULL;
 
-        if (local != NULL)
-                GF_FREE (local);
+        if (local != NULL) {
+                loc_wipe (&local->loc);
+                mem_put (local);
+        }
 
-out:
-	STACK_UNWIND_STRICT (
+        STACK_UNWIND_STRICT (
                      lookup,
                      frame,
-		      op_ret,
-		      op_errno,
-		      inode,
-		      buf,
-                      dict,
-                      postparent);
-	return 0;
+                     op_ret,
+                     op_errno,
+                     inode,
+                     buf,
+                     xdata,
+                     postparent);
+        return 0;
 }
 
 int32_t
 pl_lookup (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		dict_t *xattr_req)
+           xlator_t *this,
+           loc_t *loc,
+           dict_t *xdata)
 {
         pl_local_t *local  = NULL;
         int         ret    = -1;
@@ -1151,42 +2056,114 @@ pl_lookup (call_frame_t *frame,
         VALIDATE_OR_GOTO (this, out);
         VALIDATE_OR_GOTO (loc, out);
 
-        local = GF_CALLOC (1, sizeof (*local), gf_locks_mt_pl_local_t);
-        if (!local) {
-                ret = -1;
-                gf_log (this->name, GF_LOG_ERROR,
-                        " Out of memory");
-                goto out;
+        local = mem_get0 (this->local_pool);
+        GF_VALIDATE_OR_GOTO (this->name, local, out);
+
+        if (xdata) {
+                if (dict_get (xdata, GLUSTERFS_ENTRYLK_COUNT))
+                        local->entrylk_count_req = 1;
+                if (dict_get (xdata, GLUSTERFS_INODELK_COUNT))
+                        local->inodelk_count_req = 1;
+                if (dict_get (xdata, GLUSTERFS_INODELK_DOM_COUNT))
+                        local->inodelk_dom_count_req = 1;
+                if (dict_get (xdata, GLUSTERFS_POSIXLK_COUNT))
+                        local->posixlk_count_req = 1;
+                if (dict_get (xdata, GLUSTERFS_PARENT_ENTRYLK))
+                        local->parent_entrylk_req = 1;
         }
 
-        if (dict_get (xattr_req, GLUSTERFS_ENTRYLK_COUNT))
-                local->entrylk_count_req = 1;
-        if (dict_get (xattr_req, GLUSTERFS_INODELK_COUNT))
-                local->inodelk_count_req = 1;
-        if (dict_get (xattr_req, GLUSTERFS_POSIXLK_COUNT))
-                local->posixlk_count_req = 1;
-
         frame->local = local;
+        loc_copy (&local->loc, loc);
 
-	STACK_WIND (frame,
-		    pl_lookup_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->lookup,
-		    loc,
-		    xattr_req);
+        STACK_WIND (frame,
+                    pl_lookup_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup,
+                    loc, xdata);
         ret = 0;
 out:
         if (ret == -1)
-                STACK_UNWIND_STRICT (lookup, frame, -1, 0, NULL, NULL, NULL, NULL);
+                STACK_UNWIND_STRICT (lookup, frame, -1, 0, NULL,
+                                     NULL, NULL, NULL);
+
+        return 0;
+}
+int
+pl_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int op_ret, int op_errno, gf_dirent_t *entries, dict_t *xdata)
+{
+        pl_local_t *local  = NULL;
+        gf_dirent_t *entry = NULL;
+
+        local = frame->local;
+
+        if (op_ret <= 0)
+                goto unwind;
+
+        list_for_each_entry (entry, &entries->list, list) {
+                if (local->entrylk_count_req)
+                        pl_entrylk_xattr_fill (this, entry->inode, entry->dict);
+                if (local->inodelk_count_req)
+                        pl_inodelk_xattr_fill (this, entry->inode, entry->dict,
+                                               _gf_false);
+                if (local->inodelk_dom_count_req)
+                        pl_inodelk_xattr_fill (this, entry->inode, entry->dict,
+                                               _gf_true);
+                if (local->posixlk_count_req)
+                        pl_posixlk_xattr_fill (this, entry->inode, entry->dict);
+        }
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
+
+        if (local)
+                mem_put (local);
+
+        return 0;
+}
+
+int
+pl_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+             off_t offset, dict_t *dict)
+{
+        pl_local_t *local  = NULL;
+
+        local = mem_get0 (this->local_pool);
+        GF_VALIDATE_OR_GOTO (this->name, local, out);
+
+        if (dict) {
+                if (dict_get (dict, GLUSTERFS_ENTRYLK_COUNT))
+                        local->entrylk_count_req = 1;
+                if (dict_get (dict, GLUSTERFS_INODELK_COUNT))
+                        local->inodelk_count_req = 1;
+                if (dict_get (dict, GLUSTERFS_INODELK_DOM_COUNT))
+                        local->inodelk_dom_count_req = 1;
+                if (dict_get (dict, GLUSTERFS_POSIXLK_COUNT))
+                        local->posixlk_count_req = 1;
+        }
 
-	return 0;
+        frame->local = local;
+
+        STACK_WIND (frame, pl_readdirp_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdirp,
+                    fd, size, offset, dict);
+
+        return 0;
+out:
+        STACK_UNWIND_STRICT (readdirp, frame, -1, ENOMEM, NULL, NULL);
+        return 0;
 }
 
+
 void
-pl_dump_lock (char *str, int size, struct flock *flock,
-              uint64_t owner, void *trans)
+pl_dump_lock (char *str, int size, struct gf_flock *flock,
+              gf_lkowner_t *owner, void *trans, char *conn_id,
+              time_t *granted_time, time_t *blkd_time, gf_boolean_t active)
 {
-        char *type_str = NULL;
+        char  *type_str    = NULL;
+        char   granted[32] = {0,};
+        char   blocked[32] = {0,};
 
         switch (flock->l_type) {
         case F_RDLCK:
@@ -1203,13 +2180,35 @@ pl_dump_lock (char *str, int size, struct flock *flock,
                 break;
         }
 
-        snprintf (str, size, "type=%s, start=%llu, len=%llu, pid=%llu, lk-owner=%llu, transport=%p",
-                  type_str, (unsigned long long) flock->l_start,
-                  (unsigned long long) flock->l_len,
-                  (unsigned long long) flock->l_pid,
-                  (unsigned long long) owner,
-                  trans);
-
+        if (active) {
+                if (blkd_time && *blkd_time == 0) {
+                        snprintf (str, size, RANGE_GRNTD_FMT,
+                                  type_str, flock->l_whence,
+                                  (unsigned long long) flock->l_start,
+                                  (unsigned long long) flock->l_len,
+                                  (unsigned long long) flock->l_pid,
+                                  lkowner_utoa (owner), trans, conn_id,
+                                  ctime_r (granted_time, granted));
+                } else {
+                        snprintf (str, size, RANGE_BLKD_GRNTD_FMT,
+                                  type_str, flock->l_whence,
+                                  (unsigned long long) flock->l_start,
+                                  (unsigned long long) flock->l_len,
+                                  (unsigned long long) flock->l_pid,
+                                  lkowner_utoa (owner), trans, conn_id,
+                                  ctime_r (blkd_time, blocked),
+                                  ctime_r (granted_time, granted));
+                }
+        }
+        else {
+                snprintf (str, size, RANGE_BLKD_FMT,
+                          type_str, flock->l_whence,
+                          (unsigned long long) flock->l_start,
+                          (unsigned long long) flock->l_len,
+                          (unsigned long long) flock->l_pid,
+                          lkowner_utoa (owner), trans, conn_id,
+                          ctime_r (blkd_time, blocked));
+        }
 
 }
 
@@ -1218,8 +2217,10 @@ __dump_entrylks (pl_inode_t *pl_inode)
 {
         pl_dom_list_t   *dom  = NULL;
         pl_entry_lock_t *lock = NULL;
-        int             count = 0;
-        char            key[GF_DUMP_MAX_BUF_LEN];
+        char             blocked[32] = {0,};
+        char             granted[32] = {0,};
+        int              count = 0;
+        char             key[GF_DUMP_MAX_BUF_LEN] = {0,};
 
         char tmp[256];
 
@@ -1228,7 +2229,7 @@ __dump_entrylks (pl_inode_t *pl_inode)
                 count = 0;
 
                 gf_proc_dump_build_key(key,
-                                       "xlator.feature.locks.lock-dump.domain",
+                                       "lock-dump.domain",
                                        "domain");
                 gf_proc_dump_write(key, "%s", dom->domain);
 
@@ -1236,12 +2237,25 @@ __dump_entrylks (pl_inode_t *pl_inode)
 
                         gf_proc_dump_build_key(key,
                                                "xlator.feature.locks.lock-dump.domain.entrylk",
-                                               "entrylk[%d](ACTIVE)",count );
-                        snprintf (tmp, 256," %s on %s owner=%llu, transport=%p",
-                                  lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : 
-                                  "ENTRYLK_WRLCK", lock->basename,
-                                  (unsigned long long) lock->owner,
-                                  lock->trans);
+                                               "entrylk[%d](ACTIVE)", count );
+                        if (lock->blkd_time.tv_sec == 0 && lock->blkd_time.tv_usec == 0) {
+                                snprintf (tmp, 256, ENTRY_GRNTD_FMT,
+                                          lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" :
+                                          "ENTRYLK_WRLCK", lock->basename,
+                                          (unsigned long long) lock->client_pid,
+                                          lkowner_utoa (&lock->owner), lock->trans,
+                                          lock->connection_id,
+                                          ctime_r (&lock->granted_time.tv_sec, granted));
+                        } else {
+                                snprintf (tmp, 256, ENTRY_BLKD_GRNTD_FMT,
+                                          lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" :
+                                          "ENTRYLK_WRLCK", lock->basename,
+                                          (unsigned long long) lock->client_pid,
+                                          lkowner_utoa (&lock->owner), lock->trans,
+                                          lock->connection_id,
+                                          ctime_r (&lock->blkd_time.tv_sec, blocked),
+                                          ctime_r (&lock->granted_time.tv_sec, granted));
+                        }
 
                         gf_proc_dump_write(key, tmp);
 
@@ -1252,10 +2266,14 @@ __dump_entrylks (pl_inode_t *pl_inode)
 
                         gf_proc_dump_build_key(key,
                                                "xlator.feature.locks.lock-dump.domain.entrylk",
-                                               "entrylk[%d](BLOCKED)",count );
-                        snprintf (tmp, 256," %s on %s state = Blocked",
-                                  lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : 
-                                  "ENTRYLK_WRLCK", lock->basename);
+                                               "entrylk[%d](BLOCKED)", count );
+                        snprintf (tmp, 256, ENTRY_BLKD_FMT,
+                                  lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" :
+                                  "ENTRYLK_WRLCK", lock->basename,
+                                  (unsigned long long) lock->client_pid,
+                                  lkowner_utoa (&lock->owner), lock->trans,
+                                  lock->connection_id,
+                                  ctime_r (&lock->blkd_time.tv_sec, blocked));
 
                         gf_proc_dump_write(key, tmp);
 
@@ -1292,18 +2310,23 @@ __dump_inodelks (pl_inode_t *pl_inode)
                 count = 0;
 
                 gf_proc_dump_build_key(key,
-                                       "xlator.feature.locks.lock-dump.domain",
+                                       "lock-dump.domain",
                                        "domain");
                 gf_proc_dump_write(key, "%s", dom->domain);
 
                 list_for_each_entry (lock, &dom->inodelk_list, list) {
 
                         gf_proc_dump_build_key(key,
-                                               "xlator.feature.locks.lock-dump.domain.inodelk",
+                                               "inodelk",
                                                "inodelk[%d](ACTIVE)",count );
 
+                        SET_FLOCK_PID (&lock->user_flock, lock);
                         pl_dump_lock (tmp, 256, &lock->user_flock,
-                                      lock->owner, lock->transport);
+                                      &lock->owner,
+                                      lock->client, lock->connection_id,
+                                      &lock->granted_time.tv_sec,
+                                      &lock->blkd_time.tv_sec,
+                                      _gf_true);
                         gf_proc_dump_write(key, tmp);
 
                         count++;
@@ -1312,10 +2335,14 @@ __dump_inodelks (pl_inode_t *pl_inode)
                 list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) {
 
                         gf_proc_dump_build_key(key,
-                                               "xlator.feature.locks.lock-dump.domain.inodelk",
+                                               "inodelk",
                                                "inodelk[%d](BLOCKED)",count );
+                        SET_FLOCK_PID (&lock->user_flock, lock);
                         pl_dump_lock (tmp, 256, &lock->user_flock,
-                                      lock->owner, lock->transport);
+                                      &lock->owner,
+                                      lock->client, lock->connection_id,
+                                      0, &lock->blkd_time.tv_sec,
+                                      _gf_false);
                         gf_proc_dump_write(key, tmp);
 
                         count++;
@@ -1347,20 +2374,20 @@ __dump_posixlks (pl_inode_t *pl_inode)
 
       list_for_each_entry (lock, &pl_inode->ext_list, list) {
 
+              SET_FLOCK_PID (&lock->user_flock, lock);
               gf_proc_dump_build_key(key,
-                                     "xlator.feature.locks.lock-dump.domain.posixlk",
+                                     "posixlk",
                                      "posixlk[%d](%s)",
                                      count,
                                      lock->blocked ? "BLOCKED" : "ACTIVE");
               pl_dump_lock (tmp, 256, &lock->user_flock,
-                            lock->owner, lock->transport);
+                            &lock->owner, lock->client, NULL,
+                            &lock->granted_time.tv_sec, &lock->blkd_time.tv_sec,
+                            (lock->blocked)? _gf_false: _gf_true);
               gf_proc_dump_write(key, tmp);
 
               count++;
         }
-
-
-
 }
 
 void
@@ -1381,108 +2408,234 @@ pl_dump_inode_priv (xlator_t *this, inode_t *inode)
         int             ret = -1;
         uint64_t        tmp_pl_inode = 0;
         pl_inode_t      *pl_inode = NULL;
-        char            key[GF_DUMP_MAX_BUF_LEN];
+        char            *pathname = NULL;
+        gf_boolean_t    section_added = _gf_false;
 
         int count      = 0;
 
-        if (!inode)
-                return -1;
+        if (!inode) {
+                errno = EINVAL;
+                goto out;
+        }
 
-        ret = inode_ctx_get (inode, this, &tmp_pl_inode);
+        ret = TRY_LOCK (&inode->lock);
+        if (ret)
+                goto out;
+        {
+                ret = __inode_ctx_get (inode, this, &tmp_pl_inode);
+                if (ret)
+                        goto unlock;
+        }
+unlock:
+        UNLOCK (&inode->lock);
+        if (ret)
+                goto out;
 
-        if (ret != 0)
+        pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
+        if (!pl_inode) {
+                ret = -1;
+                goto out;
+        }
+
+        gf_proc_dump_add_section("xlator.features.locks.%s.inode", this->name);
+        section_added = _gf_true;
+
+        /*We are safe to call __inode_path since we have the
+         * inode->table->lock */
+        __inode_path (inode, NULL, &pathname);
+        if (pathname)
+                gf_proc_dump_write ("path", "%s", pathname);
+
+        gf_proc_dump_write("mandatory", "%d", pl_inode->mandatory);
+
+        ret = pthread_mutex_trylock (&pl_inode->mutex);
+        if (ret)
+                goto out;
+        {
+                count = __get_entrylk_count (this, pl_inode);
+                if (count) {
+                        gf_proc_dump_write("entrylk-count", "%d", count);
+                        __dump_entrylks (pl_inode);
+                }
+
+                count = __get_inodelk_count (this, pl_inode, NULL);
+                if (count) {
+                        gf_proc_dump_write("inodelk-count", "%d", count);
+                        __dump_inodelks (pl_inode);
+                }
+
+                count = __get_posixlk_count (this, pl_inode);
+                if (count) {
+                        gf_proc_dump_write("posixlk-count", "%d", count);
+                        __dump_posixlks (pl_inode);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+out:
+        GF_FREE (pathname);
+
+        if (ret && inode) {
+                if (!section_added)
+                        gf_proc_dump_add_section ("xlator.features.locks.%s."
+                                                  "inode", this->name);
+                gf_proc_dump_write ("Unable to print lock state", "(Lock "
+                                    "acquisition failure) %s",
+                                    uuid_utoa (inode->gfid));
+        }
+        return ret;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
                 return ret;
 
-        pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
+        ret = xlator_mem_acct_init (this, gf_locks_mt_end + 1);
 
-        if (!pl_inode)
-                return -1;
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                                "failed");
+                return ret;
+        }
 
-        gf_proc_dump_build_key(key,
-                               "xlator.feature.locks.inode",
-                               "%ld.mandatory",inode->ino);
-        gf_proc_dump_write(key, "%d", pl_inode->mandatory);
+        return ret;
+}
 
 
-                count = get_entrylk_count (this, inode);
-                gf_proc_dump_build_key(key,
-                                       "xlator.feature.locks.entrylk-count",
-                                       "%ld.entrylk-count", inode->ino);
-                gf_proc_dump_write(key, "%d", count);
+pl_ctx_t*
+pl_ctx_get (client_t *client, xlator_t *xlator)
+{
+        void *tmp = NULL;
+        pl_ctx_t *ctx = NULL;
 
-                dump_entrylks(pl_inode);
+        client_ctx_get (client, xlator, &tmp);
 
-                count = get_inodelk_count (this, inode);
-                gf_proc_dump_build_key(key,
-                                       "xlator.feature.locks.inodelk-count",
-                                       "%ld.inodelk-count", inode->ino);
-                gf_proc_dump_write(key, "%d", count);
+        ctx = tmp;
 
-                dump_inodelks(pl_inode);
+        if (ctx != NULL)
+                goto out;
 
-                count = get_posixlk_count (this, inode);
-                gf_proc_dump_build_key(key,
-                                       "xlator.feature.locks.posixlk-count",
-                                       "%ld.posixlk-count", inode->ino);
-                gf_proc_dump_write(key, "%d", count);
+        ctx = GF_CALLOC (1, sizeof (pl_ctx_t), gf_locks_mt_posix_lock_t);
 
-                dump_posixlks(pl_inode);
+        if (ctx == NULL)
+                goto out;
 
+        ctx->ltable = pl_lock_table_new();
 
-        return 0;
+        if (ctx->ltable == NULL) {
+                GF_FREE (ctx);
+                ctx = NULL;
+                goto out;
+        }
+
+        LOCK_INIT (&ctx->ltable_lock);
+
+        if (client_ctx_set (client, xlator, ctx) != 0) {
+                LOCK_DESTROY (&ctx->ltable_lock);
+                GF_FREE (ctx->ltable);
+                GF_FREE (ctx);
+                ctx = NULL;
+        }
+out:
+        return ctx;
 }
 
+static void
+ltable_delete_locks (struct _lock_table *ltable)
+{
+        struct _locker *locker = NULL;
+        struct _locker *tmp    = NULL;
+
+        list_for_each_entry_safe (locker, tmp, &ltable->inodelk_lockers, lockers) {
+                if (locker->fd)
+                        pl_del_locker (ltable, locker->volume, &locker->loc,
+                                       locker->fd, &locker->owner,
+                                       GF_FOP_INODELK);
+                GF_FREE (locker->volume);
+                GF_FREE (locker);
+        }
 
+        list_for_each_entry_safe (locker, tmp, &ltable->entrylk_lockers, lockers) {
+                if (locker->fd)
+                        pl_del_locker (ltable, locker->volume, &locker->loc,
+                                       locker->fd, &locker->owner,
+                                       GF_FOP_ENTRYLK);
+                GF_FREE (locker->volume);
+                GF_FREE (locker);
+        }
+        GF_FREE (ltable);
+}
 
-/*
- * pl_dump_inode - inode dump function for posix locks
- *
- */
-int
-pl_dump_inode (xlator_t *this)
+
+static int32_t
+destroy_cbk (xlator_t *this, client_t *client)
 {
+        void     *tmp       = NULL;
+        pl_ctx_t *locks_ctx = NULL;
 
-        assert(this);
+        client_ctx_del (client, this, &tmp);
 
-        if (this->itable) {
-                inode_table_dump(this->itable,
-                                 "xlator.features.locks.inode_table");
-        }
+        if (tmp == NULL)
+                return 0
+;
+        locks_ctx = tmp;
+        if (locks_ctx->ltable)
+                ltable_delete_locks (locks_ctx->ltable);
+
+        LOCK_DESTROY (&locks_ctx->ltable_lock);
+        GF_FREE (locks_ctx);
 
         return 0;
 }
 
-int32_t
-mem_acct_init (xlator_t *this)
+
+static int32_t
+disconnect_cbk (xlator_t *this, client_t *client)
 {
-        int     ret = -1;
+        int32_t              ret       = 0;
+        pl_ctx_t            *locks_ctx = NULL;
+        struct _lock_table  *ltable    = NULL;
 
-        if (!this)
-                return ret;
+        locks_ctx = pl_ctx_get (client, this);
+        if (locks_ctx == NULL) {
+                gf_log (this->name, GF_LOG_INFO, "pl_ctx_get() failed");
+                goto out;
+        }
 
-        ret = xlator_mem_acct_init (this, gf_locks_mt_end + 1);
-        
-        if (ret != 0) {
-                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
-                                "failed");
-                return ret;
+        LOCK (&locks_ctx->ltable_lock);
+        {
+                if (locks_ctx->ltable) {
+                        ltable = locks_ctx->ltable;
+                        locks_ctx->ltable = pl_lock_table_new ();
+                }
         }
+        UNLOCK (&locks_ctx->ltable_lock);
 
+        if (ltable)
+                ltable_delete_locks (ltable);
+
+out:
         return ret;
 }
 
+
 int
 init (xlator_t *this)
 {
         posix_locks_private_t *priv = NULL;
         xlator_list_t         *trav = NULL;
         data_t                *mandatory = NULL;
-	data_t                *trace = NULL;
+        data_t                *trace = NULL;
+        int                   ret = -1;
 
         if (!this->children || this->children->next) {
                 gf_log (this->name, GF_LOG_CRITICAL,
                         "FATAL: posix-locks should have exactly one child");
-                return -1;
+                goto out;
         }
 
         if (!this->parents) {
@@ -1498,10 +2651,10 @@ init (xlator_t *this)
                 gf_log (this->name, GF_LOG_CRITICAL,
                         "'locks' translator is not loaded over a storage "
                         "translator");
-                return -1;
+                goto out;
         }
 
-	priv = GF_CALLOC (1, sizeof (*priv),
+        priv = GF_CALLOC (1, sizeof (*priv),
                           gf_locks_mt_posix_locks_private_t);
 
         mandatory = dict_get (this->options, "mandatory-locks");
@@ -1509,18 +2662,32 @@ init (xlator_t *this)
                 gf_log (this->name, GF_LOG_WARNING,
                         "mandatory locks not supported in this minor release.");
 
-	trace = dict_get (this->options, "trace");
-	if (trace) {
-		if (gf_string2boolean (trace->data,
-				       &priv->trace) == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"'trace' takes on only boolean values.");
-			return -1;
-		}
-	}
+        trace = dict_get (this->options, "trace");
+        if (trace) {
+                if (gf_string2boolean (trace->data,
+                                       &priv->trace) == -1) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "'trace' takes on only boolean values.");
+                        goto out;
+                }
+        }
+
+        this->local_pool = mem_pool_new (pl_local_t, 32);
+        if (!this->local_pool) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto out;
+        }
 
         this->private = priv;
-        return 0;
+        ret = 0;
+
+out:
+        if (ret) {
+                GF_FREE (priv);
+        }
+        return ret;
 }
 
 
@@ -1530,6 +2697,10 @@ fini (xlator_t *this)
         posix_locks_private_t *priv = NULL;
 
         priv = this->private;
+        if (!priv)
+                return 0;
+        this->private = NULL;
+        GF_FREE (priv->brickname);
         GF_FREE (priv);
 
         return 0;
@@ -1538,21 +2709,23 @@ fini (xlator_t *this)
 
 int
 pl_inodelk (call_frame_t *frame, xlator_t *this,
-            const char *volume, loc_t *loc, int32_t cmd, struct flock *flock);
+            const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock,
+            dict_t *xdata);
 
 int
 pl_finodelk (call_frame_t *frame, xlator_t *this,
-             const char *volume, fd_t *fd, int32_t cmd, struct flock *flock);
+             const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock,
+             dict_t *xdata);
 
 int
 pl_entrylk (call_frame_t *frame, xlator_t *this,
             const char *volume, loc_t *loc, const char *basename,
-            entrylk_cmd cmd, entrylk_type type);
+            entrylk_cmd cmd, entrylk_type type, dict_t *xdata);
 
 int
 pl_fentrylk (call_frame_t *frame, xlator_t *this,
              const char *volume, fd_t *fd, const char *basename,
-             entrylk_cmd cmd, entrylk_type type);
+             entrylk_cmd cmd, entrylk_type type, dict_t *xdata);
 
 struct xlator_fops fops = {
         .lookup      = pl_lookup,
@@ -1569,6 +2742,10 @@ struct xlator_fops fops = {
         .fentrylk    = pl_fentrylk,
         .flush       = pl_flush,
         .opendir     = pl_opendir,
+        .readdirp    = pl_readdirp,
+        .getxattr    = pl_getxattr,
+        .fgetxattr   = pl_fgetxattr,
+        .fsetxattr   = pl_fsetxattr,
 };
 
 struct xlator_dumpops dumpops = {
@@ -1576,8 +2753,11 @@ struct xlator_dumpops dumpops = {
 };
 
 struct xlator_cbks cbks = {
-        .forget      = pl_forget,
-        .release     = pl_release,
+        .forget            = pl_forget,
+        .release           = pl_release,
+        .releasedir        = pl_releasedir,
+        .client_destroy    = destroy_cbk,
+        .client_disconnect = disconnect_cbk,
 };
 
 
@@ -1585,8 +2765,8 @@ struct volume_options options[] = {
         { .key  = { "mandatory-locks", "mandatory" },
           .type = GF_OPTION_TYPE_BOOL
         },
-	{ .key  = { "trace" },
-	  .type = GF_OPTION_TYPE_BOOL
-	},
+        { .key  = { "trace" },
+          .type = GF_OPTION_TYPE_BOOL
+        },
         { .key = {NULL} },
 };
diff --git a/xlators/features/locks/src/reservelk.c b/xlators/features/locks/src/reservelk.c
new file mode 100644
index 000000000..11abd26d8
--- /dev/null
+++ b/xlators/features/locks/src/reservelk.c
@@ -0,0 +1,443 @@
+/*
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "compat.h"
+#include "xlator.h"
+#include "inode.h"
+#include "logging.h"
+#include "common-utils.h"
+#include "list.h"
+
+#include "locks.h"
+#include "common.h"
+
+void
+__delete_reserve_lock (posix_lock_t *lock)
+{
+        list_del (&lock->list);
+}
+
+void
+__destroy_reserve_lock (posix_lock_t *lock)
+{
+        GF_FREE (lock);
+}
+
+/* Return true if the two reservelks have exactly same lock boundaries */
+int
+reservelks_equal (posix_lock_t *l1, posix_lock_t *l2)
+{
+        if ((l1->fl_start == l2->fl_start) &&
+            (l1->fl_end == l2->fl_end))
+                return 1;
+
+        return 0;
+}
+
+/* Determine if lock is grantable or not */
+static posix_lock_t *
+__reservelk_grantable (pl_inode_t *pl_inode, posix_lock_t *lock)
+{
+        xlator_t     *this     = NULL;
+        posix_lock_t *l        = NULL;
+        posix_lock_t *ret_lock = NULL;
+
+        this = THIS;
+
+        if (list_empty (&pl_inode->reservelk_list)) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "No reservelks in list");
+                goto out;
+        }
+        list_for_each_entry (l, &pl_inode->reservelk_list, list){
+                if (reservelks_equal (lock, l)) {
+                        ret_lock = l;
+                        break;
+                }
+        }
+out:
+        return ret_lock;
+}
+
+static inline int
+__same_owner_reservelk (posix_lock_t *l1, posix_lock_t *l2)
+{
+        return (is_same_lkowner (&l1->owner, &l2->owner));
+
+}
+
+static posix_lock_t *
+__matching_reservelk (pl_inode_t *pl_inode, posix_lock_t *lock)
+{
+        posix_lock_t *l = NULL;
+
+        if (list_empty (&pl_inode->reservelk_list)) {
+            gf_log ("posix-locks", GF_LOG_TRACE,
+                    "reservelk list empty");
+            return NULL;
+        }
+
+        list_for_each_entry (l, &pl_inode->reservelk_list, list) {
+                if (reservelks_equal (l, lock)) {
+                        gf_log ("posix-locks", GF_LOG_TRACE,
+                                "equal reservelk found");
+                        break;
+                }
+        }
+
+        return l;
+}
+
+static int
+__reservelk_conflict (xlator_t *this, pl_inode_t *pl_inode,
+                      posix_lock_t *lock)
+{
+        posix_lock_t *conf = NULL;
+        int ret = 0;
+
+        conf = __matching_reservelk (pl_inode, lock);
+        if (conf) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Matching reservelk found");
+                if (__same_owner_reservelk (lock, conf)) {
+                        list_del_init (&conf->list);
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "Removing the matching reservelk for setlk to progress");
+                        GF_FREE (conf);
+                        ret = 0;
+                } else {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "Conflicting reservelk found");
+                        ret = 1;
+                }
+
+        }
+        return ret;
+
+}
+
+int
+pl_verify_reservelk (xlator_t *this, pl_inode_t *pl_inode,
+                     posix_lock_t *lock, int can_block)
+{
+        int ret = 0;
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                if (__reservelk_conflict (this, pl_inode, lock)) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "Found conflicting reservelk. Blocking until reservelk is unlocked.");
+                        lock->blocked = can_block;
+                        list_add_tail (&lock->list, &pl_inode->blocked_calls);
+                        ret = -1;
+                        goto unlock;
+                }
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "no conflicting reservelk found. Call continuing");
+                ret = 0;
+
+        }
+unlock:
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        return ret;
+
+}
+
+
+/* Determines if lock can be granted and adds the lock. If the lock
+ * is blocking, adds it to the blocked_reservelks.
+ */
+static int
+__lock_reservelk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+                  int can_block)
+{
+        posix_lock_t *conf = NULL;
+        int ret = -EINVAL;
+
+        conf = __reservelk_grantable (pl_inode, lock);
+        if (conf){
+                ret = -EAGAIN;
+                if (can_block == 0)
+                        goto out;
+
+                list_add_tail (&lock->list, &pl_inode->blocked_reservelks);
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => Blocked",
+                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                        lock->client_pid,
+                        lkowner_utoa (&lock->owner),
+                        lock->user_flock.l_start,
+                        lock->user_flock.l_len);
+
+
+                goto out;
+        }
+
+        list_add (&lock->list, &pl_inode->reservelk_list);
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+static posix_lock_t *
+find_matching_reservelk (posix_lock_t *lock, pl_inode_t *pl_inode)
+{
+        posix_lock_t *l = NULL;
+        list_for_each_entry (l, &pl_inode->reservelk_list, list) {
+                if (reservelks_equal (l, lock))
+                        return l;
+        }
+        return NULL;
+}
+
+/* Set F_UNLCK removes a lock which has the exact same lock boundaries
+ * as the UNLCK lock specifies. If such a lock is not found, returns invalid
+ */
+static posix_lock_t *
+__reserve_unlock_lock (xlator_t *this, posix_lock_t *lock, pl_inode_t *pl_inode)
+{
+
+        posix_lock_t *conf = NULL;
+
+        conf = find_matching_reservelk (lock, pl_inode);
+        if (!conf) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        " Matching lock not found for unlock");
+                goto out;
+        }
+        __delete_reserve_lock (conf);
+        gf_log (this->name, GF_LOG_DEBUG,
+                " Matching lock found for unlock");
+
+out:
+        return conf;
+
+
+}
+
+static void
+__grant_blocked_reserve_locks (xlator_t *this, pl_inode_t *pl_inode,
+                               struct list_head *granted)
+{
+        int              bl_ret = 0;
+        posix_lock_t *bl = NULL;
+        posix_lock_t *tmp = NULL;
+
+        struct list_head blocked_list;
+
+        INIT_LIST_HEAD (&blocked_list);
+        list_splice_init (&pl_inode->blocked_reservelks, &blocked_list);
+
+        list_for_each_entry_safe (bl, tmp, &blocked_list, list) {
+
+                list_del_init (&bl->list);
+
+                bl_ret = __lock_reservelk (this, pl_inode, bl, 1);
+
+                if (bl_ret == 0) {
+                        list_add (&bl->list, granted);
+                }
+        }
+        return;
+}
+
+/* Grant all reservelks blocked on lock(s) */
+void
+grant_blocked_reserve_locks (xlator_t *this, pl_inode_t *pl_inode)
+{
+        struct list_head granted;
+        posix_lock_t *lock = NULL;
+        posix_lock_t *tmp = NULL;
+
+        INIT_LIST_HEAD (&granted);
+
+        if (list_empty (&pl_inode->blocked_reservelks)) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "No blocked locks to be granted");
+                return;
+        }
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                __grant_blocked_reserve_locks (this, pl_inode, &granted);
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (lock, tmp, &granted, list) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => Granted",
+                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                        lock->client_pid,
+                        lkowner_utoa (&lock->owner),
+                        lock->user_flock.l_start,
+                        lock->user_flock.l_len);
+
+                STACK_UNWIND_STRICT (lk, lock->frame, 0, 0, &lock->user_flock,
+                                     NULL);
+        }
+
+}
+
+static void
+__grant_blocked_lock_calls (xlator_t *this, pl_inode_t *pl_inode,
+                            struct list_head *granted)
+{
+        int              bl_ret = 0;
+        posix_lock_t *bl = NULL;
+        posix_lock_t *tmp = NULL;
+
+        struct list_head blocked_list;
+
+        INIT_LIST_HEAD (&blocked_list);
+        list_splice_init (&pl_inode->blocked_reservelks, &blocked_list);
+
+        list_for_each_entry_safe (bl, tmp, &blocked_list, list) {
+
+                list_del_init (&bl->list);
+
+                bl_ret = pl_verify_reservelk (this, pl_inode, bl, bl->blocked);
+
+                if (bl_ret == 0) {
+                        list_add_tail (&bl->list, granted);
+                }
+        }
+        return;
+}
+
+void
+grant_blocked_lock_calls (xlator_t *this, pl_inode_t *pl_inode)
+{
+        struct list_head granted;
+        posix_lock_t *lock = NULL;
+        posix_lock_t *tmp = NULL;
+        fd_t         *fd  = NULL;
+
+        int can_block = 0;
+        int32_t cmd = 0;
+        int ret = 0;
+
+        if (list_empty (&pl_inode->blocked_calls)) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "No blocked lock calls to be granted");
+                return;
+        }
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                __grant_blocked_lock_calls (this, pl_inode, &granted);
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (lock, tmp, &granted, list) {
+                fd = fd_from_fdnum (lock);
+
+                if (lock->blocked) {
+                        can_block = 1;
+                        cmd = F_SETLKW;
+                }
+                else
+                        cmd = F_SETLK;
+
+                lock->blocked = 0;
+                ret = pl_setlk (this, pl_inode, lock, can_block);
+                if (ret == -1) {
+                        if (can_block) {
+                                pl_trace_block (this, lock->frame, fd, NULL,
+                                                cmd, &lock->user_flock, NULL);
+                                continue;
+                        } else {
+                                gf_log (this->name, GF_LOG_DEBUG, "returning EAGAIN");
+                                pl_trace_out (this, lock->frame, fd, NULL, cmd,
+                                              &lock->user_flock, -1, EAGAIN, NULL);
+                                pl_update_refkeeper (this, fd->inode);
+                                STACK_UNWIND_STRICT (lk, lock->frame, -1,
+                                                     EAGAIN, &lock->user_flock,
+                                                     NULL);
+                                __destroy_lock (lock);
+                        }
+                }
+
+        }
+
+}
+
+
+int
+pl_reserve_unlock (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock)
+{
+        posix_lock_t *retlock = NULL;
+        int ret = -1;
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                retlock = __reserve_unlock_lock (this, lock, pl_inode);
+                if (!retlock) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Bad Unlock issued on Inode lock");
+                        ret = -EINVAL;
+                        goto out;
+                }
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Reservelk Unlock successful");
+                __destroy_reserve_lock (retlock);
+                ret = 0;
+        }
+out:
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        grant_blocked_reserve_locks (this, pl_inode);
+        grant_blocked_lock_calls (this, pl_inode);
+
+        return ret;
+
+}
+
+int
+pl_reserve_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+                  int can_block)
+{
+        int ret = -EINVAL;
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+
+                        ret = __lock_reservelk (this, pl_inode, lock, can_block);
+                        if (ret < 0)
+                                gf_log (this->name, GF_LOG_TRACE,
+                                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => NOK",
+                                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                                        lock->client_pid,
+                                        lkowner_utoa (&lock->owner),
+                                        lock->user_flock.l_start,
+                                        lock->user_flock.l_len);
+                        else
+                                gf_log (this->name, GF_LOG_TRACE,
+                                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => OK",
+                                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                                        lock->client_pid,
+                                        lkowner_utoa (&lock->owner),
+                                        lock->fl_start,
+                                        lock->fl_end);
+
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+        return ret;
+}
diff --git a/xlators/features/locks/tests/unit-test.c b/xlators/features/locks/tests/unit-test.c
index fc69ce8a9..d2cca32de 100644
--- a/xlators/features/locks/tests/unit-test.c
+++ b/xlators/features/locks/tests/unit-test.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
diff --git a/xlators/features/mac-compat/src/Makefile.am b/xlators/features/mac-compat/src/Makefile.am
index 915c13e30..f8567edce 100644
--- a/xlators/features/mac-compat/src/Makefile.am
+++ b/xlators/features/mac-compat/src/Makefile.am
@@ -1,13 +1,14 @@
 xlator_LTLIBRARIES = mac-compat.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
 
-mac_compat_la_LDFLAGS = -module -avoidversion
+mac_compat_la_LDFLAGS = -module -avoid-version
 
 mac_compat_la_SOURCES = mac-compat.c
 mac_compat_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/features/mac-compat/src/mac-compat.c b/xlators/features/mac-compat/src/mac-compat.c
index 649d2ad5d..7cb550ad5 100644
--- a/xlators/features/mac-compat/src/mac-compat.c
+++ b/xlators/features/mac-compat/src/mac-compat.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -57,7 +47,8 @@ static int32_t apple_xattr_len[] = {
 
 int32_t
 maccomp_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno, dict_t *dict)
+                      int32_t op_ret, int32_t op_errno, dict_t *dict,
+                      dict_t *xdata)
 {
         intptr_t ax = (intptr_t)this->private;
         int i = 0;
@@ -80,7 +71,7 @@ maccomp_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 }
          }
 
-        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict);
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
 
         return 0;
 }
@@ -88,7 +79,7 @@ maccomp_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 int32_t
 maccomp_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                  const char *name)
+                  const char *name, dict_t *xdata)
 {
         intptr_t ax = GF_XATTR_NONE;
         int i = 0;
@@ -109,14 +100,14 @@ maccomp_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
         STACK_WIND (frame, maccomp_getxattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->getxattr,
-                    loc, name);
+                    loc, name, xdata);
         return 0;
 }
 
 
 int32_t
 maccomp_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                   const char *name)
+                   const char *name, dict_t *xdata)
 {
         intptr_t ax = GF_XATTR_NONE;
         int i = 0;
@@ -137,21 +128,21 @@ maccomp_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
         STACK_WIND (frame, maccomp_getxattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fgetxattr,
-                    fd, name);
+                    fd, name, xdata);
         return 0;
 }
 
 
 int32_t
 maccomp_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno)
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         intptr_t ax = (intptr_t)this->private;
 
         if (op_ret == -1 && ax != GF_XATTR_NONE)
                 op_ret = op_errno = 0;
 
-        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno);
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
 
         return 0;
 }
@@ -159,7 +150,7 @@ maccomp_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 int32_t
 maccomp_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
-                  int32_t flags)
+                  int32_t flags, dict_t *xdata)
 {
         intptr_t ax = GF_XATTR_NONE;
         int i = 0;
@@ -177,14 +168,14 @@ maccomp_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
         STACK_WIND (frame, maccomp_setxattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->setxattr,
-                    loc, dict, flags);
+                    loc, dict, flags, xdata);
         return 0;
 }
 
 
 int32_t
 maccomp_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
-                   int32_t flags)
+                   int32_t flags, dict_t *xdata)
 {
         intptr_t ax = GF_XATTR_NONE;
         int i = 0;
@@ -202,7 +193,7 @@ maccomp_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
         STACK_WIND (frame, maccomp_setxattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fsetxattr,
-                    fd, dict, flags);
+                    fd, dict, flags, xdata);
         return 0;
 }
 
@@ -239,8 +230,7 @@ struct xlator_fops fops = {
         .fsetxattr   = maccomp_fsetxattr,
 };
 
-struct xlator_cbks cbks = {
-};
+struct xlator_cbks cbks;
 
 struct volume_options options[] = {
         { .key = {NULL} },
diff --git a/xlators/features/marker/Makefile.am b/xlators/features/marker/Makefile.am
new file mode 100644
index 000000000..a985f42a8
--- /dev/null
+++ b/xlators/features/marker/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+CLEANFILES =
diff --git a/xlators/features/marker/src/Makefile.am b/xlators/features/marker/src/Makefile.am
new file mode 100644
index 000000000..a7c676472
--- /dev/null
+++ b/xlators/features/marker/src/Makefile.am
@@ -0,0 +1,17 @@
+xlator_LTLIBRARIES = marker.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+marker_la_LDFLAGS = -module -avoid-version
+
+marker_la_SOURCES = marker.c marker-quota.c marker-quota-helper.c marker-common.c
+marker_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+noinst_HEADERS = marker-mem-types.h marker.h marker-quota.h marker-quota-helper.h marker-common.h $(top_builddir)/xlators/lib/src/libxlator.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+	-I$(top_srcdir)/xlators/lib/src
+
+AM_CFLAGS = -Wall -fno-strict-aliasing $(GF_CFLAGS)
+
+CLEANFILES =
+
diff --git a/xlators/features/marker/src/marker-common.c b/xlators/features/marker/src/marker-common.c
new file mode 100644
index 000000000..84a718add
--- /dev/null
+++ b/xlators/features/marker/src/marker-common.c
@@ -0,0 +1,69 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+#include <fnmatch.h>
+#include "marker-common.h"
+
+marker_inode_ctx_t *
+marker_inode_ctx_new ()
+{
+        marker_inode_ctx_t *ctx = NULL;
+
+        ctx = GF_CALLOC (1, sizeof (marker_inode_ctx_t),
+                         gf_marker_mt_marker_inode_ctx_t);
+        if (ctx == NULL)
+                goto out;
+
+        ctx->quota_ctx = NULL;
+out:
+        return ctx;
+}
+
+int32_t
+marker_force_inode_ctx_get (inode_t *inode, xlator_t *this,
+                            marker_inode_ctx_t **ctx)
+{
+        int32_t  ret     = -1;
+        uint64_t ctx_int = 0;
+
+        LOCK (&inode->lock);
+        {
+                ret = __inode_ctx_get (inode, this, &ctx_int);
+                if (ret == 0)
+                        *ctx = (marker_inode_ctx_t *) (unsigned long)ctx_int;
+                else {
+                        *ctx = marker_inode_ctx_new ();
+                        if (*ctx == NULL)
+                                goto unlock;
+
+                        ret = __inode_ctx_put (inode, this,
+                                               (uint64_t )(unsigned long) *ctx);
+                        if (ret == -1) {
+                                GF_FREE (*ctx);
+                                goto unlock;
+                        }
+                        ret = 0;
+                }
+        }
+unlock: UNLOCK (&inode->lock);
+
+        return ret;
+}
+
+int
+marker_filter_quota_xattr (dict_t *dict, char *key,
+                           data_t *value, void *data)
+{
+        dict_del (dict, key);
+        return 0;
+}
diff --git a/xlators/features/marker/src/marker-common.h b/xlators/features/marker/src/marker-common.h
new file mode 100644
index 000000000..23dd846cb
--- /dev/null
+++ b/xlators/features/marker/src/marker-common.h
@@ -0,0 +1,27 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _MARKER_COMMON_H
+#define _MARKER_COMMON_H
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "inode.h"
+#include "xlator.h"
+#include "marker.h"
+
+int32_t
+marker_force_inode_ctx_get (inode_t *, xlator_t *, marker_inode_ctx_t **);
+
+int
+marker_filter_quota_xattr (dict_t *, char *, data_t *, void *);
+#endif
diff --git a/xlators/features/marker/src/marker-mem-types.h b/xlators/features/marker/src/marker-mem-types.h
new file mode 100644
index 000000000..1f74d5048
--- /dev/null
+++ b/xlators/features/marker/src/marker-mem-types.h
@@ -0,0 +1,25 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef __MARKER_MEM_TYPES_H__
+#define __MARKER_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_marker_mem_types_ {
+        gf_marker_mt_marker_conf_t = gf_common_mt_end + 1,
+        gf_marker_mt_loc_t,
+        gf_marker_mt_volume_mark,
+        gf_marker_mt_int64_t,
+        gf_marker_mt_quota_inode_ctx_t,
+        gf_marker_mt_marker_inode_ctx_t,
+        gf_marker_mt_inode_contribution_t,
+        gf_marker_mt_end
+};
+#endif
diff --git a/xlators/features/marker/src/marker-quota-helper.c b/xlators/features/marker/src/marker-quota-helper.c
new file mode 100644
index 000000000..af5fed132
--- /dev/null
+++ b/xlators/features/marker/src/marker-quota-helper.c
@@ -0,0 +1,414 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "locking.h"
+#include "marker-quota.h"
+#include "marker-common.h"
+#include "marker-quota-helper.h"
+#include "marker-mem-types.h"
+
+int
+mq_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path)
+{
+        int ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", loc, out);
+        GF_VALIDATE_OR_GOTO ("marker", inode, out);
+        GF_VALIDATE_OR_GOTO ("marker", path, out);
+        /* Not checking for parent because while filling
+         * loc of root, parent will be NULL
+         */
+
+        if (inode) {
+                loc->inode = inode_ref (inode);
+        }
+
+        if (parent)
+                loc->parent = inode_ref (parent);
+
+        loc->path = gf_strdup (path);
+        if (!loc->path) {
+                gf_log ("loc fill", GF_LOG_ERROR, "strdup failed");
+                goto loc_wipe;
+        }
+
+        loc->name = strrchr (loc->path, '/');
+        if (loc->name)
+                loc->name++;
+        else
+                goto loc_wipe;
+
+        ret = 0;
+loc_wipe:
+        if (ret < 0)
+                loc_wipe (loc);
+out:
+        return ret;
+}
+
+
+int32_t
+mq_inode_loc_fill (const char *parent_gfid, inode_t *inode, loc_t *loc)
+{
+        char            *resolvedpath = NULL;
+        inode_t         *parent       = NULL;
+        int              ret          = -1;
+
+        if ((!inode) || (!loc))
+                return ret;
+
+        if ((inode) && __is_root_gfid (inode->gfid)) {
+                loc->parent = NULL;
+                goto ignore_parent;
+        }
+
+        if (parent_gfid == NULL)
+                parent = inode_parent (inode, 0, NULL);
+        else
+                parent = inode_find (inode->table,
+                                     (unsigned char *) parent_gfid);
+
+        if (parent == NULL)
+                goto err;
+
+ignore_parent:
+        ret = inode_path (inode, NULL, &resolvedpath);
+        if (ret < 0)
+                goto err;
+
+        ret = mq_loc_fill (loc, inode, parent, resolvedpath);
+        if (ret < 0)
+                goto err;
+
+err:
+        if (parent)
+                inode_unref (parent);
+
+        GF_FREE (resolvedpath);
+
+        return ret;
+}
+
+
+quota_inode_ctx_t *
+mq_alloc_inode_ctx ()
+{
+        int32_t                  ret    = -1;
+        quota_inode_ctx_t       *ctx    = NULL;
+
+        QUOTA_ALLOC (ctx, quota_inode_ctx_t, ret);
+        if (ret == -1)
+                goto out;
+
+        ctx->size = 0;
+        ctx->dirty = 0;
+        ctx->updation_status = _gf_false;
+        LOCK_INIT (&ctx->lock);
+        INIT_LIST_HEAD (&ctx->contribution_head);
+out:
+        return ctx;
+}
+
+inode_contribution_t *
+mq_get_contribution_node (inode_t *inode, quota_inode_ctx_t *ctx)
+{
+        inode_contribution_t    *contri = NULL;
+        inode_contribution_t    *temp   = NULL;
+
+        if (!inode || !ctx)
+                goto out;
+
+        list_for_each_entry (temp, &ctx->contribution_head, contri_list) {
+                if (uuid_compare (temp->gfid, inode->gfid) == 0) {
+                        contri = temp;
+                        goto out;
+                }
+        }
+out:
+        return contri;
+}
+
+
+int32_t
+mq_delete_contribution_node (dict_t *dict, char *key,
+                             inode_contribution_t *contribution)
+{
+        if (dict_get (dict, key) != NULL)
+                goto out;
+
+        QUOTA_FREE_CONTRIBUTION_NODE (contribution);
+out:
+        return 0;
+}
+
+
+inode_contribution_t *
+__mq_add_new_contribution_node (xlator_t *this, quota_inode_ctx_t *ctx, loc_t *loc)
+{
+        int32_t ret = 0;
+        inode_contribution_t *contribution = NULL;
+
+        if (!loc->parent) {
+                if (!uuid_is_null (loc->pargfid))
+                        loc->parent = inode_find (loc->inode->table,
+                                                  loc->pargfid);
+                if (!loc->parent)
+                        loc->parent = inode_parent (loc->inode, loc->pargfid,
+                                                    loc->name);
+                if (!loc->parent)
+                        goto out;
+        }
+
+        list_for_each_entry (contribution, &ctx->contribution_head, contri_list) {
+                if (loc->parent &&
+                     uuid_compare (contribution->gfid, loc->parent->gfid) == 0) {
+                        goto out;
+                }
+        }
+
+        QUOTA_ALLOC (contribution, inode_contribution_t, ret);
+        if (ret == -1)
+                goto out;
+
+        contribution->contribution = 0;
+
+        uuid_copy (contribution->gfid, loc->parent->gfid);
+
+	LOCK_INIT (&contribution->lock);
+        INIT_LIST_HEAD (&contribution->contri_list);
+
+        list_add_tail (&contribution->contri_list, &ctx->contribution_head);
+
+out:
+        return contribution;
+}
+
+
+inode_contribution_t *
+mq_add_new_contribution_node (xlator_t *this, quota_inode_ctx_t *ctx, loc_t *loc)
+{
+        inode_contribution_t *contribution = NULL;
+
+        if ((ctx == NULL) || (loc == NULL))
+                return NULL;
+
+        if (strcmp (loc->path, "/") == 0)
+                return NULL;
+
+        LOCK (&ctx->lock);
+        {
+                contribution = __mq_add_new_contribution_node (this, ctx, loc);
+        }
+        UNLOCK (&ctx->lock);
+
+        return contribution;
+}
+
+
+int32_t
+mq_dict_set_contribution (xlator_t *this, dict_t *dict,
+                          loc_t *loc)
+{
+        int32_t ret              = -1;
+        char    contri_key [512] = {0, };
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", dict, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc->parent, out);
+
+        GET_CONTRI_KEY (contri_key, loc->parent->gfid, ret);
+        if (ret < 0) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_int64 (dict, contri_key, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "unable to set dict value on %s.",
+                        loc->path);
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+int32_t
+mq_inode_ctx_get (inode_t *inode, xlator_t *this,
+                  quota_inode_ctx_t **ctx)
+{
+        int32_t             ret      = -1;
+        uint64_t            ctx_int  = 0;
+        marker_inode_ctx_t *mark_ctx = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", inode, out);
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+
+        ret = inode_ctx_get (inode, this, &ctx_int);
+        if (ret < 0) {
+                ret = -1;
+                *ctx = NULL;
+                goto out;
+        }
+
+        mark_ctx = (marker_inode_ctx_t *) (unsigned long)ctx_int;
+        if (mark_ctx->quota_ctx == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        *ctx = mark_ctx->quota_ctx;
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+
+quota_inode_ctx_t *
+__mq_inode_ctx_new (inode_t *inode, xlator_t *this)
+{
+        int32_t               ret        = -1;
+        quota_inode_ctx_t    *quota_ctx  = NULL;
+        marker_inode_ctx_t   *mark_ctx   = NULL;
+
+        ret = marker_force_inode_ctx_get (inode, this, &mark_ctx);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "marker_force_inode_ctx_get() failed");
+                goto out;
+        }
+
+        LOCK (&inode->lock);
+        {
+                if (mark_ctx->quota_ctx == NULL) {
+                        quota_ctx = mq_alloc_inode_ctx ();
+                        if (quota_ctx == NULL) {
+                                ret = -1;
+                                goto unlock;
+                        }
+                        mark_ctx->quota_ctx = quota_ctx;
+                } else {
+                        quota_ctx = mark_ctx->quota_ctx;
+                }
+
+                ret = 0;
+        }
+unlock:
+        UNLOCK (&inode->lock);
+out:
+        return quota_ctx;
+}
+
+
+quota_inode_ctx_t *
+mq_inode_ctx_new (inode_t * inode, xlator_t *this)
+{
+        return __mq_inode_ctx_new (inode, this);
+}
+
+quota_local_t *
+mq_local_new ()
+{
+        quota_local_t  *local   = NULL;
+
+        local = mem_get0 (THIS->local_pool);
+        if (!local)
+                goto out;
+
+        local->ref = 1;
+        LOCK_INIT (&local->lock);
+
+        local->ctx = NULL;
+        local->contri = NULL;
+
+out:
+        return local;
+}
+
+quota_local_t *
+mq_local_ref (quota_local_t *local)
+{
+        LOCK (&local->lock);
+        {
+                local->ref ++;
+        }
+        UNLOCK (&local->lock);
+
+        return local;
+}
+
+
+int32_t
+mq_local_unref (xlator_t *this, quota_local_t *local)
+{
+        int32_t ref = 0;
+        if (local == NULL)
+                goto out;
+
+        QUOTA_SAFE_DECREMENT (&local->lock, local->ref, ref);
+
+        if (ref != 0)
+                goto out;
+
+        if (local->fd != NULL)
+                fd_unref (local->fd);
+
+        loc_wipe (&local->loc);
+
+        loc_wipe (&local->parent_loc);
+
+        LOCK_DESTROY (&local->lock);
+
+        mem_put (local);
+out:
+        return 0;
+}
+
+
+inode_contribution_t *
+mq_get_contribution_from_loc (xlator_t *this, loc_t *loc)
+{
+        int32_t               ret          = 0;
+        quota_inode_ctx_t    *ctx          = NULL;
+        inode_contribution_t *contribution = NULL;
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret < 0) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                                  "cannot get marker-quota context from inode "
+                                  "(gfid:%s, path:%s)",
+                                  uuid_utoa (loc->inode->gfid), loc->path);
+                goto err;
+        }
+
+        contribution = mq_get_contribution_node (loc->parent, ctx);
+        if (contribution == NULL) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                                  "inode (gfid:%s, path:%s) has "
+                                  "no contribution towards parent (gfid:%s)",
+                                  uuid_utoa (loc->inode->gfid),
+                                  loc->path, uuid_utoa (loc->parent->gfid));
+                goto err;
+        }
+
+err:
+        return contribution;
+}
diff --git a/xlators/features/marker/src/marker-quota-helper.h b/xlators/features/marker/src/marker-quota-helper.h
new file mode 100644
index 000000000..6cdd14881
--- /dev/null
+++ b/xlators/features/marker/src/marker-quota-helper.h
@@ -0,0 +1,76 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _MARKER_QUOTA_HELPER_H
+#define _MARKER_QUOTA_HELPER
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "marker.h"
+
+#define QUOTA_FREE_CONTRIBUTION_NODE(_contribution)     \
+        do {                                            \
+                list_del (&_contribution->contri_list); \
+                GF_FREE (_contribution);                \
+        } while (0)
+
+#define QUOTA_SAFE_INCREMENT(lock, var)                 \
+        do {                                            \
+                LOCK (lock);                            \
+                        var ++;                         \
+                UNLOCK (lock);                          \
+        } while (0)
+
+#define QUOTA_SAFE_DECREMENT(lock, var, value)  \
+        do {                                    \
+                LOCK (lock);                    \
+                {                               \
+                      value = --var;            \
+                }                               \
+                UNLOCK (lock);                  \
+        } while (0)
+
+inode_contribution_t *
+mq_add_new_contribution_node (xlator_t *, quota_inode_ctx_t *, loc_t *);
+
+int32_t
+mq_dict_set_contribution (xlator_t *, dict_t *, loc_t *);
+
+quota_inode_ctx_t *
+mq_inode_ctx_new (inode_t *, xlator_t *);
+
+int32_t
+mq_inode_ctx_get (inode_t *, xlator_t *, quota_inode_ctx_t **);
+
+int32_t
+mq_delete_contribution_node (dict_t *, char *, inode_contribution_t *);
+
+int32_t
+mq_inode_loc_fill (const char *, inode_t *, loc_t *);
+
+quota_local_t *
+mq_local_new ();
+
+quota_local_t *
+mq_local_ref (quota_local_t *);
+
+int32_t
+mq_local_unref (xlator_t *, quota_local_t *);
+
+inode_contribution_t *
+mq_get_contribution_node (inode_t *, quota_inode_ctx_t *);
+
+inode_contribution_t *
+mq_get_contribution_from_loc (xlator_t *this, loc_t *loc);
+
+#endif
diff --git a/xlators/features/marker/src/marker-quota.c b/xlators/features/marker/src/marker-quota.c
new file mode 100644
index 000000000..6f9af6e13
--- /dev/null
+++ b/xlators/features/marker/src/marker-quota.c
@@ -0,0 +1,2520 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "dict.h"
+#include "xlator.h"
+#include "defaults.h"
+#include "libxlator.h"
+#include "common-utils.h"
+#include "byte-order.h"
+#include "marker-quota.h"
+#include "marker-quota-helper.h"
+
+int
+mq_loc_copy (loc_t *dst, loc_t *src)
+{
+        int ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", dst, out);
+        GF_VALIDATE_OR_GOTO ("marker", src, out);
+
+        if (src->inode == NULL ||
+            src->path == NULL) {
+                gf_log ("marker", GF_LOG_WARNING,
+                        "src loc is not valid");
+                goto out;
+        }
+
+        ret = loc_copy (dst, src);
+out:
+        return ret;
+}
+
+int32_t
+mq_get_local_err (quota_local_t *local,
+                  int32_t *val)
+{
+        int32_t ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", local, out);
+        GF_VALIDATE_OR_GOTO ("marker", val, out);
+
+        LOCK (&local->lock);
+        {
+                *val = local->err;
+        }
+        UNLOCK (&local->lock);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+mq_get_ctx_updation_status (quota_inode_ctx_t *ctx,
+                            gf_boolean_t *status)
+{
+        int32_t   ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+        GF_VALIDATE_OR_GOTO ("marker", status, out);
+
+        LOCK (&ctx->lock);
+        {
+                *status = ctx->updation_status;
+        }
+        UNLOCK (&ctx->lock);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+int32_t
+mq_set_ctx_updation_status (quota_inode_ctx_t *ctx,
+                            gf_boolean_t status)
+{
+        int32_t   ret = -1;
+
+        if (ctx == NULL)
+                goto out;
+
+        LOCK (&ctx->lock);
+        {
+                ctx->updation_status = status;
+        }
+        UNLOCK (&ctx->lock);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+mq_test_and_set_ctx_updation_status (quota_inode_ctx_t *ctx,
+                                     gf_boolean_t *status)
+{
+        int32_t         ret     = -1;
+        gf_boolean_t    temp    = _gf_false;
+
+        GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+        GF_VALIDATE_OR_GOTO ("marker", status, out);
+
+        LOCK (&ctx->lock);
+        {
+                temp = *status;
+                *status = ctx->updation_status;
+                ctx->updation_status = temp;
+        }
+        UNLOCK (&ctx->lock);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+void
+mq_assign_lk_owner (xlator_t *this, call_frame_t *frame)
+{
+        marker_conf_t *conf     = NULL;
+        uint64_t       lk_owner = 0;
+
+        conf = this->private;
+
+        LOCK (&conf->lock);
+        {
+                if (++conf->quota_lk_owner == 0) {
+                        ++conf->quota_lk_owner;
+                }
+
+                lk_owner = conf->quota_lk_owner;
+        }
+        UNLOCK (&conf->lock);
+
+        set_lk_owner_from_uint64 (&frame->root->lk_owner, lk_owner);
+
+        return;
+}
+
+
+int32_t
+mq_loc_fill_from_name (xlator_t *this, loc_t *newloc, loc_t *oldloc,
+                       uint64_t ino, char *name)
+{
+        int32_t   ret  = -1;
+        int32_t   len  = 0;
+        char     *path = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", newloc, out);
+        GF_VALIDATE_OR_GOTO ("marker", oldloc, out);
+        GF_VALIDATE_OR_GOTO ("marker", name, out);
+
+        newloc->inode = inode_new (oldloc->inode->table);
+
+        if (!newloc->inode) {
+                ret = -1;
+                goto out;
+        }
+
+        newloc->parent = inode_ref (oldloc->inode);
+        uuid_copy (newloc->pargfid, oldloc->inode->gfid);
+
+        len = strlen (oldloc->path);
+
+        if (oldloc->path [len - 1] == '/')
+                ret = gf_asprintf ((char **) &path, "%s%s",
+                                   oldloc->path, name);
+        else
+                ret = gf_asprintf ((char **) &path, "%s/%s",
+                                   oldloc->path, name);
+
+        if (ret < 0)
+                goto out;
+
+        newloc->path = path;
+
+        newloc->name = strrchr (newloc->path, '/');
+
+        if (newloc->name)
+                newloc->name++;
+
+        gf_log (this->name, GF_LOG_DEBUG, "path = %s name =%s",
+                newloc->path, newloc->name);
+out:
+        return ret;
+}
+
+int32_t
+mq_dirty_inode_updation_done (call_frame_t *frame, void *cookie, xlator_t *this,
+                              int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        QUOTA_STACK_DESTROY (frame, this);
+
+        return 0;
+}
+
+int32_t
+mq_release_lock_on_dirty_inode (call_frame_t *frame, void *cookie, xlator_t *this,
+                                int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        struct gf_flock   lock  = {0, };
+        quota_local_t    *local = NULL;
+        loc_t             loc = {0, };
+	int               ret = -1;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                local->err = -1;
+
+                mq_dirty_inode_updation_done (frame, NULL, this, 0, 0, NULL);
+
+                return 0;
+        }
+
+        if (op_ret == 0)
+                local->ctx->dirty = 0;
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        ret = loc_copy (&loc, &local->loc);
+	if (ret == -1) {
+                local->err = -1;
+                frame->local = NULL;
+                mq_dirty_inode_updation_done (frame, NULL, this, 0, 0, NULL);
+                return 0;
+        }
+
+        if (local->loc.inode == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Inode is NULL, so can't stackwind.");
+                goto out;
+        }
+
+        STACK_WIND (frame,
+                    mq_dirty_inode_updation_done,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &loc, F_SETLKW, &lock, NULL);
+
+        loc_wipe (&loc);
+
+        return 0;
+out:
+        mq_dirty_inode_updation_done (frame, NULL, this, -1, 0, NULL);
+
+        return 0;
+}
+
+int32_t
+mq_mark_inode_undirty (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *dict,
+                       dict_t *xdata)
+{
+        int32_t        ret     = -1;
+        int64_t       *size    = NULL;
+        dict_t        *newdict = NULL;
+        quota_local_t *local   = NULL;
+
+        local = (quota_local_t *) frame->local;
+
+        if (op_ret == -1)
+                goto err;
+
+        if (!dict)
+                goto wind;
+
+        ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+        if (ret)
+                goto wind;
+
+        LOCK (&local->ctx->lock);
+        {
+                local->ctx->size = ntoh64 (*size);
+        }
+        UNLOCK (&local->ctx->lock);
+
+wind:
+        newdict = dict_new ();
+        if (!newdict)
+                goto err;
+
+        ret = dict_set_int8 (newdict, QUOTA_DIRTY_KEY, 0);
+        if (ret)
+                goto err;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+        STACK_WIND (frame, mq_release_lock_on_dirty_inode,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr,
+                    &local->loc, newdict, 0, NULL);
+        ret = 0;
+
+err:
+        if (op_ret == -1 || ret == -1) {
+                local->err = -1;
+
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+int32_t
+mq_update_size_xattr (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, inode_t *inode,
+                      struct iatt *buf, dict_t *dict, struct iatt *postparent)
+{
+        int32_t          ret      = -1;
+        dict_t          *new_dict = NULL;
+        int64_t         *size     = NULL;
+        int64_t         *delta    = NULL;
+        quota_local_t   *local    = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1)
+                goto err;
+
+        if (dict == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Dict is null while updating the size xattr %s",
+                        local->loc.path?local->loc.path:"");
+                goto err;
+        }
+
+        ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+        if (!size) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to get the size, %s",
+                        local->loc.path?local->loc.path:"");
+                goto err;
+        }
+
+        QUOTA_ALLOC_OR_GOTO (delta, int64_t, ret, err);
+
+        *delta = hton64 (local->sum - ntoh64 (*size));
+
+        gf_log (this->name, GF_LOG_DEBUG, "calculated size = %"PRId64", "
+                "original size = %"PRIu64
+                " path = %s diff = %"PRIu64, local->sum, ntoh64 (*size),
+                local->loc.path, ntoh64 (*delta));
+
+        new_dict = dict_new ();
+        if (!new_dict);
+
+        ret = dict_set_bin (new_dict, QUOTA_SIZE_KEY, delta, 8);
+        if (ret)
+                goto err;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame, mq_mark_inode_undirty, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop, &local->loc,
+                    GF_XATTROP_ADD_ARRAY64, new_dict, NULL);
+
+        ret = 0;
+
+err:
+        if (op_ret == -1 || ret == -1) {
+                local->err = -1;
+
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (new_dict)
+                dict_unref (new_dict);
+
+        return 0;
+}
+
+int32_t
+mq_test_and_set_local_err(quota_local_t *local,
+                          int32_t *val)
+{
+        int     tmp = 0;
+        int32_t ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", local, out);
+        GF_VALIDATE_OR_GOTO ("marker", val, out);
+
+        LOCK (&local->lock);
+        {
+                tmp = local->err;
+                local->err = *val;
+                *val = tmp;
+        }
+        UNLOCK (&local->lock);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+mq_get_dirty_inode_size (call_frame_t *frame, xlator_t *this)
+{
+        int32_t        ret   = -1;
+        dict_t        *dict  = NULL;
+        quota_local_t *local = NULL;
+
+        local = (quota_local_t *) frame->local;
+
+        dict = dict_new ();
+        if (!dict) {
+                ret = -1;
+                goto err;
+        }
+
+        ret = dict_set_int64 (dict, QUOTA_SIZE_KEY, 0);
+        if (ret)
+                goto err;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame, mq_update_size_xattr, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, &local->loc, dict);
+        ret =0;
+
+err:
+        if (ret) {
+                local->err = -1;
+
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (dict)
+                dict_unref (dict);
+
+        return 0;
+}
+
+int32_t
+mq_get_child_contribution (call_frame_t *frame,
+                           void *cookie,
+                           xlator_t *this,
+                           int32_t op_ret,
+                           int32_t op_errno,
+                           inode_t *inode,
+                           struct iatt *buf,
+                           dict_t *dict,
+                           struct iatt *postparent)
+{
+        int32_t        ret                = -1;
+        int32_t        val                = 0;
+        char           contri_key [512]   = {0, };
+        int64_t       *contri             = NULL;
+        quota_local_t *local              = NULL;
+
+        local = frame->local;
+
+        frame->local = NULL;
+
+        QUOTA_STACK_DESTROY (frame, this);
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "%s",
+                        strerror (op_errno));
+                val = -2;
+                if (!mq_test_and_set_local_err (local, &val) &&
+                    val != -2)
+                        mq_release_lock_on_dirty_inode (local->frame, NULL,
+                                                        this, 0, 0, NULL);
+
+                goto exit;
+        }
+
+        ret = mq_get_local_err (local, &val);
+        if (!ret && val == -2)
+                goto exit;
+
+        GET_CONTRI_KEY (contri_key, local->loc.inode->gfid, ret);
+        if (ret < 0)
+                goto out;
+
+        if (!dict)
+                goto out;
+
+        if (dict_get_bin (dict, contri_key, (void **) &contri) == 0)
+                local->sum += ntoh64 (*contri);
+
+out:
+        LOCK (&local->lock);
+        {
+                val = --local->dentry_child_count;
+        }
+        UNLOCK (&local->lock);
+
+        if (val == 0) {
+                mq_dirty_inode_readdir (local->frame, NULL, this,
+                                        0, 0, NULL, NULL);
+        }
+        mq_local_unref (this, local);
+
+        return 0;
+exit:
+        mq_local_unref (this, local);
+        return 0;
+}
+
+int32_t
+mq_readdir_cbk (call_frame_t *frame,
+                void *cookie,
+                xlator_t *this,
+                int32_t op_ret,
+                int32_t op_errno,
+                gf_dirent_t *entries, dict_t *xdata)
+{
+        char           contri_key [512]   = {0, };
+        int32_t        ret                = 0;
+        int32_t        val                = 0;
+        off_t          offset             = 0;
+        int32_t        count              = 0;
+        dict_t        *dict               = NULL;
+        quota_local_t *local              = NULL;
+        gf_dirent_t   *entry              = NULL;
+        call_frame_t  *newframe           = NULL;
+        loc_t          loc                = {0, };
+
+        local = mq_local_ref (frame->local);
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "readdir failed %s", strerror (op_errno));
+                local->err = -1;
+
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+
+                goto end;
+        } else if (op_ret == 0) {
+                mq_get_dirty_inode_size (frame, this);
+
+                goto end;
+        }
+
+        local->dentry_child_count =  0;
+
+        list_for_each_entry (entry, (&entries->list), list) {
+                gf_log (this->name, GF_LOG_DEBUG, "entry  = %s", entry->d_name);
+
+                if ((!strcmp (entry->d_name, ".")) || (!strcmp (entry->d_name,
+                                                                ".."))) {
+                        gf_log (this->name, GF_LOG_DEBUG, "entry  = %s",
+                                entry->d_name);
+                        continue;
+                }
+
+                offset = entry->d_off;
+                count++;
+        }
+
+        if (count == 0) {
+                mq_get_dirty_inode_size (frame, this);
+                goto end;
+
+        }
+
+        local->frame = frame;
+
+        LOCK (&local->lock);
+        {
+                local->dentry_child_count = count;
+                local->d_off = offset;
+        }
+        UNLOCK (&local->lock);
+
+
+        list_for_each_entry (entry, (&entries->list), list) {
+                gf_log (this->name, GF_LOG_DEBUG, "entry  = %s", entry->d_name);
+
+                if ((!strcmp (entry->d_name, ".")) || (!strcmp (entry->d_name,
+                                                                ".."))) {
+                        gf_log (this->name, GF_LOG_DEBUG, "entry  = %s",
+                                entry->d_name);
+                        continue;
+                }
+
+                ret = mq_loc_fill_from_name (this, &loc, &local->loc,
+                                             entry->d_ino, entry->d_name);
+                if (ret < 0)
+                        goto out;
+
+                ret = 0;
+
+                LOCK (&local->lock);
+                {
+                        if (local->err != -2) {
+                                newframe = copy_frame (frame);
+                                if (!newframe) {
+                                        ret = -1;
+                                }
+                        } else
+                                ret = -1;
+                }
+                UNLOCK (&local->lock);
+
+                if (ret == -1)
+                        goto out;
+
+                newframe->local = mq_local_ref (local);
+
+                dict = dict_new ();
+                if (!dict) {
+                        ret = -1;
+                        goto out;
+                }
+
+                GET_CONTRI_KEY (contri_key, local->loc.inode->gfid, ret);
+                if (ret < 0)
+                        goto out;
+
+                ret = dict_set_int64 (dict, contri_key, 0);
+                if (ret)
+                        goto out;
+
+                STACK_WIND (newframe,
+                            mq_get_child_contribution,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->lookup,
+                            &loc, dict);
+
+                offset = entry->d_off;
+
+                loc_wipe (&loc);
+
+                newframe = NULL;
+
+        out:
+                if (dict) {
+                        dict_unref (dict);
+                        dict = NULL;
+                }
+
+                if (ret) {
+                        val = -2;
+                        mq_test_and_set_local_err (local, &val);
+
+                        if (newframe) {
+                                newframe->local = NULL;
+                                mq_local_unref(this, local);
+                                QUOTA_STACK_DESTROY (newframe, this);
+                        }
+
+                        break;
+                }
+        }
+
+        if (ret && val != -2) {
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+        }
+end:
+        mq_local_unref (this, local);
+
+        return 0;
+}
+
+int32_t
+mq_dirty_inode_readdir (call_frame_t *frame,
+                        void *cookie,
+                        xlator_t *this,
+                        int32_t op_ret,
+                        int32_t op_errno,
+                        fd_t *fd, dict_t *xdata)
+{
+        quota_local_t *local = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                local->err = -1;
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+                return 0;
+        }
+
+        if (local->fd == NULL)
+                local->fd = fd_ref (fd);
+
+        STACK_WIND (frame,
+                    mq_readdir_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdir,
+                    local->fd, READDIR_BUF, local->d_off, xdata);
+
+        return 0;
+}
+
+int32_t
+mq_check_if_still_dirty (call_frame_t *frame,
+                         void *cookie,
+                         xlator_t *this,
+                         int32_t op_ret,
+                         int32_t op_errno,
+                         inode_t *inode,
+                         struct iatt *buf,
+                         dict_t *dict,
+                         struct iatt *postparent)
+{
+        int8_t           dirty          = -1;
+        int32_t          ret            = -1;
+        fd_t            *fd             = NULL;
+        quota_local_t   *local          = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to get "
+                        "the dirty xattr for %s", local->loc.path);
+                goto err;
+        }
+
+        if (!dict) {
+                ret = -1;
+                goto err;
+        }
+
+        ret = dict_get_int8 (dict, QUOTA_DIRTY_KEY, &dirty);
+        if (ret)
+                goto err;
+
+        //the inode is not dirty anymore
+        if (dirty == 0) {
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+
+                return 0;
+        }
+
+        fd = fd_create (local->loc.inode, frame->root->pid);
+
+        local->d_off = 0;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+        STACK_WIND(frame,
+                   mq_dirty_inode_readdir,
+                   FIRST_CHILD(this),
+                   FIRST_CHILD(this)->fops->opendir,
+                   &local->loc, fd, NULL);
+
+        ret = 0;
+
+err:
+        if (op_ret == -1 || ret == -1) {
+                local->err = -1;
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (fd != NULL) {
+                fd_unref (fd);
+        }
+
+        return 0;
+}
+
+int32_t
+mq_get_dirty_xattr (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t        ret       = -1;
+        dict_t        *xattr_req = NULL;
+        quota_local_t *local     = NULL;
+
+        if (op_ret == -1) {
+                mq_dirty_inode_updation_done (frame, NULL, this, 0, 0, NULL);
+                return 0;
+        }
+
+        local = frame->local;
+
+        xattr_req = dict_new ();
+        if (xattr_req == NULL) {
+                ret = -1;
+                goto err;
+        }
+
+        ret = dict_set_int8 (xattr_req, QUOTA_DIRTY_KEY, 0);
+        if (ret)
+                goto err;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame,
+                    mq_check_if_still_dirty,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup,
+                    &local->loc,
+                    xattr_req);
+        ret = 0;
+
+err:
+        if (ret) {
+                local->err = -1;
+                mq_release_lock_on_dirty_inode(frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (xattr_req)
+                dict_unref (xattr_req);
+
+        return 0;
+}
+
+/* return 1 when dirty updation started
+ * 0 other wise
+ */
+int32_t
+mq_update_dirty_inode (xlator_t *this,
+                       loc_t *loc,
+                       quota_inode_ctx_t *ctx,
+                       inode_contribution_t *contribution)
+{
+        int32_t          ret        = -1;
+        quota_local_t   *local      = NULL;
+        gf_boolean_t    status     = _gf_false;
+        struct gf_flock  lock       = {0, };
+        call_frame_t    *frame      = NULL;
+
+        ret = mq_get_ctx_updation_status (ctx, &status);
+        if (ret == -1 || status == _gf_true) {
+                ret = 0;
+                goto out;
+        }
+
+        frame = create_frame (this, this->ctx->pool);
+        if (frame == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        mq_assign_lk_owner (this, frame);
+
+        local = mq_local_new ();
+        if (local == NULL)
+                goto fr_destroy;
+
+        frame->local = local;
+        ret = mq_loc_copy (&local->loc, loc);
+        if (ret < 0)
+                goto fr_destroy;
+
+        local->ctx = ctx;
+
+        local->contri = contribution;
+
+        lock.l_type = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start = 0;
+        lock.l_len = 0;
+
+        if (local->loc.inode == NULL) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Inode is NULL, so can't stackwind.");
+                goto fr_destroy;
+        }
+
+        STACK_WIND (frame,
+                    mq_get_dirty_xattr,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->loc, F_SETLKW, &lock, NULL);
+        return 1;
+
+fr_destroy:
+        QUOTA_STACK_DESTROY (frame, this);
+out:
+
+        return 0;
+}
+
+
+int32_t
+mq_inode_creation_done (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        quota_local_t *local = NULL;
+
+        if (frame == NULL)
+                return 0;
+
+        local = frame->local;
+
+        if (local != NULL) {
+                mq_initiate_quota_txn (this, &local->loc);
+        }
+
+        QUOTA_STACK_DESTROY (frame, this);
+
+        return 0;
+}
+
+
+int32_t
+mq_xattr_creation_release_lock (call_frame_t *frame, void *cookie,
+                                xlator_t *this, int32_t op_ret,
+                                int32_t op_errno, dict_t *xdata)
+{
+        struct gf_flock  lock  = {0, };
+        quota_local_t   *local = NULL;
+
+        local = frame->local;
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        STACK_WIND (frame,
+                    mq_inode_creation_done,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->loc,
+                    F_SETLKW, &lock, NULL);
+
+        return 0;
+}
+
+
+int32_t
+mq_create_dirty_xattr (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *dict,
+                       dict_t *xdata)
+{
+        int32_t          ret       = -1;
+        dict_t          *newdict   = NULL;
+        quota_local_t   *local     = NULL;
+
+        if (op_ret < 0) {
+                goto err;
+        }
+
+        local = frame->local;
+
+        if (local->loc.inode->ia_type == IA_IFDIR) {
+                newdict = dict_new ();
+                if (!newdict) {
+                        goto err;
+                }
+
+                ret = dict_set_int8 (newdict, QUOTA_DIRTY_KEY, 0);
+                if (ret == -1) {
+                        goto err;
+                }
+
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+                GF_UUID_ASSERT (local->loc.gfid);
+
+                STACK_WIND (frame, mq_xattr_creation_release_lock,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->setxattr,
+                            &local->loc, newdict, 0, NULL);
+        } else {
+                mq_xattr_creation_release_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        ret = 0;
+
+err:
+        if (ret < 0) {
+                mq_xattr_creation_release_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict != NULL)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+
+int32_t
+mq_create_xattr (xlator_t *this, call_frame_t *frame)
+{
+        int32_t               ret       = 0;
+        int64_t              *value     = NULL;
+        int64_t              *size      = NULL;
+        dict_t               *dict      = NULL;
+        char                  key[512]  = {0, };
+        quota_local_t        *local     = NULL;
+        quota_inode_ctx_t    *ctx       = NULL;
+        inode_contribution_t *contri    = NULL;
+
+        if (frame == NULL || this == NULL)
+                return 0;
+
+        local = frame->local;
+
+        ret = mq_inode_ctx_get (local->loc.inode, this, &ctx);
+        if (ret < 0) {
+                ctx = mq_inode_ctx_new (local->loc.inode, this);
+                if (ctx == NULL) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "mq_inode_ctx_new failed");
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        if (local->loc.inode->ia_type == IA_IFDIR) {
+                QUOTA_ALLOC_OR_GOTO (size, int64_t, ret, err);
+                ret = dict_set_bin (dict, QUOTA_SIZE_KEY, size, 8);
+                if (ret < 0)
+                        goto free_size;
+        }
+
+        if (strcmp (local->loc.path, "/") != 0) {
+                contri = mq_add_new_contribution_node (this, ctx, &local->loc);
+                if (contri == NULL)
+                        goto err;
+
+                QUOTA_ALLOC_OR_GOTO (value, int64_t, ret, err);
+                GET_CONTRI_KEY (key, local->loc.parent->gfid, ret);
+
+                ret = dict_set_bin (dict, key, value, 8);
+                if (ret < 0)
+                        goto free_value;
+        }
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame, mq_create_dirty_xattr, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop, &local->loc,
+                    GF_XATTROP_ADD_ARRAY64, dict, NULL);
+        ret = 0;
+
+free_size:
+        if (ret < 0) {
+                GF_FREE (size);
+        }
+
+free_value:
+        if (ret < 0) {
+                GF_FREE (value);
+        }
+
+err:
+        dict_unref (dict);
+
+out:
+        if (ret < 0) {
+                mq_xattr_creation_release_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        return 0;
+}
+
+
+int32_t
+mq_check_n_set_inode_xattr (call_frame_t *frame, void *cookie,
+                            xlator_t *this, int32_t op_ret, int32_t op_errno,
+                            inode_t *inode, struct iatt *buf, dict_t *dict,
+                            struct iatt *postparent)
+{
+        quota_local_t        *local           = NULL;
+        int64_t              *size            = NULL, *contri = NULL;
+        int8_t                dirty           = 0;
+        int32_t               ret             = 0;
+        char                  contri_key[512] = {0, };
+
+        if (op_ret < 0) {
+                goto out;
+        }
+
+        local = frame->local;
+
+        ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+        if (ret < 0)
+                goto create_xattr;
+
+        ret = dict_get_int8 (dict, QUOTA_DIRTY_KEY, &dirty);
+        if (ret < 0)
+                goto create_xattr;
+
+        //check contribution xattr if not root
+        if (strcmp (local->loc.path, "/") != 0) {
+                GET_CONTRI_KEY (contri_key, local->loc.parent->gfid, ret);
+                if (ret < 0)
+                        goto out;
+
+                ret = dict_get_bin (dict, contri_key, (void **) &contri);
+                if (ret < 0)
+                        goto create_xattr;
+        }
+
+out:
+        mq_xattr_creation_release_lock (frame, NULL, this, 0, 0, NULL);
+        return 0;
+
+create_xattr:
+        if (uuid_is_null (local->loc.gfid)) {
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+        }
+
+        mq_create_xattr (this, frame);
+        return 0;
+}
+
+
+int32_t
+mq_get_xattr (call_frame_t *frame, void *cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        dict_t        *xattr_req = NULL;
+        quota_local_t *local     = NULL;
+        int32_t        ret       = 0;
+
+        if (op_ret < 0) {
+                goto lock_err;
+        }
+
+        local = frame->local;
+
+        xattr_req = dict_new ();
+        if (xattr_req == NULL) {
+                goto err;
+        }
+
+        ret = mq_req_xattr (this, &local->loc, xattr_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "cannot request xattr");
+                goto err;
+        }
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame, mq_check_n_set_inode_xattr, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, &local->loc, xattr_req);
+
+        dict_unref (xattr_req);
+
+        return 0;
+
+err:
+        mq_xattr_creation_release_lock (frame, NULL, this, 0, 0, NULL);
+
+        if (xattr_req)
+                dict_unref (xattr_req);
+        return 0;
+
+lock_err:
+        mq_inode_creation_done (frame, NULL, this, 0, 0, NULL);
+        return 0;
+}
+
+
+int32_t
+mq_set_inode_xattr (xlator_t *this, loc_t *loc)
+{
+        struct gf_flock  lock  = {0, };
+        quota_local_t   *local = NULL;
+        int32_t          ret   = 0;
+        call_frame_t    *frame = NULL;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame) {
+                ret = -1;
+                goto err;
+        }
+
+        local = mq_local_new ();
+        if (local == NULL) {
+                goto err;
+        }
+
+        frame->local = local;
+
+        ret = loc_copy (&local->loc, loc);
+        if (ret < 0) {
+                goto err;
+        }
+
+        frame->local = local;
+
+        lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+        STACK_WIND (frame,
+                    mq_get_xattr,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->loc, F_SETLKW, &lock, NULL);
+
+        return 0;
+
+err:
+        QUOTA_STACK_DESTROY (frame, this);
+
+        return 0;
+}
+
+
+int32_t
+mq_get_parent_inode_local (xlator_t *this, quota_local_t *local)
+{
+        int32_t            ret = -1;
+        quota_inode_ctx_t *ctx = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", local, out);
+
+        local->contri = NULL;
+
+        loc_wipe (&local->loc);
+
+        ret = mq_loc_copy (&local->loc, &local->parent_loc);
+        if (ret < 0) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                        "loc copy failed");
+                goto out;
+        }
+
+        loc_wipe (&local->parent_loc);
+
+        ret = mq_inode_loc_fill (NULL, local->loc.parent,
+                                 &local->parent_loc);
+        if (ret < 0) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                        "failed to build parent loc of %s",
+                        local->loc.path);
+                goto out;
+        }
+
+        ret = mq_inode_ctx_get (local->loc.inode, this, &ctx);
+        if (ret < 0) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                        "inode ctx get failed");
+                goto out;
+        }
+
+        local->ctx = ctx;
+
+        if (list_empty (&ctx->contribution_head)) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                        "contribution node list is empty which "
+                        "is an error");
+                ret = -1;
+                goto out;
+        }
+
+        local->contri = (inode_contribution_t *) ctx->contribution_head.next;
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+int32_t
+mq_xattr_updation_done (call_frame_t *frame,
+                        void *cookie,
+                        xlator_t *this,
+                        int32_t op_ret,
+                        int32_t op_errno,
+                        dict_t *dict, dict_t *xdata)
+{
+        QUOTA_STACK_DESTROY (frame, this);
+        return 0;
+}
+
+
+int32_t
+mq_inodelk_cbk (call_frame_t *frame, void *cookie,
+                xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t         ret    = 0;
+        gf_boolean_t    status = _gf_false;
+        quota_local_t  *local  = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1 || local->err) {
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "unlocking failed on path (%s)(%s)",
+                                local->parent_loc.path, strerror (op_errno));
+                }
+                mq_xattr_updation_done (frame, NULL, this, 0, 0, NULL, NULL);
+
+                return 0;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "inodelk released on %s", local->parent_loc.path);
+
+        if ((strcmp (local->parent_loc.path, "/") == 0)
+            || (local->delta == 0)) {
+                mq_xattr_updation_done (frame, NULL, this, 0, 0, NULL, NULL);
+        } else {
+                ret = mq_get_parent_inode_local (this, local);
+                if (ret < 0) {
+                        mq_xattr_updation_done (frame, NULL, this, 0, 0, NULL,
+                                                NULL);
+                        goto out;
+                }
+                status = _gf_true;
+
+                ret = mq_test_and_set_ctx_updation_status (local->ctx, &status);
+                if (ret == 0 && status == _gf_false) {
+                        mq_get_lock_on_parent (frame, this);
+                } else {
+                        mq_xattr_updation_done (frame, NULL, this, 0, 0, NULL,
+                                                NULL);
+                }
+        }
+out:
+        return 0;
+}
+
+
+//now release lock on the parent inode
+int32_t
+mq_release_parent_lock (call_frame_t *frame, void *cookie,
+                        xlator_t *this, int32_t op_ret,
+                        int32_t op_errno, dict_t *xdata)
+{
+        int32_t            ret      = 0;
+        quota_local_t     *local    = NULL;
+        quota_inode_ctx_t *ctx      = NULL;
+        struct gf_flock    lock     = {0, };
+
+        local = frame->local;
+
+        if (local->err != 0) {
+                gf_log_callingfn (this->name,
+                                  (local->err == ENOENT) ? GF_LOG_DEBUG
+                                  : GF_LOG_WARNING,
+                                  "An operation during quota updation "
+                                  "of path (%s) failed (%s)", local->loc.path,
+                                  strerror (local->err));
+        }
+
+        ret = mq_inode_ctx_get (local->parent_loc.inode, this, &ctx);
+        if (ret < 0)
+                goto wind;
+
+        LOCK (&ctx->lock);
+        {
+                ctx->dirty = 0;
+        }
+        UNLOCK (&ctx->lock);
+
+        if (local->parent_loc.inode == NULL) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Invalid parent inode.");
+                goto err;
+        }
+
+wind:
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        STACK_WIND (frame,
+                    mq_inodelk_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->parent_loc,
+                    F_SETLKW, &lock, NULL);
+
+        return 0;
+err:
+        mq_xattr_updation_done (frame, NULL, this,
+                                0, 0 , NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+mq_mark_undirty (call_frame_t *frame,
+                 void *cookie,
+                 xlator_t *this,
+                 int32_t op_ret,
+                 int32_t op_errno,
+                 dict_t *dict, dict_t *xdata)
+{
+        int32_t            ret          = -1;
+        int64_t           *size         = NULL;
+        dict_t            *newdict      = NULL;
+        quota_local_t     *local        = NULL;
+        quota_inode_ctx_t *ctx          = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING, "%s occurred while"
+                        " updating the size of %s", strerror (op_errno),
+                        local->parent_loc.path);
+
+                goto err;
+        }
+
+        //update the size of the parent inode
+        if (dict != NULL) {
+                ret = mq_inode_ctx_get (local->parent_loc.inode, this, &ctx);
+                if (ret < 0) {
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
+                ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+                if (ret < 0) {
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
+                LOCK (&ctx->lock);
+                {
+                        if (size)
+                                ctx->size = ntoh64 (*size);
+                        gf_log (this->name, GF_LOG_DEBUG, "%s %"PRId64,
+                                local->parent_loc.path, ctx->size);
+                }
+                UNLOCK (&ctx->lock);
+        }
+
+        newdict = dict_new ();
+        if (!newdict) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        ret = dict_set_int8 (newdict, QUOTA_DIRTY_KEY, 0);
+
+        if (ret == -1) {
+                op_errno = -ret;
+                goto err;
+        }
+
+        uuid_copy (local->parent_loc.gfid, local->parent_loc.inode->gfid);
+        GF_UUID_ASSERT (local->parent_loc.gfid);
+
+        STACK_WIND (frame, mq_release_parent_lock,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr,
+                    &local->parent_loc, newdict, 0, NULL);
+
+        ret = 0;
+err:
+        if (op_ret == -1 || ret == -1) {
+                local->err = op_errno;
+
+                mq_release_parent_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+
+int32_t
+mq_update_parent_size (call_frame_t *frame,
+                       void *cookie,
+                       xlator_t *this,
+                       int32_t op_ret,
+                       int32_t op_errno,
+                       dict_t *dict, dict_t *xdata)
+{
+        int64_t             *size       = NULL;
+        int32_t              ret        = -1;
+        dict_t              *newdict    = NULL;
+        quota_local_t       *local      = NULL;
+        quota_inode_ctx_t   *ctx        = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, ((op_errno == ENOENT) ? GF_LOG_DEBUG :
+                                     GF_LOG_WARNING),
+                        "xattrop call failed: %s", strerror (op_errno));
+
+                goto err;
+        }
+
+        LOCK (&local->contri->lock);
+        {
+                local->contri->contribution += local->delta;
+        }
+        UNLOCK (&local->contri->lock);
+
+        gf_log (this->name, GF_LOG_DEBUG, "%s %"PRId64 "%"PRId64,
+                local->loc.path, local->ctx->size,
+                local->contri->contribution);
+
+        if (dict == NULL) {
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        ret = mq_inode_ctx_get (local->parent_loc.inode, this, &ctx);
+        if (ret < 0) {
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        newdict = dict_new ();
+        if (!newdict) {
+                op_errno = ENOMEM;
+                ret = -1;
+                goto err;
+        }
+
+        QUOTA_ALLOC_OR_GOTO (size, int64_t, ret, err);
+
+        *size = hton64 (local->delta);
+
+        ret = dict_set_bin (newdict, QUOTA_SIZE_KEY, size, 8);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto err;
+        }
+
+        if (uuid_is_null (local->parent_loc.gfid))
+                        uuid_copy (local->parent_loc.gfid,
+                                   local->parent_loc.inode->gfid);
+        GF_UUID_ASSERT (local->parent_loc.gfid);
+
+        STACK_WIND (frame,
+                    mq_mark_undirty,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop,
+                    &local->parent_loc,
+                    GF_XATTROP_ADD_ARRAY64,
+                    newdict, NULL);
+        ret = 0;
+err:
+        if (op_ret == -1 || ret < 0) {
+                local->err = op_errno;
+                mq_release_parent_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+int32_t
+mq_update_inode_contribution (call_frame_t *frame, void *cookie,
+                              xlator_t *this, int32_t op_ret,
+                              int32_t op_errno, inode_t *inode,
+                              struct iatt *buf, dict_t *dict,
+                              struct iatt *postparent)
+{
+        int32_t               ret              = -1;
+        int64_t              *size             = NULL, size_int = 0, contri_int = 0;
+        int64_t              *contri           = NULL;
+        int64_t              *delta            = NULL;
+        char                  contri_key [512] = {0, };
+        dict_t               *newdict          = NULL;
+        quota_local_t        *local            = NULL;
+        quota_inode_ctx_t    *ctx              = NULL;
+        inode_contribution_t *contribution     = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, ((op_errno == ENOENT) ? GF_LOG_DEBUG :
+                                     GF_LOG_WARNING),
+                        "failed to get size and contribution of path (%s)(%s)",
+                        local->loc.path, strerror (op_errno));
+                goto err;
+        }
+
+        ctx = local->ctx;
+        contribution = local->contri;
+
+        //prepare to update size & contribution of the inode
+        GET_CONTRI_KEY (contri_key, contribution->gfid, ret);
+        if (ret == -1) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                if (local->loc.inode->ia_type == IA_IFDIR ) {
+                        ret = dict_get_bin (dict, QUOTA_SIZE_KEY,
+                                            (void **) &size);
+                        if (ret < 0) {
+                                op_errno = EINVAL;
+                                goto unlock;
+                        }
+
+                        ctx->size = ntoh64 (*size);
+                } else
+                        ctx->size = buf->ia_blocks * 512;
+
+                size_int = ctx->size;
+        }
+unlock:
+        UNLOCK  (&ctx->lock);
+
+        if (ret < 0) {
+                goto err;
+        }
+
+        ret = dict_get_bin (dict, contri_key, (void **) &contri);
+
+        LOCK (&contribution->lock);
+        {
+                if (ret < 0)
+                        contribution->contribution = 0;
+                else
+                        contribution->contribution = ntoh64 (*contri);
+
+                contri_int = contribution->contribution;
+        }
+        UNLOCK (&contribution->lock);
+
+        gf_log (this->name, GF_LOG_DEBUG, "%s %"PRId64 "%"PRId64,
+                local->loc.path, size_int, contri_int);
+
+        local->delta = size_int - contri_int;
+
+        if (local->delta == 0) {
+                mq_mark_undirty (frame, NULL, this, 0, 0, NULL, NULL);
+                return 0;
+        }
+
+        newdict = dict_new ();
+        if (newdict == NULL) {
+                op_errno = ENOMEM;
+                ret = -1;
+                goto err;
+        }
+
+        QUOTA_ALLOC_OR_GOTO (delta, int64_t, ret, err);
+
+        *delta = hton64 (local->delta);
+
+        ret = dict_set_bin (newdict, contri_key, delta, 8);
+        if (ret < 0) {
+                op_errno = -ret;
+                ret = -1;
+                goto err;
+        }
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame,
+                    mq_update_parent_size,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop,
+                    &local->loc,
+                    GF_XATTROP_ADD_ARRAY64,
+                    newdict, NULL);
+        ret = 0;
+
+err:
+        if (op_ret == -1 || ret < 0) {
+                local->err = op_errno;
+
+                mq_release_parent_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+int32_t
+mq_fetch_child_size_and_contri (call_frame_t *frame, void *cookie,
+                                xlator_t *this, int32_t op_ret,
+                                int32_t op_errno, dict_t *xdata)
+{
+        int32_t            ret              = -1;
+        char               contri_key [512] = {0, };
+        dict_t            *newdict          = NULL;
+        quota_local_t     *local            = NULL;
+        quota_inode_ctx_t *ctx              = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, (op_errno == ENOENT) ? GF_LOG_DEBUG
+                        : GF_LOG_WARNING,
+                        "couldnt mark inode corresponding to path (%s) dirty "
+                        "(%s)", local->parent_loc.path, strerror (op_errno));
+                goto err;
+        }
+
+        VALIDATE_OR_GOTO (local->ctx, err);
+        VALIDATE_OR_GOTO (local->contri, err);
+
+        gf_log (this->name, GF_LOG_DEBUG, "%s marked dirty", local->parent_loc.path);
+
+        //update parent ctx
+        ret = mq_inode_ctx_get (local->parent_loc.inode, this, &ctx);
+        if (ret == -1) {
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->dirty = 1;
+        }
+        UNLOCK (&ctx->lock);
+
+        newdict = dict_new ();
+        if (newdict == NULL) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        if (local->loc.inode->ia_type == IA_IFDIR) {
+                ret = dict_set_int64 (newdict, QUOTA_SIZE_KEY, 0);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "dict_set failed.");
+                        goto err;
+                }
+        }
+
+        GET_CONTRI_KEY (contri_key, local->contri->gfid, ret);
+        if (ret < 0) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        ret = dict_set_int64 (newdict, contri_key, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dict_set failed.");
+                goto err;
+        }
+
+        mq_set_ctx_updation_status (local->ctx, _gf_false);
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame, mq_update_inode_contribution, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, &local->loc, newdict);
+
+        ret = 0;
+
+err:
+        if ((op_ret == -1) || (ret < 0)) {
+                local->err = op_errno;
+
+                mq_set_ctx_updation_status (local->ctx, _gf_false);
+
+                mq_release_parent_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+int32_t
+mq_markdirty (call_frame_t *frame, void *cookie,
+              xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t        ret      = -1;
+        dict_t        *dict     = NULL;
+        quota_local_t *local    = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1){
+                gf_log (this->name, (op_errno == ENOENT) ? GF_LOG_DEBUG
+                        : GF_LOG_WARNING, "acquiring locks failed on %s (%s)",
+                        local->parent_loc.path, strerror (op_errno));
+
+                local->err = op_errno;
+
+                mq_set_ctx_updation_status (local->ctx, _gf_false);
+
+                mq_inodelk_cbk (frame, NULL, this, 0, 0, NULL);
+
+                return 0;
+        }
+
+        gf_log (this->name, GF_LOG_TRACE,
+                "inodelk succeeded on  %s", local->parent_loc.path);
+
+        dict = dict_new ();
+        if (!dict) {
+                ret = -1;
+                goto err;
+        }
+
+        ret = dict_set_int8 (dict, QUOTA_DIRTY_KEY, 1);
+        if (ret == -1)
+                goto err;
+
+        uuid_copy (local->parent_loc.gfid,
+                   local->parent_loc.inode->gfid);
+        GF_UUID_ASSERT (local->parent_loc.gfid);
+
+        STACK_WIND (frame, mq_fetch_child_size_and_contri,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr,
+                    &local->parent_loc, dict, 0, NULL);
+
+        ret = 0;
+err:
+        if (ret == -1) {
+                local->err = 1;
+
+                mq_set_ctx_updation_status (local->ctx, _gf_false);
+
+                mq_release_parent_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (dict)
+                dict_unref (dict);
+
+        return 0;
+}
+
+
+int32_t
+mq_get_lock_on_parent (call_frame_t *frame, xlator_t *this)
+{
+        struct gf_flock  lock  = {0, };
+        quota_local_t   *local = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", frame, fr_destroy);
+
+        local = frame->local;
+        gf_log (this->name, GF_LOG_DEBUG, "taking lock on %s",
+                local->parent_loc.path);
+
+        if (local->parent_loc.inode == NULL) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "parent inode is not valid, aborting "
+                        "transaction.");
+                goto fr_destroy;
+        }
+
+        lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+        STACK_WIND (frame,
+                    mq_markdirty,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->parent_loc, F_SETLKW, &lock, NULL);
+
+        return 0;
+
+fr_destroy:
+        QUOTA_STACK_DESTROY (frame, this);
+
+        return -1;
+}
+
+
+int
+mq_start_quota_txn (xlator_t *this, loc_t *loc,
+                    quota_inode_ctx_t *ctx,
+                    inode_contribution_t *contri)
+{
+        int32_t        ret      = -1;
+        call_frame_t  *frame    = NULL;
+        quota_local_t *local    = NULL;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (frame == NULL)
+                goto err;
+
+        mq_assign_lk_owner (this, frame);
+
+        local = mq_local_new ();
+        if (local == NULL)
+                goto fr_destroy;
+
+        frame->local = local;
+
+        ret = mq_loc_copy (&local->loc, loc);
+        if (ret < 0)
+                goto fr_destroy;
+
+        ret = mq_inode_loc_fill (NULL, local->loc.parent,
+                                 &local->parent_loc);
+        if (ret < 0)
+                goto fr_destroy;
+
+        local->ctx = ctx;
+        local->contri = contri;
+
+        ret = mq_get_lock_on_parent (frame, this);
+        if (ret == -1)
+                goto err;
+
+        return 0;
+
+fr_destroy:
+        QUOTA_STACK_DESTROY (frame, this);
+err:
+        mq_set_ctx_updation_status (ctx, _gf_false);
+
+        return -1;
+}
+
+
+int
+mq_initiate_quota_txn (xlator_t *this, loc_t *loc)
+{
+        int32_t               ret          = -1;
+        gf_boolean_t          status       = _gf_false;
+        quota_inode_ctx_t    *ctx          = NULL;
+        inode_contribution_t *contribution = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "inode ctx get failed, aborting quota txn");
+                ret = -1;
+                goto out;
+        }
+
+        contribution = mq_get_contribution_node (loc->parent, ctx);
+        if (contribution == NULL)
+                goto out;
+
+        /* To improve performance, donot start another transaction
+         * if one is already in progress for same inode
+         */
+        status = _gf_true;
+
+        ret = mq_test_and_set_ctx_updation_status (ctx, &status);
+        if (ret < 0)
+                goto out;
+
+        if (status == _gf_false) {
+                mq_start_quota_txn (this, loc, ctx, contribution);
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+/* int32_t */
+/* validate_inode_size_contribution (xlator_t *this, loc_t *loc, int64_t size, */
+/*                                int64_t contribution) */
+/* { */
+/*   if (size != contribution) { */
+/*     mq_initiate_quota_txn (this, loc); */
+/*   } */
+
+/*   return 0; */
+/* } */
+
+
+int32_t
+mq_inspect_directory_xattr (xlator_t *this,
+                            loc_t *loc,
+                            dict_t *dict,
+                            struct iatt buf)
+{
+        int32_t               ret                 = 0;
+        int8_t                dirty               = -1;
+        int64_t              *size                = NULL, size_int = 0;
+        int64_t              *contri              = NULL, contri_int = 0;
+        char                  contri_key [512]    = {0, };
+        gf_boolean_t          not_root            = _gf_false;
+        quota_inode_ctx_t    *ctx                 = NULL;
+        inode_contribution_t *contribution        = NULL;
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret < 0) {
+                ctx = mq_inode_ctx_new (loc->inode, this);
+                if (ctx == NULL) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "mq_inode_ctx_new failed");
+                        ret = -1;
+                        goto err;
+                }
+        }
+
+        if (strcmp (loc->path, "/") != 0) {
+                contribution = mq_add_new_contribution_node (this, ctx, loc);
+                if (contribution == NULL) {
+                        if (!uuid_is_null (loc->inode->gfid))
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "cannot add a new contribution node");
+                        ret = -1;
+                        goto err;
+                }
+        }
+
+        ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+        if (ret < 0)
+                goto out;
+
+        ret = dict_get_int8 (dict, QUOTA_DIRTY_KEY, &dirty);
+        if (ret < 0)
+                goto out;
+
+        if (strcmp (loc->path, "/") != 0) {
+                not_root = _gf_true;
+
+                GET_CONTRI_KEY (contri_key, contribution->gfid, ret);
+                if (ret < 0)
+                        goto out;
+
+                ret = dict_get_bin (dict, contri_key, (void **) &contri);
+                if (ret < 0)
+                        goto out;
+
+                LOCK (&contribution->lock);
+                {
+                        contribution->contribution = ntoh64 (*contri);
+                        contri_int = contribution->contribution;
+                }
+                UNLOCK (&contribution->lock);
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->size = ntoh64 (*size);
+                ctx->dirty = dirty;
+                size_int = ctx->size;
+        }
+        UNLOCK (&ctx->lock);
+
+        gf_log (this->name, GF_LOG_DEBUG, "size=%"PRId64
+                " contri=%"PRId64, size_int, contri_int);
+
+        if (dirty) {
+                ret = mq_update_dirty_inode (this, loc, ctx, contribution);
+        }
+
+        if ((!dirty || ret == 0) && (not_root == _gf_true) &&
+            (size_int != contri_int)) {
+                mq_initiate_quota_txn (this, loc);
+        }
+
+        ret = 0;
+out:
+        if (ret)
+                mq_set_inode_xattr (this, loc);
+err:
+        return ret;
+}
+
+int32_t
+mq_inspect_file_xattr (xlator_t *this,
+                       loc_t *loc,
+                       dict_t *dict,
+                       struct iatt buf)
+{
+        int32_t               ret              = -1;
+        uint64_t              contri_int       = 0, size = 0;
+        int64_t              *contri_ptr       = NULL;
+        char                  contri_key [512] = {0, };
+        quota_inode_ctx_t    *ctx              = NULL;
+        inode_contribution_t *contribution     = NULL;
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret < 0) {
+                ctx = mq_inode_ctx_new (loc->inode, this);
+                if (ctx == NULL) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "mq_inode_ctx_new failed");
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        contribution = mq_add_new_contribution_node (this, ctx, loc);
+        if (contribution == NULL)
+                goto out;
+
+        LOCK (&ctx->lock);
+        {
+                ctx->size = 512 * buf.ia_blocks;
+                size = ctx->size;
+        }
+        UNLOCK (&ctx->lock);
+
+        list_for_each_entry (contribution, &ctx->contribution_head,
+                             contri_list) {
+                GET_CONTRI_KEY (contri_key, contribution->gfid, ret);
+                if (ret < 0)
+                        continue;
+
+                ret = dict_get_bin (dict, contri_key, (void **) &contri_int);
+                if (ret == 0) {
+                        contri_ptr = (int64_t *)(unsigned long)contri_int;
+
+                        LOCK (&contribution->lock);
+                        {
+                                contribution->contribution = ntoh64 (*contri_ptr);
+                                contri_int = contribution->contribution;
+                        }
+                        UNLOCK (&contribution->lock);
+
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "size=%"PRId64 " contri=%"PRId64, size, contri_int);
+
+                        if (size != contri_int) {
+                                mq_initiate_quota_txn (this, loc);
+                        }
+                } else
+                        mq_initiate_quota_txn (this, loc);
+        }
+
+out:
+        return ret;
+}
+
+int32_t
+mq_xattr_state (xlator_t *this,
+                loc_t *loc,
+                dict_t *dict,
+                struct iatt buf)
+{
+        if (buf.ia_type == IA_IFREG ||
+            buf.ia_type == IA_IFLNK) {
+                mq_inspect_file_xattr (this, loc, dict, buf);
+        } else if (buf.ia_type == IA_IFDIR)
+                mq_inspect_directory_xattr (this, loc, dict, buf);
+
+        return 0;
+}
+
+int32_t
+mq_req_xattr (xlator_t *this,
+              loc_t *loc,
+              dict_t *dict)
+{
+        int32_t               ret       = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", dict, out);
+
+        if (!loc)
+                goto set_size;
+
+        //if not "/" then request contribution
+        if (strcmp (loc->path, "/") == 0)
+                goto set_size;
+
+        ret = mq_dict_set_contribution (this, dict, loc);
+        if (ret == -1)
+                goto out;
+
+set_size:
+        ret = dict_set_uint64 (dict, QUOTA_SIZE_KEY, 0);
+        if (ret < 0) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_int8 (dict, QUOTA_DIRTY_KEY, 0);
+        if (ret < 0) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+
+int32_t
+mq_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        QUOTA_STACK_DESTROY (frame, this);
+
+        return 0;
+}
+
+int32_t
+_mq_inode_remove_done (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t        ret                = 0;
+        char           contri_key [512]   = {0, };
+        quota_local_t *local              = NULL;
+
+        local = (quota_local_t *) frame->local;
+
+        if (op_ret == -1 || local->err == -1) {
+                mq_removexattr_cbk (frame, NULL, this, -1, 0, NULL);
+                return 0;
+        }
+
+        frame->local = NULL;
+
+        if (local->hl_count > 1) {
+                GET_CONTRI_KEY (contri_key, local->contri->gfid, ret);
+
+                STACK_WIND (frame, mq_removexattr_cbk, FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->removexattr,
+                            &local->loc, contri_key, NULL);
+                ret = 0;
+        } else {
+                mq_removexattr_cbk (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (strcmp (local->parent_loc.path, "/") != 0) {
+                ret = mq_get_parent_inode_local (this, local);
+                if (ret < 0)
+                        goto out;
+
+                mq_start_quota_txn (this, &local->loc, local->ctx, local->contri);
+        }
+out:
+        mq_local_unref (this, local);
+
+        return 0;
+}
+
+int32_t
+mq_inode_remove_done (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *dict,
+                      dict_t *xdata)
+{
+        int32_t            ret   = -1;
+        struct gf_flock    lock  = {0, };
+        quota_inode_ctx_t *ctx   = NULL;
+        quota_local_t     *local = NULL;
+        int64_t contribution = 0;
+
+        local = frame->local;
+        if (op_ret == -1)
+                local->err = -1;
+
+        ret = mq_inode_ctx_get (local->parent_loc.inode, this, &ctx);
+
+        LOCK (&local->contri->lock);
+        {
+                contribution = local->contri->contribution;
+        }
+        UNLOCK (&local->contri->lock);
+
+        if (contribution == local->size) {
+                if (ret == 0) {
+                        LOCK (&ctx->lock);
+                        {
+                                ctx->size -= contribution;
+                        }
+                        UNLOCK (&ctx->lock);
+
+                        LOCK (&local->contri->lock);
+                        {
+                                local->contri->contribution = 0;
+                        }
+                        UNLOCK (&local->contri->lock);
+                }
+        }
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        STACK_WIND (frame,
+                    _mq_inode_remove_done,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->parent_loc,
+                    F_SETLKW, &lock, NULL);
+        return 0;
+}
+
+int32_t
+mq_reduce_parent_size_xattr (call_frame_t *frame, void *cookie, xlator_t *this,
+                             int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t                  ret               = -1;
+        int64_t                 *size              = NULL;
+        dict_t                  *dict              = NULL;
+        quota_local_t           *local             = NULL;
+
+        local = frame->local;
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "inodelk set failed on %s", local->parent_loc.path);
+                QUOTA_STACK_DESTROY (frame, this);
+                return 0;
+        }
+
+        VALIDATE_OR_GOTO (local->contri, err);
+
+        dict = dict_new ();
+        if (dict == NULL) {
+                ret = -1;
+                goto err;
+        }
+
+        QUOTA_ALLOC_OR_GOTO (size, int64_t, ret, err);
+
+        *size = hton64 (-local->size);
+
+        ret = dict_set_bin (dict, QUOTA_SIZE_KEY, size, 8);
+        if (ret < 0)
+                goto err;
+
+        uuid_copy (local->parent_loc.gfid,
+                   local->parent_loc.inode->gfid);
+        GF_UUID_ASSERT (local->parent_loc.gfid);
+
+        STACK_WIND (frame, mq_inode_remove_done, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop, &local->parent_loc,
+                    GF_XATTROP_ADD_ARRAY64, dict, NULL);
+        dict_unref (dict);
+        return 0;
+
+err:
+        local->err = 1;
+        mq_inode_remove_done (frame, NULL, this, -1, 0, NULL, NULL);
+        if (dict)
+                dict_unref (dict);
+        return 0;
+}
+
+int32_t
+mq_reduce_parent_size (xlator_t *this, loc_t *loc, int64_t contri)
+{
+        int32_t                  ret           = -1;
+        struct gf_flock          lock          = {0,};
+        call_frame_t            *frame         = NULL;
+        quota_local_t           *local         = NULL;
+        quota_inode_ctx_t       *ctx           = NULL;
+        inode_contribution_t    *contribution  = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc, out);
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret < 0)
+                goto out;
+
+        contribution = mq_get_contribution_node (loc->parent, ctx);
+        if (contribution == NULL)
+                goto out;
+
+        local = mq_local_new ();
+        if (local == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        if (contri >= 0) {
+                local->size = contri;
+        } else {
+                LOCK (&contribution->lock);
+                {
+                        local->size = contribution->contribution;
+                }
+                UNLOCK (&contribution->lock);
+        }
+
+        if (local->size == 0) {
+                ret = 0;
+                goto out;
+        }
+
+        ret = mq_loc_copy (&local->loc, loc);
+        if (ret < 0)
+                goto out;
+
+        local->ctx = ctx;
+        local->contri = contribution;
+
+        ret = mq_inode_loc_fill (NULL, loc->parent, &local->parent_loc);
+        if (ret < 0)
+                goto out;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame) {
+                ret = -1;
+                goto out;
+        }
+
+        mq_assign_lk_owner (this, frame);
+
+        frame->local = local;
+
+        lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+        if (local->parent_loc.inode == NULL) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Inode is NULL, so can't stackwind.");
+                goto out;
+        }
+
+        STACK_WIND (frame,
+                    mq_reduce_parent_size_xattr,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->parent_loc, F_SETLKW, &lock, NULL);
+        local = NULL;
+        ret = 0;
+
+out:
+        if (local != NULL)
+                mq_local_unref (this, local);
+
+        return ret;
+}
+
+
+int32_t
+init_quota_priv (xlator_t *this)
+{
+        return 0;
+}
+
+
+int32_t
+mq_rename_update_newpath (xlator_t *this, loc_t *loc)
+{
+        int32_t               ret          = -1;
+        quota_inode_ctx_t    *ctx          = NULL;
+        inode_contribution_t *contribution = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret < 0)
+                goto out;
+
+        contribution = mq_add_new_contribution_node (this, ctx, loc);
+        if (contribution == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        mq_initiate_quota_txn (this, loc);
+out:
+        return ret;
+}
+
+int32_t
+mq_forget (xlator_t *this, quota_inode_ctx_t *ctx)
+{
+        inode_contribution_t *contri = NULL;
+        inode_contribution_t *next   = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+
+        list_for_each_entry_safe (contri, next, &ctx->contribution_head,
+                                  contri_list) {
+                list_del (&contri->contri_list);
+                GF_FREE (contri);
+        }
+
+        LOCK_DESTROY (&ctx->lock);
+        GF_FREE (ctx);
+out:
+        return 0;
+}
diff --git a/xlators/features/marker/src/marker-quota.h b/xlators/features/marker/src/marker-quota.h
new file mode 100644
index 000000000..385760ac4
--- /dev/null
+++ b/xlators/features/marker/src/marker-quota.h
@@ -0,0 +1,130 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _MARKER_QUOTA_H
+#define _MARKER_QUOTA_H
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "marker-mem-types.h"
+
+#define QUOTA_XATTR_PREFIX "trusted.glusterfs"
+#define QUOTA_DIRTY_KEY "trusted.glusterfs.quota.dirty"
+
+#define CONTRIBUTION "contri"
+#define CONTRI_KEY_MAX 512
+#define READDIR_BUF 4096
+
+
+#define QUOTA_STACK_DESTROY(_frame, _this)              \
+        do {                                            \
+                quota_local_t *_local = NULL;           \
+                _local = _frame->local;                 \
+                _frame->local = NULL;                   \
+                STACK_DESTROY (_frame->root);           \
+                mq_local_unref (_this, _local);         \
+        } while (0)
+
+
+#define QUOTA_ALLOC(var, type, ret)                     \
+        do {                                            \
+                ret = 0;                                \
+                var = GF_CALLOC (sizeof (type), 1,      \
+                                 gf_marker_mt_##type);  \
+                if (!var) {                             \
+                        gf_log ("", GF_LOG_ERROR,       \
+                                "out of memory");       \
+                        ret = -1;                       \
+                }                                       \
+        } while (0);
+
+#define QUOTA_ALLOC_OR_GOTO(var, type, ret, label)      \
+        do {                                            \
+                var = GF_CALLOC (sizeof (type), 1,      \
+                                 gf_marker_mt_##type);  \
+                if (!var) {                             \
+                        gf_log ("", GF_LOG_ERROR,       \
+                                "out of memory");       \
+                        ret = -1;                       \
+                        goto label;                     \
+                }                                       \
+                ret = 0;                                \
+        } while (0);
+
+#define GET_CONTRI_KEY(var, _gfid, _ret)        \
+        do {                                    \
+                char _gfid_unparsed[40];        \
+                uuid_unparse (_gfid, _gfid_unparsed); \
+                _ret = snprintf (var, CONTRI_KEY_MAX, QUOTA_XATTR_PREFIX \
+                                 ".%s.%s." CONTRIBUTION, "quota", \
+                                 _gfid_unparsed); \
+        } while (0);
+
+#define QUOTA_SAFE_INCREMENT(lock, var)         \
+        do {                                    \
+                LOCK (lock);                    \
+                var ++;                         \
+                UNLOCK (lock);                  \
+        } while (0)
+
+struct quota_inode_ctx {
+        int64_t                size;
+        int8_t                 dirty;
+        gf_boolean_t           updation_status;
+        gf_lock_t              lock;
+        struct list_head       contribution_head;
+};
+typedef struct quota_inode_ctx quota_inode_ctx_t;
+
+struct inode_contribution {
+        struct list_head contri_list;
+        int64_t          contribution;
+        uuid_t           gfid;
+  gf_lock_t lock;
+};
+typedef struct inode_contribution inode_contribution_t;
+
+int32_t
+mq_get_lock_on_parent (call_frame_t *, xlator_t *);
+
+int32_t
+mq_req_xattr (xlator_t *, loc_t *, dict_t *);
+
+int32_t
+init_quota_priv (xlator_t *);
+
+int32_t
+mq_xattr_state (xlator_t *, loc_t *, dict_t *, struct iatt);
+
+int32_t
+mq_set_inode_xattr (xlator_t *, loc_t *);
+
+int
+mq_initiate_quota_txn (xlator_t *, loc_t *);
+
+int32_t
+mq_dirty_inode_readdir (call_frame_t *, void *, xlator_t *,
+                        int32_t, int32_t, fd_t *, dict_t *);
+
+int32_t
+mq_reduce_parent_size (xlator_t *, loc_t *, int64_t);
+
+int32_t
+mq_rename_update_newpath (xlator_t *, loc_t *);
+
+int32_t
+mq_inspect_file_xattr (xlator_t *this, loc_t *loc, dict_t *dict, struct iatt buf);
+
+int32_t
+mq_forget (xlator_t *, quota_inode_ctx_t *);
+#endif
diff --git a/xlators/features/marker/src/marker.c b/xlators/features/marker/src/marker.c
new file mode 100644
index 000000000..6a2c85691
--- /dev/null
+++ b/xlators/features/marker/src/marker.c
@@ -0,0 +1,2862 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "libxlator.h"
+#include "marker.h"
+#include "marker-mem-types.h"
+#include "marker-quota.h"
+#include "marker-quota-helper.h"
+#include "marker-common.h"
+#include "byte-order.h"
+
+#define _GF_UID_GID_CHANGED 1
+
+void
+fini (xlator_t *this);
+
+int32_t
+marker_start_setxattr (call_frame_t *, xlator_t *);
+
+marker_local_t *
+marker_local_ref (marker_local_t *local)
+{
+        GF_VALIDATE_OR_GOTO ("marker", local, err);
+
+        LOCK (&local->lock);
+        {
+                local->ref++;
+        }
+        UNLOCK (&local->lock);
+
+        return local;
+err:
+        return NULL;
+}
+
+int
+marker_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path)
+{
+        int     ret = -1;
+
+        if (!loc)
+                return ret;
+
+        if (inode) {
+                loc->inode = inode_ref (inode);
+                if (uuid_is_null (loc->gfid)) {
+                        uuid_copy (loc->gfid, loc->inode->gfid);
+                }
+        }
+
+        if (parent)
+                loc->parent = inode_ref (parent);
+
+        if (path) {
+                loc->path = gf_strdup (path);
+                if (!loc->path) {
+                        gf_log ("loc fill", GF_LOG_ERROR, "strdup failed");
+                        goto loc_wipe;
+                }
+
+                loc->name = strrchr (loc->path, '/');
+                if (loc->name)
+                        loc->name++;
+        }
+
+        ret = 0;
+loc_wipe:
+        if (ret < 0)
+                loc_wipe (loc);
+
+        return ret;
+}
+
+int
+marker_inode_loc_fill (inode_t *inode, loc_t *loc)
+{
+        char            *resolvedpath = NULL;
+        int              ret          = -1;
+	inode_t         *parent = NULL;
+
+        if ((!inode) || (!loc))
+                return ret;
+
+	parent = inode_parent (inode, NULL, NULL);
+
+        ret = inode_path (inode, NULL, &resolvedpath);
+        if (ret < 0)
+                goto err;
+
+        ret = marker_loc_fill (loc, inode, parent, resolvedpath);
+        if (ret < 0)
+                goto err;
+
+err:
+	if (parent)
+		inode_unref (parent);
+
+        GF_FREE (resolvedpath);
+
+        return ret;
+}
+
+int32_t
+marker_trav_parent (marker_local_t *local)
+{
+        int32_t ret = 0;
+        loc_t   loc = {0, };
+        inode_t *parent = NULL;
+        int8_t  need_unref = 0;
+
+        if (!local->loc.parent) {
+                parent = inode_parent (local->loc.inode, NULL, NULL);
+                if (parent)
+                        need_unref = 1;
+        } else
+                parent = local->loc.parent;
+
+        ret = marker_inode_loc_fill (parent, &loc);
+
+        if (ret < 0) {
+                ret = -1;
+                goto out;
+        }
+
+        loc_wipe (&local->loc);
+
+        local->loc = loc;
+out:
+        if (need_unref)
+                inode_unref (parent);
+
+        return ret;
+}
+
+int32_t
+marker_error_handler (xlator_t *this, marker_local_t *local, int32_t op_errno)
+{
+         marker_conf_t *priv = NULL;
+         const char    *path = NULL;
+
+         priv = (marker_conf_t *) this->private;
+         path = local
+                 ? (local->loc.path
+                    ? local->loc.path : uuid_utoa(local->loc.gfid))
+                 : "<nul>";
+
+         gf_log (this->name, GF_LOG_CRITICAL,
+                 "Indexing gone corrupt at %s (reason: %s)."
+                 " Geo-replication slave content needs to be revalidated",
+                 path, strerror (op_errno));
+        unlink (priv->timestamp_file);
+
+        return 0;
+}
+
+int32_t
+marker_local_unref (marker_local_t *local)
+{
+        int32_t   var = 0;
+
+        if (local == NULL)
+                return -1;
+
+        LOCK (&local->lock);
+        {
+                var = --local->ref;
+        }
+        UNLOCK (&local->lock);
+
+        if (var != 0)
+                goto out;
+
+        loc_wipe (&local->loc);
+        loc_wipe (&local->parent_loc);
+        if (local->xdata)
+                dict_unref (local->xdata);
+
+        if (local->oplocal) {
+                marker_local_unref (local->oplocal);
+                local->oplocal = NULL;
+        }
+        mem_put (local);
+out:
+        return 0;
+}
+
+int32_t
+stat_stampfile (xlator_t *this, marker_conf_t *priv,
+                struct volume_mark **status)
+{
+        struct stat          buf      = {0, };
+        struct volume_mark  *vol_mark = NULL;
+
+        vol_mark = GF_CALLOC (sizeof (struct volume_mark), 1,
+                              gf_marker_mt_volume_mark);
+
+        vol_mark->major = 1;
+        vol_mark->minor = 0;
+
+        GF_ASSERT (sizeof (priv->volume_uuid_bin) == 16);
+        memcpy (vol_mark->uuid, priv->volume_uuid_bin, 16);
+
+        if (stat (priv->timestamp_file, &buf) != -1) {
+                vol_mark->retval = 0;
+                vol_mark->sec = htonl (buf.st_ctime);
+                vol_mark->usec = htonl (ST_CTIM_NSEC (&buf)/1000);
+        } else
+                vol_mark->retval = 1;
+
+        *status = vol_mark;
+
+        return 0;
+}
+
+int32_t
+marker_getxattr_stampfile_cbk (call_frame_t *frame, xlator_t *this,
+                               const char *name, struct volume_mark *vol_mark,
+                               dict_t *xdata)
+{
+        int32_t   ret  = -1;
+        dict_t   *dict = NULL;
+
+        if (vol_mark == NULL){
+                STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM, NULL, NULL);
+
+                goto out;
+        }
+
+        dict = dict_new ();
+
+        ret = dict_set_bin (dict, (char *)name, vol_mark,
+                            sizeof (struct volume_mark));
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING, "failed to set key %s",
+                        name);
+
+        STACK_UNWIND_STRICT (getxattr, frame, 0, 0, dict, xdata);
+
+        dict_unref (dict);
+out:
+        return 0;
+}
+
+int32_t
+call_from_special_client (call_frame_t *frame, xlator_t *this, const char *name)
+{
+        struct volume_mark     *vol_mark   = NULL;
+        marker_conf_t          *priv       = NULL;
+        gf_boolean_t            ret        = _gf_true;
+
+        priv = (marker_conf_t *)this->private;
+
+        if (frame->root->pid != GF_CLIENT_PID_GSYNCD || name == NULL ||
+            strcmp (name, MARKER_XATTR_PREFIX "." VOLUME_MARK) != 0) {
+                ret = _gf_false;
+                goto out;
+        }
+
+        stat_stampfile (this, priv, &vol_mark);
+
+        marker_getxattr_stampfile_cbk (frame, this, name, vol_mark, NULL);
+out:
+        return ret;
+}
+
+int32_t
+marker_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *dict,
+                     dict_t *xdata)
+{
+        if (cookie) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Filtering the quota extended attributes");
+
+                dict_foreach_fnmatch (dict, "trusted.glusterfs.quota*",
+                                      marker_filter_quota_xattr, NULL);
+        }
+
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+int32_t
+marker_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                 const char *name, dict_t *xdata)
+{
+        gf_boolean_t   ret    = _gf_false;
+        marker_conf_t *priv   = NULL;
+        unsigned long  cookie = 0;
+
+        priv = this->private;
+
+        if (priv == NULL || (priv->feature_enabled & GF_XTIME) == 0)
+                goto wind;
+
+        gf_log (this->name, GF_LOG_DEBUG, "USER:PID = %d", frame->root->pid);
+
+        ret = call_from_special_client (frame, this, name);
+wind:
+        if (ret == _gf_false) {
+                if (name == NULL) {
+                        /* Signifies that marker translator
+                         * has to filter the quota's xattr's,
+                         * this is to prevent afr from performing
+                         * self healing on marker-quota xattrs'
+                         */
+                        cookie = 1;
+                }
+                STACK_WIND_COOKIE (frame, marker_getxattr_cbk, (void *)cookie,
+                                   FIRST_CHILD(this),
+                                   FIRST_CHILD(this)->fops->getxattr, loc,
+                                   name, xdata);
+        }
+
+        return 0;
+}
+
+
+int32_t
+marker_setxattr_done (call_frame_t *frame)
+{
+        marker_local_t *local = NULL;
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_DESTROY (frame->root);
+
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int
+marker_specific_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                              int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t         ret   = 0;
+        int32_t         done  = 0;
+        marker_local_t *local = NULL;
+
+        local = (marker_local_t*) frame->local;
+
+        if (op_ret == -1 && op_errno == ENOSPC) {
+                marker_error_handler (this, local, op_errno);
+                done = 1;
+                goto out;
+        }
+
+        if (local) {
+                if (local->loc.path && strcmp (local->loc.path, "/") == 0) {
+                        done = 1;
+                        goto out;
+                }
+                if (__is_root_gfid (local->loc.gfid)) {
+                        done = 1;
+                        goto out;
+                }
+        }
+
+        ret = marker_trav_parent (local);
+
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "Error occurred "
+                        "while traversing to the parent, stopping marker");
+
+                done = 1;
+
+                goto out;
+        }
+
+        marker_start_setxattr (frame, this);
+
+out:
+        if (done) {
+                marker_setxattr_done (frame);
+        }
+
+        return 0;
+}
+
+int32_t
+marker_start_setxattr (call_frame_t *frame, xlator_t *this)
+{
+        int32_t          ret   = -1;
+        dict_t          *dict  = NULL;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        local = (marker_local_t*) frame->local;
+
+        if (!local)
+                goto out;
+
+        dict = dict_new ();
+
+        if (!dict)
+                goto out;
+
+        if (local->loc.inode && uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        ret = dict_set_static_bin (dict, priv->marker_xattr,
+                                   (void *)local->timebuf, 8);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set marker xattr (%s)", local->loc.path);
+                goto out;
+        }
+
+        STACK_WIND (frame, marker_specific_setxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr, &local->loc, dict, 0,
+                    NULL);
+
+        ret = 0;
+out:
+        if (dict)
+                dict_unref (dict);
+
+        return ret;
+}
+
+void
+marker_gettimeofday (marker_local_t *local)
+{
+        struct timeval tv = {0, };
+
+        gettimeofday (&tv, NULL);
+
+        local->timebuf [0] = htonl (tv.tv_sec);
+        local->timebuf [1] = htonl (tv.tv_usec);
+
+        return;
+}
+
+int32_t
+marker_create_frame (xlator_t *this, marker_local_t *local)
+{
+        call_frame_t *frame = NULL;
+
+        frame = create_frame (this, this->ctx->pool);
+
+        frame->local = (void *) local;
+
+        marker_start_setxattr (frame, this);
+
+        return 0;
+}
+
+int32_t
+marker_xtime_update_marks (xlator_t *this, marker_local_t *local)
+{
+        marker_conf_t *priv = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, local, out);
+
+        priv = this->private;
+
+        if ((local->pid == GF_CLIENT_PID_GSYNCD
+             && !(priv->feature_enabled & GF_XTIME_GSYNC_FORCE))
+            || (local->pid == GF_CLIENT_PID_DEFRAG))
+                goto out;
+
+        marker_gettimeofday (local);
+
+        marker_local_ref (local);
+
+        marker_create_frame (this, local);
+out:
+        return 0;
+}
+
+
+int32_t
+marker_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, inode_t *inode,
+                  struct iatt *buf, struct iatt *preparent,
+                  struct iatt *postparent, dict_t *xdata)
+{
+        marker_conf_t      *priv    = NULL;
+        marker_local_t     *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "error occurred "
+                        "while Creating a file %s", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode,
+                             buf, preparent, postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_set_inode_xattr (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int
+marker_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+              mode_t umask, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_mkdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (mkdir, frame, -1, ENOMEM, NULL,
+                             NULL, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+                   struct iatt *buf, struct iatt *preparent,
+                   struct iatt *postparent, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "error occurred "
+                        "while Creating a file %s", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
+                             preparent, postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_set_inode_xattr (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+               mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_create_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->create, loc, flags, mode, umask,
+                    fd, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (create, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+                             NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                   struct iatt *postbuf, dict_t *xdata)
+{
+        marker_conf_t      *priv    = NULL;
+        marker_local_t     *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "error occurred "
+                        "while write, %s", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf,
+                             xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_writev (call_frame_t *frame,
+               xlator_t *this,
+               fd_t *fd,
+               struct iovec *vector,
+               int32_t count,
+               off_t offset, uint32_t flags,
+               struct iobref *iobref, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_writev_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
+                    flags, iobref, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (writev, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                  struct iatt *postparent, dict_t *xdata)
+{
+        marker_conf_t      *priv    = NULL;
+        marker_local_t     *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "error occurred "
+                        "rmdir %s", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno, preparent,
+                             postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_reduce_parent_size (this, &local->loc, -1);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+              dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_rmdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (rmdir, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                   struct iatt *postparent, dict_t *xdata)
+{
+        marker_conf_t      *priv    = NULL;
+        marker_local_t     *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "%s occurred in unlink", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno, preparent,
+                             postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if ((priv->feature_enabled & GF_QUOTA) && (local->ia_nlink == 1))
+                mq_reduce_parent_size (this, &local->loc, -1);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+
+int32_t
+marker_unlink_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                        dict_t *xdata)
+{
+        marker_local_t *local = NULL;
+
+        local = frame->local;
+        if (op_ret < 0) {
+                goto err;
+        }
+
+        if (local == NULL) {
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->ia_nlink = buf->ia_nlink;
+
+        STACK_WIND (frame, marker_unlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->unlink, &local->loc, local->xflag,
+                    local->xdata);
+        return 0;
+err:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL, NULL, NULL);
+        marker_local_unref (local);
+        return 0;
+}
+
+
+int32_t
+marker_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+               dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto unlink_wind;
+
+        local = mem_get0 (this->local_pool);
+        local->xflag = xflag;
+        if (xdata)
+                local->xdata = dict_ref (xdata);
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+
+        if (uuid_is_null (loc->gfid) && loc->inode)
+                uuid_copy (loc->gfid, loc->inode->gfid);
+
+        STACK_WIND (frame, marker_unlink_stat_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->stat, loc, xdata);
+        return 0;
+
+unlink_wind:
+        STACK_WIND (frame, marker_unlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+        return 0;
+err:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (unlink, frame, -1, ENOMEM, NULL, NULL, NULL);
+        marker_local_unref (local);
+        return 0;
+}
+
+
+int32_t
+marker_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, inode_t *inode,
+                 struct iatt *buf, struct iatt *preparent,
+                 struct iatt *postparent, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "linking a file ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+             dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, newloc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_link_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (link, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+                             NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_rename_done (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t *local  = NULL, *oplocal = NULL;
+        loc_t           newloc = {0, };
+        marker_conf_t  *priv   = NULL;
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        priv = this->private;
+
+        frame->local = NULL;
+
+        if (op_ret < 0) {
+                if (local->err == 0) {
+                        local->err = op_errno;
+                }
+
+                gf_log (this->name, GF_LOG_WARNING,
+                        "inodelk (UNLOCK) failed on path:%s (gfid:%s) (%s)",
+                        local->parent_loc.path,
+                        uuid_utoa (local->parent_loc.inode->gfid),
+                        strerror (op_errno));
+        }
+
+        if (local->stub != NULL) {
+                call_resume (local->stub);
+                local->stub = NULL;
+        } else if (local->err != 0) {
+                STACK_UNWIND_STRICT (rename, frame, -1, local->err, NULL, NULL,
+                                     NULL, NULL, NULL, NULL);
+        }
+
+        mq_reduce_parent_size (this, &oplocal->loc, oplocal->contribution);
+
+        if (local->loc.inode != NULL) {
+                mq_reduce_parent_size (this, &local->loc, local->contribution);
+        }
+
+        newloc.inode = inode_ref (oplocal->loc.inode);
+        newloc.path = gf_strdup (local->loc.path);
+        newloc.name = strrchr (newloc.path, '/');
+        if (newloc.name)
+                newloc.name++;
+        newloc.parent = inode_ref (local->loc.parent);
+
+        mq_rename_update_newpath (this, &newloc);
+
+        loc_wipe (&newloc);
+
+        if (priv->feature_enabled & GF_XTIME) {
+                //update marks on oldpath
+                uuid_copy (local->loc.gfid, oplocal->loc.inode->gfid);
+                marker_xtime_update_marks (this, oplocal);
+                marker_xtime_update_marks (this, local);
+        }
+
+        marker_local_unref (local);
+        marker_local_unref (oplocal);
+        return 0;
+}
+
+
+int32_t
+marker_rename_release_newp_lock (call_frame_t *frame, void *cookie,
+                                 xlator_t *this, int32_t op_ret,
+                                 int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t  *local = NULL, *oplocal = NULL;
+        struct gf_flock  lock  = {0, };
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        if (op_ret < 0) {
+                if (local->err == 0) {
+                        local->err = op_errno;
+                }
+
+                gf_log (this->name, GF_LOG_WARNING,
+                        "inodelk (UNLOCK) failed on %s (gfid:%s) (%s)",
+                        oplocal->parent_loc.path,
+                        uuid_utoa (oplocal->parent_loc.inode->gfid),
+                        strerror (op_errno));
+        }
+
+        if (local->next_lock_on == NULL) {
+                marker_rename_done (frame, NULL, this, 0, 0, NULL);
+                goto out;
+        }
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        STACK_WIND (frame,
+                    marker_rename_done,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->parent_loc, F_SETLKW, &lock, NULL);
+
+out:
+        return 0;
+}
+
+
+int32_t
+marker_rename_release_oldp_lock (call_frame_t *frame, void *cookie,
+                                 xlator_t *this, int32_t op_ret,
+                                 int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t  *local = NULL, *oplocal = NULL;
+        struct gf_flock  lock  = {0, };
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        if ((op_ret < 0) && (op_errno != ENOATTR)) {
+                local->err = op_errno;
+        }
+
+        //Reset frame uid and gid if set.
+        if (cookie == (void *) _GF_UID_GID_CHANGED)
+                MARKER_RESET_UID_GID (frame, frame->root, local);
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        STACK_WIND (frame,
+                    marker_rename_release_newp_lock,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &oplocal->parent_loc, F_SETLKW, &lock, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                   struct iatt *preoldparent, struct iatt *postoldparent,
+                   struct iatt *prenewparent, struct iatt *postnewparent,
+                   dict_t *xdata)
+{
+        marker_conf_t  *priv                 = NULL;
+        marker_local_t *local                = NULL;
+        marker_local_t *oplocal              = NULL;
+        call_stub_t    *stub                 = NULL;
+        int32_t         ret                  = 0;
+        char            contri_key [512]     = {0, };
+        loc_t           newloc               = {0, };
+
+        local = (marker_local_t *) frame->local;
+
+        if (local != NULL) {
+                oplocal = local->oplocal;
+        }
+
+        priv = this->private;
+
+        if (op_ret < 0) {
+                if (local != NULL) {
+                        local->err = op_errno;
+                }
+
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "renaming a file ", strerror (op_errno));
+        }
+
+        if (priv->feature_enabled & GF_QUOTA) {
+                if ((op_ret < 0) || (local == NULL)) {
+                        goto quota_err;
+                }
+
+                stub = fop_rename_cbk_stub (frame, default_rename_cbk, op_ret,
+                                            op_errno, buf, preoldparent,
+                                            postoldparent, prenewparent,
+                                            postnewparent, xdata);
+                if (stub == NULL) {
+                        local->err = ENOMEM;
+                        goto quota_err;
+                }
+
+                local->stub = stub;
+
+                GET_CONTRI_KEY (contri_key, oplocal->loc.parent->gfid, ret);
+                if (ret < 0) {
+                        local->err = ENOMEM;
+                        goto quota_err;
+                }
+
+                /* Removexattr requires uid and gid to be 0,
+                 * reset them in the callback.
+                 */
+                MARKER_SET_UID_GID (frame, local, frame->root);
+
+                newloc.inode = inode_ref (oplocal->loc.inode);
+                newloc.path = gf_strdup (local->loc.path);
+                newloc.name = strrchr (newloc.path, '/');
+                if (newloc.name)
+                        newloc.name++;
+                newloc.parent = inode_ref (local->loc.parent);
+                uuid_copy (newloc.gfid, oplocal->loc.inode->gfid);
+
+                STACK_WIND_COOKIE (frame, marker_rename_release_oldp_lock,
+                                   frame->cookie, FIRST_CHILD(this),
+                                   FIRST_CHILD(this)->fops->removexattr,
+                                   &newloc, contri_key, NULL);
+
+                loc_wipe (&newloc);
+        } else {
+                frame->local = NULL;
+
+                STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf,
+                                     preoldparent, postoldparent, prenewparent,
+                                     postnewparent, xdata);
+
+                if ((op_ret < 0) || (local == NULL)) {
+                        goto out;
+                }
+
+                if (priv->feature_enabled & GF_XTIME) {
+                        //update marks on oldpath
+                        uuid_copy (local->loc.gfid, oplocal->loc.inode->gfid);
+                        marker_xtime_update_marks (this, oplocal);
+                        marker_xtime_update_marks (this, local);
+                }
+        }
+
+out:
+        if (!(priv->feature_enabled & GF_QUOTA)) {
+                marker_local_unref (local);
+                marker_local_unref (oplocal);
+        }
+
+        return 0;
+
+quota_err:
+        marker_rename_release_oldp_lock (frame, NULL, this, 0, 0, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_do_rename (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        marker_local_t       *local           = NULL, *oplocal = NULL;
+        char                  contri_key[512] = {0, };
+        int32_t               ret             = 0;
+        int64_t              *contribution    = 0;
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        //Reset frame uid and gid if set.
+        if (cookie == (void *) _GF_UID_GID_CHANGED)
+                MARKER_RESET_UID_GID (frame, frame->root, local);
+
+        if ((op_ret < 0) && (op_errno != ENOATTR)) {
+                local->err = op_errno;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "fetching contribution values from %s (gfid:%s) "
+                        "failed (%s)", local->loc.path,
+                        uuid_utoa (local->loc.inode->gfid),
+                        strerror (op_errno));
+                goto err;
+        }
+
+        if (local->loc.inode != NULL) {
+                GET_CONTRI_KEY (contri_key, local->loc.parent->gfid, ret);
+                if (ret < 0) {
+                        local->err = errno;
+                        goto err;
+                }
+
+                if (dict_get_bin (dict, contri_key,
+                                  (void **) &contribution) == 0) {
+                        local->contribution = ntoh64 (*contribution);
+                }
+        }
+
+        STACK_WIND (frame, marker_rename_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rename, &oplocal->loc,
+                    &local->loc, NULL);
+
+        return 0;
+
+err:
+        marker_rename_release_oldp_lock (frame, NULL, this, 0, 0, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_get_newpath_contribution (call_frame_t *frame, void *cookie,
+                                 xlator_t *this, int32_t op_ret,
+                                 int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        marker_local_t *local           = NULL, *oplocal = NULL;
+        char            contri_key[512] = {0, };
+        int32_t         ret             = 0;
+        int64_t        *contribution    = 0;
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        //Reset frame uid and gid if set.
+        if (cookie == (void *) _GF_UID_GID_CHANGED)
+                MARKER_RESET_UID_GID (frame, frame->root, local);
+
+        if ((op_ret < 0) && (op_errno != ENOATTR)) {
+                local->err = op_errno;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "fetching contribution values from %s (gfid:%s) "
+                        "failed (%s)", oplocal->loc.path,
+                        uuid_utoa (oplocal->loc.inode->gfid),
+                        strerror (op_errno));
+                goto err;
+        }
+
+        GET_CONTRI_KEY (contri_key, oplocal->loc.parent->gfid, ret);
+        if (ret < 0) {
+                local->err = errno;
+                goto err;
+        }
+
+        if (dict_get_bin (dict, contri_key, (void **) &contribution) == 0)
+                oplocal->contribution = ntoh64 (*contribution);
+
+        if (local->loc.inode != NULL) {
+                GET_CONTRI_KEY (contri_key, local->loc.parent->gfid, ret);
+                if (ret < 0) {
+                        local->err = errno;
+                        goto err;
+                }
+
+                /* getxattr requires uid and gid to be 0,
+                 * reset them in the callback.
+                 */
+                MARKER_SET_UID_GID (frame, local, frame->root);
+                if (uuid_is_null (local->loc.gfid))
+                        uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+                GF_UUID_ASSERT (local->loc.gfid);
+
+                STACK_WIND_COOKIE (frame, marker_do_rename,
+                                   frame->cookie, FIRST_CHILD(this),
+                                   FIRST_CHILD(this)->fops->getxattr,
+                                   &local->loc, contri_key, NULL);
+        } else {
+                marker_do_rename (frame, NULL, this, 0, 0, NULL, NULL);
+        }
+
+        return 0;
+err:
+        marker_rename_release_oldp_lock (frame, NULL, this, 0, 0, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_get_oldpath_contribution (call_frame_t *frame, void *cookie,
+                                 xlator_t *this, int32_t op_ret,
+                                 int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t *local           = NULL, *oplocal = NULL;
+        char            contri_key[512] = {0, };
+        int32_t         ret             = 0;
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        if (op_ret < 0) {
+                local->err = op_errno;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "cannot hold inodelk on %s (gfid:%s) (%s)",
+                        local->next_lock_on->path,
+                        uuid_utoa (local->next_lock_on->inode->gfid),
+                        strerror (op_errno));
+                goto lock_err;
+        }
+
+        GET_CONTRI_KEY (contri_key, oplocal->loc.parent->gfid, ret);
+        if (ret < 0) {
+                local->err = errno;
+                goto quota_err;
+        }
+
+        /* getxattr requires uid and gid to be 0,
+         * reset them in the callback.
+         */
+        MARKER_SET_UID_GID (frame, local, frame->root);
+
+        if (uuid_is_null (oplocal->loc.gfid))
+                        uuid_copy (oplocal->loc.gfid,
+                                   oplocal->loc.inode->gfid);
+
+        GF_UUID_ASSERT (oplocal->loc.gfid);
+
+        STACK_WIND_COOKIE (frame, marker_get_newpath_contribution,
+                           frame->cookie, FIRST_CHILD(this),
+                           FIRST_CHILD(this)->fops->getxattr,
+                           &oplocal->loc, contri_key, NULL);
+        return 0;
+
+quota_err:
+        marker_rename_release_oldp_lock (frame, NULL, this, 0, 0, NULL);
+        return 0;
+
+lock_err:
+        if ((local->next_lock_on == NULL)
+            || (local->next_lock_on == &local->parent_loc)) {
+                local->next_lock_on = NULL;
+                marker_rename_release_oldp_lock (frame, NULL, this, 0, 0, NULL);
+        } else {
+                marker_rename_release_newp_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        return 0;
+}
+
+
+int32_t
+marker_rename_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t  *local = NULL, *oplocal = NULL;
+        loc_t           *loc   = NULL;
+        struct gf_flock  lock  = {0, };
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        if (op_ret < 0) {
+                if (local->next_lock_on != &oplocal->parent_loc) {
+                        loc = &oplocal->parent_loc;
+                } else {
+                        loc = &local->parent_loc;
+                }
+
+                local->err = op_errno;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "cannot hold inodelk on %s (gfid:%s) (%s)",
+                        loc->path, uuid_utoa (loc->inode->gfid),
+                        strerror (op_errno));
+                goto err;
+        }
+
+        if (local->next_lock_on != NULL) {
+                lock.l_len    = 0;
+                lock.l_start  = 0;
+                lock.l_type   = F_WRLCK;
+                lock.l_whence = SEEK_SET;
+
+                STACK_WIND (frame,
+                            marker_get_oldpath_contribution,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->inodelk,
+                            this->name, local->next_lock_on,
+                            F_SETLKW, &lock, NULL);
+        } else {
+                marker_get_oldpath_contribution (frame, 0, this, 0, 0, NULL);
+        }
+
+        return 0;
+
+err:
+        marker_rename_done (frame, NULL, this, 0, 0, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+               loc_t *newloc, dict_t *xdata)
+{
+        int32_t         ret              = 0;
+        marker_local_t *local            = NULL;
+        marker_local_t *oplocal          = NULL;
+        marker_conf_t  *priv             = NULL;
+        struct gf_flock lock             = {0, };
+        loc_t          *lock_on          = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto rename_wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        oplocal = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, oplocal);
+
+        frame->local = local;
+
+        local->oplocal = marker_local_ref (oplocal);
+
+        ret = loc_copy (&local->loc, newloc);
+        if (ret < 0)
+                goto err;
+
+        ret = loc_copy (&oplocal->loc, oldloc);
+        if (ret < 0)
+                goto err;
+
+        if (!(priv->feature_enabled & GF_QUOTA)) {
+                goto rename_wind;
+        }
+
+        ret = mq_inode_loc_fill (NULL, newloc->parent, &local->parent_loc);
+        if (ret < 0)
+                goto err;
+
+        ret = mq_inode_loc_fill (NULL, oldloc->parent, &oplocal->parent_loc);
+        if (ret < 0)
+                goto err;
+
+        if ((newloc->inode != NULL) && (newloc->parent != oldloc->parent)
+            && (uuid_compare (newloc->parent->gfid,
+                              oldloc->parent->gfid) < 0)) {
+                lock_on = &local->parent_loc;
+                local->next_lock_on = &oplocal->parent_loc;
+        } else {
+                lock_on = &oplocal->parent_loc;
+                if ((newloc->inode != NULL) && (newloc->parent
+                                                != oldloc->parent)) {
+                        local->next_lock_on = &local->parent_loc;
+                }
+        }
+
+        lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+        STACK_WIND (frame,
+                    marker_rename_inodelk_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, lock_on,
+                    F_SETLKW, &lock, NULL);
+
+        return 0;
+
+rename_wind:
+        STACK_WIND (frame, marker_rename_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (rename, frame, -1, ENOMEM, NULL,
+                             NULL, NULL, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "truncating a file ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+                 dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_truncate_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                      struct iatt *postbuf, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "truncating a file ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+                  dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_ftruncate_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, inode_t *inode,
+                    struct iatt *buf, struct iatt *preparent,
+                    struct iatt *postparent, dict_t *xdata)
+{
+        marker_conf_t      *priv    = NULL;
+        marker_local_t     *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "creating symlinks ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_set_inode_xattr (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int
+marker_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
+                loc_t *loc, mode_t umask, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_symlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->symlink, linkpath, loc, umask,
+                    xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (symlink, frame, -1, ENOMEM, NULL,
+                             NULL, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, inode_t *inode,
+                  struct iatt *buf, struct iatt *preparent,
+                  struct iatt *postparent, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "creating symlinks ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode,
+                             buf, preparent, postparent, xdata);
+
+        if (op_ret == -1 ||  local == NULL)
+                goto out;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        priv = this->private;
+
+        if ((priv->feature_enabled & GF_QUOTA) && (S_ISREG (local->mode))) {
+                mq_set_inode_xattr (this, &local->loc);
+        }
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int
+marker_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+              dev_t rdev, mode_t umask, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        local->mode = mode;
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_mknod_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask,
+                    xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (mknod, frame, -1, ENOMEM, NULL,
+                             NULL, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_fallocate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                      struct iatt *postbuf, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "fallocating a file ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (fallocate, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+		 off_t offset, size_t len, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_fallocate_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
+		    xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (fallocate, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                   struct iatt *postbuf, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred during discard",
+                        strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (discard, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	       size_t len, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_discard_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (discard, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+int32_t
+marker_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                   struct iatt *postbuf, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred during zerofill",
+                        strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (zerofill, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+               size_t len, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_zerofill_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (zerofill, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+/* when a call from the special client is received on
+ * key trusted.glusterfs.volume-mark with value "RESET"
+ * or if the value is 0length, update the change the
+ * access time and modification time via touching the
+ * timestamp file.
+ */
+int32_t
+call_from_sp_client_to_reset_tmfile (call_frame_t *frame,
+                                     xlator_t     *this,
+                                     dict_t       *dict)
+{
+        int32_t          fd       = 0;
+        int32_t          op_ret   = 0;
+        int32_t          op_errno = 0;
+        data_t          *data     = NULL;
+        marker_conf_t   *priv     = NULL;
+
+        if (frame == NULL || this == NULL || dict == NULL)
+                return -1;
+
+        priv = this->private;
+
+        data = dict_get (dict, "trusted.glusterfs.volume-mark");
+        if (data == NULL)
+                return -1;
+
+        if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
+                op_ret = -1;
+                op_errno = EPERM;
+
+                goto out;
+        }
+
+        if (data->len == 0 || (data->len == 5 &&
+                               memcmp (data->data, "RESET", 5) == 0)) {
+                fd = open (priv->timestamp_file, O_WRONLY|O_TRUNC);
+                if (fd != -1) {
+                        /* TODO check  whether the O_TRUNC would update the
+                         * timestamps on a zero length file on all machies.
+                         */
+                        close (fd);
+                }
+
+                if (fd != -1 || errno == ENOENT) {
+                        op_ret = 0;
+                        op_errno = 0;
+                } else {
+                        op_ret = -1;
+                        op_errno = errno;
+                }
+        } else {
+                op_ret = -1;
+                op_errno = EINVAL;
+        }
+out:
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred in "
+                        "setxattr ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+                 int32_t flags, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        ret = call_from_sp_client_to_reset_tmfile (frame, this, dict);
+        if (ret == 0)
+                return 0;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_setxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (setxattr, frame, -1, ENOMEM, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "creating symlinks ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+                  int32_t flags, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        ret = call_from_sp_client_to_reset_tmfile (frame, this, dict);
+        if (ret == 0)
+                return 0;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_fsetxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (fsetxattr, frame, -1, ENOMEM, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                     struct iatt *statpost, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "%s occurred while "
+                        "creating symlinks ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (fsetattr, frame, op_ret, op_errno, statpre,
+                             statpost, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+
+int32_t
+marker_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_fsetattr_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->fsetattr, fd, stbuf, valid, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (fsetattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                    struct iatt *statpost, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, ((op_errno == ENOENT) ? GF_LOG_DEBUG :
+                                     GF_LOG_ERROR),
+                        "%s occurred during setattr of %s",
+                        strerror (op_errno),
+                        (local ? local->loc.path : "<nul>"));
+        }
+
+        STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, statpre,
+                             statpost, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_setattr_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->setattr, loc, stbuf, valid, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (setattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "%s occurred while "
+                        "creating symlinks ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                    const char *name, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_removexattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (removexattr, frame, -1, ENOMEM, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, inode_t *inode,
+                   struct iatt *buf, dict_t *dict, struct iatt *postparent)
+{
+        marker_conf_t  *priv    = NULL;
+        marker_local_t *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "lookup failed with %s",
+                        strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
+                             dict, postparent);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        /* copy the gfid from the stat structure instead of inode,
+         * since if the lookup is fresh lookup, then the inode
+         * would have not yet linked to the inode table which happens
+         * in protocol/server.
+         */
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA) {
+                mq_xattr_state (this, &local->loc, dict, *buf);
+        }
+
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_lookup (call_frame_t *frame, xlator_t *this,
+               loc_t *loc, dict_t *xattr_req)
+{
+        int32_t         ret     = 0;
+        marker_local_t *local   = NULL;
+        marker_conf_t  *priv    = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+        if (ret == -1)
+                goto err;
+
+        if ((priv->feature_enabled & GF_QUOTA) && xattr_req)
+                mq_req_xattr (this, loc, xattr_req);
+wind:
+        STACK_WIND (frame, marker_lookup_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (lookup, frame, -1, 0, NULL, NULL, NULL, NULL);
+
+        return 0;
+}
+
+int
+marker_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int op_ret, int op_errno, gf_dirent_t *entries,
+                     dict_t *xdata)
+{
+        gf_dirent_t *entry = NULL;
+
+        if (op_ret <= 0)
+                goto unwind;
+
+        list_for_each_entry (entry, &entries->list, list) {
+                /* TODO: fill things */
+        }
+
+unwind:
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
+
+        return 0;
+}
+
+int
+marker_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+                 off_t offset, dict_t *dict)
+{
+        marker_conf_t  *priv    = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        if ((priv->feature_enabled & GF_QUOTA) && dict)
+                mq_req_xattr (this, NULL, dict);
+
+wind:
+        STACK_WIND (frame, marker_readdirp_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdirp,
+                    fd, size, offset, dict);
+
+        return 0;
+}
+
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_marker_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
+                       "failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+
+int32_t
+init_xtime_priv (xlator_t *this, dict_t *options)
+{
+        data_t          *data    = NULL;
+        int32_t          ret     = -1;
+        marker_conf_t   *priv    = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, options, out);
+        GF_VALIDATE_OR_GOTO (this->name, this->private, out);
+
+        priv = this->private;
+
+        if((data = dict_get (options, VOLUME_UUID)) != NULL) {
+                priv->volume_uuid = data->data;
+
+                ret = uuid_parse (priv->volume_uuid, priv->volume_uuid_bin);
+                if (ret == -1) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "invalid volume uuid %s", priv->volume_uuid);
+                        goto out;
+                }
+
+                ret = gf_asprintf (& (priv->marker_xattr), "%s.%s.%s",
+                                   MARKER_XATTR_PREFIX, priv->volume_uuid,
+                                   XTIME);
+
+                if (ret == -1){
+                        priv->marker_xattr = NULL;
+
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to allocate memory");
+                        goto out;
+                }
+
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "the volume-uuid = %s", priv->volume_uuid);
+        } else {
+                priv->volume_uuid = NULL;
+
+                gf_log (this->name, GF_LOG_ERROR,
+                        "please specify the volume-uuid"
+                        "in the translator options");
+
+                return -1;
+        }
+
+        if ((data = dict_get (options, TIMESTAMP_FILE)) != NULL) {
+                priv->timestamp_file = data->data;
+
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "the timestamp-file is = %s",
+                        priv->timestamp_file);
+
+        } else {
+                priv->timestamp_file = NULL;
+
+                gf_log (this->name, GF_LOG_ERROR,
+                        "please specify the timestamp-file"
+                        "in the translator options");
+
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+void
+marker_xtime_priv_cleanup (xlator_t *this)
+{
+        marker_conf_t *priv = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+
+        priv = (marker_conf_t *) this->private;
+
+        GF_VALIDATE_OR_GOTO (this->name, priv, out);
+
+        GF_FREE (priv->volume_uuid);
+
+        GF_FREE (priv->timestamp_file);
+
+        GF_FREE (priv->marker_xattr);
+out:
+        return;
+}
+
+void
+marker_priv_cleanup (xlator_t *this)
+{
+        marker_conf_t *priv = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+
+        priv = (marker_conf_t *) this->private;
+
+        GF_VALIDATE_OR_GOTO (this->name, priv, out);
+
+        marker_xtime_priv_cleanup (this);
+
+        LOCK_DESTROY (&priv->lock);
+
+        GF_FREE (priv);
+out:
+        return;
+}
+
+int32_t
+reconfigure (xlator_t *this, dict_t *options)
+{
+        int32_t         ret     = 0;
+        data_t         *data    = NULL;
+        gf_boolean_t    flag    = _gf_false;
+        marker_conf_t  *priv    = NULL;
+
+        GF_ASSERT (this);
+        GF_ASSERT (this->private);
+
+        priv = this->private;
+
+        priv->feature_enabled = 0;
+
+        GF_VALIDATE_OR_GOTO (this->name, options, out);
+
+        data = dict_get (options, "quota");
+        if (data) {
+                ret = gf_string2boolean (data->data, &flag);
+                if (ret == 0 && flag == _gf_true) {
+                        ret = init_quota_priv (this);
+                        if (ret < 0) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "failed to initialize quota private");
+                        } else {
+                                priv->feature_enabled |= GF_QUOTA;
+                        }
+                }
+        }
+
+        data = dict_get (options, "xtime");
+        if (data) {
+                ret = gf_string2boolean (data->data, &flag);
+                if (ret == 0 && flag == _gf_true) {
+                        marker_xtime_priv_cleanup (this);
+
+                        ret = init_xtime_priv (this, options);
+                        if (ret < 0) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "failed to initialize xtime private, "
+                                        "xtime updation will fail");
+                        } else {
+                                priv->feature_enabled |= GF_XTIME;
+                                data = dict_get (options, "gsync-force-xtime");
+                                if (!data)
+                                        goto out;
+                                ret = gf_string2boolean (data->data, &flag);
+                                if (ret == 0 && flag)
+                                        priv->feature_enabled |= GF_XTIME_GSYNC_FORCE;
+                        }
+                }
+        }
+out:
+        return ret;
+}
+
+
+int32_t
+init (xlator_t *this)
+{
+        dict_t        *options = NULL;
+        data_t        *data    = NULL;
+        int32_t        ret     = 0;
+        gf_boolean_t   flag    = _gf_false;
+        marker_conf_t *priv    = NULL;
+
+        if (!this->children) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "marker translator needs subvolume defined.");
+                return -1;
+        }
+
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Volume is dangling.");
+                return -1;
+        }
+
+        options = this->options;
+
+        ALLOCATE_OR_GOTO (this->private, marker_conf_t, err);
+
+        priv = this->private;
+
+        priv->feature_enabled = 0;
+
+        LOCK_INIT (&priv->lock);
+
+        data = dict_get (options, "quota");
+        if (data) {
+                ret = gf_string2boolean (data->data, &flag);
+                if (ret == 0 && flag == _gf_true) {
+                        ret = init_quota_priv (this);
+                        if (ret < 0)
+                                goto err;
+
+                        priv->feature_enabled |= GF_QUOTA;
+                }
+        }
+
+        data = dict_get (options, "xtime");
+        if (data) {
+                ret = gf_string2boolean (data->data, &flag);
+                if (ret == 0 && flag == _gf_true) {
+                        ret = init_xtime_priv (this, options);
+                        if (ret < 0)
+                                goto err;
+
+                        priv->feature_enabled |= GF_XTIME;
+                        data = dict_get (options, "gsync-force-xtime");
+                        if (!data)
+                                goto cont;
+                        ret = gf_string2boolean (data->data, &flag);
+                        if (ret == 0 && flag)
+                                priv->feature_enabled |= GF_XTIME_GSYNC_FORCE;
+                }
+        }
+
+ cont:
+        this->local_pool = mem_pool_new (marker_local_t, 128);
+        if (!this->local_pool) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto err;
+        }
+
+        return 0;
+err:
+        marker_priv_cleanup (this);
+
+        return -1;
+}
+
+int32_t
+marker_forget (xlator_t *this, inode_t *inode)
+{
+        marker_inode_ctx_t *ctx   = NULL;
+        uint64_t            value = 0;
+
+        if (inode_ctx_del (inode, this, &value) != 0)
+                goto out;
+
+        ctx = (marker_inode_ctx_t *)(unsigned long)value;
+        if (ctx == NULL) {
+                goto out;
+        }
+
+        mq_forget (this, ctx->quota_ctx);
+
+        GF_FREE (ctx);
+out:
+        return 0;
+}
+
+void
+fini (xlator_t *this)
+{
+        marker_priv_cleanup (this);
+}
+
+struct xlator_fops fops = {
+        .lookup      = marker_lookup,
+        .create      = marker_create,
+        .mkdir       = marker_mkdir,
+        .writev      = marker_writev,
+        .truncate    = marker_truncate,
+        .ftruncate   = marker_ftruncate,
+        .symlink     = marker_symlink,
+        .link        = marker_link,
+        .unlink      = marker_unlink,
+        .rmdir       = marker_rmdir,
+        .rename      = marker_rename,
+        .mknod       = marker_mknod,
+        .setxattr    = marker_setxattr,
+        .fsetxattr   = marker_fsetxattr,
+        .setattr     = marker_setattr,
+        .fsetattr    = marker_fsetattr,
+        .removexattr = marker_removexattr,
+        .getxattr    = marker_getxattr,
+        .readdirp    = marker_readdirp,
+	.fallocate   = marker_fallocate,
+	.discard     = marker_discard,
+        .zerofill    = marker_zerofill,
+};
+
+struct xlator_cbks cbks = {
+        .forget = marker_forget
+};
+
+struct volume_options options[] = {
+        {.key = {"volume-uuid"}},
+        {.key = {"timestamp-file"}},
+        {.key = {"quota"}},
+        {.key = {"xtime"}},
+        {.key = {"gsync-force-xtime"}},
+        {.key = {NULL}}
+};
diff --git a/xlators/features/marker/src/marker.h b/xlators/features/marker/src/marker.h
new file mode 100644
index 000000000..1a58f8cfc
--- /dev/null
+++ b/xlators/features/marker/src/marker.h
@@ -0,0 +1,138 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _MARKER_H
+#define _MARKER_H
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "marker-quota.h"
+#include "xlator.h"
+#include "defaults.h"
+#include "uuid.h"
+#include "call-stub.h"
+
+#define MARKER_XATTR_PREFIX "trusted.glusterfs"
+#define XTIME               "xtime"
+#define VOLUME_MARK         "volume-mark"
+#define VOLUME_UUID         "volume-uuid"
+#define TIMESTAMP_FILE      "timestamp-file"
+
+enum {
+        GF_QUOTA             = 1,
+        GF_XTIME             = 2,
+        GF_XTIME_GSYNC_FORCE = 4,
+};
+
+/*initialize the local variable*/
+#define MARKER_INIT_LOCAL(_frame,_local) do {                   \
+                _frame->local = _local;                         \
+                _local->pid = _frame->root->pid;                \
+                memset (&_local->loc, 0, sizeof (loc_t));       \
+                _local->ref = 1;                                \
+                _local->uid = -1;                               \
+                _local->gid = -1;                               \
+                LOCK_INIT (&_local->lock);                      \
+                _local->oplocal = NULL;                         \
+        } while (0)
+
+/* try alloc and if it fails, goto label */
+#define ALLOCATE_OR_GOTO(var, type, label) do {                  \
+                var = GF_CALLOC (sizeof (type), 1,               \
+                                 gf_marker_mt_##type);           \
+                if (!var) {                                      \
+                        gf_log (this->name, GF_LOG_ERROR,        \
+                                "out of memory :(");             \
+                        goto label;                              \
+                }                                                \
+        } while (0)
+
+#define _MARKER_SET_UID_GID(dest, src)          \
+        do {                                    \
+                if (src->uid != -1 &&           \
+                    src->gid != -1) {           \
+                        dest->uid = src->uid;   \
+                        dest->gid = src->gid;   \
+                }                               \
+        } while (0)
+
+#define MARKER_SET_UID_GID(frame, dest, src)                    \
+        do {                                                    \
+                _MARKER_SET_UID_GID (dest, src);                \
+                frame->root->uid = 0;                           \
+                frame->root->gid = 0;                           \
+                frame->cookie = (void *) _GF_UID_GID_CHANGED;   \
+        } while (0)
+
+#define MARKER_RESET_UID_GID(frame, dest, src)                  \
+        do {                                                    \
+                _MARKER_SET_UID_GID (dest, src);                \
+                frame->cookie = NULL;                           \
+        } while (0)
+
+struct marker_local{
+        uint32_t        timebuf[2];
+        pid_t           pid;
+        loc_t           loc;
+        loc_t           parent_loc;
+        loc_t          *next_lock_on;
+        uid_t           uid;
+        gid_t           gid;
+        int32_t         ref;
+        int32_t         ia_nlink;
+        gf_lock_t       lock;
+        mode_t          mode;
+        int32_t         err;
+        call_stub_t    *stub;
+        int64_t         contribution;
+        struct marker_local *oplocal;
+
+        /* marker quota specific */
+        int64_t delta;
+        int64_t d_off;
+        int64_t sum;
+        int64_t size;
+        int32_t hl_count;
+        int32_t dentry_child_count;
+
+        fd_t         *fd;
+        call_frame_t *frame;
+
+        quota_inode_ctx_t    *ctx;
+        inode_contribution_t *contri;
+
+        int xflag;
+        dict_t *xdata;
+};
+typedef struct marker_local marker_local_t;
+
+#define quota_local_t marker_local_t
+
+struct marker_inode_ctx {
+        struct quota_inode_ctx *quota_ctx;
+};
+typedef struct marker_inode_ctx marker_inode_ctx_t;
+
+struct marker_conf{
+        char         feature_enabled;
+        char        *size_key;
+        char        *dirty_key;
+        char        *volume_uuid;
+        uuid_t      volume_uuid_bin;
+        char        *timestamp_file;
+        char        *marker_xattr;
+        uint64_t     quota_lk_owner;
+        gf_lock_t    lock;
+};
+typedef struct marker_conf marker_conf_t;
+
+#endif
diff --git a/xlators/features/path-convertor/src/Makefile.am b/xlators/features/path-convertor/src/Makefile.am
index 58cfed0f9..393a7bd08 100644
--- a/xlators/features/path-convertor/src/Makefile.am
+++ b/xlators/features/path-convertor/src/Makefile.am
@@ -2,13 +2,14 @@
 xlator_LTLIBRARIES = path-converter.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/testing/features
 
-path_converter_la_LDFLAGS = -module -avoidversion 
+path_converter_la_LDFLAGS = -module -avoid-version 
 
 path_converter_la_SOURCES = path.c
 path_converter_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la 
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/features/path-convertor/src/path-mem-types.h b/xlators/features/path-convertor/src/path-mem-types.h
index 99f794679..77ada8d53 100644
--- a/xlators/features/path-convertor/src/path-mem-types.h
+++ b/xlators/features/path-convertor/src/path-mem-types.h
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __PATH_MEM_TYPES_H__
 #define __PATH_MEM_TYPES_H__
 
diff --git a/xlators/features/path-convertor/src/path.c b/xlators/features/path-convertor/src/path.c
index 0a86baa87..5c52e0a8d 100644
--- a/xlators/features/path-convertor/src/path.c
+++ b/xlators/features/path-convertor/src/path.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 /* TODO: add gf_log to all the cases returning errors */
 
 #ifndef _CONFIG_H
@@ -52,7 +42,7 @@ static char *
 name_this_to_that (xlator_t *xl, const char *path, const char *name)
 {
 	path_private_t *priv = xl->private;
-	char priv_path[ZR_PATH_MAX] = {0,};
+	char priv_path[PATH_MAX] = {0,};
 	char *tmp_name = NULL;
 	int32_t path_len = strlen (path);
 	int32_t name_len = strlen (name) - ZR_FILE_CONTENT_STRLEN;
@@ -848,8 +838,7 @@ path_setxattr (call_frame_t *frame,
 	if (tmp_path != loc_path)
 		GF_FREE (tmp_path);
 
-	if (tmp_name)
-		GF_FREE (tmp_name);
+	GF_FREE (tmp_name);
 
 	return 0;
 }
@@ -1057,7 +1046,7 @@ path_entrylk (call_frame_t *frame, xlator_t *this,
 
 int32_t
 path_inodelk (call_frame_t *frame, xlator_t *this,
-              const char *volume, loc_t *loc, int32_t cmd, struct flock *lock)
+              const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *lock)
 {
 	char *loc_path = (char *)loc->path;
 	char *tmp_path = NULL;
diff --git a/xlators/protocol/legacy/client/Makefile.am b/xlators/features/protect/Makefile.am
index d471a3f92..d471a3f92 100644
--- a/xlators/protocol/legacy/client/Makefile.am
+++ b/xlators/features/protect/Makefile.am
diff --git a/xlators/features/protect/src/Makefile.am b/xlators/features/protect/src/Makefile.am
new file mode 100644
index 000000000..7eb93f32e
--- /dev/null
+++ b/xlators/features/protect/src/Makefile.am
@@ -0,0 +1,21 @@
+xlator_LTLIBRARIES = prot_dht.la prot_client.la prot_server.la
+
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+prot_dht_la_LDFLAGS = -module -avoidversion
+prot_dht_la_SOURCES = prot_dht.c
+prot_dht_la_LIBADD  = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+prot_client_la_LDFLAGS = -module -avoidversion
+prot_client_la_SOURCES = prot_client.c
+prot_client_la_LIBADD  = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+prot_server_la_LDFLAGS = -module -avoidversion
+prot_server_la_SOURCES = prot_server.c
+prot_server_la_LIBADD  = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES = 
+
diff --git a/xlators/features/protect/src/prot_client.c b/xlators/features/protect/src/prot_client.c
new file mode 100644
index 000000000..a27216d0a
--- /dev/null
+++ b/xlators/features/protect/src/prot_client.c
@@ -0,0 +1,215 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+
+#include <execinfo.h>
+
+#define NUM_FRAMES 20
+
+static char PROTECT_KEY[] = "trusted.glusterfs.protect";
+
+enum {
+        PROT_ACT_NONE = 0,
+        PROT_ACT_LOG,
+        PROT_ACT_REJECT,
+};
+
+void
+pcli_print_trace (char *name, call_frame_t *frame)
+{
+        void    *frames[NUM_FRAMES];
+        char    **symbols;
+        int     size;
+        int     i;
+
+        gf_log (name, GF_LOG_INFO, "Translator stack:");
+        while (frame) {
+                gf_log (name, GF_LOG_INFO, "%s (%s)",
+                        frame->wind_from, frame->this->name);
+                frame = frame->next;
+        }
+
+        size = backtrace(frames,NUM_FRAMES);
+        if (size <= 0) {
+                return;
+        }
+        symbols = backtrace_symbols(frames,size);
+        if (!symbols) {
+                return;
+        }
+
+        gf_log(name, GF_LOG_INFO, "Processor stack:");
+        for (i = 0; i < size; ++i) {
+                gf_log (name, GF_LOG_INFO, "%s", symbols[i]);
+        }
+        free(symbols);
+}
+
+int32_t
+pcli_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+                loc_t *newloc, dict_t *xdata)
+{
+        uint64_t        value;
+
+        if (newloc->parent == oldloc->parent) {
+                gf_log (this->name, GF_LOG_DEBUG, "rename in same directory");
+                goto simple_unwind;
+        }
+        if (!oldloc->parent) {
+                goto simple_unwind;
+        }
+        if (inode_ctx_get(oldloc->parent,this,&value) != 0) {
+                goto simple_unwind;
+        }
+
+        if (value != PROT_ACT_NONE) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "got rename for protected %s", oldloc->path);
+                pcli_print_trace(this->name,frame->next);
+                if (value == PROT_ACT_REJECT) {
+                        STACK_UNWIND_STRICT (rename, frame, -1, EPERM,
+                                             NULL, NULL, NULL, NULL, NULL,
+                                             xdata);
+                        return 0;
+                }
+        }
+
+simple_unwind:
+        STACK_WIND_TAIL (frame, FIRST_CHILD(this),
+                         FIRST_CHILD(this)->fops->rename, oldloc, newloc,
+                         xdata);
+        return 0;
+}
+
+int32_t
+pcli_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+               int32_t flags, dict_t *xdata)
+{
+        data_t          *data;
+        uint64_t        value;
+
+        /*
+         * We can't use dict_get_str and strcmp here, because the value comes
+         * directly from the user and might not be NUL-terminated (it would
+         * be if we had set it ourselves.
+         */
+
+        data = dict_get(dict,PROTECT_KEY);
+        if (!data) {
+                goto simple_wind;
+        }
+
+        if (dict->count > 1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "attempted to mix %s with other keys", PROTECT_KEY);
+                goto simple_wind;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "got %s request", PROTECT_KEY);
+        if (!strncmp(data->data,"log",data->len)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "logging removals on %s", loc->path);
+                value = PROT_ACT_LOG;
+        }
+        else if (!strncmp(data->data,"reject",data->len)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "rejecting removals on %s", loc->path);
+                value = PROT_ACT_REJECT;
+        }
+        else {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "removing protection on %s", loc->path);
+                value = PROT_ACT_NONE;
+        }
+        /* Right now the value doesn't matter - just the presence. */
+        if (inode_ctx_set(loc->inode,this,&value) != 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set protection status for %s", loc->path);
+        }
+        STACK_UNWIND_STRICT (setxattr, frame, 0, 0, NULL);
+        return 0;
+
+simple_wind:
+        STACK_WIND_TAIL (frame,
+                         FIRST_CHILD(this), FIRST_CHILD(this)->fops->setxattr,
+                         loc, dict, flags, xdata);
+        return 0;
+}
+
+int32_t
+pcli_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+             dict_t *xdata)
+{
+        uint64_t        value;
+
+        if (!loc->parent || (inode_ctx_get(loc->parent,this,&value) != 0)) {
+                goto simple_unwind;
+        }
+
+        if (value != PROT_ACT_NONE) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "got unlink for protected %s", loc->path);
+                pcli_print_trace(this->name,frame->next);
+                if (value == PROT_ACT_REJECT) {
+                        STACK_UNWIND_STRICT (unlink, frame, -1, EPERM,
+                                             NULL, NULL, NULL);
+                        return 0;
+                }
+        }
+
+simple_unwind:
+        STACK_WIND_TAIL (frame, FIRST_CHILD(this),
+                         FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+        return 0;
+}
+
+int32_t
+init (xlator_t *this)
+{
+	if (!this->children || this->children->next) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"translator not configured with exactly one child");
+		return -1;
+	}
+
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+
+	return 0;
+}
+
+
+void
+fini (xlator_t *this)
+{
+	return;
+}
+
+
+struct xlator_fops fops = {
+        .rename         = pcli_rename,
+        .setxattr       = pcli_setxattr,
+        .unlink         = pcli_unlink,
+};
+
+struct xlator_cbks cbks = {
+};
+
+struct volume_options options[] = {
+	{ .key = {NULL} },
+};
diff --git a/xlators/features/protect/src/prot_dht.c b/xlators/features/protect/src/prot_dht.c
new file mode 100644
index 000000000..feec6ffd6
--- /dev/null
+++ b/xlators/features/protect/src/prot_dht.c
@@ -0,0 +1,168 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+
+enum gf_pdht_mem_types_ {
+        gf_pdht_mt_coord_t = gf_common_mt_end + 1,
+        gf_pdht_mt_end
+};
+
+typedef struct {
+        pthread_mutex_t         lock;
+        uint16_t                refs;
+        int32_t                 op_ret;
+        int32_t                 op_errno;
+        dict_t                  *xdata;
+} pdht_coord_t;
+
+static char PROTECT_KEY[] = "trusted.glusterfs.protect";
+
+void
+pdht_unref_and_unlock (call_frame_t *frame, xlator_t *this,
+                       pdht_coord_t *coord)
+{
+        gf_boolean_t    should_unwind;
+
+        should_unwind = (--(coord->refs) == 0);
+        pthread_mutex_unlock(&coord->lock);
+
+        if (should_unwind) {
+                STACK_UNWIND_STRICT (setxattr, frame, 
+                                     coord->op_ret, coord->op_errno,
+                                     coord->xdata);
+                if (coord->xdata) {
+                        dict_unref(coord->xdata);
+                }
+                GF_FREE(coord);
+        }
+}
+
+int32_t
+pdht_recurse_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        pdht_coord_t    *coord = cookie;
+
+        pthread_mutex_lock(&coord->lock);
+        if (op_ret) {
+                coord->op_ret = op_ret;
+                coord->op_errno = op_errno;
+        }
+        if (xdata) {
+                if (coord->xdata) {
+                        dict_unref(coord->xdata);
+                }
+                coord->xdata = dict_ref(xdata);
+        }
+        pdht_unref_and_unlock(frame,this,coord);
+                
+        return 0;
+}
+
+void
+pdht_recurse (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+              int32_t flags, dict_t *xdata, xlator_t *xl, pdht_coord_t *coord)
+{
+        xlator_list_t   *iter;
+
+        if (!strcmp(xl->type,"features/prot_client")) {
+                pthread_mutex_lock(&coord->lock);
+                ++(coord->refs);
+                pthread_mutex_unlock(&coord->lock);
+                STACK_WIND_COOKIE (frame, pdht_recurse_cbk, coord, xl,
+                                   xl->fops->setxattr, loc, dict, flags, xdata);
+        }
+
+        else for (iter = xl->children; iter; iter = iter->next) {
+                pdht_recurse (frame, this, loc, dict, flags, xdata,
+                              iter->xlator, coord);
+        }
+}
+
+int32_t
+pdht_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+               int32_t flags, dict_t *xdata)
+{
+        pdht_coord_t    *coord;
+
+        if (!dict_get(dict,PROTECT_KEY)) {
+                goto simple_wind;
+        }
+
+        if (dict->count > 1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "attempted to mix %s with other keys", PROTECT_KEY);
+                goto simple_wind;
+        }
+
+        coord = GF_CALLOC(1,sizeof(*coord),gf_pdht_mt_coord_t);
+        if (!coord) {
+                gf_log (this->name, GF_LOG_WARNING, "allocation failed");
+                goto simple_wind;
+        }
+
+        pthread_mutex_init(&coord->lock,NULL);
+        coord->refs = 1;
+        coord->op_ret = 0;
+        coord->xdata = NULL;
+
+        pdht_recurse(frame,this,loc,dict,flags,xdata,this,coord);
+        pthread_mutex_lock(&coord->lock);
+        pdht_unref_and_unlock(frame,this,coord);
+
+        return 0;
+
+simple_wind:
+        STACK_WIND_TAIL (frame,
+                         FIRST_CHILD(this), FIRST_CHILD(this)->fops->setxattr,
+                         loc, dict, flags, xdata);
+        return 0;
+}
+
+int32_t
+init (xlator_t *this)
+{
+	if (!this->children || this->children->next) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"translator not configured with exactly one child");
+		return -1;
+	}
+
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+
+	return 0;
+}
+
+
+void
+fini (xlator_t *this)
+{
+	return;
+}
+
+struct xlator_fops fops = {
+        .setxattr = pdht_setxattr,
+};
+
+struct xlator_cbks cbks = {
+};
+
+struct volume_options options[] = {
+	{ .key = {NULL} },
+};
diff --git a/xlators/features/protect/src/prot_server.c b/xlators/features/protect/src/prot_server.c
new file mode 100644
index 000000000..beaee0889
--- /dev/null
+++ b/xlators/features/protect/src/prot_server.c
@@ -0,0 +1,51 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+
+int32_t
+init (xlator_t *this)
+{
+	if (!this->children || this->children->next) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"translator not configured with exactly one child");
+		return -1;
+	}
+
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+
+	return 0;
+}
+
+
+void
+fini (xlator_t *this)
+{
+	return;
+}
+
+
+struct xlator_fops fops = {
+};
+
+struct xlator_cbks cbks = {
+};
+
+struct volume_options options[] = {
+	{ .key = {NULL} },
+};
diff --git a/xlators/features/qemu-block/Makefile.am b/xlators/features/qemu-block/Makefile.am
new file mode 100644
index 000000000..af437a64d
--- /dev/null
+++ b/xlators/features/qemu-block/Makefile.am
@@ -0,0 +1 @@
+SUBDIRS = src
diff --git a/xlators/features/qemu-block/src/Makefile.am b/xlators/features/qemu-block/src/Makefile.am
new file mode 100644
index 000000000..08a7b62a0
--- /dev/null
+++ b/xlators/features/qemu-block/src/Makefile.am
@@ -0,0 +1,155 @@
+if ENABLE_QEMU_BLOCK
+xlator_LTLIBRARIES = qemu-block.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+qemu_block_la_LDFLAGS = -module -avoid-version
+qemu_block_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la $(GLIB_LIBS) -lz -lrt
+
+qemu_block_la_SOURCES_qemu = \
+	$(CONTRIBDIR)/qemu/qemu-coroutine.c \
+	$(CONTRIBDIR)/qemu/qemu-coroutine-lock.c \
+	$(CONTRIBDIR)/qemu/qemu-coroutine-sleep.c \
+	$(CONTRIBDIR)/qemu/coroutine-ucontext.c \
+	$(CONTRIBDIR)/qemu/block.c \
+	$(CONTRIBDIR)/qemu/nop-symbols.c
+
+qemu_block_la_SOURCES_qemu_util = \
+	$(CONTRIBDIR)/qemu/util/aes.c \
+	$(CONTRIBDIR)/qemu/util/bitmap.c \
+	$(CONTRIBDIR)/qemu/util/bitops.c \
+	$(CONTRIBDIR)/qemu/util/cutils.c \
+	$(CONTRIBDIR)/qemu/util/error.c \
+	$(CONTRIBDIR)/qemu/util/hbitmap.c \
+	$(CONTRIBDIR)/qemu/util/iov.c \
+	$(CONTRIBDIR)/qemu/util/module.c \
+	$(CONTRIBDIR)/qemu/util/oslib-posix.c \
+	$(CONTRIBDIR)/qemu/util/qemu-option.c \
+	$(CONTRIBDIR)/qemu/util/qemu-error.c \
+	$(CONTRIBDIR)/qemu/util/qemu-thread-posix.c \
+	$(CONTRIBDIR)/qemu/util/unicode.c \
+	$(CONTRIBDIR)/qemu/util/hexdump.c
+
+qemu_block_la_SOURCES_qemu_block = \
+	$(CONTRIBDIR)/qemu/block/snapshot.c \
+	$(CONTRIBDIR)/qemu/block/qcow2-cache.c \
+	$(CONTRIBDIR)/qemu/block/qcow2-cluster.c \
+	$(CONTRIBDIR)/qemu/block/qcow2-refcount.c \
+	$(CONTRIBDIR)/qemu/block/qcow2-snapshot.c \
+	$(CONTRIBDIR)/qemu/block/qcow2.c \
+	$(CONTRIBDIR)/qemu/block/qed-check.c \
+	$(CONTRIBDIR)/qemu/block/qed-cluster.c \
+	$(CONTRIBDIR)/qemu/block/qed-gencb.c \
+	$(CONTRIBDIR)/qemu/block/qed-l2-cache.c \
+	$(CONTRIBDIR)/qemu/block/qed-table.c \
+	$(CONTRIBDIR)/qemu/block/qed.c
+
+qemu_block_la_SOURCES_qemu_qobject = \
+	$(CONTRIBDIR)/qemu/qobject/json-lexer.c \
+	$(CONTRIBDIR)/qemu/qobject/json-parser.c \
+	$(CONTRIBDIR)/qemu/qobject/json-streamer.c \
+	$(CONTRIBDIR)/qemu/qobject/qbool.c \
+	$(CONTRIBDIR)/qemu/qobject/qdict.c \
+	$(CONTRIBDIR)/qemu/qobject/qerror.c \
+	$(CONTRIBDIR)/qemu/qobject/qfloat.c \
+	$(CONTRIBDIR)/qemu/qobject/qint.c \
+	$(CONTRIBDIR)/qemu/qobject/qjson.c \
+	$(CONTRIBDIR)/qemu/qobject/qlist.c \
+	$(CONTRIBDIR)/qemu/qobject/qstring.c
+
+qemu_block_la_SOURCES = \
+	$(qemu_block_la_SOURCES_qemu) \
+	$(qemu_block_la_SOURCES_qemu_util) \
+	$(qemu_block_la_SOURCES_qemu_block) \
+	$(qemu_block_la_SOURCES_qemu_qobject) \
+	bdrv-xlator.c \
+	coroutine-synctask.c \
+	bh-syncop.c \
+	monitor-logging.c \
+	clock-timer.c \
+	qemu-block.c \
+	qb-coroutines.c
+
+noinst_HEADERS_qemu = \
+	$(CONTRIBDIR)/qemu/config-host.h \
+	$(CONTRIBDIR)/qemu/qapi-types.h \
+	$(CONTRIBDIR)/qemu/qmp-commands.h \
+	$(CONTRIBDIR)/qemu/trace/generated-tracers.h \
+	$(CONTRIBDIR)/qemu/include/config.h \
+	$(CONTRIBDIR)/qemu/include/glib-compat.h \
+	$(CONTRIBDIR)/qemu/include/qemu-common.h \
+	$(CONTRIBDIR)/qemu/include/trace.h \
+	$(CONTRIBDIR)/qemu/include/block/coroutine.h \
+	$(CONTRIBDIR)/qemu/include/block/aio.h \
+	$(CONTRIBDIR)/qemu/include/block/block.h \
+	$(CONTRIBDIR)/qemu/include/block/block_int.h \
+	$(CONTRIBDIR)/qemu/include/block/blockjob.h \
+	$(CONTRIBDIR)/qemu/include/block/coroutine.h \
+	$(CONTRIBDIR)/qemu/include/block/coroutine_int.h \
+	$(CONTRIBDIR)/qemu/include/block/snapshot.h \
+	$(CONTRIBDIR)/qemu/include/exec/cpu-common.h \
+	$(CONTRIBDIR)/qemu/include/exec/hwaddr.h \
+	$(CONTRIBDIR)/qemu/include/exec/poison.h \
+	$(CONTRIBDIR)/qemu/include/fpu/softfloat.h \
+	$(CONTRIBDIR)/qemu/include/migration/migration.h \
+	$(CONTRIBDIR)/qemu/include/migration/qemu-file.h \
+	$(CONTRIBDIR)/qemu/include/migration/vmstate.h \
+	$(CONTRIBDIR)/qemu/include/monitor/monitor.h \
+	$(CONTRIBDIR)/qemu/include/monitor/readline.h \
+	$(CONTRIBDIR)/qemu/include/qapi/error.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/json-lexer.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/json-parser.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/json-streamer.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qbool.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qdict.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qerror.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qfloat.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qint.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qjson.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qlist.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qobject.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qstring.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/types.h \
+	$(CONTRIBDIR)/qemu/include/qemu/aes.h \
+	$(CONTRIBDIR)/qemu/include/qemu/atomic.h \
+	$(CONTRIBDIR)/qemu/include/qemu/bitmap.h \
+	$(CONTRIBDIR)/qemu/include/qemu/bitops.h \
+	$(CONTRIBDIR)/qemu/include/qemu/bswap.h \
+	$(CONTRIBDIR)/qemu/include/qemu/compiler.h \
+	$(CONTRIBDIR)/qemu/include/qemu/error-report.h \
+	$(CONTRIBDIR)/qemu/include/qemu/event_notifier.h \
+	$(CONTRIBDIR)/qemu/include/qemu/hbitmap.h \
+	$(CONTRIBDIR)/qemu/include/qemu/host-utils.h \
+	$(CONTRIBDIR)/qemu/include/qemu/iov.h \
+	$(CONTRIBDIR)/qemu/include/qemu/main-loop.h \
+	$(CONTRIBDIR)/qemu/include/qemu/module.h \
+	$(CONTRIBDIR)/qemu/include/qemu/notify.h \
+	$(CONTRIBDIR)/qemu/include/qemu/option.h \
+	$(CONTRIBDIR)/qemu/include/qemu/option_int.h \
+	$(CONTRIBDIR)/qemu/include/qemu/osdep.h \
+	$(CONTRIBDIR)/qemu/include/qemu/queue.h \
+	$(CONTRIBDIR)/qemu/include/qemu/sockets.h \
+	$(CONTRIBDIR)/qemu/include/qemu/thread-posix.h \
+	$(CONTRIBDIR)/qemu/include/qemu/thread.h \
+	$(CONTRIBDIR)/qemu/include/qemu/timer.h \
+	$(CONTRIBDIR)/qemu/include/qemu/typedefs.h \
+	$(CONTRIBDIR)/qemu/include/sysemu/sysemu.h \
+	$(CONTRIBDIR)/qemu/include/sysemu/os-posix.h \
+	$(CONTRIBDIR)/qemu/block/qcow2.h \
+	$(CONTRIBDIR)/qemu/block/qed.h
+
+noinst_HEADERS = \
+	$(noinst_HEADERS_qemu) \
+	qemu-block.h \
+	qemu-block-memory-types.h \
+	qb-coroutines.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+            -I$(CONTRIBDIR)/qemu \
+            -I$(CONTRIBDIR)/qemu/include \
+            -DGLUSTER_XLATOR
+
+AM_CFLAGS = -fno-strict-aliasing -Wall $(GF_CFLAGS) $(GLIB_CFLAGS)
+
+CLEANFILES =
+
+endif
diff --git a/xlators/features/qemu-block/src/bdrv-xlator.c b/xlators/features/qemu-block/src/bdrv-xlator.c
new file mode 100644
index 000000000..106c59775
--- /dev/null
+++ b/xlators/features/qemu-block/src/bdrv-xlator.c
@@ -0,0 +1,397 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "inode.h"
+#include "syncop.h"
+#include "qemu-block.h"
+#include "block/block_int.h"
+
+typedef struct BDRVGlusterState {
+	inode_t *inode;
+} BDRVGlusterState;
+
+static QemuOptsList runtime_opts = {
+	.name = "gluster",
+	.head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head),
+	.desc = {
+		{
+			.name = "filename",
+			.type = QEMU_OPT_STRING,
+			.help = "GFID of file",
+		},
+		{ /* end of list */ }
+	},
+};
+
+inode_t *
+qb_inode_from_filename (const char *filename)
+{
+	const char *iptr = NULL;
+	inode_t *inode = NULL;
+
+	iptr = filename + 17;
+	sscanf (iptr, "%p", &inode);
+
+	return inode;
+}
+
+
+int
+qb_inode_to_filename (inode_t *inode, char *filename, int size)
+{
+	return snprintf (filename, size, "gluster://inodep:%p", inode);
+}
+
+
+static fd_t *
+fd_from_bs (BlockDriverState *bs)
+{
+	BDRVGlusterState *s = bs->opaque;
+
+	return fd_anonymous (s->inode);
+}
+
+
+static int
+qemu_gluster_open (BlockDriverState *bs, QDict *options, int bdrv_flags)
+{
+	inode_t *inode = NULL;
+	BDRVGlusterState *s = bs->opaque;
+	QemuOpts *opts = NULL;
+	Error *local_err = NULL;
+	const char *filename = NULL;
+	char gfid_str[128];
+	int ret;
+	qb_conf_t *conf = THIS->private;
+
+	opts = qemu_opts_create_nofail(&runtime_opts);
+	qemu_opts_absorb_qdict(opts, options, &local_err);
+	if (error_is_set(&local_err)) {
+		qerror_report_err(local_err);
+		error_free(local_err);
+		return -EINVAL;
+	}
+
+	filename = qemu_opt_get(opts, "filename");
+
+	/*
+	 * gfid:<gfid> format means we're opening a backing image.
+	 */
+	ret = sscanf(filename, "gluster://gfid:%s", gfid_str);
+	if (ret) {
+		loc_t loc = {0,};
+		struct iatt buf = {0,};
+		uuid_t gfid;
+
+		uuid_parse(gfid_str, gfid);
+
+		loc.inode = inode_find(conf->root_inode->table, gfid);
+		if (!loc.inode) {
+			loc.inode = inode_new(conf->root_inode->table);
+			uuid_copy(loc.inode->gfid, gfid);
+		}
+
+		uuid_copy(loc.gfid, loc.inode->gfid);
+		ret = syncop_lookup(FIRST_CHILD(THIS), &loc, NULL, &buf, NULL,
+				    NULL);
+		if (ret) {
+			loc_wipe(&loc);
+			return -errno;
+		}
+
+		s->inode = inode_ref(loc.inode);
+		loc_wipe(&loc);
+	} else {
+		inode = qb_inode_from_filename (filename);
+		if (!inode)
+			return -EINVAL;
+
+		s->inode = inode_ref(inode);
+	}
+
+	return 0;
+}
+
+
+static int
+qemu_gluster_create (const char *filename, QEMUOptionParameter *options)
+{
+	uint64_t total_size = 0;
+	inode_t *inode = NULL;
+	fd_t *fd = NULL;
+	struct iatt stat = {0, };
+	int ret = 0;
+
+	inode = qb_inode_from_filename (filename);
+	if (!inode)
+		return -EINVAL;
+
+	while (options && options->name) {
+		if (!strcmp(options->name, BLOCK_OPT_SIZE)) {
+			total_size = options->value.n / BDRV_SECTOR_SIZE;
+		}
+		options++;
+	}
+
+	fd = fd_anonymous (inode);
+	if (!fd)
+		return -ENOMEM;
+
+	ret = syncop_fstat (FIRST_CHILD(THIS), fd, &stat);
+	if (ret) {
+		fd_unref (fd);
+		return -errno;
+	}
+
+	if (stat.ia_size) {
+		/* format ONLY if the filesize is 0 bytes */
+		fd_unref (fd);
+		return -EFBIG;
+	}
+
+	if (total_size) {
+		ret = syncop_ftruncate (FIRST_CHILD(THIS), fd, total_size);
+		if (ret) {
+			fd_unref (fd);
+			return -errno;
+		}
+	}
+
+	fd_unref (fd);
+	return 0;
+}
+
+
+static int
+qemu_gluster_co_readv (BlockDriverState *bs, int64_t sector_num, int nb_sectors,
+		       QEMUIOVector *qiov)
+{
+	fd_t *fd = NULL;
+	off_t offset = 0;
+	size_t size = 0;
+	struct iovec *iov = NULL;
+	int count = 0;
+	struct iobref *iobref = NULL;
+	int ret = 0;
+
+	fd = fd_from_bs (bs);
+	if (!fd)
+		return -EIO;
+
+	offset = sector_num * BDRV_SECTOR_SIZE;
+	size = nb_sectors * BDRV_SECTOR_SIZE;
+
+	ret = syncop_readv (FIRST_CHILD(THIS), fd, size, offset, 0,
+			    &iov, &count, &iobref);
+	if (ret < 0) {
+		ret = -errno;
+		goto out;
+	}
+
+	iov_copy (qiov->iov, qiov->niov, iov, count); /* *choke!* */
+
+out:
+	GF_FREE (iov);
+	if (iobref)
+		iobref_unref (iobref);
+	fd_unref (fd);
+	return ret;
+}
+
+
+static int
+qemu_gluster_co_writev (BlockDriverState *bs, int64_t sector_num, int nb_sectors,
+			QEMUIOVector *qiov)
+{
+	fd_t *fd = NULL;
+	off_t offset = 0;
+	size_t size = 0;
+	struct iobref *iobref = NULL;
+	struct iobuf *iobuf = NULL;
+	struct iovec iov = {0, };
+	int ret = -ENOMEM;
+
+	fd = fd_from_bs (bs);
+	if (!fd)
+		return -EIO;
+
+	offset = sector_num * BDRV_SECTOR_SIZE;
+	size = nb_sectors * BDRV_SECTOR_SIZE;
+
+	iobuf = iobuf_get2 (THIS->ctx->iobuf_pool, size);
+	if (!iobuf)
+		goto out;
+
+	iobref = iobref_new ();
+	if (!iobref) {
+		iobuf_unref (iobuf);
+		goto out;
+	}
+
+	iobref_add (iobref, iobuf);
+
+	iov_unload (iobuf_ptr (iobuf), qiov->iov, qiov->niov); /* *choke!* */
+
+	iov.iov_base = iobuf_ptr (iobuf);
+	iov.iov_len = size;
+
+	ret = syncop_writev (FIRST_CHILD(THIS), fd, &iov, 1, offset, iobref, 0);
+	if (ret < 0)
+		ret = -errno;
+
+out:
+	if (iobuf)
+		iobuf_unref (iobuf);
+	if (iobref)
+		iobref_unref (iobref);
+	fd_unref (fd);
+	return ret;
+}
+
+
+static int
+qemu_gluster_co_flush (BlockDriverState *bs)
+{
+	fd_t *fd = NULL;
+	int ret = 0;
+
+	fd = fd_from_bs (bs);
+
+	ret = syncop_flush (FIRST_CHILD(THIS), fd);
+
+	fd_unref (fd);
+
+	return ret;
+}
+
+
+static int
+qemu_gluster_co_fsync (BlockDriverState *bs)
+{
+	fd_t *fd = NULL;
+	int ret = 0;
+
+	fd = fd_from_bs (bs);
+
+	ret = syncop_fsync (FIRST_CHILD(THIS), fd, 0);
+
+	fd_unref (fd);
+
+	return ret;
+}
+
+
+static int
+qemu_gluster_truncate (BlockDriverState *bs, int64_t offset)
+{
+	fd_t *fd = NULL;
+	int ret = 0;
+
+	fd = fd_from_bs (bs);
+
+	ret = syncop_ftruncate (FIRST_CHILD(THIS), fd, offset);
+
+	fd_unref (fd);
+
+	if (ret < 0)
+		return ret;
+
+	return ret;
+}
+
+
+static int64_t
+qemu_gluster_getlength (BlockDriverState *bs)
+{
+	fd_t *fd = NULL;
+	int ret = 0;
+	struct iatt iatt = {0, };
+
+	fd = fd_from_bs (bs);
+
+	ret = syncop_fstat (FIRST_CHILD(THIS), fd, &iatt);
+	if (ret < 0)
+		return -1;
+
+	return iatt.ia_size;
+}
+
+
+static int64_t
+qemu_gluster_allocated_file_size (BlockDriverState *bs)
+{
+	fd_t *fd = NULL;
+	int ret = 0;
+	struct iatt iatt = {0, };
+
+	fd = fd_from_bs (bs);
+
+	ret = syncop_fstat (FIRST_CHILD(THIS), fd, &iatt);
+	if (ret < 0)
+		return -1;
+
+	return iatt.ia_blocks * 512;
+}
+
+
+static void
+qemu_gluster_close (BlockDriverState *bs)
+{
+	BDRVGlusterState *s = NULL;
+
+	s = bs->opaque;
+
+	inode_unref (s->inode);
+
+	return;
+}
+
+
+static QEMUOptionParameter qemu_gluster_create_options[] = {
+	{
+		.name = BLOCK_OPT_SIZE,
+		.type = OPT_SIZE,
+		.help = "Virtual disk size"
+	},
+	{ NULL }
+};
+
+
+static BlockDriver bdrv_gluster = {
+	.format_name                  = "gluster",
+	.protocol_name                = "gluster",
+	.instance_size                = sizeof(BDRVGlusterState),
+	.bdrv_file_open               = qemu_gluster_open,
+	.bdrv_close                   = qemu_gluster_close,
+	.bdrv_create                  = qemu_gluster_create,
+	.bdrv_getlength               = qemu_gluster_getlength,
+	.bdrv_get_allocated_file_size = qemu_gluster_allocated_file_size,
+	.bdrv_co_readv                = qemu_gluster_co_readv,
+	.bdrv_co_writev               = qemu_gluster_co_writev,
+	.bdrv_co_flush_to_os          = qemu_gluster_co_flush,
+	.bdrv_co_flush_to_disk        = qemu_gluster_co_fsync,
+	.bdrv_truncate                = qemu_gluster_truncate,
+	.create_options               = qemu_gluster_create_options,
+};
+
+
+static void bdrv_gluster_init(void)
+{
+	bdrv_register(&bdrv_gluster);
+}
+
+
+block_init(bdrv_gluster_init);
diff --git a/xlators/features/qemu-block/src/bh-syncop.c b/xlators/features/qemu-block/src/bh-syncop.c
new file mode 100644
index 000000000..e8686f6d4
--- /dev/null
+++ b/xlators/features/qemu-block/src/bh-syncop.c
@@ -0,0 +1,48 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "syncop.h"
+#include "qemu-block-memory-types.h"
+
+#include "block/aio.h"
+
+void
+qemu_bh_schedule (QEMUBH *bh)
+{
+	return;
+}
+
+void
+qemu_bh_cancel (QEMUBH *bh)
+{
+	return;
+}
+
+void
+qemu_bh_delete (QEMUBH *bh)
+{
+
+}
+
+QEMUBH *
+qemu_bh_new (QEMUBHFunc *cb, void *opaque)
+{
+	return NULL;
+}
diff --git a/xlators/features/qemu-block/src/clock-timer.c b/xlators/features/qemu-block/src/clock-timer.c
new file mode 100644
index 000000000..fcbec6ad1
--- /dev/null
+++ b/xlators/features/qemu-block/src/clock-timer.c
@@ -0,0 +1,60 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "syncop.h"
+#include "qemu-block-memory-types.h"
+
+#include "qemu/timer.h"
+
+QEMUClock *vm_clock;
+int use_rt_clock = 0;
+
+QEMUTimer *qemu_new_timer (QEMUClock *clock, int scale,
+			   QEMUTimerCB *cb, void *opaque)
+{
+	return NULL;
+}
+
+int64_t qemu_get_clock_ns (QEMUClock *clock)
+{
+	return 0;
+}
+
+void qemu_mod_timer (QEMUTimer *ts, int64_t expire_time)
+{
+	return;
+}
+
+void qemu_free_timer (QEMUTimer *ts)
+{
+
+}
+
+void qemu_del_timer (QEMUTimer *ts)
+{
+
+}
+
+bool qemu_aio_wait()
+{
+	synctask_wake (synctask_get());
+	synctask_yield (synctask_get());
+	return 0;
+}
diff --git a/xlators/features/qemu-block/src/coroutine-synctask.c b/xlators/features/qemu-block/src/coroutine-synctask.c
new file mode 100644
index 000000000..e43988a95
--- /dev/null
+++ b/xlators/features/qemu-block/src/coroutine-synctask.c
@@ -0,0 +1,116 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "syncop.h"
+#include "qemu-block-memory-types.h"
+
+#include "qemu-block.h"
+
+/*
+ * This code serves as the bridge from the main glusterfs context to the qemu
+ * coroutine context via synctask. We create a single threaded syncenv with a
+ * single synctask responsible for processing a queue of coroutines. The qemu
+ * code invoked from within the synctask function handlers uses the ucontext
+ * coroutine implementation and scheduling logic internal to qemu. This
+ * effectively donates a thread of execution to qemu and its internal coroutine
+ * management.
+ *
+ * NOTE: The existence of concurrent synctasks has proven quite racy with regard
+ * to qemu coroutine management, particularly related to the lifecycle
+ * differences with top-level synctasks and internally created coroutines and
+ * interactions with qemu-internal queues (and locks, in turn). We explicitly
+ * disallow this scenario, via the queue, until it is more well supported.
+ */
+
+static struct {
+	struct list_head	queue;
+	gf_lock_t		lock;
+	struct synctask		*task;
+} qb_co;
+
+static void
+init_qbco()
+{
+	INIT_LIST_HEAD(&qb_co.queue);
+	LOCK_INIT(&qb_co.lock);
+}
+
+static int
+synctask_nop_cbk (int ret, call_frame_t *frame, void *opaque)
+{
+	return 0;
+}
+
+static int
+qb_synctask_wrap (void *opaque)
+{
+	qb_local_t *qb_local, *tmp;
+
+	LOCK(&qb_co.lock);
+
+	while (!list_empty(&qb_co.queue)) {
+		list_for_each_entry_safe(qb_local, tmp, &qb_co.queue, list) {
+			list_del_init(&qb_local->list);
+			break;
+		}
+
+		UNLOCK(&qb_co.lock);
+
+		qb_local->synctask_fn(qb_local);
+		/* qb_local is now unwound and gone! */
+
+		LOCK(&qb_co.lock);
+	}
+
+	qb_co.task = NULL;
+
+	UNLOCK(&qb_co.lock);
+
+	return 0;
+}
+
+int
+qb_coroutine (call_frame_t *frame, synctask_fn_t fn)
+{
+	qb_local_t *qb_local = NULL;
+	qb_conf_t *qb_conf = NULL;
+	static int init = 0;
+
+	qb_local = frame->local;
+	qb_local->synctask_fn = fn;
+	qb_conf = frame->this->private;
+
+	if (!init) {
+		init = 1;
+		init_qbco();
+	}
+
+	LOCK(&qb_co.lock);
+
+	if (!qb_co.task)
+		qb_co.task = synctask_create(qb_conf->env, qb_synctask_wrap,
+					     synctask_nop_cbk, frame, NULL);
+
+	list_add_tail(&qb_local->list, &qb_co.queue);
+
+	UNLOCK(&qb_co.lock);
+
+	return 0;
+}
diff --git a/xlators/features/qemu-block/src/monitor-logging.c b/xlators/features/qemu-block/src/monitor-logging.c
new file mode 100644
index 000000000..d37c37f0f
--- /dev/null
+++ b/xlators/features/qemu-block/src/monitor-logging.c
@@ -0,0 +1,50 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "qemu-block-memory-types.h"
+
+#include "block/block_int.h"
+
+Monitor *cur_mon;
+
+int
+monitor_cur_is_qmp()
+{
+	/* No QMP support here */
+	return 0;
+}
+
+void
+monitor_set_error (Monitor *mon, QError *qerror)
+{
+	/* NOP here */
+	return;
+}
+
+
+void
+monitor_vprintf(Monitor *mon, const char *fmt, va_list ap)
+{
+	char buf[4096];
+
+	vsnprintf(buf, sizeof(buf), fmt, ap);
+
+	gf_log (THIS->name, GF_LOG_ERROR, "%s", buf);
+}
diff --git a/xlators/features/qemu-block/src/qb-coroutines.c b/xlators/features/qemu-block/src/qb-coroutines.c
new file mode 100644
index 000000000..7c52adb21
--- /dev/null
+++ b/xlators/features/qemu-block/src/qb-coroutines.c
@@ -0,0 +1,662 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "inode.h"
+#include "call-stub.h"
+#include "defaults.h"
+#include "qemu-block-memory-types.h"
+#include "qemu-block.h"
+#include "qb-coroutines.h"
+
+
+int
+qb_format_and_resume (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	char filename[64];
+	char base_filename[128];
+	int use_base = 0;
+	qb_inode_t *qb_inode = NULL;
+	Error *local_err = NULL;
+	fd_t *fd = NULL;
+	dict_t *xattr = NULL;
+	qb_conf_t *qb_conf = NULL;
+	int ret = -1;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+	qb_conf = frame->this->private;
+
+	qb_inode_to_filename (inode, filename, 64);
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+
+	/*
+	 * See if the caller specified a backing image.
+	 */
+	if (!uuid_is_null(qb_inode->backing_gfid) || qb_inode->backing_fname) {
+		loc_t loc = {0,};
+		char gfid_str[64];
+		struct iatt buf;
+
+		if (!uuid_is_null(qb_inode->backing_gfid)) {
+			loc.inode = inode_find(qb_conf->root_inode->table,
+					qb_inode->backing_gfid);
+			if (!loc.inode) {
+				loc.inode = inode_new(qb_conf->root_inode->table);
+				uuid_copy(loc.inode->gfid,
+					qb_inode->backing_gfid);
+			}
+			uuid_copy(loc.gfid, loc.inode->gfid);
+		} else if (qb_inode->backing_fname) {
+			loc.inode = inode_new(qb_conf->root_inode->table);
+			loc.name = qb_inode->backing_fname;
+			loc.parent = inode_parent(inode, NULL, NULL);
+			loc_path(&loc, loc.name);
+		}
+
+		/*
+		 * Lookup the backing image. Verify existence and/or get the
+		 * gfid if we don't already have it.
+		 */
+		ret = syncop_lookup(FIRST_CHILD(frame->this), &loc, NULL, &buf,
+				    NULL, NULL);
+		GF_FREE(qb_inode->backing_fname);
+		if (ret) {
+			loc_wipe(&loc);
+			ret = errno;
+			goto err;
+		}
+
+		uuid_copy(qb_inode->backing_gfid, buf.ia_gfid);
+		loc_wipe(&loc);
+
+		/*
+		 * We pass the filename of the backing image into the qemu block
+		 * subsystem as the associated gfid. This is embedded into the
+		 * clone image and passed along to the gluster bdrv backend when
+		 * the block subsystem needs to operate on the backing image on
+		 * behalf of the clone.
+		 */
+		uuid_unparse(qb_inode->backing_gfid, gfid_str);
+		snprintf(base_filename, sizeof(base_filename),
+			 "gluster://gfid:%s", gfid_str);
+		use_base = 1;
+	}
+
+	bdrv_img_create (filename, qb_inode->fmt,
+			 use_base ? base_filename : NULL, 0, 0, qb_inode->size,
+			 0, &local_err, true);
+
+	if (error_is_set (&local_err)) {
+		gf_log (frame->this->name, GF_LOG_ERROR, "%s",
+			error_get_pretty (local_err));
+		error_free (local_err);
+		QB_STUB_UNWIND (stub, -1, EIO);
+		return 0;
+	}
+
+	fd = fd_anonymous (inode);
+	if (!fd) {
+		gf_log (frame->this->name, GF_LOG_ERROR,
+			"could not create anonymous fd for %s",
+			uuid_utoa (inode->gfid));
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		return 0;
+	}
+
+	xattr = dict_new ();
+	if (!xattr) {
+		gf_log (frame->this->name, GF_LOG_ERROR,
+			"could not allocate xattr dict for %s",
+			uuid_utoa (inode->gfid));
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		fd_unref (fd);
+		return 0;
+	}
+
+	ret = dict_set_str (xattr, qb_conf->qb_xattr_key, local->fmt);
+	if (ret) {
+		gf_log (frame->this->name, GF_LOG_ERROR,
+			"could not dict_set for %s",
+			uuid_utoa (inode->gfid));
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		fd_unref (fd);
+		dict_unref (xattr);
+		return 0;
+	}
+
+	ret = syncop_fsetxattr (FIRST_CHILD(THIS), fd, xattr, 0);
+	if (ret) {
+		ret = errno;
+		gf_log (frame->this->name, GF_LOG_ERROR,
+			"failed to setxattr for %s",
+			uuid_utoa (inode->gfid));
+		QB_STUB_UNWIND (stub, -1, ret);
+		fd_unref (fd);
+		dict_unref (xattr);
+		return 0;
+	}
+
+	fd_unref (fd);
+	dict_unref (xattr);
+
+	QB_STUB_UNWIND (stub, 0, 0);
+
+	return 0;
+
+err:
+	QB_STUB_UNWIND(stub, -1, ret);
+	return 0;
+}
+
+
+static BlockDriverState *
+qb_bs_create (inode_t *inode, const char *fmt)
+{
+	char filename[64];
+	BlockDriverState *bs = NULL;
+	BlockDriver *drv = NULL;
+	int op_errno = 0;
+	int ret = 0;
+
+	bs = bdrv_new (uuid_utoa (inode->gfid));
+	if (!bs) {
+		op_errno = ENOMEM;
+		gf_log (THIS->name, GF_LOG_ERROR,
+			"could not allocate @bdrv for gfid:%s",
+			uuid_utoa (inode->gfid));
+		goto err;
+	}
+
+	drv = bdrv_find_format (fmt);
+	if (!drv) {
+		op_errno = EINVAL;
+		gf_log (THIS->name, GF_LOG_ERROR,
+			"Unknown file format: %s for gfid:%s",
+			fmt, uuid_utoa (inode->gfid));
+		goto err;
+	}
+
+	qb_inode_to_filename (inode, filename, 64);
+
+	ret = bdrv_open (bs, filename, NULL, BDRV_O_RDWR, drv);
+	if (ret < 0) {
+		op_errno = -ret;
+		gf_log (THIS->name, GF_LOG_ERROR,
+			"Unable to bdrv_open() gfid:%s (%s)",
+			uuid_utoa (inode->gfid), strerror (op_errno));
+		goto err;
+	}
+
+	return bs;
+err:
+	errno = op_errno;
+	return NULL;
+}
+
+
+int
+qb_co_open (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+	qb_inode->refcnt++;
+
+	QB_STUB_RESUME (stub);
+
+	return 0;
+}
+
+
+int
+qb_co_writev (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	QEMUIOVector qiov = {0, };
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	qemu_iovec_init_external (&qiov, stub->args.vector, stub->args.count);
+
+	ret = bdrv_pwritev (qb_inode->bs, stub->args.offset, &qiov);
+
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
+
+
+int
+qb_co_readv (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	struct iobuf *iobuf = NULL;
+	struct iobref *iobref = NULL;
+	struct iovec iov = {0, };
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	if (stub->args.offset >= qb_inode->size) {
+		QB_STUB_UNWIND (stub, 0, 0);
+		return 0;
+	}
+
+	iobuf = iobuf_get2 (frame->this->ctx->iobuf_pool, stub->args.size);
+	if (!iobuf) {
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		return 0;
+	}
+
+	iobref = iobref_new ();
+	if (!iobref) {
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		iobuf_unref (iobuf);
+		return 0;
+	}
+
+	if (iobref_add (iobref, iobuf) < 0) {
+		iobuf_unref (iobuf);
+		iobref_unref (iobref);
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		return 0;
+	}
+
+	ret = bdrv_pread (qb_inode->bs, stub->args.offset, iobuf_ptr (iobuf),
+			  stub->args.size);
+
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+		iobref_unref (iobref);
+		return 0;
+	}
+
+	iov.iov_base = iobuf_ptr (iobuf);
+	iov.iov_len = ret;
+
+	stub->args_cbk.vector = iov_dup (&iov, 1);
+	stub->args_cbk.count = 1;
+	stub->args_cbk.iobref = iobref;
+
+	QB_STUB_UNWIND (stub, ret, 0);
+
+	return 0;
+}
+
+
+int
+qb_co_fsync (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	ret = bdrv_flush (qb_inode->bs);
+
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
+
+
+static void
+qb_update_size_xattr (xlator_t *this, fd_t *fd, const char *fmt, off_t offset)
+{
+	char val[QB_XATTR_VAL_MAX];
+	qb_conf_t *qb_conf = NULL;
+	dict_t *xattr = NULL;
+
+	qb_conf = this->private;
+
+	snprintf (val, QB_XATTR_VAL_MAX, "%s:%llu",
+		  fmt, (long long unsigned) offset);
+
+	xattr = dict_new ();
+	if (!xattr)
+		return;
+
+	if (dict_set_str (xattr, qb_conf->qb_xattr_key, val) != 0) {
+		dict_unref (xattr);
+		return;
+	}
+
+	syncop_fsetxattr (FIRST_CHILD(this), fd, xattr, 0);
+	dict_unref (xattr);
+}
+
+
+int
+qb_co_truncate (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	int ret = 0;
+	off_t offset = 0;
+	xlator_t *this = NULL;
+
+	this = THIS;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	syncop_fstat (FIRST_CHILD(this), local->fd, &stub->args_cbk.prestat);
+	stub->args_cbk.prestat.ia_size = qb_inode->size;
+
+	ret = bdrv_truncate (qb_inode->bs, stub->args.offset);
+	if (ret < 0)
+		goto out;
+
+	offset = bdrv_getlength (qb_inode->bs);
+
+	qb_inode->size = offset;
+
+	syncop_fstat (FIRST_CHILD(this), local->fd, &stub->args_cbk.poststat);
+	stub->args_cbk.poststat.ia_size = qb_inode->size;
+
+	qb_update_size_xattr (this, local->fd, qb_inode->fmt, qb_inode->size);
+
+out:
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
+
+
+int
+qb_co_close (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	BlockDriverState *bs = NULL;
+
+	local = opaque;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (THIS, inode);
+
+	if (!--qb_inode->refcnt) {
+		bs = qb_inode->bs;
+		qb_inode->bs = NULL;
+		bdrv_delete (bs);
+	}
+
+	frame = local->frame;
+	frame->local = NULL;
+	qb_local_free (THIS, local);
+	STACK_DESTROY (frame->root);
+
+	return 0;
+}
+
+
+int
+qb_snapshot_create (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	QEMUSnapshotInfo sn;
+	struct timeval tv = {0, };
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	memset (&sn, 0, sizeof (sn));
+        pstrcpy (sn.name, sizeof(sn.name), local->name);
+        gettimeofday (&tv, NULL);
+        sn.date_sec = tv.tv_sec;
+        sn.date_nsec = tv.tv_usec * 1000;
+
+        ret = bdrv_snapshot_create (qb_inode->bs, &sn);
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
+
+
+int
+qb_snapshot_delete (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	ret = bdrv_snapshot_delete (qb_inode->bs, local->name);
+
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
+
+
+int
+qb_snapshot_goto (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	ret = bdrv_snapshot_goto (qb_inode->bs, local->name);
+
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
diff --git a/xlators/features/qemu-block/src/qb-coroutines.h b/xlators/features/qemu-block/src/qb-coroutines.h
new file mode 100644
index 000000000..583319f3b
--- /dev/null
+++ b/xlators/features/qemu-block/src/qb-coroutines.h
@@ -0,0 +1,30 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef __QB_COROUTINES_H
+#define __QB_COROUTINES_H
+
+#include "syncop.h"
+#include "call-stub.h"
+#include "block/block_int.h"
+#include "monitor/monitor.h"
+
+int qb_format_and_resume (void *opaque);
+int qb_snapshot_create (void *opaque);
+int qb_snapshot_delete (void *opaque);
+int qb_snapshot_goto (void *opaque);
+int qb_co_open (void *opaque);
+int qb_co_close (void *opaque);
+int qb_co_writev (void *opaque);
+int qb_co_readv (void *opaque);
+int qb_co_fsync (void *opaque);
+int qb_co_truncate (void *opaque);
+
+#endif /* __QB_COROUTINES_H */
diff --git a/xlators/features/qemu-block/src/qemu-block-memory-types.h b/xlators/features/qemu-block/src/qemu-block-memory-types.h
new file mode 100644
index 000000000..267b3893f
--- /dev/null
+++ b/xlators/features/qemu-block/src/qemu-block-memory-types.h
@@ -0,0 +1,25 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef __QB_MEM_TYPES_H__
+#define __QB_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_qb_mem_types_ {
+        gf_qb_mt_qb_conf_t   = gf_common_mt_end + 1,
+        gf_qb_mt_qb_inode_t,
+        gf_qb_mt_qb_local_t,
+	gf_qb_mt_coroutinesynctask_t,
+        gf_qb_mt_end
+};
+#endif
+
diff --git a/xlators/features/qemu-block/src/qemu-block.c b/xlators/features/qemu-block/src/qemu-block.c
new file mode 100644
index 000000000..48bbf3140
--- /dev/null
+++ b/xlators/features/qemu-block/src/qemu-block.c
@@ -0,0 +1,1140 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "inode.h"
+#include "call-stub.h"
+#include "defaults.h"
+#include "qemu-block-memory-types.h"
+#include "qemu-block.h"
+#include "qb-coroutines.h"
+
+
+qb_inode_t *
+__qb_inode_ctx_get (xlator_t *this, inode_t *inode)
+{
+        uint64_t    value    = 0;
+        qb_inode_t *qb_inode = NULL;
+
+        __inode_ctx_get (inode, this, &value);
+        qb_inode = (qb_inode_t *)(unsigned long) value;
+
+        return qb_inode;
+}
+
+
+qb_inode_t *
+qb_inode_ctx_get (xlator_t *this, inode_t *inode)
+{
+        qb_inode_t *qb_inode = NULL;
+
+        LOCK (&inode->lock);
+        {
+                qb_inode = __qb_inode_ctx_get (this, inode);
+        }
+        UNLOCK (&inode->lock);
+
+        return qb_inode;
+}
+
+
+qb_inode_t *
+qb_inode_ctx_del (xlator_t *this, inode_t *inode)
+{
+        uint64_t    value    = 0;
+        qb_inode_t *qb_inode = NULL;
+
+        inode_ctx_del (inode, this, &value);
+        qb_inode = (qb_inode_t *)(unsigned long) value;
+
+        return qb_inode;
+}
+
+
+int
+qb_inode_cleanup (xlator_t *this, inode_t *inode, int warn)
+{
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_del (this, inode);
+
+	if (!qb_inode)
+		return 0;
+
+	if (warn)
+		gf_log (this->name, GF_LOG_WARNING,
+			"inode %s no longer block formatted",
+			uuid_utoa (inode->gfid));
+
+	/* free (qb_inode->bs); */
+
+	GF_FREE (qb_inode);
+
+	return 0;
+}
+
+
+int
+qb_iatt_fixup (xlator_t *this, inode_t *inode, struct iatt *iatt)
+{
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, inode);
+	if (!qb_inode)
+		return 0;
+
+	iatt->ia_size = qb_inode->size;
+
+	return 0;
+}
+
+
+int
+qb_format_extract (xlator_t *this, char *format, inode_t *inode)
+{
+	char       *s, *save;
+	uint64_t    size = 0;
+	char        fmt[QB_XATTR_VAL_MAX+1] = {0, };
+	qb_inode_t *qb_inode = NULL;
+	char *formatstr = NULL;
+	uuid_t gfid = {0,};
+	char gfid_str[64] = {0,};
+	int ret;
+
+	strncpy(fmt, format, QB_XATTR_VAL_MAX);
+
+	s = strtok_r(fmt, ":", &save);
+	if (!s)
+		goto invalid;
+	formatstr = gf_strdup(s);
+
+	s = strtok_r(NULL, ":", &save);
+	if (!s)
+		goto invalid;
+	if (gf_string2bytesize (s, &size))
+		goto invalid;
+	if (!size)
+		goto invalid;
+
+	s = strtok_r(NULL, "\0", &save);
+	if (s && !strncmp(s, "<gfid:", strlen("<gfid:"))) {
+		/*
+		 * Check for valid gfid backing image specifier.
+		 */
+		if (strlen(s) + 1 > sizeof(gfid_str))
+			goto invalid;
+		ret = sscanf(s, "<gfid:%[^>]s", gfid_str);
+		if (ret == 1) {
+			ret = uuid_parse(gfid_str, gfid);
+			if (ret < 0)
+				goto invalid;
+		}
+	}
+
+	qb_inode = qb_inode_ctx_get (this, inode);
+	if (!qb_inode)
+		qb_inode = GF_CALLOC (1, sizeof (*qb_inode),
+				      gf_qb_mt_qb_inode_t);
+	if (!qb_inode) {
+		GF_FREE(formatstr);
+		return ENOMEM;
+	}
+
+	strncpy(qb_inode->fmt, formatstr, QB_XATTR_VAL_MAX);
+	qb_inode->size = size;
+
+	/*
+	 * If a backing gfid was not specified, interpret any remaining bytes
+	 * associated with a backing image as a filename local to the parent
+	 * directory. The format processing will validate further.
+	 */
+	if (!uuid_is_null(gfid))
+		uuid_copy(qb_inode->backing_gfid, gfid);
+	else if (s)
+		qb_inode->backing_fname = gf_strdup(s);
+
+	inode_ctx_set (inode, this, (void *)&qb_inode);
+
+	GF_FREE(formatstr);
+
+	return 0;
+
+invalid:
+	GF_FREE(formatstr);
+
+	gf_log (this->name, GF_LOG_WARNING,
+		"invalid format '%s' in inode %s", format,
+		uuid_utoa (inode->gfid));
+	return EINVAL;
+}
+
+
+void
+qb_local_free (xlator_t *this, qb_local_t *local)
+{
+	if (local->inode)
+		inode_unref (local->inode);
+	if (local->fd)
+		fd_unref (local->fd);
+	GF_FREE (local);
+}
+
+
+int
+qb_local_init (call_frame_t *frame)
+{
+	qb_local_t *qb_local = NULL;
+
+	qb_local = GF_CALLOC (1, sizeof (*qb_local), gf_qb_mt_qb_local_t);
+	if (!qb_local)
+		return -1;
+	INIT_LIST_HEAD(&qb_local->list);
+
+	qb_local->frame = frame;
+	frame->local = qb_local;
+
+	return 0;
+}
+
+
+int
+qb_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+	       int op_ret, int op_errno, inode_t *inode, struct iatt *buf,
+	       dict_t *xdata, struct iatt *postparent)
+{
+	char *format = NULL;
+	qb_conf_t *conf = NULL;
+
+	conf = this->private;
+
+	if (op_ret == -1)
+		goto out;
+
+	/*
+	 * Cache the root inode for dealing with backing images. The format
+	 * coroutine and the gluster qemu backend driver both use the root inode
+	 * table to verify and/or redirect I/O to the backing image via
+	 * anonymous fd's.
+	 */
+	if (!conf->root_inode && __is_root_gfid(inode->gfid))
+		conf->root_inode = inode_ref(inode);
+
+	if (!xdata)
+		goto out;
+
+	if (dict_get_str (xdata, conf->qb_xattr_key, &format))
+		goto out;
+
+	if (!format) {
+		qb_inode_cleanup (this, inode, 1);
+		goto out;
+	}
+
+	op_errno = qb_format_extract (this, format, inode);
+	if (op_errno)
+		op_ret = -1;
+
+	qb_iatt_fixup (this, inode, buf);
+out:
+	QB_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, buf,
+			 xdata, postparent);
+	return 0;
+}
+
+
+int
+qb_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+	qb_conf_t *conf = NULL;
+
+	conf = this->private;
+
+	xdata = xdata ? dict_ref (xdata) : dict_new ();
+
+	if (!xdata)
+		goto enomem;
+
+	if (dict_set_int32 (xdata, conf->qb_xattr_key, 0))
+		goto enomem;
+
+	STACK_WIND (frame, qb_lookup_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->lookup, loc, xdata);
+	dict_unref (xdata);
+	return 0;
+enomem:
+	QB_STACK_UNWIND (lookup, frame, -1, ENOMEM, 0, 0, 0, 0);
+	if (xdata)
+		dict_unref (xdata);
+	return 0;
+}
+
+
+int
+qb_setxattr_format (call_frame_t *frame, xlator_t *this, call_stub_t *stub,
+		    dict_t *xattr, inode_t *inode)
+{
+	char *format = NULL;
+	int op_errno = 0;
+	qb_local_t *qb_local = NULL;
+	data_t *data = NULL;
+	qb_inode_t *qb_inode;
+
+	if (!(data = dict_get (xattr, "trusted.glusterfs.block-format"))) {
+		QB_STUB_RESUME (stub);
+		return 0;
+	}
+
+	format = alloca (data->len + 1);
+	memcpy (format, data->data, data->len);
+	format[data->len] = 0;
+
+	op_errno = qb_format_extract (this, format, inode);
+	if (op_errno) {
+		QB_STUB_UNWIND (stub, -1, op_errno);
+		return 0;
+	}
+	qb_inode = qb_inode_ctx_get(this, inode);
+
+	qb_local = frame->local;
+
+	qb_local->stub = stub;
+	qb_local->inode = inode_ref (inode);
+
+	snprintf(qb_local->fmt, QB_XATTR_VAL_MAX, "%s:%lu", qb_inode->fmt,
+		 qb_inode->size);
+
+	qb_coroutine (frame, qb_format_and_resume);
+
+	return 0;
+}
+
+
+int
+qb_setxattr_snapshot_create (call_frame_t *frame, xlator_t *this,
+			     call_stub_t *stub, dict_t *xattr, inode_t *inode)
+{
+	qb_local_t *qb_local = NULL;
+	char *name = NULL;
+	data_t *data = NULL;
+
+	if (!(data = dict_get (xattr, "trusted.glusterfs.block-snapshot-create"))) {
+		QB_STUB_RESUME (stub);
+		return 0;
+	}
+
+	name = alloca (data->len + 1);
+	memcpy (name, data->data, data->len);
+	name[data->len] = 0;
+
+	qb_local = frame->local;
+
+	qb_local->stub = stub;
+	qb_local->inode = inode_ref (inode);
+	strncpy (qb_local->name, name, 128);
+
+	qb_coroutine (frame, qb_snapshot_create);
+
+	return 0;
+}
+
+
+int
+qb_setxattr_snapshot_delete (call_frame_t *frame, xlator_t *this,
+			     call_stub_t *stub, dict_t *xattr, inode_t *inode)
+{
+	qb_local_t *qb_local = NULL;
+	char *name = NULL;
+	data_t *data = NULL;
+
+	if (!(data = dict_get (xattr, "trusted.glusterfs.block-snapshot-delete"))) {
+		QB_STUB_RESUME (stub);
+		return 0;
+	}
+
+	name = alloca (data->len + 1);
+	memcpy (name, data->data, data->len);
+	name[data->len] = 0;
+
+	qb_local = frame->local;
+
+	qb_local->stub = stub;
+	qb_local->inode = inode_ref (inode);
+	strncpy (qb_local->name, name, 128);
+
+	qb_coroutine (frame, qb_snapshot_delete);
+
+	return 0;
+}
+
+int
+qb_setxattr_snapshot_goto (call_frame_t *frame, xlator_t *this,
+			   call_stub_t *stub, dict_t *xattr, inode_t *inode)
+{
+	qb_local_t *qb_local = NULL;
+	char *name = NULL;
+	data_t *data = NULL;
+
+	if (!(data = dict_get (xattr, "trusted.glusterfs.block-snapshot-goto"))) {
+		QB_STUB_RESUME (stub);
+		return 0;
+	}
+
+	name = alloca (data->len + 1);
+	memcpy (name, data->data, data->len);
+	name[data->len] = 0;
+
+	qb_local = frame->local;
+
+	qb_local->stub = stub;
+	qb_local->inode = inode_ref (inode);
+	strncpy (qb_local->name, name, 128);
+
+	qb_coroutine (frame, qb_snapshot_goto);
+
+	return 0;
+}
+
+
+int
+qb_setxattr_common (call_frame_t *frame, xlator_t *this, call_stub_t *stub,
+		    dict_t *xattr, inode_t *inode)
+{
+	data_t *data = NULL;
+
+	if ((data = dict_get (xattr, "trusted.glusterfs.block-format"))) {
+		qb_setxattr_format (frame, this, stub, xattr, inode);
+		return 0;
+	}
+
+	if ((data = dict_get (xattr, "trusted.glusterfs.block-snapshot-create"))) {
+		qb_setxattr_snapshot_create (frame, this, stub, xattr, inode);
+		return 0;
+	}
+
+	if ((data = dict_get (xattr, "trusted.glusterfs.block-snapshot-delete"))) {
+		qb_setxattr_snapshot_delete (frame, this, stub, xattr, inode);
+		return 0;
+	}
+
+	if ((data = dict_get (xattr, "trusted.glusterfs.block-snapshot-goto"))) {
+		qb_setxattr_snapshot_goto (frame, this, stub, xattr, inode);
+		return 0;
+	}
+
+	QB_STUB_RESUME (stub);
+
+	return 0;
+}
+
+
+int
+qb_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr,
+	     int flags, dict_t *xdata)
+{
+	call_stub_t *stub = NULL;
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	stub = fop_setxattr_stub (frame, default_setxattr_resume, loc, xattr,
+				  flags, xdata);
+	if (!stub)
+		goto enomem;
+
+	qb_setxattr_common (frame, this, stub, xattr, loc->inode);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (setxattr, frame, -1, ENOMEM, 0);
+	return 0;
+}
+
+
+int
+qb_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xattr,
+	      int flags, dict_t *xdata)
+{
+	call_stub_t *stub = NULL;
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	stub = fop_fsetxattr_stub (frame, default_fsetxattr_resume, fd, xattr,
+				   flags, xdata);
+	if (!stub)
+		goto enomem;
+
+	qb_setxattr_common (frame, this, stub, xattr, fd->inode);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (fsetxattr, frame, -1, ENOMEM, 0);
+	return 0;
+}
+
+
+int
+qb_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+	     int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
+{
+	call_stub_t *stub = NULL;
+	qb_local_t *qb_local = NULL;
+
+	qb_local = frame->local;
+
+	if (op_ret < 0)
+		goto unwind;
+
+	if (!qb_inode_ctx_get (this, qb_local->inode))
+		goto unwind;
+
+	stub = fop_open_cbk_stub (frame, NULL, op_ret, op_errno, fd, xdata);
+	if (!stub) {
+		op_ret = -1;
+		op_errno = ENOMEM;
+		goto unwind;
+	}
+
+	qb_local->stub = stub;
+
+	qb_coroutine (frame, qb_co_open);
+
+	return 0;
+unwind:
+	QB_STACK_UNWIND (open, frame, op_ret, op_errno, fd, xdata);
+	return 0;
+}
+
+
+int
+qb_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+	 fd_t *fd, dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, loc->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_open_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->open, loc, flags, fd,
+			    xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (loc->inode);
+	qb_local->fd = fd_ref (fd);
+
+	STACK_WIND (frame, qb_open_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+	return 0;
+enomem:
+	QB_STACK_UNWIND (open, frame, -1, ENOMEM, 0, 0);
+	return 0;
+}
+
+
+int
+qb_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+	   int count, off_t offset, uint32_t flags, struct iobref *iobref,
+	   dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, fd->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_writev_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->writev, fd, vector, count,
+			    offset, flags, iobref, xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (fd->inode);
+	qb_local->fd = fd_ref (fd);
+
+	qb_local->stub = fop_writev_stub (frame, NULL, fd, vector, count,
+					  offset, flags, iobref, xdata);
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_writev);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (writev, frame, -1, ENOMEM, 0, 0, 0);
+	return 0;
+}
+
+
+int
+qb_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+	  off_t offset, uint32_t flags, dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, fd->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_readv_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->readv, fd, size, offset,
+			    flags, xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (fd->inode);
+	qb_local->fd = fd_ref (fd);
+
+	qb_local->stub = fop_readv_stub (frame, NULL, fd, size, offset,
+					  flags, xdata);
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_readv);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (readv, frame, -1, ENOMEM, 0, 0, 0, 0, 0);
+	return 0;
+}
+
+
+int
+qb_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int dsync,
+	  dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, fd->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_fsync_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->fsync, fd, dsync, xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (fd->inode);
+	qb_local->fd = fd_ref (fd);
+
+	qb_local->stub = fop_fsync_stub (frame, NULL, fd, dsync, xdata);
+
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_fsync);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (fsync, frame, -1, ENOMEM, 0, 0, 0);
+	return 0;
+}
+
+
+int
+qb_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, fd->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_flush_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->flush, fd, xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (fd->inode);
+	qb_local->fd = fd_ref (fd);
+
+	qb_local->stub = fop_flush_stub (frame, NULL, fd, xdata);
+
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_fsync);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (flush, frame, -1, ENOMEM, 0);
+	return 0;
+}
+
+static int32_t
+qb_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+		int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+		dict_t *xdata)
+{
+	qb_conf_t *conf = this->private;
+	gf_dirent_t *entry;
+	char *format;
+
+	list_for_each_entry(entry, &entries->list, list) {
+		if (!entry->inode || !entry->dict)
+			continue;
+
+		format = NULL;
+		if (dict_get_str(entry->dict, conf->qb_xattr_key, &format))
+			continue;
+
+		if (!format) {
+			qb_inode_cleanup(this, entry->inode, 1);
+			continue;
+		}
+
+		if (qb_format_extract(this, format, entry->inode))
+			continue;
+
+		qb_iatt_fixup(this, entry->inode, &entry->d_stat);
+	}
+
+	STACK_UNWIND_STRICT(readdirp, frame, op_ret, op_errno, entries, xdata);
+	return 0;
+}
+
+static int32_t
+qb_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+	    off_t off, dict_t *xdata)
+{
+	qb_conf_t *conf = this->private;
+
+	xdata = xdata ? dict_ref(xdata) : dict_new();
+	if (!xdata)
+		goto enomem;
+
+	if (dict_set_int32 (xdata, conf->qb_xattr_key, 0))
+		goto enomem;
+
+	STACK_WIND(frame, qb_readdirp_cbk, FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->readdirp, fd, size, off, xdata);
+
+	dict_unref(xdata);
+	return 0;
+
+enomem:
+	QB_STACK_UNWIND(readdirp, frame, -1, ENOMEM, NULL, NULL);
+	if (xdata)
+		dict_unref(xdata);
+	return 0;
+}
+
+int
+qb_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+	     dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, loc->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_truncate_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->truncate, loc, offset,
+			    xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (loc->inode);
+	qb_local->fd = fd_anonymous (loc->inode);
+
+	qb_local->stub = fop_truncate_stub (frame, NULL, loc, offset, xdata);
+
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_truncate);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (truncate, frame, -1, ENOMEM, 0, 0, 0);
+	return 0;
+}
+
+
+int
+qb_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	     dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, fd->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_ftruncate_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->ftruncate, fd, offset,
+			    xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (fd->inode);
+	qb_local->fd = fd_ref (fd);
+
+	qb_local->stub = fop_ftruncate_stub (frame, NULL, fd, offset, xdata);
+
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_truncate);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (ftruncate, frame, -1, ENOMEM, 0, 0, 0);
+	return 0;
+}
+
+
+int
+qb_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+	     int op_ret, int op_errno, struct iatt *iatt, dict_t *xdata)
+{
+	inode_t *inode = NULL;
+
+	inode = frame->local;
+	frame->local = NULL;
+
+	if (inode) {
+		qb_iatt_fixup (this, inode, iatt);
+		inode_unref (inode);
+	}
+
+	QB_STACK_UNWIND (stat, frame, op_ret, op_errno, iatt, xdata);
+
+	return 0;
+}
+
+int
+qb_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+	if (qb_inode_ctx_get (this, loc->inode))
+		frame->local = inode_ref (loc->inode);
+
+	STACK_WIND (frame, qb_stat_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->stat, loc, xdata);
+	return 0;
+}
+
+
+int
+qb_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+	      int op_ret, int op_errno, struct iatt *iatt, dict_t *xdata)
+{
+	inode_t *inode = NULL;
+
+	inode = frame->local;
+	frame->local = NULL;
+
+	if (inode) {
+		qb_iatt_fixup (this, inode, iatt);
+		inode_unref (inode);
+	}
+
+	QB_STACK_UNWIND (fstat, frame, op_ret, op_errno, iatt, xdata);
+
+	return 0;
+}
+
+
+int
+qb_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+	if (qb_inode_ctx_get (this, fd->inode))
+		frame->local = inode_ref (fd->inode);
+
+	STACK_WIND (frame, qb_fstat_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->fstat, fd, xdata);
+	return 0;
+}
+
+
+int
+qb_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		int op_ret, int op_errno, struct iatt *pre, struct iatt *post,
+		dict_t *xdata)
+{
+	inode_t *inode = NULL;
+
+	inode = frame->local;
+	frame->local = NULL;
+
+	if (inode) {
+		qb_iatt_fixup (this, inode, pre);
+		qb_iatt_fixup (this, inode, post);
+		inode_unref (inode);
+	}
+
+	QB_STACK_UNWIND (setattr, frame, op_ret, op_errno, pre, post, xdata);
+
+	return 0;
+}
+
+
+int
+qb_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *buf,
+	    int valid, dict_t *xdata)
+{
+	if (qb_inode_ctx_get (this, loc->inode))
+		frame->local = inode_ref (loc->inode);
+
+	STACK_WIND (frame, qb_setattr_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->setattr, loc, buf, valid, xdata);
+	return 0;
+}
+
+
+int
+qb_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		 int op_ret, int op_errno, struct iatt *pre, struct iatt *post,
+		 dict_t *xdata)
+{
+	inode_t *inode = NULL;
+
+	inode = frame->local;
+	frame->local = NULL;
+
+	if (inode) {
+		qb_iatt_fixup (this, inode, pre);
+		qb_iatt_fixup (this, inode, post);
+		inode_unref (inode);
+	}
+
+	QB_STACK_UNWIND (fsetattr, frame, op_ret, op_errno, pre, post, xdata);
+
+	return 0;
+}
+
+
+int
+qb_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *buf,
+	     int valid, dict_t *xdata)
+{
+	if (qb_inode_ctx_get (this, fd->inode))
+		frame->local = inode_ref (fd->inode);
+
+	STACK_WIND (frame, qb_setattr_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->fsetattr, fd, buf, valid, xdata);
+	return 0;
+}
+
+
+int
+qb_forget (xlator_t *this, inode_t *inode)
+{
+	return qb_inode_cleanup (this, inode, 0);
+}
+
+
+int
+qb_release (xlator_t *this, fd_t *fd)
+{
+	call_frame_t *frame = NULL;
+
+	frame = create_frame (this, this->ctx->pool);
+	if (!frame) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"Could not allocate frame. "
+			"Leaking QEMU BlockDriverState");
+		return -1;
+	}
+
+	if (qb_local_init (frame) != 0) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"Could not allocate local. "
+			"Leaking QEMU BlockDriverState");
+		STACK_DESTROY (frame->root);
+		return -1;
+	}
+
+	if (qb_coroutine (frame, qb_co_close) != 0) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"Could not allocate coroutine. "
+			"Leaking QEMU BlockDriverState");
+		qb_local_free (this, frame->local);
+		frame->local = NULL;
+		STACK_DESTROY (frame->root);
+	}
+
+	return 0;
+}
+
+int
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        ret = xlator_mem_acct_init (this, gf_qb_mt_end + 1);
+
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init "
+                        "failed");
+        return ret;
+}
+
+
+int
+reconfigure (xlator_t *this, dict_t *options)
+{
+	return 0;
+}
+
+
+int
+init (xlator_t *this)
+{
+        qb_conf_t *conf    = NULL;
+        int32_t    ret     = -1;
+	static int bdrv_inited = 0;
+
+        if (!this->children || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "FATAL: qemu-block (%s) not configured with exactly "
+                        "one child", this->name);
+                goto out;
+        }
+
+        conf = GF_CALLOC (1, sizeof (*conf), gf_qb_mt_qb_conf_t);
+        if (!conf)
+                goto out;
+
+        /* configure 'option window-size <size>' */
+        GF_OPTION_INIT ("default-password", conf->default_password, str, out);
+
+	/* qemu coroutines use "co_mutex" for synchronizing among themselves.
+	   However "co_mutex" itself is not threadsafe if the coroutine framework
+	   is multithreaded (which usually is not). However synctasks are
+	   fundamentally multithreaded, so for now create a syncenv which has
+	   scaling limits set to max 1 thread so that the qemu coroutines can
+	   execute "safely".
+
+	   Future work: provide an implementation of "co_mutex" which is
+	   threadsafe and use the global multithreaded ctx->env syncenv.
+	*/
+	conf->env = syncenv_new (0, 1, 1);
+
+        this->private = conf;
+
+        ret = 0;
+
+	snprintf (conf->qb_xattr_key, QB_XATTR_KEY_MAX, QB_XATTR_KEY_FMT,
+		  this->name);
+
+	cur_mon = (void *) 1;
+
+	if (!bdrv_inited) {
+		bdrv_init ();
+		bdrv_inited = 1;
+	}
+
+out:
+        if (ret)
+                GF_FREE (conf);
+
+        return ret;
+}
+
+
+void
+fini (xlator_t *this)
+{
+        qb_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        this->private = NULL;
+
+	if (conf->root_inode)
+		inode_unref(conf->root_inode);
+        GF_FREE (conf);
+
+	return;
+}
+
+
+struct xlator_fops fops = {
+	.lookup      = qb_lookup,
+	.fsetxattr   = qb_fsetxattr,
+	.setxattr    = qb_setxattr,
+	.open        = qb_open,
+        .writev      = qb_writev,
+        .readv       = qb_readv,
+        .fsync       = qb_fsync,
+        .truncate    = qb_truncate,
+        .ftruncate   = qb_ftruncate,
+        .stat        = qb_stat,
+        .fstat       = qb_fstat,
+	.setattr     = qb_setattr,
+	.fsetattr    = qb_fsetattr,
+        .flush       = qb_flush,
+/*
+	.getxattr    = qb_getxattr,
+	.fgetxattr   = qb_fgetxattr
+*/
+	.readdirp    = qb_readdirp,
+};
+
+
+struct xlator_cbks cbks = {
+        .forget   = qb_forget,
+	.release  = qb_release,
+};
+
+
+struct xlator_dumpops dumpops = {
+};
+
+
+struct volume_options options[] = {
+        { .key  = {"default-password"},
+          .type = GF_OPTION_TYPE_STR,
+          .default_value = "",
+          .description = "Default password for the AES encrypted block images."
+        },
+        { .key = {NULL} },
+};
diff --git a/xlators/features/qemu-block/src/qemu-block.h b/xlators/features/qemu-block/src/qemu-block.h
new file mode 100644
index 000000000..c95f2799a
--- /dev/null
+++ b/xlators/features/qemu-block/src/qemu-block.h
@@ -0,0 +1,109 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef __QEMU_BLOCK_H
+#define __QEMU_BLOCK_H
+
+#include "syncop.h"
+#include "call-stub.h"
+#include "block/block_int.h"
+#include "monitor/monitor.h"
+
+/* QB_XATTR_KEY_FMT is the on-disk xattr stored in the inode which
+   indicates that the file must be "interpreted" by the block format
+   logic. The value of the key is of the pattern:
+
+   "format:virtual_size"
+
+   e.g
+
+   "qcow2:20GB" or "qed:100GB"
+
+   The format and virtual size are colon separated. The format is
+   a case sensitive string which qemu recognizes. virtual_size is
+   specified as a size which glusterfs recognizes as size (i.e.,
+   value accepted by gf_string2bytesize())
+*/
+#define QB_XATTR_KEY_FMT "trusted.glusterfs.%s.format"
+
+#define QB_XATTR_KEY_MAX 64
+
+#define QB_XATTR_VAL_MAX 64
+
+
+typedef struct qb_inode {
+	char fmt[QB_XATTR_VAL_MAX]; /* this is only the format, not "format:size" */
+	size_t size; /* virtual size in bytes */
+	BlockDriverState *bs;
+	int refcnt;
+	uuid_t backing_gfid;
+	char *backing_fname;
+} qb_inode_t;
+
+
+typedef struct qb_conf {
+	Monitor *mon;
+	struct syncenv *env;
+	char qb_xattr_key[QB_XATTR_KEY_MAX];
+	char *default_password;
+	inode_t *root_inode;
+} qb_conf_t;
+
+
+typedef struct qb_local {
+	call_frame_t      *frame; /* backpointer */
+	call_stub_t       *stub;
+	inode_t           *inode;
+	fd_t              *fd;
+	char               fmt[QB_XATTR_VAL_MAX+1];
+	char               name[256];
+	synctask_fn_t      synctask_fn;
+	struct list_head   list;
+} qb_local_t;
+
+void qb_local_free (xlator_t *this, qb_local_t *local);
+int qb_coroutine (call_frame_t *frame, synctask_fn_t fn);
+inode_t *qb_inode_from_filename (const char *filename);
+int qb_inode_to_filename (inode_t *inode, char *filename, int size);
+int qb_format_extract (xlator_t *this, char *format, inode_t *inode);
+
+qb_inode_t *qb_inode_ctx_get (xlator_t *this, inode_t *inode);
+
+#define QB_STACK_UNWIND(typ, frame, args ...) do {	\
+	qb_local_t *__local = frame->local;		\
+	xlator_t *__this = frame->this;			\
+							\
+	frame->local = NULL;				\
+	STACK_UNWIND_STRICT (typ, frame, args);		\
+	if (__local)					\
+		qb_local_free (__this, __local);	\
+	} while (0)
+
+#define QB_STUB_UNWIND(stub, op_ret, op_errno) do {	\
+	qb_local_t *__local = stub->frame->local;	\
+	xlator_t *__this = stub->frame->this;		\
+							\
+	stub->frame->local = NULL;			\
+	call_unwind_error (stub, op_ret, op_errno);	\
+	if (__local)					\
+		qb_local_free (__this, __local);	\
+	} while (0)
+
+#define QB_STUB_RESUME(stub_errno) do {			\
+	qb_local_t *__local = stub->frame->local;	\
+	xlator_t *__this = stub->frame->this;		\
+							\
+	stub->frame->local = NULL;			\
+	call_resume (stub);				\
+	if (__local)					\
+		qb_local_free (__this, __local);	\
+	} while (0)
+
+#endif /* !__QEMU_BLOCK_H */
diff --git a/xlators/features/quiesce/Makefile.am b/xlators/features/quiesce/Makefile.am
new file mode 100644
index 000000000..a985f42a8
--- /dev/null
+++ b/xlators/features/quiesce/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+CLEANFILES =
diff --git a/xlators/features/quiesce/src/Makefile.am b/xlators/features/quiesce/src/Makefile.am
new file mode 100644
index 000000000..15e46629e
--- /dev/null
+++ b/xlators/features/quiesce/src/Makefile.am
@@ -0,0 +1,15 @@
+xlator_LTLIBRARIES = quiesce.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+quiesce_la_LDFLAGS = -module -avoid-version
+
+quiesce_la_SOURCES = quiesce.c
+quiesce_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+noinst_HEADERS = quiesce.h quiesce-mem-types.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
diff --git a/xlators/features/quiesce/src/quiesce-mem-types.h b/xlators/features/quiesce/src/quiesce-mem-types.h
new file mode 100644
index 000000000..6e582f424
--- /dev/null
+++ b/xlators/features/quiesce/src/quiesce-mem-types.h
@@ -0,0 +1,20 @@
+/*
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __QUIESCE_MEM_TYPES_H__
+#define __QUIESCE_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_quiesce_mem_types_ {
+        gf_quiesce_mt_priv_t = gf_common_mt_end + 1,
+        gf_quiesce_mt_end
+};
+#endif
diff --git a/xlators/features/quiesce/src/quiesce.c b/xlators/features/quiesce/src/quiesce.c
new file mode 100644
index 000000000..24c7dc6ed
--- /dev/null
+++ b/xlators/features/quiesce/src/quiesce.c
@@ -0,0 +1,2610 @@
+/*
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "quiesce.h"
+#include "defaults.h"
+#include "call-stub.h"
+
+/* TODO: */
+/* Think about 'writev/_*_lk/setattr/xattrop/' fops to do re-transmittion */
+
+
+/* Quiesce Specific Functions */
+void
+gf_quiesce_local_wipe (xlator_t *this, quiesce_local_t *local)
+{
+        if (!local || !this || !this->private)
+                return;
+
+        if (local->loc.inode)
+                loc_wipe (&local->loc);
+        if (local->fd)
+                fd_unref (local->fd);
+        GF_FREE (local->name);
+        GF_FREE (local->volname);
+        if (local->dict)
+                dict_unref (local->dict);
+        if (local->iobref)
+                iobref_unref (local->iobref);
+        GF_FREE (local->vector);
+
+        mem_put (local);
+}
+
+call_stub_t *
+gf_quiesce_dequeue (xlator_t *this)
+{
+        call_stub_t  *stub = NULL;
+        quiesce_priv_t *priv = NULL;
+
+        priv = this->private;
+
+        if (!priv || list_empty (&priv->req))
+                return NULL;
+
+        LOCK (&priv->lock);
+        {
+                stub = list_entry (priv->req.next, call_stub_t, list);
+                list_del_init (&stub->list);
+                priv->queue_size--;
+        }
+        UNLOCK (&priv->lock);
+
+        return stub;
+}
+
+void *
+gf_quiesce_dequeue_start (void *data)
+{
+        xlator_t       *this = NULL;
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        this = data;
+        priv = this->private;
+        THIS = this;
+
+        while (!list_empty (&priv->req)) {
+                stub = gf_quiesce_dequeue (this);
+                if (stub) {
+                        call_resume (stub);
+                }
+        }
+
+        return 0;
+}
+
+
+void
+gf_quiesce_timeout (void *data)
+{
+        xlator_t       *this = NULL;
+        quiesce_priv_t *priv = NULL;
+
+        this = data;
+        priv = this->private;
+        THIS = this;
+
+        LOCK (&priv->lock);
+        {
+                priv->pass_through = _gf_true;
+        }
+        UNLOCK (&priv->lock);
+
+        gf_quiesce_dequeue_start (this);
+
+        return;
+}
+
+void
+gf_quiesce_enqueue (xlator_t *this, call_stub_t *stub)
+{
+        quiesce_priv_t *priv    = NULL;
+        struct timespec timeout = {0,};
+
+        priv = this->private;
+        if (!priv) {
+                gf_log_callingfn (this->name, GF_LOG_ERROR,
+                                  "this->private == NULL");
+                return;
+        }
+
+        LOCK (&priv->lock);
+        {
+                list_add_tail (&stub->list, &priv->req);
+                priv->queue_size++;
+        }
+        UNLOCK (&priv->lock);
+
+        if (!priv->timer) {
+                timeout.tv_sec = 20;
+                timeout.tv_nsec = 0;
+
+                priv->timer = gf_timer_call_after (this->ctx,
+                                                   timeout,
+                                                   gf_quiesce_timeout,
+                                                   (void *) this);
+        }
+
+        return;
+}
+
+
+
+/* _CBK function section */
+
+int32_t
+quiesce_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, inode_t *inode,
+                    struct iatt *buf, dict_t *dict, struct iatt *postparent)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_lookup_stub (frame, default_lookup_resume,
+                                        &local->loc, local->dict);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (lookup, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
+                             dict, postparent);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                  dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_stat_stub (frame, default_stat_resume,
+                                      &local->loc, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (stat, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_access_stub (frame, default_access_resume,
+                                        &local->loc, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (access, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (access, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, const char *path,
+                      struct iatt *buf, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_readlink_stub (frame, default_readlink_resume,
+                                          &local->loc, local->size, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (readlink, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path, buf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_open_stub (frame, default_open_resume,
+                                      &local->loc, local->flag, local->fd,
+                                      xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (open, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iovec *vector,
+                   int32_t count, struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_readv_stub (frame, default_readv_resume,
+                                       local->fd, local->size, local->offset,
+                                       local->io_flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (readv, frame, -1, ENOMEM,
+                                             NULL, 0, NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector, count,
+                             stbuf, iobref, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_flush_stub (frame, default_flush_resume,
+                                       local->fd, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (flush, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+
+int32_t
+quiesce_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                   struct iatt *postbuf, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fsync_stub (frame, default_fsync_resume,
+                                       local->fd, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fsync, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fstat_stub (frame, default_fstat_resume,
+                                       local->fd, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fstat, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_opendir_stub (frame, default_opendir_resume,
+                                         &local->loc, local->fd, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (opendir, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fsyncdir_stub (frame, default_fsyncdir_resume,
+                                          local->fd, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fsyncdir, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct statvfs *buf, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_statfs_stub (frame, default_statfs_resume,
+                                        &local->loc, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (statfs, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fgetxattr_stub (frame, default_fgetxattr_resume,
+                                           local->fd, local->name, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fgetxattr, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_getxattr_stub (frame, default_getxattr_resume,
+                                          &local->loc, local->name, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_rchecksum_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, uint32_t weak_checksum,
+                       uint8_t *strong_checksum, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_rchecksum_stub (frame, default_rchecksum_resume,
+                                           local->fd, local->offset, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (rchecksum, frame, -1, ENOMEM,
+                                             0, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (rchecksum, frame, op_ret, op_errno, weak_checksum,
+                             strong_checksum, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, gf_dirent_t *entries, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_readdir_stub (frame, default_readdir_resume,
+                                         local->fd, local->size, local->offset, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (readdir, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, entries, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, gf_dirent_t *entries, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_readdirp_stub (frame, default_readdirp_resume,
+                                          local->fd, local->size, local->offset,
+                                          local->dict);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (readdirp, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+#if 0
+
+int32_t
+quiesce_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                    struct iatt *postbuf, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_writev_stub (frame, default_writev_resume,
+                                        local->fd, local->vector, local->flag,
+                                        local->offset, local->io_flags,
+                                        local->iobref, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (writev, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_xattrop_stub (frame, default_xattrop_resume,
+                                         &local->loc, local->xattrop_flags,
+                                         local->dict, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (xattrop, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (xattrop, frame, op_ret, op_errno, dict, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fxattrop_stub (frame, default_fxattrop_resume,
+                                          local->fd, local->xattrop_flags,
+                                          local->dict, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fxattrop, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fxattrop, frame, op_ret, op_errno, dict, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_lk_stub (frame, default_lk_resume,
+                                    local->fd, local->flag, &local->flock, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (lk, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_inodelk_stub (frame, default_inodelk_resume,
+                                         local->volname, &local->loc,
+                                         local->flag, &local->flock, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (inodelk, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_finodelk_stub (frame, default_finodelk_resume,
+                                         local->volname, local->fd,
+                                         local->flag, &local->flock, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (finodelk, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (finodelk, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_entrylk_stub (frame, default_entrylk_resume,
+                                         local->volname, &local->loc,
+                                         local->name, local->cmd, local->type, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (entrylk, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fentrylk_stub (frame, default_fentrylk_resume,
+                                          local->volname, local->fd,
+                                          local->name, local->cmd, local->type, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fentrylk, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fentrylk, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                     struct iatt *statpost, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_setattr_stub (frame, default_setattr_resume,
+                                         &local->loc, &local->stbuf, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (setattr, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, statpre,
+                             statpost, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                      struct iatt *statpost, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fsetattr_stub (frame, default_fsetattr_resume,
+                                          local->fd, &local->stbuf, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fsetattr, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fsetattr, frame, op_ret, op_errno, statpre,
+                             statpost, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+#endif /* if 0 */
+
+
+/* FOP */
+
+/* No retransmittion */
+
+int32_t
+quiesce_removexattr (call_frame_t *frame,
+		     xlator_t *this,
+		     loc_t *loc,
+		     const char *name, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_removexattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->removexattr,
+                            loc,
+                            name, xdata);
+                return 0;
+        }
+
+        stub = fop_removexattr_stub (frame, default_removexattr_resume,
+                                     loc, name, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (removexattr, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_truncate (call_frame_t *frame,
+		  xlator_t *this,
+		  loc_t *loc,
+		  off_t offset, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_truncate_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->truncate,
+                            loc,
+                            offset, xdata);
+                return 0;
+        }
+
+        stub = fop_truncate_stub (frame, default_truncate_resume, loc, offset, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_fsetxattr (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t *fd,
+                   dict_t *dict,
+                   int32_t flags, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_fsetxattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->fsetxattr,
+                            fd,
+                            dict,
+                            flags, xdata);
+                return 0;
+        }
+
+        stub = fop_fsetxattr_stub (frame, default_fsetxattr_resume,
+                                   fd, dict, flags, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (fsetxattr, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_setxattr (call_frame_t *frame,
+		  xlator_t *this,
+		  loc_t *loc,
+		  dict_t *dict,
+		  int32_t flags, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_setxattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->setxattr,
+                            loc,
+                            dict,
+                            flags, xdata);
+                return 0;
+        }
+
+        stub = fop_setxattr_stub (frame, default_setxattr_resume,
+                                  loc, dict, flags, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (setxattr, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_create (call_frame_t *frame, xlator_t *this,
+		loc_t *loc, int32_t flags, mode_t mode,
+                mode_t umask, fd_t *fd, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                /* Don't send O_APPEND below, as write() re-transmittions can
+                   fail with O_APPEND */
+                STACK_WIND (frame, default_create_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->create,
+                            loc, (flags & ~O_APPEND), mode, umask, fd, xdata);
+                return 0;
+        }
+
+        stub = fop_create_stub (frame, default_create_resume,
+                                loc, (flags & ~O_APPEND), mode, umask, fd, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (create, frame, -1, ENOMEM,
+                                     NULL, NULL, NULL, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_link (call_frame_t *frame,
+	      xlator_t *this,
+	      loc_t *oldloc,
+	      loc_t *newloc, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_link_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->link,
+                            oldloc, newloc, xdata);
+                return 0;
+        }
+
+        stub = fop_link_stub (frame, default_link_resume, oldloc, newloc, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (link, frame, -1, ENOMEM,
+                                     NULL, NULL, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_rename (call_frame_t *frame,
+		xlator_t *this,
+		loc_t *oldloc,
+		loc_t *newloc, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_rename_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->rename,
+                            oldloc, newloc, xdata);
+                return 0;
+        }
+
+        stub = fop_rename_stub (frame, default_rename_resume, oldloc, newloc, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (rename, frame, -1, ENOMEM,
+                                     NULL, NULL, NULL, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+
+int
+quiesce_symlink (call_frame_t *frame, xlator_t *this,
+		 const char *linkpath, loc_t *loc, mode_t umask, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame, default_symlink_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->symlink,
+                            linkpath, loc, umask, xdata);
+                return 0;
+        }
+
+        stub = fop_symlink_stub (frame, default_symlink_resume,
+                                 linkpath, loc, umask, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (symlink, frame, -1, ENOMEM,
+                                     NULL, NULL, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+
+int
+quiesce_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame, default_rmdir_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->rmdir,
+                            loc, flags, xdata);
+                return 0;
+        }
+
+        stub = fop_rmdir_stub (frame, default_rmdir_resume, loc, flags, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (rmdir, frame, -1, ENOMEM, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_unlink (call_frame_t *frame,
+		xlator_t *this,
+		loc_t *loc, int xflag, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_unlink_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->unlink,
+                            loc, xflag, xdata);
+                return 0;
+        }
+
+        stub = fop_unlink_stub (frame, default_unlink_resume, loc, xflag, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (unlink, frame, -1, ENOMEM, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int
+quiesce_mkdir (call_frame_t *frame, xlator_t *this,
+	       loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame, default_mkdir_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->mkdir,
+                            loc, mode, umask, xdata);
+                return 0;
+        }
+
+        stub = fop_mkdir_stub (frame, default_mkdir_resume,
+                               loc, mode, umask, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (mkdir, frame, -1, ENOMEM,
+                                     NULL, NULL, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+
+int
+quiesce_mknod (call_frame_t *frame, xlator_t *this,
+	       loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame, default_mknod_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->mknod,
+                            loc, mode, rdev, umask, xdata);
+                return 0;
+        }
+
+        stub = fop_mknod_stub (frame, default_mknod_resume,
+                               loc, mode, rdev, umask, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (mknod, frame, -1, ENOMEM,
+                                     NULL, NULL, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_ftruncate (call_frame_t *frame,
+		   xlator_t *this,
+		   fd_t *fd,
+		   off_t offset, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_ftruncate_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->ftruncate,
+                            fd,
+                            offset, xdata);
+                return 0;
+        }
+
+        stub = fop_ftruncate_stub (frame, default_ftruncate_resume, fd, offset, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+/* Re-transmittion */
+
+int32_t
+quiesce_readlink (call_frame_t *frame,
+		  xlator_t *this,
+		  loc_t *loc,
+		  size_t size, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                loc_dup (loc, &local->loc);
+                local->size = size;
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_readlink_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->readlink,
+                            loc,
+                            size, xdata);
+                return 0;
+        }
+
+        stub = fop_readlink_stub (frame, default_readlink_resume, loc, size, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (readlink, frame, -1, ENOMEM, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_access (call_frame_t *frame,
+		xlator_t *this,
+		loc_t *loc,
+		int32_t mask, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                loc_dup (loc, &local->loc);
+                local->flag = mask;
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_access_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->access,
+                            loc,
+                            mask, xdata);
+                return 0;
+        }
+
+        stub = fop_access_stub (frame, default_access_resume, loc, mask, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (access, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_fgetxattr (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t *fd,
+                   const char *name, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                local->fd = fd_ref (fd);
+                if (name)
+                        local->name = gf_strdup (name);
+
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_fgetxattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->fgetxattr,
+                            fd,
+                            name, xdata);
+                return 0;
+        }
+
+        stub = fop_fgetxattr_stub (frame, default_fgetxattr_resume, fd, name, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (fgetxattr, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_statfs (call_frame_t *frame,
+		xlator_t *this,
+		loc_t *loc, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                loc_dup (loc, &local->loc);
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_statfs_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->statfs,
+                            loc, xdata);
+                return 0;
+        }
+
+        stub = fop_statfs_stub (frame, default_statfs_resume, loc, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (statfs, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_fsyncdir (call_frame_t *frame,
+		  xlator_t *this,
+		  fd_t *fd,
+		  int32_t flags, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                local->fd = fd_ref (fd);
+                local->flag = flags;
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_fsyncdir_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->fsyncdir,
+                            fd,
+                            flags, xdata);
+                return 0;
+        }
+
+        stub = fop_fsyncdir_stub (frame, default_fsyncdir_resume, fd, flags, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (fsyncdir, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_opendir (call_frame_t *frame,
+		 xlator_t *this,
+		 loc_t *loc, fd_t *fd, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                loc_dup (loc, &local->loc);
+                local->fd = fd_ref (fd);
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_opendir_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->opendir,
+                            loc, fd, xdata);
+                return 0;
+        }
+
+        stub = fop_opendir_stub (frame, default_opendir_resume, loc, fd, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (opendir, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_fstat (call_frame_t *frame,
+	       xlator_t *this,
+	       fd_t *fd, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                local->fd = fd_ref (fd);
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_fstat_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->fstat,
+                            fd, xdata);
+                return 0;
+        }
+
+        stub = fop_fstat_stub (frame, default_fstat_resume, fd, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (fstat, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_fsync (call_frame_t *frame,
+	       xlator_t *this,
+	       fd_t *fd,
+	       int32_t flags, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                local->fd = fd_ref (fd);
+                local->flag = flags;
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_fsync_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->fsync,
+                            fd,
+                            flags, xdata);
+                return 0;
+        }
+
+        stub = fop_fsync_stub (frame, default_fsync_resume, fd, flags, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (fsync, frame, -1, ENOMEM, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_flush (call_frame_t *frame,
+	       xlator_t *this,
+	       fd_t *fd, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                local->fd = fd_ref (fd);
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_flush_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->flush,
+                            fd, xdata);
+                return 0;
+        }
+
+        stub = fop_flush_stub (frame, default_flush_resume, fd, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (flush, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_writev (call_frame_t *frame,
+		xlator_t *this,
+		fd_t *fd,
+		struct iovec *vector,
+		int32_t count,
+		off_t off, uint32_t flags,
+                struct iobref *iobref, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                STACK_WIND (frame,
+                            default_writev_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->writev,
+                            fd,
+                            vector,
+                            count,
+                            off, flags,
+                            iobref, xdata);
+                return 0;
+        }
+
+        stub = fop_writev_stub (frame, default_writev_resume,
+                                fd, vector, count, off, flags, iobref, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (writev, frame, -1, ENOMEM, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_readv (call_frame_t *frame,
+	       xlator_t *this,
+	       fd_t *fd,
+	       size_t size,
+	       off_t offset, uint32_t flags, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                local->fd = fd_ref (fd);
+                local->size = size;
+                local->offset = offset;
+                local->io_flag = flags;
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_readv_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->readv,
+                            fd,
+                            size,
+                            offset, flags, xdata);
+                return 0;
+        }
+
+        stub = fop_readv_stub (frame, default_readv_resume, fd, size, offset,
+                               flags, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (readv, frame, -1, ENOMEM,
+                                     NULL, 0, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_open (call_frame_t *frame,
+	      xlator_t *this,
+	      loc_t *loc,
+	      int32_t flags, fd_t *fd,
+              dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                loc_dup (loc, &local->loc);
+                local->fd = fd_ref (fd);
+
+                /* Don't send O_APPEND below, as write() re-transmittions can
+                   fail with O_APPEND */
+                local->flag = (flags & ~O_APPEND);
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_open_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->open,
+                            loc, (flags & ~O_APPEND), fd, xdata);
+                return 0;
+        }
+
+        stub = fop_open_stub (frame, default_open_resume, loc,
+                              (flags & ~O_APPEND), fd, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (open, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_getxattr (call_frame_t *frame,
+		  xlator_t *this,
+		  loc_t *loc,
+		  const char *name, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                loc_dup (loc, &local->loc);
+                if (name)
+                        local->name = gf_strdup (name);
+
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_getxattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->getxattr,
+                            loc,
+                            name, xdata);
+                return 0;
+        }
+
+        stub = fop_getxattr_stub (frame, default_getxattr_resume, loc, name, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_xattrop (call_frame_t *frame,
+		 xlator_t *this,
+		 loc_t *loc,
+		 gf_xattrop_flags_t flags,
+		 dict_t *dict, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                STACK_WIND (frame,
+                            default_xattrop_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->xattrop,
+                            loc,
+                            flags,
+                            dict, xdata);
+                return 0;
+        }
+
+        stub = fop_xattrop_stub (frame, default_xattrop_resume,
+                                 loc, flags, dict, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (xattrop, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_fxattrop (call_frame_t *frame,
+		  xlator_t *this,
+		  fd_t *fd,
+		  gf_xattrop_flags_t flags,
+		  dict_t *dict, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                STACK_WIND (frame,
+                            default_fxattrop_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->fxattrop,
+                            fd,
+                            flags,
+                            dict, xdata);
+                return 0;
+        }
+
+        stub = fop_fxattrop_stub (frame, default_fxattrop_resume,
+                                  fd, flags, dict, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (fxattrop, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_lk (call_frame_t *frame,
+	    xlator_t *this,
+	    fd_t *fd,
+	    int32_t cmd,
+	    struct gf_flock *lock, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                STACK_WIND (frame,
+                            default_lk_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->lk,
+                            fd,
+                            cmd,
+                            lock, xdata);
+                return 0;
+        }
+
+        stub = fop_lk_stub (frame, default_lk_resume, fd, cmd, lock, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (lk, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_inodelk (call_frame_t *frame, xlator_t *this,
+		 const char *volume, loc_t *loc, int32_t cmd,
+                 struct gf_flock *lock, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                STACK_WIND (frame,
+                            default_inodelk_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->inodelk,
+                            volume, loc, cmd, lock, xdata);
+                return 0;
+        }
+
+        stub = fop_inodelk_stub (frame, default_inodelk_resume,
+                                 volume, loc, cmd, lock, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (inodelk, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_finodelk (call_frame_t *frame, xlator_t *this,
+		  const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                STACK_WIND (frame,
+                            default_finodelk_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->finodelk,
+                            volume, fd, cmd, lock, xdata);
+                return 0;
+        }
+
+        stub = fop_finodelk_stub (frame, default_finodelk_resume,
+                                  volume, fd, cmd, lock, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (finodelk, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_entrylk (call_frame_t *frame, xlator_t *this,
+		 const char *volume, loc_t *loc, const char *basename,
+		 entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                STACK_WIND (frame, default_entrylk_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->entrylk,
+                            volume, loc, basename, cmd, type, xdata);
+                return 0;
+        }
+
+        stub = fop_entrylk_stub (frame, default_entrylk_resume,
+                                 volume, loc, basename, cmd, type, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (entrylk, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_fentrylk (call_frame_t *frame, xlator_t *this,
+		  const char *volume, fd_t *fd, const char *basename,
+		  entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                STACK_WIND (frame, default_fentrylk_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->fentrylk,
+                            volume, fd, basename, cmd, type, xdata);
+                return 0;
+        }
+
+        stub = fop_fentrylk_stub (frame, default_fentrylk_resume,
+                                  volume, fd, basename, cmd, type, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (fentrylk, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_rchecksum (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t *fd, off_t offset,
+                   int32_t len, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                local->fd = fd_ref (fd);
+                local->offset = offset;
+                local->flag = len;
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_rchecksum_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->rchecksum,
+                            fd, offset, len, xdata);
+                return 0;
+        }
+
+        stub = fop_rchecksum_stub (frame, default_rchecksum_resume,
+                                   fd, offset, len, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (rchecksum, frame, -1, ENOMEM, 0, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_readdir (call_frame_t *frame,
+		 xlator_t *this,
+		 fd_t *fd,
+		 size_t size,
+		 off_t off, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                local->fd = fd_ref (fd);
+                local->size = size;
+                local->offset = off;
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_readdir_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->readdir,
+                            fd, size, off, xdata);
+                return 0;
+        }
+
+        stub = fop_readdir_stub (frame, default_readdir_resume, fd, size, off, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (readdir, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_readdirp (call_frame_t *frame,
+		  xlator_t *this,
+		  fd_t *fd,
+		  size_t size,
+		  off_t off, dict_t *dict)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                local->fd = fd_ref (fd);
+                local->size = size;
+                local->offset = off;
+                local->dict = dict_ref (dict);
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_readdirp_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->readdirp,
+                            fd, size, off, dict);
+                return 0;
+        }
+
+        stub = fop_readdirp_stub (frame, default_readdirp_resume, fd, size,
+                                  off, dict);
+        if (!stub) {
+                STACK_UNWIND_STRICT (readdirp, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_setattr (call_frame_t *frame,
+                 xlator_t *this,
+                 loc_t *loc,
+                 struct iatt *stbuf,
+                 int32_t valid, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                STACK_WIND (frame,
+                            default_setattr_cbk,
+                            FIRST_CHILD (this),
+                            FIRST_CHILD (this)->fops->setattr,
+                            loc, stbuf, valid, xdata);
+                return 0;
+        }
+
+        stub = fop_setattr_stub (frame, default_setattr_resume,
+                                   loc, stbuf, valid, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (setattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_stat (call_frame_t *frame,
+	      xlator_t *this,
+	      loc_t *loc, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                loc_dup (loc, &local->loc);
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_stat_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->stat,
+                            loc, xdata);
+                return 0;
+        }
+
+        stub = fop_stat_stub (frame, default_stat_resume, loc, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (stat, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_lookup (call_frame_t *frame,
+		xlator_t *this,
+		loc_t *loc,
+		dict_t *xattr_req)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                local = mem_get0 (priv->local_pool);
+                loc_dup (loc, &local->loc);
+                local->dict = dict_ref (xattr_req);
+                frame->local = local;
+
+                STACK_WIND (frame,
+                            quiesce_lookup_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->lookup,
+                            loc, xattr_req);
+                return 0;
+        }
+
+        stub = fop_lookup_stub (frame, default_lookup_resume, loc, xattr_req);
+        if (!stub) {
+                STACK_UNWIND_STRICT (lookup, frame, -1, ENOMEM,
+                                     NULL, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_fsetattr (call_frame_t *frame,
+                  xlator_t *this,
+                  fd_t *fd,
+                  struct iatt *stbuf,
+                  int32_t valid, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv && priv->pass_through) {
+                STACK_WIND (frame,
+                            default_fsetattr_cbk,
+                            FIRST_CHILD (this),
+                            FIRST_CHILD (this)->fops->fsetattr,
+                            fd, stbuf, valid, xdata);
+                return 0;
+        }
+
+        stub = fop_fsetattr_stub (frame, default_fsetattr_resume,
+                                  fd, stbuf, valid, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (fsetattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        ret = xlator_mem_acct_init (this, gf_quiesce_mt_end + 1);
+
+        return ret;
+}
+
+int
+init (xlator_t *this)
+{
+        int ret = -1;
+        quiesce_priv_t *priv = NULL;
+
+	if (!this->children || this->children->next) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"'quiesce' not configured with exactly one child");
+                goto out;
+	}
+
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+
+        priv = GF_CALLOC (1, sizeof (*priv), gf_quiesce_mt_priv_t);
+        if (!priv)
+                goto out;
+
+        priv->local_pool =  mem_pool_new (quiesce_local_t,
+                                          GF_FOPS_EXPECTED_IN_PARALLEL);
+
+        LOCK_INIT (&priv->lock);
+        priv->pass_through = _gf_false;
+
+        INIT_LIST_HEAD (&priv->req);
+
+        this->private = priv;
+        ret = 0;
+out:
+        return ret;
+}
+
+void
+fini (xlator_t *this)
+{
+        quiesce_priv_t *priv = NULL;
+
+        priv = this->private;
+        if (!priv)
+                goto out;
+        this->private = NULL;
+
+        mem_pool_destroy (priv->local_pool);
+        LOCK_DESTROY (&priv->lock);
+        GF_FREE (priv);
+out:
+        return;
+}
+
+int
+notify (xlator_t *this, int event, void *data, ...)
+{
+        int             ret     = 0;
+        quiesce_priv_t *priv    = NULL;
+        struct timespec timeout = {0,};
+
+        priv = this->private;
+        if (!priv)
+                goto out;
+
+        switch (event) {
+        case GF_EVENT_CHILD_UP:
+        {
+                ret = pthread_create (&priv->thr, NULL, gf_quiesce_dequeue_start,
+                                      this);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to create the quiesce-dequeue thread");
+                }
+
+                LOCK (&priv->lock);
+                {
+                        priv->pass_through = _gf_true;
+                }
+                UNLOCK (&priv->lock);
+                break;
+        }
+        case GF_EVENT_CHILD_DOWN:
+                LOCK (&priv->lock);
+                {
+                        priv->pass_through = _gf_false;
+                }
+                UNLOCK (&priv->lock);
+
+                if (priv->timer)
+                        break;
+                timeout.tv_sec = 20;
+                timeout.tv_nsec = 0;
+
+                priv->timer = gf_timer_call_after (this->ctx,
+                                                   timeout,
+                                                   gf_quiesce_timeout,
+                                                   (void *) this);
+
+                if (priv->timer == NULL) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Cannot create timer");
+                }
+
+                break;
+        default:
+                break;
+        }
+
+        ret = default_notify (this, event, data);
+out:
+        return ret;
+}
+
+
+struct xlator_fops fops = {
+        /* write/modifying fops */
+	.mknod       = quiesce_mknod,
+	.create      = quiesce_create,
+	.truncate    = quiesce_truncate,
+	.ftruncate   = quiesce_ftruncate,
+	.setxattr    = quiesce_setxattr,
+	.removexattr = quiesce_removexattr,
+	.symlink     = quiesce_symlink,
+	.unlink      = quiesce_unlink,
+	.link        = quiesce_link,
+	.mkdir       = quiesce_mkdir,
+	.rmdir       = quiesce_rmdir,
+	.rename      = quiesce_rename,
+
+        /* The below calls are known to change state, hence
+           re-transmittion is not advised */
+	.lk          = quiesce_lk,
+	.inodelk     = quiesce_inodelk,
+	.finodelk    = quiesce_finodelk,
+	.entrylk     = quiesce_entrylk,
+	.fentrylk    = quiesce_fentrylk,
+	.xattrop     = quiesce_xattrop,
+	.fxattrop    = quiesce_fxattrop,
+        .setattr     = quiesce_setattr,
+        .fsetattr    = quiesce_fsetattr,
+
+        /* Special case, re-transmittion is not harmful *
+         * as offset is properly sent from above layers */
+        /* TODO: not re-transmitted as of now */
+	.writev      = quiesce_writev,
+
+        /* re-transmittable fops */
+	.lookup      = quiesce_lookup,
+	.stat        = quiesce_stat,
+	.fstat       = quiesce_fstat,
+	.access      = quiesce_access,
+	.readlink    = quiesce_readlink,
+	.getxattr    = quiesce_getxattr,
+	.open        = quiesce_open,
+	.readv       = quiesce_readv,
+	.flush       = quiesce_flush,
+	.fsync       = quiesce_fsync,
+	.statfs      = quiesce_statfs,
+	.opendir     = quiesce_opendir,
+	.readdir     = quiesce_readdir,
+	.readdirp    = quiesce_readdirp,
+	.fsyncdir    = quiesce_fsyncdir,
+
+};
+
+struct xlator_dumpops dumpops;
+
+
+struct xlator_cbks cbks;
+
+
+struct volume_options options[] = {
+	{ .key  = {NULL} },
+};
diff --git a/xlators/features/quiesce/src/quiesce.h b/xlators/features/quiesce/src/quiesce.h
new file mode 100644
index 000000000..878ed77e9
--- /dev/null
+++ b/xlators/features/quiesce/src/quiesce.h
@@ -0,0 +1,51 @@
+/*
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __QUIESCE_H__
+#define __QUIESCE_H__
+
+#include "quiesce-mem-types.h"
+#include "xlator.h"
+#include "timer.h"
+
+#define GF_FOPS_EXPECTED_IN_PARALLEL 512
+
+typedef struct {
+        gf_timer_t       *timer;
+        gf_boolean_t      pass_through;
+        gf_lock_t         lock;
+        struct list_head  req;
+        int               queue_size;
+        pthread_t         thr;
+        struct mem_pool  *local_pool;
+} quiesce_priv_t;
+
+typedef struct {
+        fd_t               *fd;
+        char               *name;
+        char               *volname;
+        loc_t               loc;
+        off_t               size;
+        off_t               offset;
+        mode_t              mode;
+        int32_t             flag;
+        struct iatt         stbuf;
+        struct iovec       *vector;
+        struct iobref      *iobref;
+        dict_t             *dict;
+        struct gf_flock     flock;
+        entrylk_cmd         cmd;
+        entrylk_type        type;
+        gf_xattrop_flags_t  xattrop_flags;
+        int32_t             wbflags;
+        uint32_t            io_flag;
+} quiesce_local_t;
+
+#endif
diff --git a/xlators/features/quota/src/Makefile.am b/xlators/features/quota/src/Makefile.am
index fe373c8b5..9546f4276 100644
--- a/xlators/features/quota/src/Makefile.am
+++ b/xlators/features/quota/src/Makefile.am
@@ -1,15 +1,17 @@
 xlator_LTLIBRARIES = quota.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
 
-quota_la_LDFLAGS = -module -avoidversion
+quota_la_LDFLAGS = -module -avoid-version
 
 quota_la_SOURCES = quota.c
 quota_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = quota-mem-types.h
+noinst_HEADERS = quota-mem-types.h quota.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+	-I$(top_srcdir)/xlators/cluster/dht/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/features/quota/src/quota-mem-types.h b/xlators/features/quota/src/quota-mem-types.h
index b71314ed8..3082865da 100644
--- a/xlators/features/quota/src/quota-mem-types.h
+++ b/xlators/features/quota/src/quota-mem-types.h
@@ -1,30 +1,26 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __QUOTA_MEM_TYPES_H__
 #define __QUOTA_MEM_TYPES_H__
 
 #include "mem-types.h"
 
 enum gf_quota_mem_types_ {
-        gf_quota_mt_quota_local = gf_common_mt_end + 1,
-        gf_quota_mt_quota_priv,
+        gf_quota_mt_quota_priv_t = gf_common_mt_end + 1,
+        gf_quota_mt_quota_inode_ctx_t,
+        gf_quota_mt_loc_t,
+        gf_quota_mt_char,
+        gf_quota_mt_int64_t,
+        gf_quota_mt_int32_t,
+        gf_quota_mt_limits_t,
+        gf_quota_mt_quota_dentry_t,
         gf_quota_mt_end
 };
 #endif
diff --git a/xlators/features/quota/src/quota.c b/xlators/features/quota/src/quota.c
index 094b62d4a..c527e7ca7 100644
--- a/xlators/features/quota/src/quota.c
+++ b/xlators/features/quota/src/quota.c
@@ -1,1028 +1,3146 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
+#include <fnmatch.h>
 
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
+#include "quota.h"
+#include "common-utils.h"
+#include "defaults.h"
 
-#include <sys/time.h>
+int32_t
+quota_check_limit (call_frame_t *frame, inode_t *inode, xlator_t *this,
+                   char *name, uuid_t par);
+struct volume_options options[];
 
-#include "xlator.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "quota-mem-types.h"
-
-#ifndef MAX_IOVEC
-#define MAX_IOVEC 16
-#endif
-
-struct quota_local {
-	struct iatt    stbuf;
-	inode_t       *inode;
-	char          *path;
-	fd_t          *fd;
-	off_t          offset;
-	int32_t        count;
-	struct iovec   vector[MAX_IOVEC];
-	struct iobref *iobref;
-	loc_t          loc;
-};
+int
+quota_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path)
+{
+        int ret = -1;
 
+        if (!loc) {
+                return ret;
+        }
 
-struct quota_priv {
-	char       only_first_time;          /* Used to make sure a call is done only one time */
-	gf_lock_t  lock;                     /* Used while updating variables */
+        if (inode) {
+                loc->inode = inode_ref (inode);
+        }
 
-	uint64_t   disk_usage_limit;         /* Used for Disk usage quota */
-	uint64_t   current_disk_usage;       /* Keep the current usage value */
+        if (parent) {
+                loc->parent = inode_ref (parent);
+        }
 
-	uint32_t   min_free_disk_limit;        /* user specified limit, in %*/
-	uint32_t   current_free_disk;          /* current free disk space available, in % */
-	uint32_t   refresh_interval;           /* interval in seconds */
-	uint32_t   min_disk_last_updated_time; /* used for interval calculation */	
+        loc->path = gf_strdup (path);
+        if (!loc->path) {
+                goto loc_wipe;
+        }
 
-	loc_t      root_loc;		     /* Store '/' loc_t to make xattr calls */
-};
+        loc->name = strrchr (loc->path, '/');
+        if (loc->name) {
+                loc->name++;
+        } else {
+                goto loc_wipe;
+        }
+
+        ret = 0;
+
+loc_wipe:
+        if (ret < 0) {
+                loc_wipe (loc);
+        }
+
+        return ret;
+}
 
 
 int
-quota_statvfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno, struct statvfs *stbuf)
-{
-	struct quota_priv *priv = this->private;
-	
-	if (op_ret >= 0) {
-		priv->current_free_disk =
-			(stbuf->f_bavail * 100) / stbuf->f_blocks;
-	}
+quota_inode_loc_fill (inode_t *inode, loc_t *loc)
+{
+        char            *resolvedpath = NULL;
+        inode_t         *parent       = NULL;
+        int              ret          = -1;
+        xlator_t        *this         = NULL;
+
+        if ((!inode) || (!loc)) {
+                return ret;
+        }
+
+        this = THIS;
+
+        if ((inode) && __is_root_gfid (inode->gfid)) {
+                loc->parent = NULL;
+                goto ignore_parent;
+        }
 
-	STACK_DESTROY (frame->root);
-	return 0;
+        parent = inode_parent (inode, 0, NULL);
+        if (!parent) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "cannot find parent for inode (gfid:%s)",
+                        uuid_utoa (inode->gfid));
+                goto err;
+        }
+
+ignore_parent:
+        ret = inode_path (inode, NULL, &resolvedpath);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "cannot construct path for inode (gfid:%s)",
+                        uuid_utoa (inode->gfid));
+                goto err;
+        }
+
+        ret = quota_loc_fill (loc, inode, parent, resolvedpath);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "cannot fill loc");
+                goto err;
+        }
+
+err:
+        if (parent) {
+                inode_unref (parent);
+        }
+
+        GF_FREE (resolvedpath);
+
+        return ret;
 }
 
 
-void
-gf_quota_usage_subtract (xlator_t *this, size_t size)
+int32_t
+quota_local_cleanup (xlator_t *this, quota_local_t *local)
 {
-	struct quota_priv *priv = NULL;
+        if (local == NULL) {
+                goto out;
+        }
 
-	priv = this->private;
+        loc_wipe (&local->loc);
+        loc_wipe (&local->newloc);
+        loc_wipe (&local->oldloc);
+        loc_wipe (&local->validate_loc);
 
-	LOCK (&priv->lock);
-	{
-		if (priv->current_disk_usage < size)
-			priv->current_disk_usage = 0;
-		else
-			priv->current_disk_usage -= size;
-	}
-	UNLOCK (&priv->lock);
+        inode_unref (local->inode);
+        LOCK_DESTROY (&local->lock);
+
+        mem_put (local);
+out:
+        return 0;
+}
+
+
+static inline quota_local_t *
+quota_local_new ()
+{
+        quota_local_t *local = NULL;
+        local = mem_get0 (THIS->local_pool);
+        if (local)
+                LOCK_INIT (&local->lock);
+        return local;
+}
+
+
+quota_dentry_t *
+__quota_dentry_new (quota_inode_ctx_t *ctx, char *name, uuid_t par)
+{
+        quota_dentry_t    *dentry = NULL;
+        GF_UNUSED int32_t  ret    = 0;
+
+        QUOTA_ALLOC_OR_GOTO (dentry, quota_dentry_t, err);
+
+        INIT_LIST_HEAD (&dentry->next);
+
+        dentry->name = gf_strdup (name);
+        if (dentry->name == NULL) {
+                GF_FREE (dentry);
+                goto err;
+        }
+
+        uuid_copy (dentry->par, par);
+
+        list_add_tail (&dentry->next, &ctx->parents);
+err:
+        return dentry;
 }
 
 
 void
-gf_quota_usage_add (xlator_t *this, size_t size)
+__quota_dentry_free (quota_dentry_t *dentry)
 {
-	struct quota_priv *priv = this->private;
+        if (dentry == NULL) {
+                goto out;
+        }
 
-	LOCK (&priv->lock);
-	{
-		priv->current_disk_usage += size;
-	}
-	UNLOCK (&priv->lock);
+        list_del_init (&dentry->next);
+
+        GF_FREE (dentry->name);
+        GF_FREE (dentry);
+out:
+        return;
 }
 
 
-void 
-gf_quota_update_current_free_disk (xlator_t *this)
+int32_t
+quota_validate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *dict,
+                    dict_t *xdata)
 {
-	call_frame_t *frame = NULL;
-	call_pool_t   *pool = NULL;
+        quota_local_t     *local          = NULL;
+        uint32_t           validate_count = 0, link_count = 0;
+        int32_t            ret            = 0;
+        quota_inode_ctx_t *ctx            = NULL;
+        int64_t           *size           = 0;
+        uint64_t           value          = 0;
+        call_stub_t       *stub           = NULL;
+
+        local = frame->local;
+
+        if (op_ret < 0) {
+                goto unwind;
+        }
+
+        GF_ASSERT (local);
+        GF_ASSERT (frame);
+        GF_VALIDATE_OR_GOTO_WITH_ERROR ("quota", this, unwind, op_errno,
+                                        EINVAL);
+        GF_VALIDATE_OR_GOTO_WITH_ERROR (this->name, dict, unwind, op_errno,
+                                        EINVAL);
+
+        ret = inode_ctx_get (local->validate_loc.inode, this, &value);
+
+        ctx = (quota_inode_ctx_t *)(unsigned long)value;
+        if ((ret == -1) || (ctx == NULL)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context is not present in inode (gfid:%s)",
+                        uuid_utoa (local->validate_loc.inode->gfid));
+                op_errno = EINVAL;
+                goto unwind;
+        }
+
+        ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "size key not present in dict");
+                op_errno = EINVAL;
+                goto unwind;
+        }
+
+        local->just_validated = 1; /* so that we don't go into infinite
+                                    * loop of validation and checking
+                                    * limit when timeout is zero.
+                                    */
+        LOCK (&ctx->lock);
+        {
+                ctx->size = ntoh64 (*size);
+                gettimeofday (&ctx->tv, NULL);
+        }
+        UNLOCK (&ctx->lock);
+
+        quota_check_limit (frame, local->validate_loc.inode, this, NULL, NULL);
+        return 0;
+
+unwind:
+        LOCK (&local->lock);
+        {
+                local->op_ret = -1;
+                local->op_errno = op_errno;
+
+                validate_count = --local->validate_count;
+                link_count = local->link_count;
+
+                if ((validate_count == 0) && (link_count == 0)) {
+                        stub = local->stub;
+                        local->stub = NULL;
+                }
+        }
+        UNLOCK (&local->lock);
 
-	struct quota_priv *priv = NULL;
+        if (stub != NULL) {
+                call_resume (stub);
+        }
 
-	pool  = this->ctx->pool;
-	frame = create_frame (this, pool);
-  
-	priv = this->private;
+        return 0;
+}
 
-	STACK_WIND (frame, quota_statvfs_cbk,
-		    this->children->xlator,
-		    this->children->xlator->fops->statfs, &(priv->root_loc));
 
-	return ;
+static inline uint64_t
+quota_time_elapsed (struct timeval *now, struct timeval *then)
+{
+        return (now->tv_sec - then->tv_sec);
 }
 
 
-int
-gf_quota_check_free_disk (xlator_t *this) 
-{
-        struct quota_priv * priv = NULL;
-	struct timeval tv = {0, 0};
-
-	priv = this->private;
-	if (priv->min_free_disk_limit) {
-		gettimeofday (&tv, NULL);
-		if (tv.tv_sec > (priv->refresh_interval + 
-				 priv->min_disk_last_updated_time)) {
-			priv->min_disk_last_updated_time = tv.tv_sec;
-			gf_quota_update_current_free_disk (this);
-		}
-		if (priv->current_free_disk <= priv->min_free_disk_limit)
-			return -1;
-	}
+int32_t
+quota_timeout (struct timeval *tv, int32_t timeout)
+{
+        struct timeval now       = {0,};
+        int32_t        timed_out = 0;
+
+        gettimeofday (&now, NULL);
+
+        if (quota_time_elapsed (&now, tv) >= timeout) {
+                timed_out = 1;
+        }
 
-	return 0;
+        return timed_out;
 }
 
 
-int
-quota_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                    struct iatt *postbuf)
+int32_t
+quota_check_limit (call_frame_t *frame, inode_t *inode, xlator_t *this,
+                   char *name, uuid_t par)
 {
-	struct quota_priv *priv = this->private;
-	struct quota_local *local = NULL;
+        int32_t               ret            = -1;
+        inode_t              *_inode         = NULL, *parent = NULL;
+        quota_inode_ctx_t    *ctx            = NULL;
+        quota_priv_t         *priv           = NULL;
+        quota_local_t        *local          = NULL;
+        char                  need_validate  = 0, need_unwind = 0;
+        int64_t               delta          = 0;
+        call_stub_t          *stub           = NULL;
+        int32_t               validate_count = 0, link_count = 0;
+        uint64_t              value          = 0;
+        char                  just_validated = 0;
+        uuid_t                trav_uuid      = {0,};
 
-	local = frame->local;
+        GF_VALIDATE_OR_GOTO ("quota", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, frame, out);
+        GF_VALIDATE_OR_GOTO (this->name, inode, out);
 
-	if ((op_ret >= 0) && priv->disk_usage_limit) {
-		gf_quota_usage_subtract (this, (local->stbuf.ia_blocks -
-						postbuf->ia_blocks) * 512);
-		loc_wipe (&local->loc);
-	}
+        local = frame->local;
+        GF_VALIDATE_OR_GOTO (this->name, local, out);
+
+        delta = local->delta;
+
+        GF_VALIDATE_OR_GOTO (this->name, local->stub, out);
+
+        priv = this->private;
+
+        inode_ctx_get (inode, this, &value);
+        ctx = (quota_inode_ctx_t *)(unsigned long)value;
+
+        _inode = inode_ref (inode);
+
+        LOCK (&local->lock);
+        {
+                just_validated = local->just_validated;
+                local->just_validated = 0;
+
+                if (just_validated) {
+                        local->validate_count--;
+                }
+        }
+        UNLOCK (&local->lock);
+
+        if ( par != NULL ) {
+                uuid_copy (trav_uuid, par);
+        }
+
+        do {
+                if (ctx != NULL) {
+                        LOCK (&ctx->lock);
+                        {
+                                if (ctx->limit >= 0) {
+                                        if (!just_validated
+                                            && quota_timeout (&ctx->tv,
+                                                              priv->timeout)) {
+                                                need_validate = 1;
+                                        } else if ((ctx->size + delta)
+                                                   >= ctx->limit) {
+                                                local->op_ret = -1;
+                                                local->op_errno = EDQUOT;
+                                                need_unwind = 1;
+                                        }
+                                }
+                        }
+                        UNLOCK (&ctx->lock);
+
+                        if (need_validate) {
+                                goto validate;
+                        }
+
+                        if (need_unwind) {
+                                break;
+                        }
+                }
+
+                if (__is_root_gfid (_inode->gfid)) {
+                        break;
+                }
+
+                parent = inode_parent (_inode, trav_uuid, name);
+
+                if (name != NULL) {
+                        name = NULL;
+                        uuid_clear (trav_uuid);
+                }
+
+                if (parent == NULL) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "cannot find parent for inode (gfid:%s), hence "
+                                "aborting enforcing quota-limits and continuing"
+                                " with the fop", uuid_utoa (_inode->gfid));
+                }
+
+                inode_unref (_inode);
+                _inode = parent;
+                just_validated = 0;
+
+                if (_inode == NULL) {
+                        break;
+                }
+
+                value = 0;
+                inode_ctx_get (_inode, this, &value);
+                ctx = (quota_inode_ctx_t *)(unsigned long)value;
+        } while (1);
+
+        ret = 0;
+
+        if (_inode != NULL) {
+                inode_unref (_inode);
+        }
+
+        LOCK (&local->lock);
+        {
+                validate_count = local->validate_count;
+                link_count = local->link_count;
+                if ((validate_count == 0) && (link_count == 0)) {
+                        stub = local->stub;
+                        local->stub = NULL;
+                }
+        }
+        UNLOCK (&local->lock);
+
+        if (stub != NULL) {
+                call_resume (stub);
+        }
+
+out:
+        return ret;
+
+validate:
+        LOCK (&local->lock);
+        {
+                loc_wipe (&local->validate_loc);
+
+                if (just_validated) {
+                        local->validate_count--;
+                }
+
+                local->validate_count++;
+                ret = quota_inode_loc_fill (_inode, &local->validate_loc);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "cannot fill loc for inode (gfid:%s), hence "
+                                "aborting quota-checks and continuing with fop",
+                                uuid_utoa (_inode->gfid));
+                        local->validate_count--;
+                }
+        }
+        UNLOCK (&local->lock);
+
+        if (ret < 0) {
+                goto loc_fill_failed;
+        }
+
+        STACK_WIND (frame, quota_validate_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, &local->validate_loc,
+                    QUOTA_SIZE_KEY, NULL);
 
-	STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno,
-                             prebuf, postbuf);
-	return 0;
+loc_fill_failed:
+        inode_unref (_inode);
+        return 0;
 }
 
 
-int
-quota_truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			 int32_t op_ret, int32_t op_errno, struct iatt *buf)
+int32_t
+quota_get_limit_value (inode_t *inode, xlator_t *this, int64_t *n)
 {
-	struct quota_local *local = NULL;
+        int32_t       ret        = 0;
+        char         *path       = NULL;
+        limits_t     *limit_node = NULL;
+        quota_priv_t *priv       = NULL;
+
+        if (inode == NULL || n == NULL) {
+                ret = -1;
+                goto out;
+        }
 
-	local = frame->local;
+        *n = 0;
 
-	if (op_ret >= 0) {
-		local->stbuf = *buf;
-	}
+        ret = inode_path (inode, NULL, &path);
+        if (ret < 0) {
+                ret = -1;
+                goto out;
+        }
+
+        priv = this->private;
 
-	STACK_WIND (frame, quota_truncate_cbk,
-		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->truncate,
-		    &local->loc, local->offset);
-	return 0;
+        list_for_each_entry (limit_node, &priv->limit_head, limit_list) {
+                if (strcmp (limit_node->path, path) == 0) {
+                        *n = limit_node->value;
+                        break;
+                }
+        }
+
+out:
+        GF_FREE (path);
+
+        return ret;
 }
 
 
-int
-quota_truncate (call_frame_t *frame, xlator_t *this,
-		loc_t *loc, off_t offset)
+static int32_t
+__quota_init_inode_ctx (inode_t *inode, int64_t limit, xlator_t *this,
+                        dict_t *dict, struct iatt *buf,
+                        quota_inode_ctx_t **context)
 {
-	struct quota_local *local = NULL;
-	struct quota_priv  *priv = NULL;
+        int32_t            ret  = -1;
+        int64_t           *size = 0;
+        quota_inode_ctx_t *ctx  = NULL;
 
-	priv = this->private;
+        if (inode == NULL) {
+                goto out;
+        }
 
-	if (priv->disk_usage_limit) {
-		local = GF_CALLOC (1, sizeof (struct quota_local),
-                                   gf_quota_mt_quota_local);
-		frame->local  = local;
+        QUOTA_ALLOC_OR_GOTO (ctx, quota_inode_ctx_t, out);
 
-		loc_copy (&local->loc, loc);
-		local->offset = offset;
+        ctx->limit = limit;
+        if (buf)
+                ctx->buf = *buf;
 
-		STACK_WIND (frame, quota_truncate_stat_cbk,
-			    FIRST_CHILD(this),
-			    FIRST_CHILD(this)->fops->stat, loc);
-		return 0;
-	}
+        LOCK_INIT(&ctx->lock);
+
+        if (context != NULL) {
+                *context = ctx;
+        }
+
+        INIT_LIST_HEAD (&ctx->parents);
+
+        if (dict != NULL) {
+                ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+                if (ret == 0) {
+                        ctx->size = ntoh64 (*size);
+                        gettimeofday (&ctx->tv, NULL);
+                }
+        }
 
-	STACK_WIND (frame, quota_truncate_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->truncate,
-		    loc, offset);
-	return 0;
+        ret = __inode_ctx_put (inode, this, (uint64_t )(long)ctx);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "cannot set quota context in inode (gfid:%s)",
+                        uuid_utoa (inode->gfid));
+        }
+out:
+        return ret;
 }
 
 
-int
-quota_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                     struct iatt *postbuf)
+static int32_t
+quota_inode_ctx_get (inode_t *inode, int64_t limit, xlator_t *this,
+                     dict_t *dict, struct iatt *buf, quota_inode_ctx_t **ctx,
+                     char create_if_absent)
 {
-	struct quota_priv  *priv = NULL;
-	struct quota_local *local = NULL;
+        int32_t  ret = 0;
+        uint64_t ctx_int;
+
+        LOCK (&inode->lock);
+        {
+                ret = __inode_ctx_get (inode, this, &ctx_int);
+
+                if ((ret == 0) && (ctx != NULL)) {
+                        *ctx = (quota_inode_ctx_t *) (unsigned long)ctx_int;
+                } else if (create_if_absent) {
+                        ret = __quota_init_inode_ctx (inode, limit, this, dict,
+                                                      buf, ctx);
+                }
+        }
+        UNLOCK (&inode->lock);
 
-	local = frame->local;
-	priv = this->private;
+        return ret;
+}
 
-	if ((op_ret >= 0) && priv->disk_usage_limit) {
-		gf_quota_usage_subtract (this, (local->stbuf.ia_blocks -
-						postbuf->ia_blocks) * 512);
-		fd_unref (local->fd);
-	}
 
-	STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno,
-                             prebuf, postbuf);
-	return 0;
+int32_t
+quota_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, inode_t *inode,
+                  struct iatt *buf, dict_t *dict, struct iatt *postparent)
+{
+        int32_t            ret        = -1;
+        char               found      = 0;
+        quota_local_t     *local      = NULL;
+        quota_inode_ctx_t *ctx        = NULL;
+        quota_dentry_t    *dentry     = NULL;
+        int64_t           *size       = 0;
+        uint64_t           value      = 0;
+        limits_t          *limit_node = NULL;
+        quota_priv_t      *priv       = NULL;
+
+        local = frame->local;
+
+        priv = this->private;
+
+        inode_ctx_get (inode, this, &value);
+        ctx = (quota_inode_ctx_t *)(unsigned long)value;
+
+        if ((op_ret < 0) || (local == NULL)
+            || (((ctx == NULL) || (ctx->limit == local->limit))
+                && (local->limit < 0) && !((IA_ISREG (buf->ia_type))
+                                           || (IA_ISLNK (buf->ia_type))))) {
+                goto unwind;
+        }
+
+        LOCK (&priv->lock);
+        {
+                list_for_each_entry (limit_node, &priv->limit_head,
+                                     limit_list) {
+                        if (strcmp (local->loc.path, limit_node->path) == 0) {
+                                uuid_copy (limit_node->gfid, buf->ia_gfid);
+                                break;
+                        }
+                }
+        }
+        UNLOCK (&priv->lock);
+
+        ret = quota_inode_ctx_get (local->loc.inode, local->limit, this, dict,
+                                   buf, &ctx, 1);
+        if ((ret == -1) || (ctx == NULL)) {
+                gf_log (this->name, GF_LOG_WARNING, "cannot create quota "
+                        "context in inode(gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                op_ret = -1;
+                op_errno = ENOMEM;
+                goto unwind;
+        }
+
+        LOCK (&ctx->lock);
+        {
+
+                if (dict != NULL) {
+                        ret = dict_get_bin (dict, QUOTA_SIZE_KEY,
+                                            (void **) &size);
+                        if (ret == 0) {
+                                ctx->size = ntoh64 (*size);
+                                gettimeofday (&ctx->tv, NULL);
+                        }
+                }
+
+                if (local->limit != ctx->limit) {
+                        ctx->limit = local->limit;
+                }
+
+                ctx->buf = *buf;
+
+                if (!(IA_ISREG (buf->ia_type) || IA_ISLNK (buf->ia_type))) {
+                        goto unlock;
+                }
+
+                if (local->loc.name == NULL)
+                        goto unlock;
+
+                list_for_each_entry (dentry, &ctx->parents, next) {
+                        if ((strcmp (dentry->name, local->loc.name) == 0) &&
+                            (uuid_compare (local->loc.parent->gfid,
+                                           dentry->par) == 0)) {
+                                found = 1;
+                                break;
+                        }
+                }
+
+                if (!found) {
+                        dentry = __quota_dentry_new (ctx,
+                                                     (char *)local->loc.name,
+                                                     local->loc.parent->gfid);
+                        if (dentry == NULL) {
+                                /*
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "cannot create a new dentry (par:%"
+                                        PRId64", name:%s) for inode(ino:%"
+                                        PRId64", gfid:%s)",
+                                        uuid_utoa (local->loc.inode->gfid));
+                                */
+                                op_ret = -1;
+                                op_errno = ENOMEM;
+                                goto unlock;
+                        }
+                }
+        }
+unlock:
+        UNLOCK (&ctx->lock);
+
+unwind:
+        QUOTA_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, buf,
+                            dict, postparent);
+        return 0;
 }
 
 
-int
-quota_ftruncate_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			   int32_t op_ret, int32_t op_errno, struct iatt *buf)
+int32_t
+quota_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
+              dict_t *xattr_req)
 {
-	struct quota_local *local = NULL;
+        int32_t             ret         = -1;
+        int64_t             limit       = -1;
+        limits_t           *limit_node  = NULL;
+        gf_boolean_t        dict_newed  = _gf_false;
+        quota_priv_t       *priv        = NULL;
+        quota_local_t      *local       = NULL;
+
+        priv = this->private;
+
+        list_for_each_entry (limit_node, &priv->limit_head, limit_list) {
+                if (strcmp (limit_node->path, loc->path) == 0) {
+                        limit = limit_node->value;
+                }
+        }
 
-	local = frame->local;
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto err;
+        }
 
-	if (op_ret >= 0) {
-		local->stbuf = *buf;
-	}
+        ret = loc_copy (&local->loc, loc);
+        if (ret == -1) {
+                goto err;
+        }
+
+        frame->local = local;
 
-	STACK_WIND (frame, quota_ftruncate_cbk,
-		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->ftruncate,
-		    local->fd, local->offset);
-	return 0;
+        local->limit = limit;
+
+        if (limit < 0) {
+                goto wind;
+        }
+
+        if (xattr_req == NULL) {
+                xattr_req  = dict_new ();
+                dict_newed = _gf_true;
+        }
+
+        ret = dict_set_uint64 (xattr_req, QUOTA_SIZE_KEY, 0);
+        if (ret < 0) {
+                goto err;
+        }
+
+wind:
+        STACK_WIND (frame, quota_lookup_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+
+        ret = 0;
+
+err:
+        if (ret < 0) {
+                QUOTA_STACK_UNWIND (lookup, frame, -1, ENOMEM,
+                                    NULL, NULL, NULL, NULL);
+        }
+
+        if (dict_newed == _gf_true) {
+                dict_unref (xattr_req);
+        }
+
+        return 0;
 }
 
 
-int
-quota_ftruncate (call_frame_t *frame, xlator_t *this,
-		 fd_t *fd, off_t offset)
+void
+quota_update_size (xlator_t *this, inode_t *inode, char *name, uuid_t par,
+                   int64_t delta)
 {
-	struct quota_local *local = NULL;
-	struct quota_priv  *priv = NULL;
+        inode_t           *_inode    = NULL;
+        inode_t           *parent    = NULL;
+        uint64_t           value     = 0;
+        quota_inode_ctx_t *ctx       = NULL;
+        uuid_t             trav_uuid = {0,};
 
+        GF_VALIDATE_OR_GOTO ("quota", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, inode, out);
 
-	priv = this->private;
+        inode_ctx_get (inode, this, &value);
+        ctx = (quota_inode_ctx_t *)(unsigned long)value;
 
-	if (priv->disk_usage_limit) {
-		local = GF_CALLOC (1, sizeof (struct quota_local),
-                                   gf_quota_mt_quota_local);
-		frame->local  = local;
+        _inode = inode_ref (inode);
 
-		local->fd = fd_ref (fd);
-		local->offset = offset;
+        if ( par != NULL ) {
+                uuid_copy (trav_uuid, par);
+        }
 
-		STACK_WIND (frame, quota_ftruncate_fstat_cbk,
-			    FIRST_CHILD(this),
-			    FIRST_CHILD(this)->fops->fstat, fd);
-		return 0;
-	}
+        do {
+                if ((ctx != NULL) && (ctx->limit >= 0)) {
+                        LOCK (&ctx->lock);
+                        {
+                                ctx->size += delta;
+                        }
+                        UNLOCK (&ctx->lock);
+                }
+
+                if (__is_root_gfid (_inode->gfid)) {
+                        break;
+                }
+
+                parent = inode_parent (_inode, trav_uuid, name);
+                if (parent == NULL) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "cannot find parent for inode (gfid:%s), hence "
+                                "aborting size updation of parents",
+                                uuid_utoa (_inode->gfid));
+                }
 
-	STACK_WIND (frame, quota_ftruncate_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->ftruncate,
-		    fd, offset);
-	return 0;
+                if (name != NULL) {
+                        name = NULL;
+                        uuid_clear (trav_uuid);
+                }
+
+                inode_unref (_inode);
+                _inode = parent;
+
+                if (_inode == NULL) {
+                        break;
+                }
+
+                inode_ctx_get (_inode, this, &value);
+                ctx = (quota_inode_ctx_t *)(unsigned long)value;
+        } while (1);
+
+out:
+        return;
 }
 
 
-int
-quota_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int32_t op_ret, int32_t op_errno,
-                 inode_t *inode, struct iatt *buf, struct iatt *preparent,
-                 struct iatt *postparent)
+int32_t
+quota_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                  struct iatt *postbuf, dict_t *xdata)
 {
-	struct quota_priv *priv = NULL;
+        int32_t                  ret            = 0;
+        uint64_t                 ctx_int        = 0;
+        quota_inode_ctx_t       *ctx            = NULL;
+        quota_local_t           *local          = NULL;
+        quota_dentry_t          *dentry         = NULL;
+        int64_t                  delta          = 0;
 
-	priv = this->private;
+        local = frame->local;
 
-	if ((op_ret >= 0) && priv->disk_usage_limit) {
-		gf_quota_usage_add (this, buf->ia_blocks * 512);
-	}
+        if ((op_ret < 0) || (local == NULL)) {
+                goto out;
+        }
+
+        ret = inode_ctx_get (local->loc.inode, this, &ctx_int);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to get the context", local->loc.path);
+                goto out;
+        }
+
+        ctx = (quota_inode_ctx_t *)(unsigned long) ctx_int;
+
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in %s (gfid:%s)",
+                        local->loc.path, uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->buf = *postbuf;
+        }
+        UNLOCK (&ctx->lock);
+
+        list_for_each_entry (dentry, &ctx->parents, next) {
+                delta = (postbuf->ia_blocks - prebuf->ia_blocks) * 512;
+                quota_update_size (this, local->loc.inode,
+                                   dentry->name, dentry->par, delta);
+        }
+
+out:
+        QUOTA_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf,
+                            xdata);
 
-	STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
-	return 0;
+        return 0;
 }
 
 
-int
-quota_mknod (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, mode_t mode, dev_t rdev, dict_t *params)
+int32_t
+quota_writev_helper (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                     struct iovec *vector, int32_t count, off_t off,
+                     uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
-	struct quota_priv *priv = NULL;
+        quota_local_t *local    = NULL;
+        int32_t        op_errno = EINVAL;
 
-	priv = this->private;
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto unwind;
+        }
 
-	if (gf_quota_check_free_disk (this) == -1) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"min-free-disk limit (%u) crossed, current available is %u",
-			priv->min_free_disk_limit, priv->current_free_disk);
-		STACK_UNWIND_STRICT (mknod, frame, -1, ENOSPC, NULL, NULL,
-                                     NULL, NULL);
-		return 0;
-	}
+        if (local->op_ret == -1) {
+                op_errno = local->op_errno;
+                goto unwind;
+        }
+
+        STACK_WIND (frame, quota_writev_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->writev, fd, vector, count, off,
+                    flags, iobref, xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+quota_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+              struct iovec *vector, int32_t count, off_t off,
+              uint32_t flags, struct iobref *iobref, dict_t *xdata)
+{
+        int32_t            ret     = -1, op_errno = EINVAL;
+        int32_t            parents = 0;
+        uint64_t           size    = 0;
+        quota_local_t     *local   = NULL;
+        quota_inode_ctx_t *ctx     = NULL;
+        quota_priv_t      *priv    = NULL;
+        call_stub_t       *stub    = NULL;
+        quota_dentry_t    *dentry  = NULL;
+
+        GF_ASSERT (frame);
+        GF_VALIDATE_OR_GOTO ("quota", this, unwind);
+        GF_VALIDATE_OR_GOTO (this->name, fd, unwind);
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto unwind;
+        }
+
+        frame->local = local;
+        local->loc.inode = inode_ref (fd->inode);
+
+        ret = quota_inode_ctx_get (fd->inode, -1, this, NULL, NULL, &ctx, 0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (fd->inode->gfid));
+                goto unwind;
+        }
+
+        stub = fop_writev_stub (frame, quota_writev_helper, fd, vector, count,
+                                off, flags, iobref, xdata);
+        if (stub == NULL) {
+                op_errno = ENOMEM;
+                goto unwind;
+        }
+
+        priv = this->private;
+        GF_VALIDATE_OR_GOTO (this->name, priv, unwind);
+
+        size = iov_length (vector, count);
+        LOCK (&ctx->lock);
+        {
+                list_for_each_entry (dentry, &ctx->parents, next) {
+                        parents++;
+                }
+        }
+        UNLOCK (&ctx->lock);
+
+        local->delta = size;
+        local->stub = stub;
+        local->link_count = parents;
+
+        list_for_each_entry (dentry, &ctx->parents, next) {
+                ret = quota_check_limit (frame, fd->inode, this, dentry->name,
+                                         dentry->par);
+                if (ret == -1) {
+                        break;
+                }
+        }
+
+        stub = NULL;
+
+        LOCK (&local->lock);
+        {
+                local->link_count = 0;
+                if (local->validate_count == 0) {
+                        stub = local->stub;
+                        local->stub = NULL;
+                }
+        }
+        UNLOCK (&local->lock);
 
-        if (priv->current_disk_usage > priv->disk_usage_limit) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"Disk usage limit (%"PRIu64") crossed, current usage is %"PRIu64"",
-			priv->disk_usage_limit, priv->current_disk_usage);
-		STACK_UNWIND_STRICT (mknod, frame, -1, ENOSPC, NULL, NULL,
-                                     NULL, NULL);
-		return 0;
+        if (stub != NULL) {
+                call_resume (stub);
         }
 
-	STACK_WIND (frame, quota_mknod_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->mknod,
-		    loc, mode, rdev, params);
-	return 0;
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
 }
 
 
-int
+int32_t
 quota_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int32_t op_ret, int32_t op_errno, inode_t *inode,
+                 int32_t op_ret, int32_t op_errno, inode_t *inode,
                  struct iatt *buf, struct iatt *preparent,
-                 struct iatt *postparent)
+                 struct iatt *postparent, dict_t *xdata)
 {
-	struct quota_priv *priv = NULL;
+        QUOTA_STACK_UNWIND (mkdir, frame, op_ret, op_errno, inode,
+                            buf, preparent, postparent, xdata);
+        return 0;
+}
 
-	priv = this->private;
 
-	if ((op_ret >= 0) && priv->disk_usage_limit) {
-		gf_quota_usage_subtract (this, buf->ia_blocks * 512);
-	}
+int32_t
+quota_mkdir_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                    mode_t mode, mode_t umask, dict_t *xdata)
+{
+        quota_local_t *local    = NULL;
+        int32_t        op_errno = EINVAL;
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto unwind;
+        }
+
+        op_errno = local->op_errno;
+
+        if (local->op_ret == -1) {
+                goto unwind;
+        }
 
-	STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
-	return 0;
+        STACK_WIND (frame, quota_mkdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL,
+                            NULL, NULL, NULL);
+        return 0;
 }
 
 
-int
+int32_t
 quota_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-             dict_t *params)
+             mode_t umask, dict_t *xdata)
 {
-	struct quota_priv *priv = NULL;
+        int32_t        ret            = 0, op_errno = 0;
+        quota_local_t *local          = NULL;
+        call_stub_t   *stub           = NULL;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	priv = this->private;
+        frame->local = local;
 
-	if (gf_quota_check_free_disk (this) == -1) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"min-free-disk limit (%u) crossed, current available is %u",
-			priv->min_free_disk_limit, priv->current_free_disk);
-		STACK_UNWIND_STRICT (mkdir, frame, -1, ENOSPC, NULL, NULL,
-                                     NULL, NULL);
-		return 0;
-		
-	}
+        local->link_count = 1;
+
+        ret = loc_copy (&local->loc, loc);
+        if (ret) {
+                op_errno = ENOMEM;
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto err;
+        }
 
-        if (priv->current_disk_usage > priv->disk_usage_limit) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"Disk usage limit (%"PRIu64") crossed, current usage is %"PRIu64"",
-			priv->disk_usage_limit, priv->current_disk_usage);
-		STACK_UNWIND_STRICT (mkdir, frame, -1, ENOSPC, NULL, NULL,
-                                     NULL, NULL);
-		return 0;
+        stub = fop_mkdir_stub (frame, quota_mkdir_helper, loc, mode, umask,
+                               xdata);
+        if (stub == NULL) {
+                op_errno = ENOMEM;
+                goto err;
         }
 
-	STACK_WIND (frame, quota_mkdir_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->mkdir,
-		    loc, mode, params);
+        local->stub = stub;
+        local->delta = 0;
 
-	return 0;
+        quota_check_limit (frame, loc->parent, this, NULL, NULL);
+
+        stub = NULL;
+
+        LOCK (&local->lock);
+        {
+                if (local->validate_count == 0) {
+                        stub = local->stub;
+                        local->stub = NULL;
+                }
+
+                local->link_count = 0;
+        }
+        UNLOCK (&local->lock);
+
+        if (stub != NULL) {
+                call_resume (stub);
+        }
+
+        return 0;
+err:
+        QUOTA_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
+                            NULL, NULL);
+
+        return 0;
 }
 
 
-int
+int32_t
+quota_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+                  struct iatt *buf, struct iatt *preparent,
+                  struct iatt *postparent, dict_t *xdata)
+{
+        int32_t            ret    = -1;
+        quota_local_t     *local  = NULL;
+        quota_inode_ctx_t *ctx    = NULL;
+        quota_dentry_t    *dentry = NULL;
+
+        local = frame->local;
+        if (op_ret < 0) {
+                goto unwind;
+        }
+
+        ret = quota_inode_ctx_get (inode, -1, this, NULL, buf, &ctx, 1);
+        if ((ret == -1) || (ctx == NULL)) {
+                gf_log (this->name, GF_LOG_WARNING, "cannot create quota "
+                        "context in inode(gfid:%s)",
+                        uuid_utoa (inode->gfid));
+                op_ret = -1;
+                op_errno = ENOMEM;
+                goto unwind;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->buf = *buf;
+
+                dentry = __quota_dentry_new (ctx, (char *)local->loc.name,
+                                             local->loc.parent->gfid);
+                if (dentry == NULL) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "cannot create a new dentry (name:%s) for "
+                                "inode(gfid:%s)", local->loc.name,
+                                uuid_utoa (local->loc.inode->gfid));
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        goto unlock;
+                }
+        }
+unlock:
+        UNLOCK (&ctx->lock);
+
+unwind:
+        QUOTA_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, buf,
+                            preparent, postparent, xdata);
+        return 0;
+}
+
+
+int32_t
+quota_create_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                     int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
+                     dict_t *xdata)
+{
+        quota_local_t *local    = NULL;
+        int32_t        op_errno = EINVAL;
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto unwind;
+        }
+
+        if (local->op_ret == -1) {
+                op_errno = local->op_errno;
+                goto unwind;
+        }
+
+        STACK_WIND (frame, quota_create_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->create, loc, flags, mode, umask,
+                    fd, xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL,
+                            NULL, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+quota_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+              mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        int32_t            ret            = -1;
+        quota_local_t     *local          = NULL;
+        call_stub_t       *stub           = NULL;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto err;
+        }
+
+        frame->local = local;
+
+        ret = loc_copy (&local->loc, loc);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto err;
+        }
+
+        stub = fop_create_stub (frame, quota_create_helper, loc, flags, mode,
+                                umask, fd, xdata);
+        if (stub == NULL) {
+                goto err;
+        }
+
+        local->link_count = 1;
+        local->stub = stub;
+        local->delta = 0;
+
+        quota_check_limit (frame, loc->parent, this, NULL, NULL);
+
+        stub = NULL;
+
+        LOCK (&local->lock);
+        {
+                local->link_count = 0;
+                if (local->validate_count == 0) {
+                        stub = local->stub;
+                        local->stub = NULL;
+                }
+        }
+        UNLOCK (&local->lock);
+
+        if (stub != NULL) {
+                call_resume (stub);
+        }
+
+        return 0;
+err:
+        QUOTA_STACK_UNWIND (create, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+                            NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
 quota_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                  struct iatt *postparent)
+                  int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                  struct iatt *postparent, dict_t *xdata)
 {
-	struct quota_local *local = NULL;
+        quota_local_t     *local = NULL;
+        quota_inode_ctx_t *ctx   = NULL;
+        uint64_t           value = 0;
 
-	local = frame->local;
+        if (op_ret < 0) {
+                goto out;
+        }
 
-	if (local) {
-		if (op_ret >= 0) {
-			gf_quota_usage_subtract (this,
-						 local->stbuf.ia_blocks * 512);
-		}
-		loc_wipe (&local->loc);
-	}
+        local = (quota_local_t *) frame->local;
 
-	STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno, preparent, postparent);
-	return 0;
+        inode_ctx_get (local->loc.inode, this, &value);
+        ctx = (quota_inode_ctx_t *)(unsigned long)value;
+
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        quota_update_size (this, local->loc.inode, (char *)local->loc.name,
+                           local->loc.parent->gfid,
+                           (-(ctx->buf.ia_blocks * 512)));
+
+out:
+        QUOTA_STACK_UNWIND (unlink, frame, op_ret, op_errno, preparent,
+                            postparent, xdata);
+        return 0;
 }
 
 
-int
-quota_unlink_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno, struct iatt *buf)
+int32_t
+quota_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+              dict_t *xdata)
 {
-	struct quota_local *local = NULL;
+        int32_t        ret = 0;
+        quota_local_t *local = NULL;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto err;
+        }
 
-	local = frame->local;
+        frame->local = local;
 
-	if (op_ret >= 0) {
-		if (buf->ia_nlink == 1) {
-			local->stbuf = *buf;
-		}
-	}
+        ret = loc_copy (&local->loc, loc);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto err;
+        }
+
+        STACK_WIND (frame, quota_unlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
 
-	STACK_WIND (frame, quota_unlink_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->unlink,
-		    &local->loc);
+        ret = 0;
+
+err:
+        if (ret == -1) {
+                QUOTA_STACK_UNWIND (unlink, frame, -1, 0, NULL, NULL, NULL);
+        }
 
-	return 0;
+        return 0;
 }
 
 
-int
-quota_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
+int32_t
+quota_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, inode_t *inode,
+                struct iatt *buf, struct iatt *preparent,
+                struct iatt *postparent, dict_t *xdata)
 {
-	struct quota_local *local = NULL;
-	struct quota_priv  *priv = NULL;
+        int32_t               ret          = -1;
+        quota_local_t        *local        = NULL;
+        quota_inode_ctx_t    *ctx          = NULL;
+        quota_dentry_t       *dentry       = NULL;
+        char                  found        = 0;
+
+        if (op_ret < 0) {
+                goto out;
+        }
 
-	priv = this->private;
+        local = (quota_local_t *) frame->local;
 
-	if (priv->disk_usage_limit) {
-		local = GF_CALLOC (1, sizeof (struct quota_local),
-                                   gf_quota_mt_quota_local);
-		frame->local = local;
+        quota_update_size (this, local->loc.parent, NULL, NULL,
+                           (buf->ia_blocks * 512));
 
-		loc_copy (&local->loc, loc);
+        ret = quota_inode_ctx_get (inode, -1, this, NULL, NULL, &ctx, 0);
+        if ((ret == -1) || (ctx == NULL)) {
+                gf_log (this->name, GF_LOG_WARNING, "cannot find quota "
+                        "context in %s (gfid:%s)", local->loc.path,
+                        uuid_utoa (inode->gfid));
+                op_ret = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
 
-		STACK_WIND (frame,
-			    quota_unlink_stat_cbk,
-			    FIRST_CHILD(this),
-			    FIRST_CHILD(this)->fops->stat,
-			    loc);
-		return 0;
-	}
+        LOCK (&ctx->lock);
+        {
+                list_for_each_entry (dentry, &ctx->parents, next) {
+                        if ((strcmp (dentry->name, local->loc.name) == 0) &&
+                            (uuid_compare (local->loc.parent->gfid,
+                                           dentry->par) == 0)) {
+                                found = 1;
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "new entry being linked (name:%s) for "
+                                        "inode (gfid:%s) is already present "
+                                        "in inode-dentry-list", dentry->name,
+                                        uuid_utoa (local->loc.inode->gfid));
+                                break;
+                        }
+                }
+
+                if (!found) {
+                        dentry = __quota_dentry_new (ctx,
+                                                     (char *)local->loc.name,
+                                                     local->loc.parent->gfid);
+                        if (dentry == NULL) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "cannot create a new dentry (name:%s) "
+                                        "for inode(gfid:%s)", local->loc.name,
+                                        uuid_utoa (local->loc.inode->gfid));
+                                op_ret = -1;
+                                op_errno = ENOMEM;
+                                goto unlock;
+                        }
+                }
+
+                ctx->buf = *buf;
+        }
+unlock:
+        UNLOCK (&ctx->lock);
+
+out:
+        QUOTA_STACK_UNWIND (link, frame, op_ret, op_errno, inode, buf,
+                            preparent, postparent, xdata);
 
-	STACK_WIND (frame, quota_unlink_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->unlink,
-		    loc);
-	return 0;
+        return 0;
 }
 
 
-int
-quota_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                 struct iatt *postparent)
+int32_t
+quota_link_helper (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+                   loc_t *newloc, dict_t *xdata)
 {
-	struct quota_local *local = NULL;
+        quota_local_t *local    = NULL;
+        int32_t        op_errno = EINVAL;
 
-	local = frame->local;
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto unwind;
+        }
 
-	if (local) {
-		if (op_ret >= 0) {
-			gf_quota_usage_subtract (this, local->stbuf.ia_blocks * 512);
-		}
-		loc_wipe (&local->loc);
-	}
+        op_errno = local->op_errno;
 
-	STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno, preparent, postparent);
-	return 0;
+        if (local->op_ret == -1) {
+                goto unwind;
+        }
+
+        STACK_WIND (frame, quota_link_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL,
+                            NULL, NULL, NULL);
+        return 0;
 }
 
 
-int
-quota_rmdir_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		      int32_t op_ret, int32_t op_errno, struct iatt *buf)
+int32_t
+quota_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+            dict_t *xdata)
 {
-	struct quota_local *local = NULL;
+        int32_t            ret   = -1, op_errno = ENOMEM;
+        quota_local_t     *local = NULL;
+        call_stub_t       *stub  = NULL;
+        quota_inode_ctx_t *ctx = NULL;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto err;
+        }
 
-	local = frame->local;
+        frame->local = (void *) local;
 
-	if (op_ret >= 0) {
-		local->stbuf = *buf;
-	}
+        ret = loc_copy (&local->loc, newloc);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto err;
+        }
+
+        stub = fop_link_stub (frame, quota_link_helper, oldloc, newloc, xdata);
+        if (stub == NULL) {
+                goto err;
+        }
+
+        local->link_count = 1;
+        local->stub = stub;
+
+        ret = quota_inode_ctx_get (oldloc->inode, -1, this, NULL, NULL, &ctx,
+                                   0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        oldloc->inode ? uuid_utoa (oldloc->inode->gfid) : "0");
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->delta = ctx->buf.ia_blocks * 512;
+
+        quota_check_limit (frame, newloc->parent, this, NULL, NULL);
 
-	STACK_WIND (frame, quota_rmdir_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->rmdir,
-		    &local->loc);
+        stub = NULL;
 
-	return 0;
+        LOCK (&local->lock);
+        {
+                if (local->validate_count == 0) {
+                        stub = local->stub;
+                        local->stub = NULL;
+                }
+
+                local->link_count = 0;
+        }
+        UNLOCK (&local->lock);
+
+        if (stub != NULL) {
+                call_resume (stub);
+        }
+
+        ret = 0;
+err:
+        if (ret < 0) {
+                QUOTA_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL,
+                                    NULL, NULL, NULL);
+        }
+
+        return 0;
 }
 
 
-int
-quota_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
+int32_t
+quota_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                  struct iatt *preoldparent, struct iatt *postoldparent,
+                  struct iatt *prenewparent, struct iatt *postnewparent,
+                  dict_t *xdata)
 {
-	struct quota_local *local = NULL;
-	struct quota_priv  *priv = NULL;
+        int32_t               ret              = -1;
+        quota_local_t        *local            = NULL;
+        quota_inode_ctx_t    *ctx              = NULL;
+        quota_dentry_t       *old_dentry       = NULL, *dentry = NULL;
+        char                  new_dentry_found = 0;
+        int64_t               size             = 0;
+
+        if (op_ret < 0) {
+                goto out;
+        }
+
+        local = frame->local;
+        if (local == NULL) {
+                op_ret = -1;
+                op_errno = EINVAL;
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto out;
+        }
+
+        if (IA_ISREG (local->oldloc.inode->ia_type)
+            || IA_ISLNK (local->oldloc.inode->ia_type)) {
+                size = buf->ia_blocks * 512;
+        }
 
-	priv = this->private;
+        if (local->oldloc.parent != local->newloc.parent) {
+                quota_update_size (this, local->oldloc.parent, NULL, NULL, (-size));
+                quota_update_size (this, local->newloc.parent, NULL, NULL, size);
+        }
 
-	if (priv->disk_usage_limit) {
-		local = GF_CALLOC (1, sizeof (struct quota_local),
-                                   gf_quota_mt_quota_local);
-		frame->local = local;
+        if (!(IA_ISREG (local->oldloc.inode->ia_type)
+              || IA_ISLNK (local->oldloc.inode->ia_type))) {
+                goto out;
+        }
 
-		loc_copy (&local->loc, loc);
+        ret = quota_inode_ctx_get (local->oldloc.inode, -1, this, NULL, NULL,
+                                   &ctx, 0);
+        if ((ret == -1) || (ctx == NULL)) {
+                gf_log (this->name, GF_LOG_WARNING, "quota context not"
+                        "set in inode(gfid:%s)",
+                        uuid_utoa (local->oldloc.inode->gfid));
+                op_ret = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
 
-		STACK_WIND (frame, quota_rmdir_stat_cbk,
-			    FIRST_CHILD(this),
-			    FIRST_CHILD(this)->fops->stat, loc);
-		return 0;
-	}
+        LOCK (&ctx->lock);
+        {
+                /* decision of whether to create a context in newloc->inode
+                 * depends on fuse_rename_cbk's choice of inode it retains
+                 * after rename. currently it just associates oldloc->inode
+                 * with new parent and name. If this changes, following code
+                 * should be changed to set a new context in newloc->inode.
+                 */
+                list_for_each_entry (dentry, &ctx->parents, next) {
+                        if ((strcmp (dentry->name, local->oldloc.name) == 0) &&
+                            (uuid_compare (local->oldloc.parent->gfid,
+                                           dentry->par) == 0)) {
+                                old_dentry = dentry;
+                        } else if ((strcmp (dentry->name,
+                                            local->newloc.name) == 0) &&
+                                   (uuid_compare (local->oldloc.parent->gfid,
+                                                  dentry->par) == 0)) {
+                                new_dentry_found = 1;
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "new entry being linked (name:%s) for "
+                                        "inode (gfid:%s) is already present "
+                                        "in inode-dentry-list", dentry->name,
+                                        uuid_utoa (local->newloc.inode->gfid));
+                                break;
+                        }
+                }
+
+                if (old_dentry != NULL) {
+                        __quota_dentry_free (old_dentry);
+                } else {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "dentry corresponding to the path just renamed "
+                                "(name:%s) is not present", local->oldloc.name);
+                }
+
+                if (!new_dentry_found) {
+                        dentry = __quota_dentry_new (ctx,
+                                                     (char *)local->newloc.name,
+                                                     local->newloc.parent->gfid);
+                        if (dentry == NULL) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "cannot create a new dentry (name:%s) "
+                                        "for inode(gfid:%s)", local->newloc.name,
+                                        uuid_utoa (local->newloc.inode->gfid));
+                                op_ret = -1;
+                                op_errno = ENOMEM;
+                                goto unlock;
+                        }
+                }
+
+                ctx->buf = *buf;
+        }
+unlock:
+        UNLOCK (&ctx->lock);
+
+out:
+        QUOTA_STACK_UNWIND (rename, frame, op_ret, op_errno, buf, preoldparent,
+                            postoldparent, prenewparent, postnewparent, xdata);
 
-	STACK_WIND (frame, quota_rmdir_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->rmdir,
-		    loc);
-	return 0;
+        return 0;
 }
 
 
-int
+int32_t
+quota_rename_helper (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+                     loc_t *newloc, dict_t *xdata)
+{
+        quota_local_t *local    = NULL;
+        int32_t        op_errno = EINVAL;
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto unwind;
+        }
+
+        op_errno = local->op_errno;
+
+        if (local->op_ret == -1) {
+                goto unwind;
+        }
+
+        STACK_WIND (frame, quota_rename_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL,
+                            NULL, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+quota_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+              loc_t *newloc, dict_t *xdata)
+{
+        int32_t            ret            = -1, op_errno = ENOMEM;
+        quota_local_t     *local          = NULL;
+        call_stub_t       *stub           = NULL;
+        quota_inode_ctx_t *ctx            = NULL;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto err;
+        }
+
+        frame->local = local;
+
+        ret = loc_copy (&local->oldloc, oldloc);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto err;
+        }
+
+        ret = loc_copy (&local->newloc, newloc);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto err;
+        }
+
+        stub = fop_rename_stub (frame, quota_rename_helper, oldloc, newloc,
+                                xdata);
+        if (stub == NULL) {
+                goto err;
+        }
+
+        local->link_count = 1;
+        local->stub = stub;
+
+        if (IA_ISREG (oldloc->inode->ia_type)
+            || IA_ISLNK (oldloc->inode->ia_type)) {
+                ret = quota_inode_ctx_get (oldloc->inode, -1, this, NULL, NULL,
+                                           &ctx, 0);
+                if (ctx == NULL) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "quota context not set in inode (gfid:%s)",
+                                oldloc->inode ? uuid_utoa (oldloc->inode->gfid)
+                                : "0");
+                        op_errno = EINVAL;
+                        goto err;
+                }
+                local->delta = ctx->buf.ia_blocks * 512;
+        } else {
+                local->delta = 0;
+        }
+
+        quota_check_limit (frame, newloc->parent, this, NULL, NULL);
+
+        stub = NULL;
+
+        LOCK (&local->lock);
+        {
+                if (local->validate_count == 0) {
+                        stub = local->stub;
+                        local->stub = NULL;
+                }
+
+                local->link_count = 0;
+        }
+        UNLOCK (&local->lock);
+
+        if (stub != NULL) {
+                call_resume (stub);
+        }
+
+        ret = 0;
+err:
+        if (ret == -1) {
+                QUOTA_STACK_UNWIND (rename, frame, -1, op_errno, NULL,
+                                    NULL, NULL, NULL, NULL, NULL);
+        }
+
+        return 0;
+}
+
+
+int32_t
 quota_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno, inode_t *inode,
+                   int32_t op_ret, int32_t op_errno, inode_t *inode,
                    struct iatt *buf, struct iatt *preparent,
-                   struct iatt *postparent)
+                   struct iatt *postparent, dict_t *xdata)
 {
-	struct quota_priv *priv = NULL;
+        int64_t            size   = 0;
+        quota_local_t     *local  = NULL;
+        quota_inode_ctx_t *ctx    = NULL;
+        quota_dentry_t    *dentry = NULL;
 
-	priv = this->private;
+        if (op_ret < 0) {
+                goto out;
+        }
 
-	if ((op_ret >= 0) && priv->disk_usage_limit) {
-		gf_quota_usage_add (this, buf->ia_blocks * 512);
-	}
+        local = frame->local;
+        size = buf->ia_blocks * 512;
+
+        quota_update_size (this, local->loc.parent, NULL, NULL, size);
+
+        quota_inode_ctx_get (local->loc.inode, -1, this, NULL, NULL,
+                             &ctx, 1);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->buf = *buf;
+
+                dentry = __quota_dentry_new (ctx, (char *)local->loc.name,
+                                             local->loc.parent->gfid);
+                if (dentry == NULL) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "cannot create a new dentry (name:%s) for "
+                                "inode(gfid:%s)", local->loc.name,
+                                uuid_utoa (local->loc.inode->gfid));
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                }
+        }
+        UNLOCK (&ctx->lock);
 
-	STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
-	return 0;
+out:
+        QUOTA_STACK_UNWIND (symlink, frame, op_ret, op_errno, inode, buf,
+                            preparent, postparent, xdata);
+
+        return 0;
 }
 
 
 int
-quota_symlink (call_frame_t *frame, xlator_t *this,
-	       const char *linkpath, loc_t *loc, dict_t *params)
+quota_symlink_helper (call_frame_t *frame, xlator_t *this, const char *linkpath,
+                      loc_t *loc, mode_t umask, dict_t *xdata)
 {
-	struct quota_priv *priv = NULL;
+        quota_local_t *local    = NULL;
+        int32_t        op_errno = EINVAL;
 
-	priv = this->private;
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto unwind;
+        }
 
-	if (gf_quota_check_free_disk (this) == -1) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"min-free-disk limit (%u) crossed, current available is %u",
-			priv->min_free_disk_limit, priv->current_free_disk);
-		STACK_UNWIND_STRICT (symlink, frame, -1, ENOSPC, NULL, NULL,
-                                     NULL, NULL);
-		return 0;
-		
-	}
-        if (priv->current_disk_usage > priv->disk_usage_limit) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"Disk usage limit (%"PRIu64") crossed, current usage is %"PRIu64"",
-			priv->disk_usage_limit, priv->current_disk_usage);
-		STACK_UNWIND_STRICT (symlink, frame, -1, ENOSPC, NULL, NULL,
-                                     NULL, NULL);
-		return 0;
+        if (local->op_ret == -1) {
+                op_errno = local->op_errno;
+                goto unwind;
         }
 
-	STACK_WIND (frame, quota_symlink_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->symlink,
-		    linkpath, loc, params);
-	return 0;
+        STACK_WIND (frame, quota_symlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->symlink, linkpath, loc, umask,
+                    xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (symlink, frame, -1, op_errno, NULL, NULL,
+                            NULL, NULL, NULL);
+        return 0;
 }
 
 
 int
-quota_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int32_t op_ret, int32_t op_errno,
-		  fd_t *fd, inode_t *inode, struct iatt *buf,
-                  struct iatt *preparent, struct iatt *postparent)
+quota_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
+               loc_t *loc, mode_t umask, dict_t *xdata)
 {
-	struct quota_priv *priv = this->private;
+        int32_t          ret      = -1;
+        int32_t          op_errno = ENOMEM;
+        quota_local_t   *local    = NULL;
+        call_stub_t     *stub     = NULL;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto err;
+        }
 
-	if ((op_ret >= 0) && priv->disk_usage_limit) {
-		gf_quota_usage_add (this, buf->ia_blocks * 512);
+        frame->local = local;
 
-		fd_ctx_set (fd, this, 1);
-	}
+        ret = loc_copy (&local->loc, loc);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto err;
+        }
+
+        local->link_count = 1;
+
+        stub = fop_symlink_stub (frame, quota_symlink_helper, linkpath, loc,
+                                 umask, xdata);
+        if (stub == NULL) {
+                goto err;
+        }
+
+        local->stub = stub;
+        local->delta = strlen (linkpath);
+
+        quota_check_limit (frame, loc->parent, this, NULL, NULL);
+
+        stub = NULL;
+
+        LOCK (&local->lock);
+        {
+                if (local->validate_count == 0) {
+                        stub = local->stub;
+                        local->stub = NULL;
+                }
+
+                local->link_count = 0;
+        }
+        UNLOCK (&local->lock);
+
+        if (stub != NULL) {
+                call_resume (stub);
+        }
+
+        return 0;
 
-	STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
-                             preparent, postparent);
-	return 0;
+err:
+        QUOTA_STACK_UNWIND (symlink, frame, -1, op_errno, NULL, NULL, NULL,
+                            NULL, NULL);
+
+        return 0;
 }
 
 
-int
-quota_create (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, int32_t flags, mode_t mode, fd_t *fd, dict_t *params)
+int32_t
+quota_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                    struct iatt *postbuf, dict_t *xdata)
 {
-	struct quota_priv *priv = NULL;
+        quota_local_t     *local = NULL;
+        int64_t            delta = 0;
+        quota_inode_ctx_t *ctx   = NULL;
 
-	priv = this->private;
+        if (op_ret < 0) {
+                goto out;
+        }
 
-	if (gf_quota_check_free_disk (this) == -1) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"min-free-disk limit (%u) crossed, current available is %u",
-			priv->min_free_disk_limit, priv->current_free_disk);
-		STACK_UNWIND_STRICT (create, frame, -1, ENOSPC, NULL, NULL, NULL,
-                                     NULL, NULL);
-		return 0;
-		
-	}
-        if (priv->current_disk_usage > priv->disk_usage_limit) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"Disk usage limit (%"PRIu64") crossed, current usage is %"PRIu64"",
-			priv->disk_usage_limit, priv->current_disk_usage);
-		STACK_UNWIND_STRICT (create, frame, -1, ENOSPC, NULL, NULL, NULL,
-                                     NULL, NULL);
-		return 0;
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto out;
+        }
+
+        delta = (postbuf->ia_blocks - prebuf->ia_blocks) * 512;
+
+        quota_update_size (this, local->loc.inode, NULL, NULL, delta);
+
+        quota_inode_ctx_get (local->loc.inode, -1, this, NULL, NULL,
+                             &ctx, 0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->buf = *postbuf;
         }
+        UNLOCK (&ctx->lock);
 
-	STACK_WIND (frame, quota_create_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->create,
-		    loc, flags, mode, fd, params);
-	return 0;
+out:
+        QUOTA_STACK_UNWIND (truncate, frame, op_ret, op_errno, prebuf,
+                            postbuf, xdata);
+        return 0;
 }
 
 
-int
-quota_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int32_t op_ret, int32_t op_errno, fd_t *fd)
+int32_t
+quota_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+                dict_t *xdata)
 {
+        int32_t          ret   = -1;
+        quota_local_t   *local = NULL;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto err;
+        }
+
+        frame->local = local;
+
+        ret =  loc_copy (&local->loc, loc);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto err;
+        }
 
-	if (op_ret >= 0)
-		fd_ctx_set (fd, this, 1);
+        STACK_WIND (frame, quota_truncate_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
 
-	STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd);
-	return 0;
+        return 0;
+err:
+        QUOTA_STACK_UNWIND (truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
 }
 
 
-int
-quota_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-            fd_t *fd, int32_t wbflags)
+int32_t
+quota_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
 {
-	STACK_WIND (frame, quota_open_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->open,
-		    loc, flags, fd, wbflags);
-	return 0;
+        quota_local_t     *local = NULL;
+        int64_t            delta = 0;
+        quota_inode_ctx_t *ctx   = NULL;
+
+        if (op_ret < 0) {
+                goto out;
+        }
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto out;
+        }
+
+        delta = (postbuf->ia_blocks - prebuf->ia_blocks) * 512;
+
+        quota_update_size (this, local->loc.inode, NULL, NULL, delta);
+
+        quota_inode_ctx_get (local->loc.inode, -1, this, NULL, NULL,
+                             &ctx, 0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->buf = *postbuf;
+        }
+        UNLOCK (&ctx->lock);
+
+out:
+        QUOTA_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, prebuf,
+                            postbuf, xdata);
+        return 0;
 }
 
 
-int
-quota_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                  struct iatt *postbuf)
+int32_t
+quota_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+                 dict_t *xdata)
 {
-	struct quota_priv *priv = NULL;
-	struct quota_local *local = NULL;
+        quota_local_t   *local = NULL;
 
+        local = quota_local_new ();
+        if (local == NULL)
+                goto err;
 
-	priv = this->private;
-	local = frame->local;
+        frame->local = local;
 
-	if (priv->disk_usage_limit) {
-		if (op_ret >= 0) { 
-			gf_quota_usage_add (this, (postbuf->ia_blocks -
-						   prebuf->ia_blocks) * 512);
-		}
-		fd_unref (local->fd);
-		iobref_unref (local->iobref);
-	}
+        local->loc.inode = inode_ref (fd->inode);
 
-	STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf);
-	return 0;
+        STACK_WIND (frame, quota_ftruncate_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+
+        return 0;
+err:
+        QUOTA_STACK_UNWIND (ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
 }
 
 
-int
-quota_writev_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int32_t op_ret,	int32_t op_errno, struct iatt *buf)
+int32_t
+quota_send_dir_limit_to_cli (call_frame_t *frame, xlator_t *this,
+                             inode_t *inode, const char *name)
 {
-	struct quota_local *local = NULL;
-	struct quota_priv  *priv = NULL;
-	int                 iovlen = 0;
+        int32_t            ret               = 0;
+        char               dir_limit [1024]  = {0, };
+        dict_t            *dict              = NULL;
+        quota_inode_ctx_t *ctx               = NULL;
+        uint64_t           value             = 0;
+
+        ret = inode_ctx_get (inode, this, &value);
+        if (ret < 0)
+                goto out;
+
+        ctx = (quota_inode_ctx_t *)(unsigned long)value;
+        snprintf (dir_limit, 1024, "%"PRId64",%"PRId64, ctx->size, ctx->limit);
+
+        dict = dict_new ();
+        if (dict == NULL) {
+                ret = -1;
+                goto out;
+        }
 
+        ret = dict_set_str (dict, (char *) name, dir_limit);
+        if (ret < 0)
+                goto out;
 
-	local = frame->local;
-	priv = this->private;
+        gf_log (this->name, GF_LOG_INFO, "str = %s", dir_limit);
 
-	if (op_ret >= 0) {
-		if (priv->current_disk_usage > priv->disk_usage_limit) {
-			iovlen = iov_length (local->vector, local->count);
+        QUOTA_STACK_UNWIND (getxattr, frame, 0, 0, dict, NULL);
 
-			if (iovlen > (buf->ia_blksize - (buf->ia_size % buf->ia_blksize))) {
-				fd_unref (local->fd);
-				iobref_unref (local->iobref);
-				STACK_UNWIND_STRICT (writev, frame, -1, ENOSPC,
-                                                     NULL, NULL);
-				return 0;
-			}
-		}
-		local->stbuf = *buf;
-	}
-	
-	STACK_WIND (frame, quota_writev_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->writev,
-		    local->fd, local->vector, local->count, local->offset,
-                    local->iobref);
+        ret = 0;
 
-	return 0;
+out:
+        return ret;
 }
 
 
-int
-quota_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
-	      struct iovec *vector, int32_t count, off_t off,
-              struct iobref *iobref)
-{
-	struct quota_local *local = NULL;
-	struct quota_priv  *priv  = NULL;
-        int                 i     = 0;
-
-	priv = this->private;
-
-	if (gf_quota_check_free_disk (this) == -1) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"min-free-disk limit (%u) crossed, current available is %u",
-			priv->min_free_disk_limit, priv->current_free_disk);
-		STACK_UNWIND_STRICT (writev, frame, -1, ENOSPC,
-                                     NULL, NULL);
-		return 0;
-	}
+int32_t
+quota_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 const char *name, dict_t *xdata)
+{
+        int32_t ret     = 0;
 
-	if (priv->disk_usage_limit) {
-		local = GF_CALLOC (1, sizeof (struct quota_local),
-                                   gf_quota_mt_quota_local);
-		local->fd     = fd_ref (fd);
-		local->iobref = iobref_ref (iobref);
-                for (i = 0; i < count; i++) {
-                        local->vector[i].iov_base = vector[i].iov_base;
-                        local->vector[i].iov_len = vector[i].iov_len;
+        if (name && strcasecmp (name, "trusted.limit.list") == 0) {
+                ret = quota_send_dir_limit_to_cli (frame, this, fd->inode,
+                                                   name);
+                if (ret == 0) {
+                        return 0;
                 }
+        }
 
-                local->count = count;
-		local->offset = off;
-		frame->local  = local;
+        STACK_WIND (frame, default_fgetxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+        return 0;
+}
 
-		STACK_WIND (frame, quota_writev_fstat_cbk,
-			    FIRST_CHILD(this),
-			    FIRST_CHILD(this)->fops->fstat, fd);
-		return 0;
-	}
 
-	STACK_WIND (frame, quota_writev_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->writev,
-		    fd, vector, count, off, iobref);
-	return 0;
+int32_t
+quota_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                const char *name, dict_t *xdata)
+{
+        int32_t ret     = 0;
+
+        if ((name != NULL) && strcasecmp (name, "trusted.limit.list") == 0) {
+                ret = quota_send_dir_limit_to_cli (frame, this, loc->inode,
+                                                   name);
+                if (ret == 0)
+                        return 0;
+        }
+
+        STACK_WIND (frame, default_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+        return 0;
 }
 
 
-int
-quota_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno)
+int32_t
+quota_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_CRITICAL, 
-			"failed to remove the disk-usage value: %s",
-			strerror (op_errno));
-	} 
-	
-	STACK_DESTROY (frame->root);
-	return 0;
+        quota_local_t     *local = NULL;
+        quota_inode_ctx_t *ctx   = NULL;
+
+        if (op_ret < 0) {
+                goto out;
+        }
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto out;
+        }
+
+        quota_inode_ctx_get (local->loc.inode, -1, this, NULL, NULL,
+                             &ctx, 0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                if (buf)
+                        ctx->buf = *buf;
+        }
+        UNLOCK (&ctx->lock);
+
+out:
+        QUOTA_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
 }
 
 
-int
-quota_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno)
+int32_t
+quota_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-        dict_t            *dict = NULL;
+        quota_local_t *local = NULL;
+        int32_t        ret   = -1;
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_CRITICAL, 
-			"failed to set the disk-usage value: %s",
-			strerror (op_errno));
-	} 
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto unwind;
+        }
 
-        if (cookie) {
-                dict = (dict_t *) cookie;
-                dict_unref (dict);
+        frame->local = local;
+        ret = loc_copy (&local->loc, loc);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto unwind;
         }
 
-	STACK_DESTROY (frame->root);
-	return 0;
+        STACK_WIND (frame, quota_stat_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->stat, loc, xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (stat, frame, -1, ENOMEM, NULL, NULL);
+        return 0;
 }
 
 
-int
-quota_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int32_t op_ret, int32_t op_errno, struct statvfs *statvfs)
+int32_t
+quota_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                 dict_t *xdata)
 {
-	struct quota_priv *priv = NULL;
-	uint64_t           f_blocks = 0;
-	int64_t            f_bfree = 0;
-	uint64_t           f_bused = 0;
+        quota_local_t     *local = NULL;
+        quota_inode_ctx_t *ctx   = NULL;
 
+        if (op_ret < 0) {
+                goto out;
+        }
 
-	priv = this->private;
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto out;
+        }
+
+        quota_inode_ctx_get (local->loc.inode, -1, this, NULL, NULL,
+                             &ctx, 0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                if (buf)
+                        ctx->buf = *buf;
+        }
+        UNLOCK (&ctx->lock);
+
+out:
+        QUOTA_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+
+int32_t
+quota_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+        quota_local_t *local = NULL;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto unwind;
+        }
+
+        frame->local = local;
+
+        local->loc.inode = inode_ref (fd->inode);
+
+        STACK_WIND (frame, quota_fstat_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fstat, fd, xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (fstat, frame, -1, ENOMEM, NULL, NULL);
+        return 0;
+}
 
-	if (op_ret != 0)
-		goto unwind;
 
-	f_blocks = priv->disk_usage_limit / statvfs->f_frsize;
-	f_bused = priv->current_disk_usage / statvfs->f_frsize;
+int32_t
+quota_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, const char *path,
+                    struct iatt *buf, dict_t *xdata)
+{
+        quota_local_t     *local = NULL;
+        quota_inode_ctx_t *ctx   = NULL;
+
+        if (op_ret < 0) {
+                goto out;
+        }
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto out;
+        }
+
+        quota_inode_ctx_get (local->loc.inode, -1, this, NULL, NULL,
+                             &ctx, 0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->buf = *buf;
+        }
+        UNLOCK (&ctx->lock);
+
+out:
+        QUOTA_STACK_UNWIND (readlink, frame, op_ret, op_errno, path, buf, xdata);
+        return 0;
+}
+
+
+int32_t
+quota_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+                dict_t *xdata)
+{
+        quota_local_t *local = NULL;
+        int32_t        ret   = -1;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto unwind;
+        }
+
+        frame->local = local;
+
+        ret = loc_copy (&local->loc, loc);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto unwind;
+        }
+
+        STACK_WIND (frame, quota_readlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readlink, loc, size, xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (readlink, frame, -1, ENOMEM, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+quota_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, struct iovec *vector,
+                 int32_t count, struct iatt *buf, struct iobref *iobref,
+                 dict_t *xdata)
+{
+        quota_local_t     *local = NULL;
+        quota_inode_ctx_t *ctx   = NULL;
+
+        if (op_ret < 0) {
+                goto out;
+        }
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto out;
+        }
+
+        quota_inode_ctx_get (local->loc.inode, -1, this, NULL, NULL,
+                             &ctx, 0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->buf = *buf;
+        }
+        UNLOCK (&ctx->lock);
+
+out:
+        QUOTA_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count,
+                            buf, iobref, xdata);
+        return 0;
+}
+
+
+int32_t
+quota_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+             off_t offset, uint32_t flags, dict_t *xdata)
+{
+        quota_local_t *local = NULL;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto unwind;
+        }
 
-	if (f_blocks && (f_blocks < statvfs->f_blocks))
-		statvfs->f_blocks = f_blocks;
+        frame->local = local;
 
-	f_bfree = (statvfs->f_blocks - f_bused);
+        local->loc.inode = inode_ref (fd->inode);
 
-	if (f_bfree >= 0)
-		statvfs->f_bfree = statvfs->f_bavail = f_bfree;
-	else
-		statvfs->f_bfree = statvfs->f_bavail = 0;
+        STACK_WIND (frame, quota_readv_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readv, fd, size, offset, flags,
+                    xdata);
+        return 0;
 
 unwind:
-	STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, statvfs);
-	return 0;
+        QUOTA_STACK_UNWIND (readv, frame, -1, ENOMEM, NULL, -1, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+quota_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                 struct iatt *postbuf, dict_t *xdata)
+{
+        quota_local_t     *local = NULL;
+        quota_inode_ctx_t *ctx   = NULL;
+
+        if (op_ret < 0) {
+                goto out;
+        }
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto out;
+        }
+
+        quota_inode_ctx_get (local->loc.inode, -1, this, NULL, NULL,
+                             &ctx, 0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->buf = *postbuf;
+        }
+        UNLOCK (&ctx->lock);
+
+out:
+        QUOTA_STACK_UNWIND (fsync, frame, op_ret, op_errno, prebuf, postbuf,
+                            xdata);
+        return 0;
+}
+
+
+int32_t
+quota_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+             dict_t *xdata)
+{
+        quota_local_t *local = NULL;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto unwind;
+        }
+
+        local->loc.inode = inode_ref (fd->inode);
+
+        frame->local = local;
+
+        STACK_WIND (frame, quota_fsync_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsync, fd, flags, xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (fsync, frame, -1, ENOMEM, NULL, NULL, NULL);
+        return 0;
+
+}
+
+
+int32_t
+quota_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                   struct iatt *statpost, dict_t *xdata)
+{
+        quota_local_t     *local = NULL;
+        quota_inode_ctx_t *ctx   = NULL;
+
+        if (op_ret < 0) {
+                goto out;
+        }
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto out;
+        }
+
+        quota_inode_ctx_get (local->loc.inode, -1, this, NULL, NULL,
+                             &ctx, 0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                if (statpost)
+                        ctx->buf = *statpost;
+        }
+        UNLOCK (&ctx->lock);
+
+out:
+        QUOTA_STACK_UNWIND (setattr, frame, op_ret, op_errno, statpre,
+                            statpost, xdata);
+        return 0;
+}
+
+
+int32_t
+quota_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+               struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        quota_local_t *local = NULL;
+        int32_t        ret   = -1;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto unwind;
+        }
+
+        frame->local = local;
+
+        ret = loc_copy (&local->loc, loc);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto unwind;
+        }
+
+        STACK_WIND (frame, quota_setattr_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->setattr, loc, stbuf, valid, xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (setattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+quota_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                    struct iatt *statpost, dict_t *xdata)
+{
+        quota_local_t     *local = NULL;
+        quota_inode_ctx_t *ctx   = NULL;
+
+        if (op_ret < 0) {
+                goto out;
+        }
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto out;
+        }
+
+        quota_inode_ctx_get (local->loc.inode, -1, this, NULL, NULL,
+                             &ctx, 0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->buf = *statpost;
+        }
+        UNLOCK (&ctx->lock);
+
+out:
+        QUOTA_STACK_UNWIND (fsetattr, frame, op_ret, op_errno, statpre,
+                            statpost, xdata);
+        return 0;
+}
+
+
+int32_t
+quota_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        quota_local_t *local = NULL;
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto unwind;
+        }
+
+        frame->local = local;
+
+        local->loc.inode = inode_ref (fd->inode);
+
+        STACK_WIND (frame, quota_fsetattr_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->fsetattr, fd, stbuf, valid, xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (fsetattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+quota_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, inode_t *inode,
+                 struct iatt *buf, struct iatt *preparent,
+                 struct iatt *postparent, dict_t *xdata)
+{
+        int32_t            ret    = -1;
+        quota_local_t     *local  = NULL;
+        quota_inode_ctx_t *ctx    = NULL;
+        quota_dentry_t    *dentry = NULL;
+
+        local = frame->local;
+        if (op_ret < 0) {
+                goto unwind;
+        }
+
+        ret = quota_inode_ctx_get (inode, -1, this, NULL, buf, &ctx, 1);
+        if ((ret == -1) || (ctx == NULL)) {
+                gf_log (this->name, GF_LOG_WARNING, "cannot create quota "
+                        "context in inode (gfid:%s)", uuid_utoa (inode->gfid));
+                op_ret = -1;
+                op_errno = ENOMEM;
+                goto unwind;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->buf = *buf;
+
+                dentry = __quota_dentry_new (ctx, (char *)local->loc.name,
+                                             local->loc.parent->gfid);
+                if (dentry == NULL) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "cannot create a new dentry (name:%s) for "
+                                "inode(gfid:%s)", local->loc.name,
+                                uuid_utoa (local->loc.inode->gfid));
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        goto unlock;
+                }
+        }
+unlock:
+        UNLOCK (&ctx->lock);
+
+unwind:
+        QUOTA_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode,
+                            buf, preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
-quota_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
+quota_mknod_helper (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                    mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
 {
-	STACK_WIND (frame, quota_statfs_cbk,
-		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->statfs, loc);
+        quota_local_t *local    = NULL;
+        int32_t        op_errno = EINVAL;
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto unwind;
+        }
 
-	return 0;
+        if (local->op_ret == -1) {
+                op_errno = local->op_errno;
+                goto unwind;
+        }
+
+        STACK_WIND (frame, quota_mknod_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask,
+                    xdata);
+
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL,
+                            NULL, NULL, NULL);
+        return 0;
 }
 
 
 int
-quota_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno, dict_t *value)
-{
-	data_t *data = NULL;
-	struct quota_priv *priv = this->private;
-	
-	if (op_ret >= 0) {
-		data = dict_get (value, "trusted.glusterfs-quota-du");
-		if (data) {
-			LOCK (&priv->lock);
-			{
-				priv->current_disk_usage = data_to_uint64 (data);
-			}
-			UNLOCK (&priv->lock);
+quota_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+             dev_t rdev, mode_t umask, dict_t *xdata)
+{
+        int32_t            ret            = -1;
+        quota_local_t     *local          = NULL;
+        call_stub_t       *stub           = NULL;
 
-			return 0;
-		}
-	} 
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto err;
+        }
+
+        frame->local = local;
+
+        ret = loc_copy (&local->loc, loc);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "loc_copy failed");
+                goto err;
+        }
+
+        stub = fop_mknod_stub (frame, quota_mknod_helper, loc, mode, rdev,
+                               umask, xdata);
+        if (stub == NULL) {
+                goto err;
+        }
+
+        local->link_count = 1;
+        local->stub = stub;
+        local->delta = 0;
+
+        quota_check_limit (frame, loc->parent, this, NULL, NULL);
+
+        stub = NULL;
+
+        LOCK (&local->lock);
+        {
+                local->link_count = 0;
+                if (local->validate_count == 0) {
+                        stub = local->stub;
+                        local->stub = NULL;
+                }
+        }
+        UNLOCK (&local->lock);
+
+        if (stub != NULL) {
+                call_resume (stub);
+        }
+
+        return 0;
+err:
+        QUOTA_STACK_UNWIND (mknod, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+                            NULL);
 
-	STACK_DESTROY (frame->root);
+        return 0;
+}
 
-	return 0;
+int
+quota_setxattr_cbk (call_frame_t *frame, void *cookie,
+                    xlator_t *this, int op_ret, int op_errno, dict_t *xdata)
+{
+        QUOTA_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
+int
+quota_setxattr (call_frame_t *frame, xlator_t *this,
+                loc_t *loc, dict_t *dict, int flags, dict_t *xdata)
+{
+        int             op_errno = EINVAL;
+        int             op_ret   = -1;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.quota*", dict,
+                                   op_errno, err);
+
+        STACK_WIND (frame, quota_setxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr,
+                    loc, dict, flags, xdata);
+        return 0;
+err:
+        QUOTA_STACK_UNWIND (setxattr, frame, op_ret, op_errno, NULL);
+        return 0;
+}
 
-void
-gf_quota_get_disk_usage (xlator_t *this)
+int
+quota_fsetxattr_cbk (call_frame_t *frame, void *cookie,
+                     xlator_t *this, int op_ret, int op_errno, dict_t *xdata)
 {
-	call_frame_t *frame = NULL;
-	call_pool_t  *pool = NULL;
+        QUOTA_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
 
-	struct quota_priv *priv = NULL;
+int
+quota_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 dict_t *dict, int flags, dict_t *xdata)
+{
+        int32_t         op_ret   = -1;
+        int32_t         op_errno = EINVAL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.quota*", dict,
+                                   op_errno, err);
+
+        STACK_WIND (frame, quota_fsetxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr,
+                    fd, dict, flags, xdata);
+        return 0;
+ err:
+        QUOTA_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, NULL);
+        return 0;
+}
 
-	pool = this->ctx->pool;
-	frame = create_frame (this, pool);
-	priv = this->private;
 
-	STACK_WIND (frame, quota_getxattr_cbk,
-		    this->children->xlator,
-		    this->children->xlator->fops->getxattr,
-		    &(priv->root_loc),
-		    "trusted.glusterfs-quota-du");
-	return ;
+int
+quota_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        QUOTA_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
+int
+quota_removexattr (call_frame_t *frame, xlator_t *this,
+                   loc_t *loc, const char *name, dict_t *xdata)
+{
+        int32_t         op_errno = EINVAL;
 
-void
-gf_quota_cache_sync (xlator_t *this)
+        VALIDATE_OR_GOTO (this, err);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.quota*",
+                                 name, op_errno, err);
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (loc, err);
+
+        STACK_WIND (frame, quota_removexattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->removexattr,
+                    loc, name, xdata);
+        return 0;
+err:
+        QUOTA_STACK_UNWIND (removexattr, frame, -1,  op_errno, NULL);
+        return 0;
+}
+
+
+int
+quota_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	struct quota_priv *priv = NULL;
-	call_frame_t      *frame = NULL;
-	dict_t            *dict = get_new_dict ();
+        QUOTA_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+quota_fremovexattr (call_frame_t *frame, xlator_t *this,
+                    fd_t *fd, const char *name, dict_t *xdata)
+{
+        int32_t         op_ret   = -1;
+        int32_t         op_errno = EINVAL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.quota*",
+                                 name, op_errno, err);
+
+        STACK_WIND (frame, quota_fremovexattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fremovexattr,
+                    fd, name, xdata);
+        return 0;
+ err:
+        QUOTA_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, NULL);
+        return 0;
+}
 
 
+int32_t
+quota_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct statvfs *buf,
+                  dict_t *xdata)
+{
+	inode_t           *root_inode = NULL;
+        quota_priv_t      *priv       = NULL;
+        uint64_t           value      = 0;
+        quota_inode_ctx_t *ctx        = NULL;
+        limits_t          *limit_node = NULL;
+	int64_t            usage      = -1;
+	int64_t            avail      = -1;
+        int64_t            blocks     = 0;
+
+	root_inode = cookie;
+
+        /* This fop will fail mostly in case of client disconnect's,
+         * which is already logged. Hence, not logging here */
+        if (op_ret == -1)
+                goto unwind;
+	/*
+	 * We should never get here unless quota_statfs (below) sent us a
+	 * cookie, and it would only do so if the value was non-NULL.  This
+	 * check is therefore just routine defensive coding.
+	 */
+	if (!root_inode) {
+		gf_log(this->name,GF_LOG_WARNING,
+		       "null inode, cannot adjust for quota");
+		goto unwind;
+	}
+	if (!root_inode->table || (root_inode != root_inode->table->root)) {
+		gf_log(this->name,GF_LOG_WARNING,
+		       "non-root inode, cannot adjust for quota");
+		goto unwind;
+	}
 
-	priv = this->private;
+        inode_ctx_get (root_inode, this, &value);
+	if (!value) {
+		goto unwind;
+	}
+        ctx = (quota_inode_ctx_t *)(unsigned long)value;
+	usage = (ctx->size) / buf->f_bsize;
+        priv = this->private;
+
+        list_for_each_entry (limit_node, &priv->limit_head, limit_list) {
+		/* Notice that this only works for volume-level quota. */
+                if (strcmp (limit_node->path, "/") == 0) {
+                        blocks = limit_node->value / buf->f_bsize;
+                        if (usage > blocks) {
+                                break;
+                        }
+
+			buf->f_blocks = blocks;
+			avail = buf->f_blocks - usage;
+			if (buf->f_bfree > avail) {
+				buf->f_bfree = avail;
+			}
+			/*
+			 * We have to assume that the total assigned quota
+			 * won't cause us to dip into the reserved space,
+			 * because dealing with the overcommitted cases is
+			 * just too hairy (especially when different bricks
+			 * might be using different reserved percentages and
+			 * such).
+			 */
+			buf->f_bavail = buf->f_bfree;
+			break;
+                }
+        }
 
-	frame = create_frame (this, this->ctx->pool);
-	dict_set (dict, "trusted.glusterfs-quota-du", 
-		  data_from_uint64 (priv->current_disk_usage));
+unwind:
+	if (root_inode) {
+		inode_unref(root_inode);
+	}
+        STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
 
-        dict_ref (dict);
 
-	STACK_WIND_COOKIE (frame, quota_setxattr_cbk,
-                           (void *) (dict_t *) dict,
-                           this->children->xlator,
-                           this->children->xlator->fops->setxattr,
-                           &(priv->root_loc), dict, 0);
+int32_t
+quota_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+	inode_t *root_inode = NULL;
+        quota_priv_t    *priv = NULL;
+
+        priv = this->private;
+
+	if (priv->consider_statfs && loc->inode) {
+		root_inode = loc->inode->table->root;
+		inode_ref(root_inode);
+		STACK_WIND_COOKIE (frame, quota_statfs_cbk, root_inode,
+				   FIRST_CHILD(this),
+				   FIRST_CHILD(this)->fops->statfs, loc, xdata);
+	}
+	else {
+		/*
+		 * We have to make sure that we never get to quota_statfs_cbk
+		 * with a cookie that points to something other than an inode,
+		 * which is exactly what would happen with STACK_UNWIND using
+		 * that as a callback.  Therefore, use default_statfs_cbk in
+		 * this case instead.
+                 *
+                 * Also if the option deem-statfs is not set to "on" don't
+                 * bother calculating quota limit on / in statfs_cbk.
+		 */
+                if (priv->consider_statfs)
+                        gf_log(this->name,GF_LOG_WARNING,
+                               "missing inode, cannot adjust for quota");
+		STACK_WIND (frame, default_statfs_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->statfs, loc, xdata);
+	}
+        return 0;
 }
 
 
 int
-quota_release (xlator_t *this, fd_t *fd)
+quota_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int op_ret, int op_errno, gf_dirent_t *entries,
+                    dict_t *xdata)
 {
-	gf_quota_cache_sync (this);
+        gf_dirent_t *entry = NULL;
 
-	return 0;
+        if (op_ret <= 0)
+                goto unwind;
+
+        list_for_each_entry (entry, &entries->list, list) {
+                /* TODO: fill things */
+        }
+
+unwind:
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
+
+        return 0;
 }
+int
+quota_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+                off_t offset, dict_t *dict)
+{
+        int ret = 0;
 
+        if (dict) {
+                ret = dict_set_uint64 (dict, QUOTA_SIZE_KEY, 0);
+                if (ret < 0) {
+                        goto err;
+                }
+        }
+
+        STACK_WIND (frame, quota_readdirp_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdirp,
+                    fd, size, offset, dict);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (readdirp, frame, -1, EINVAL, NULL, NULL);
+        return 0;
+}
 
-/* notify */
 int32_t
-notify (xlator_t *this,
-	int32_t event,
-	void *data,
-	...)
+quota_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                    struct iatt *postbuf, dict_t *xdata)
 {
-	default_notify (this, event, data);
-	return 0;
+        int32_t                  ret            = 0;
+        uint64_t                 ctx_int        = 0;
+        quota_inode_ctx_t       *ctx            = NULL;
+        quota_local_t           *local          = NULL;
+        quota_dentry_t          *dentry         = NULL;
+        int64_t                  delta          = 0;
+
+        local = frame->local;
+
+        if ((op_ret < 0) || (local == NULL)) {
+                goto out;
+        }
+
+        ret = inode_ctx_get (local->loc.inode, this, &ctx_int);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to get the context", local->loc.path);
+                goto out;
+        }
+
+        ctx = (quota_inode_ctx_t *)(unsigned long) ctx_int;
+
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in %s (gfid:%s)",
+                        local->loc.path, uuid_utoa (local->loc.inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->buf = *postbuf;
+        }
+        UNLOCK (&ctx->lock);
+
+        list_for_each_entry (dentry, &ctx->parents, next) {
+                delta = (postbuf->ia_blocks - prebuf->ia_blocks) * 512;
+                quota_update_size (this, local->loc.inode,
+                                   dentry->name, dentry->par, delta);
+        }
+
+out:
+        QUOTA_STACK_UNWIND (fallocate, frame, op_ret, op_errno, prebuf, postbuf,
+                            xdata);
+
+        return 0;
 }
 
 int32_t
-quota_lookup_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    inode_t *inode,
-		    struct iatt *buf,
-                    dict_t *dict,
-                    struct iatt *postparent)
-{
-	STACK_UNWIND_STRICT (
-                     lookup, 
-                     frame,
-		      op_ret,
-		      op_errno,
-		      inode,
-		      buf,
-                      dict,
-                      postparent);
-	return 0;
+quota_fallocate_helper(call_frame_t *frame, xlator_t *this, fd_t *fd,
+		       int32_t mode, off_t offset, size_t len, dict_t *xdata)
+{
+        quota_local_t *local    = NULL;
+        int32_t        op_errno = EINVAL;
+
+        local = frame->local;
+        if (local == NULL) {
+                gf_log (this->name, GF_LOG_WARNING, "local is NULL");
+                goto unwind;
+        }
+
+        if (local->op_ret == -1) {
+                op_errno = local->op_errno;
+                goto unwind;
+        }
+
+        STACK_WIND (frame, quota_fallocate_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
+		    xdata);
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
 }
 
 int32_t
-quota_lookup (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		dict_t *xattr_req)
-{
-	struct quota_priv *priv = NULL;
-
-	priv = this->private;
-
-	if (priv->only_first_time) {
-		if (strcmp (loc->path, "/") == 0) {
-			loc_copy(&(priv->root_loc), loc); 
-			priv->only_first_time = 0;
-			if (priv->disk_usage_limit)
-				gf_quota_get_disk_usage (this);
-		}
-	}
+quota_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+		off_t offset, size_t len, dict_t *xdata)
+{
+        int32_t            ret     = -1, op_errno = EINVAL;
+        int32_t            parents = 0;
+        quota_local_t     *local   = NULL;
+        quota_inode_ctx_t *ctx     = NULL;
+        quota_priv_t      *priv    = NULL;
+        call_stub_t       *stub    = NULL;
+        quota_dentry_t    *dentry  = NULL;
+
+        GF_ASSERT (frame);
+        GF_VALIDATE_OR_GOTO ("quota", this, unwind);
+        GF_VALIDATE_OR_GOTO (this->name, fd, unwind);
+
+        local = quota_local_new ();
+        if (local == NULL) {
+                goto unwind;
+        }
+
+        frame->local = local;
+        local->loc.inode = inode_ref (fd->inode);
+
+        ret = quota_inode_ctx_get (fd->inode, -1, this, NULL, NULL, &ctx, 0);
+        if (ctx == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "quota context not set in inode (gfid:%s)",
+                        uuid_utoa (fd->inode->gfid));
+                goto unwind;
+        }
+
+        stub = fop_fallocate_stub(frame, quota_fallocate_helper, fd, mode, offset, len,
+				  xdata); 
+        if (stub == NULL) {
+                op_errno = ENOMEM;
+                goto unwind;
+        }
+
+        priv = this->private;
+        GF_VALIDATE_OR_GOTO (this->name, priv, unwind);
+
+        LOCK (&ctx->lock);
+        {
+                list_for_each_entry (dentry, &ctx->parents, next) {
+                        parents++;
+                }
+        }
+        UNLOCK (&ctx->lock);
+
+	/*
+	 * Note that by using len as the delta we're assuming the range from
+	 * offset to offset+len has not already been allocated. This can result
+	 * in ENOSPC errors attempting to allocate an already allocated range.
+	 */
+        local->delta = len;
+        local->stub = stub;
+        local->link_count = parents;
+
+        list_for_each_entry (dentry, &ctx->parents, next) {
+                ret = quota_check_limit (frame, fd->inode, this, dentry->name,
+                                         dentry->par);
+                if (ret == -1) {
+                        break;
+                }
+        }
+
+        stub = NULL;
+
+        LOCK (&local->lock);
+        {
+                local->link_count = 0;
+                if (local->validate_count == 0) {
+                        stub = local->stub;
+                        local->stub = NULL;
+                }
+        }
+        UNLOCK (&local->lock);
+
+        if (stub != NULL) {
+                call_resume (stub);
+        }
 
-	STACK_WIND (frame,
-		    quota_lookup_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->lookup,
-		    loc,
-		    xattr_req);
-	return 0;
+        return 0;
+
+unwind:
+        QUOTA_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
 }
 
+
 int32_t
 mem_acct_init (xlator_t *this)
 {
@@ -1032,174 +3150,354 @@ mem_acct_init (xlator_t *this)
                 return ret;
 
         ret = xlator_mem_acct_init (this, gf_quota_mt_end + 1);
-        
+
         if (ret != 0) {
-                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
-                        "failed");
+                gf_log (this->name, GF_LOG_WARNING, "Memory accounting"
+                        "init failed");
                 return ret;
         }
 
         return ret;
 }
 
+
+int32_t
+quota_forget (xlator_t *this, inode_t *inode)
+{
+        int32_t               ret     = 0;
+        uint64_t              ctx_int = 0;
+        quota_inode_ctx_t    *ctx     = NULL;
+        quota_dentry_t       *dentry  = NULL, *tmp;
+
+        ret = inode_ctx_del (inode, this, &ctx_int);
+
+        if (ret < 0) {
+                return 0;
+        }
+
+        ctx = (quota_inode_ctx_t *) (long)ctx_int;
+
+        LOCK (&ctx->lock);
+        {
+                list_for_each_entry_safe (dentry, tmp, &ctx->parents, next) {
+                        __quota_dentry_free (dentry);
+                }
+        }
+        UNLOCK (&ctx->lock);
+
+        LOCK_DESTROY (&ctx->lock);
+
+        GF_FREE (ctx);
+
+        return 0;
+}
+
+
 int
-reconfigure (xlator_t *this, dict_t *options)
+quota_parse_limits (quota_priv_t *priv, xlator_t *this, dict_t *xl_options,
+                    struct list_head *old_list)
 {
+        int32_t       ret       = -1;
+        char         *str       = NULL;
+        char         *str_val   = NULL;
+        char         *path      = NULL, *saveptr = NULL;
+        uint64_t      value     = 0;
+        limits_t     *quota_lim = NULL, *old = NULL;
+        char         *last_colon= NULL;
+
+        ret = dict_get_str (xl_options, "limit-set", &str);
+
+        if (str) {
+                path = strtok_r (str, ",", &saveptr);
+
+                while (path) {
+                        last_colon = strrchr (path, ':');
+                        *last_colon = '\0';
+                        str_val = last_colon + 1;
+
+                        ret = gf_string2bytesize (str_val, &value);
+                        if (ret != 0)
+                                goto err;
+
+                        QUOTA_ALLOC_OR_GOTO (quota_lim, limits_t, err);
+
+                        quota_lim->path = path;
+
+                        quota_lim->value = value;
+
+                        gf_log (this->name, GF_LOG_INFO, "%s:%"PRId64,
+                                quota_lim->path, quota_lim->value);
+
+                        if (old_list != NULL) {
+                                list_for_each_entry (old, old_list,
+                                                     limit_list) {
+                                        if (strcmp (old->path, quota_lim->path)
+                                            == 0) {
+                                                uuid_copy (quota_lim->gfid,
+                                                           old->gfid);
+                                                break;
+                                        }
+                                }
+                        }
+
+                        LOCK (&priv->lock);
+                        {
+                                list_add_tail (&quota_lim->limit_list,
+                                               &priv->limit_head);
+                        }
+                        UNLOCK (&priv->lock);
+
+                        path = strtok_r (NULL, ",", &saveptr);
+                }
+        } else {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no \"limit-set\" option provided");
+        }
 
-	struct quota_priv *_private = NULL;
-	uint64_t	   disk_usage_limit;	
-	uint32_t   	   min_free_disk_limit;
-	data_t 		  *data = NULL;
-	int		   ret = 0;
-	
-	_private = this->private;
-
-	data = dict_get (options, "disk-usage-limit");
-	if (data) {
-		if (gf_string2bytesize (data->data, &disk_usage_limit) != 0) {
-                        gf_log (this->name, GF_LOG_ERROR, 
-                                "Reconfigure: Invalid number '%s' "
-				"for disk-usage limit", data->data);
-			//return -1;
-			ret = -1;
-			goto out;
+        LOCK (&priv->lock);
+        {
+                list_for_each_entry (quota_lim, &priv->limit_head, limit_list) {
+                        gf_log (this->name, GF_LOG_INFO, "%s:%"PRId64,
+                                quota_lim->path, quota_lim->value);
                 }
-		_private->disk_usage_limit = disk_usage_limit;
-		gf_log (this->name, GF_LOG_TRACE,
-                        "Reconfiguring disk-usage-limit %"PRIu64"",
-			disk_usage_limit);
+        }
+        UNLOCK (&priv->lock);
 
-	}
-	
-    
-        data = dict_get (options, "min-free-disk-limit");
-        if (data) {
-		if (gf_string2percent (data->data, &min_free_disk_limit) != 0){
-                        gf_log (this->name, GF_LOG_ERROR, 
-                                "Reconfigure : Invalid percent '%s'  for"
-				" min-free-disk-limit", data->data);
-			ret = -1;
-			goto out;
-		}
-
-		_private->min_free_disk_limit = min_free_disk_limit;
-		gf_log (this->name, GF_LOG_TRACE,
-                        "Reconfiguring min-free-disk-limit %d \%",
-			min_free_disk_limit);
-		
-        }
-out:	
-	return ret;
-
-}
-
-int32_t 
+        ret = 0;
+err:
+        return ret;
+}
+
+
+int32_t
 init (xlator_t *this)
 {
-	int     ret  = 0;
-        data_t *data = NULL;
-	struct quota_priv *_private = NULL;
+        int32_t       ret       = -1;
+        quota_priv_t *priv      = NULL;
+
+        if ((this->children == NULL)
+            || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "FATAL: quota (%s) not configured with "
+                        "exactly one child", this->name);
+                return -1;
+        }
 
-	if (!this->children || this->children->next) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"FATAL: quota should have exactly one child");
-		return -1;
-	}
-	
-	if (!this->parents) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"dangling volume. check volfile ");
-	}
+        if (this->parents == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dangling volume. check volfile");
+        }
+
+        QUOTA_ALLOC_OR_GOTO (priv, quota_priv_t, err);
+
+        INIT_LIST_HEAD (&priv->limit_head);
+
+        LOCK_INIT (&priv->lock);
+
+        this->private = priv;
+
+        ret = quota_parse_limits (priv, this, this->options, NULL);
+
+        if (ret) {
+                goto err;
+        }
+
+        GF_OPTION_INIT ("timeout", priv->timeout, int64, err);
+        GF_OPTION_INIT ("deem-statfs", priv->consider_statfs, bool, err);
+
+        this->local_pool = mem_pool_new (quota_local_t, 64);
+        if (!this->local_pool) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto err;
+        }
 
-	_private = GF_CALLOC (1, sizeof (struct quota_priv),
-                              gf_quota_mt_quota_priv);
-        _private->disk_usage_limit = 0;
-        data = dict_get (this->options, "disk-usage-limit");
-        if (data) {
-		if (gf_string2bytesize (data->data, &_private->disk_usage_limit) != 0) {
-                        gf_log (this->name, GF_LOG_ERROR, 
-                                "invalid number '%s' for disk-usage limit", data->data);
-			ret = -1;
-			goto out;
+        ret = 0;
+err:
+        return ret;
+}
+
+
+void
+__quota_reconfigure_inode_ctx (xlator_t *this, inode_t *inode, limits_t *limit)
+{
+        int                ret = -1;
+        quota_inode_ctx_t *ctx = NULL;
+
+        GF_VALIDATE_OR_GOTO ("quota", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, inode, out);
+        GF_VALIDATE_OR_GOTO (this->name, limit, out);
+
+        ret = quota_inode_ctx_get (inode, limit->value, this, NULL, NULL, &ctx,
+                                   1);
+        if ((ret == -1) || (ctx == NULL)) {
+                gf_log (this->name, GF_LOG_WARNING, "cannot create quota "
+                        "context in inode(gfid:%s)",
+                        uuid_utoa (inode->gfid));
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->limit = limit->value;
+        }
+        UNLOCK (&ctx->lock);
+
+out:
+        return;
+}
+
+
+void
+__quota_reconfigure (xlator_t *this, inode_table_t *itable, limits_t *limit)
+{
+        inode_t *inode = NULL;
+
+        if ((this == NULL) || (itable == NULL) || (limit == NULL)) {
+                goto out;
+        }
+
+        if (!uuid_is_null (limit->gfid)) {
+                inode = inode_find (itable, limit->gfid);
+        } else {
+                inode = inode_resolve (itable, limit->path);
+        }
+
+        if (inode != NULL) {
+                __quota_reconfigure_inode_ctx (this, inode, limit);
+        }
+
+out:
+        return;
+}
+
+
+int
+reconfigure (xlator_t *this, dict_t *options)
+{
+        int32_t           ret   = -1;
+        quota_priv_t     *priv  = NULL;
+        limits_t         *limit = NULL, *next = NULL, *new = NULL;
+        struct list_head  head  = {0, };
+        xlator_t         *top   = NULL;
+        char              found = 0;
+
+        priv = this->private;
+
+        INIT_LIST_HEAD (&head);
+
+        LOCK (&priv->lock);
+        {
+                list_splice_init (&priv->limit_head, &head);
+        }
+        UNLOCK (&priv->lock);
+
+        ret = quota_parse_limits (priv, this, options, &head);
+        if (ret == -1) {
+                gf_log ("quota", GF_LOG_WARNING,
+                        "quota reconfigure failed, "
+                        "new changes will not take effect");
+                goto out;
+        }
+
+        LOCK (&priv->lock);
+        {
+                top = ((glusterfs_ctx_t *)this->ctx)->active->top;
+                GF_ASSERT (top);
+
+                list_for_each_entry (limit, &priv->limit_head, limit_list) {
+                        __quota_reconfigure (this, top->itable, limit);
                 }
 
-		LOCK_INIT (&_private->lock);
-		_private->current_disk_usage = 0;
-	}
-	
-        _private->min_free_disk_limit = 0;
-        data = dict_get (this->options, "min-free-disk-limit");
-        if (data) {
-		if (gf_string2percent (data->data, &_private->min_free_disk_limit) != 0) {
-                        gf_log (this->name, GF_LOG_ERROR, 
-                                "invalid percent '%s' for min-free-disk limit", data->data);
-			ret = -1;
-			goto out;
+                list_for_each_entry_safe (limit, next, &head, limit_list) {
+                        found = 0;
+                        list_for_each_entry (new, &priv->limit_head,
+                                             limit_list) {
+                                if (strcmp (new->path, limit->path) == 0) {
+                                        found = 1;
+                                        break;
+                                }
+                        }
+
+                        if (!found) {
+                                limit->value = -1;
+                                __quota_reconfigure (this, top->itable, limit);
+                        }
+
+                        list_del_init (&limit->limit_list);
+                        GF_FREE (limit);
                 }
-		_private->refresh_interval = 20; /* 20seconds is default */
-		data = dict_get (this->options, "refresh-interval");
-		if (data) {
-			if (gf_string2time (data->data, 
-					    &_private->refresh_interval)!= 0) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"invalid time '%s' for refresh "
-					"interval", data->data);
-				ret = -1;
-				goto out;
-			}
-		}
         }
+        UNLOCK (&priv->lock);
 
-	_private->only_first_time = 1;
-        this->private = (void *)_private;
-	ret = 0;
- out:
-	return ret;
-}
+        GF_OPTION_RECONF ("timeout", priv->timeout, options, int64, out);
+        GF_OPTION_RECONF ("deem-statfs", priv->consider_statfs, options, bool,
+                          out);
 
+        ret = 0;
+out:
+        return ret;
+}
 
 
-void 
+void
 fini (xlator_t *this)
 {
-	struct quota_priv *_private = this->private;
-
-	if (_private) {
-		gf_quota_cache_sync (this);
-		this->private = NULL;
-	}
-	
-	return ;
+        return;
 }
 
+
 struct xlator_fops fops = {
-	.create      = quota_create,
-	.open        = quota_open,
-	.lookup	     = quota_lookup,
-	.truncate    = quota_truncate,
-	.ftruncate   = quota_ftruncate,
-	.writev      = quota_writev,
-	.unlink      = quota_unlink,
-	.rmdir       = quota_rmdir,
-	.mknod       = quota_mknod,
-	.mkdir       = quota_mkdir,
-	.symlink     = quota_symlink,
-	.statfs      = quota_statfs,
+        .statfs       = quota_statfs,
+        .lookup       = quota_lookup,
+        .writev       = quota_writev,
+        .create       = quota_create,
+        .mkdir        = quota_mkdir,
+        .truncate     = quota_truncate,
+        .ftruncate    = quota_ftruncate,
+        .unlink       = quota_unlink,
+        .symlink      = quota_symlink,
+        .link         = quota_link,
+        .rename       = quota_rename,
+        .getxattr     = quota_getxattr,
+        .fgetxattr    = quota_fgetxattr,
+        .stat         = quota_stat,
+        .fstat        = quota_fstat,
+        .readlink     = quota_readlink,
+        .readv        = quota_readv,
+        .fsync        = quota_fsync,
+        .setattr      = quota_setattr,
+        .fsetattr     = quota_fsetattr,
+        .mknod        = quota_mknod,
+        .setxattr     = quota_setxattr,
+        .fsetxattr    = quota_fsetxattr,
+        .removexattr  = quota_removexattr,
+        .fremovexattr = quota_fremovexattr,
+        .readdirp     = quota_readdirp,
+	.fallocate    = quota_fallocate,
 };
 
 struct xlator_cbks cbks = {
-	.release     = quota_release
+        .forget = quota_forget
 };
 
 struct volume_options options[] = {
-	{ .key  = {"min-free-disk-limit"}, 
-	  .type = GF_OPTION_TYPE_PERCENT
-	},
-	{ .key  = {"refresh-interval"}, 
-	  .type = GF_OPTION_TYPE_TIME
-	},
-	{ .key  = {"disk-usage-limit"}, 
-	  .type = GF_OPTION_TYPE_SIZET 
-	},
-	{ .key = {NULL} },
+        {.key = {"limit-set"}},
+        {.key = {"timeout"},
+         .type = GF_OPTION_TYPE_SIZET,
+         .min = 0,
+         .max = 60,
+         .default_value = "0",
+         .description = "quota caches the directory sizes on client. Timeout "
+                        "indicates the timeout for the cache to be revalidated."
+        },
+        {.key = {"deem-statfs"},
+         .type = GF_OPTION_TYPE_BOOL,
+         .default_value = "off",
+         .description = "If set to on, it takes quota limits into"
+                        "consideration while estimating fs size. (df command)"
+                        " (Default is off)."
+        },
+        {.key = {NULL}}
 };
diff --git a/xlators/features/quota/src/quota.h b/xlators/features/quota/src/quota.h
new file mode 100644
index 000000000..84ecbb308
--- /dev/null
+++ b/xlators/features/quota/src/quota.h
@@ -0,0 +1,151 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "call-stub.h"
+#include "defaults.h"
+#include "byte-order.h"
+#include "common-utils.h"
+#include "quota-mem-types.h"
+
+#define QUOTA_XATTR_PREFIX      "trusted."
+#define DIRTY                   "dirty"
+#define SIZE                    "size"
+#define CONTRIBUTION            "contri"
+#define VAL_LENGTH              8
+#define READDIR_BUF             4096
+
+#define QUOTA_SAFE_INCREMENT(lock, var)         \
+        do {                                    \
+                LOCK (lock);                    \
+                var ++;                         \
+                UNLOCK (lock);                  \
+        } while (0)
+
+#define QUOTA_SAFE_DECREMENT(lock, var)         \
+        do {                                    \
+                LOCK (lock);                    \
+                var --;                         \
+                UNLOCK (lock);                  \
+        } while (0)
+
+#define QUOTA_ALLOC_OR_GOTO(var, type, label)           \
+        do {                                            \
+                var = GF_CALLOC (sizeof (type), 1,      \
+                                 gf_quota_mt_##type);   \
+                if (!var) {                             \
+                        gf_log ("", GF_LOG_ERROR,       \
+                                "out of memory :(");    \
+                        ret = -1;                       \
+                        goto label;                     \
+                }                                       \
+        } while (0);
+
+#define QUOTA_STACK_UNWIND(fop, frame, params...)                       \
+        do {                                                            \
+                quota_local_t *_local = NULL;                           \
+                xlator_t      *_this  = NULL;                           \
+                if (frame) {                                            \
+                        _local = frame->local;                          \
+                        _this  = frame->this;                           \
+                        frame->local = NULL;                            \
+                }                                                       \
+                STACK_UNWIND_STRICT (fop, frame, params);               \
+                quota_local_cleanup (_this, _local);                    \
+        } while (0)
+
+#define QUOTA_FREE_CONTRIBUTION_NODE(_contribution)     \
+        do {                                            \
+                list_del (&_contribution->contri_list); \
+                GF_FREE (_contribution);                \
+        } while (0)
+
+#define GET_CONTRI_KEY(var, _vol_name, _gfid, _ret)             \
+        do {                                                    \
+                char _gfid_unparsed[40];                        \
+                uuid_unparse (_gfid, _gfid_unparsed);           \
+                _ret = gf_asprintf (var, QUOTA_XATTR_PREFIX     \
+                                    "%s.%s." CONTRIBUTION,      \
+                                    _vol_name, _gfid_unparsed); \
+        } while (0)
+
+
+#define GET_CONTRI_KEY_OR_GOTO(var, _vol_name, _gfid, label)    \
+        do {                                                    \
+                GET_CONTRI_KEY(var, _vol_name, _gfid, ret);     \
+                if (ret == -1)                                  \
+                        goto label;                             \
+        } while (0)
+
+#define GET_DIRTY_KEY_OR_GOTO(var, _vol_name, label)            \
+        do {                                                    \
+                ret = gf_asprintf (var, QUOTA_XATTR_PREFIX      \
+                                   "%s." DIRTY, _vol_name);     \
+                if (ret == -1)                                  \
+                        goto label;                             \
+        } while (0)
+
+struct quota_dentry {
+        char            *name;
+        uuid_t           par;
+        struct list_head next;
+};
+typedef struct quota_dentry quota_dentry_t;
+
+struct quota_inode_ctx {
+        int64_t          size;
+        int64_t          limit;
+        struct iatt      buf;
+        struct list_head parents;
+        struct timeval   tv;
+        gf_lock_t        lock;
+};
+typedef struct quota_inode_ctx quota_inode_ctx_t;
+
+struct quota_local {
+        gf_lock_t    lock;
+        uint32_t     validate_count;
+        uint32_t     link_count;
+        loc_t        loc;
+        loc_t        oldloc;
+        loc_t        newloc;
+        loc_t        validate_loc;
+        int64_t      delta;
+        int32_t      op_ret;
+        int32_t      op_errno;
+        int64_t      size;
+        int64_t      limit;
+        char         just_validated;
+        inode_t     *inode;
+        call_stub_t *stub;
+};
+typedef struct quota_local quota_local_t;
+
+struct quota_priv {
+        int64_t           timeout;
+        gf_boolean_t      consider_statfs;
+        struct list_head  limit_head;
+        gf_lock_t         lock;
+};
+typedef struct quota_priv quota_priv_t;
+
+struct limits {
+        struct list_head  limit_list;
+        char             *path;
+        int64_t           value;
+        uuid_t            gfid;
+};
+typedef struct limits     limits_t;
+
+uint64_t cn = 1;
diff --git a/xlators/features/read-only/src/Makefile.am b/xlators/features/read-only/src/Makefile.am
index 15f49966f..4c1462137 100644
--- a/xlators/features/read-only/src/Makefile.am
+++ b/xlators/features/read-only/src/Makefile.am
@@ -1,13 +1,22 @@
-xlator_LTLIBRARIES = read-only.la
+xlator_LTLIBRARIES = read-only.la worm.la
+
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
 
-read_only_la_LDFLAGS = -module -avoidversion
+noinst_HEADERS = read-only-common.h
+
+read_only_la_LDFLAGS = -module -avoid-version
 
-read_only_la_SOURCES = read-only.c
+read_only_la_SOURCES = read-only.c read-only-common.c
 read_only_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la 
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+worm_la_LDFLAGS = -module -avoid-version
+
+worm_la_SOURCES = read-only-common.c worm.c
+worm_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/features/read-only/src/read-only-common.c b/xlators/features/read-only/src/read-only-common.c
new file mode 100644
index 000000000..56a7a7176
--- /dev/null
+++ b/xlators/features/read-only/src/read-only-common.c
@@ -0,0 +1,239 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+
+int32_t
+ro_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (xattrop, frame, -1, EROFS, NULL, xdata);
+        return 0;
+}
+
+int32_t
+ro_fxattrop (call_frame_t *frame, xlator_t *this,
+             fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (fxattrop, frame, -1, EROFS, NULL, xdata);
+        return 0;
+}
+
+int32_t
+ro_entrylk (call_frame_t *frame, xlator_t *this, const char *volume,
+            loc_t *loc, const char *basename, entrylk_cmd cmd,
+            entrylk_type type, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (entrylk, frame, -1, EROFS, xdata);
+        return 0;
+}
+
+int32_t
+ro_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume,
+             fd_t *fd, const char *basename, entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (fentrylk, frame, -1, EROFS, xdata);
+        return 0;
+}
+
+int32_t
+ro_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+            loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (inodelk, frame, -1, EROFS, xdata);
+        return 0;
+}
+
+int32_t
+ro_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+             fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (finodelk, frame, -1, EROFS, xdata);
+        return 0;
+}
+
+int32_t
+ro_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int cmd,
+       struct gf_flock *flock, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (lk, frame, -1, EROFS, NULL, xdata);
+        return 0;
+}
+
+int32_t
+ro_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (setattr, frame, -1, EROFS, NULL, NULL, xdata);
+	return 0;
+}
+
+int32_t
+ro_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+             struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (fsetattr, frame, -1, EROFS, NULL, NULL, xdata);
+        return 0;
+}
+
+
+int32_t
+ro_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (truncate, frame, -1, EROFS, NULL, NULL, xdata);
+	return 0;
+}
+
+int32_t
+ro_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (ftruncate, frame, -1, EROFS, NULL, NULL, xdata);
+	return 0;
+}
+
+int
+ro_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+          dev_t rdev, mode_t umask, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (mknod, frame, -1, EROFS, NULL, NULL, NULL, NULL, xdata);
+        return 0;
+}
+
+
+int
+ro_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+          mode_t umask, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (mkdir, frame, -1, EROFS, NULL, NULL, NULL, NULL, xdata);
+        return 0;
+}
+
+int32_t
+ro_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+           dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (unlink, frame, -1, EROFS, NULL, NULL, xdata);
+        return 0;
+}
+
+
+int
+ro_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+          dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (rmdir, frame, -1, EROFS, NULL, NULL, xdata);
+        return 0;
+}
+
+
+int
+ro_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
+            loc_t *loc, mode_t umask, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (symlink, frame, -1, EROFS, NULL, NULL, NULL,
+                             NULL, xdata);
+        return 0;
+}
+
+
+
+int32_t
+ro_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (rename, frame, -1, EROFS, NULL, NULL, NULL, NULL,
+                             NULL, xdata);
+        return 0;
+}
+
+
+int32_t
+ro_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (link, frame, -1, EROFS, NULL, NULL, NULL, NULL, xdata);
+        return 0;
+}
+
+int32_t
+ro_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+           mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (create, frame, -1, EROFS, NULL, NULL, NULL,
+                             NULL, NULL, xdata);
+        return 0;
+}
+
+
+static int32_t
+ro_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+             int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+	return 0;
+}
+
+int32_t
+ro_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+         fd_t *fd, dict_t *xdata)
+{
+        if (((flags & O_ACCMODE) == O_WRONLY) ||
+              ((flags & O_ACCMODE) == O_RDWR)) {
+                STACK_UNWIND_STRICT (open, frame, -1, EROFS, NULL, xdata);
+                return 0;
+	}
+
+	STACK_WIND (frame, ro_open_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+	return 0;
+}
+
+int32_t
+ro_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+              int32_t flags, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (fsetxattr, frame, -1, EROFS, xdata);
+	return 0;
+}
+
+int32_t
+ro_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (fsyncdir, frame, -1, EROFS, xdata);
+	return 0;
+}
+
+int32_t
+ro_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+           int32_t count, off_t off, uint32_t flags, struct iobref *iobref, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (writev, frame, -1, EROFS, NULL, NULL, xdata);
+        return 0;
+}
+
+
+int32_t
+ro_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+             int32_t flags, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (setxattr, frame, -1, EROFS, xdata);
+        return 0;
+}
+
+int32_t
+ro_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                const char *name, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (removexattr, frame, -1, EROFS, xdata);
+        return 0;
+}
diff --git a/xlators/features/read-only/src/read-only-common.h b/xlators/features/read-only/src/read-only-common.h
new file mode 100644
index 000000000..5d4c7e260
--- /dev/null
+++ b/xlators/features/read-only/src/read-only-common.h
@@ -0,0 +1,115 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+
+int32_t
+ro_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata);
+
+int32_t
+ro_fxattrop (call_frame_t *frame, xlator_t *this,
+             fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata);
+
+int32_t
+ro_entrylk (call_frame_t *frame, xlator_t *this, const char *volume,
+            loc_t *loc, const char *basename, entrylk_cmd cmd,
+            entrylk_type type, dict_t *xdata);
+
+int32_t
+ro_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume,
+             fd_t *fd, const char *basename, entrylk_cmd cmd, entrylk_type
+             type, dict_t *xdata);
+
+int32_t
+ro_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+            loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata);
+
+int32_t
+ro_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+             fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata);
+
+int32_t
+ro_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int cmd,
+       struct gf_flock *flock, dict_t *xdata);
+
+int32_t
+ro_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            struct iatt *stbuf, int32_t valid, dict_t *xdata);
+
+int32_t
+ro_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+             struct iatt *stbuf, int32_t valid, dict_t *xdata);
+
+
+int32_t
+ro_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset, dict_t *xdata);
+
+int32_t
+ro_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset, dict_t *xdata);
+
+int
+ro_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+          dev_t rdev, mode_t umask, dict_t *xdata);
+
+int
+ro_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+          mode_t umask, dict_t *xdata);
+
+int32_t
+ro_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+           dict_t *xdata);
+
+int
+ro_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+          dict_t *xdata);
+
+
+int
+ro_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
+            loc_t *loc, mode_t umask, dict_t *xdata);
+
+int32_t
+ro_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc, dict_t *xdata);
+
+int32_t
+ro_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc, dict_t *xdata);
+
+int32_t
+ro_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+           mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata);
+
+int32_t
+ro_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+         fd_t *fd, dict_t *xdata);
+
+int32_t
+ro_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+              int32_t flags, dict_t *xdata);
+
+int32_t
+ro_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata);
+
+int32_t
+ro_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+           int32_t count, off_t off, uint32_t flags, struct iobref *iobref, dict_t *xdata);
+
+int32_t
+ro_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+             int32_t flags, dict_t *xdata);
+
+int32_t
+ro_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                const char *name, dict_t *xdata);
diff --git a/xlators/features/read-only/src/read-only.c b/xlators/features/read-only/src/read-only.c
index 49c2ce58d..e49e54a1b 100644
--- a/xlators/features/read-only/src/read-only.c
+++ b/xlators/features/read-only/src/read-only.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -24,225 +14,7 @@
 
 #include "xlator.h"
 #include "defaults.h"
-
-int32_t
-ro_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
-            gf_xattrop_flags_t flags, dict_t *dict)
-{
-        STACK_UNWIND_STRICT (xattrop, frame, -1, EROFS, NULL);
-        return 0;
-}
-
-int32_t
-ro_fxattrop (call_frame_t *frame, xlator_t *this,
-             fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict)
-{
-        STACK_UNWIND_STRICT (fxattrop, frame, -1, EROFS, NULL);
-        return 0;
-}
-
-int32_t
-ro_entrylk (call_frame_t *frame, xlator_t *this, const char *volume,
-            loc_t *loc, const char *basename, entrylk_cmd cmd,
-            entrylk_type type)
-{
-        STACK_UNWIND_STRICT (entrylk, frame, -1, EROFS);
-        return 0;
-}
-
-int32_t
-ro_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume,
-             fd_t *fd, const char *basename, entrylk_cmd cmd, entrylk_type type)
-{
-        STACK_UNWIND_STRICT (fentrylk, frame, -1, EROFS);
-        return 0;
-}
-
-int32_t
-ro_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
-            loc_t *loc, int32_t cmd, struct flock *lock)
-{
-        STACK_UNWIND_STRICT (inodelk, frame, -1, EROFS);
-        return 0;
-}
-
-int32_t
-ro_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
-             fd_t *fd, int32_t cmd, struct flock *lock)
-{
-        STACK_UNWIND_STRICT (finodelk, frame, -1, EROFS);
-        return 0;
-}
-
-int32_t
-ro_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int cmd,
-       struct flock *flock)
-{
-        STACK_UNWIND_STRICT (lk, frame, -1, EROFS, NULL);
-        return 0;
-}
-
-int32_t
-ro_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-            struct iatt *stbuf, int32_t valid)
-{
-        STACK_UNWIND_STRICT (setattr, frame, -1, EROFS, NULL, NULL);
-	return 0;
-}
-
-int32_t
-ro_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
-             struct iatt *stbuf, int32_t valid)
-{
-        STACK_UNWIND_STRICT (fsetattr, frame, -1, EROFS, NULL, NULL);
-        return 0;
-}
-
-
-int32_t
-ro_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
-{
-        STACK_UNWIND_STRICT (truncate, frame, -1, EROFS, NULL, NULL);
-	return 0;
-}
-
-int32_t
-ro_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
-{
-        STACK_UNWIND_STRICT (ftruncate, frame, -1, EROFS, NULL, NULL);
-	return 0;
-}
-
-int
-ro_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-          dev_t rdev, dict_t *params)
-{
-        STACK_UNWIND_STRICT (mknod, frame, -1, EROFS, NULL, NULL, NULL, NULL);
-        return 0;
-}
-
-
-int
-ro_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-          dict_t *params)
-{
-        STACK_UNWIND_STRICT (mkdir, frame, -1, EROFS, NULL, NULL, NULL, NULL);
-        return 0;
-}
-
-int32_t
-ro_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
-{
-        STACK_UNWIND_STRICT (unlink, frame, -1, EROFS, NULL, NULL);
-        return 0;
-}
-
-int32_t
-ro_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
-{
-        STACK_UNWIND_STRICT (rmdir, frame, -1, EROFS, NULL, NULL);
-        return 0;
-}
-
-
-int
-ro_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
-            loc_t *loc, dict_t *params)
-{
-        STACK_UNWIND_STRICT (symlink, frame, -1, EROFS, NULL, NULL, NULL, NULL);
-        return 0;
-}
-
-
-
-int32_t
-ro_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
-{
-        STACK_UNWIND_STRICT (rename, frame, -1, EROFS, NULL, NULL, NULL, NULL,
-                             NULL);
-        return 0;
-}
-
-
-int32_t
-ro_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
-{
-        STACK_UNWIND_STRICT (link, frame, -1, EROFS, NULL, NULL, NULL, NULL);
-        return 0;
-}
-
-int32_t
-ro_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-           mode_t mode, fd_t *fd, dict_t *params)
-{
-        STACK_UNWIND_STRICT (create, frame, -1, EROFS, NULL, NULL, NULL,
-                             NULL, NULL);
-        return 0;
-}
-
-
-static int32_t
-ro_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
-             int32_t op_errno, fd_t *fd)
-{
-	STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd);
-	return 0;
-}
-
-int32_t
-ro_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-         fd_t *fd, int32_t wbflags)
-{
-        if (((flags & O_ACCMODE) == O_WRONLY) ||
-              ((flags & O_ACCMODE) == O_RDWR)) {
-                STACK_UNWIND_STRICT (open, frame, -1, EROFS, NULL);
-                return 0;
-	}
-
-	STACK_WIND (frame, ro_open_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->open, loc, flags, fd, wbflags);
-	return 0;
-}
-
-int32_t
-ro_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
-              int32_t flags)
-{
-        STACK_UNWIND_STRICT (fsetxattr, frame, -1, EROFS);
-	return 0;
-}
-
-int32_t
-ro_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
-{
-        STACK_UNWIND_STRICT (fsyncdir, frame, -1, EROFS);
-	return 0;
-}
-
-int32_t
-ro_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
-           int32_t count, off_t off, struct iobref *iobref)
-{
-        STACK_UNWIND_STRICT (writev, frame, -1, EROFS, NULL, NULL);
-        return 0;
-}
-
-
-int32_t
-ro_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
-             int32_t flags)
-{
-        STACK_UNWIND_STRICT (setxattr, frame, -1, EROFS);
-        return 0;
-}
-
-int32_t
-ro_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                const char *name)
-{
-        STACK_UNWIND_STRICT (removexattr, frame, -1, EROFS);
-        return 0;
-}
+#include "read-only-common.h"
 
 int32_t
 init (xlator_t *this)
diff --git a/xlators/features/read-only/src/worm.c b/xlators/features/read-only/src/worm.c
new file mode 100644
index 000000000..16c3eb3da
--- /dev/null
+++ b/xlators/features/read-only/src/worm.c
@@ -0,0 +1,89 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "read-only-common.h"
+
+static int32_t
+worm_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int32_t op_ret,
+               int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+        return 0;
+}
+
+int32_t
+worm_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+           fd_t *fd, dict_t *xdata)
+{
+        if ((((flags & O_ACCMODE) == O_WRONLY) ||
+              ((flags & O_ACCMODE) == O_RDWR)) &&
+              !(flags & O_APPEND)) {
+                STACK_UNWIND_STRICT (open, frame, -1, EROFS, NULL, NULL);
+                return 0;
+        }
+
+        STACK_WIND (frame, worm_open_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+        return 0;
+}
+
+int32_t
+init (xlator_t *this)
+{
+        if (!this->children || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "translator not configured with exactly one child");
+                return -1;
+        }
+
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dangling volume. check volfile ");
+        }
+
+        return 0;
+}
+
+
+void
+fini (xlator_t *this)
+{
+        return;
+}
+
+struct xlator_fops fops = {
+        .open        = worm_open,
+
+        .unlink      = ro_unlink,
+        .rmdir       = ro_rmdir,
+        .rename      = ro_rename,
+        .truncate    = ro_truncate,
+        .removexattr = ro_removexattr,
+        .fsyncdir    = ro_fsyncdir,
+        .xattrop     = ro_xattrop,
+        .inodelk     = ro_inodelk,
+        .finodelk    = ro_finodelk,
+        .entrylk     = ro_entrylk,
+        .fentrylk    = ro_fentrylk,
+        .lk          = ro_lk,
+};
+
+struct xlator_cbks cbks;
+
+struct volume_options options[] = {
+        { .key = {NULL} },
+};
+
diff --git a/xlators/features/trash/src/Makefile.am b/xlators/features/trash/src/Makefile.am
index 4671d06d3..5251eb082 100644
--- a/xlators/features/trash/src/Makefile.am
+++ b/xlators/features/trash/src/Makefile.am
@@ -1,15 +1,16 @@
 xlator_LTLIBRARIES = trash.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/testing/features
 
-trash_la_LDFLAGS = -module -avoidversion 
+trash_la_LDFLAGS = -module -avoid-version 
 
 trash_la_SOURCES = trash.c
 trash_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la 
 
 noinst_HEADERS = trash.h trash-mem-types.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/features/trash/src/trash-mem-types.h b/xlators/features/trash/src/trash-mem-types.h
index 48613d1e8..0e6ef572f 100644
--- a/xlators/features/trash/src/trash-mem-types.h
+++ b/xlators/features/trash/src/trash-mem-types.h
@@ -1,30 +1,19 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __TRASH_MEM_TYPES_H__
 #define __TRASH_MEM_TYPES_H__
 
 #include "mem-types.h"
 
 enum gf_trash_mem_types_ {
-        gf_trash_mt_trash_local_t = gf_common_mt_end + 1,
-        gf_trash_mt_trash_private_t,
+        gf_trash_mt_trash_private_t = gf_common_mt_end + 1,
         gf_trash_mt_char,
         gf_trash_mt_trash_elim_pattern_t,
         gf_trash_mt_end
diff --git a/xlators/features/trash/src/trash.c b/xlators/features/trash/src/trash.c
index 8a11f6e55..addeb66a0 100644
--- a/xlators/features/trash/src/trash.c
+++ b/xlators/features/trash/src/trash.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -63,7 +53,7 @@ trash_local_wipe (trash_local_t *local)
         if (local->newfd)
                 fd_unref (local->newfd);
 
-        GF_FREE (local);
+        mem_put (local);
 out:
         return;
 }
@@ -170,8 +160,7 @@ trash_unlink_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 out:
         GF_FREE (cookie);
-        if (tmp_str)
-                GF_FREE (tmp_str);
+        GF_FREE (tmp_str);
 
         return 0;
 }
@@ -442,8 +431,7 @@ trash_rename_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 out:
         GF_FREE (cookie); /* strdup (dir_name) was sent here :) */
-        if (tmp_str)
-                GF_FREE (tmp_str);
+        GF_FREE (tmp_str);
 
         return 0;
 }
@@ -504,9 +492,7 @@ trash_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
         trash_elim_pattern_t  *trav  = NULL;
         trash_private_t *priv = NULL;
         trash_local_t   *local = NULL;
-        struct tm       *tm = NULL;
-        char             timestr[256] = {0,};
-        time_t           utime = 0;
+        char             timestr[64] = {0,};
         int32_t          match = 0;
 
         priv = this->private;
@@ -533,8 +519,7 @@ trash_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
                 return 0;
         }
 
-        local = GF_CALLOC (1, sizeof (trash_local_t), 
-                           gf_trash_mt_trash_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 TRASH_STACK_UNWIND (rename, frame, -1, ENOMEM,
@@ -554,9 +539,8 @@ trash_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
         {
                 /* append timestamp to file name */
                 /* TODO: can we make it optional? */
-                utime = time (NULL);
-                tm    = localtime (&utime);
-                strftime (timestr, 256, ".%Y-%m-%d-%H%M%S", tm);
+                gf_time_ftm (timestr, sizeof timestr, time (NULL),
+                             gf_timefmt_F_HMS);
                 strcat (local->newpath, timestr);
         }
 
@@ -575,9 +559,7 @@ trash_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
         trash_elim_pattern_t  *trav  = NULL;
         trash_private_t *priv = NULL;
         trash_local_t   *local = NULL;
-        struct tm       *tm = NULL;
-        char             timestr[256] = {0,};
-        time_t           utime = 0;
+        char             timestr[64] = {0,};
         int32_t          match = 0;
 
         priv = this->private;
@@ -610,8 +592,7 @@ trash_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
                 return 0;
         }
 
-        local = GF_CALLOC (1, sizeof (trash_local_t),
-                           gf_trash_mt_trash_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 gf_log (this->name, GF_LOG_DEBUG, "out of memory");
                 TRASH_STACK_UNWIND (unlink, frame, -1, ENOMEM, NULL, NULL);
@@ -627,9 +608,8 @@ trash_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
         {
                 /* append timestamp to file name */
                 /* TODO: can we make it optional? */
-                utime = time (NULL);
-                tm    = localtime (&utime);
-                strftime (timestr, 256, ".%Y-%m-%d-%H%M%S", tm);
+                gf_time_fmt (timestr, sizeof timestr, time (NULL),
+                             gf_timefmt_F_HMS);
                 strcat (local->newpath, timestr);
         }
 
@@ -690,7 +670,7 @@ trash_truncate_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         local->fsize = stbuf->ia_size;
         STACK_WIND (frame, trash_truncate_writev_cbk, FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->writev,
-                    local->newfd, vector, count, local->cur_offset, iobuf);
+                    local->newfd, vector, count, local->cur_offset, 0, iobuf);
 
 out:
         return 0;
@@ -723,7 +703,7 @@ trash_truncate_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 STACK_WIND (frame, trash_truncate_readv_cbk,
                             FIRST_CHILD(this), FIRST_CHILD(this)->fops->readv,
                             local->fd, (size_t)GF_BLOCK_READV_SIZE,
-                            local->cur_offset);
+                            local->cur_offset, 0);
                 goto out;
         }
 
@@ -763,7 +743,7 @@ trash_truncate_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         STACK_WIND (frame, trash_truncate_readv_cbk,
                     FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv,
-                    local->fd, (size_t)GF_BLOCK_READV_SIZE, local->cur_offset);
+                    local->fd, (size_t)GF_BLOCK_READV_SIZE, local->cur_offset, 0);
 
 out:
         return 0;
@@ -932,8 +912,7 @@ trash_truncate_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 out:
         GF_FREE (cookie); /* strdup (dir_name) was sent here :) */
-        if (tmp_str)
-                GF_FREE (tmp_str);
+        GF_FREE (tmp_str);
 
         return 0;
 }
@@ -945,10 +924,8 @@ trash_truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 {
         trash_private_t     *priv  = NULL;
         trash_local_t       *local = NULL;
-        struct tm           *tm = NULL;
-        char                 timestr[256] = {0,};
+        char                 timestr[64] = {0,};
         char                 loc_newname[PATH_MAX] = {0,};
-        time_t               utime = 0;
         int32_t              flags = 0;
 
         priv = this->private;
@@ -980,9 +957,8 @@ trash_truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         strcat (local->newpath, local->loc.path);
 
         {
-                utime = time (NULL);
-                tm    = localtime (&utime);
-                strftime (timestr, 256, ".%Y-%m-%d-%H%M%S", tm);
+                gf_time_fmt (timestr, sizeof timestr, time (NULL),
+                             gf_timefmt_F_HMS);
                 strcat (local->newpath, timestr);
         }
         strcpy (loc_newname,local->loc.name);
@@ -991,7 +967,6 @@ trash_truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         local->newloc.name = gf_strdup (loc_newname);
         local->newloc.path = gf_strdup (local->newpath);
         local->newloc.inode = inode_new (local->loc.inode->table);
-        local->newloc.ino   = local->newloc.inode->ino;
         local->newfd = fd_create (local->newloc.inode, frame->root->pid);
 
         flags = O_CREAT|O_EXCL|O_WRONLY;
@@ -1045,8 +1020,7 @@ trash_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
 
         LOCK_INIT (&frame->lock);
 
-        local = GF_CALLOC (1, sizeof (trash_local_t), 
-                           gf_trash_mt_trash_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 gf_log (this->name, GF_LOG_DEBUG, "out of memory");
                 TRASH_STACK_UNWIND (truncate, frame, -1, ENOMEM, NULL, NULL);
@@ -1111,7 +1085,7 @@ trash_ftruncate_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 STACK_WIND (frame, trash_ftruncate_readv_cbk,
                             FIRST_CHILD(this), FIRST_CHILD(this)->fops->readv,
                             local->fd, (size_t)GF_BLOCK_READV_SIZE,
-                            local->cur_offset);
+                            local->cur_offset, 0);
                 return 0;
         }
 
@@ -1143,7 +1117,7 @@ trash_ftruncate_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         STACK_WIND (frame, trash_ftruncate_writev_cbk,
                     FIRST_CHILD(this), FIRST_CHILD(this)->fops->writev,
-                    local->newfd, vector, count, local->cur_offset, NULL);
+                    local->newfd, vector, count, local->cur_offset, 0, NULL);
 
         return 0;
 }
@@ -1195,7 +1169,7 @@ trash_ftruncate_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         STACK_WIND (frame, trash_ftruncate_readv_cbk, FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readv, local->fd,
-                    (size_t)GF_BLOCK_READV_SIZE, local->cur_offset);
+                    (size_t)GF_BLOCK_READV_SIZE, local->cur_offset, 0);
 
         return 0;
 }
@@ -1300,8 +1274,7 @@ trash_ftruncate_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 out:
         GF_FREE (cookie); /* strdup (dir_name) was sent here :) */
-        if (tmp_str)
-                GF_FREE (tmp_str);
+        GF_FREE (tmp_str);
 
         return 0;
 }
@@ -1350,11 +1323,9 @@ trash_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
         trash_private_t       *priv = NULL;
         trash_local_t         *local = NULL;
         dentry_t              *dir_entry = NULL;
-        struct tm             *tm = NULL;
         char                  *pathbuf = NULL;
         inode_t               *newinode = NULL;
-        time_t                 utime = 0;
-        char                   timestr[256];
+        char                   timestr[64];
         int32_t                retval = 0;
         int32_t                match = 0;
 
@@ -1386,18 +1357,14 @@ trash_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
                 return 0;
         }
 
-        local = GF_CALLOC (1, sizeof (trash_local_t), 
-                           gf_trash_mt_trash_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 gf_log (this->name, GF_LOG_DEBUG, "out of memory");
                 TRASH_STACK_UNWIND (ftruncate, frame, -1, ENOMEM, NULL, NULL);
                 return 0;
         }
 
-        utime = time (NULL);
-        tm    = localtime (&utime);
-        strftime (timestr, 256, ".%Y-%m-%d-%H%M%S", tm);
-
+        gf_time_fmt (timestr, sizeof timestr, time (NULL), gf_timefmt_F_HMS);
         strcpy (local->newpath, priv->trash_dir);
         strcat (local->newpath, pathbuf);
         strcat (local->newpath, timestr);
@@ -1411,7 +1378,6 @@ trash_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
         local->newloc.path = local->newpath;
 
         local->loc.inode = inode_ref (fd->inode);
-        local->loc.ino   = fd->inode->ino;
         local->loc.path  = pathbuf;
 
         local->fop_offset = offset;
@@ -1458,7 +1424,7 @@ init (xlator_t *this)
 
         data = dict_get (this->options, "trash-dir");
         if (!data) {
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "no option specified for 'trash-dir', "
                         "using \"/.trashcan/\"");
                 _priv->trash_dir = gf_strdup ("/.trashcan");
@@ -1524,6 +1490,14 @@ init (xlator_t *this)
                         _priv->max_trash_file_size);
         }
 
+        this->local_pool = mem_pool_new (trash_local_t, 64);
+        if (!this->local_pool) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                return -1;
+        }
+
+
         this->private = (void *)_priv;
         return 0;
 }
@@ -1534,8 +1508,7 @@ fini (xlator_t *this)
         trash_private_t *priv = NULL;
 
         priv = this->private;
-        if (priv)
-                GF_FREE (priv);
+        GF_FREE (priv);
 
         return;
 }
diff --git a/xlators/features/trash/src/trash.h b/xlators/features/trash/src/trash.h
index ea393fe67..9a7c03361 100644
--- a/xlators/features/trash/src/trash.h
+++ b/xlators/features/trash/src/trash.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __TRASH_H__
 #define __TRASH_H__
 
diff --git a/xlators/lib/src/libxlator.c b/xlators/lib/src/libxlator.c
new file mode 100644
index 000000000..9e5357255
--- /dev/null
+++ b/xlators/lib/src/libxlator.c
@@ -0,0 +1,470 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#include "mem-types.h"
+#include "libxlator.h"
+
+
+int marker_xtime_default_gauge[] = {
+        [MCNT_FOUND]    =  1,
+        [MCNT_NOTFOUND] = -1,
+        [MCNT_ENODATA]  = -1,
+        [MCNT_ENOTCONN] = -1,
+        [MCNT_ENOENT]   = -1,
+        [MCNT_EOTHER]   = -1,
+};
+
+int marker_uuid_default_gauge[] = {
+        [MCNT_FOUND]    =  1,
+        [MCNT_NOTFOUND] =  0,
+        [MCNT_ENODATA]  =  0,
+        [MCNT_ENOTCONN] =  0,
+        [MCNT_ENOENT]   =  0,
+        [MCNT_EOTHER]   =  0,
+};
+
+static int marker_idx_errno_map[] = {
+        [MCNT_FOUND]    = EINVAL,
+        [MCNT_NOTFOUND] = EINVAL,
+        [MCNT_ENOENT]   = ENOENT,
+        [MCNT_ENOTCONN] = ENOTCONN,
+        [MCNT_ENODATA]  = ENODATA,
+        [MCNT_EOTHER]   = EINVAL,
+        [MCNT_MAX]      = 0,
+};
+
+/*Copy the contents of oldtimebuf to newtimbuf*/
+static void
+update_timebuf (uint32_t *oldtimbuf, uint32_t *newtimebuf)
+{
+        newtimebuf[0] =  (oldtimbuf[0]);
+        newtimebuf[1] =  (oldtimbuf[1]);
+}
+
+/* Convert Timebuf in network order to host order */
+static void
+get_hosttime (uint32_t *oldtimbuf, uint32_t *newtimebuf)
+{
+        newtimebuf[0] = ntohl (oldtimbuf[0]);
+        newtimebuf[1] = ntohl (oldtimbuf[1]);
+}
+
+
+
+/* Match the Incoming trusted.glusterfs.<uuid>.xtime against volume uuid */
+int
+match_uuid_local (const char *name, char *uuid)
+{
+        name = strtail ((char *)name, MARKER_XATTR_PREFIX);
+        if (!name || name++[0] != '.')
+                return -1;
+
+        name = strtail ((char *)name, uuid);
+        if (!name || strcmp (name, ".xtime") != 0)
+                return -1;
+
+        return 0;
+}
+
+static void
+marker_local_incr_errcount (xl_marker_local_t *local, int op_errno)
+{
+        marker_result_idx_t i = -1;
+
+        if (!local)
+                return;
+
+        switch (op_errno) {
+                case ENODATA:
+                        i = MCNT_ENODATA;
+                        break;
+                case ENOENT:
+                        i = MCNT_ENOENT;
+                        break;
+                case ENOTCONN:
+                        i = MCNT_ENOTCONN;
+                        break;
+                default:
+                        i = MCNT_EOTHER;
+                        break;
+        }
+
+        local->count[i]++;
+}
+
+static int
+evaluate_marker_results (int *gauge, int *count)
+{
+        int i = 0;
+        int op_errno = 0;
+        gf_boolean_t sane = _gf_true;
+
+        /* check if the policy of the gauge is violated;
+         * if yes, try to get the best errno, ie. look
+         * for the first position where there is a more
+         * specific kind of vioilation than the generic EINVAL
+         */
+        for (i = 0; i < MCNT_MAX; i++) {
+                if (sane) {
+                        if ((gauge[i] > 0 && count[i] < gauge[i]) ||
+                            (gauge[i] < 0 && count[i] >= -gauge[i])) {
+                                sane = _gf_false;
+                                /* generic action: adopt corresponding errno */
+                                op_errno = marker_idx_errno_map[i];
+                        }
+                } else {
+                        /* already insane; trying to get a more informative
+                         * errno by checking subsequent counters
+                         */
+                        if (count[i] > 0)
+                                op_errno = marker_idx_errno_map[i];
+                }
+                if (op_errno && op_errno != EINVAL)
+                        break;
+        }
+
+        return op_errno;
+}
+
+/* Aggregate all the <volid>.xtime attrs of the cluster and send the max*/
+int32_t
+cluster_markerxtime_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+
+{
+
+        int32_t           callcnt         = 0;
+        int               ret             = -1;
+        uint32_t          *net_timebuf    = NULL;
+        uint32_t          host_timebuf[2] = {0,};
+        char              *marker_xattr   = NULL;
+        xl_marker_local_t *local          = NULL;
+        char              *vol_uuid       = NULL;
+        char              need_unwind     = 0;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("", GF_LOG_DEBUG, "possible NULL deref");
+                need_unwind = 1;
+                goto out;
+        }
+
+        local = frame->local;
+        if (!local || !local->vol_uuid) {
+                gf_log (this->name, GF_LOG_DEBUG, "possible NULL deref");
+                need_unwind = 1;
+                goto out;
+        }
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
+
+                vol_uuid = local->vol_uuid;
+
+                if (op_ret) {
+                        marker_local_incr_errcount (local, op_errno);
+                        goto unlock;
+                }
+
+                if (!gf_asprintf (&marker_xattr, "%s.%s.%s",
+                                MARKER_XATTR_PREFIX, vol_uuid, XTIME)) {
+                        op_errno = ENOMEM;
+                        goto unlock;
+                }
+
+
+                if (dict_get_ptr (dict, marker_xattr, (void **)&net_timebuf)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Unable to get <uuid>.xtime attr");
+                        local->count[MCNT_NOTFOUND]++;
+                        goto unlock;
+                }
+
+                if (local->count[MCNT_FOUND]) {
+                        get_hosttime (net_timebuf, host_timebuf);
+                        if ( (host_timebuf[0]>local->host_timebuf[0]) ||
+                                (host_timebuf[0] == local->host_timebuf[0] &&
+                                 host_timebuf[1] >= local->host_timebuf[1])) {
+                                update_timebuf (net_timebuf, local->net_timebuf);
+                                update_timebuf (host_timebuf, local->host_timebuf);
+                        }
+
+                } else {
+                        get_hosttime (net_timebuf, local->host_timebuf);
+                        update_timebuf (net_timebuf, local->net_timebuf);
+                        local->count[MCNT_FOUND]++;
+                }
+
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                op_ret = 0;
+                op_errno = 0;
+                need_unwind = 1;
+
+                if (local->count[MCNT_FOUND]) {
+                        if (!dict)
+                                dict = dict_new();
+
+                        ret = dict_set_static_bin (dict, marker_xattr,
+                                           (void *)local->net_timebuf, 8);
+                        if (ret) {
+                                op_ret = -1;
+                                op_errno = ENOMEM;
+                                goto out;
+                        }
+                }
+
+                op_errno = evaluate_marker_results (local->gauge, local->count);
+                if (op_errno)
+                        op_ret = -1;
+        }
+
+out:
+        if (need_unwind && local && local->xl_specf_unwind) {
+                frame->local = local->xl_local;
+                local->xl_specf_unwind (frame, op_ret,
+                                        op_errno, dict, xdata);
+        } else if (need_unwind) {
+                STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno,
+                                     dict, xdata);
+        }
+
+        GF_FREE (marker_xattr);
+        return 0;
+
+}
+
+int32_t
+cluster_markeruuid_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+{
+        int32_t             callcnt     = 0;
+        struct volume_mark  *volmark    = NULL;
+        xl_marker_local_t   *local      = NULL;
+        int32_t             ret         = -1;
+        char                need_unwind = 0;
+        char                *vol_uuid   = NULL;
+
+        if (!this || !frame || !cookie) {
+                gf_log ("", GF_LOG_DEBUG, "possible NULL deref");
+                need_unwind = 1;
+                goto out;
+        }
+
+        local = frame->local;
+
+        if (!local) {
+                gf_log (this->name, GF_LOG_DEBUG, "possible NULL deref");
+                need_unwind = 1;
+                goto out;
+        }
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
+                vol_uuid = local->vol_uuid;
+
+                if (op_ret) {
+                        marker_local_incr_errcount (local, op_errno);
+                        goto unlock;
+                }
+
+                ret = dict_get_bin (dict, GF_XATTR_MARKER_KEY,
+                                    (void *)&volmark);
+                if (ret)
+                        goto unlock;
+
+                if (local->count[MCNT_FOUND]) {
+                        if ((local->volmark->major != volmark->major) ||
+                            (local->volmark->minor != volmark->minor)) {
+                                op_ret = -1;
+                                op_errno = EINVAL;
+                                goto unlock;
+                        }
+
+                        if (local->retval)
+                                goto unlock;
+                        else if (volmark->retval) {
+                                GF_FREE (local->volmark);
+                                local->volmark =
+                                        memdup (volmark, sizeof (*volmark));
+                                local->retval = volmark->retval;
+                        } else if ((volmark->sec > local->volmark->sec) ||
+                                   ((volmark->sec == local->volmark->sec) &&
+                                    (volmark->usec >= local->volmark->usec))) {
+                                GF_FREE (local->volmark);
+                                local->volmark =
+                                      memdup (volmark, sizeof (*volmark));
+                        }
+
+                } else {
+                        local->volmark = memdup (volmark, sizeof (*volmark));
+                        VALIDATE_OR_GOTO (local->volmark, unlock);
+                        uuid_unparse (volmark->uuid, vol_uuid);
+                        if (volmark->retval)
+                                local->retval = volmark->retval;
+                        local->count[MCNT_FOUND]++;
+                }
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                op_ret = 0;
+                op_errno = 0;
+                need_unwind = 1;
+
+                if (local->count[MCNT_FOUND]) {
+                        if (!dict)
+                                dict = dict_new();
+
+                        if (dict_set_bin (dict, GF_XATTR_MARKER_KEY,
+                                          local->volmark,
+                                          sizeof (struct volume_mark))) {
+                                op_ret = -1;
+                                op_errno = ENOMEM;
+                        }
+                }
+                op_errno = evaluate_marker_results (local->gauge, local->count);
+                if (op_errno)
+                        op_ret = -1;
+        }
+
+ out:
+        if (need_unwind && local && local->xl_specf_unwind) {
+                frame->local = local->xl_local;
+                local->xl_specf_unwind (frame, op_ret,
+                                        op_errno, dict, xdata);
+                return 0;
+        } else if (need_unwind){
+                STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno,
+                                     dict, xdata);
+        }
+        return 0;
+}
+
+
+int32_t
+cluster_getmarkerattr (call_frame_t *frame,xlator_t *this, loc_t *loc,
+                       const char *name, void *xl_local,
+                       xlator_specf_unwind_t xl_specf_getxattr_unwind,
+                       xlator_t **sub_volumes, int count, int type,
+                       int *gauge, char *vol_uuid)
+{
+        int                i     = 0;
+        xl_marker_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (name, err);
+        VALIDATE_OR_GOTO (xl_specf_getxattr_unwind, err);
+
+        local = GF_CALLOC (sizeof (struct marker_str), 1,
+                            gf_common_mt_libxl_marker_local);
+
+        if (!local)
+                goto err;
+
+        local->xl_local = xl_local;
+        local->call_count = count;
+        local->xl_specf_unwind = xl_specf_getxattr_unwind;
+        local->vol_uuid = vol_uuid;
+        memcpy (local->gauge, gauge, sizeof (local->gauge));
+
+        frame->local = local;
+
+        for (i=0; i < count; i++) {
+                if (MARKER_UUID_TYPE == type)
+                        STACK_WIND (frame, cluster_markeruuid_cbk,
+                                    *(sub_volumes + i),
+                                    (*(sub_volumes + i))->fops->getxattr,
+                                    loc, name, NULL);
+                else if (MARKER_XTIME_TYPE == type)
+                        STACK_WIND (frame, cluster_markerxtime_cbk,
+                                    *(sub_volumes + i),
+                                    (*(sub_volumes + i))->fops->getxattr,
+                                    loc, name, NULL);
+                else {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                 "Unrecognized type (%d) of marker attr "
+                                 "received", type);
+                        STACK_WIND (frame, default_getxattr_cbk,
+                                    *(sub_volumes + i),
+                                    (*(sub_volumes + i))->fops->getxattr,
+                                    loc, name, NULL);
+                        break;
+                }
+        }
+
+        return 0;
+err:
+        return -1;
+
+}
+
+int
+gf_get_min_stime (xlator_t *this, dict_t *dst, char *key, data_t *value)
+{
+        int ret = -1;
+        uint32_t *net_timebuf = NULL;
+        uint32_t *value_timebuf = NULL;
+        uint32_t host_timebuf[2] = {0,};
+        uint32_t host_value_timebuf[2] = {0,};
+
+        /* stime should be minimum of all the other nodes */
+        ret = dict_get_bin (dst, key, (void **)&net_timebuf);
+        if (ret < 0) {
+                net_timebuf = GF_CALLOC (1, sizeof (int64_t),
+                                           gf_common_mt_char);
+                if (!net_timebuf)
+                        goto out;
+
+                ret = dict_set_bin (dst, key, net_timebuf, sizeof (int64_t));
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "key=%s: dict set failed", key);
+                        goto error;
+                }
+        }
+
+        value_timebuf = data_to_bin (value);
+        if (!value_timebuf) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "key=%s: getting value of stime failed", key);
+                ret = -1;
+                goto out;
+        }
+
+        get_hosttime (value_timebuf, host_value_timebuf);
+        get_hosttime (net_timebuf, host_timebuf);
+
+        /* can't use 'min()' macro here as we need to compare two fields
+           in the array, selectively */
+        if ((host_value_timebuf[0] > host_timebuf[0]) ||
+            ((host_value_timebuf[0] == host_timebuf[0]) &&
+             (host_value_timebuf[1] > host_timebuf[1]))) {
+                update_timebuf (value_timebuf, net_timebuf);
+        }
+
+        ret = 0;
+out:
+        return ret;
+error:
+        /* To be used only when net_timebuf is not set in the dict */
+        if (net_timebuf)
+                GF_FREE (net_timebuf);
+
+        return ret;
+}
diff --git a/xlators/lib/src/libxlator.h b/xlators/lib/src/libxlator.h
new file mode 100644
index 000000000..1d5e1657f
--- /dev/null
+++ b/xlators/lib/src/libxlator.h
@@ -0,0 +1,153 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _LIBXLATOR_H
+#define _LIBXLATOR_H
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "logging.h"
+#include "defaults.h"
+#include "common-utils.h"
+#include "compat.h"
+#include "compat-errno.h"
+
+
+#define MARKER_XATTR_PREFIX "trusted.glusterfs"
+#define XTIME               "xtime"
+#define VOLUME_MARK         "volume-mark"
+#define GF_XATTR_MARKER_KEY MARKER_XATTR_PREFIX "." VOLUME_MARK
+#define UUID_SIZE 36
+#define MARKER_UUID_TYPE    1
+#define MARKER_XTIME_TYPE   2
+#define GF_XATTR_QUOTA_SIZE_KEY "trusted.glusterfs.quota.size"
+
+
+typedef int32_t (*xlator_specf_unwind_t) (call_frame_t *frame,
+                                          int op_ret, int op_errno,
+                                          dict_t *dict, dict_t *xdata);
+
+
+struct volume_mark {
+        uint8_t major;
+        uint8_t minor;
+        uint8_t uuid[16];
+        uint8_t retval;
+        uint32_t sec;
+        uint32_t usec;
+}__attribute__ ((__packed__));
+
+
+/*
+ * The enumerated type here
+ * is used to index two kind
+ * of integer arrays:
+ * - gauges
+ * - counters
+
+ * A counter is used internally,
+ * in getxattr callbacks, to count
+ * the results, categorized as
+ * the enum names suggest. So values
+ * in the counter are always non-negative.
+
+ * Gauges are part of the API.
+ * The caller passes one to the
+ * top-level aggregator function,
+ * cluster_getmarkerattr(). The gauge
+ * defines an evaluation policy for the
+ * counter. That is, at the
+ * end of the aggregation process
+ * the gauge is matched against the
+ * counter, and the policy
+ * represented by the gauge decides
+ * whether to return with success or failure,
+ * and in latter case, what particular failure
+ * case (errno).
+
+ * The rules are the following: for some index i,
+ * - if gauge[i] == 0, no requirement is set
+ *   against counter[i];
+ * - if gauge[i] > 0, counter[i] >= gauge[i]
+ *   is required;
+ * - if gauge[i] < 0, counter[i] < |gauge[i]|
+ *   is required.
+
+ * If the requirement is not met, then i is mapped
+ * to the respective errno (MCNT_ENOENT -> ENOENT),
+ * or in lack of that, EINVAL.
+
+ * Cf. evaluate_marker_results() and marker_idx_errno_map[]
+ * in libxlator.c
+
+ * We provide two default gauges, one inteded for xtime
+ * aggregation, other for volume mark aggregation. The
+ * policies they represent agree with the hard-coded
+ * one prior to gauges. Cf. marker_xtime_default_gauge
+ * and marker_uuid_default_gauge in libxlator.c
+ */
+
+typedef enum {
+        MCNT_FOUND,
+        MCNT_NOTFOUND,
+        MCNT_ENODATA,
+        MCNT_ENOTCONN,
+        MCNT_ENOENT,
+        MCNT_EOTHER,
+        MCNT_MAX
+} marker_result_idx_t;
+
+extern int marker_xtime_default_gauge[];
+extern int marker_uuid_default_gauge[];
+
+struct marker_str {
+        struct volume_mark    *volmark;
+        data_t                *data;
+
+        uint32_t               host_timebuf[2];
+        uint32_t               net_timebuf[2];
+        int32_t                call_count;
+        int                    gauge[MCNT_MAX];
+        int                    count[MCNT_MAX];
+
+        xlator_specf_unwind_t  xl_specf_unwind;
+        void                  *xl_local;
+        char                  *vol_uuid;
+        uint8_t                retval;
+};
+
+typedef struct marker_str xl_marker_local_t;
+
+int32_t
+cluster_markerxtime_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int op_ret, int op_errno, dict_t *dict, dict_t *xdata);
+
+int32_t
+cluster_markeruuid_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int op_ret, int op_errno, dict_t *dict, dict_t *xdata);
+
+int32_t
+cluster_getmarkerattr (call_frame_t *frame,xlator_t *this, loc_t *loc,
+                       const char *name, void *xl_local,
+                       xlator_specf_unwind_t xl_specf_getxattr_unwind,
+                       xlator_t **sub_volumes, int count, int type,
+                       int *gauge, char *vol_uuid);
+
+int
+match_uuid_local (const char *name, char *uuid);
+
+int
+gf_get_min_stime (xlator_t *this, dict_t *dst, char *key, data_t *value);
+
+#endif /* !_LIBXLATOR_H */
diff --git a/xlators/meta/src/Makefile.am b/xlators/meta/src/Makefile.am
index 385ff553f..f8fa7d4cb 100644
--- a/xlators/meta/src/Makefile.am
+++ b/xlators/meta/src/Makefile.am
@@ -4,7 +4,8 @@ xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/
 meta_so_SOURCES = meta.c tree.c misc.c view.c
 noinst_HEADERS = meta.h tree.h misc.h view.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall
 
 CLEANFILES = 
diff --git a/xlators/meta/src/meta-mem-types.h b/xlators/meta/src/meta-mem-types.h
index a9ec9435a..62028b246 100644
--- a/xlators/meta/src/meta-mem-types.h
+++ b/xlators/meta/src/meta-mem-types.h
@@ -1,23 +1,13 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
 
-
 #ifndef __META_MEM_TYPES_H__
 #define __META_MEM_TYPES_H__
 
diff --git a/xlators/meta/src/meta.c b/xlators/meta/src/meta.c
index face2fe70..e69719f3c 100644
--- a/xlators/meta/src/meta.c
+++ b/xlators/meta/src/meta.c
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #include <sys/types.h>
 #include <sys/stat.h>
 #include <unistd.h>
@@ -1136,7 +1126,7 @@ meta_lk_cbk (call_frame_t *frame,
 	     xlator_t *this,
 	     int32_t op_ret,
 	     int32_t op_errno,
-	     struct flock *lock)
+	     struct gf_flock *lock)
 {
   STACK_UNWIND (frame,
 		op_ret,
@@ -1150,7 +1140,7 @@ meta_lk (call_frame_t *frame,
 	 xlator_t *this,
 	 dict_t *file,
 	 int32_t cmd,
-	 struct flock *lock)
+	 struct gf_flock *lock)
 {
   STACK_WIND (frame,
 	      meta_lk_cbk,
diff --git a/xlators/meta/src/meta.h b/xlators/meta/src/meta.h
index 7f44162cc..73e0e50db 100644
--- a/xlators/meta/src/meta.h
+++ b/xlators/meta/src/meta.h
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __META_H__
 #define __META_H__
 
diff --git a/xlators/meta/src/misc.c b/xlators/meta/src/misc.c
index 062e741f2..1a8dfa806 100644
--- a/xlators/meta/src/misc.c
+++ b/xlators/meta/src/misc.c
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #include <unistd.h>
 #include <sys/uio.h>
 
diff --git a/xlators/meta/src/misc.h b/xlators/meta/src/misc.h
index 8ede1328b..30dd10e34 100644
--- a/xlators/meta/src/misc.h
+++ b/xlators/meta/src/misc.h
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __MISC_H__
 #define __MISC_H__
 
diff --git a/xlators/meta/src/tree.c b/xlators/meta/src/tree.c
index 787f27da4..dacbd665a 100644
--- a/xlators/meta/src/tree.c
+++ b/xlators/meta/src/tree.c
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -123,7 +113,7 @@ lookup_meta_entry (meta_dirent_t *root, const char *path,
 	    gf_asprintf (remain, "/%s/%s", *remain, piece);
 	  else
 	    gf_asprintf (remain, "/%s", piece);
-	  if (tmp) GF_FREE (tmp);
+	  GF_FREE (tmp);
 	  piece = strtok (NULL, "/");
 	}
       }
diff --git a/xlators/meta/src/tree.h b/xlators/meta/src/tree.h
index bb2ffb976..985df3bd7 100644
--- a/xlators/meta/src/tree.h
+++ b/xlators/meta/src/tree.h
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __TREE_H__
 #define __TREE_H__
 
diff --git a/xlators/meta/src/view.c b/xlators/meta/src/view.c
index cbb0b710a..b4e2d64a2 100644
--- a/xlators/meta/src/view.c
+++ b/xlators/meta/src/view.c
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
diff --git a/xlators/meta/src/view.h b/xlators/meta/src/view.h
index d26d42e26..2eff6126e 100644
--- a/xlators/meta/src/view.h
+++ b/xlators/meta/src/view.h
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __VIEW_H__
 #define __VIEW_H__
 
diff --git a/xlators/mgmt/glusterd/src/Makefile.am b/xlators/mgmt/glusterd/src/Makefile.am
index 81e542d65..933c44019 100644
--- a/xlators/mgmt/glusterd/src/Makefile.am
+++ b/xlators/mgmt/glusterd/src/Makefile.am
@@ -1,21 +1,50 @@
 xlator_LTLIBRARIES = glusterd.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/mgmt
-glusterd_la_LDFLAGS = -module -avoidversion
-glusterd_la_SOURCES = glusterd.c glusterd-handler.c glusterd-sm.c glusterd-op-sm.c \
-	glusterd-utils.c glusterd3_1-mops.c glusterd-store.c glusterd-handshake.c \
-	glusterd-pmap.c glusterd-volgen.c
+glusterd_la_CPPFLAGS = $(AM_CPPFLAGS) "-DFILTERDIR=\"$(libdir)/glusterfs/$(PACKAGE_VERSION)/filter\""
+glusterd_la_LDFLAGS = -module -avoid-version
+if ENABLE_BD_XLATOR
+glusterd_la_LDFLAGS += -llvm2app
+endif
+glusterd_la_SOURCES = glusterd.c glusterd-handler.c glusterd-sm.c \
+	glusterd-op-sm.c glusterd-utils.c glusterd-rpc-ops.c \
+	glusterd-store.c glusterd-handshake.c glusterd-pmap.c \
+	glusterd-volgen.c glusterd-rebalance.c glusterd-quota.c \
+	glusterd-geo-rep.c glusterd-replace-brick.c glusterd-log-ops.c \
+	glusterd-volume-ops.c glusterd-brick-ops.c glusterd-mountbroker.c \
+	glusterd-syncop.c glusterd-hooks.c glusterd-volume-set.c \
+        glusterd-locks.c glusterd-snapshot.c glusterd-mgmt-handler.c \
+        glusterd-mgmt.c
 
-glusterd_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la\
-		     $(top_builddir)/rpc/xdr/src/libgfxdr.la\
-		     $(top_builddir)/rpc/rpc-lib/src/libgfrpc.la
+glusterd_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \
+		     $(top_builddir)/rpc/xdr/src/libgfxdr.la \
+		     $(top_builddir)/rpc/rpc-lib/src/libgfrpc.la \
+		     $(XML_LIBS) -lcrypto
 
-noinst_HEADERS = glusterd.h glusterd-utils.h glusterd-op-sm.h glusterd-sm.h \
-	glusterd-store.h glusterd-mem-types.h glusterd-pmap.h glusterd-volgen.h
+noinst_HEADERS = glusterd.h glusterd-utils.h glusterd-op-sm.h \
+	glusterd-sm.h glusterd-store.h glusterd-mem-types.h \
+	glusterd-pmap.h glusterd-volgen.h glusterd-mountbroker.h \
+	glusterd-syncop.h glusterd-hooks.h glusterd-locks.h \
+	glusterd-mgmt.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)\
-	-I$(rpclibdir) -L$(xlatordir)/ -I$(CONTRIBDIR)/rbtree -I$(top_srcdir)/rpc/xdr/src\
-	-I$(top_srcdir)/rpc/rpc-lib/src -I$(CONTRIBDIR)/uuid -DGFS_PREFIX=\"$(prefix)\"
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+	-I$(rpclibdir) -I$(CONTRIBDIR)/rbtree \
+	-I$(top_srcdir)/rpc/xdr/src -I$(top_srcdir)/rpc/rpc-lib/src \
+	-I$(CONTRIBDIR)/uuid \
+	-DSBIN_DIR=\"$(sbindir)\" -DDATADIR=\"$(localstatedir)\" \
+	-DGSYNCD_PREFIX=\"$(libexecdir)/glusterfs\"\
+        -DSYNCDAEMON_COMPILE=$(SYNCDAEMON_COMPILE) $(XML_CPPFLAGS)
 
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+AM_LDFLAGS = -L$(xlatordir)
 
 CLEANFILES =
+
+install-data-hook:
+
+if GF_INSTALL_VAR_LIB_GLUSTERD
+	$(mkdir_p) $(localstatedir)/lib/
+	(stat $(sysconfdir)/glusterd && \
+	    mv $(sysconfdir)/glusterd $(localstatedir)/lib/) || true;
+	(ln -sf $(localstatedir)/lib/glusterd $(sysconfdir)/glusterd) || true;
+endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-brick-ops.c b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c
new file mode 100644
index 000000000..596503c21
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c
@@ -0,0 +1,1953 @@
+/*
+   Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "common-utils.h"
+#include "cli1-xdr.h"
+#include "xdr-generic.h"
+#include "glusterd.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-store.h"
+#include "glusterd-utils.h"
+#include "glusterd-volgen.h"
+#include "run.h"
+#include <sys/signal.h>
+
+/* misc */
+
+/* In this function, we decide, based on the 'count' of the brick,
+   where to add it in the current volume. 'count' tells us already
+   how many of the given bricks are added. other argument are self-
+   descriptive. */
+int
+add_brick_at_right_order (glusterd_brickinfo_t *brickinfo,
+                          glusterd_volinfo_t *volinfo, int count,
+                          int32_t stripe_cnt, int32_t replica_cnt)
+{
+        int                   idx     = 0;
+        int                   i       = 0;
+        int                   sub_cnt = 0;
+        glusterd_brickinfo_t *brick   = NULL;
+
+        /* The complexity of the function is in deciding at which index
+           to add new brick. Even though it can be defined with a complex
+           single formula for all volume, it is seperated out to make it
+           more readable */
+        if (stripe_cnt) {
+                /* common formula when 'stripe_count' is set */
+                /* idx = ((count / ((stripe_cnt * volinfo->replica_count) -
+                   volinfo->dist_leaf_count)) * volinfo->dist_leaf_count) +
+                   (count + volinfo->dist_leaf_count);
+                */
+
+                sub_cnt = volinfo->dist_leaf_count;
+
+                idx = ((count / ((stripe_cnt * volinfo->replica_count) -
+                                 sub_cnt)) * sub_cnt) +
+                        (count + sub_cnt);
+
+                goto insert_brick;
+        }
+
+        /* replica count is set */
+        /* common formula when 'replica_count' is set */
+        /* idx = ((count / (replica_cnt - existing_replica_count)) *
+           existing_replica_count) +
+           (count + existing_replica_count);
+        */
+
+        sub_cnt = volinfo->replica_count;
+        idx = (count / (replica_cnt - sub_cnt) * sub_cnt) +
+                (count + sub_cnt);
+
+insert_brick:
+        i = 0;
+        list_for_each_entry (brick, &volinfo->bricks, brick_list) {
+                i++;
+                if (i < idx)
+                        continue;
+                gf_log (THIS->name, GF_LOG_DEBUG, "brick:%s index=%d, count=%d",
+                        brick->path, idx, count);
+
+                list_add (&brickinfo->brick_list, &brick->brick_list);
+                break;
+        }
+
+        return 0;
+}
+
+
+static int
+gd_addbr_validate_stripe_count (glusterd_volinfo_t *volinfo, int stripe_count,
+                                int total_bricks, int *type, char *err_str,
+                                size_t err_len)
+{
+        int ret = -1;
+
+        switch (volinfo->type) {
+        case GF_CLUSTER_TYPE_NONE:
+                if ((volinfo->brick_count * stripe_count) == total_bricks) {
+                        /* Change the volume type */
+                        *type = GF_CLUSTER_TYPE_STRIPE;
+                        gf_log (THIS->name, GF_LOG_INFO,
+                                "Changing the type of volume %s from "
+                                "'distribute' to 'stripe'", volinfo->volname);
+                        ret = 0;
+                        goto out;
+                } else {
+                        snprintf (err_str, err_len, "Incorrect number of "
+                                  "bricks (%d) supplied for stripe count (%d).",
+                                  (total_bricks - volinfo->brick_count),
+                                  stripe_count);
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str);
+                        goto out;
+                }
+                break;
+        case GF_CLUSTER_TYPE_REPLICATE:
+                if (!(total_bricks % (volinfo->replica_count * stripe_count))) {
+                        /* Change the volume type */
+                        *type = GF_CLUSTER_TYPE_STRIPE_REPLICATE;
+                        gf_log (THIS->name, GF_LOG_INFO,
+                                "Changing the type of volume %s from "
+                                "'replicate' to 'replicate-stripe'",
+                                volinfo->volname);
+                        ret = 0;
+                        goto out;
+                } else {
+                        snprintf (err_str, err_len, "Incorrect number of "
+                                  "bricks (%d) supplied for changing volume's "
+                                  "stripe count to %d, need at least %d bricks",
+                                  (total_bricks - volinfo->brick_count),
+                                  stripe_count,
+                                  (volinfo->replica_count * stripe_count));
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str);
+                        goto out;
+                }
+                break;
+        case GF_CLUSTER_TYPE_STRIPE:
+        case GF_CLUSTER_TYPE_STRIPE_REPLICATE:
+                if (stripe_count < volinfo->stripe_count) {
+                        snprintf (err_str, err_len,
+                                  "Incorrect stripe count (%d) supplied. "
+                                  "Volume already has stripe count (%d)",
+                                  stripe_count, volinfo->stripe_count);
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str);
+                        goto out;
+                }
+                if (stripe_count == volinfo->stripe_count) {
+                        if (!(total_bricks % volinfo->dist_leaf_count)) {
+                                /* its same as the one which exists */
+                                ret = 1;
+                                goto out;
+                        }
+                }
+                if (stripe_count > volinfo->stripe_count) {
+                        /* We have to make sure before and after 'add-brick',
+                           the number or subvolumes for distribute will remain
+                           same, when stripe count is given */
+                        if ((volinfo->brick_count * (stripe_count *
+                                                     volinfo->replica_count)) ==
+                            (total_bricks * volinfo->dist_leaf_count)) {
+                                /* Change the dist_leaf_count */
+                                gf_log (THIS->name, GF_LOG_INFO,
+                                        "Changing the stripe count of "
+                                        "volume %s from %d to %d",
+                                        volinfo->volname,
+                                        volinfo->stripe_count, stripe_count);
+                                ret = 0;
+                                goto out;
+                        }
+                }
+                break;
+        }
+
+out:
+        return ret;
+}
+
+static int
+gd_addbr_validate_replica_count (glusterd_volinfo_t *volinfo, int replica_count,
+                                 int total_bricks, int *type, char *err_str,
+                                 int err_len)
+{
+        int ret = -1;
+
+        /* replica count is set */
+        switch (volinfo->type) {
+        case GF_CLUSTER_TYPE_NONE:
+                if ((volinfo->brick_count * replica_count) == total_bricks) {
+                        /* Change the volume type */
+                        *type = GF_CLUSTER_TYPE_REPLICATE;
+                        gf_log (THIS->name, GF_LOG_INFO,
+                                "Changing the type of volume %s from "
+                                "'distribute' to 'replica'", volinfo->volname);
+                        ret = 0;
+                        goto out;
+
+                } else {
+                        snprintf (err_str, err_len, "Incorrect number of "
+                                  "bricks (%d) supplied for replica count (%d).",
+                                  (total_bricks - volinfo->brick_count),
+                                  replica_count);
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str);
+                        goto out;
+                }
+                break;
+        case GF_CLUSTER_TYPE_STRIPE:
+                if (!(total_bricks % (volinfo->dist_leaf_count * replica_count))) {
+                        /* Change the volume type */
+                        *type = GF_CLUSTER_TYPE_STRIPE_REPLICATE;
+                        gf_log (THIS->name, GF_LOG_INFO,
+                                "Changing the type of volume %s from "
+                                "'stripe' to 'replicate-stripe'",
+                                volinfo->volname);
+                        ret = 0;
+                        goto out;
+                } else {
+                        snprintf (err_str, err_len, "Incorrect number of "
+                                  "bricks (%d) supplied for changing volume's "
+                                  "replica count to %d, need at least %d "
+                                  "bricks",
+                                  (total_bricks - volinfo->brick_count),
+                                  replica_count, (volinfo->dist_leaf_count *
+                                                  replica_count));
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str);
+                        goto out;
+                }
+                break;
+        case GF_CLUSTER_TYPE_REPLICATE:
+        case GF_CLUSTER_TYPE_STRIPE_REPLICATE:
+                if (replica_count < volinfo->replica_count) {
+                        snprintf (err_str, err_len,
+                                  "Incorrect replica count (%d) supplied. "
+                                  "Volume already has (%d)",
+                                  replica_count, volinfo->replica_count);
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str);
+                        goto out;
+                }
+                if (replica_count == volinfo->replica_count) {
+                        if (!(total_bricks % volinfo->dist_leaf_count)) {
+                                ret = 1;
+                                goto out;
+                        }
+                }
+                if (replica_count > volinfo->replica_count) {
+                        /* We have to make sure before and after 'add-brick',
+                           the number or subvolumes for distribute will remain
+                           same, when replica count is given */
+                        if ((total_bricks * volinfo->dist_leaf_count) ==
+                            (volinfo->brick_count * (replica_count *
+                                                     volinfo->stripe_count))) {
+                                /* Change the dist_leaf_count */
+                                gf_log (THIS->name, GF_LOG_INFO,
+                                        "Changing the replica count of "
+                                        "volume %s from %d to %d",
+                                        volinfo->volname, volinfo->replica_count,
+                                        replica_count);
+                                ret = 0;
+                                goto out;
+                        }
+                }
+                break;
+        }
+out:
+        return ret;
+}
+
+static int
+gd_rmbr_validate_replica_count (glusterd_volinfo_t *volinfo,
+                                int32_t replica_count,
+                                int32_t brick_count, char *err_str,
+                                size_t err_len)
+{
+        int ret = -1;
+        int replica_nodes = 0;
+
+        switch (volinfo->type) {
+        case GF_CLUSTER_TYPE_NONE:
+        case GF_CLUSTER_TYPE_STRIPE:
+                snprintf (err_str, err_len,
+                          "replica count (%d) option given for non replicate "
+                          "volume %s", replica_count, volinfo->volname);
+                gf_log (THIS->name, GF_LOG_WARNING, "%s", err_str);
+                goto out;
+
+        case GF_CLUSTER_TYPE_REPLICATE:
+        case GF_CLUSTER_TYPE_STRIPE_REPLICATE:
+                /* in remove brick, you can only reduce the replica count */
+                if (replica_count > volinfo->replica_count) {
+                        snprintf (err_str, err_len,
+                                  "given replica count (%d) option is more "
+                                  "than volume %s's replica count (%d)",
+                                  replica_count, volinfo->volname,
+                                  volinfo->replica_count);
+                        gf_log (THIS->name, GF_LOG_WARNING, "%s", err_str);
+                        goto out;
+                }
+                if (replica_count == volinfo->replica_count) {
+                        /* This means the 'replica N' option on CLI was
+                           redundant. Check if the total number of bricks given
+                           for removal is same as 'dist_leaf_count' */
+                        if (brick_count % volinfo->dist_leaf_count) {
+                                snprintf (err_str, err_len,
+                                          "number of bricks provided (%d) is "
+                                          "not valid. need at least %d "
+                                          "(or %dxN)", brick_count,
+                                          volinfo->dist_leaf_count,
+                                          volinfo->dist_leaf_count);
+                                gf_log (THIS->name, GF_LOG_WARNING, "%s",
+                                        err_str);
+                                goto out;
+                        }
+                        ret = 1;
+                        goto out;
+                }
+
+                replica_nodes = ((volinfo->brick_count /
+                                  volinfo->replica_count) *
+                                 (volinfo->replica_count - replica_count));
+
+                if (brick_count % replica_nodes) {
+                        snprintf (err_str, err_len,
+                                  "need %d(xN) bricks for reducing replica "
+                                  "count of the volume from %d to %d",
+                                  replica_nodes, volinfo->replica_count,
+                                  replica_count);
+                        goto out;
+                }
+                break;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+/* Handler functions */
+int
+__glusterd_handle_add_brick (rpcsvc_request_t *req)
+{
+        int32_t                         ret = -1;
+        gf_cli_req                      cli_req = {{0,}};
+        dict_t                          *dict = NULL;
+        char                            *bricks = NULL;
+        char                            *volname = NULL;
+        int                             brick_count = 0;
+        void                            *cli_rsp = NULL;
+        char                            err_str[2048] = {0,};
+        gf_cli_rsp                      rsp = {0,};
+        glusterd_volinfo_t              *volinfo = NULL;
+        xlator_t                        *this = NULL;
+        int                             total_bricks = 0;
+        int32_t                         replica_count = 0;
+        int32_t                         stripe_count = 0;
+        int                             type = 0;
+
+        this = THIS;
+        GF_ASSERT(this);
+
+        GF_ASSERT (req);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req,
+                              (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                snprintf (err_str, sizeof (err_str), "Garbage args received");
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_INFO, "Received add brick req");
+
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
+                        goto out;
+                }
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get volume "
+                          "name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        if (!(ret = glusterd_check_volume_exists (volname))) {
+                ret = -1;
+                snprintf (err_str, sizeof (err_str), "Volume %s does not exist",
+                          volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "count", &brick_count);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get volume "
+                          "brick count");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "replica-count", &replica_count);
+        if (!ret) {
+                gf_log (this->name, GF_LOG_INFO, "replica-count is %d",
+                        replica_count);
+        }
+
+        ret = dict_get_int32 (dict, "stripe-count", &stripe_count);
+        if (!ret) {
+                gf_log (this->name, GF_LOG_INFO, "stripe-count is %d",
+                        stripe_count);
+        }
+
+        if (!dict_get (dict, "force")) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get flag");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get volinfo "
+                          "for volume name %s", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+
+        }
+
+        total_bricks = volinfo->brick_count + brick_count;
+
+        if (!stripe_count && !replica_count) {
+                if (volinfo->type == GF_CLUSTER_TYPE_NONE)
+                        goto brick_val;
+
+                if ((volinfo->brick_count < volinfo->dist_leaf_count) &&
+                    (total_bricks <= volinfo->dist_leaf_count))
+                        goto brick_val;
+
+                if ((brick_count % volinfo->dist_leaf_count) != 0) {
+                        snprintf (err_str, sizeof (err_str), "Incorrect number "
+                                  "of bricks supplied %d with count %d",
+                                 brick_count, volinfo->dist_leaf_count);
+                        gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                        ret = -1;
+                        goto out;
+                }
+                goto brick_val;
+                /* done with validation.. below section is if stripe|replica
+                   count is given */
+        }
+
+        /* These bricks needs to be added one per a replica or stripe volume */
+        if (stripe_count) {
+                ret = gd_addbr_validate_stripe_count (volinfo, stripe_count,
+                                                      total_bricks, &type,
+                                                      err_str,
+                                                      sizeof (err_str));
+                if (ret == -1) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                        goto out;
+                }
+
+                /* if stripe count is same as earlier, set it back to 0 */
+                if (ret == 1)
+                        stripe_count = 0;
+
+                ret = dict_set_int32 (dict, "stripe-count", stripe_count);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to set the stripe-count in dict");
+                        goto out;
+                }
+                goto brick_val;
+        }
+
+        ret = gd_addbr_validate_replica_count (volinfo, replica_count,
+                                               total_bricks,
+                                               &type, err_str,
+                                               sizeof (err_str));
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        /* if replica count is same as earlier, set it back to 0 */
+        if (ret == 1)
+                replica_count = 0;
+
+        ret = dict_set_int32 (dict, "replica-count", replica_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to set the replica-count in dict");
+                goto out;
+        }
+
+brick_val:
+        ret = dict_get_str (dict, "bricks", &bricks);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get volume "
+                          "bricks");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        if (type != volinfo->type) {
+                ret = dict_set_int32 (dict, "type", type);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to set the new type in dict");
+        }
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_ADD_BRICK, dict);
+
+out:
+        if (ret) {
+                rsp.op_ret = -1;
+                rsp.op_errno = 0;
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str), "Operation failed");
+                rsp.op_errstr = err_str;
+                cli_rsp = &rsp;
+                glusterd_to_cli (req, cli_rsp, NULL, 0, NULL,
+                                 (xdrproc_t)xdr_gf_cli_rsp, dict);
+                ret = 0; //sent error to cli, prevent second reply
+        }
+
+        free (cli_req.dict.dict_val); //its malloced by xdr
+
+        return ret;
+}
+
+int
+glusterd_handle_add_brick (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __glusterd_handle_add_brick);
+}
+
+static int
+subvol_matcher_init (int **subvols, int count)
+{
+        int ret = -1;
+
+        *subvols = GF_CALLOC (count, sizeof(int), gf_gld_mt_int);
+        if (*subvols)
+                ret = 0;
+
+        return ret;
+}
+
+static void
+subvol_matcher_update (int *subvols, glusterd_volinfo_t *volinfo,
+                       glusterd_brickinfo_t *brickinfo)
+{
+        glusterd_brickinfo_t *tmp        = NULL;
+        int32_t               sub_volume = 0;
+        int                   pos        = 0;
+
+        list_for_each_entry (tmp, &volinfo->bricks, brick_list) {
+
+                if (strcmp (tmp->hostname, brickinfo->hostname) ||
+                    strcmp (tmp->path, brickinfo->path)) {
+                        pos++;
+                        continue;
+                }
+                gf_log (THIS->name, GF_LOG_DEBUG, LOGSTR_FOUND_BRICK,
+                                        brickinfo->hostname, brickinfo->path,
+                                        volinfo->volname);
+                sub_volume = (pos / volinfo->dist_leaf_count);
+                subvols[sub_volume]++;
+                break;
+        }
+
+}
+
+static int
+subvol_matcher_verify (int *subvols, glusterd_volinfo_t *volinfo, char *err_str,
+                       size_t err_len, char *vol_type)
+{
+        int i = 0;
+        int ret = 0;
+
+       do {
+
+                if (subvols[i] % volinfo->dist_leaf_count == 0) {
+                        continue;
+                } else {
+                        ret = -1;
+                        snprintf (err_str, err_len,
+                                "Bricks not from same subvol for %s", vol_type);
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", err_str);
+                        break;
+                }
+        } while (++i < volinfo->subvol_count);
+
+        return ret;
+}
+
+static void
+subvol_matcher_destroy (int *subvols)
+{
+        GF_FREE (subvols);
+}
+
+int
+__glusterd_handle_remove_brick (rpcsvc_request_t *req)
+{
+        int32_t                   ret              = -1;
+        gf_cli_req                cli_req          = {{0,}};
+        dict_t                   *dict             = NULL;
+        int32_t                   count            = 0;
+        char                     *brick            = NULL;
+        char                      key[256]         = {0,};
+        char                     *brick_list       = NULL;
+        int                       i                = 1;
+        glusterd_volinfo_t       *volinfo          = NULL;
+        glusterd_brickinfo_t     *brickinfo        = NULL;
+        int                      *subvols          = NULL;
+        glusterd_brickinfo_t     *tmp              = NULL;
+        char                      err_str[2048]    = {0};
+        gf_cli_rsp                rsp              = {0,};
+        void                     *cli_rsp          = NULL;
+        char                      vol_type[256]    = {0,};
+        int32_t                   replica_count    = 0;
+        int32_t                   brick_index      = 0;
+        int32_t                   tmp_brick_idx    = 0;
+        int                       found            = 0;
+        int                       diff_count       = 0;
+        char                     *volname          = 0;
+        xlator_t                 *this             = NULL;
+
+        GF_ASSERT (req);
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req,
+                              (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                snprintf (err_str, sizeof (err_str), "Received garbage args");
+                goto out;
+        }
+
+
+        gf_log (this->name, GF_LOG_INFO, "Received rem brick req");
+
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
+                        goto out;
+                }
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get volume "
+                          "name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "count", &count);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get brick "
+                          "count");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                 snprintf (err_str, sizeof (err_str),"Volume %s does not exist",
+                           volname);
+                 gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                 goto out;
+        }
+
+        ret = dict_get_int32 (dict, "replica-count", &replica_count);
+        if (!ret) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "request to change replica-count to %d", replica_count);
+                ret = gd_rmbr_validate_replica_count (volinfo, replica_count,
+                                                      count, err_str,
+                                                      sizeof (err_str));
+                if (ret < 0) {
+                        /* logging and error msg are done in above function
+                           itself */
+                        goto out;
+                }
+                dict_del (dict, "replica-count");
+                if (ret) {
+                        replica_count = 0;
+                } else {
+                        ret = dict_set_int32 (dict, "replica-count",
+                                              replica_count);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "failed to set the replica_count "
+                                        "in dict");
+                                goto out;
+                        }
+                }
+        }
+
+        /* 'vol_type' is used for giving the meaning full error msg for user */
+        if (volinfo->type == GF_CLUSTER_TYPE_REPLICATE) {
+                strcpy (vol_type, "replica");
+        } else if (volinfo->type == GF_CLUSTER_TYPE_STRIPE) {
+                strcpy (vol_type, "stripe");
+        } else if (volinfo->type == GF_CLUSTER_TYPE_STRIPE_REPLICATE) {
+                strcpy (vol_type, "stripe-replicate");
+        } else {
+                strcpy (vol_type, "distribute");
+        }
+
+	/* Do not allow remove-brick if the volume is a stripe volume*/
+	if ((volinfo->type == GF_CLUSTER_TYPE_STRIPE) &&
+            (volinfo->brick_count == volinfo->stripe_count)) {
+                snprintf (err_str, sizeof (err_str),
+                          "Removing brick from a stripe volume is not allowed");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                ret = -1;
+                goto out;
+	}
+
+	if (!replica_count &&
+            (volinfo->type == GF_CLUSTER_TYPE_STRIPE_REPLICATE) &&
+            (volinfo->brick_count == volinfo->dist_leaf_count)) {
+                snprintf (err_str, sizeof(err_str),
+                          "Removing bricks from stripe-replicate"
+                          " configuration is not allowed without reducing "
+                          "replica or stripe count explicitly.");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                ret = -1;
+                goto out;
+        }
+
+	if (!replica_count &&
+            (volinfo->type == GF_CLUSTER_TYPE_REPLICATE) &&
+            (volinfo->brick_count == volinfo->dist_leaf_count)) {
+                snprintf (err_str, sizeof (err_str),
+                          "Removing bricks from replicate configuration "
+                          "is not allowed without reducing replica count "
+                          "explicitly.");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                ret = -1;
+                goto out;
+        }
+
+	/* Do not allow remove-brick if the bricks given is less than
+           the replica count or stripe count */
+        if (!replica_count && (volinfo->type != GF_CLUSTER_TYPE_NONE)) {
+                if (volinfo->dist_leaf_count &&
+                    (count % volinfo->dist_leaf_count)) {
+                        snprintf (err_str, sizeof (err_str), "Remove brick "
+                                  "incorrect brick count of %d for %s %d",
+                                  count, vol_type, volinfo->dist_leaf_count);
+                        gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        brick_list = GF_MALLOC (120000 * sizeof(*brick_list),gf_common_mt_char);
+
+        if (!brick_list) {
+                ret = -1;
+                goto out;
+        }
+
+        strcpy (brick_list, " ");
+
+        if ((volinfo->type != GF_CLUSTER_TYPE_NONE) &&
+            (volinfo->subvol_count > 1)) {
+                ret = subvol_matcher_init (&subvols, volinfo->subvol_count);
+                if (ret)
+                        goto out;
+        }
+
+        while ( i <= count) {
+                snprintf (key, sizeof (key), "brick%d", i);
+                ret = dict_get_str (dict, key, &brick);
+                if (ret) {
+                        snprintf (err_str, sizeof (err_str), "Unable to get %s",
+                                  key);
+                        gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                        goto out;
+                }
+                gf_log (this->name, GF_LOG_DEBUG, "Remove brick count %d brick:"
+                        " %s", i, brick);
+
+                ret = glusterd_volume_brickinfo_get_by_brick(brick, volinfo,
+                                                             &brickinfo);
+                if (ret) {
+                        snprintf (err_str, sizeof (err_str), "Incorrect brick "
+                                  "%s for volume %s", brick, volname);
+                        gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                        goto out;
+                }
+                strcat(brick_list, brick);
+                strcat(brick_list, " ");
+
+                i++;
+                if ((volinfo->type == GF_CLUSTER_TYPE_NONE) ||
+                    (volinfo->brick_count <= volinfo->dist_leaf_count))
+                        continue;
+
+                if (replica_count) {
+                        /* do the validation of bricks here */
+                        /* -2 because i++ is already done, and i starts with 1,
+                           instead of 0 */
+                        diff_count = (volinfo->replica_count - replica_count);
+                        brick_index = (((i -2) / diff_count) * volinfo->replica_count);
+                        tmp_brick_idx = 0;
+                        found = 0;
+                        list_for_each_entry (tmp, &volinfo->bricks, brick_list) {
+                                tmp_brick_idx++;
+                                gf_log (this->name, GF_LOG_TRACE,
+                                        "validate brick %s:%s (%d %d %d)",
+                                        tmp->hostname, tmp->path, tmp_brick_idx,
+                                        brick_index, volinfo->replica_count);
+                                if (tmp_brick_idx <= brick_index)
+                                        continue;
+                                if (tmp_brick_idx >
+                                    (brick_index + volinfo->replica_count))
+                                        break;
+                                if ((!strcmp (tmp->hostname,brickinfo->hostname)) &&
+                                    !strcmp (tmp->path, brickinfo->path)) {
+                                        found = 1;
+                                        break;
+                                }
+                        }
+                        if (found)
+                                continue;
+
+                        snprintf (err_str, sizeof (err_str), "Bricks are from "
+                                  "same subvol");
+                        gf_log (this->name, GF_LOG_INFO,
+                                "failed to validate brick %s:%s (%d %d %d)",
+                                tmp->hostname, tmp->path, tmp_brick_idx,
+                                brick_index, volinfo->replica_count);
+                        ret = -1;
+                        /* brick order is not valid */
+                        goto out;
+                }
+
+                /* Find which subvolume the brick belongs to */
+                subvol_matcher_update (subvols, volinfo, brickinfo);
+        }
+
+        /* Check if the bricks belong to the same subvolumes.*/
+        if ((volinfo->type != GF_CLUSTER_TYPE_NONE) &&
+            (volinfo->subvol_count > 1)) {
+                ret = subvol_matcher_verify (subvols, volinfo,
+                                             err_str, sizeof(err_str),
+                                             vol_type);
+                if (ret)
+                        goto out;
+        }
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_REMOVE_BRICK, dict);
+
+out:
+        if (ret) {
+                rsp.op_ret = -1;
+                rsp.op_errno = 0;
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                rsp.op_errstr = err_str;
+                cli_rsp = &rsp;
+                glusterd_to_cli (req, cli_rsp, NULL, 0, NULL,
+                                 (xdrproc_t)xdr_gf_cli_rsp, dict);
+
+                ret = 0; //sent error to cli, prevent second reply
+
+        }
+
+        GF_FREE (brick_list);
+        subvol_matcher_destroy (subvols);
+        free (cli_req.dict.dict_val); //its malloced by xdr
+
+        return ret;
+}
+
+int
+glusterd_handle_remove_brick (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_remove_brick);
+}
+
+static int
+_glusterd_restart_gsync_session (dict_t *this, char *key,
+                                 data_t *value, void *data)
+{
+        char                          *slave      = NULL;
+        char                          *slave_buf  = NULL;
+        char                          *path_list  = NULL;
+        char                          *slave_vol  = NULL;
+        char                          *slave_ip   = NULL;
+        char                          *conf_path  = NULL;
+        char                         **errmsg     = NULL;
+        int                            ret        = -1;
+        glusterd_gsync_status_temp_t  *param      = NULL;
+        gf_boolean_t                   is_running = _gf_false;
+
+        param = (glusterd_gsync_status_temp_t *)data;
+
+        GF_ASSERT (param);
+        GF_ASSERT (param->volinfo);
+
+        slave = strchr(value->data, ':');
+        if (slave) {
+                slave++;
+                slave_buf = gf_strdup (slave);
+                if (!slave_buf) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Failed to gf_strdup");
+                        ret = -1;
+                        goto out;
+                }
+        }
+        else
+                return 0;
+
+        ret = dict_set_dynstr (param->rsp_dict, "slave", slave_buf);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to store slave");
+                if (slave_buf)
+                        GF_FREE(slave_buf);
+                goto out;
+        }
+
+        ret = glusterd_get_slave_details_confpath (param->volinfo,
+                                                   param->rsp_dict,
+                                                   &slave_ip, &slave_vol,
+                                                   &conf_path, errmsg);
+        if (ret) {
+                if (*errmsg)
+                        gf_log ("", GF_LOG_ERROR, "%s", *errmsg);
+                else
+                        gf_log ("", GF_LOG_ERROR,
+                                "Unable to fetch slave or confpath details.");
+                goto out;
+        }
+
+        /* In cases that gsyncd is not running, we will not invoke it
+         * because of add-brick. */
+        ret = glusterd_check_gsync_running_local (param->volinfo->volname,
+                                                  slave, conf_path,
+                                                  &is_running);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "gsync running validation failed.");
+                goto out;
+        }
+        if (_gf_false == is_running) {
+                gf_log ("", GF_LOG_DEBUG, "gsync session for %s and %s is"
+                        " not running on this node. Hence not restarting.",
+                        param->volinfo->volname, slave);
+                ret = 0;
+                goto out;
+        }
+
+        ret = glusterd_get_local_brickpaths (param->volinfo, &path_list);
+        if (!path_list) {
+                gf_log ("", GF_LOG_DEBUG, "This node not being part of"
+                        " volume should not be running gsyncd. Hence"
+                        " no gsyncd process to restart.");
+                ret = 0;
+                goto out;
+        }
+
+        ret = glusterd_check_restart_gsync_session (param->volinfo, slave,
+                                                    param->rsp_dict, path_list,
+                                                    conf_path, 0);
+        if (ret)
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to restart gsync session.");
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d.", ret);
+        return ret;
+}
+
+/* op-sm */
+
+int
+glusterd_op_perform_add_bricks (glusterd_volinfo_t *volinfo, int32_t count,
+                                char  *bricks, dict_t *dict)
+{
+        char                         *brick          = NULL;
+        int32_t                       i              = 1;
+        char                         *brick_list     = NULL;
+        char                         *free_ptr1      = NULL;
+        char                         *free_ptr2      = NULL;
+        char                         *saveptr        = NULL;
+        int32_t                       ret            = -1;
+        int32_t                       stripe_count   = 0;
+        int32_t                       replica_count  = 0;
+        int32_t                       type           = 0;
+        glusterd_brickinfo_t         *brickinfo      = NULL;
+        glusterd_gsync_status_temp_t  param          = {0, };
+        gf_boolean_t                  restart_needed = 0;
+        char                          msg[1024] __attribute__((unused)) = {0, };
+        int                           caps           = 0;
+
+        GF_ASSERT (volinfo);
+
+        if (bricks) {
+                brick_list = gf_strdup (bricks);
+                free_ptr1 = brick_list;
+        }
+
+        if (count)
+                brick = strtok_r (brick_list+1, " \n", &saveptr);
+
+        if (dict) {
+                ret = dict_get_int32 (dict, "stripe-count", &stripe_count);
+                if (!ret)
+                        gf_log (THIS->name, GF_LOG_INFO,
+                                "stripe-count is set %d", stripe_count);
+
+                ret = dict_get_int32 (dict, "replica-count", &replica_count);
+                if (!ret)
+                        gf_log (THIS->name, GF_LOG_INFO,
+                                "replica-count is set %d", replica_count);
+                ret = dict_get_int32 (dict, "type", &type);
+                if (!ret)
+                        gf_log (THIS->name, GF_LOG_INFO,
+                                "type is set %d, need to change it", type);
+        }
+
+        while ( i <= count) {
+                ret = glusterd_brickinfo_new_from_brick (brick, &brickinfo);
+                if (ret)
+                        goto out;
+
+                ret = glusterd_resolve_brick (brickinfo);
+                if (ret)
+                        goto out;
+                if (stripe_count || replica_count) {
+                        add_brick_at_right_order (brickinfo, volinfo, (i - 1),
+                                                  stripe_count, replica_count);
+                } else {
+                        list_add_tail (&brickinfo->brick_list, &volinfo->bricks);
+                }
+                brick = strtok_r (NULL, " \n", &saveptr);
+                i++;
+                volinfo->brick_count++;
+
+        }
+
+
+        /* Gets changed only if the options are given in add-brick cli */
+        if (type)
+                volinfo->type = type;
+        if (replica_count) {
+                volinfo->replica_count = replica_count;
+        }
+        if (stripe_count) {
+                volinfo->stripe_count = stripe_count;
+        }
+        volinfo->dist_leaf_count = glusterd_get_dist_leaf_count (volinfo);
+
+        /* backward compatibility */
+        volinfo->sub_count = ((volinfo->dist_leaf_count == 1) ? 0:
+                              volinfo->dist_leaf_count);
+
+        volinfo->subvol_count = (volinfo->brick_count /
+                                 volinfo->dist_leaf_count);
+
+        ret = glusterd_create_volfiles_and_notify_services (volinfo);
+        if (ret)
+                goto out;
+
+        ret = 0;
+        if (GLUSTERD_STATUS_STARTED != volinfo->status)
+                goto out;
+
+        brick_list = gf_strdup (bricks);
+        free_ptr2 = brick_list;
+        i = 1;
+
+        if (count)
+                brick = strtok_r (brick_list+1, " \n", &saveptr);
+#ifdef HAVE_BD_XLATOR
+        if (brickinfo->vg[0])
+                caps = CAPS_BD | CAPS_THIN |
+                        CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT;
+#endif
+
+        while (i <= count) {
+                ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo,
+                                                              &brickinfo);
+                if (ret)
+                        goto out;
+#ifdef HAVE_BD_XLATOR
+                /* Check for VG/thin pool if its BD volume */
+                if (brickinfo->vg[0]) {
+                        ret = glusterd_is_valid_vg (brickinfo, 0, msg);
+                        if (ret) {
+                                gf_log (THIS->name, GF_LOG_CRITICAL, "%s", msg);
+                                goto out;
+                        }
+                        /* if anyone of the brick does not have thin support,
+                           disable it for entire volume */
+                        caps &= brickinfo->caps;
+                } else
+                        caps = 0;
+#endif
+
+                if (uuid_is_null (brickinfo->uuid)) {
+                        ret = glusterd_resolve_brick (brickinfo);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK,
+                                        brickinfo->hostname, brickinfo->path);
+                                goto out;
+                        }
+                }
+
+                ret = glusterd_brick_start (volinfo, brickinfo,
+                                            _gf_true);
+                if (ret)
+                        goto out;
+                i++;
+                brick = strtok_r (NULL, " \n", &saveptr);
+
+                /* Check if the brick is added in this node, and set
+                 * the restart_needed flag. */
+                if ((!uuid_compare (brickinfo->uuid, MY_UUID)) &&
+                    !restart_needed) {
+                        restart_needed = 1;
+                        gf_log ("", GF_LOG_DEBUG,
+                                "Restart gsyncd session, if it's already "
+                                "running.");
+                }
+        }
+
+        /* If the restart_needed flag is set, restart gsyncd sessions for that
+         * particular master with all the slaves. */
+        if (restart_needed) {
+                param.rsp_dict = dict;
+                param.volinfo = volinfo;
+                dict_foreach (volinfo->gsync_slaves,
+                              _glusterd_restart_gsync_session, &param);
+        }
+        volinfo->caps = caps;
+out:
+        GF_FREE (free_ptr1);
+        GF_FREE (free_ptr2);
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+
+int
+glusterd_op_perform_remove_brick (glusterd_volinfo_t  *volinfo, char *brick,
+                                  int force, int *need_migrate)
+{
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        int32_t                  ret = -1;
+        glusterd_conf_t         *priv = NULL;
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (brick);
+
+        priv = THIS->private;
+        GF_ASSERT (priv);
+
+        ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo,
+                                                      &brickinfo);
+        if (ret)
+                goto out;
+
+        ret = glusterd_resolve_brick (brickinfo);
+        if (ret)
+                goto out;
+
+        glusterd_volinfo_reset_defrag_stats (volinfo);
+
+        if (!uuid_compare (brickinfo->uuid, MY_UUID)) {
+                /* Only if the brick is in this glusterd, do the rebalance */
+                if (need_migrate)
+                        *need_migrate = 1;
+        }
+
+        if (force) {
+                ret = glusterd_brick_stop (volinfo, brickinfo,
+                                           _gf_true);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_ERROR, "Unable to stop "
+                                "glusterfs, ret: %d", ret);
+                }
+                goto out;
+        }
+
+        brickinfo->decommissioned = 1;
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr)
+{
+        int                                     ret = 0;
+        char                                    *volname = NULL;
+        int                                     count = 0;
+        int                                     i = 0;
+        char                                    *bricks    = NULL;
+        char                                    *brick_list = NULL;
+        char                                    *saveptr = NULL;
+        char                                    *free_ptr = NULL;
+        char                                    *brick = NULL;
+        glusterd_brickinfo_t                    *brickinfo = NULL;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        glusterd_conf_t                         *priv = NULL;
+        char                                    msg[2048] = {0,};
+        gf_boolean_t                            brick_alloc = _gf_false;
+        char                                    *all_bricks = NULL;
+        char                                    *str_ret = NULL;
+        gf_boolean_t                            is_force = _gf_false;
+
+        priv = THIS->private;
+        if (!priv)
+                goto out;
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "Unable to get volume name");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "Unable to find volume: %s", volname);
+                goto out;
+        }
+
+        ret = glusterd_validate_volume_id (dict, volinfo);
+        if (ret)
+                goto out;
+
+        if (glusterd_is_rb_ongoing (volinfo)) {
+                snprintf (msg, sizeof (msg), "Replace brick is in progress on "
+                          "volume %s. Please retry after replace-brick "
+                          "operation is committed or aborted", volname);
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                ret = -1;
+                goto out;
+        }
+
+        if (glusterd_is_defrag_on(volinfo)) {
+                snprintf (msg, sizeof(msg), "Volume name %s rebalance is in "
+                          "progress. Please retry after completion", volname);
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                ret = -1;
+                goto out;
+        }
+        ret = dict_get_int32 (dict, "count", &count);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get count");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "bricks", &bricks);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Unable to get bricks");
+                goto out;
+        }
+
+        is_force = dict_get_str_boolean (dict, "force", _gf_false);
+
+        if (bricks) {
+                brick_list = gf_strdup (bricks);
+                all_bricks = gf_strdup (bricks);
+                free_ptr = brick_list;
+        }
+
+        if (count)
+                brick = strtok_r (brick_list+1, " \n", &saveptr);
+
+
+        while ( i < count) {
+                if (!glusterd_store_is_valid_brickpath (volname, brick) ||
+                    !glusterd_is_valid_volfpath (volname, brick)) {
+                        snprintf (msg, sizeof (msg), "brick path %s is "
+                                  "too long", brick);
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                        *op_errstr = gf_strdup (msg);
+
+                        ret = -1;
+                        goto out;
+
+                }
+
+                ret = glusterd_brickinfo_new_from_brick (brick, &brickinfo);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_ERROR,
+                                "Add-brick: Unable"
+                                " to get brickinfo");
+                        goto out;
+                }
+                brick_alloc = _gf_true;
+
+                ret = glusterd_new_brick_validate (brick, brickinfo, msg,
+                                                   sizeof (msg));
+                if (ret) {
+                        *op_errstr = gf_strdup (msg);
+                        ret = -1;
+                        goto out;
+                }
+
+                if (!uuid_compare (brickinfo->uuid, MY_UUID)) {
+#ifdef HAVE_BD_XLATOR
+                        if (brickinfo->vg[0]) {
+                                ret = glusterd_is_valid_vg (brickinfo, 1, msg);
+                                if (ret) {
+                                        gf_log (THIS->name, GF_LOG_ERROR, "%s",
+                                                msg);
+                                        *op_errstr = gf_strdup (msg);
+                                        goto out;
+                                }
+                        }
+#endif
+
+                        ret = glusterd_validate_and_create_brickpath (brickinfo,
+                                                          volinfo->volume_id,
+                                                          op_errstr, is_force);
+                        if (ret)
+                                goto out;
+                }
+
+                glusterd_brickinfo_delete (brickinfo);
+                brick_alloc = _gf_false;
+                brickinfo = NULL;
+                brick = strtok_r (NULL, " \n", &saveptr);
+                i++;
+        }
+
+out:
+        GF_FREE (free_ptr);
+        if (brick_alloc && brickinfo)
+                glusterd_brickinfo_delete (brickinfo);
+        GF_FREE (str_ret);
+        GF_FREE (all_bricks);
+
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+int
+glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr)
+{
+        int                 ret         = -1;
+        char               *volname     = NULL;
+        glusterd_volinfo_t *volinfo     = NULL;
+        char               *errstr      = NULL;
+        int32_t             brick_count = 0;
+        char                msg[2048]   = {0,};
+        int32_t             flag        = 0;
+        gf1_op_commands     cmd         = GF_OP_CMD_NONE;
+        char               *task_id_str = NULL;
+        xlator_t           *this        = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Volume %s does not exist", volname);
+                goto out;
+        }
+
+        ret = glusterd_validate_volume_id (dict, volinfo);
+        if (ret)
+                goto out;
+
+        if (glusterd_is_rb_ongoing (volinfo)) {
+                snprintf (msg, sizeof (msg), "Replace brick is in progress on "
+                          "volume %s. Please retry after replace-brick "
+                          "operation is committed or aborted", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "command", &flag);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get brick count");
+                goto out;
+        }
+        cmd = flag;
+
+        ret = -1;
+        switch (cmd) {
+        case GF_OP_CMD_NONE:
+                errstr = gf_strdup ("no remove-brick command issued");
+                goto out;
+
+        case GF_OP_CMD_STATUS:
+                ret = 0;
+                goto out;
+
+        case GF_OP_CMD_START:
+        {
+                if ((volinfo->type == GF_CLUSTER_TYPE_REPLICATE) &&
+                    dict_get (dict, "replica-count")) {
+                        snprintf (msg, sizeof(msg), "Migration of data is not "
+                                  "needed when reducing replica count. Use the"
+                                  " 'force' option");
+                        errstr = gf_strdup (msg);
+                        gf_log (this->name, GF_LOG_ERROR, "%s", errstr);
+                        goto out;
+                }
+
+                if (GLUSTERD_STATUS_STARTED != volinfo->status) {
+                        snprintf (msg, sizeof (msg), "Volume %s needs to be "
+                                  "started before remove-brick (you can use "
+                                  "'force' or 'commit' to override this "
+                                  "behavior)", volinfo->volname);
+                        errstr = gf_strdup (msg);
+                        gf_log (this->name, GF_LOG_ERROR, "%s", errstr);
+                        goto out;
+                }
+                if (!gd_is_remove_brick_committed (volinfo)) {
+                        snprintf (msg, sizeof (msg), "An earlier remove-brick "
+                                  "task exists for volume %s. Either commit it"
+                                  " or stop it before starting a new task.",
+                                  volinfo->volname);
+                        errstr = gf_strdup (msg);
+                        gf_log (this->name, GF_LOG_ERROR, "Earlier remove-brick"
+                                " task exists for volume %s.",
+                                volinfo->volname);
+                        goto out;
+                }
+                if (glusterd_is_defrag_on(volinfo)) {
+                        errstr = gf_strdup("Rebalance is in progress. Please "
+                                           "retry after completion");
+                        gf_log (this->name, GF_LOG_ERROR, "%s", errstr);
+                        goto out;
+                }
+
+                if (is_origin_glusterd (dict)) {
+                        ret = glusterd_generate_and_set_task_id
+                                (dict, GF_REMOVE_BRICK_TID_KEY);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to generate task-id");
+                                goto out;
+                        }
+                } else {
+                        ret = dict_get_str (dict, GF_REMOVE_BRICK_TID_KEY,
+                                            &task_id_str);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "Missing remove-brick-id");
+                                ret = 0;
+                        }
+                }
+                break;
+        }
+
+        case GF_OP_CMD_STOP:
+                ret = 0;
+                break;
+
+        case GF_OP_CMD_COMMIT:
+                if (volinfo->decommission_in_progress) {
+                        errstr = gf_strdup ("use 'force' option as migration "
+                                            "is in progress");
+                        goto out;
+                }
+                break;
+
+        case GF_OP_CMD_COMMIT_FORCE:
+                break;
+        }
+
+        ret = dict_get_int32 (dict, "count", &brick_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get brick count");
+                goto out;
+        }
+
+        ret = 0;
+        if (volinfo->brick_count == brick_count) {
+                errstr = gf_strdup ("Deleting all the bricks of the "
+                                    "volume is not allowed");
+                ret = -1;
+                goto out;
+        }
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        if (ret && errstr) {
+                if (op_errstr)
+                        *op_errstr = errstr;
+        }
+
+        return ret;
+}
+
+int
+glusterd_remove_brick_migrate_cbk (glusterd_volinfo_t *volinfo,
+                                   gf_defrag_status_t status)
+{
+        int                   ret = 0;
+
+#if 0  /* TODO: enable this behavior once cluster-wide awareness comes for
+          defrag cbk function */
+        glusterd_brickinfo_t *brickinfo = NULL;
+        glusterd_brickinfo_t *tmp = NULL;
+
+        switch (status) {
+        case GF_DEFRAG_STATUS_PAUSED:
+        case GF_DEFRAG_STATUS_FAILED:
+                /* No changes required in the volume file.
+                   everything should remain as is */
+                break;
+        case GF_DEFRAG_STATUS_STOPPED:
+                /* Fall back to the old volume file */
+                list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks, brick_list) {
+                        if (!brickinfo->decommissioned)
+                                continue;
+                        brickinfo->decommissioned = 0;
+                }
+                break;
+
+        case GF_DEFRAG_STATUS_COMPLETE:
+                /* Done with the task, you can remove the brick from the
+                   volume file */
+                list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks, brick_list) {
+                        if (!brickinfo->decommissioned)
+                                continue;
+                        gf_log (THIS->name, GF_LOG_INFO, "removing the brick %s",
+                                brickinfo->path);
+                        brickinfo->decommissioned = 0;
+                        if (GLUSTERD_STATUS_STARTED == volinfo->status) {
+                            /*TODO: use the 'atomic' flavour of brick_stop*/
+                                ret = glusterd_brick_stop (volinfo, brickinfo);
+                                if (ret) {
+                                        gf_log (THIS->name, GF_LOG_ERROR,
+                                                "Unable to stop glusterfs (%d)", ret);
+                                }
+                        }
+                        glusterd_delete_brick (volinfo, brickinfo);
+                }
+                break;
+
+        default:
+                GF_ASSERT (!"cbk function called with wrong status");
+                break;
+        }
+
+        ret = glusterd_create_volfiles_and_notify_services (volinfo);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "Unable to write volume files (%d)", ret);
+
+        ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "Unable to store volume info (%d)", ret);
+
+
+        if (GLUSTERD_STATUS_STARTED == volinfo->status) {
+                ret = glusterd_check_generate_start_nfs ();
+                if (ret)
+                        gf_log (THIS->name, GF_LOG_ERROR,
+                                "Unable to start nfs process (%d)", ret);
+        }
+
+#endif
+
+        volinfo->decommission_in_progress = 0;
+        return ret;
+}
+
+
+int
+glusterd_op_add_brick (dict_t *dict, char **op_errstr)
+{
+        int                                     ret = 0;
+        char                                    *volname = NULL;
+        glusterd_conf_t                         *priv = NULL;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        xlator_t                                *this = NULL;
+        char                                    *bricks = NULL;
+        int32_t                                 count = 0;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_str (dict, "volname", &volname);
+
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "count", &count);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get count");
+                goto out;
+        }
+
+
+        ret = dict_get_str (dict, "bricks", &bricks);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get bricks");
+                goto out;
+        }
+
+        ret = glusterd_op_perform_add_bricks (volinfo, count, bricks, dict);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to add bricks");
+                goto out;
+        }
+
+        ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret)
+                goto out;
+
+        if (GLUSTERD_STATUS_STARTED == volinfo->status)
+                ret = glusterd_nodesvcs_handle_graph_change (volinfo);
+
+out:
+        return ret;
+}
+
+int
+glusterd_op_remove_brick (dict_t *dict, char **op_errstr)
+{
+        int                     ret            = -1;
+        char                    *volname       = NULL;
+        glusterd_volinfo_t      *volinfo       = NULL;
+        char                    *brick         = NULL;
+        int32_t                 count          = 0;
+        int32_t                 i              = 1;
+        char                    key[256]       = {0,};
+        int32_t                 flag           = 0;
+        char                    err_str[4096]  = {0,};
+        int                     need_rebalance = 0;
+        int                     force          = 0;
+        gf1_op_commands         cmd            = 0;
+        int32_t                 replica_count  = 0;
+        glusterd_brickinfo_t    *brickinfo     = NULL;
+        glusterd_brickinfo_t    *tmp           = NULL;
+        char                    *task_id_str   = NULL;
+        xlator_t                *this          = NULL;
+        dict_t                  *bricks_dict   = NULL;
+        char                    *brick_tmpstr  = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = dict_get_str (dict, "volname", &volname);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to allocate memory");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "command", &flag);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get command");
+                goto out;
+        }
+        cmd = flag;
+
+        /* Set task-id, if available, in ctx dict for operations other than
+         * start
+         */
+        if (is_origin_glusterd (dict) && (cmd != GF_OP_CMD_START)) {
+                if (!uuid_is_null (volinfo->rebal.rebalance_id)) {
+                        ret = glusterd_copy_uuid_to_dict
+                                (volinfo->rebal.rebalance_id, dict,
+                                 GF_REMOVE_BRICK_TID_KEY);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set remove-brick-id");
+                                goto out;
+                        }
+                }
+        }
+
+        /* Clear task-id, rebal.op and stored bricks on commmitting/stopping
+         * remove-brick */
+        if ((cmd != GF_OP_CMD_START) || (cmd != GF_OP_CMD_STATUS)) {
+                uuid_clear (volinfo->rebal.rebalance_id);
+                volinfo->rebal.op = GD_OP_NONE;
+                dict_unref (volinfo->rebal.dict);
+                volinfo->rebal.dict = NULL;
+        }
+
+        ret = -1;
+        switch (cmd) {
+        case GF_OP_CMD_NONE:
+                goto out;
+
+        case GF_OP_CMD_STATUS:
+                ret = 0;
+                goto out;
+
+        case GF_OP_CMD_STOP:
+        {
+                /* Fall back to the old volume file */
+                list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks,
+                                          brick_list) {
+                        if (!brickinfo->decommissioned)
+                                continue;
+                        brickinfo->decommissioned = 0;
+                }
+                ret = glusterd_create_volfiles_and_notify_services (volinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "failed to create volfiles");
+                        goto out;
+                }
+
+                ret = glusterd_store_volinfo (volinfo,
+                                             GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "failed to store volinfo");
+                        goto out;
+                }
+
+                ret = 0;
+                goto out;
+        }
+
+        case GF_OP_CMD_START:
+                ret = dict_get_str (dict, GF_REMOVE_BRICK_TID_KEY, &task_id_str);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Missing remove-brick-id");
+                        ret = 0;
+                } else {
+                        uuid_parse (task_id_str, volinfo->rebal.rebalance_id) ;
+                        volinfo->rebal.op = GD_OP_REMOVE_BRICK;
+                }
+                force = 0;
+                break;
+
+        case GF_OP_CMD_COMMIT:
+                force = 1;
+                break;
+
+        case GF_OP_CMD_COMMIT_FORCE:
+
+                if (volinfo->decommission_in_progress) {
+                        if (volinfo->rebal.defrag) {
+                                LOCK (&volinfo->rebal.defrag->lock);
+                                /* Fake 'rebalance-complete' so the graph change
+                                   happens right away */
+                                volinfo->rebal.defrag_status =
+                                                GF_DEFRAG_STATUS_COMPLETE;
+
+                                UNLOCK (&volinfo->rebal.defrag->lock);
+                        }
+                        /* Graph change happens in rebalance _cbk function,
+                           no need to do anything here */
+                        /* TODO: '_cbk' function is not doing anything for now */
+                }
+
+                ret = 0;
+                force = 1;
+                break;
+        }
+
+        ret = dict_get_int32 (dict, "count", &count);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get count");
+                goto out;
+        }
+
+        /* Save the list of bricks for later usage. Right now this is required
+         * for displaying the task parameters with task status in volume status.
+         */
+        bricks_dict = dict_new ();
+        if (!bricks_dict) {
+                ret = -1;
+                goto out;
+        }
+        ret = dict_set_int32 (bricks_dict, "count", count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to save remove-brick count");
+                goto out;
+        }
+        while ( i <= count) {
+                snprintf (key, 256, "brick%d", i);
+                ret = dict_get_str (dict, key, &brick);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Unable to get %s",
+                                key);
+                        goto out;
+                }
+
+                brick_tmpstr = gf_strdup (brick);
+                if (!brick_tmpstr) {
+                        ret = -1;
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to duplicate brick name");
+                        goto out;
+                }
+                ret = dict_set_dynstr (bricks_dict, key, brick_tmpstr);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to add brick to dict");
+                        goto out;
+                }
+                brick_tmpstr = NULL;
+
+                ret = glusterd_op_perform_remove_brick (volinfo, brick, force,
+                                                        &need_rebalance);
+                if (ret)
+                        goto out;
+                i++;
+        }
+        ret = dict_get_int32 (dict, "replica-count", &replica_count);
+        if (!ret) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "changing replica count %d to %d on volume %s",
+                        volinfo->replica_count, replica_count,
+                        volinfo->volname);
+                volinfo->replica_count = replica_count;
+                volinfo->sub_count = replica_count;
+                volinfo->dist_leaf_count = glusterd_get_dist_leaf_count (volinfo);
+                volinfo->subvol_count = (volinfo->brick_count /
+                                         volinfo->dist_leaf_count);
+
+                if (replica_count == 1) {
+                        if (volinfo->type == GF_CLUSTER_TYPE_REPLICATE) {
+                                volinfo->type = GF_CLUSTER_TYPE_NONE;
+                                /* backward compatibility */
+                                volinfo->sub_count = 0;
+                        } else {
+                                volinfo->type = GF_CLUSTER_TYPE_STRIPE;
+                                /* backward compatibility */
+                                volinfo->sub_count = volinfo->dist_leaf_count;
+                        }
+                }
+        }
+        volinfo->rebal.dict = bricks_dict;
+        bricks_dict = NULL;
+
+        ret = glusterd_create_volfiles_and_notify_services (volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "failed to create volfiles");
+                goto out;
+        }
+
+        ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "failed to store volinfo");
+                goto out;
+        }
+
+        /* Need to reset the defrag/rebalance status accordingly */
+        switch (volinfo->rebal.defrag_status) {
+        case GF_DEFRAG_STATUS_FAILED:
+        case GF_DEFRAG_STATUS_COMPLETE:
+                volinfo->rebal.defrag_status = 0;
+        default:
+                break;
+        }
+        if (!force && need_rebalance) {
+                /* perform the rebalance operations */
+                ret = glusterd_handle_defrag_start
+                        (volinfo, err_str, sizeof (err_str),
+                         GF_DEFRAG_CMD_START_FORCE,
+                         glusterd_remove_brick_migrate_cbk, GD_OP_REMOVE_BRICK);
+
+                if (!ret)
+                        volinfo->decommission_in_progress = 1;
+
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to start the rebalance");
+                }
+        } else {
+                if (GLUSTERD_STATUS_STARTED == volinfo->status)
+                        ret = glusterd_nodesvcs_handle_graph_change (volinfo);
+        }
+
+out:
+        if (ret && err_str[0] && op_errstr)
+                *op_errstr = gf_strdup (err_str);
+
+        GF_FREE (brick_tmpstr);
+        if (bricks_dict)
+                dict_unref (bricks_dict);
+
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-geo-rep.c b/xlators/mgmt/glusterd/src/glusterd-geo-rep.c
new file mode 100644
index 000000000..5786694bd
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-geo-rep.c
@@ -0,0 +1,4236 @@
+/*
+   Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "common-utils.h"
+#include "cli1-xdr.h"
+#include "xdr-generic.h"
+#include "glusterd.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-store.h"
+#include "glusterd-utils.h"
+#include "glusterd-volgen.h"
+#include "run.h"
+#include "syscall.h"
+
+#include <signal.h>
+
+static int
+dict_get_param (dict_t *dict, char *key, char **param);
+
+static int
+glusterd_get_statefile_name (glusterd_volinfo_t *volinfo, char *slave,
+                             char *conf_path, char **statefile);
+
+static int
+glusterd_get_slave_info (char *slave, char **slave_ip,
+                         char **slave_vol, char **op_errstr);
+
+static int
+glusterd_gsync_read_frm_status (char *path, char *buf, size_t blen);
+
+struct gsync_config_opt_vals_ gsync_confopt_vals[] = {
+        {.op_name        = "change_detector",
+         .no_of_pos_vals = 2,
+         .case_sensitive = _gf_true,
+         .values         = {"xsync", "changelog"},
+        },
+        {.op_name        = "special_sync_mode",
+         .no_of_pos_vals = 2,
+         .case_sensitive = _gf_true,
+         .values         = {"partial", "recover"}
+        },
+        {.op_name        = "log-level",
+         .no_of_pos_vals = 5,
+         .case_sensitive = _gf_false,
+         .values         = {"critical", "error", "warning", "info", "debug"}
+        },
+        {.op_name = NULL,
+        },
+};
+
+static char *gsync_reserved_opts[] = {
+        "gluster-command-dir",
+        "pid-file",
+        "remote-gsyncd"
+        "state-file",
+        "session-owner",
+        "state-socket-unencoded",
+        "socketdir",
+        "ignore-deletes",
+        "local-id",
+        "local-path",
+        "slave-id",
+        NULL
+};
+
+int
+__glusterd_handle_sys_exec (rpcsvc_request_t *req)
+{
+        int32_t                 ret     = 0;
+        dict_t                  *dict   = NULL;
+        gf_cli_req              cli_req = {{0},};
+        glusterd_op_t           cli_op = GD_OP_SYS_EXEC;
+        glusterd_conf_t         *priv   = NULL;
+        char                    *host_uuid = NULL;
+        char                    err_str[2048] = {0,};
+        xlator_t                *this = NULL;
+
+        GF_ASSERT (req);
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req,
+                              (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (cli_req.dict.dict_len) {
+                dict = dict_new ();
+                if (!dict)
+                        goto out;
+
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
+                        goto out;
+               } else {
+                        dict->extra_stdfree = cli_req.dict.dict_val;
+                }
+
+                host_uuid = gf_strdup (uuid_utoa(MY_UUID));
+                if (host_uuid == NULL) {
+                        snprintf (err_str, sizeof (err_str), "Failed to get "
+                                  "the uuid of local glusterd");
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = dict_set_dynstr (dict, "host-uuid", host_uuid);
+                if (ret)
+                        goto out;
+        }
+
+        ret = glusterd_op_begin_synctask (req, cli_op, dict);
+
+out:
+        if (ret) {
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
+        }
+        return ret;
+}
+
+int
+__glusterd_handle_copy_file (rpcsvc_request_t *req)
+{
+        int32_t                 ret     = 0;
+        dict_t                  *dict   = NULL;
+        gf_cli_req              cli_req = {{0},};
+        glusterd_op_t           cli_op = GD_OP_COPY_FILE;
+        glusterd_conf_t         *priv   = NULL;
+        char                    *host_uuid = NULL;
+        char                    err_str[2048] = {0,};
+        xlator_t                *this = NULL;
+
+        GF_ASSERT (req);
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req,
+                              (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (cli_req.dict.dict_len) {
+                dict = dict_new ();
+                if (!dict)
+                        goto out;
+
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
+                        goto out;
+               } else {
+                        dict->extra_stdfree = cli_req.dict.dict_val;
+                }
+
+                host_uuid = gf_strdup (uuid_utoa(MY_UUID));
+                if (host_uuid == NULL) {
+                        snprintf (err_str, sizeof (err_str), "Failed to get "
+                                  "the uuid of local glusterd");
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = dict_set_dynstr (dict, "host-uuid", host_uuid);
+                if (ret)
+                        goto out;
+        }
+
+        ret = glusterd_op_begin_synctask (req, cli_op, dict);
+
+out:
+        if (ret) {
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
+        }
+        return ret;
+}
+
+int
+__glusterd_handle_gsync_set (rpcsvc_request_t *req)
+{
+        int32_t                 ret     = 0;
+        dict_t                  *dict   = NULL;
+        gf_cli_req              cli_req = {{0},};
+        glusterd_op_t           cli_op = GD_OP_GSYNC_SET;
+        char                    *master = NULL;
+        char                    *slave = NULL;
+        char                    operation[256] = {0,};
+        int                     type = 0;
+        glusterd_conf_t         *priv   = NULL;
+        char                    *host_uuid = NULL;
+        char                    err_str[2048] = {0,};
+        xlator_t                *this = NULL;
+
+        GF_ASSERT (req);
+
+       this = THIS;
+       GF_ASSERT (this);
+       priv = this->private;
+       GF_ASSERT (priv);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req,
+                              (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (cli_req.dict.dict_len) {
+                dict = dict_new ();
+                if (!dict)
+                        goto out;
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
+                        goto out;
+                } else {
+                        dict->extra_stdfree = cli_req.dict.dict_val;
+                }
+
+                host_uuid = gf_strdup (uuid_utoa(MY_UUID));
+                if (host_uuid == NULL) {
+                        snprintf (err_str, sizeof (err_str), "Failed to get "
+                                  "the uuid of local glusterd");
+                        ret = -1;
+                        goto out;
+                }
+                ret = dict_set_dynstr (dict, "host-uuid", host_uuid);
+                if (ret)
+                        goto out;
+
+        }
+
+        ret = dict_get_str (dict, "master", &master);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_INFO, "master not found, while "
+                        "handling "GEOREP" options");
+                master = "(No Master)";
+        }
+
+        ret = dict_get_str (dict, "slave", &slave);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_INFO, "slave not found, while "
+                        "handling "GEOREP" options");
+                slave = "(No Slave)";
+        }
+
+        ret = dict_get_int32 (dict, "type", &type);
+        if (ret < 0) {
+                snprintf (err_str, sizeof (err_str), "Command type not found "
+                          "while handling "GEOREP" options");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        switch (type) {
+        case GF_GSYNC_OPTION_TYPE_CREATE:
+                strncpy (operation, "create", sizeof (operation));
+                cli_op = GD_OP_GSYNC_CREATE;
+                break;
+
+        case GF_GSYNC_OPTION_TYPE_START:
+                strncpy (operation, "start", sizeof (operation));
+                break;
+
+        case GF_GSYNC_OPTION_TYPE_STOP:
+                strncpy (operation, "stop", sizeof (operation));
+                break;
+
+        case GF_GSYNC_OPTION_TYPE_CONFIG:
+                strncpy (operation, "config", sizeof (operation));
+                break;
+
+        case GF_GSYNC_OPTION_TYPE_STATUS:
+                strncpy (operation, "status", sizeof (operation));
+                break;
+        }
+
+        ret = glusterd_op_begin_synctask (req, cli_op, dict);
+
+out:
+        if (ret) {
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
+        }
+        return ret;
+}
+
+int
+glusterd_handle_sys_exec (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __glusterd_handle_sys_exec);
+}
+
+int
+glusterd_handle_copy_file (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __glusterd_handle_copy_file);
+}
+
+int
+glusterd_handle_gsync_set (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __glusterd_handle_gsync_set);
+}
+
+/*****
+ *
+ * glusterd_urltransform* internal API
+ *
+ *****/
+
+static void
+glusterd_urltransform_init (runner_t *runner, const char *transname)
+{
+        runinit (runner);
+        runner_add_arg (runner, GSYNCD_PREFIX"/gsyncd");
+        runner_argprintf (runner, "--%s-url", transname);
+}
+
+static void
+glusterd_urltransform_add (runner_t *runner, const char *url)
+{
+        runner_add_arg (runner, url);
+}
+
+static int
+_glusterd_urltransform_add_iter (dict_t *dict, char *key, data_t *value, void *data)
+{
+        runner_t *runner = (runner_t *)data;
+        char *slave = NULL;
+
+        slave = strchr (value->data, ':');
+        GF_ASSERT (slave);
+        slave++;
+        runner_add_arg (runner, slave);
+
+        return 0;
+}
+
+static void
+glusterd_urltransform_free (char **linearr, unsigned n)
+{
+        int i = 0;
+
+        for (; i < n; i++)
+                GF_FREE (linearr[i]);
+
+        GF_FREE (linearr);
+}
+
+static int
+glusterd_urltransform (runner_t *runner, char ***linearrp)
+{
+        char **linearr = NULL;
+        char *line = NULL;
+        unsigned arr_len = 32;
+        unsigned arr_idx = 0;
+        gf_boolean_t error = _gf_false;
+
+        linearr = GF_CALLOC (arr_len, sizeof (char *), gf_gld_mt_linearr);
+        if (!linearr) {
+                error = _gf_true;
+                goto out;
+        }
+
+        runner_redir (runner, STDOUT_FILENO, RUN_PIPE);
+        if (runner_start (runner) != 0) {
+                gf_log ("", GF_LOG_ERROR, "spawning child failed");
+
+                error = _gf_true;
+                goto out;
+        }
+
+        arr_idx = 0;
+        for (;;) {
+                size_t len;
+                line = GF_MALLOC (1024, gf_gld_mt_linebuf);
+                if (!line) {
+                        error = _gf_true;
+                        goto out;
+                }
+
+                if (fgets (line, 1024, runner_chio (runner, STDOUT_FILENO)) ==
+                    NULL)
+                        break;
+
+                len = strlen (line);
+                if (len == 0 || line[len - 1] != '\n') {
+                        GF_FREE (line);
+                        error = _gf_true;
+                        goto out;
+                }
+                line[len - 1] = '\0';
+
+                if (arr_idx == arr_len) {
+                        void *p = linearr;
+                        arr_len <<= 1;
+                        p = GF_REALLOC (linearr, arr_len);
+                        if (!p) {
+                                GF_FREE (line);
+                                error = _gf_true;
+                                goto out;
+                        }
+                        linearr = p;
+                }
+                linearr[arr_idx] = line;
+
+                arr_idx++;
+        }
+
+ out:
+
+        /* XXX chpid field is not exported by run API
+         * but runner_end() does not abort the invoked
+         * process (ie. it might block in waitpid(2))
+         * so we resort to a manual kill a the private field
+         */
+        if (error && runner->chpid > 0)
+                kill (runner->chpid, SIGKILL);
+
+        if (runner_end (runner) != 0)
+                error = _gf_true;
+
+        if (error) {
+                gf_log ("", GF_LOG_ERROR, "reading data from child failed");
+                glusterd_urltransform_free (linearr, arr_idx);
+                return -1;
+        }
+
+        *linearrp = linearr;
+        return arr_idx;
+}
+
+static int
+glusterd_urltransform_single (const char *url, const char *transname,
+                              char ***linearrp)
+{
+        runner_t runner = {0,};
+
+        glusterd_urltransform_init (&runner, transname);
+        glusterd_urltransform_add (&runner, url);
+        return glusterd_urltransform (&runner, linearrp);
+}
+
+
+struct dictidxmark {
+        unsigned isrch;
+        unsigned ithis;
+        char *ikey;
+};
+
+static int
+_dict_mark_atindex (dict_t *dict, char *key, data_t *value, void *data)
+{
+        struct dictidxmark *dim = data;
+
+        if (dim->isrch == dim->ithis)
+                dim->ikey = key;
+
+        dim->ithis++;
+        return 0;
+}
+
+static char *
+dict_get_by_index (dict_t *dict, unsigned i)
+{
+        struct dictidxmark dim = {0,};
+
+        dim.isrch = i;
+        dict_foreach (dict, _dict_mark_atindex, &dim);
+
+        return dim.ikey;
+}
+
+static int
+glusterd_get_slave (glusterd_volinfo_t *vol, const char *slaveurl, char **slavekey)
+{
+        runner_t runner = {0,};
+        int n = 0;
+        int i = 0;
+        char **linearr = NULL;
+
+        glusterd_urltransform_init (&runner, "canonicalize");
+        dict_foreach (vol->gsync_slaves, _glusterd_urltransform_add_iter, &runner);
+        glusterd_urltransform_add (&runner, slaveurl);
+
+        n = glusterd_urltransform (&runner, &linearr);
+        if (n == -1)
+                return -2;
+
+        for (i = 0; i < n - 1; i++) {
+                if (strcmp (linearr[i], linearr[n - 1]) == 0)
+                        break;
+        }
+        glusterd_urltransform_free (linearr, i);
+
+        if (i < n - 1)
+                *slavekey = dict_get_by_index (vol->gsync_slaves, i);
+        else
+                i = -1;
+
+        return i;
+}
+
+
+static int
+glusterd_query_extutil_generic (char *resbuf, size_t blen, runner_t *runner, void *data,
+                                int (*fcbk)(char *resbuf, size_t blen, FILE *fp, void *data))
+{
+        int                 ret = 0;
+
+        runner_redir (runner, STDOUT_FILENO, RUN_PIPE);
+        if (runner_start (runner) != 0) {
+                gf_log ("", GF_LOG_ERROR, "spawning child failed");
+
+                return -1;
+        }
+
+        ret = fcbk (resbuf, blen, runner_chio (runner, STDOUT_FILENO), data);
+
+        ret |= runner_end (runner);
+        if (ret)
+                gf_log ("", GF_LOG_ERROR, "reading data from child failed");
+
+        return ret ? -1 : 0;
+}
+
+static int
+_fcbk_singleline(char *resbuf, size_t blen, FILE *fp, void *data)
+{
+        char *ptr = NULL;
+
+        errno = 0;
+        ptr = fgets (resbuf, blen, fp);
+        if (ptr) {
+                size_t len = strlen(resbuf);
+                if (len && resbuf[len-1] == '\n')
+                        resbuf[len-1] = '\0'; //strip off \n
+        }
+
+        return errno ? -1 : 0;
+}
+
+static int
+glusterd_query_extutil (char *resbuf, runner_t *runner)
+{
+        return glusterd_query_extutil_generic (resbuf, PATH_MAX, runner, NULL,
+                                               _fcbk_singleline);
+}
+
+static int
+_fcbk_conftodict (char *resbuf, size_t blen, FILE *fp, void *data)
+{
+        char   *ptr = NULL;
+        dict_t *dict = data;
+        char   *v = NULL;
+
+        for (;;) {
+                errno = 0;
+                ptr = fgets (resbuf, blen, fp);
+                if (!ptr)
+                        break;
+                v = resbuf + strlen(resbuf) - 1;
+                while (isspace (*v))
+                        /* strip trailing space */
+                        *v-- = '\0';
+                if (v == resbuf)
+                        /* skip empty line */
+                        continue;
+                v = strchr (resbuf, ':');
+                if (!v)
+                        return -1;
+                *v++ = '\0';
+                while (isspace (*v))
+                        v++;
+                v = gf_strdup (v);
+                if (!v)
+                        return -1;
+                if (dict_set_dynstr (dict, resbuf, v) != 0) {
+                        GF_FREE (v);
+                        return -1;
+                }
+        }
+
+        return errno ? -1 : 0;
+}
+
+static int
+glusterd_gsync_get_config (char *master, char *slave, char *conf_path, dict_t *dict)
+{
+        /* key + value, where value must be able to accommodate a path */
+        char resbuf[256 + PATH_MAX] = {0,};
+        runner_t             runner = {0,};
+
+        runinit (&runner);
+        runner_add_args  (&runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL);
+        runner_argprintf (&runner, "%s", conf_path);
+        runner_argprintf (&runner, ":%s", master);
+        runner_add_args  (&runner, slave, "--config-get-all", NULL);
+
+        return glusterd_query_extutil_generic (resbuf, sizeof (resbuf),
+                                               &runner, dict, _fcbk_conftodict);
+}
+
+static int
+glusterd_gsync_get_param_file (char *prmfile, const char *param, char *master,
+                               char *slave, char *conf_path)
+{
+        runner_t            runner = {0,};
+
+        runinit (&runner);
+        runner_add_args  (&runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL);
+        runner_argprintf (&runner, "%s", conf_path);
+        runner_argprintf (&runner, ":%s", master);
+        runner_add_args  (&runner, slave, "--config-get", NULL);
+        runner_argprintf (&runner, "%s-file", param);
+
+        return glusterd_query_extutil (prmfile, &runner);
+}
+
+static int
+gsyncd_getpidfile (char *master, char *slave, char *pidfile, char *conf_path)
+{
+        int                ret             = -1;
+        glusterd_conf_t    *priv  = NULL;
+        char               *confpath = NULL;
+        char                conf_buf[PATH_MAX] = "";
+        struct stat         stbuf = {0,};
+
+
+        GF_ASSERT (THIS);
+        GF_ASSERT (THIS->private);
+
+        priv = THIS->private;
+
+        GF_VALIDATE_OR_GOTO ("gsync", master, out);
+        GF_VALIDATE_OR_GOTO ("gsync", slave, out);
+
+        ret = lstat (conf_path, &stbuf);
+        if (!ret) {
+                gf_log ("", GF_LOG_DEBUG, "Using passed config template(%s).",
+                        conf_path);
+                confpath = conf_path;
+        } else {
+                ret = snprintf (conf_buf, sizeof(conf_buf) - 1,
+                                "%s/"GSYNC_CONF_TEMPLATE, priv->workdir);
+                conf_buf[ret] = '\0';
+                confpath = conf_buf;
+                gf_log ("", GF_LOG_DEBUG, "Using default config template(%s).",
+                        confpath);
+        }
+
+        ret = glusterd_gsync_get_param_file (pidfile, "pid", master,
+                                              slave, confpath);
+        if (ret == -1) {
+                ret = -2;
+                gf_log ("", GF_LOG_WARNING, "failed to create the pidfile string");
+                goto out;
+        }
+
+        ret = open (pidfile, O_RDWR);
+
+ out:
+        return ret;
+}
+
+static int
+gsync_status_byfd (int fd)
+{
+        GF_ASSERT (fd >= -1);
+
+        if (lockf (fd, F_TEST, 0) == -1 &&
+            (errno == EAGAIN || errno == EACCES))
+                /* gsyncd keeps the pidfile locked */
+                return 0;
+
+        return -1;
+}
+
+/* status: return 0 when gsync is running
+ * return -1 when not running
+ */
+int
+gsync_status (char *master, char *slave, char *conf_path, int *status)
+{
+        char pidfile[PATH_MAX] = {0,};
+        int  fd                = -1;
+
+        fd = gsyncd_getpidfile (master, slave, pidfile, conf_path);
+        if (fd == -2)
+                return -1;
+
+        *status = gsync_status_byfd (fd);
+
+        sys_close (fd);
+
+        return 0;
+}
+
+
+static int32_t
+glusterd_gsync_volinfo_dict_set (glusterd_volinfo_t *volinfo,
+                                 char *key, char *value)
+{
+        int32_t  ret            = -1;
+        char    *gsync_status   = NULL;
+
+        gsync_status = gf_strdup (value);
+        if (!gsync_status) {
+                gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
+                goto out;
+        }
+
+        ret = dict_set_dynstr (volinfo->dict, key, gsync_status);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to set dict");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return 0;
+}
+
+static int
+glusterd_verify_gsyncd_spawn (char *master, char *slave)
+{
+        int                 ret = 0;
+        runner_t            runner = {0,};
+
+        runinit (&runner);
+        runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd",
+                         "--verify", "spawning", NULL);
+        runner_argprintf (&runner, ":%s", master);
+        runner_add_args  (&runner, slave, NULL);
+        runner_redir (&runner, STDOUT_FILENO, RUN_PIPE);
+        ret = runner_start (&runner);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "spawning child failed");
+                ret = -1;
+                goto out;
+        }
+
+        if (runner_end (&runner) != 0)
+                ret = -1;
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+static int
+gsync_verify_config_options (dict_t *dict, char **op_errstr, char *volname)
+{
+        char                         **resopt           = NULL;
+        int                            i                = 0;
+        int                            ret              = -1;
+        char                          *subop            = NULL;
+        char                          *slave            = NULL;
+        char                          *op_name          = NULL;
+        char                          *op_value         = NULL;
+        char                          *t                = NULL;
+        char                           errmsg[PATH_MAX] = "";
+        gf_boolean_t                   banned           = _gf_true;
+        gf_boolean_t                   op_match         = _gf_true;
+        gf_boolean_t                   val_match        = _gf_true;
+        struct gsync_config_opt_vals_ *conf_vals        = NULL;
+
+        if (dict_get_str (dict, "subop", &subop) != 0) {
+                gf_log ("", GF_LOG_WARNING, "missing subop");
+                *op_errstr = gf_strdup ("Invalid config request");
+                return -1;
+        }
+
+        if (dict_get_str (dict, "slave", &slave) != 0) {
+                gf_log ("", GF_LOG_WARNING, GEOREP" CONFIG: no slave given");
+                *op_errstr = gf_strdup ("Slave required");
+                return -1;
+        }
+
+        if (strcmp (subop, "get-all") == 0)
+                return 0;
+
+        if (dict_get_str (dict, "op_name", &op_name) != 0) {
+                gf_log ("", GF_LOG_WARNING, "option name missing");
+                *op_errstr = gf_strdup ("Option name missing");
+                return -1;
+        }
+
+        if (runcmd (GSYNCD_PREFIX"/gsyncd", "--config-check", op_name, NULL)) {
+                ret = glusterd_verify_gsyncd_spawn (volname, slave);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to spawn gsyncd");
+                        return 0;
+                }
+
+                gf_log ("", GF_LOG_WARNING, "Invalid option %s", op_name);
+                *op_errstr = gf_strdup ("Invalid option");
+
+                return -1;
+        }
+
+        if (strcmp (subop, "get") == 0)
+                return 0;
+
+        t = strtail (subop, "set");
+        if (!t)
+                t = strtail (subop, "del");
+        if (!t || (t[0] && strcmp (t, "-glob") != 0)) {
+                gf_log ("", GF_LOG_WARNING, "unknown subop %s", subop);
+                *op_errstr = gf_strdup ("Invalid config request");
+                return -1;
+        }
+
+        if (strtail (subop, "set") &&
+            dict_get_str (dict, "op_value", &op_value) != 0) {
+                gf_log ("", GF_LOG_WARNING, "missing value for set");
+                *op_errstr = gf_strdup ("missing value");
+        }
+
+        /* match option name against reserved options, modulo -/_
+         * difference
+         */
+        for (resopt = gsync_reserved_opts; *resopt; resopt++) {
+                banned = _gf_true;
+                for (i = 0; (*resopt)[i] && op_name[i]; i++) {
+                        if ((*resopt)[i] == op_name[i] ||
+                            ((*resopt)[i] == '-' && op_name[i] == '_'))
+                                continue;
+                        banned = _gf_false;
+                }
+                if (banned) {
+                        gf_log ("", GF_LOG_WARNING, "Reserved option %s", op_name);
+                        *op_errstr = gf_strdup ("Reserved option");
+
+                        return -1;
+                        break;
+                }
+        }
+
+        /* Check options in gsync_confopt_vals for invalid values */
+        for (conf_vals = gsync_confopt_vals; conf_vals->op_name; conf_vals++) {
+                op_match = _gf_true;
+                for (i = 0; conf_vals->op_name[i] && op_name[i]; i++) {
+                        if (conf_vals->op_name[i] == op_name[i] ||
+                            (conf_vals->op_name[i] == '_' && op_name[i] == '-'))
+                                continue;
+                        op_match = _gf_false;
+                }
+
+                if (op_match) {
+                        val_match = _gf_false;
+                        for (i = 0; i < conf_vals->no_of_pos_vals; i++) {
+                                if(conf_vals->case_sensitive){
+                                        if (!strcmp (conf_vals->values[i], op_value))
+                                                val_match = _gf_true;
+                                } else {
+                                        if (!strcasecmp (conf_vals->values[i], op_value))
+                                                val_match = _gf_true;
+                                }
+                        }
+
+                        if (!val_match) {
+                                ret = snprintf (errmsg, sizeof(errmsg) - 1,
+                                                "Invalid values (%s) for"
+                                                " option %s", op_value,
+                                                op_name);
+                                errmsg[ret] = '\0';
+
+                                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                                *op_errstr = gf_strdup (errmsg);
+                                return -1;
+                        }
+                }
+        }
+
+        return 0;
+}
+
+static int
+glusterd_get_gsync_status_mst_slv (glusterd_volinfo_t *volinfo,
+                                   char *slave, char *conf_path,
+                                   dict_t *rsp_dict, char *node);
+
+static int
+_get_status_mst_slv (dict_t *this, char *key, data_t *value, void *data)
+{
+        glusterd_gsync_status_temp_t  *param = NULL;
+        char                          *slave = NULL;
+        char                          *slave_buf = NULL;
+        char                          *slave_ip  = NULL;
+        char                          *slave_vol = NULL;
+        char                          *errmsg    = NULL;
+        char                           conf_path[PATH_MAX] = "";
+        int                           ret = -1;
+        glusterd_conf_t              *priv = NULL;
+
+        param = (glusterd_gsync_status_temp_t *)data;
+
+        GF_ASSERT (param);
+        GF_ASSERT (param->volinfo);
+
+        if (THIS)
+                priv = THIS->private;
+        if (priv == NULL) {
+                gf_log ("", GF_LOG_ERROR, "priv of glusterd not present");
+                goto out;
+        }
+
+        slave = strchr(value->data, ':');
+        if (!slave)
+                return 0;
+        slave++;
+
+        ret = glusterd_get_slave_info (slave, &slave_ip, &slave_vol, &errmsg);
+        if (ret) {
+                if (errmsg)
+                        gf_log ("", GF_LOG_ERROR, "Unable to fetch "
+                                "slave details. Error: %s", errmsg);
+                else
+                        gf_log ("", GF_LOG_ERROR,
+                                "Unable to fetch slave details.");
+                ret = -1;
+                goto out;
+        }
+
+        ret = snprintf (conf_path, sizeof(conf_path) - 1,
+                        "%s/"GEOREP"/%s_%s_%s/gsyncd.conf",
+                        priv->workdir, param->volinfo->volname,
+                        slave_ip, slave_vol);
+        conf_path[ret] = '\0';
+
+        ret = glusterd_get_gsync_status_mst_slv(param->volinfo,
+                                                slave, conf_path,
+                                                param->rsp_dict,
+                                                param->node);
+out:
+
+        if (slave_buf)
+                GF_FREE(slave_buf);
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d.", ret);
+        return ret;
+}
+
+
+static int
+_get_max_gsync_slave_num (dict_t *this, char *key, data_t *value, void *data)
+{
+        int  tmp_slvnum = 0;
+        int *slvnum = (int *)data;
+
+        sscanf (key, "slave%d", &tmp_slvnum);
+        if (tmp_slvnum > *slvnum)
+                *slvnum = tmp_slvnum;
+
+        return 0;
+}
+
+static int
+glusterd_remove_slave_in_info (glusterd_volinfo_t *volinfo, char *slave,
+                               char **op_errstr)
+{
+        int   zero_slave_entries = _gf_true;
+        int   ret = 0;
+        char *slavekey = NULL;
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (slave);
+
+        do {
+                ret = glusterd_get_slave (volinfo, slave, &slavekey);
+                if (ret < 0 && zero_slave_entries) {
+                        ret++;
+                        goto out;
+                }
+                zero_slave_entries = _gf_false;
+                dict_del (volinfo->gsync_slaves, slavekey);
+        } while (ret >= 0);
+
+        ret = glusterd_store_volinfo (volinfo,
+                                      GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret) {
+                 *op_errstr = gf_strdup ("Failed to store the Volume"
+                                         "information");
+                goto out;
+        }
+ out:
+        gf_log ("", GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+
+}
+
+static int
+glusterd_gsync_get_uuid (char *slave, glusterd_volinfo_t *vol,
+                         uuid_t uuid)
+{
+        int   ret = 0;
+        char *slavekey = NULL;
+        char *slaveentry = NULL;
+        char *t = NULL;
+
+        GF_ASSERT (vol);
+        GF_ASSERT (slave);
+
+        ret = glusterd_get_slave (vol, slave, &slavekey);
+        if (ret < 0) {
+                /* XXX colliding cases of failure and non-extant
+                 * slave... now just doing this as callers of this
+                 * function can make sense only of -1 and 0 as retvals;
+                 * getting at the proper semanticals will involve
+                 * fixing callers as well.
+                 */
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_str (vol->gsync_slaves, slavekey, &slaveentry);
+        GF_ASSERT (ret == 0);
+
+        t = strchr (slaveentry, ':');
+        GF_ASSERT (t);
+        *t = '\0';
+        ret = uuid_parse (slaveentry, uuid);
+        *t = ':';
+
+ out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_check_gsync_running_local (char *master, char *slave,
+                                    char *conf_path,
+                                    gf_boolean_t *is_run)
+{
+        int                 ret    = -1;
+        int                 ret_status = 0;
+
+        GF_ASSERT (master);
+        GF_ASSERT (slave);
+        GF_ASSERT (is_run);
+
+        *is_run = _gf_false;
+        ret = gsync_status (master, slave, conf_path, &ret_status);
+        if (ret == 0 && ret_status == 0) {
+                *is_run = _gf_true;
+        } else if (ret == -1) {
+                gf_log ("", GF_LOG_WARNING, GEOREP" validation "
+                        " failed");
+                goto out;
+        }
+        ret = 0;
+ out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+
+}
+
+static int
+glusterd_store_slave_in_info (glusterd_volinfo_t *volinfo, char *slave,
+                              char *host_uuid, char **op_errstr,
+                              gf_boolean_t is_force)
+{
+        int    ret = 0;
+        int    maxslv = 0;
+        char **linearr = NULL;
+        char  *value = NULL;
+        char  *slavekey = NULL;
+        char  *slaveentry = NULL;
+        char   key[512] = {0, };
+        char  *t = NULL;
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (slave);
+        GF_ASSERT (host_uuid);
+
+        ret = glusterd_get_slave (volinfo, slave, &slavekey);
+        switch (ret) {
+        case -2:
+                ret = -1;
+                goto out;
+        case -1:
+                                break;
+        default:
+                if (!is_force)
+                        GF_ASSERT (ret > 0);
+                ret = dict_get_str (volinfo->gsync_slaves, slavekey, &slaveentry);
+                GF_ASSERT (ret == 0);
+
+                /* same-name + same-uuid slave entries should have been filtered
+                 * out in glusterd_op_verify_gsync_start_options(), so we can
+                 * assert an uuid mismatch
+                 */
+                t = strtail (slaveentry, host_uuid);
+                if (!is_force)
+                        GF_ASSERT (!t || *t != ':');
+
+                if (is_force) {
+                        gf_log ("", GF_LOG_DEBUG, GEOREP" has already been "
+                                "invoked for the %s (master) and %s (slave)."
+                                " Allowing without saving info again due to"
+                                " force command.", volinfo->volname, slave);
+                        ret = 0;
+                        goto out;
+                }
+
+                gf_log ("", GF_LOG_ERROR, GEOREP" has already been invoked for "
+                                          "the %s (master) and %s (slave) "
+                                          "from a different machine",
+                                           volinfo->volname, slave);
+                *op_errstr = gf_strdup (GEOREP" already running in "
+                                        "another machine");
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_urltransform_single (slave, "normalize", &linearr);
+        if (ret == -1)
+                goto out;
+
+        ret = gf_asprintf (&value,  "%s:%s", host_uuid, linearr[0]);
+        glusterd_urltransform_free (linearr, 1);
+        if (ret  == -1)
+                goto out;
+
+        dict_foreach (volinfo->gsync_slaves, _get_max_gsync_slave_num, &maxslv);
+        snprintf (key, 512, "slave%d", maxslv + 1);
+        ret = dict_set_dynstr (volinfo->gsync_slaves, key, value);
+        if (ret)
+                goto out;
+
+        ret = glusterd_store_volinfo (volinfo,
+                                      GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret) {
+                *op_errstr = gf_strdup ("Failed to store the Volume "
+                                        "information");
+                goto out;
+        }
+        ret = 0;
+ out:
+        return ret;
+}
+
+static int
+glusterd_op_verify_gsync_start_options (glusterd_volinfo_t *volinfo,
+                                        char *slave, char *conf_path,
+                                        char *statefile, char **op_errstr,
+                                        gf_boolean_t is_force)
+{
+        int                     ret = -1;
+        gf_boolean_t            is_running = _gf_false;
+        char                    msg[2048] = {0};
+        uuid_t                  uuid = {0};
+        glusterd_conf_t        *priv = NULL;
+        xlator_t               *this = NULL;
+        struct stat             stbuf = {0,};
+
+        this = THIS;
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (slave);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (conf_path);
+        GF_ASSERT (this && this->private);
+
+        priv  = this->private;
+
+        if (GLUSTERD_STATUS_STARTED != volinfo->status) {
+                snprintf (msg, sizeof (msg), "Volume %s needs to be started "
+                          "before "GEOREP" start", volinfo->volname);
+                goto out;
+        }
+
+        ret = lstat (statefile, &stbuf);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Session between %s and %s has"
+                          " not been created. Please create session and retry.",
+                          volinfo->volname, slave);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        /* Check if the gsync slave info is stored. If not
+         * session has not been created */
+        ret = glusterd_gsync_get_uuid (slave, volinfo, uuid);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Session between %s and %s has"
+                          " not been created. Please create session and retry.",
+                          volinfo->volname, slave);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        if (is_force) {
+                ret = 0;
+                goto out;
+        }
+
+        /*Check if the gsync is already started in cmd. inited host
+         * If so initiate add it into the glusterd's priv*/
+        ret = glusterd_check_gsync_running_local (volinfo->volname,
+                                                  slave, conf_path,
+                                                  &is_running);
+        if (ret) {
+                snprintf (msg, sizeof (msg), GEOREP" start option "
+                          "validation failed ");
+                goto out;
+        }
+        if (_gf_true == is_running) {
+                snprintf (msg, sizeof (msg), GEOREP " session between"
+                          " %s & %s already started", volinfo->volname,
+                          slave);
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_verify_gsyncd_spawn (volinfo->volname, slave);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Unable to spawn gsyncd");
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+        }
+out:
+        if (ret && (msg[0] != '\0')) {
+                *op_errstr = gf_strdup (msg);
+        }
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_check_gsync_running (glusterd_volinfo_t *volinfo, gf_boolean_t *flag)
+{
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (flag);
+
+        if (volinfo->gsync_slaves->count)
+                *flag = _gf_true;
+        else
+                *flag = _gf_false;
+
+        return 0;
+}
+
+static int
+glusterd_op_verify_gsync_running (glusterd_volinfo_t *volinfo,
+                                  char *slave, char *conf_path,
+                                  char **op_errstr)
+{
+        int        pfd               = -1;
+        int        ret               = -1;
+        char       msg[2048]         = {0};
+        char       pidfile[PATH_MAX] = {0,};
+
+        GF_ASSERT (THIS && THIS->private);
+        GF_ASSERT (volinfo);
+        GF_ASSERT (slave);
+        GF_ASSERT (op_errstr);
+
+        if (GLUSTERD_STATUS_STARTED != volinfo->status) {
+                snprintf (msg, sizeof (msg), "Volume %s needs to be started "
+                          "before "GEOREP" start", volinfo->volname);
+
+                goto out;
+        }
+
+        pfd = gsyncd_getpidfile (volinfo->volname, slave, pidfile, conf_path);
+        if (pfd == -2) {
+                gf_log ("", GF_LOG_ERROR, GEOREP" stop validation "
+                        "failed for %s & %s", volinfo->volname, slave);
+                ret = -1;
+                goto out;
+        }
+        if (gsync_status_byfd (pfd) == -1) {
+                snprintf (msg, sizeof (msg), GEOREP" session b/w %s & %s is not"
+                          " running on this node.", volinfo->volname, slave);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                ret = -1;
+                /* monitor gsyncd already dead */
+                goto out;
+        }
+
+        if (pfd < 0)
+                goto out;
+
+        ret = 0;
+out:
+        if (ret && (msg[0] != '\0')) {
+                *op_errstr = gf_strdup (msg);
+        }
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_verify_gsync_status_opts (dict_t *dict, char **op_errstr)
+{
+        char               *slave  = NULL;
+        char               *volname = NULL;
+        char               errmsg[PATH_MAX] = {0, };
+        gf_boolean_t       exists = _gf_false;
+        glusterd_volinfo_t *volinfo = NULL;
+        int                ret = 0;
+        char               *conf_path = NULL;
+        char               *slave_ip = NULL;
+        char               *slave_vol = NULL;
+        glusterd_conf_t    *priv = NULL;
+
+        if (THIS)
+                priv = THIS->private;
+        if (priv == NULL) {
+                gf_log ("", GF_LOG_ERROR, "priv of glusterd not present");
+                *op_errstr = gf_strdup ("glusterd defunct");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "master", &volname);
+        if (ret < 0) {
+                ret = 0;
+                goto out;
+        }
+
+        exists = glusterd_check_volume_exists (volname);
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if ((ret) || (!exists)) {
+                gf_log ("", GF_LOG_WARNING, "volume name does not exist");
+                snprintf (errmsg, sizeof(errmsg), "Volume name %s does not"
+                          " exist", volname);
+                *op_errstr = gf_strdup (errmsg);
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "slave", &slave);
+        if (ret < 0) {
+                ret = 0;
+                goto out;
+        }
+
+        ret = glusterd_get_slave_details_confpath (volinfo, dict, &slave_ip,
+                                                   &slave_vol, &conf_path,
+                                                   op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to fetch slave  or confpath details.");
+                ret = -1;
+                goto out;
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+
+int
+glusterd_op_gsync_args_get (dict_t *dict, char **op_errstr,
+                            char **master, char **slave, char **host_uuid)
+{
+
+        int             ret = -1;
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+
+        if (master) {
+                ret = dict_get_str (dict, "master", master);
+                if (ret < 0) {
+                        gf_log ("", GF_LOG_WARNING, "master not found");
+                        *op_errstr = gf_strdup ("master not found");
+                        goto out;
+                }
+        }
+
+        if (slave) {
+                ret = dict_get_str (dict, "slave", slave);
+                if (ret < 0) {
+                        gf_log ("", GF_LOG_WARNING, "slave not found");
+                        *op_errstr = gf_strdup ("slave not found");
+                        goto out;
+                }
+        }
+
+        if (host_uuid) {
+                ret = dict_get_str (dict, "host-uuid", host_uuid);
+                if (ret < 0) {
+                        gf_log ("", GF_LOG_WARNING, "host_uuid not found");
+                        *op_errstr = gf_strdup ("host_uuid not found");
+                        goto out;
+                }
+        }
+
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_stage_sys_exec (dict_t *dict, char **op_errstr)
+{
+        char             errmsg[PATH_MAX]          = "";
+        char            *command                   = NULL;
+        char             command_path[PATH_MAX]    = "";
+        struct stat      st                        = {0,};
+        int              ret                       = -1;
+        glusterd_conf_t *conf                      = NULL;
+        xlator_t        *this                      = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        if (conf->op_version < 2) {
+                gf_log ("", GF_LOG_ERROR, "Op Version not supported.");
+                snprintf (errmsg, sizeof(errmsg), "One or more nodes do not"
+                          " support the required op version.");
+                *op_errstr = gf_strdup (errmsg);
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "command", &command);
+        if (ret) {
+                strcpy (errmsg, "internal error");
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to get command from dict");
+                goto out;
+        }
+
+        /* enforce local occurrence of the command */
+        if (strchr (command, '/')) {
+                strcpy (errmsg, "invalid command name");
+                ret = -1;
+                goto out;
+        }
+
+        sprintf (command_path, GSYNCD_PREFIX"/peer_%s", command);
+        /* check if it's executable */
+        ret = access (command_path, X_OK);
+        if (!ret)
+                /* check if it's a regular file */
+                ret = stat (command_path, &st);
+        if (!ret && !S_ISREG (st.st_mode))
+                ret = -1;
+
+out:
+        if (ret) {
+                if (errmsg[0] == '\0')
+                        snprintf (errmsg, sizeof (errmsg), "%s not found.",
+                                  command);
+                *op_errstr = gf_strdup (errmsg);
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+        }
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_stage_copy_file (dict_t *dict, char **op_errstr)
+{
+        char             abs_filename[PATH_MAX] = "";
+        char             errmsg[PATH_MAX]       = "";
+        char            *filename               = NULL;
+        char            *host_uuid              = NULL;
+        char             uuid_str [64]          = {0};
+        int              ret                    = -1;
+        glusterd_conf_t *priv                   = NULL;
+        struct stat      stbuf                  = {0,};
+
+        if (THIS)
+                priv = THIS->private;
+        if (priv == NULL) {
+                gf_log ("", GF_LOG_ERROR, "priv of glusterd not present");
+                *op_errstr = gf_strdup ("glusterd defunct");
+                goto out;
+        }
+
+        if (priv->op_version < 2) {
+                gf_log ("", GF_LOG_ERROR, "Op Version not supported.");
+                snprintf (errmsg, sizeof(errmsg), "One or more nodes do not"
+                          " support the required op version.");
+                *op_errstr = gf_strdup (errmsg);
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "host-uuid", &host_uuid);
+        if (ret < 0) {
+                gf_log ("", GF_LOG_ERROR, "Unable to fetch"
+                        " host-uuid from dict.");
+                goto out;
+        }
+
+        uuid_utoa_r (MY_UUID, uuid_str);
+        if (!strcmp (uuid_str, host_uuid)) {
+                ret = dict_get_str (dict, "source", &filename);
+                if (ret < 0) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to fetch"
+                                " filename from dict.");
+                        *op_errstr = gf_strdup ("command unsuccessful");
+                        goto out;
+                }
+                snprintf (abs_filename, sizeof(abs_filename),
+                          "%s/%s", priv->workdir, filename);
+
+                ret = lstat (abs_filename, &stbuf);
+                if (ret) {
+                        snprintf (errmsg, sizeof (errmsg), "Source file"
+                                  " does not exist in %s", priv->workdir);
+                        *op_errstr = gf_strdup (errmsg);
+                        goto out;
+                }
+
+                if (!S_ISREG(stbuf.st_mode)) {
+                        snprintf (errmsg, sizeof (errmsg), "Source file"
+                                 " is not a regular file.");
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_get_statefile_name (glusterd_volinfo_t *volinfo, char *slave,
+                             char *conf_path, char **statefile)
+{
+        glusterd_conf_t *priv = NULL;
+        int              ret = -1;
+        char            *master    = NULL;
+        char            *buf       = NULL;
+        dict_t          *confd = NULL;
+        char            *confpath = NULL;
+        char             conf_buf[PATH_MAX] = "";
+        struct stat      stbuf = {0,};
+
+        GF_ASSERT (THIS);
+        GF_ASSERT (THIS->private);
+        GF_ASSERT (volinfo);
+
+        master = volinfo->volname;
+
+        confd = dict_new ();
+        if (!confd) {
+                gf_log ("", GF_LOG_ERROR, "Unable to create new dict");
+                goto out;
+        }
+
+        priv = THIS->private;
+
+        ret = lstat (conf_path, &stbuf);
+        if (!ret) {
+                gf_log ("", GF_LOG_INFO, "Using passed config template(%s).",
+                        conf_path);
+                confpath = conf_path;
+        } else {
+                ret = snprintf (conf_buf, sizeof(conf_buf) - 1,
+                                "%s/"GSYNC_CONF_TEMPLATE, priv->workdir);
+                conf_buf[ret] = '\0';
+                confpath = conf_buf;
+                gf_log ("", GF_LOG_INFO, "Using default config template(%s).",
+                        confpath);
+        }
+
+        ret = glusterd_gsync_get_config (master, slave, confpath,
+                                         confd);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get configuration data"
+                        "for %s(master), %s(slave)", master, slave);
+                goto out;
+
+        }
+
+        ret = dict_get_param (confd, "state_file", &buf);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get state_file's name.");
+                goto out;
+        }
+
+        *statefile = gf_strdup(buf);
+        if (!*statefile) {
+                gf_log ("", GF_LOG_ERROR, "Unable to gf_strdup.");
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+ out:
+        if (confd)
+                dict_destroy (confd);
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d ", ret);
+        return ret;
+}
+
+static int
+glusterd_create_status_file (char *master, char *slave, char *slave_ip,
+                             char *slave_vol, char *status)
+{
+        int                ret    = -1;
+        runner_t           runner = {0,};
+        glusterd_conf_t   *priv   = NULL;
+
+        if (THIS)
+                priv = THIS->private;
+        if (priv == NULL) {
+                gf_log ("", GF_LOG_ERROR, "priv of glusterd not present");
+                goto out;
+        }
+
+        if (!status) {
+                gf_log ("", GF_LOG_ERROR, "Status Empty");
+                goto out;
+        }
+        gf_log ("", GF_LOG_DEBUG, "slave = %s", slave);
+
+        runinit (&runner);
+        runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "--create",
+                         status, "-c", NULL);
+        runner_argprintf (&runner, "%s/"GEOREP"/%s_%s_%s/gsyncd.conf",
+                          priv->workdir, master, slave_ip, slave_vol);
+        runner_argprintf (&runner, ":%s", master);
+        runner_add_args  (&runner, slave, NULL);
+        synclock_unlock (&priv->big_lock);
+        ret = runner_run (&runner);
+        synclock_lock (&priv->big_lock);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Creating status file failed.");
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_verify_slave (char *volname, char *slave_ip, char *slave,
+                       char **op_errstr, gf_boolean_t *is_force_blocker)
+{
+        int32_t          ret                     = -1;
+        runner_t         runner                  = {0,};
+        char             log_file_path[PATH_MAX] = "";
+        char             buf[PATH_MAX]           = "";
+        char            *tmp                     = NULL;
+        char            *save_ptr                = NULL;
+        glusterd_conf_t *priv                    = NULL;
+
+        GF_ASSERT (volname);
+        GF_ASSERT (slave_ip);
+        GF_ASSERT (slave);
+
+        if (THIS)
+                priv = THIS->private;
+        if (priv == NULL) {
+                gf_log ("", GF_LOG_ERROR, "priv of glusterd not present");
+                goto out;
+        }
+
+        snprintf (log_file_path, sizeof(log_file_path),
+                  DEFAULT_LOG_FILE_DIRECTORY"/create_verify_log");
+
+        runinit (&runner);
+        runner_add_args  (&runner, GSYNCD_PREFIX"/gverify.sh", NULL);
+        runner_argprintf (&runner, "%s", volname);
+        runner_argprintf (&runner, "%s", slave_ip);
+        runner_argprintf (&runner, "%s", slave);
+        runner_argprintf (&runner, "%s", log_file_path);
+        runner_redir (&runner, STDOUT_FILENO, RUN_PIPE);
+        synclock_unlock (&priv->big_lock);
+        ret = runner_run (&runner);
+        synclock_lock (&priv->big_lock);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Not a valid slave");
+                ret = glusterd_gsync_read_frm_status (log_file_path,
+                                                      buf, sizeof(buf));
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to read from %s",
+                                log_file_path);
+                        goto out;
+                }
+
+                /* Tokenize the error message from gverify.sh to figure out
+                 * if the error is a force blocker or not. */
+                tmp = strtok_r (buf, "|", &save_ptr);
+                if (!strcmp (tmp, "FORCE_BLOCKER"))
+                        *is_force_blocker = 1;
+                else {
+                        /* No FORCE_BLOCKER flag present so all that is
+                         * present is the error message. */
+                        *is_force_blocker = 0;
+                        if (tmp)
+                                *op_errstr = gf_strdup (tmp);
+                        ret = -1;
+                        goto out;
+                }
+
+                /* Copy rest of the error message to op_errstr */
+                tmp = strtok_r (NULL, "|", &save_ptr);
+                if (tmp)
+                        *op_errstr = gf_strdup (tmp);
+                ret = -1;
+                goto out;
+        }
+        ret = 0;
+out:
+        unlink (log_file_path);
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_mountbroker_check (char **slave_ip, char **op_errstr)
+{
+        int   ret             = -1;
+        char *tmp             = NULL;
+        char *save_ptr        = NULL;
+        char *username        = NULL;
+        char *host            = NULL;
+        char errmsg[PATH_MAX] = "";
+
+        GF_ASSERT (slave_ip);
+        GF_ASSERT (*slave_ip);
+
+        /* Checking if hostname has user specified */
+        host = strstr (*slave_ip, "@");
+        if (!host) {
+                gf_log ("", GF_LOG_DEBUG, "No username provided.");
+                ret = 0;
+                goto out;
+        } else {
+                /* Moving the host past the '@' and checking if the
+                 * actual hostname also has '@' */
+                host++;
+                if (strstr (host, "@")) {
+                        gf_log ("", GF_LOG_DEBUG, "host = %s", host);
+                        ret = snprintf (errmsg, sizeof(errmsg) - 1,
+                                        "Invalid Hostname (%s).", host);
+                        errmsg[ret] = '\0';
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        ret = -1;
+                        if (op_errstr)
+                                *op_errstr = gf_strdup (errmsg);
+                        goto out;
+                }
+
+                /* Fetching the username and hostname
+                 * and checking if the username is non-root */
+                username = strtok_r (*slave_ip, "@", &save_ptr);
+                tmp = strtok_r (NULL, "@", &save_ptr);
+                if (strcmp (username, "root")) {
+                        ret = snprintf (errmsg, sizeof(errmsg) - 1,
+                                        "Non-root username (%s@%s) not allowed.",
+                                        username, tmp);
+                        errmsg[ret] = '\0';
+                        if (op_errstr)
+                                *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR,
+                                "Non-Root username not allowed.");
+                        ret = -1;
+                        goto out;
+                }
+
+                *slave_ip = gf_strdup (tmp);
+                if (!*slave_ip) {
+                        gf_log ("", GF_LOG_ERROR, "Out of memory");
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr)
+{
+        char               *down_peerstr              = NULL;
+        char               *slave                     = NULL;
+        char               *volname                   = NULL;
+        char               *host_uuid                 = NULL;
+        char               *statefile                 = NULL;
+        char               *slave_ip                  = NULL;
+        char               *slave_vol                 = NULL;
+        char               *conf_path                 = NULL;
+        char                errmsg[PATH_MAX]          = "";
+        char                common_pem_file[PATH_MAX] = "";
+        char                hook_script[PATH_MAX]     = "";
+        char                uuid_str [64]             = "";
+        int                 ret                       = -1;
+        int                 is_pem_push               = -1;
+        gf_boolean_t        is_force                  = -1;
+        gf_boolean_t        is_force_blocker          = -1;
+        gf_boolean_t        exists                    = _gf_false;
+        glusterd_conf_t    *conf                      = NULL;
+        glusterd_volinfo_t *volinfo                   = NULL;
+        struct stat         stbuf                     = {0,};
+        xlator_t           *this                      = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        ret = glusterd_op_gsync_args_get (dict, op_errstr, &volname,
+                                          &slave, &host_uuid);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to fetch arguments");
+                gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+                return -1;
+        }
+
+        if (conf->op_version < 2) {
+                gf_log ("", GF_LOG_ERROR, "Op Version not supported.");
+                snprintf (errmsg, sizeof(errmsg), "One or more nodes do not"
+                          " support the required op version.");
+                *op_errstr = gf_strdup (errmsg);
+                ret = -1;
+                goto out;
+        }
+
+        exists = glusterd_check_volume_exists (volname);
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if ((ret) || (!exists)) {
+                gf_log ("", GF_LOG_WARNING, "volume name does not exist");
+                snprintf (errmsg, sizeof(errmsg), "Volume name %s does not"
+                          " exist", volname);
+                *op_errstr = gf_strdup (errmsg);
+                gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+                return -1;
+        }
+
+        ret = glusterd_get_slave_details_confpath (volinfo, dict, &slave_ip,
+                                                   &slave_vol, &conf_path,
+                                                   op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to fetch slave or confpath details.");
+                ret = -1;
+                goto out;
+        }
+
+        is_force = dict_get_str_boolean (dict, "force", _gf_false);
+
+        uuid_utoa_r (MY_UUID, uuid_str);
+        if (!strcmp (uuid_str, host_uuid)) {
+                ret = glusterd_are_vol_all_peers_up (volinfo,
+                                                     &conf->peers,
+                                                     &down_peerstr);
+                if ((ret == _gf_false) && !is_force) {
+                        snprintf (errmsg, sizeof (errmsg), "Peer %s,"
+                                  " which is a part of %s volume, is"
+                                  " down. Please bring up the peer and"
+                                  " retry.", down_peerstr,
+                                  volinfo->volname);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        *op_errstr = gf_strdup (errmsg);
+                        GF_FREE (down_peerstr);
+                        down_peerstr = NULL;
+                        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+                        return -1;
+                } else if (ret == _gf_false) {
+                        gf_log ("", GF_LOG_INFO, "Peer %s,"
+                                " which is a part of %s volume, is"
+                                " down. Force creating geo-rep session."
+                                " On bringing up the peer, re-run"
+                                " \"gluster system:: execute"
+                                " gsec_create\" and \"gluster volume"
+                                " geo-replication %s %s create push-pem"
+                                " force\"", down_peerstr, volinfo->volname,
+                                volinfo->volname, slave);
+                }
+
+                /* Checking if slave host is pingable, has proper passwordless
+                 * ssh login setup, slave volume is created, slave vol is empty,
+                 * and if it has enough memory and bypass in case of force if
+                 * the error is not a force blocker */
+                ret = glusterd_verify_slave (volname, slave_ip, slave_vol,
+                                             op_errstr, &is_force_blocker);
+                if (ret) {
+                        if (is_force && !is_force_blocker) {
+                                gf_log ("", GF_LOG_INFO, "%s is not a valid slave"
+                                        " volume. Error: %s. Force creating geo-rep"
+                                        " session.", slave, *op_errstr);
+                        } else {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "%s is not a valid slave volume. Error: %s",
+                                        slave, *op_errstr);
+                                ret = -1;
+                                goto out;
+                        }
+                }
+
+                ret = dict_get_int32 (dict, "push_pem", &is_pem_push);
+                if (!ret && is_pem_push) {
+                        ret = snprintf (common_pem_file,
+                                        sizeof(common_pem_file) - 1,
+                                        "%s"GLUSTERD_COMMON_PEM_PUB_FILE,
+                                        conf->workdir);
+                        common_pem_file[ret] = '\0';
+
+                        ret = snprintf (hook_script, sizeof(hook_script) - 1,
+                                        "%s"GLUSTERD_CREATE_HOOK_SCRIPT,
+                                        conf->workdir);
+                        hook_script[ret] = '\0';
+
+                        ret = lstat (common_pem_file, &stbuf);
+                        if (ret) {
+                                snprintf (errmsg, sizeof (errmsg), "%s"
+                                          " required for push-pem is"
+                                          " not present. Please run"
+                                          " \"gluster system:: execute"
+                                          " gsec_create\"", common_pem_file);
+                                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                                *op_errstr = gf_strdup (errmsg);
+                                ret = -1;
+                                goto out;
+                        }
+
+                        ret = lstat (hook_script, &stbuf);
+                        if (ret) {
+                                snprintf (errmsg, sizeof (errmsg),
+                                          "The hook-script (%s) required "
+                                          "for push-pem is not present. "
+                                          "Please install the hook-script "
+                                          "and retry", hook_script);
+                                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                                *op_errstr = gf_strdup (errmsg);
+                                ret = -1;
+                                goto out;
+                        }
+
+                        if (!S_ISREG(stbuf.st_mode)) {
+                                snprintf (errmsg, sizeof (errmsg), "%s"
+                                          " required for push-pem is"
+                                          " not a regular file. Please run"
+                                          " \"gluster system:: execute"
+                                          " gsec_create\"", common_pem_file);
+                                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                                ret = -1;
+                                goto out;
+                        }
+                }
+        }
+
+        ret = glusterd_get_statefile_name (volinfo, slave, conf_path, &statefile);
+        if (ret) {
+                if (!strstr(slave, "::"))
+                        snprintf (errmsg, sizeof (errmsg),
+                                  "%s is not a valid slave url.", slave);
+                else
+                        snprintf (errmsg, sizeof (errmsg), "Please check gsync "
+                                  "config file. Unable to get statefile's name");
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_str (dict, "statefile", statefile);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to store statefile path");
+                goto out;
+        }
+
+        ret = lstat (statefile, &stbuf);
+        if (!ret && !is_force) {
+                snprintf (errmsg, sizeof (errmsg), "Session between %s"
+                          " and %s is already created.",
+                          volinfo->volname, slave);
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                ret = -1;
+                goto out;
+        } else if (!ret)
+                gf_log ("", GF_LOG_INFO, "Session between %s"
+                        " and %s is already created. Force"
+                        " creating again.", volinfo->volname, slave);
+
+        ret = glusterd_verify_gsyncd_spawn (volinfo->volname, slave);
+        if (ret) {
+                snprintf (errmsg, sizeof (errmsg), "Unable to spawn gsyncd.");
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                goto out;
+        }
+
+        ret = 0;
+out:
+
+        if (ret && errmsg[0] != '\0')
+                *op_errstr = gf_strdup (errmsg);
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr)
+{
+        int                     ret     = 0;
+        int                     type    = 0;
+        char                    *volname = NULL;
+        char                    *slave   = NULL;
+        char                    *slave_ip = NULL;
+        char                    *slave_vol = NULL;
+        char                    *down_peerstr = NULL;
+        char                    *statefile    = NULL;
+        char                    *path_list    = NULL;
+        char                    *conf_path    = NULL;
+        gf_boolean_t            exists   = _gf_false;
+        glusterd_volinfo_t      *volinfo = NULL;
+        char                    errmsg[PATH_MAX] = {0,};
+        dict_t                  *ctx = NULL;
+        gf_boolean_t            is_force = 0;
+        gf_boolean_t            is_force_blocker = -1;
+        gf_boolean_t            is_running = _gf_false;
+        uuid_t                  uuid = {0};
+        char                    uuid_str [64] = {0};
+        char                    *host_uuid = NULL;
+        xlator_t                *this = NULL;
+        glusterd_conf_t         *conf = NULL;
+        struct stat             stbuf = {0,};
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        ret = dict_get_int32 (dict, "type", &type);
+        if (ret < 0) {
+                gf_log ("", GF_LOG_WARNING, "command type not found");
+                *op_errstr = gf_strdup ("command unsuccessful");
+                goto out;
+        }
+
+        if (type == GF_GSYNC_OPTION_TYPE_STATUS) {
+                ret = glusterd_verify_gsync_status_opts (dict, op_errstr);
+                goto out;
+        }
+
+        ret = glusterd_op_gsync_args_get (dict, op_errstr,
+                                          &volname, &slave, &host_uuid);
+        if (ret)
+                goto out;
+
+        uuid_utoa_r (MY_UUID, uuid_str);
+
+        if (conf->op_version < 2) {
+                gf_log ("", GF_LOG_ERROR, "Op Version not supported.");
+                snprintf (errmsg, sizeof(errmsg), "One or more nodes do not"
+                          " support the required op version.");
+               *op_errstr = gf_strdup (errmsg);
+                ret = -1;
+                goto out;
+        }
+
+        exists = glusterd_check_volume_exists (volname);
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if ((ret) || (!exists)) {
+                gf_log ("", GF_LOG_WARNING, "volume name does not exist");
+                snprintf (errmsg, sizeof(errmsg), "Volume name %s does not"
+                          " exist", volname);
+                *op_errstr = gf_strdup (errmsg);
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_get_slave_details_confpath (volinfo, dict, &slave_ip,
+                                                   &slave_vol, &conf_path,
+                                                   op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to fetch slave or confpath details.");
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_get_statefile_name (volinfo, slave, conf_path, &statefile);
+        if (ret) {
+                /* Checking if slave host is pingable, has proper passwordless
+                 * ssh login setup */
+                ret = glusterd_verify_slave (volname, slave_ip, slave_vol,
+                                             op_errstr, &is_force_blocker);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "%s is not a valid slave volume. Error: %s",
+                                slave, *op_errstr);
+                        goto out;
+                }
+
+                if (!strstr(slave, "::"))
+                        snprintf (errmsg, sizeof (errmsg),
+                                  "%s is not a valid slave url.", slave);
+                else
+                        snprintf (errmsg, sizeof (errmsg),
+                                  "Unable to get statefile's name");
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_str (dict, "statefile", statefile);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to store statefile path");
+                goto out;
+        }
+
+        is_force = dict_get_str_boolean (dict, "force", _gf_false);
+
+        /* Allowing stop force to bypass the statefile check
+         * as this command acts as a fail safe method to stop geo-rep
+         * session. */
+        if ((type == GF_GSYNC_OPTION_TYPE_CONFIG) ||
+            ((type == GF_GSYNC_OPTION_TYPE_STOP) && !is_force) ||
+            (type == GF_GSYNC_OPTION_TYPE_DELETE)) {
+                ret = lstat (statefile, &stbuf);
+                if (ret) {
+                        snprintf (errmsg, sizeof(errmsg), "Geo-replication"
+                                  " session between %s and %s does not exist.",
+                                  volinfo->volname, slave);
+                        gf_log ("", GF_LOG_ERROR, "%s. statefile = %s",
+                                errmsg, statefile);
+                        *op_errstr = gf_strdup (errmsg);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        /* Check if all peers that are a part of the volume are up or not */
+        if ((type == GF_GSYNC_OPTION_TYPE_DELETE) ||
+            ((type == GF_GSYNC_OPTION_TYPE_STOP) && !is_force)) {
+                if (!strcmp (uuid_str, host_uuid)) {
+                        ret = glusterd_are_vol_all_peers_up (volinfo,
+                                                             &conf->peers,
+                                                             &down_peerstr);
+                        if (ret == _gf_false) {
+                                snprintf (errmsg, sizeof (errmsg), "Peer %s,"
+                                          " which is a part of %s volume, is"
+                                          " down. Please bring up the peer and"
+                                          " retry.", down_peerstr,
+                                          volinfo->volname);
+                                *op_errstr = gf_strdup (errmsg);
+                                ret = -1;
+                                GF_FREE (down_peerstr);
+                                down_peerstr = NULL;
+                                goto out;
+                        }
+                }
+        }
+
+        switch (type) {
+        case GF_GSYNC_OPTION_TYPE_START:
+                /* don't attempt to start gsync if replace-brick is
+                 * in progress */
+                if (glusterd_is_rb_ongoing (volinfo)) {
+                        snprintf (errmsg, sizeof(errmsg), "replace-brick is in"
+                                   " progress, not starting geo-replication");
+                        *op_errstr = gf_strdup (errmsg);
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = glusterd_op_verify_gsync_start_options (volinfo, slave,
+                                                              conf_path, statefile,
+                                                              op_errstr, is_force);
+                if (ret)
+                        goto out;
+                ctx = glusterd_op_get_ctx();
+                if (ctx) {
+                        /* gsyncd does a fuse mount to start
+                         * the geo-rep session */
+                        if (!glusterd_is_fuse_available ()) {
+                                gf_log ("glusterd", GF_LOG_ERROR, "Unable to "
+                                        "open /dev/fuse (%s), geo-replication "
+                                        "start failed", strerror (errno));
+                                snprintf (errmsg, sizeof(errmsg),
+                                          "fuse unvailable");
+                                *op_errstr = gf_strdup (errmsg);
+                                ret = -1;
+                                goto out;
+                        }
+                }
+                break;
+
+        case GF_GSYNC_OPTION_TYPE_STOP:
+                if (!is_force) {
+                        ret = glusterd_op_verify_gsync_running (volinfo, slave,
+                                                                conf_path,
+                                                                op_errstr);
+                        if (ret) {
+                                ret = glusterd_get_local_brickpaths (volinfo,
+                                                                     &path_list);
+                                if (path_list)
+                                        ret = -1;
+                        }
+                }
+                break;
+
+        case GF_GSYNC_OPTION_TYPE_CONFIG:
+                ret = gsync_verify_config_options (dict, op_errstr, volname);
+                goto out;
+                break;
+
+        case GF_GSYNC_OPTION_TYPE_DELETE:
+                /* Check if the gsync session is still running
+                 * If so ask the user to stop geo-replication first.*/
+                ret = glusterd_gsync_get_uuid (slave, volinfo, uuid);
+                if (ret) {
+                        snprintf (errmsg, sizeof(errmsg), "Geo-replication"
+                                  " session between %s and %s does not exist.",
+                                  volinfo->volname, slave);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        *op_errstr = gf_strdup (errmsg);
+                        ret = -1;
+                        goto out;
+                } else {
+                        ret = glusterd_check_gsync_running_local (volinfo->volname,
+                                                                  slave, conf_path,
+                                                                  &is_running);
+                        if (_gf_true == is_running) {
+                                snprintf (errmsg, sizeof (errmsg), GEOREP
+                                          " session between %s & %s is "
+                                          "still active. Please stop the "
+                                          "session and retry.",
+                                          volinfo->volname, slave);
+                                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                                *op_errstr = gf_strdup (errmsg);
+                                ret = -1;
+                                goto out;
+                        }
+                }
+
+                ret = glusterd_verify_gsyncd_spawn (volinfo->volname, slave);
+                if (ret) {
+                        snprintf (errmsg, sizeof (errmsg),
+                                  "Unable to spawn gsyncd");
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                }
+
+                break;
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+static int
+stop_gsync (char *master, char *slave, char **msg,
+            char *conf_path, gf_boolean_t is_force)
+{
+        int32_t         ret     = 0;
+        int             pfd     = -1;
+        pid_t           pid     = 0;
+        char            pidfile[PATH_MAX] = {0,};
+        char            buf [1024] = {0,};
+        int             i       = 0;
+
+        GF_ASSERT (THIS);
+        GF_ASSERT (THIS->private);
+
+        pfd = gsyncd_getpidfile (master, slave, pidfile, conf_path);
+        if (pfd == -2 && !is_force) {
+                gf_log ("", GF_LOG_ERROR, GEOREP" stop validation "
+                        " failed for %s & %s", master, slave);
+                ret = -1;
+                goto out;
+        }
+        if (gsync_status_byfd (pfd) == -1 && !is_force) {
+                gf_log ("", GF_LOG_ERROR, "gsyncd b/w %s & %s is not"
+                        " running", master, slave);
+                /* monitor gsyncd already dead */
+                goto out;
+        }
+
+        if (pfd < 0)
+                goto out;
+
+        ret = read (pfd, buf, 1024);
+        if (ret > 0) {
+                pid = strtol (buf, NULL, 10);
+                ret = kill (-pid, SIGTERM);
+                if (ret) {
+                        gf_log ("", GF_LOG_WARNING,
+                                "failed to kill gsyncd");
+                        goto out;
+                }
+                for (i = 0; i < 20; i++) {
+                        if (gsync_status_byfd (pfd) == -1) {
+                                /* monitor gsyncd is dead but worker may
+                                 * still be alive, give some more time
+                                 * before SIGKILL (hack)
+                                 */
+                                usleep (50000);
+                                break;
+                        }
+                        usleep (50000);
+                }
+                kill (-pid, SIGKILL);
+                unlink (pidfile);
+        }
+        ret = 0;
+
+out:
+        sys_close (pfd);
+
+        if (is_force)
+                ret = 0;
+        return ret;
+}
+
+static int
+glusterd_gsync_configure (glusterd_volinfo_t *volinfo, char *slave,
+                          char *path_list, dict_t *dict,
+                          dict_t *resp_dict, char **op_errstr)
+{
+        int32_t         ret     = -1;
+        char            *op_name = NULL;
+        char            *op_value = NULL;
+        runner_t        runner    = {0,};
+        glusterd_conf_t *priv   = NULL;
+        char            *subop  = NULL;
+        char            *master = NULL;
+        char            *conf_path = NULL;
+        char            *slave_ip  = NULL;
+        char            *slave_vol = NULL;
+        struct stat      stbuf     = {0, };
+
+        GF_ASSERT (slave);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (dict);
+        GF_ASSERT (resp_dict);
+
+        ret = dict_get_str (dict, "subop", &subop);
+        if (ret != 0)
+                goto out;
+
+        if (strcmp (subop, "get") == 0 || strcmp (subop, "get-all") == 0) {
+                /* deferred to cli */
+                gf_log ("", GF_LOG_DEBUG, "Returning 0");
+                return 0;
+        }
+
+        ret = dict_get_str (dict, "op_name", &op_name);
+        if (ret != 0)
+                goto out;
+
+        if (strtail (subop, "set")) {
+                ret = dict_get_str (dict, "op_value", &op_value);
+                if (ret != 0)
+                        goto out;
+        }
+
+        if (THIS)
+                priv = THIS->private;
+        if (priv == NULL) {
+                gf_log ("", GF_LOG_ERROR, "priv of glusterd not present");
+                *op_errstr = gf_strdup ("glusterd defunct");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "conf_path", &conf_path);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to fetch conf file path.");
+                goto out;
+        }
+
+        master = "";
+        runinit (&runner);
+        runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL);
+        runner_argprintf (&runner, "%s", conf_path);
+        if (volinfo) {
+                master = volinfo->volname;
+                runner_argprintf (&runner, ":%s", master);
+        }
+        runner_add_arg (&runner, slave);
+        runner_argprintf (&runner, "--config-%s", subop);
+        runner_add_arg (&runner, op_name);
+        if (op_value)
+                runner_add_arg (&runner, op_value);
+        synclock_unlock (&priv->big_lock);
+        ret = runner_run (&runner);
+        synclock_lock (&priv->big_lock);
+        if (ret) {
+                gf_log ("", GF_LOG_WARNING, "gsyncd failed to "
+                        "%s %s option for %s %s peers",
+                        subop, op_name, master, slave);
+
+                gf_asprintf (op_errstr, GEOREP" config-%s failed for %s %s",
+                             subop, master, slave);
+
+                goto out;
+        }
+
+        if (!strcmp (op_name, "state_file")) {
+
+                ret = lstat (op_value, &stbuf);
+                if (ret) {
+                        ret = dict_get_str (dict, "slave_ip", &slave_ip);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to fetch slave IP.");
+                                goto out;
+                        }
+
+                        ret = dict_get_str (dict, "slave_vol", &slave_vol);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to fetch slave volume name.");
+                                goto out;
+                        }
+
+                        ret = glusterd_create_status_file (volinfo->volname, slave,
+                                                           slave_ip, slave_vol,
+                                                           "Switching Status File");
+                        if (ret || lstat (op_value, &stbuf)) {
+                                gf_log ("", GF_LOG_ERROR, "Unable to create %s"
+                                        ". Error : %s", op_value,
+                                        strerror (errno));
+                                ret = -1;
+                                goto out;
+                        }
+                }
+        }
+
+        ret = 0;
+        gf_asprintf (op_errstr, "config-%s successful", subop);
+
+out:
+        if (!ret && volinfo) {
+                ret = glusterd_check_restart_gsync_session (volinfo, slave,
+                                                            resp_dict, path_list,
+                                                            conf_path, 0);
+                if (ret)
+                        *op_errstr = gf_strdup ("internal error");
+        }
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_gsync_read_frm_status (char *path, char *buf, size_t blen)
+{
+        int                 ret = 0;
+        int           status_fd = -1;
+
+        GF_ASSERT (path);
+        GF_ASSERT (buf);
+        status_fd = open (path, O_RDONLY);
+        if (status_fd == -1) {
+                gf_log ("", GF_LOG_ERROR, "Unable to read gsyncd status"
+                        " file");
+                return -1;
+        }
+        ret = read (status_fd, buf, blen - 1);
+        if (ret > 0) {
+                size_t len = strnlen (buf, ret);
+                /* Ensure there is a NUL byte and that it's not the first.  */
+                if (len == 0 || len == blen - 1) {
+                        ret = -1;
+                } else {
+                        char *p = buf + len - 1;
+                        while (isspace (*p))
+                                *p-- = '\0';
+                        ret = 0;
+                }
+        } else if (ret < 0)
+                gf_log ("", GF_LOG_ERROR, "Status file of gsyncd is corrupt");
+
+        close (status_fd);
+        return ret;
+}
+
+static int
+glusterd_gsync_fetch_status_extra (char *path, char *buf, size_t blen)
+{
+        char sockpath[PATH_MAX] = {0,};
+        struct sockaddr_un   sa = {0,};
+        size_t                l = 0;
+        int                   s = -1;
+        struct pollfd       pfd = {0,};
+        int                 ret = 0;
+
+        l = strlen (buf);
+        /* seek to end of data in buf */
+        buf += l;
+        blen -= l;
+
+        glusterd_set_socket_filepath (path, sockpath, sizeof (sockpath));
+
+        strncpy(sa.sun_path, sockpath, sizeof(sa.sun_path));
+        if (sa.sun_path[sizeof (sa.sun_path) - 1])
+                return -1;
+        sa.sun_family = AF_UNIX;
+
+        s = socket(AF_UNIX, SOCK_STREAM, 0);
+        if (s == -1)
+                return -1;
+        ret = fcntl (s, F_GETFL);
+        if (ret != -1)
+                ret = fcntl (s, F_SETFL, ret | O_NONBLOCK);
+        if (ret == -1)
+                goto out;
+
+        ret = connect (s, (struct sockaddr *)&sa, sizeof (sa));
+        if (ret == -1)
+                goto out;
+        pfd.fd = s;
+        pfd.events = POLLIN;
+        /* we don't want to hang on gsyncd */
+        if (poll (&pfd, 1, 5000) < 1 ||
+            !(pfd.revents & POLLIN)) {
+                ret = -1;
+                goto out;
+        }
+        ret = read(s, buf, blen);
+        /* we expect a terminating 0 byte */
+        if (ret == 0 || (ret > 0 && buf[ret - 1]))
+                ret = -1;
+        if (ret > 0)
+                ret = 0;
+
+ out:
+        close (s);
+        return ret;
+}
+
+static int
+dict_get_param (dict_t *dict, char *key, char **param)
+{
+        char  *dk = NULL;
+        char   *s = NULL;
+        char    x = '\0';
+        int   ret = 0;
+
+        if (dict_get_str (dict, key, param) == 0)
+                return 0;
+
+        dk = gf_strdup (key);
+        if (!key)
+                return -1;
+
+        s = strpbrk (dk, "-_");
+        if (!s)
+                return -1;
+        x = (*s == '-') ? '_' : '-';
+        *s++ = x;
+        while ((s = strpbrk (s, "-_")))
+                *s++ = x;
+
+        ret = dict_get_str (dict, dk, param);
+
+        GF_FREE (dk);
+        return ret;
+}
+
+static int
+glusterd_read_status_file (glusterd_volinfo_t *volinfo, char *slave,
+                           char *conf_path, dict_t *dict, char *node)
+{
+        glusterd_conf_t *priv = NULL;
+        int              ret = 0;
+        char            *statefile = NULL;
+        char            *master    = NULL;
+        char             buf[1024] = "defunct";
+        char             nds[1024] = {0, };
+        char             mst[1024] = {0, };
+        char             slv[1024] = {0, };
+        char             sts[1024] = {0, };
+        char            *bufp = NULL;
+        dict_t          *confd = NULL;
+        int              gsync_count = 0;
+        int              status = 0;
+        char *dyn_node = NULL;
+        char *path_list = NULL;
+
+        GF_ASSERT (THIS);
+        GF_ASSERT (THIS->private);
+        GF_ASSERT (volinfo);
+
+        master = volinfo->volname;
+
+        confd = dict_new ();
+        if (!dict) {
+                gf_log ("", GF_LOG_ERROR, "Not able to create dict.");
+                return -1;
+        }
+
+        priv = THIS->private;
+
+        ret = glusterd_gsync_get_config (master, slave, conf_path,
+                                         confd);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get configuration data"
+                        "for %s(master), %s(slave)", master, slave);
+                goto done;
+
+        }
+
+        ret = dict_get_param (confd, "state_file", &statefile);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get state_file's name "
+                        "for %s(master), %s(slave). Please check gsync "
+                        "config file.", master, slave);
+                goto done;
+        }
+        ret = glusterd_gsync_read_frm_status (statefile, buf, sizeof (buf));
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to read the status"
+                        "file for %s(master), %s(slave)", master, slave);
+                strncpy (buf, "defunct", sizeof (buf));
+                goto done;
+        }
+
+        ret = gsync_status (master, slave, conf_path, &status);
+        if (ret == 0 && status == -1) {
+                if ((strcmp (buf, "Not Started")) &&
+                    (strcmp (buf, "Stopped")))
+                    strncpy (buf, "defunct", sizeof (buf));
+                goto done;
+        } else if (ret == -1) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get gsync status");
+                goto done;
+        }
+
+        if (strcmp (buf, "Stable") != 0)
+                goto done;
+
+        ret = dict_get_param (confd, "state_socket_unencoded", &statefile);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get state_socket_unencoded"
+                        " filepath. Please check gsync config file.");
+                goto done;
+        }
+        ret = glusterd_gsync_fetch_status_extra (statefile, buf, sizeof (buf));
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to fetch extra status"
+                        "for %s(master), %s(slave)", master, slave);
+                /* there is a slight chance that this occurs due to race
+                 * -- in that case, the following options all seem bad:
+                 *
+                 * - suppress irregurlar behavior by just leaving status
+                 *   on "OK"
+                 * - freak out users with a misleading "defunct"
+                 * - overload the meaning of the regular error signal
+                 *   mechanism of gsyncd, that is, when status is "faulty"
+                 *
+                 * -- so we just come up with something new...
+                 */
+                strncpy (buf, "N/A", sizeof (buf));
+                goto done;
+        }
+
+ done:
+        if ((!strcmp (buf, "defunct")) ||
+            (!strcmp (buf, "Not Started")) ||
+            (!strcmp (buf, "Stopped"))) {
+                ret = glusterd_get_local_brickpaths (volinfo, &path_list);
+                if (!path_list) {
+                        gf_log ("", GF_LOG_DEBUG, "This node not being part of"
+                                " volume should not be running gsyncd. Hence"
+                                " shouldn't display status for this node.");
+                        ret = 0;
+                        goto out;
+                }
+        }
+
+        ret = dict_get_int32 (dict, "gsync-count", &gsync_count);
+
+        if (ret)
+                gsync_count = 1;
+        else
+                gsync_count++;
+
+        (void) snprintf (nds, sizeof (nds), "node%d", gsync_count);
+        dyn_node = gf_strdup (node);
+        if (!dyn_node)
+                goto out;
+        ret = dict_set_dynstr (dict, nds, dyn_node);
+        if (ret) {
+                GF_FREE (dyn_node);
+                goto out;
+        }
+
+        snprintf (mst, sizeof (mst), "master%d", gsync_count);
+        master = gf_strdup (master);
+        if (!master)
+                goto out;
+        ret = dict_set_dynstr (dict, mst, master);
+        if (ret) {
+                GF_FREE (master);
+                goto out;
+        }
+
+        snprintf (slv, sizeof (slv), "slave%d", gsync_count);
+        slave = gf_strdup (slave);
+        if (!slave)
+                goto out;
+        ret = dict_set_dynstr (dict, slv, slave);
+        if (ret) {
+                GF_FREE (slave);
+                goto out;
+        }
+
+        snprintf (sts, sizeof (slv), "status%d", gsync_count);
+        bufp = gf_strdup (buf);
+        if (!bufp)
+                goto out;
+        ret = dict_set_dynstr (dict, sts, bufp);
+        if (ret) {
+                GF_FREE (bufp);
+                goto out;
+        }
+        ret = dict_set_int32 (dict, "gsync-count", gsync_count);
+        if (ret)
+                goto out;
+
+ out:
+        dict_destroy (confd);
+
+        return 0;
+}
+
+int
+glusterd_check_restart_gsync_session (glusterd_volinfo_t *volinfo, char *slave,
+                                      dict_t *resp_dict, char *path_list,
+                                      char *conf_path, gf_boolean_t is_force)
+{
+
+        int                    ret = 0;
+        glusterd_conf_t        *priv = NULL;
+        char                   *status_msg = NULL;
+        gf_boolean_t           is_running = _gf_false;
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (slave);
+        GF_ASSERT (THIS);
+        GF_ASSERT (THIS->private);
+
+        priv = THIS->private;
+
+        ret = glusterd_check_gsync_running_local (volinfo->volname,
+                                                  slave, conf_path,
+                                                  &is_running);
+        if (!ret && (_gf_true != is_running))
+                /* gsynd not running, nothing to do */
+                goto out;
+
+        ret = stop_gsync (volinfo->volname, slave, &status_msg,
+                          conf_path, is_force);
+        if (ret == 0 && status_msg)
+                ret = dict_set_str (resp_dict, "gsync-status",
+                                    status_msg);
+        if (ret == 0)
+                ret = glusterd_start_gsync (volinfo, slave, path_list,
+                                            conf_path, uuid_utoa(MY_UUID),
+                                            NULL);
+
+ out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+static int32_t
+glusterd_marker_changelog_create_volfile (glusterd_volinfo_t *volinfo)
+{
+        int32_t          ret     = 0;
+
+        ret = glusterd_create_volfiles_and_notify_services (volinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to create volfile"
+                        " for setting of marker while '"GEOREP" start'");
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret)
+                goto out;
+
+        if (GLUSTERD_STATUS_STARTED == volinfo->status)
+                ret = glusterd_nodesvcs_handle_graph_change (volinfo);
+        ret = 0;
+out:
+        return ret;
+}
+
+static int
+glusterd_set_gsync_knob (glusterd_volinfo_t *volinfo, char *key, int *vc)
+{
+        int   ret          = -1;
+        int   conf_enabled = _gf_false;
+        char *knob_on      = NULL;
+
+        GF_ASSERT (THIS);
+        GF_ASSERT (THIS->private);
+
+        conf_enabled = glusterd_volinfo_get_boolean (volinfo, key);
+        if (conf_enabled == -1) {
+                gf_log ("", GF_LOG_ERROR,
+                        "failed to get key %s from volinfo", key);
+                goto out;
+        }
+
+        ret = 0;
+        if (conf_enabled == _gf_false) {
+                *vc = 1;
+                knob_on = gf_strdup ("on");
+                if (knob_on == NULL) {
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = glusterd_gsync_volinfo_dict_set (volinfo,
+                                                       key, knob_on);
+        }
+
+ out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_set_gsync_confs (glusterd_volinfo_t *volinfo)
+{
+        int ret             = -1;
+        int volfile_changed = 0;
+
+        ret = glusterd_set_gsync_knob (volinfo,
+                                       VKEY_MARKER_XTIME, &volfile_changed);
+        if (ret)
+                goto out;
+
+        /**
+         * enable ignore-pid-check blindly as it could be needed for
+         * cascading setups.
+         */
+        ret = glusterd_set_gsync_knob (volinfo, VKEY_MARKER_XTIME_FORCE,
+                                       &volfile_changed);
+        if (ret)
+                goto out;
+
+        ret = glusterd_set_gsync_knob (volinfo,
+                                       VKEY_CHANGELOG, &volfile_changed);
+        if (ret)
+                goto out;
+
+        if (volfile_changed)
+                ret = glusterd_marker_changelog_create_volfile (volinfo);
+
+ out:
+        return ret;
+}
+
+static int
+glusterd_get_gsync_status_mst_slv (glusterd_volinfo_t *volinfo,
+                                   char *slave, char *conf_path,
+                                   dict_t *rsp_dict, char *node)
+{
+        char              *statefile = NULL;
+        uuid_t             uuid = {0, };
+        glusterd_conf_t    *priv = NULL;
+        int                ret = 0;
+        struct stat        stbuf = {0, };
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (slave);
+        GF_ASSERT (THIS);
+        GF_ASSERT (THIS->private);
+
+        priv = THIS->private;
+
+        ret = glusterd_gsync_get_uuid (slave, volinfo, uuid);
+        if (ret) {
+                gf_log ("", GF_LOG_INFO, "geo-replication status %s %s :"
+                        "session is not active", volinfo->volname, slave);
+
+                ret = glusterd_get_statefile_name (volinfo, slave,
+                                                   conf_path, &statefile);
+                if (ret) {
+                        if (!strstr(slave, "::"))
+                                gf_log ("", GF_LOG_INFO,
+                                        "%s is not a valid slave url.", slave);
+                        else
+                                gf_log ("", GF_LOG_INFO, "Unable to get"
+                                        " statefile's name");
+                        ret = 0;
+                        goto out;
+                }
+
+                ret = lstat (statefile, &stbuf);
+                if (ret) {
+                        gf_log ("", GF_LOG_INFO, "%s statefile not present.",
+                                statefile);
+                        ret = 0;
+                        goto out;
+                }
+        }
+
+        ret = glusterd_read_status_file (volinfo, slave, conf_path,
+                                         rsp_dict, node);
+out:
+        if (statefile)
+                GF_FREE (statefile);
+
+        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
+
+static int
+glusterd_get_gsync_status_mst (glusterd_volinfo_t *volinfo, dict_t *rsp_dict,
+                               char *node)
+{
+        glusterd_gsync_status_temp_t  param = {0, };
+
+        GF_ASSERT (volinfo);
+
+        param.rsp_dict = rsp_dict;
+        param.volinfo = volinfo;
+        param.node = node;
+        dict_foreach (volinfo->gsync_slaves, _get_status_mst_slv, &param);
+
+        return 0;
+}
+
+static int
+glusterd_get_gsync_status_all (dict_t *rsp_dict, char *node)
+{
+
+        int32_t                 ret = 0;
+        glusterd_conf_t         *priv = NULL;
+        glusterd_volinfo_t      *volinfo = NULL;
+
+        GF_ASSERT (THIS);
+        priv = THIS->private;
+
+        GF_ASSERT (priv);
+
+        list_for_each_entry (volinfo, &priv->volumes, vol_list) {
+                ret = glusterd_get_gsync_status_mst (volinfo, rsp_dict, node);
+                if (ret)
+                        goto out;
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+
+}
+
+static int
+glusterd_get_gsync_status (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
+{
+        char               *slave  = NULL;
+        char               *volname = NULL;
+        char               *conf_path = NULL;
+        char               errmsg[PATH_MAX] = {0, };
+        gf_boolean_t       exists = _gf_false;
+        glusterd_volinfo_t *volinfo = NULL;
+        int                ret = 0;
+        char my_hostname[256] = {0,};
+
+        ret = gethostname(my_hostname, 256);
+        if (ret) {
+                /* stick to N/A */
+                (void) strcpy (my_hostname, "N/A");
+        }
+
+        ret = dict_get_str (dict, "master", &volname);
+        if (ret < 0){
+                ret = glusterd_get_gsync_status_all (rsp_dict, my_hostname);
+                goto out;
+        }
+
+        exists = glusterd_check_volume_exists (volname);
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if ((ret) || (!exists)) {
+                gf_log ("", GF_LOG_WARNING, "volume name does not exist");
+                snprintf (errmsg, sizeof(errmsg), "Volume name %s does not"
+                          " exist", volname);
+                *op_errstr = gf_strdup (errmsg);
+                ret = -1;
+                goto out;
+        }
+
+
+        ret = dict_get_str (dict, "slave", &slave);
+        if (ret < 0) {
+                ret = glusterd_get_gsync_status_mst (volinfo,
+                                                     rsp_dict, my_hostname);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "conf_path", &conf_path);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to fetch conf file path.");
+                goto out;
+        }
+
+        ret = glusterd_get_gsync_status_mst_slv (volinfo, slave, conf_path,
+                                                 rsp_dict, my_hostname);
+
+ out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_gsync_delete (glusterd_volinfo_t *volinfo, char *slave, char *slave_ip,
+                       char *slave_vol, char *path_list, dict_t *dict,
+                       dict_t *resp_dict, char **op_errstr)
+{
+        int32_t         ret     = -1;
+        runner_t        runner    = {0,};
+        glusterd_conf_t *priv   = NULL;
+        char            *master = NULL;
+        char            *gl_workdir = NULL;
+        char             geo_rep_dir[PATH_MAX] = "";
+        char            *conf_path = NULL;
+
+        GF_ASSERT (slave);
+        GF_ASSERT (slave_ip);
+        GF_ASSERT (slave_vol);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (dict);
+        GF_ASSERT (resp_dict);
+
+        if (THIS)
+                priv = THIS->private;
+        if (priv == NULL) {
+                gf_log ("", GF_LOG_ERROR, "priv of glusterd not present");
+                *op_errstr = gf_strdup ("glusterd defunct");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "conf_path", &conf_path);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to fetch conf file path.");
+                goto out;
+        }
+
+        gl_workdir = priv->workdir;
+        master = "";
+        runinit (&runner);
+        runner_add_args  (&runner, GSYNCD_PREFIX"/gsyncd",
+                          "--delete", "-c", NULL);
+        runner_argprintf (&runner, "%s", conf_path);
+
+        if (volinfo) {
+                master = volinfo->volname;
+                runner_argprintf (&runner, ":%s", master);
+        }
+        runner_add_arg (&runner, slave);
+        runner_redir (&runner, STDOUT_FILENO, RUN_PIPE);
+        synclock_unlock (&priv->big_lock);
+        ret = runner_run (&runner);
+        synclock_lock (&priv->big_lock);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "gsyncd failed to "
+                        "delete session info for %s and %s peers",
+                        master, slave);
+
+                gf_asprintf (op_errstr, "gsyncd failed to "
+                             "delete session info for %s and %s peers",
+                             master, slave);
+
+                goto out;
+        }
+
+        ret = snprintf (geo_rep_dir, sizeof(geo_rep_dir) - 1,
+                        "%s/"GEOREP"/%s_%s_%s", gl_workdir,
+                        volinfo->volname, slave_ip, slave_vol);
+        geo_rep_dir[ret] = '\0';
+
+        ret = rmdir (geo_rep_dir);
+        if (ret) {
+                if (errno == ENOENT)
+                        gf_log ("", GF_LOG_DEBUG, "Geo Rep Dir(%s) Not Present.",
+                                geo_rep_dir);
+                else {
+                        gf_log ("", GF_LOG_ERROR, "Unable to delete "
+                                "Geo Rep Dir(%s). Error: %s", geo_rep_dir,
+                                strerror (errno));
+                        goto out;
+                }
+        }
+
+        ret = 0;
+
+        gf_asprintf (op_errstr, "delete successful");
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_sys_exec (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
+{
+        char                   buf[PATH_MAX]          = "";
+        char                   cmd_arg_name[PATH_MAX] = "";
+        char                   output_name[PATH_MAX]  = "";
+        char                   errmsg[PATH_MAX]       = "";
+        char                  *ptr                    = NULL;
+        char                  *bufp                   = NULL;
+        char                  *command                = NULL;
+        char                 **cmd_args               = NULL;
+        int                    ret                    = -1;
+        int                    i                      = -1;
+        int                    cmd_args_count         = 0;
+        int                    output_count           = 0;
+        glusterd_conf_t       *priv                   = NULL;
+        runner_t               runner                 = {0,};
+
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (rsp_dict);
+
+        if (THIS)
+                priv = THIS->private;
+        if (priv == NULL) {
+                gf_log ("", GF_LOG_ERROR, "priv of glusterd not present");
+                *op_errstr = gf_strdup ("glusterd defunct");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "command", &command);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to get command from dict");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "cmd_args_count", &cmd_args_count);
+        if (ret)
+                gf_log ("", GF_LOG_INFO, "No cmd_args_count");
+
+        if (cmd_args_count) {
+                cmd_args = GF_CALLOC (cmd_args_count, sizeof (char*),
+                                      gf_common_mt_char);
+                if (!cmd_args) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to calloc. "
+                                "Errno = %s", strerror(errno));
+                        goto out;
+                }
+
+                for (i=1; i <= cmd_args_count; i++) {
+                        memset (cmd_arg_name, '\0', sizeof(cmd_arg_name));
+                        snprintf (cmd_arg_name, sizeof(cmd_arg_name),
+                                  "cmd_arg_%d", i);
+                        ret = dict_get_str (dict, cmd_arg_name, &cmd_args[i-1]);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to get %s in dict",
+                                        cmd_arg_name);
+                                goto out;
+                        }
+                }
+        }
+
+        runinit (&runner);
+        runner_argprintf (&runner, GSYNCD_PREFIX"/peer_%s", command);
+        for (i=0; i < cmd_args_count; i++)
+                runner_add_arg (&runner, cmd_args[i]);
+        runner_redir (&runner, STDOUT_FILENO, RUN_PIPE);
+        synclock_unlock (&priv->big_lock);
+        ret = runner_start (&runner);
+        if (ret == -1) {
+                snprintf (errmsg, sizeof (errmsg), "Unable to "
+                          "execute command. Error : %s",
+                          strerror (errno));
+                *op_errstr = gf_strdup (errmsg);
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                ret = -1;
+                synclock_lock (&priv->big_lock);
+                goto out;
+        }
+
+        ptr = fgets(buf, sizeof(buf), runner_chio (&runner, STDOUT_FILENO));
+        if (ptr) {
+                ret = dict_get_int32 (rsp_dict, "output_count", &output_count);
+                if (ret)
+                        output_count = 1;
+                else
+                        output_count++;
+                memset (output_name, '\0', sizeof (output_name));
+                snprintf (output_name, sizeof (output_name),
+                          "output_%d", output_count);
+                if (buf[strlen(buf) - 1] == '\n')
+                        buf[strlen(buf) - 1] = '\0';
+                bufp = gf_strdup (buf);
+                if (!bufp)
+                        gf_log ("", GF_LOG_ERROR, "gf_strdup failed.");
+                ret = dict_set_dynstr (rsp_dict, output_name, bufp);
+                if (ret) {
+                        GF_FREE (bufp);
+                        gf_log ("", GF_LOG_ERROR, "output set failed.");
+                }
+                ret = dict_set_int32 (rsp_dict, "output_count", output_count);
+                if (ret)
+                        gf_log ("", GF_LOG_ERROR, "output_count set failed.");
+        }
+
+        ret = runner_end (&runner);
+        if (ret) {
+                snprintf (errmsg, sizeof (errmsg), "Unable to "
+                          "end. Error : %s",
+                          strerror (errno));
+                *op_errstr = gf_strdup (errmsg);
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                ret = -1;
+                synclock_lock (&priv->big_lock);
+                goto out;
+        }
+        synclock_lock (&priv->big_lock);
+
+        ret = 0;
+out:
+        if (cmd_args) {
+                GF_FREE (cmd_args);
+                cmd_args = NULL;
+        }
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_copy_file (dict_t *dict, char **op_errstr)
+{
+        char             abs_filename[PATH_MAX] = "";
+        char             errmsg[PATH_MAX]       = "";
+        char            *filename               = NULL;
+        char            *host_uuid              = NULL;
+        char             uuid_str [64]          = {0};
+        char            *contents               = NULL;
+        char             buf[1024]              = "";
+        int              ret                    = -1;
+        int              fd                     = -1;
+        int              bytes_writen           = 0;
+        int              bytes_read             = 0;
+        int              contents_size          = -1;
+        int              file_mode              = -1;
+        glusterd_conf_t *priv                   = NULL;
+        struct stat      stbuf                  = {0,};
+
+
+        if (THIS)
+                priv = THIS->private;
+        if (priv == NULL) {
+                gf_log ("", GF_LOG_ERROR, "priv of glusterd not present");
+                *op_errstr = gf_strdup ("glusterd defunct");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "host-uuid", &host_uuid);
+        if (ret < 0)
+                goto out;
+
+        ret = dict_get_str (dict, "source", &filename);
+        if (ret < 0) {
+               gf_log ("", GF_LOG_ERROR, "Unable to fetch"
+                       " filename from dict.");
+               *op_errstr = gf_strdup ("command unsuccessful");
+               goto out;
+        }
+        snprintf (abs_filename, sizeof(abs_filename),
+                  "%s/%s", priv->workdir, filename);
+
+        uuid_utoa_r (MY_UUID, uuid_str);
+        if (!strcmp (uuid_str, host_uuid)) {
+                ret = lstat (abs_filename, &stbuf);
+                if (ret) {
+                        snprintf (errmsg, sizeof (errmsg), "Source file"
+                                 " does not exist in %s", priv->workdir);
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        goto out;
+                }
+
+                contents = GF_CALLOC(1, stbuf.st_size+1, gf_common_mt_char);
+                if (!contents) {
+                        snprintf (errmsg, sizeof (errmsg),
+                                  "Unable to allocate memory");
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        ret = -1;
+                        goto out;
+                }
+
+                fd = open (abs_filename, O_RDONLY);
+                if (fd < 0) {
+                        snprintf (errmsg, sizeof (errmsg), "Unable to open %s",
+                                  abs_filename);
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        ret = -1;
+                        goto out;
+                }
+
+                do {
+                        ret = read (fd, buf, sizeof(buf));
+                        if (ret > 0) {
+                                memcpy (contents+bytes_read, buf, ret);
+                                bytes_read += ret;
+                                memset (buf, '\0', sizeof(buf));
+                        }
+                } while (ret > 0);
+
+                if (bytes_read != stbuf.st_size) {
+                        snprintf (errmsg, sizeof (errmsg), "Unable to read all "
+                                  "the data from %s", abs_filename);
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = dict_set_int32 (dict, "contents_size", stbuf.st_size);
+                if (ret) {
+                        snprintf (errmsg, sizeof (errmsg), "Unable to set"
+                                  " contents size in dict.");
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        goto out;
+                }
+
+                ret = dict_set_int32 (dict, "file_mode",
+                                      (int32_t)stbuf.st_mode);
+                if (ret) {
+                        snprintf (errmsg, sizeof (errmsg), "Unable to set"
+                                  " file mode in dict.");
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        goto out;
+                }
+
+                ret = dict_set_bin (dict, "common_pem_contents",
+                                    contents, stbuf.st_size);
+                if (ret) {
+                        snprintf (errmsg, sizeof (errmsg), "Unable to set"
+                                  " pem contents in dict.");
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        goto out;
+                }
+                close (fd);
+        } else {
+                ret = dict_get_bin (dict, "common_pem_contents",
+                                    (void **) &contents);
+                if (ret) {
+                        snprintf (errmsg, sizeof (errmsg), "Unable to get"
+                                  " pem contents in dict.");
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        goto out;
+                }
+
+                ret = dict_get_int32 (dict, "contents_size", &contents_size);
+                if (ret) {
+                        snprintf (errmsg, sizeof (errmsg), "Unable to set"
+                                  " contents size in dict.");
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        goto out;
+                }
+
+                ret = dict_get_int32 (dict, "file_mode", &file_mode);
+                if (ret) {
+                        snprintf (errmsg, sizeof (errmsg), "Unable to get"
+                                  " file mode in dict.");
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        goto out;
+                }
+
+                fd = open (abs_filename, O_WRONLY | O_TRUNC | O_CREAT, 0600);
+                if (fd < 0) {
+                        snprintf (errmsg, sizeof (errmsg), "Unable to open %s",
+                                  abs_filename);
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        ret = -1;
+                        goto out;
+                }
+
+                bytes_writen = write (fd, contents, contents_size);
+
+                if (bytes_writen != contents_size) {
+                        snprintf (errmsg, sizeof (errmsg), "Failed to write"
+                                  " to %s", abs_filename);
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        ret = -1;
+                        goto out;
+                }
+
+                fchmod (fd, file_mode);
+                close (fd);
+        }
+
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
+{
+        int32_t             ret     = -1;
+        int32_t             type    = -1;
+        dict_t             *ctx    = NULL;
+        dict_t             *resp_dict = NULL;
+        char               *host_uuid = NULL;
+        char               *slave  = NULL;
+        char               *slave_ip  = NULL;
+        char               *slave_vol = NULL;
+        char               *volname = NULL;
+        char               *path_list = NULL;
+        glusterd_volinfo_t *volinfo = NULL;
+        glusterd_conf_t    *priv = NULL;
+        gf_boolean_t        is_force = _gf_false;
+        char               *status_msg = NULL;
+        gf_boolean_t        is_running = _gf_false;
+        char               *conf_path = NULL;
+
+        GF_ASSERT (THIS);
+        GF_ASSERT (THIS->private);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+
+        priv = THIS->private;
+
+        ret = dict_get_int32 (dict, "type", &type);
+        if (ret < 0)
+                goto out;
+
+        ret = dict_get_str (dict, "host-uuid", &host_uuid);
+        if (ret < 0)
+                goto out;
+
+        ctx = glusterd_op_get_ctx ();
+        resp_dict = ctx ? ctx : rsp_dict;
+        GF_ASSERT (resp_dict);
+
+        if (type == GF_GSYNC_OPTION_TYPE_STATUS) {
+                ret = glusterd_get_gsync_status (dict, op_errstr, resp_dict);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "slave", &slave);
+        if (ret < 0)
+                goto out;
+
+        ret = dict_get_str (dict, "slave_ip", &slave_ip);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to fetch slave volume name.");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "slave_vol", &slave_vol);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to fetch slave volume name.");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "conf_path", &conf_path);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to fetch conf file path.");
+                goto out;
+        }
+
+        if (dict_get_str (dict, "master", &volname) == 0) {
+                ret = glusterd_volinfo_find (volname, &volinfo);
+                if (ret) {
+                        gf_log ("", GF_LOG_WARNING, "Volinfo for %s (master) not found",
+                                volname);
+                        goto out;
+                }
+
+                ret = glusterd_get_local_brickpaths (volinfo, &path_list);
+        }
+
+        if (type == GF_GSYNC_OPTION_TYPE_CONFIG) {
+                ret = glusterd_gsync_configure (volinfo, slave, path_list,
+                                                dict, resp_dict, op_errstr);
+
+                ret = dict_set_str (resp_dict, "conf_path", conf_path);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Unable to store conf_file_path.");
+                        goto out;
+                }
+                goto out;
+        }
+
+        if (type == GF_GSYNC_OPTION_TYPE_DELETE) {
+                ret = glusterd_remove_slave_in_info(volinfo, slave, op_errstr);
+                if (ret && !is_force && path_list)
+                        goto out;
+
+                ret = glusterd_gsync_delete (volinfo, slave, slave_ip,
+                                             slave_vol, path_list, dict,
+                                             resp_dict, op_errstr);
+                goto out;
+        }
+
+        if (!volinfo) {
+                ret = -1;
+                goto out;
+        }
+
+        is_force = dict_get_str_boolean (dict, "force", _gf_false);
+
+        if (type == GF_GSYNC_OPTION_TYPE_START) {
+
+                ret = glusterd_set_gsync_confs (volinfo);
+                if (ret != 0) {
+                        gf_log ("", GF_LOG_WARNING, "marker/changelog start failed");
+                        *op_errstr = gf_strdup ("failed to initialize indexing");
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = glusterd_start_gsync (volinfo, slave, path_list,
+                                            conf_path, host_uuid, op_errstr);
+        }
+
+        if (type == GF_GSYNC_OPTION_TYPE_STOP) {
+                ret = glusterd_check_gsync_running_local (volinfo->volname,
+                                                          slave, conf_path,
+                                                          &is_running);
+                if (!ret && !is_force && path_list &&
+                    (_gf_true != is_running)) {
+                        gf_log ("", GF_LOG_WARNING, GEOREP" is not set up for"
+                                "%s(master) and %s(slave)", volname, slave);
+                        *op_errstr = strdup (GEOREP" is not set up");
+                        goto out;
+                }
+
+                ret = stop_gsync (volname, slave, &status_msg, conf_path, is_force);
+                if (ret == 0 && status_msg)
+                        ret = dict_set_str (resp_dict, "gsync-status",
+                                            status_msg);
+                if (ret != 0 && !is_force && path_list)
+                        *op_errstr = gf_strdup ("internal error");
+
+                if (!ret) {
+                        ret = glusterd_create_status_file (volinfo->volname,
+                                                           slave, slave_ip,
+                                                           slave_vol, "Stopped");
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR, "Unable to update"
+                                        "state_file. Error : %s",
+                                        strerror (errno));
+                        }
+                }
+        }
+
+out:
+        if (path_list) {
+                GF_FREE (path_list);
+                path_list = NULL;
+        }
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_get_slave_details_confpath (glusterd_volinfo_t *volinfo, dict_t *dict,
+                                     char **slave_ip, char **slave_vol,
+                                     char **conf_path, char **op_errstr)
+{
+        int                ret                = -1;
+        char               confpath[PATH_MAX] = "";
+        glusterd_conf_t   *priv               = NULL;
+        char              *slave              = NULL;
+
+        GF_ASSERT (THIS);
+        priv = THIS->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_str (dict, "slave", &slave);
+        if (ret || !slave) {
+                gf_log ("", GF_LOG_ERROR, "Unable to fetch slave from dict");
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_get_slave_info (slave, slave_ip, slave_vol, op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to fetch slave details.");
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_str (dict, "slave_ip", *slave_ip);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to store slave IP.");
+                goto out;
+        }
+
+        ret = dict_set_str (dict, "slave_vol", *slave_vol);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to store slave volume name.");
+                goto out;
+        }
+
+        ret = snprintf (confpath, sizeof(confpath) - 1,
+                        "%s/"GEOREP"/%s_%s_%s/gsyncd.conf",
+                        priv->workdir, volinfo->volname,
+                        *slave_ip, *slave_vol);
+        confpath[ret] = '\0';
+        *conf_path = gf_strdup (confpath);
+        if (!(*conf_path)) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to gf_strdup. Error: %s", strerror (errno));
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_str (dict, "conf_path", *conf_path);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to store conf_path");
+                goto out;
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG,"Returning %d", ret);
+        return ret;
+
+}
+
+static int
+glusterd_get_slave_info (char *slave, char **slave_ip,
+                         char **slave_vol, char **op_errstr)
+{
+        char     *tmp              = NULL;
+        char     *save_ptr         = NULL;
+        char    **linearr          = NULL;
+        int32_t   ret              = -1;
+        char      errmsg[PATH_MAX] = "";
+
+        ret = glusterd_urltransform_single (slave, "normalize",
+                                            &linearr);
+        if (ret == -1) {
+                ret = snprintf (errmsg, sizeof(errmsg) - 1,
+                                "Invalid Url: %s", slave);
+                errmsg[ret] = '\0';
+                *op_errstr = gf_strdup (errmsg);
+                gf_log ("", GF_LOG_ERROR, "Failed to normalize url");
+                goto out;
+        }
+
+        tmp = strtok_r (linearr[0], "/", &save_ptr);
+        tmp = strtok_r (NULL, "/", &save_ptr);
+        slave = strtok_r (tmp, ":", &save_ptr);
+        if (slave) {
+                ret = glusterd_mountbroker_check (&slave, op_errstr);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Invalid slave url: %s", *op_errstr);
+                        goto out;
+                }
+
+                *slave_ip = gf_strdup (slave);
+                if (!*slave_ip) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Failed to gf_strdup");
+                        ret = -1;
+                        goto out;
+                }
+                gf_log ("", GF_LOG_DEBUG, "Slave IP : %s", *slave_ip);
+                ret = 0;
+        } else {
+                gf_log ("", GF_LOG_ERROR, "Invalid slave name");
+                goto out;
+        }
+
+        slave = strtok_r (NULL, ":", &save_ptr);
+        if (slave) {
+                *slave_vol = gf_strdup (slave);
+                if (!*slave_vol) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Failed to gf_strdup");
+                        ret = -1;
+                        goto out;
+                }
+                gf_log ("", GF_LOG_DEBUG, "Slave Vol : %s", *slave_vol);
+                ret = 0;
+        } else {
+                gf_log ("", GF_LOG_ERROR, "Invalid slave name");
+                goto out;
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+static void
+runinit_gsyncd_setrx (runner_t *runner, char *conf_path)
+{
+        runinit (runner);
+        runner_add_args (runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL);
+        runner_argprintf (runner, "%s", conf_path);
+        runner_add_arg (runner, "--config-set-rx");
+}
+
+static int
+glusterd_check_gsync_present (int *valid_state)
+{
+        char                buff[PATH_MAX] = {0, };
+        runner_t            runner = {0,};
+        char               *ptr = NULL;
+        int                 ret = 0;
+
+        runinit (&runner);
+        runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "--version", NULL);
+        runner_redir (&runner, STDOUT_FILENO, RUN_PIPE);
+        ret = runner_start (&runner);
+        if (ret == -1) {
+                if (errno == ENOENT) {
+                        gf_log ("glusterd", GF_LOG_INFO, GEOREP
+                                 " module not installed in the system");
+                        *valid_state = 0;
+                }
+                else {
+                        gf_log ("glusterd", GF_LOG_ERROR, GEOREP
+                                  " module not working as desired");
+                        *valid_state = -1;
+                }
+                goto out;
+        }
+
+        ptr = fgets(buff, sizeof(buff), runner_chio (&runner, STDOUT_FILENO));
+        if (ptr) {
+                if (!strstr (buff, "gsyncd")) {
+                        ret = -1;
+                        gf_log ("glusterd", GF_LOG_ERROR, GEOREP" module not "
+                                 "working as desired");
+                        *valid_state = -1;
+                        goto out;
+                }
+        } else {
+                ret = -1;
+                gf_log ("glusterd", GF_LOG_ERROR, GEOREP" module not "
+                         "working as desired");
+                *valid_state = -1;
+                goto out;
+        }
+
+        ret = 0;
+ out:
+
+        runner_end (&runner);
+
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+
+}
+
+static int
+create_conf_file (glusterd_conf_t *conf, char *conf_path)
+#define RUN_GSYNCD_CMD do {                                                          \
+        ret = runner_run_reuse (&runner);                                            \
+        if (ret == -1) {                                                             \
+                runner_log (&runner, "glusterd", GF_LOG_ERROR, "command failed");    \
+                runner_end (&runner);                                                \
+                goto out;                                                            \
+        }                                                                            \
+        runner_end (&runner);                                                        \
+} while (0)
+{
+        int ret = 0;
+        runner_t runner = {0,};
+        char georepdir[PATH_MAX] = {0,};
+        int valid_state = 0;
+
+        valid_state = -1;
+        ret = glusterd_check_gsync_present (&valid_state);
+        if (-1 == ret) {
+                ret = valid_state;
+                goto out;
+        }
+
+        ret = snprintf (georepdir, sizeof(georepdir) - 1, "%s/"GEOREP,
+                        conf->workdir);
+        georepdir[ret] = '\0';
+
+        /************
+         * master pre-configuration
+         ************/
+
+        /* remote-gsyncd */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner, "remote-gsyncd", GSYNCD_PREFIX"/gsyncd", ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner, "remote-gsyncd", "/nonexistent/gsyncd",
+                         ".", "^ssh:", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* gluster-command-dir */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner, "gluster-command-dir", SBIN_DIR"/",
+                         ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* gluster-params */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner, "gluster-params",
+                         "aux-gfid-mount xlator-option=*-dht.assert-no-child-down=true",
+                         ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* ssh-command */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_arg (&runner, "ssh-command");
+        runner_argprintf (&runner,
+                          "ssh -oPasswordAuthentication=no "
+                           "-oStrictHostKeyChecking=no "
+                           "-i %s/secret.pem", georepdir);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* pid-file */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_arg (&runner, "pid-file");
+        runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.pid", georepdir);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* state-file */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_arg (&runner, "state-file");
+        runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.status", georepdir);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* state-detail-file */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_arg (&runner, "state-detail-file");
+        runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}-detail.status", georepdir);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* state-socket */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_arg (&runner, "state-socket-unencoded");
+        runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.socket", georepdir);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* socketdir */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner, "socketdir", GLUSTERD_SOCK_DIR, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* log-file */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner,
+                         "log-file",
+                         DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"/${mastervol}/${eSlave}.log",
+                         ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* gluster-log-file */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner,
+                         "gluster-log-file",
+                         DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"/${mastervol}/${eSlave}${local_id}.gluster.log",
+                         ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* ignore-deletes */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner, "ignore-deletes", "true", ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* special-sync-mode */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner, "special-sync-mode", "partial", ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* change-detector == changelog */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args(&runner, "change-detector", "changelog", ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_arg(&runner, "working-dir");
+        runner_argprintf(&runner, "%s/${mastervol}/${eSlave}",
+                         DEFAULT_VAR_RUN_DIRECTORY);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /************
+         * slave pre-configuration
+         ************/
+
+        /* gluster-command-dir */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner, "gluster-command-dir", SBIN_DIR"/",
+                         ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* gluster-params */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner, "gluster-params",
+                         "aux-gfid-mount xlator-option=*-dht.assert-no-child-down=true",
+                         ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* log-file */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner,
+                         "log-file",
+                         DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/${session_owner}:${eSlave}.log",
+                         ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* MountBroker log-file */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner,
+                         "log-file-mbr",
+                         DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/mbr/${session_owner}:${eSlave}.log",
+                         ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* gluster-log-file */
+        runinit_gsyncd_setrx (&runner, conf_path);
+        runner_add_args (&runner,
+                         "gluster-log-file",
+                         DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/${session_owner}:${eSlave}.gluster.log",
+                         ".", NULL);
+        RUN_GSYNCD_CMD;
+
+ out:
+        return ret ? -1 : 0;
+}
+
+static int
+glusterd_create_essential_dir_files (glusterd_volinfo_t *volinfo, dict_t *dict,
+                                     char *slave, char *slave_ip,
+                                     char *slave_vol, char **op_errstr)
+{
+        int                ret              = -1;
+        char              *conf_path        = NULL;
+        char              *statefile        = NULL;
+        char               buf[PATH_MAX]    = "";
+        char               errmsg[PATH_MAX] = "";
+        glusterd_conf_t   *conf             = NULL;
+        struct stat        stbuf            = {0,};
+
+        GF_ASSERT (THIS);
+        conf = THIS->private;
+
+        ret = dict_get_str (dict, "conf_path", &conf_path);
+        if (ret) {
+                snprintf (errmsg, sizeof (errmsg),
+                          "Unable to fetch conf file path.");
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "statefile", &statefile);
+        if (ret) {
+                snprintf (errmsg, sizeof (errmsg),
+                          "Unable to fetch statefile path.");
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                goto out;
+        }
+
+        ret = snprintf (buf, sizeof(buf) - 1, "%s/"GEOREP"/%s_%s_%s",
+                        conf->workdir, volinfo->volname, slave_ip, slave_vol);
+        buf[ret] = '\0';
+        ret = mkdir_p (buf, 0777, _gf_true);
+        if (ret) {
+                snprintf (errmsg, sizeof (errmsg), "Unable to create %s"
+                          ". Error : %s", buf, strerror (errno));
+                *op_errstr = gf_strdup (errmsg);
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                goto out;
+        }
+
+        ret = snprintf (buf, PATH_MAX, DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"/%s",
+                        volinfo->volname);
+        buf[ret] = '\0';
+        ret = mkdir_p (buf, 0777, _gf_true);
+        if (ret) {
+                snprintf (errmsg, sizeof (errmsg), "Unable to create %s"
+                          ". Error : %s", buf, strerror (errno));
+                *op_errstr = gf_strdup (errmsg);
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                goto out;
+        }
+
+        ret = lstat (conf_path, &stbuf);
+        if (!ret) {
+                gf_log ("", GF_LOG_DEBUG, "Session already running."
+                        " Not creating config file again.");
+        } else {
+                ret = create_conf_file (conf, conf_path);
+                if (ret || lstat (conf_path, &stbuf)) {
+                        snprintf (errmsg, sizeof (errmsg), "Failed to create"
+                                  " config file(%s).", conf_path);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        goto out;
+                }
+        }
+
+        ret = lstat (statefile, &stbuf);
+        if (!ret) {
+                gf_log ("", GF_LOG_DEBUG, "Session already running."
+                        " Not creating status file again.");
+                goto out;
+        } else {
+                ret = glusterd_create_status_file (volinfo->volname, slave,
+                                                   slave_ip, slave_vol,
+                                                   "Not Started");
+                if (ret || lstat (statefile, &stbuf)) {
+                        snprintf (errmsg, sizeof (errmsg), "Unable to create %s"
+                                  ". Error : %s", statefile, strerror (errno));
+                        *op_errstr = gf_strdup (errmsg);
+                        gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG,"Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_gsync_create (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
+{
+        char                common_pem_file[PATH_MAX] = "";
+        char                errmsg[PATH_MAX]          = "";
+        char                hooks_args[PATH_MAX]      = "";
+        char                uuid_str [64]             = "";
+        char               *host_uuid                 = NULL;
+        char               *slave_ip                  = NULL;
+        char               *slave_vol                 = NULL;
+        char               *arg_buf                   = NULL;
+        char               *volname                   = NULL;
+        char               *slave                     = NULL;
+        int32_t             ret                       = -1;
+        int32_t             is_pem_push               = -1;
+        gf_boolean_t        is_force                  = -1;
+        glusterd_conf_t    *conf                      = NULL;
+        glusterd_volinfo_t *volinfo                   = NULL;
+
+        GF_ASSERT (THIS);
+        conf = THIS->private;
+        GF_ASSERT (conf);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+
+        ret = glusterd_op_gsync_args_get (dict, op_errstr,
+                                          &volname, &slave, &host_uuid);
+        if (ret)
+                goto out;
+
+        snprintf (common_pem_file, sizeof(common_pem_file),
+                  "%s"GLUSTERD_COMMON_PEM_PUB_FILE, conf->workdir);
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Volinfo for %s"
+                        " (master) not found", volname);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "slave_vol", &slave_vol);
+        if (ret) {
+                snprintf (errmsg, sizeof (errmsg),
+                          "Unable to fetch slave volume name.");
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "slave_ip", &slave_ip);
+        if (ret) {
+                snprintf (errmsg, sizeof (errmsg),
+                          "Unable to fetch slave IP.");
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                ret = -1;
+                goto out;
+        }
+
+        is_force = dict_get_str_boolean (dict, "force", _gf_false);
+
+        uuid_utoa_r (MY_UUID, uuid_str);
+        if (!strcmp (uuid_str, host_uuid)) {
+                ret = dict_get_int32 (dict, "push_pem", &is_pem_push);
+                if (!ret && is_pem_push) {
+                        gf_log ("", GF_LOG_DEBUG, "Trying to setup"
+                                " pem files in slave");
+                        is_pem_push = 1;
+                } else
+                        is_pem_push = 0;
+
+                snprintf(hooks_args, sizeof(hooks_args),
+                         "is_push_pem=%d pub_file=%s slave_ip=%s",
+                         is_pem_push, common_pem_file, slave_ip);
+
+        } else
+                snprintf(hooks_args, sizeof(hooks_args),
+                         "This argument will stop the hooks script");
+
+        arg_buf = gf_strdup (hooks_args);
+        if (!arg_buf) {
+                gf_log ("", GF_LOG_ERROR, "Failed to"
+                        " gf_strdup");
+                if (is_force) {
+                        ret = 0;
+                        goto create_essentials;
+                }
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_str (dict, "hooks_args", arg_buf);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Failed to set"
+                        " hooks_args in dict.");
+                if (is_force) {
+                        ret = 0;
+                        goto create_essentials;
+                }
+                goto out;
+        }
+
+create_essentials:
+
+        ret = glusterd_create_essential_dir_files (volinfo, dict, slave,
+                                                   slave_ip, slave_vol,
+                                                   op_errstr);
+        if (ret)
+                goto out;
+
+        ret = glusterd_store_slave_in_info (volinfo, slave,
+                                            host_uuid, op_errstr,
+                                            is_force);
+        if (ret) {
+                snprintf (errmsg, sizeof (errmsg), "Unable to store"
+                          " slave info.");
+                gf_log ("", GF_LOG_ERROR, "%s", errmsg);
+                goto out;
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG,"Returning %d", ret);
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-handler.c b/xlators/mgmt/glusterd/src/glusterd-handler.c
index 4d9a662b3..71d076624 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handler.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handler.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -36,17 +26,21 @@
 #include "compat.h"
 #include "compat-errno.h"
 #include "statedump.h"
+#include "run.h"
 #include "glusterd-mem-types.h"
 #include "glusterd.h"
 #include "glusterd-sm.h"
 #include "glusterd-op-sm.h"
 #include "glusterd-utils.h"
 #include "glusterd-store.h"
+#include "glusterd-locks.h"
 
-#include "glusterd1.h"
-#include "cli1.h"
-#include "rpc-clnt.h"
 #include "glusterd1-xdr.h"
+#include "cli1-xdr.h"
+#include "xdr-generic.h"
+#include "rpc-clnt.h"
+#include "glusterd-volgen.h"
+#include "glusterd-mountbroker.h"
 
 #include <sys/resource.h>
 #include <inttypes.h>
@@ -54,143 +48,67 @@
 #include "defaults.c"
 #include "common-utils.h"
 
-static int
-glusterd_friend_find_by_hostname (const char *hoststr,
-                                  glusterd_peerinfo_t  **peerinfo)
-{
-        int                     ret = -1;
-        glusterd_conf_t         *priv = NULL;
-        glusterd_peerinfo_t     *entry = NULL;
-        glusterd_peer_hostname_t *name = NULL;
-        struct addrinfo         *addr, *p;
-        char                    *host = NULL;
-        struct sockaddr_in6     *s6 = NULL;
-        struct sockaddr_in      *s4 = NULL;
-        struct in_addr          *in_addr = NULL;
-        char                    hname[1024] = {0,};
-
-        GF_ASSERT (hoststr);
-        GF_ASSERT (peerinfo);
-
-        *peerinfo = NULL;
-        priv    = THIS->private;
-
-        GF_ASSERT (priv);
-
-        list_for_each_entry (entry, &priv->peers, uuid_list) {
-                list_for_each_entry (name, &entry->hostnames, hostname_list) {
-                        if (!strncmp (name->hostname, hoststr,
-                                        1024)) {
-
-                        gf_log ("glusterd", GF_LOG_NORMAL,
-                                 "Friend %s found.. state: %d", hoststr,
-                                  entry->state.state);
-                        *peerinfo = entry;
-                        return 0;
-                        }
-                }
-        }
-
-        ret = getaddrinfo(hoststr, NULL, NULL, &addr);
-        if (ret != 0) {
-                gf_log ("", GF_LOG_ERROR, "error in getaddrinfo: %s\n",
-                        gai_strerror(ret));
-                goto out;
-        }
-
-        for (p = addr; p != NULL; p = p->ai_next) {
-                switch (p->ai_family) {
-                        case AF_INET:
-                                s4 = (struct sockaddr_in *) p->ai_addr;
-                                in_addr = &s4->sin_addr;
-                                break;
-                        case AF_INET6:
-                                s6 = (struct sockaddr_in6 *) p->ai_addr;
-                                in_addr =(struct in_addr *) &s6->sin6_addr;
-                                break;
-                       default: ret = -1;
-                                goto out;
-                }
-                host = inet_ntoa(*in_addr);
+#include "globals.h"
+#include "glusterd-syncop.h"
 
-                ret = getnameinfo (p->ai_addr, p->ai_addrlen, hname,
-                                   1024, NULL, 0, 0);
-                if (ret)
-                        goto out;
+#ifdef HAVE_BD_XLATOR
+#include <lvm2app.h>
+#endif
 
-                list_for_each_entry (entry, &priv->peers, uuid_list) {
-                        list_for_each_entry (name, &entry->hostnames,
-                                             hostname_list) {
-                                if (!strncmp (name->hostname, host,
-                                    1024) || !strncmp (name->hostname,hname,
-                                    1024)) {
-                                        gf_log ("glusterd", GF_LOG_NORMAL,
-                                                "Friend %s found.. state: %d", 
-                                                hoststr, entry->state.state);
-                                        *peerinfo = entry;
-                                        freeaddrinfo (addr);
-                                        return 0;
-                                }
-                        } 
-                } 
-        }
+extern uuid_t global_txn_id;
 
-out:
-        if (addr)
-                freeaddrinfo (addr); 
-        return -1;
+int glusterd_big_locked_notify (struct rpc_clnt *rpc, void *mydata,
+                                rpc_clnt_event_t event,
+                                void *data, rpc_clnt_notify_t notify_fn)
+{
+        glusterd_conf_t *priv = THIS->private;
+        int             ret   = -1;
+        synclock_lock (&priv->big_lock);
+        ret = notify_fn (rpc, mydata, event, data);
+        synclock_unlock (&priv->big_lock);
+        return ret;
 }
 
-static int
-glusterd_friend_find_by_uuid (uuid_t uuid,
-                              glusterd_peerinfo_t  **peerinfo)
+int glusterd_big_locked_handler (rpcsvc_request_t *req, rpcsvc_actor actor_fn)
 {
-        int                     ret = -1;
-        glusterd_conf_t         *priv = NULL;
-        glusterd_peerinfo_t     *entry = NULL;
-
-        GF_ASSERT (peerinfo);
+        glusterd_conf_t *priv = THIS->private;
+        int             ret   = -1;
 
-        *peerinfo = NULL;
-        priv    = THIS->private;
-
-        GF_ASSERT (priv);
-
-        if (uuid_is_null (uuid))
-                return -1;
-
-        list_for_each_entry (entry, &priv->peers, uuid_list) {
-                if (!uuid_compare (entry->uuid, uuid)) {
-
-                        gf_log ("glusterd", GF_LOG_NORMAL,
-                                 "Friend found.. state: %d",
-                                  entry->state.state);
-                        *peerinfo = entry;
-                        return 0;
-                }
-        }
+        synclock_lock (&priv->big_lock);
+        ret = actor_fn (req);
+        synclock_unlock (&priv->big_lock);
 
         return ret;
 }
 
 static int
 glusterd_handle_friend_req (rpcsvc_request_t *req, uuid_t  uuid,
-                            char *hostname, int port, dict_t *dict)
+                            char *hostname, int port,
+                            gd1_mgmt_friend_req *friend_req)
 {
         int                             ret = -1;
         glusterd_peerinfo_t             *peerinfo = NULL;
         glusterd_friend_sm_event_t      *event = NULL;
         glusterd_friend_req_ctx_t       *ctx = NULL;
+        char                            rhost[UNIX_PATH_MAX + 1] = {0};
+        uuid_t                          friend_uuid = {0};
+        dict_t                          *dict = NULL;
 
+        uuid_parse (uuid_utoa (uuid), friend_uuid);
         if (!port)
-                port = 6969; // TODO: use define values.
+                port = GF_DEFAULT_BASE_PORT;
 
-        ret = glusterd_friend_find (uuid, hostname, &peerinfo);
+        ret = glusterd_remote_hostname_get (req, rhost, sizeof (rhost));
+        ret = glusterd_friend_find (uuid, rhost, &peerinfo);
 
         if (ret) {
-                gf_log ("glusterd", GF_LOG_NORMAL,
-                         "Unable to find peer");
-
+                ret = glusterd_xfer_friend_add_resp (req, hostname, rhost, port,
+                                                     -1, GF_PROBE_UNKNOWN_PEER);
+                if (friend_req->vols.vols_val) {
+                        free (friend_req->vols.vols_val);
+                        friend_req->vols.vols_val = NULL;
+                }
+                goto out;
         }
 
         ret = glusterd_friend_sm_new_event
@@ -215,12 +133,26 @@ glusterd_handle_friend_req (rpcsvc_request_t *req, uuid_t  uuid,
         if (hostname)
                 ctx->hostname = gf_strdup (hostname);
         ctx->req = req;
-        ctx->vols = dict;
 
+        dict = dict_new ();
+        if (!dict) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_unserialize (friend_req->vols.vols_val,
+                                friend_req->vols.vols_len,
+                                &dict);
+
+        if (ret)
+                goto out;
+        else
+                dict->extra_stdfree = friend_req->vols.vols_val;
+
+        ctx->vols = dict;
         event->ctx = ctx;
 
         ret = glusterd_friend_sm_inject_event (event);
-
         if (ret) {
                 gf_log ("glusterd", GF_LOG_ERROR, "Unable to inject event %d, "
                         "ret = %d", event->event, ret);
@@ -233,16 +165,23 @@ out:
         if (0 != ret) {
                 if (ctx && ctx->hostname)
                         GF_FREE (ctx->hostname);
-                if (ctx && ctx->vols)
-                        dict_destroy (ctx->vols);
-                if (ctx)
-                        GF_FREE (ctx);
+                GF_FREE (ctx);
+                if (dict) {
+                        if ((!dict->extra_stdfree) &&
+                            friend_req->vols.vols_val)
+                                free (friend_req->vols.vols_val);
+                        dict_unref (dict);
+                } else {
+                    free (friend_req->vols.vols_val);
+                }
+                GF_FREE (event);
+        } else {
+                if (peerinfo && (0 == peerinfo->connected))
+                        ret = GLUSTERD_CONNECTION_AWAITED;
         }
-
         return ret;
 }
 
-
 static int
 glusterd_handle_unfriend_req (rpcsvc_request_t *req, uuid_t  uuid,
                               char *hostname, int port)
@@ -253,14 +192,17 @@ glusterd_handle_unfriend_req (rpcsvc_request_t *req, uuid_t  uuid,
         glusterd_friend_req_ctx_t       *ctx = NULL;
 
         if (!port)
-                port = 6969; //TODO: use define'd macro
+                port = GF_DEFAULT_BASE_PORT;
 
         ret = glusterd_friend_find (uuid, hostname, &peerinfo);
 
         if (ret) {
-                gf_log ("glusterd", GF_LOG_NORMAL,
-                         "Unable to find peer");
-
+                gf_log ("glusterd", GF_LOG_CRITICAL,
+                        "Received remove-friend from unknown peer %s",
+                        hostname);
+                ret = glusterd_xfer_friend_remove_resp (req, hostname,
+                                                        port);
+                goto out;
         }
 
         ret = glusterd_friend_sm_new_event
@@ -302,8 +244,7 @@ out:
         if (0 != ret) {
                 if (ctx && ctx->hostname)
                         GF_FREE (ctx->hostname);
-                if (ctx)
-                        GF_FREE (ctx);
+                GF_FREE (ctx);
         }
 
         return ret;
@@ -316,13 +257,14 @@ glusterd_add_peer_detail_to_dict (glusterd_peerinfo_t   *peerinfo,
 
         int             ret = -1;
         char            key[256] = {0, };
+        char           *peer_uuid_str = NULL;
 
         GF_ASSERT (peerinfo);
         GF_ASSERT (friends);
 
         snprintf (key, 256, "friend%d.uuid", count);
-        uuid_unparse (peerinfo->uuid, peerinfo->uuid_str);
-        ret = dict_set_str (friends, key, peerinfo->uuid_str);
+        peer_uuid_str = gd_peer_uuid_str (peerinfo);
+        ret = dict_set_str (friends, key, peer_uuid_str);
         if (ret)
                 goto out;
 
@@ -336,8 +278,14 @@ glusterd_add_peer_detail_to_dict (glusterd_peerinfo_t   *peerinfo,
         if (ret)
                 goto out;
 
+        snprintf (key, 256, "friend%d.stateId", count);
+        ret = dict_set_int32 (friends, key, peerinfo->state.state);
+        if (ret)
+                goto out;
+
         snprintf (key, 256, "friend%d.state", count);
-        ret = dict_set_int32 (friends, key, (int32_t)peerinfo->state.state);
+        ret = dict_set_str (friends, key,
+                    glusterd_friend_sm_state_name_get(peerinfo->state.state));
         if (ret)
                 goto out;
 
@@ -350,9 +298,34 @@ out:
         return ret;
 }
 
+struct args_pack {
+    dict_t *dict;
+    int vol_count;
+    int opt_count;
+};
+
+static int
+_build_option_key (dict_t *d, char *k, data_t *v, void *tmp)
+{
+        char                    reconfig_key[256] = {0, };
+        struct args_pack        *pack             = NULL;
+        int                     ret               = -1;
+
+        pack = tmp;
+        if (strcmp (k, GLUSTERD_GLOBAL_OPT_VERSION) == 0)
+                return 0;
+        snprintf (reconfig_key, 256, "volume%d.option.%s",
+                  pack->vol_count, k);
+        ret = dict_set_str (pack->dict, reconfig_key, v->data);
+        if (0 == ret)
+                pack->opt_count++;
+
+        return 0;
+}
+
 int
 glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo,
-                                    dict_t  *volumes, int   count)
+                                    dict_t  *volumes, int count)
 {
 
         int                     ret = -1;
@@ -360,10 +333,23 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo,
         glusterd_brickinfo_t    *brickinfo = NULL;
         char                    *buf = NULL;
         int                     i = 1;
+        dict_t                  *dict = NULL;
+        glusterd_conf_t         *priv = NULL;
+        char                    *volume_id_str  = NULL;
+        struct args_pack        pack = {0,};
+        xlator_t                *this = NULL;
+#ifdef HAVE_BD_XLATOR
+        int                     caps = 0;
+#endif
 
         GF_ASSERT (volinfo);
         GF_ASSERT (volumes);
 
+        this = THIS;
+        priv = this->private;
+
+        GF_ASSERT (priv);
+
         snprintf (key, 256, "volume%d.name", count);
         ret = dict_set_str (volumes, key, volinfo->volname);
         if (ret)
@@ -379,18 +365,131 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo,
         if (ret)
                 goto out;
 
+        /* As of now, the snap volumes are also displayed as part of
+           volume info command. So this change is to display whether
+           the volume is original volume or the snap_volume. If
+           displaying of snap volumes in volume info o/p is not needed
+           this should be removed.
+        */
+        snprintf (key, 256, "volume%d.snap_volume", count);
+        ret = dict_set_int32 (volumes, key, volinfo->is_snap_volume);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "failed to set whether "
+                        "the volume is a snap volume or actual volume (%s)",
+                        volinfo->volname);
+                goto out;
+        }
+
         snprintf (key, 256, "volume%d.brick_count", count);
         ret = dict_set_int32 (volumes, key, volinfo->brick_count);
         if (ret)
                 goto out;
 
+        snprintf (key, 256, "volume%d.dist_count", count);
+        ret = dict_set_int32 (volumes, key, volinfo->dist_leaf_count);
+        if (ret)
+                goto out;
+
+        snprintf (key, 256, "volume%d.stripe_count", count);
+        ret = dict_set_int32 (volumes, key, volinfo->stripe_count);
+        if (ret)
+                goto out;
+
+        snprintf (key, 256, "volume%d.replica_count", count);
+        ret = dict_set_int32 (volumes, key, volinfo->replica_count);
+        if (ret)
+                goto out;
+
         snprintf (key, 256, "volume%d.transport", count);
         ret = dict_set_int32 (volumes, key, volinfo->transport_type);
         if (ret)
                 goto out;
 
+        volume_id_str = gf_strdup (uuid_utoa (volinfo->volume_id));
+        if (!volume_id_str)
+                goto out;
+
+        snprintf (key, sizeof (key), "volume%d.volume_id", count);
+        ret = dict_set_dynstr (volumes, key, volume_id_str);
+        if (ret)
+                goto out;
+
+        snprintf (key, 256, "volume%d.rebalance", count);
+        ret = dict_set_int32 (volumes, key, volinfo->rebal.defrag_cmd);
+        if (ret)
+                goto out;
+
+#ifdef HAVE_BD_XLATOR
+        if (volinfo->caps) {
+                caps = 0;
+                snprintf (key, 256, "volume%d.xlator0", count);
+                buf = GF_MALLOC (256, gf_common_mt_char);
+                if (!buf) {
+                        ret = ENOMEM;
+                        goto out;
+                }
+                if (volinfo->caps & CAPS_BD)
+                        snprintf (buf, 256, "BD");
+                ret = dict_set_dynstr (volumes, key, buf);
+                if (ret) {
+                        GF_FREE (buf);
+                        goto out;
+                }
+
+                if (volinfo->caps & CAPS_THIN) {
+                        snprintf (key, 256, "volume%d.xlator0.caps%d", count,
+                                  caps++);
+                        buf = GF_MALLOC (256, gf_common_mt_char);
+                        if (!buf) {
+                                ret = ENOMEM;
+                                goto out;
+                        }
+                        snprintf (buf, 256, "thin");
+                        ret = dict_set_dynstr (volumes, key, buf);
+                        if (ret) {
+                                GF_FREE (buf);
+                                goto out;
+                        }
+                }
+
+                if (volinfo->caps & CAPS_OFFLOAD_COPY) {
+                        snprintf (key, 256, "volume%d.xlator0.caps%d", count,
+                                  caps++);
+                        buf = GF_MALLOC (256, gf_common_mt_char);
+                        if (!buf) {
+                                ret = ENOMEM;
+                                goto out;
+                        }
+                        snprintf (buf, 256, "offload_copy");
+                        ret = dict_set_dynstr (volumes, key, buf);
+                        if (ret) {
+                                GF_FREE (buf);
+                                goto out;
+                        }
+                }
+
+                if (volinfo->caps & CAPS_OFFLOAD_SNAPSHOT) {
+                        snprintf (key, 256, "volume%d.xlator0.caps%d", count,
+                                  caps++);
+                        buf = GF_MALLOC (256, gf_common_mt_char);
+                        if (!buf) {
+                                ret = ENOMEM;
+                                goto out;
+                        }
+                        snprintf (buf, 256, "offload_snapshot");
+                        ret = dict_set_dynstr (volumes, key, buf);
+                        if (ret)  {
+                                GF_FREE (buf);
+                                goto out;
+                        }
+                }
+
+        }
+#endif
+
         list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
                 char    brick[1024] = {0,};
+                char    brick_uuid[64] = {0,};
                 snprintf (key, 256, "volume%d.brick%d", count, i);
                 snprintf (brick, 1024, "%s:%s", brickinfo->hostname,
                           brickinfo->path);
@@ -398,8 +497,42 @@ glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo,
                 ret = dict_set_dynstr (volumes, key, buf);
                 if (ret)
                         goto out;
+                snprintf (key, 256, "volume%d.brick%d.uuid", count, i);
+                snprintf (brick_uuid, 64, "%s", uuid_utoa (brickinfo->uuid));
+                buf = gf_strdup (brick_uuid);
+                if (!buf)
+                        goto out;
+                ret = dict_set_dynstr (volumes, key, buf);
+                if (ret)
+                        goto out;
+
+#ifdef HAVE_BD_XLATOR
+                if (volinfo->caps & CAPS_BD) {
+                        snprintf (key, 256, "volume%d.vg%d", count, i);
+                        snprintf (brick, 1024, "%s", brickinfo->vg);
+                        buf = gf_strdup (brick);
+                        ret = dict_set_dynstr (volumes, key, buf);
+                        if (ret)
+                                goto out;
+                }
+#endif
                 i++;
         }
+
+        dict = volinfo->dict;
+        if (!dict) {
+                ret = 0;
+                goto out;
+        }
+
+        pack.dict = volumes;
+        pack.vol_count = count;
+        pack.opt_count = 0;
+        dict_foreach (dict, _build_option_key, (void *) &pack);
+        dict_foreach (priv->opts, _build_option_key, &pack);
+
+        snprintf (key, 256, "volume%d.opt_count", pack.vol_count);
+        ret = dict_set_int32 (volumes, key, pack.opt_count);
 out:
         return ret;
 }
@@ -409,13 +542,18 @@ glusterd_friend_find (uuid_t uuid, char *hostname,
                       glusterd_peerinfo_t **peerinfo)
 {
         int     ret = -1;
+        xlator_t *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
 
         if (uuid) {
                 ret = glusterd_friend_find_by_uuid (uuid, peerinfo);
 
                 if (ret) {
-                        gf_log ("glusterd", GF_LOG_NORMAL,
-                                 "Unable to find peer by uuid");
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                 "Unable to find peer by uuid: %s",
+                                 uuid_utoa (uuid));
                 } else {
                         goto out;
                 }
@@ -426,7 +564,7 @@ glusterd_friend_find (uuid_t uuid, char *hostname,
                 ret = glusterd_friend_find_by_hostname (hostname, peerinfo);
 
                 if (ret) {
-                        gf_log ("glusterd", GF_LOG_NORMAL,
+                        gf_log (this->name, GF_LOG_DEBUG,
                                 "Unable to find hostname: %s", hostname);
                 } else {
                         goto out;
@@ -437,649 +575,705 @@ out:
         return ret;
 }
 
-int
-glusterd_handle_cluster_lock (rpcsvc_request_t *req)
+int32_t
+glusterd_op_txn_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx,
+                       char *err_str, size_t err_len)
 {
-        gd1_mgmt_cluster_lock_req       lock_req = {{0},};
-        int32_t                         ret = -1;
-        char                            str[50] = {0,};
-        glusterd_op_lock_ctx_t          *ctx = NULL;
+        int32_t                     ret             = -1;
+        dict_t                     *dict            = NULL;
+        xlator_t                   *this            = NULL;
+        glusterd_conf_t            *priv            = NULL;
+        int32_t                     locked          = 0;
+        char                       *tmp             = NULL;
+        char                       *volname         = NULL;
+        uuid_t                     *txn_id          = NULL;
+        uuid_t                     *originator_uuid = NULL;
+        glusterd_op_info_t          txn_op_info     = {{0},};
+        glusterd_op_sm_event_type_t event_type      = GD_OP_EVENT_NONE;
 
         GF_ASSERT (req);
+        GF_ASSERT ((op > GD_OP_NONE) && (op < GD_OP_MAX));
+        GF_ASSERT (NULL != ctx);
 
-        if (!gd_xdr_to_mgmt_cluster_lock_req (req->msg[0], &lock_req)) {
-                //failed to decode msg;
-                req->rpc_err = GARBAGE_ARGS;
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        dict = ctx;
+
+        /* Generate a transaction-id for this operation and
+         * save it in the dict. This transaction id distinguishes
+         * each transaction, and helps separate opinfos in the
+         * op state machine. */
+        txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t);
+        if (!txn_id)
+                goto out;
+
+        uuid_generate (*txn_id);
+
+        ret = dict_set_bin (dict, "transaction_id",
+                            txn_id, sizeof(*txn_id));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                       "Failed to set transaction id.");
                 goto out;
         }
-        uuid_unparse (lock_req.uuid, str);
 
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received LOCK from uuid: %s", str);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Transaction_id = %s", uuid_utoa (*txn_id));
 
+        /* Save the MY_UUID as the originator_uuid. This originator_uuid
+         * will be used by is_origin_glusterd() to determine if a node
+         * is the originator node for a command. */
+        originator_uuid = GF_CALLOC (1, sizeof(uuid_t),
+                                     gf_common_mt_uuid_t);
+        if (!originator_uuid) {
+                ret = -1;
+                goto out;
+        }
 
-        ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_op_lock_ctx_t);
+        uuid_copy (*originator_uuid, MY_UUID);
+        ret = dict_set_bin (dict, "originator_uuid",
+                            originator_uuid, sizeof (uuid_t));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set originator uuid.");
+                goto out;
+        }
 
-        if (!ctx) {
-                //respond here
-                return -1;
+        /* Based on the op_version, acquire a cluster or mgmt_v3 lock */
+        if (priv->op_version < 3) {
+                ret = glusterd_lock (MY_UUID);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to acquire lock on localhost, ret: %d",
+                                ret);
+                        snprintf (err_str, err_len,
+                                  "Another transaction is in progress. "
+                                  "Please try again after sometime.");
+                        goto out;
+                }
+        } else {
+                /* If no volname is given as a part of the command, locks will
+                 * not be held */
+                ret = dict_get_str (dict, "volname", &tmp);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Failed to get volume "
+                                "name");
+                        goto local_locking_done;
+                } else {
+                        /* Use a copy of volname, as cli response will be
+                         * sent before the unlock, and the volname in the
+                         * dict, might be removed */
+                        volname = gf_strdup (tmp);
+                        if (!volname)
+                                goto out;
+                }
+
+                ret = glusterd_mgmt_v3_lock (volname, MY_UUID, "vol");
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to acquire lock for %s", volname);
+                        snprintf (err_str, err_len,
+                                  "Another transaction is in progress for %s. "
+                                  "Please try again after sometime.", volname);
+                        goto out;
+                }
         }
 
-        uuid_copy (ctx->uuid, lock_req.uuid);
-        ctx->req = req;
+        locked = 1;
+        gf_log (this->name, GF_LOG_DEBUG, "Acquired lock on localhost");
 
-        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_LOCK, ctx);
+local_locking_done:
+
+        /* If no volname is given as a part of the command, locks will
+         * not be held, hence sending stage event. */
+        if (volname)
+                event_type = GD_OP_EVENT_START_LOCK;
+        else {
+                txn_op_info.state.state = GD_OP_STATE_LOCK_SENT;
+                event_type = GD_OP_EVENT_ALL_ACC;
+        }
+
+        /* Save opinfo for this transaction with the transaction id */
+        glusterd_txn_opinfo_init (&txn_op_info, NULL, &op, ctx, req);
+
+        ret = glusterd_set_txn_opinfo (txn_id, &txn_op_info);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to set transaction's opinfo");
+                if (ctx)
+                        dict_unref (ctx);
+                goto out;
+        }
+
+        ret = glusterd_op_sm_inject_event (event_type, txn_id, ctx);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to acquire cluster"
+                        " lock.");
+                goto out;
+        }
 
 out:
-        gf_log ("", GF_LOG_NORMAL, "Returning %d", ret);
+        if (locked && ret) {
+                /* Based on the op-version, we release the
+                 * cluster or mgmt_v3 lock */
+                if (priv->op_version < 3)
+                        glusterd_unlock (MY_UUID);
+                else {
+                        ret = glusterd_mgmt_v3_unlock (volname, MY_UUID,
+                                                       "vol");
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to release lock for %s",
+                                        volname);
+                        ret = -1;
+                }
+        }
+
+        if (volname)
+                GF_FREE (volname);
 
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
 int
-glusterd_handle_stage_op (rpcsvc_request_t *req)
+__glusterd_handle_cluster_lock (rpcsvc_request_t *req)
 {
-        int32_t                         ret = -1;
-        char                            str[50] = {0,};
-        gd1_mgmt_stage_op_req           stage_req = {{0,}};
-        glusterd_op_stage_ctx_t         *ctx = NULL;
+        dict_t                         *op_ctx      = NULL;
+        int32_t                         ret         = -1;
+        gd1_mgmt_cluster_lock_req       lock_req    = {{0},};
+        glusterd_op_lock_ctx_t         *ctx         = NULL;
+        glusterd_op_t                   op          = GD_OP_EVENT_LOCK;
+        glusterd_peerinfo_t            *peerinfo    = NULL;
+        glusterd_op_info_t              txn_op_info = {{0},};
+        uuid_t                         *txn_id      = &global_txn_id;
+        xlator_t                       *this        = NULL;
 
+        this = THIS;
+        GF_ASSERT (this);
         GF_ASSERT (req);
 
-        if (!gd_xdr_to_mgmt_stage_op_req (req->msg[0], &stage_req)) {
-                //failed to decode msg;
+        ret = xdr_to_generic (req->msg[0], &lock_req,
+                              (xdrproc_t)xdr_gd1_mgmt_cluster_lock_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode lock "
+                        "request received from peer");
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        uuid_unparse (stage_req.uuid, str);
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received stage op from uuid: %s", str);
+        gf_log (this->name, GF_LOG_DEBUG, "Received LOCK from uuid: %s",
+                uuid_utoa (lock_req.uuid));
+
+        if (glusterd_friend_find_by_uuid (lock_req.uuid, &peerinfo)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s doesn't "
+                        "belong to the cluster. Ignoring request.",
+                        uuid_utoa (lock_req.uuid));
+                ret = -1;
+                goto out;
+        }
 
-        ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_op_stage_ctx_t);
+        ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_op_lock_ctx_t);
 
         if (!ctx) {
                 //respond here
                 return -1;
         }
 
-        //CHANGE THIS
-        uuid_copy (ctx->stage_req.uuid, stage_req.uuid);
-        ctx->stage_req.op = stage_req.op;
-        ctx->stage_req.buf.buf_len = stage_req.buf.buf_len;
-        ctx->stage_req.buf.buf_val = GF_CALLOC (1, stage_req.buf.buf_len,
-                                                gf_gld_mt_string);
-        if (!ctx->stage_req.buf.buf_val)
+        uuid_copy (ctx->uuid, lock_req.uuid);
+        ctx->req = req;
+        ctx->dict = NULL;
+
+        op_ctx =  dict_new ();
+        if (!op_ctx) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to set new dict");
                 goto out;
+        }
 
-        memcpy (ctx->stage_req.buf.buf_val, stage_req.buf.buf_val,
-                stage_req.buf.buf_len);
+        glusterd_txn_opinfo_init (&txn_op_info, NULL, &op, op_ctx, req);
 
-        ctx->req   = req;
+        ret = glusterd_set_txn_opinfo (txn_id, &txn_op_info);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to set transaction's opinfo");
+                dict_unref (txn_op_info.op_ctx);
+                goto out;
+        }
 
-        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_OP, ctx);
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_LOCK, txn_id, ctx);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to inject event GD_OP_EVENT_LOCK");
 
 out:
-        if (stage_req.buf.buf_val)
-                free (stage_req.buf.buf_val);//malloced by xdr
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
         return ret;
 }
 
 int
-glusterd_handle_commit_op (rpcsvc_request_t *req)
+glusterd_handle_cluster_lock (rpcsvc_request_t *req)
 {
-        int32_t                         ret = -1;
-        char                            str[50] = {0,};
-        gd1_mgmt_commit_op_req          commit_req = {{0},};
-        glusterd_op_commit_ctx_t        *ctx = NULL;
-
-        GF_ASSERT (req);
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cluster_lock);
+}
 
-        if (!gd_xdr_to_mgmt_commit_op_req (req->msg[0], &commit_req)) {
-                //failed to decode msg;
-                req->rpc_err = GARBAGE_ARGS;
-                goto out;
-        }
+int
+glusterd_req_ctx_create (rpcsvc_request_t *rpc_req,
+                         glusterd_op_t op, uuid_t uuid,
+                         char *buf_val, size_t buf_len,
+                         gf_gld_mem_types_t mem_type,
+                         glusterd_req_ctx_t **req_ctx_out)
+{
+        int                 ret     = -1;
+        char                str[50] = {0,};
+        glusterd_req_ctx_t *req_ctx = NULL;
+        dict_t             *dict    = NULL;
+        xlator_t           *this    = NULL;
 
-        uuid_unparse (commit_req.uuid, str);
+        this = THIS;
+        GF_ASSERT (this);
 
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received commit op from uuid: %s", str);
+        uuid_unparse (uuid, str);
+        gf_log (this->name, GF_LOG_DEBUG, "Received op from uuid %s", str);
 
-        ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_op_commit_ctx_t);
+        dict = dict_new ();
+        if (!dict)
+                goto out;
 
-        if (!ctx) {
-                //respond here
-                return -1;
+        req_ctx = GF_CALLOC (1, sizeof (*req_ctx), mem_type);
+        if (!req_ctx) {
+                goto out;
         }
 
-        ctx->req = req;
-        //CHANGE THIS
-        uuid_copy (ctx->stage_req.uuid, commit_req.uuid);
-        ctx->stage_req.op = commit_req.op;
-        ctx->stage_req.buf.buf_len = commit_req.buf.buf_len;
-        ctx->stage_req.buf.buf_val = GF_CALLOC (1, commit_req.buf.buf_len,
-                                                gf_gld_mt_string);
-        if (!ctx->stage_req.buf.buf_val)
+        uuid_copy (req_ctx->uuid, uuid);
+        req_ctx->op = op;
+        ret = dict_unserialize (buf_val, buf_len, &dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to unserialize the dictionary");
                 goto out;
+        }
 
-        memcpy (ctx->stage_req.buf.buf_val, commit_req.buf.buf_val,
-                commit_req.buf.buf_len);
-
-        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_OP, ctx);
-
+        req_ctx->dict = dict;
+        req_ctx->req = rpc_req;
+        *req_ctx_out = req_ctx;
+        ret = 0;
 out:
-        if (commit_req.buf.buf_val)
-                free (commit_req.buf.buf_val);//malloced by xdr
+        if (ret) {
+                if (dict)
+                        dict_unref (dict);
+                GF_FREE (req_ctx);
+        }
         return ret;
 }
 
 int
-glusterd_handle_cli_probe (rpcsvc_request_t *req)
+__glusterd_handle_stage_op (rpcsvc_request_t *req)
 {
         int32_t                         ret = -1;
-        gf1_cli_probe_req               cli_req = {0,};
+        glusterd_req_ctx_t              *req_ctx = NULL;
+        gd1_mgmt_stage_op_req           op_req = {{0},};
         glusterd_peerinfo_t             *peerinfo = NULL;
+        xlator_t                        *this = NULL;
+        uuid_t                          *txn_id = &global_txn_id;
+        glusterd_op_info_t              txn_op_info = {{0},};
+        glusterd_op_sm_state_info_t     state;
+
+        this = THIS;
+        GF_ASSERT (this);
         GF_ASSERT (req);
 
-        if (!gf_xdr_to_cli_probe_req (req->msg[0], &cli_req)) {
-                //failed to decode msg;
-                gf_log ("", GF_LOG_ERROR, "xdr decoding error");
+        ret = xdr_to_generic (req->msg[0], &op_req,
+                              (xdrproc_t)xdr_gd1_mgmt_stage_op_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode stage "
+                        "request received from peer");
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        gf_cmd_log ("peer probe", " on host %s:%d", cli_req.hostname,
-                    cli_req.port);
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received CLI probe req %s %d",
-                cli_req.hostname, cli_req.port);
-
-        if (!(ret = glusterd_is_local_addr(cli_req.hostname))) {
-                glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_LOCALHOST,
-                                              cli_req.hostname, cli_req.port);
+        if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s doesn't "
+                        "belong to the cluster. Ignoring request.",
+                        uuid_utoa (op_req.uuid));
+                ret = -1;
                 goto out;
         }
-        if (!(ret = glusterd_friend_find_by_hostname(cli_req.hostname,
-                                         &peerinfo))) {
-                if ((peerinfo->state.state != GD_FRIEND_STATE_REQ_RCVD)
-                    || (peerinfo->state.state != GD_FRIEND_STATE_DEFAULT)) {
 
-                        gf_log ("glusterd", GF_LOG_NORMAL, "Probe host %s port %d"
-                               "already a friend", cli_req.hostname, cli_req.port);
-                        glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_FRIEND,
-                                                      cli_req.hostname, cli_req.port);
+        ret = glusterd_req_ctx_create (req, op_req.op, op_req.uuid,
+                                       op_req.buf.buf_val, op_req.buf.buf_len,
+                                       gf_gld_mt_op_stage_ctx_t, &req_ctx);
+        if (ret)
+                goto out;
+
+        ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id);
+
+        gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id));
+
+        /* In cases where there is no volname, the receivers won't have a
+         * transaction opinfo created, as for those operations, the locking
+         * phase where the transaction opinfos are created, won't be called. */
+        ret = glusterd_get_txn_opinfo (txn_id, &txn_op_info);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "No transaction's opinfo set");
+
+                state.state = GD_OP_STATE_LOCKED;
+                glusterd_txn_opinfo_init (&txn_op_info, &state,
+                                          &op_req.op, req_ctx->dict, req);
+
+                ret = glusterd_set_txn_opinfo (txn_id, &txn_op_info);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to set transaction's opinfo");
+                        dict_unref (req_ctx->dict);
                         goto out;
                 }
         }
-        ret = glusterd_probe_begin (req, cli_req.hostname, cli_req.port);
 
-        gf_cmd_log ("peer probe","on host %s:%d %s",cli_req.hostname, cli_req.port,
-                    (ret) ? "FAILED" : "SUCCESS");
-out:
-        if (cli_req.hostname)
-                free (cli_req.hostname);//its malloced by xdr
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_OP,
+                                           txn_id, req_ctx);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to inject event GD_OP_EVENT_STAGE_OP");
+
+ out:
+        free (op_req.buf.buf_val);//malloced by xdr
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
         return ret;
 }
 
 int
-glusterd_handle_cli_deprobe (rpcsvc_request_t *req)
+glusterd_handle_stage_op (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __glusterd_handle_stage_op);
+}
+
+
+int
+__glusterd_handle_commit_op (rpcsvc_request_t *req)
 {
         int32_t                         ret = -1;
-        gf1_cli_probe_req               cli_req = {0,};
+        glusterd_req_ctx_t              *req_ctx = NULL;
+        gd1_mgmt_commit_op_req          op_req = {{0},};
+        glusterd_peerinfo_t             *peerinfo = NULL;
+        xlator_t                        *this = NULL;
+        uuid_t                          *txn_id = &global_txn_id;
 
+        this = THIS;
+        GF_ASSERT (this);
         GF_ASSERT (req);
 
-        if (!gf_xdr_to_cli_probe_req (req->msg[0], &cli_req)) {
-                //failed to decode msg;
+        ret = xdr_to_generic (req->msg[0], &op_req,
+                              (xdrproc_t)xdr_gd1_mgmt_commit_op_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode commit "
+                        "request received from peer");
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received CLI deprobe req");
+        if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s doesn't "
+                        "belong to the cluster. Ignoring request.",
+                        uuid_utoa (op_req.uuid));
+                ret = -1;
+                goto out;
+        }
+
+        //the structures should always be equal
+        GF_ASSERT (sizeof (gd1_mgmt_commit_op_req) == sizeof (gd1_mgmt_stage_op_req));
+        ret = glusterd_req_ctx_create (req, op_req.op, op_req.uuid,
+                                       op_req.buf.buf_val, op_req.buf.buf_len,
+                                       gf_gld_mt_op_commit_ctx_t, &req_ctx);
+        if (ret)
+                goto out;
 
+        ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id);
 
-        ret = glusterd_deprobe_begin (req, cli_req.hostname, cli_req.port);
+        gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id));
+
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_OP,
+                                           txn_id, req_ctx);
 
-        gf_cmd_log ("peer deprobe", "on host %s:%d %s", cli_req.hostname,
-                    cli_req.port, (ret) ? "FAILED" : "SUCCESS");
 out:
-        if (cli_req.hostname)
-                free (cli_req.hostname);//malloced by xdr
+        free (op_req.buf.buf_val);//malloced by xdr
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
         return ret;
 }
 
 int
-glusterd_handle_cli_list_friends (rpcsvc_request_t *req)
+glusterd_handle_commit_op (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __glusterd_handle_commit_op);
+}
+
+int
+__glusterd_handle_cli_probe (rpcsvc_request_t *req)
 {
         int32_t                         ret = -1;
-        gf1_cli_peer_list_req           cli_req = {0,};
-        dict_t                          *dict = NULL;
+        gf_cli_req                  cli_req = {{0,},};
+        glusterd_peerinfo_t       *peerinfo = NULL;
+        gf_boolean_t                run_fsm = _gf_true;
+        xlator_t                      *this = NULL;
+        char                     *bind_name = NULL;
+        dict_t                        *dict = NULL;
+        char                      *hostname = NULL;
+        int                            port = 0;
 
         GF_ASSERT (req);
+        this = THIS;
 
-        if (!gf_xdr_to_cli_peer_list_req (req->msg[0], &cli_req)) {
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0)  {
                 //failed to decode msg;
+                gf_log ("", GF_LOG_ERROR, "xdr decoding error");
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received cli list req");
-
         if (cli_req.dict.dict_len) {
-                /* Unserialize the dictionary */
-                dict  = dict_new ();
+                dict = dict_new ();
 
                 ret = dict_unserialize (cli_req.dict.dict_val,
-                                        cli_req.dict.dict_len,
-                                        &dict);
+                                        cli_req.dict.dict_len, &dict);
                 if (ret < 0) {
-                        gf_log ("glusterd", GF_LOG_ERROR,
-                                "failed to "
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
                                 "unserialize req-buffer to dictionary");
                         goto out;
-                } else {
-                        dict->extra_stdfree = cli_req.dict.dict_val;
                 }
         }
 
-        ret = glusterd_list_friends (req, dict, cli_req.flags);
-
-out:
-        if (dict)
-                dict_unref (dict);
-        return ret;
-}
-
-int
-glusterd_check_and_rebalance (glusterd_volinfo_t *volinfo, char *dir)
-{
-        int                     ret                = -1;
-        int                     dst_fd             = -1;
-        int                     src_fd             = -1;
-        DIR                    *fd                 = NULL;
-        glusterd_defrag_info_t *defrag             = NULL;
-        struct dirent          *entry              = NULL;
-        struct stat             stbuf              = {0,};
-        struct stat             new_stbuf          = {0,};
-        char                    full_path[1024]    = {0,};
-        char                    tmp_filename[1024] = {0,};
-        char                    value[128]         = {0,};
-
-        defrag = volinfo->defrag;
-        if (!defrag)
+        ret = dict_get_str (dict, "hostname", &hostname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get hostname");
                 goto out;
+        }
 
-
-        /* Fix files at this level */
-        fd = opendir (dir);
-        if (!fd)
+        ret = dict_get_int32 (dict, "port", &port);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get port");
                 goto out;
-        while ((entry = readdir (fd))) {
-                if (!entry)
-                        break;
-
-                if (!strcmp (entry->d_name, ".") || !strcmp (entry->d_name, ".."))
-                        continue;
-
-                snprintf (full_path, 1024, "%s/%s", dir, entry->d_name);
-
-                ret = stat (full_path, &stbuf);
-                if (ret == -1)
-                        continue;
-
-                if (S_ISDIR (stbuf.st_mode)) {
-                        /* Fix the layout of the directory */
-                        getxattr (full_path, "trusted.distribute.fix.layout",
-                                  &value, 128);
-                        continue;
-                }
-                if (S_ISREG (stbuf.st_mode) && ((stbuf.st_mode & 01000) == 01000)) {
-                        /* TODO: run the defrag */
-                        snprintf (tmp_filename, 1024, "%s/.%s.gfs%llu", dir,
-                                  entry->d_name,
-                                  (unsigned long long)stbuf.st_size);
-
-                        dst_fd = creat (tmp_filename, (stbuf.st_mode & ~01000));
-                        if (dst_fd == -1)
-                                continue;
-
-                        src_fd = open (full_path, O_RDONLY);
-                        if (src_fd == -1) {
-                                close (dst_fd);
-                                continue;
-                        }
-
-                        while (1) {
-                                ret = read (src_fd, defrag->databuf, 131072);
-                                if (!ret || (ret < 0)) {
-                                        close (dst_fd);
-                                        close (src_fd);
-                                        break;
-                                }
-                                ret = write (dst_fd, defrag->databuf, ret);
-                                if (ret < 0) {
-                                        close (dst_fd);
-                                        close (src_fd);
-                                        break;
-                                }
-                        }
-
-                        ret = stat (full_path, &new_stbuf);
-                        if (ret < 0)
-                                continue;
-                        if (new_stbuf.st_mtime != stbuf.st_mtime)
-                                continue;
-
-                        ret = rename (tmp_filename, full_path);
-                        if (ret != -1) {
-                                LOCK (&defrag->lock);
-                                {
-                                        defrag->total_files += 1;
-                                        defrag->total_data += stbuf.st_size;
-                                }
-                                UNLOCK (&defrag->lock);
-                        }
-                } else {
-                        LOCK (&defrag->lock);
-                        {
-                                if (S_ISREG (stbuf.st_mode))
-                                        defrag->num_files_lookedup += 1;
-                        }
-                        UNLOCK (&defrag->lock);
-                }
-
-                if (volinfo->defrag_status == GF_DEFRAG_STATUS_STOPED) {
-                        closedir (fd);
-                        goto out;
-                }
         }
-        closedir (fd);
 
-        /* Iterate over directories */
-        fd = opendir (dir);
-        if (!fd)
+        if (glusterd_is_any_volume_in_server_quorum (this) &&
+            !does_gd_meet_server_quorum (this)) {
+                glusterd_xfer_cli_probe_resp (req, -1, GF_PROBE_QUORUM_NOT_MET,
+                                              NULL, hostname, port, dict);
+                gf_log (this->name, GF_LOG_ERROR, "Quorum does not meet, "
+                        "rejecting operation");
+                ret = 0;
                 goto out;
-        while ((entry = readdir (fd))) {
-                if (!entry)
-                        break;
-
-                if (!strcmp (entry->d_name, ".") || !strcmp (entry->d_name, ".."))
-                        continue;
-
-                snprintf (full_path, 1024, "%s/%s", dir, entry->d_name);
-
-                ret = stat (full_path, &stbuf);
-                if (ret == -1)
-                        continue;
-
-                if (S_ISDIR (stbuf.st_mode)) {
-                        /* iterate in subdirectories */
-                        ret = glusterd_check_and_rebalance (volinfo, full_path);
-                        if (ret)
-                                break;
-                }
         }
 
-        closedir (fd);
+        gf_log ("glusterd", GF_LOG_INFO, "Received CLI probe req %s %d",
+                hostname, port);
 
-        if (!entry)
+        if (dict_get_str(this->options,"transport.socket.bind-address",
+                         &bind_name) == 0) {
+                gf_log ("glusterd", GF_LOG_DEBUG,
+                        "only checking probe address vs. bind address");
+                ret = gf_is_same_address (bind_name, hostname);
+        }
+        else {
+                ret = gf_is_local_addr (hostname);
+        }
+        if (ret) {
+                glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_LOCALHOST,
+                                              NULL, hostname, port, dict);
                 ret = 0;
-out:
-        return ret;
-}
+                goto out;
+        }
 
-void *
-glusterd_defrag_start (void *data)
-{
-        glusterd_volinfo_t     *volinfo = data;
-        glusterd_defrag_info_t *defrag  = NULL;
-        char                    cmd_str[1024] = {0,};
-        int                     ret     = -1;
-        struct stat             stbuf   = {0,};
-        char                    value[128] = {0,};
-
-        defrag = volinfo->defrag;
-        if (!defrag)
-                goto out;
-
-        sleep (1);
-        ret = stat (defrag->mount, &stbuf);
-        if ((ret == -1) && (errno == ENOTCONN)) {
-                /* Wait for some more time before starting rebalance */
-                sleep (2);
-                ret = stat (defrag->mount, &stbuf);
-                if (ret == -1) {
-                        volinfo->defrag_status   = GF_DEFRAG_STATUS_FAILED;
-                        volinfo->rebalance_files = 0;
-                        volinfo->rebalance_data  = 0;
-                        volinfo->lookedup_files  = 0;
+        if (!(ret = glusterd_friend_find_by_hostname (hostname, &peerinfo))) {
+                if (strcmp (peerinfo->hostname, hostname) == 0) {
+
+                        gf_log ("glusterd", GF_LOG_DEBUG, "Probe host %s port "
+                                "%d already a peer", hostname, port);
+                        glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_FRIEND,
+                                                      NULL, hostname, port,
+                                                      dict);
                         goto out;
                 }
         }
+        ret = glusterd_probe_begin (req, hostname, port, dict);
 
-        /* Fix the root ('/') first */
-        getxattr (defrag->mount, "trusted.distribute.fix.layout", &value, 128);
-
-        ret = glusterd_check_and_rebalance (volinfo, defrag->mount);
-
-        /* TODO: This should run in a thread, and finish the thread when
-           the task is complete. While defrag is running, keep updating
-           files */
-
-        volinfo->defrag_status   = GF_DEFRAG_STATUS_COMPLETE;
-        volinfo->rebalance_files = defrag->total_files;
-        volinfo->rebalance_data  = defrag->total_data;
-        volinfo->lookedup_files  = defrag->num_files_lookedup;
-out:
-        if (defrag) {
-                gf_log ("defrag", GF_LOG_NORMAL, "defrag on %s complete",
-                        defrag->mount);
-
-                snprintf (cmd_str, 1024, "umount -l %s", defrag->mount);
-                ret = system (cmd_str);
-                LOCK_DESTROY (&defrag->lock);
-                GF_FREE (defrag);
+        if (ret == GLUSTERD_CONNECTION_AWAITED) {
+                //fsm should be run after connection establishes
+                run_fsm = _gf_false;
+                ret = 0;
         }
-        volinfo->defrag = NULL;
 
-        return NULL;
-}
-
-int
-glusterd_defrag_stop (glusterd_volinfo_t *volinfo,
-                      gf1_cli_defrag_vol_rsp *rsp)
-{
-        /* TODO: set a variaeble 'stop_defrag' here, it should be checked
-           in defrag loop */
-        if (!volinfo || !volinfo->defrag)
-                goto out;
+out:
+        free (cli_req.dict.dict_val);
 
-        LOCK (&volinfo->defrag->lock);
-        {
-                volinfo->defrag_status = GF_DEFRAG_STATUS_STOPED;
-                rsp->files = volinfo->defrag->total_files;
-                rsp->size = volinfo->defrag->total_data;
+        if (run_fsm) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
         }
-        UNLOCK (&volinfo->defrag->lock);
 
-        rsp->op_ret = 0;
-out:
-        return 0;
+        return ret;
 }
 
 int
-glusterd_defrag_status_get (glusterd_volinfo_t *volinfo,
-                            gf1_cli_defrag_vol_rsp *rsp)
+glusterd_handle_cli_probe (rpcsvc_request_t *req)
 {
-        if (!volinfo)
-                goto out;
-
-        if (volinfo->defrag) {
-                LOCK (&volinfo->defrag->lock);
-                {
-                        rsp->files = volinfo->defrag->total_files;
-                        rsp->size = volinfo->defrag->total_data;
-                        rsp->lookedup_files = volinfo->defrag->num_files_lookedup;
-                }
-                UNLOCK (&volinfo->defrag->lock);
-        } else {
-                rsp->files = volinfo->rebalance_files;
-                rsp->size  = volinfo->rebalance_data;
-                rsp->lookedup_files = volinfo->lookedup_files;
-        }
-
-        rsp->op_errno = volinfo->defrag_status;
-        rsp->op_ret = 0;
-out:
-        return 0;
+        return glusterd_big_locked_handler (req, __glusterd_handle_cli_probe);
 }
 
 int
-glusterd_handle_defrag_volume (rpcsvc_request_t *req)
+__glusterd_handle_cli_deprobe (rpcsvc_request_t *req)
 {
-        int32_t                ret           = -1;
-        gf1_cli_defrag_vol_req cli_req       = {0,};
-        glusterd_conf_t         *priv = NULL;
-        char                   cmd_str[4096] = {0,};
-        glusterd_volinfo_t      *volinfo = NULL;
-        glusterd_defrag_info_t *defrag =  NULL;
-        gf1_cli_defrag_vol_rsp rsp = {0,};
+        int32_t                         ret = -1;
+        gf_cli_req                  cli_req = {{0,},};
+        uuid_t                         uuid = {0};
+        int                        op_errno = 0;
+        xlator_t                      *this = NULL;
+        glusterd_conf_t               *priv = NULL;
+        dict_t                        *dict = NULL;
+        char                      *hostname = NULL;
+        int                            port = 0;
+        int                           flags = 0;
 
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
         GF_ASSERT (req);
 
-        priv    = THIS->private;
-        if (!gf_xdr_to_cli_defrag_vol_req (req->msg[0], &cli_req)) {
+        ret = xdr_to_generic (req->msg[0], &cli_req,
+                              (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        switch (cli_req.cmd) {
-                case GF_DEFRAG_CMD_START:
-                        gf_cmd_log ("Volume rebalance"," on volname: %s "
-                                    "cmd: start, attempted", cli_req.volname);
-                        break;
-                case GF_DEFRAG_CMD_STOP:
-                        gf_cmd_log ("Volume rebalance"," on volname: %s "
-                                    "cmd: stop, attempted", cli_req.volname);
-                        break;
-                default:
-                        break;
+        if (cli_req.dict.dict_len) {
+                dict = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len, &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                "unserialize req-buffer to dictionary");
+                        goto out;
+                }
         }
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received defrag volume on %s",
-                cli_req.volname);
 
-        rsp.volname = cli_req.volname;
-        rsp.op_ret = -1;
-        if (glusterd_volinfo_find(cli_req.volname, &volinfo)) {
-                gf_log ("glusterd", GF_LOG_NORMAL, "Received defrag on invalid"
-                        " volname %s", cli_req.volname);
+        gf_log ("glusterd", GF_LOG_INFO, "Received CLI deprobe req");
+
+        ret = dict_get_str (dict, "hostname", &hostname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get hostname");
                 goto out;
         }
 
-        if (volinfo->status != GLUSTERD_STATUS_STARTED) {
-                gf_log ("glusterd", GF_LOG_NORMAL, "Received defrag on stopped"
-                        " volname %s", cli_req.volname);
+        ret = dict_get_int32 (dict, "port", &port);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get port");
                 goto out;
         }
 
-        switch (cli_req.cmd) {
-        case GF_DEFRAG_CMD_START:
-        {
-                if (volinfo->defrag) {
-                        gf_log ("glusterd", GF_LOG_DEBUG,
-                                "defrag on volume %s already started",
-                                cli_req.volname);
-                        goto out;
-                }
-
-                volinfo->defrag = GF_CALLOC (1, sizeof (glusterd_defrag_info_t),
-                                             gf_gld_mt_defrag_info);
-                if (!volinfo->defrag)
-                        goto out;
+        ret = dict_get_int32 (dict, "flags", &flags);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get flags");
+                goto out;
+        }
 
-                defrag = volinfo->defrag;
+        ret = glusterd_hostname_to_uuid (hostname, uuid);
+        if (ret) {
+                op_errno = GF_DEPROBE_NOT_FRIEND;
+                goto out;
+        }
 
-                LOCK_INIT (&defrag->lock);
-                snprintf (defrag->mount, 1024, "%s/mount/%s",
-                          priv->workdir, cli_req.volname);
-                /* Create a directory, mount glusterfs over it, start glusterfs-defrag */
-                snprintf (cmd_str, 4096, "mkdir -p %s", defrag->mount);
-                ret = system (cmd_str);
+        if (!uuid_compare (uuid, MY_UUID)) {
+                op_errno = GF_DEPROBE_LOCALHOST;
+                ret = -1;
+                goto out;
+        }
 
-                if (ret) {
-                        gf_log("glusterd", GF_LOG_DEBUG, "command: %s failed", cmd_str);
-                        goto out;
+        if (!(flags & GF_CLI_FLAG_OP_FORCE)) {
+                if (!uuid_is_null (uuid)) {
+                        /* Check if peers are connected, except peer being detached*/
+                        if (!glusterd_chk_peers_connected_befriended (uuid)) {
+                                ret = -1;
+                                op_errno = GF_DEPROBE_FRIEND_DOWN;
+                                goto out;
+                        }
+                        ret = glusterd_all_volume_cond_check (
+                                                 glusterd_friend_brick_belongs,
+                                                 -1, &uuid);
+                        if (ret) {
+                                op_errno = GF_DEPROBE_BRICK_EXIST;
+                                goto out;
+                        }
                 }
 
-                snprintf (cmd_str, 4096, "%s/sbin/glusterfs -s localhost "
-                          "--volfile-id %s --volume-name %s-quick-read "
-                          "--xlator-option *dht.unhashed-sticky-bit=yes "
-                          "--xlator-option *dht.use-readdirp=yes "
-                          "--xlator-option *dht.lookup-unhashed=yes %s",
-                          GFS_PREFIX, cli_req.volname, cli_req.volname,
-                          defrag->mount);
-                ret = gf_system (cmd_str);
-                if (ret) {
-                        gf_log("glusterd", GF_LOG_DEBUG, "command: %s failed", cmd_str);
+                if (glusterd_is_any_volume_in_server_quorum (this) &&
+                    !does_gd_meet_server_quorum (this)) {
+                        gf_log (this->name, GF_LOG_ERROR, "Quorum does not "
+                                "meet, rejecting operation");
+                        ret = -1;
+                        op_errno = GF_DEPROBE_QUORUM_NOT_MET;
                         goto out;
                 }
-
-                volinfo->defrag_status = GF_DEFRAG_STATUS_STARTED;
-                rsp.op_ret = 0;
-
-                ret = pthread_create (&defrag->th, NULL, glusterd_defrag_start,
-                                      volinfo);
-                if (ret) {
-                        snprintf (cmd_str, 1024, "umount -l %s", defrag->mount);
-                        ret = system (cmd_str);
-                        rsp.op_ret = -1;
-                }
-                break;
-        }
-        case GF_DEFRAG_CMD_STOP:
-                ret = glusterd_defrag_stop (volinfo, &rsp);
-                break;
-        case GF_DEFRAG_CMD_STATUS:
-                ret = glusterd_defrag_status_get (volinfo, &rsp);
-                break;
-        default:
-                break;
         }
-        if (ret)
-                gf_log("glusterd", GF_LOG_DEBUG, "command: %s failed",cmd_str);
 
-        if (cli_req.cmd != GF_DEFRAG_CMD_STATUS) {
-                gf_cmd_log ("volume rebalance"," on volname: %s %d %s",
-                            cli_req.volname,
-                            cli_req.cmd, ((ret)?"FAILED":"SUCCESS"));
+        if (!uuid_is_null (uuid)) {
+                ret = glusterd_deprobe_begin (req, hostname, port, uuid, dict);
+        } else {
+                ret = glusterd_deprobe_begin (req, hostname, port, NULL, dict);
         }
 
 out:
+        free (cli_req.dict.dict_val);
+
+        if (ret) {
+                ret = glusterd_xfer_cli_deprobe_resp (req, ret, op_errno, NULL,
+                                                      hostname, dict);
+        }
+
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
 
-        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gf_xdr_serialize_cli_defrag_vol_rsp);
-        if (cli_req.volname)
-                free (cli_req.volname);//malloced by xdr
         return ret;
 }
 
 int
-glusterd_handle_cli_get_volume (rpcsvc_request_t *req)
+glusterd_handle_cli_deprobe (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __glusterd_handle_cli_deprobe);
+}
+
+int
+__glusterd_handle_cli_list_friends (rpcsvc_request_t *req)
 {
         int32_t                         ret = -1;
-        gf1_cli_get_vol_req             cli_req = {0,};
+        gf1_cli_peer_list_req           cli_req = {0,};
         dict_t                          *dict = NULL;
 
         GF_ASSERT (req);
 
-        if (!gf_xdr_to_cli_get_vol_req (req->msg[0], &cli_req)) {
+        ret = xdr_to_generic (req->msg[0], &cli_req,
+                              (xdrproc_t)xdr_gf1_cli_peer_list_req);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received get vol req");
+        gf_log ("glusterd", GF_LOG_INFO, "Received cli list req");
 
         if (cli_req.dict.dict_len) {
                 /* Unserialize the dictionary */
@@ -1098,61 +1292,50 @@ glusterd_handle_cli_get_volume (rpcsvc_request_t *req)
                 }
         }
 
-        ret = glusterd_get_volumes (req, dict, cli_req.flags);
+        ret = glusterd_list_friends (req, dict, cli_req.flags);
 
 out:
         if (dict)
                 dict_unref (dict);
+
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
         return ret;
 }
 
 int
-glusterd_handle_create_volume (rpcsvc_request_t *req)
-{
-        int32_t                 ret         = -1;
-        gf1_cli_create_vol_req  cli_req     = {0,};
-        dict_t                 *dict        = NULL;
-        glusterd_brickinfo_t   *brickinfo   = NULL;
-        char                   *brick       = NULL;
-        char                   *bricks      = NULL;
-        char                   *volname     = NULL;
-        int                    brick_count = 0;
-        char                   *tmpptr      = NULL;
-        int                    i           = 0;
-        glusterd_peerinfo_t    *peerinfo    = NULL;
-        char                   *brick_list  = NULL;
-        void                   *cli_rsp     = NULL;
-        char                    err_str[1048];
-        gf1_cli_create_vol_rsp  rsp         = {0,};
-        glusterd_conf_t        *priv        = NULL;
-        int                     err_ret     = 0;
-        xlator_t               *this        = NULL;
-        char                   *free_ptr    = NULL;
-        char                   *trans_type  = NULL;
-        uuid_t                  volume_id   = {0,};
-        char                    volid[64]   = {0,};
-
-        GF_ASSERT (req);
+glusterd_handle_cli_list_friends (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_list_friends);
+}
 
-        this = THIS;
-        GF_ASSERT(this);
+int
+__glusterd_handle_cli_get_volume (rpcsvc_request_t *req)
+{
+        int32_t                         ret = -1;
+        gf_cli_req                      cli_req = {{0,}};
+        dict_t                          *dict = NULL;
+        int32_t                         flags = 0;
 
-        priv = this->private;
+        GF_ASSERT (req);
 
-        if (!gf_xdr_to_cli_create_vol_req (req->msg[0], &cli_req)) {
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received create volume req");
+        gf_log ("glusterd", GF_LOG_INFO, "Received get vol req");
 
-        if (cli_req.bricks.bricks_len) {
+        if (cli_req.dict.dict_len) {
                 /* Unserialize the dictionary */
                 dict  = dict_new ();
 
-                ret = dict_unserialize (cli_req.bricks.bricks_val,
-                                        cli_req.bricks.bricks_len,
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
                                         &dict);
                 if (ret < 0) {
                         gf_log ("glusterd", GF_LOG_ERROR,
@@ -1160,526 +1343,413 @@ glusterd_handle_create_volume (rpcsvc_request_t *req)
                                 "unserialize req-buffer to dictionary");
                         goto out;
                 } else {
-                        dict->extra_stdfree = cli_req.bricks.bricks_val;
+                        dict->extra_stdfree = cli_req.dict.dict_val;
                 }
         }
 
-        ret = dict_get_str (dict, "volname", &volname);
-
+        ret = dict_get_int32 (dict, "flags", &flags);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                gf_log (THIS->name, GF_LOG_ERROR, "failed to get flags");
                 goto out;
         }
-        gf_cmd_log ("Volume create", "on volname: %s attempted", volname);
-
-        if ((ret = glusterd_check_volume_exists (volname))) {
-                snprintf(err_str, 1048, "Volname %s already exists",
-                         volname);
-                gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str);
-                err_ret = 1;
-                goto out;
-        }
-
-        ret = dict_get_int32 (dict, "count", &brick_count);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get count");
-                goto out;
-        }
-
-        ret = dict_get_str (dict, "transport", &trans_type);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get transport-type");
-                goto out;
-        }
-        ret = dict_get_str (dict, "bricks", &bricks);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get bricks");
-                goto out;
-        }
-
-        uuid_generate (volume_id);
-        uuid_unparse (volume_id, volid);
-        free_ptr = gf_strdup (volid);
-        ret = dict_set_dynstr (dict, "volume-id", free_ptr);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "unable to set volume-id");
-                goto out;
-        }
-        free_ptr = NULL;
-
-        if (bricks) {
-                brick_list = gf_strdup (bricks);
-                free_ptr = brick_list;
-        }
-
-        gf_cmd_log ("Volume create", "on volname: %s type:%s count:%d bricks:%s",
-                    cli_req.volname, ((cli_req.type == 0)? "DEFAULT":
-                    ((cli_req.type == 1)? "STRIPE":"REPLICATE")), cli_req.count,
-                    bricks);
-
-        while ( i < brick_count) {
-                i++;
-                brick= strtok_r (brick_list, " \n", &tmpptr);
-                brick_list = tmpptr;
-                if (brickinfo)
-                        glusterd_brickinfo_delete (brickinfo);
-                ret = glusterd_brickinfo_from_brick (brick, &brickinfo);
-                if (ret)
-                        goto out;
-
-                if(!(ret = glusterd_is_local_addr (brickinfo->hostname)))
-                        goto brick_validation;        //localhost, continue without validation
-
-                ret = glusterd_friend_find_by_hostname (brickinfo->hostname,
-                                                        &peerinfo);
-                if (ret) {
-                        snprintf (err_str, 1048, "Host %s not a friend",
-                                  brickinfo->hostname);
-                        gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str);
-                        err_ret = 1;
-                        goto out;
-                }
-                if ((!peerinfo->connected) ||
-                    (peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED)) {
-                        snprintf(err_str, 1048, "Host %s not connected",
-                                 brickinfo->hostname);
-                        gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str);
-                        err_ret = 1;
-                        goto out;
-                }
-brick_validation:
-                err_ret = glusterd_is_exisiting_brick (brickinfo->hostname,
-                                                       brickinfo->path);
-                if (err_ret) {
-                        snprintf(err_str, 1048, "Brick: %s already in use",
-                                 brick);
-                        goto out;
-                }
-        }
-        ret = glusterd_create_volume (req, dict);
 
-        gf_cmd_log ("Volume create", "on volname: %s %s", volname,
-                    ((ret || err_ret) != 0) ? "FAILED": "SUCCESS");
+        ret = glusterd_get_volumes (req, dict, flags);
 
 out:
-        if ((err_ret || ret) && dict)
+        if (dict)
                 dict_unref (dict);
-        if (err_ret) {
-                rsp.op_ret = -1;
-                rsp.op_errno = 0;
-                rsp.volname = "";
-                rsp.op_errstr = err_str;
-                cli_rsp = &rsp;
-                glusterd_submit_reply(req, cli_rsp, NULL, 0, NULL,
-                                      gf_xdr_serialize_cli_create_vol_rsp);
-                if (!glusterd_opinfo_unlock())
-                        gf_log ("glusterd", GF_LOG_ERROR, "Unlock on opinfo"
-                                " failed");
-                ret = 0; //Client response sent, prevent second response
-        }
-
-        if (free_ptr)
-                GF_FREE(free_ptr);
-        if (brickinfo)
-                glusterd_brickinfo_delete (brickinfo);
-        if (cli_req.volname)
-                free (cli_req.volname); // its a malloced by xdr
+
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
         return ret;
 }
 
 int
-glusterd_handle_cli_start_volume (rpcsvc_request_t *req)
+glusterd_handle_cli_get_volume (rpcsvc_request_t *req)
 {
-        int32_t                         ret = -1;
-        gf1_cli_start_vol_req           cli_req = {0,};
-        int32_t                         flags = 0;
-
-        GF_ASSERT (req);
-
-        if (!gf_xdr_to_cli_start_vol_req (req->msg[0], &cli_req)) {
-                //failed to decode msg;
-                req->rpc_err = GARBAGE_ARGS;
-                goto out;
-        }
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received start vol req"
-                "for volume %s", cli_req.volname);
-
-        ret = glusterd_start_volume (req, cli_req.volname, flags);
-
-        gf_cmd_log ("volume start","on volname: %s %s", cli_req.volname,
-                    ((ret == 0) ? "SUCCESS": "FAILED"));
-
-out:
-        if (cli_req.volname)
-                free (cli_req.volname); //its malloced by xdr
-        return ret;
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_get_volume);
 }
 
-
 int
-glusterd_handle_cli_stop_volume (rpcsvc_request_t *req)
+__glusterd_handle_cli_uuid_reset (rpcsvc_request_t *req)
 {
-        int32_t                         ret = -1;
-        gf1_cli_stop_vol_req           cli_req = {0,};
+        int                     ret     = -1;
+        dict_t                  *dict   = NULL;
+        xlator_t                *this   = NULL;
+        glusterd_conf_t         *priv   = NULL;
+        uuid_t                  uuid    = {0};
+        gf_cli_rsp              rsp     = {0,};
+        gf_cli_req              cli_req = {{0,}};
+        char                    msg_str[2048] = {0,};
 
         GF_ASSERT (req);
 
-        if (!gf_xdr_to_cli_stop_vol_req (req->msg[0], &cli_req)) {
+        this = THIS;
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received stop vol req"
-                "for volume %s", cli_req.volname);
+        gf_log ("glusterd", GF_LOG_DEBUG, "Received uuid reset req");
 
-        ret = glusterd_stop_volume (req, cli_req.volname, cli_req.flags);
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
 
-        gf_cmd_log ("Volume stop","on volname: %s %s", cli_req.volname,
-                    ((ret)?"FAILED":"SUCCESS"));
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log ("glusterd", GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (msg_str, sizeof (msg_str), "Unable to decode "
+                                  "the buffer");
+                        goto out;
+                } else {
+                        dict->extra_stdfree = cli_req.dict.dict_val;
+                }
+        }
 
-out:
-        if (cli_req.volname)
-                free (cli_req.volname); //its malloced by xdr
-        return ret;
-}
+        /* In the above section if dict_unserialize is successful, ret is set
+         * to zero.
+         */
+        ret = -1;
+        // Do not allow peer reset if there are any volumes in the cluster
+        if (!list_empty (&priv->volumes)) {
+                snprintf (msg_str, sizeof (msg_str), "volumes are already "
+                          "present in the cluster. Resetting uuid is not "
+                          "allowed");
+                gf_log (this->name, GF_LOG_WARNING, "%s", msg_str);
+                goto out;
+        }
 
-int
-glusterd_handle_cli_delete_volume (rpcsvc_request_t *req)
-{
-        int32_t                         ret = -1;
-        gf1_cli_delete_vol_req          cli_req = {0,};
-        int32_t                         flags = 0;
+        // Do not allow peer reset if trusted storage pool is already formed
+        if (!list_empty (&priv->peers)) {
+                snprintf (msg_str, sizeof (msg_str),"trusted storage pool "
+                          "has been already formed. Please detach this peer "
+                          "from the pool and reset its uuid.");
+                gf_log (this->name, GF_LOG_WARNING, "%s", msg_str);
+                goto out;
+        }
 
-        GF_ASSERT (req);
+        uuid_copy (uuid, priv->uuid);
+        ret = glusterd_uuid_generate_save ();
 
-        if (!gf_xdr_to_cli_delete_vol_req (req->msg[0], &cli_req)) {
-                //failed to decode msg;
-                req->rpc_err = GARBAGE_ARGS;
+        if (!uuid_compare (uuid, MY_UUID)) {
+                snprintf (msg_str, sizeof (msg_str), "old uuid and the new uuid"
+                          " are same. Try gluster peer reset again");
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg_str);
+                ret = -1;
                 goto out;
         }
-        gf_cmd_log ("Volume delete","on volname: %s attempted", cli_req.volname);
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received delete vol req"
-                "for volume %s", cli_req.volname);
 
-        ret = glusterd_delete_volume (req, cli_req.volname, flags);
+out:
+        if (ret) {
+                rsp.op_ret = -1;
+                if (msg_str[0] == '\0')
+                        snprintf (msg_str, sizeof (msg_str), "Operation "
+                                  "failed");
+                rsp.op_errstr = msg_str;
+                ret = 0;
+        } else {
+                rsp.op_errstr = "";
+        }
 
-        gf_cmd_log ("Volume delete", "on volname: %s %s", cli_req.volname,
-                   ((ret) ? "FAILED" : "SUCCESS"));
+        glusterd_to_cli (req, &rsp, NULL, 0, NULL,
+                         (xdrproc_t)xdr_gf_cli_rsp, dict);
 
-out:
-        if (cli_req.volname)
-                free (cli_req.volname); //its malloced by xdr
         return ret;
 }
 
 int
-glusterd_handle_add_brick (rpcsvc_request_t *req)
+glusterd_handle_cli_uuid_reset (rpcsvc_request_t *req)
 {
-        int32_t                         ret = -1;
-        gf1_cli_add_brick_req          cli_req = {0,};
-        dict_t                          *dict = NULL;
-        glusterd_brickinfo_t            *brickinfo = NULL;
-        char                            *brick = NULL;
-        char                            *bricks = NULL;
-        char                            *volname = NULL;
-        int                             brick_count = 0;
-        char                            *tmpptr = NULL;
-        int                             i = 0;
-        glusterd_peerinfo_t             *peerinfo = NULL;
-        char                            *brick_list = NULL;
-        void                            *cli_rsp = NULL;
-        char                            err_str[1048];
-        gf1_cli_add_brick_rsp           rsp = {0,};
-        glusterd_volinfo_t              *volinfo = NULL;
-        int32_t                         err_ret = 0;
-        glusterd_conf_t                 *priv = NULL;
-        xlator_t                        *this = NULL;
-        char                            *free_ptr = NULL;
-
-        this = THIS;
-        GF_ASSERT(this);
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_uuid_reset);
+}
 
-        priv = this->private;
+int
+__glusterd_handle_cli_uuid_get (rpcsvc_request_t *req)
+{
+        int                     ret         = -1;
+        dict_t                  *dict       = NULL;
+        dict_t                  *rsp_dict   = NULL;
+        xlator_t                *this       = NULL;
+        glusterd_conf_t         *priv       = NULL;
+        gf_cli_rsp              rsp         = {0,};
+        gf_cli_req              cli_req     = {{0,}};
+        char                    msg_str[2048] = {0,};
+        char                    uuid_str[64] = {0,};
 
         GF_ASSERT (req);
 
-        if (!gf_xdr_to_cli_add_brick_req (req->msg[0], &cli_req)) {
-                //failed to decode msg;
+        this = THIS;
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        gf_cmd_log ("Volume add-brick", "on volname: %s attempted",
-                    cli_req.volname);
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received add brick req");
+        gf_log ("glusterd", GF_LOG_DEBUG, "Received uuid get req");
 
-        if (cli_req.bricks.bricks_len) {
-                /* Unserialize the dictionary */
+        if (cli_req.dict.dict_len) {
                 dict  = dict_new ();
+                if (!dict) {
+                        ret = -1;
+                        goto out;
+                }
 
-                ret = dict_unserialize (cli_req.bricks.bricks_val,
-                                        cli_req.bricks.bricks_len,
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
                                         &dict);
                 if (ret < 0) {
                         gf_log ("glusterd", GF_LOG_ERROR,
                                 "failed to "
                                 "unserialize req-buffer to dictionary");
+                        snprintf (msg_str, sizeof (msg_str), "Unable to decode "
+                                  "the buffer");
                         goto out;
+
                 } else {
-                        dict->extra_stdfree = cli_req.bricks.bricks_val;
+                        dict->extra_stdfree = cli_req.dict.dict_val;
+
                 }
         }
 
-        ret = dict_get_str (dict, "volname", &volname);
-
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                ret = -1;
                 goto out;
         }
 
-        if (!(ret = glusterd_check_volume_exists (volname))) {
-                snprintf(err_str, 1048, "Volname %s does not exist",
-                         volname);
-                gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str);
-                err_ret = -1;
+        uuid_utoa_r (MY_UUID, uuid_str);
+        ret = dict_set_str (rsp_dict, "uuid", uuid_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set uuid in "
+                        "dictionary.");
                 goto out;
         }
 
-        ret = dict_get_int32 (dict, "count", &brick_count);
+        ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val,
+                                           &rsp.dict.dict_len);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get count");
+                gf_log (this->name, GF_LOG_ERROR, "Failed to serialize "
+                        "dictionary.");
                 goto out;
         }
+        ret = 0;
+out:
+        if (ret) {
+                rsp.op_ret = -1;
+                if (msg_str[0] == '\0')
+                        snprintf (msg_str, sizeof (msg_str), "Operation "
+                                  "failed");
+                rsp.op_errstr = msg_str;
 
-        if (!(ret = glusterd_volinfo_find (volname, &volinfo))) {
-                if (volinfo->type == GF_CLUSTER_TYPE_NONE)
-                        goto brick_val;
-                if (!brick_count || !volinfo->sub_count)
-                        goto brick_val;
-
-                if (volinfo->brick_count < volinfo->sub_count) {
-                        if ((volinfo->sub_count - volinfo->brick_count) == brick_count)
-                                goto brick_val;
-                }
-
-                if ((brick_count % volinfo->sub_count) != 0) {
-                        snprintf(err_str, 2048, "Incorrect number of bricks"
-                                " supplied %d for type %s with count %d",
-                                brick_count, (volinfo->type == 1)? "STRIPE":
-                                "REPLICATE", volinfo->sub_count);
-                        gf_log("glusterd", GF_LOG_ERROR, "%s", err_str);
-                        err_ret = 1;
-                        goto out;
-                }
         } else {
-                gf_log("", GF_LOG_ERROR, "Unable to get volinfo for volname"
-                       " %s", volname);
-                goto out;
-        }
+                rsp.op_errstr = "";
 
-brick_val:
-        ret = dict_get_str (dict, "bricks", &bricks);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get bricks");
-                goto out;
         }
 
-        if (bricks)
-                brick_list = gf_strdup (bricks);
-        if (!brick_list) {
-                gf_log ("", GF_LOG_ERROR, "Out of memory");
-                ret = -1;
-                goto out;
-        } else {
-                free_ptr = brick_list;
-        }
+        glusterd_to_cli (req, &rsp, NULL, 0, NULL,
+                         (xdrproc_t)xdr_gf_cli_rsp, dict);
+
+        return 0;
+}
+int
+glusterd_handle_cli_uuid_get (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_uuid_get);
+}
+
+int
+__glusterd_handle_cli_list_volume (rpcsvc_request_t *req)
+{
+        int                     ret = -1;
+        dict_t                  *dict = NULL;
+        glusterd_conf_t         *priv = NULL;
+        glusterd_volinfo_t      *volinfo = NULL;
+        int                     count = 0;
+        char                    key[1024] = {0,};
+        gf_cli_rsp              rsp = {0,};
+
+        GF_ASSERT (req);
 
-        gf_cmd_log ("Volume add-brick", "volname: %s type %s count:%d bricks:%s"
-                    ,volname, ((volinfo->type == 0)? "DEFAULT" : ((volinfo->type
-                    == 1)? "STRIPE": "REPLICATE")), brick_count, brick_list);
+        priv = THIS->private;
+        GF_ASSERT (priv);
 
-        while ( i < brick_count) {
-                i++;
-                brick= strtok_r (brick_list, " \n", &tmpptr);
-                brick_list = tmpptr;
-                if (brickinfo)
-                        glusterd_brickinfo_delete (brickinfo);
-                ret = glusterd_brickinfo_from_brick (brick, &brickinfo);
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        list_for_each_entry (volinfo, &priv->volumes, vol_list) {
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), "volume%d", count);
+                ret = dict_set_str (dict, key, volinfo->volname);
                 if (ret)
                         goto out;
-                if(!(ret = glusterd_is_local_addr(brickinfo->hostname)))
-                        goto brick_validation;       //localhost, continue without validation
-                ret = glusterd_friend_find_by_hostname(brickinfo->hostname,
-                                                        &peerinfo);
-                if (ret) {
-                        snprintf(err_str, 1048, "Host %s not a friend",
-                                 brickinfo->hostname);
-                        gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str);
-                        err_ret = 1;
-                        goto out;
-                }
-                if ((!peerinfo->connected) ||
-                    (peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED)) {
-                        snprintf(err_str, 1048, "Host %s not connected",
-                                 brickinfo->hostname);
-                        gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str);
-                        err_ret = 1;
-                        goto out;
-                }
-brick_validation:
-                err_ret = glusterd_is_exisiting_brick (brickinfo->hostname,
-                                                       brickinfo->path);
-                if (err_ret) {
-                        snprintf(err_str, 1048, "Brick: %s already in use",
-                                 brick);
-                        goto out;
-                }
+                count++;
         }
 
-        ret = glusterd_add_brick (req, dict);
+        ret = dict_set_int32 (dict, "count", count);
+        if (ret)
+                goto out;
 
-        gf_cmd_log ("Volume add-brick","on volname: %s %s", volname,
-                   ((ret || err_ret) != 0)? "FAILED" : "SUCCESS");
+        ret = dict_allocate_and_serialize (dict, &rsp.dict.dict_val,
+                                           &rsp.dict.dict_len);
+        if (ret)
+                goto out;
+
+        ret = 0;
 
 out:
-        if ((err_ret || ret) && dict)
+        rsp.op_ret = ret;
+        if (ret)
+                rsp.op_errstr = "Error listing volumes";
+        else
+                rsp.op_errstr = "";
+
+        glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gf_cli_rsp);
+        ret = 0;
+
+        if (dict)
                 dict_unref (dict);
-        if (err_ret) {
-                rsp.op_ret = -1;
-                rsp.op_errno = 0;
-                rsp.volname = "";
-                rsp.op_errstr = err_str;
-                cli_rsp = &rsp;
-                glusterd_submit_reply(req, cli_rsp, NULL, 0, NULL,
-                                      gf_xdr_serialize_cli_add_brick_rsp);
-                if (!glusterd_opinfo_unlock())
-                        gf_log ("glusterd", GF_LOG_ERROR, "Unlock on "
-                               "opinfo failed");
 
-                ret = 0; //sent error to cli, prevent second reply
-        }
-        if (brickinfo)
-                glusterd_brickinfo_delete (brickinfo);
-        if (free_ptr)
-                GF_FREE (free_ptr);
-        if (cli_req.volname)
-                free (cli_req.volname); //its malloced by xdr
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
         return ret;
 }
 
 int
-glusterd_handle_replace_brick (rpcsvc_request_t *req)
+glusterd_handle_cli_list_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_list_volume);
+}
+
+int32_t
+glusterd_op_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx,
+                   char *err_str, size_t err_len)
+{
+        int             ret = -1;
+
+        ret = glusterd_op_txn_begin (req, op, ctx, err_str, err_len);
+
+        return ret;
+}
+
+int
+__glusterd_handle_reset_volume (rpcsvc_request_t *req)
 {
         int32_t                         ret = -1;
-        gf1_cli_replace_brick_req          cli_req = {0,};
+        gf_cli_req                      cli_req = {{0,}};
         dict_t                          *dict = NULL;
-        char                            *src_brick = NULL;
-        char                            *dst_brick = NULL;
-        int32_t                         op = 0;
-        char                            operation[8];
+        glusterd_op_t                   cli_op = GD_OP_RESET_VOLUME;
+        char                            *volname = NULL;
+        char                            err_str[2048] = {0,};
+        xlator_t                        *this = NULL;
 
         GF_ASSERT (req);
+        this = THIS;
+        GF_ASSERT (this);
 
-        if (!gf_xdr_to_cli_replace_brick_req (req->msg[0], &cli_req)) {
-                //failed to decode msg;
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                snprintf (err_str, sizeof (err_str), "Failed to decode request "
+                          "received from cli");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        gf_cmd_log ("Volume replace-brick","on volname: %s attempted", cli_req.volname);
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received replace brick req");
-
-        if (cli_req.bricks.bricks_len) {
+        if (cli_req.dict.dict_len) {
                 /* Unserialize the dictionary */
                 dict  = dict_new ();
 
-                ret = dict_unserialize (cli_req.bricks.bricks_val,
-                                        cli_req.bricks.bricks_len,
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
                                         &dict);
                 if (ret < 0) {
-                        gf_log ("glusterd", GF_LOG_ERROR,
-                                "failed to "
-                                "unserialize req-buffer to dictionary");
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                    "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
                         goto out;
                 } else {
-                        dict->extra_stdfree = cli_req.bricks.bricks_val;
+                        dict->extra_stdfree = cli_req.dict.dict_val;
                 }
         }
 
-        ret = dict_get_int32 (dict, "operation", &op);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "dict_get on operation failed");
-                goto out;
-        }
-
-        ret = dict_get_str (dict, "src-brick", &src_brick);
-
+        ret = dict_get_str (dict, "volname", &volname);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get src brick");
+                snprintf (err_str, sizeof (err_str), "Failed to get volume "
+                          "name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
                 goto out;
         }
-        gf_log ("", GF_LOG_DEBUG,
-                "src brick=%s", src_brick);
+        gf_log (this->name, GF_LOG_DEBUG, "Received volume reset request for "
+                "volume %s", volname);
 
-        ret = dict_get_str (dict, "dst-brick", &dst_brick);
+        ret = glusterd_op_begin_synctask (req, GD_OP_RESET_VOLUME, dict);
 
+out:
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get dest brick");
-                goto out;
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
         }
 
-        gf_log ("", GF_LOG_DEBUG,
-                "dst brick=%s", dst_brick);
-
-        switch (op) {
-                case GF_REPLACE_OP_START: strcpy (operation, "start");
-                        break;
-                case GF_REPLACE_OP_COMMIT: strcpy (operation, "commit");
-                        break;
-                case GF_REPLACE_OP_PAUSE:  strcpy (operation, "pause");
-                        break;
-                case GF_REPLACE_OP_ABORT:  strcpy (operation, "abort");
-                        break;
-                case GF_REPLACE_OP_STATUS: strcpy (operation, "status");
-                        break;
-                default:strcpy (operation, "unknown");
-                        break;
-        }
-
-        gf_cmd_log ("Volume replace-brick","volname: %s src_brick:%s"
-                    " dst_brick:%s op:%s",cli_req.volname, src_brick, dst_brick
-                    ,operation);
-
-
-        ret = glusterd_replace_brick (req, dict);
-
-        gf_cmd_log ("Volume replace-brick","on volname: %s %s", cli_req.volname,
-                   (ret) ? "FAILED" : "SUCCESS");
-
-out:
-        if (ret && dict)
-                dict_unref (dict);
-        if (cli_req.volname)
-                free (cli_req.volname);//malloced by xdr
         return ret;
 }
 
 int
-glusterd_handle_set_volume (rpcsvc_request_t *req)
+glusterd_handle_reset_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_reset_volume);
+}
+
+int
+__glusterd_handle_set_volume (rpcsvc_request_t *req)
 {
-        int32_t                           ret = -1;
-        gf1_cli_set_vol_req           cli_req = {0,};
+        int32_t                         ret = -1;
+        gf_cli_req                      cli_req = {{0,}};
         dict_t                          *dict = NULL;
+        glusterd_op_t                   cli_op = GD_OP_SET_VOLUME;
+        char                            *key = NULL;
+        char                            *value = NULL;
+        char                            *volname = NULL;
+        char                            *op_errstr = NULL;
+        gf_boolean_t                    help = _gf_false;
+        char                            err_str[2048] = {0,};
+        xlator_t                        *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
 
         GF_ASSERT (req);
 
-        if (!gf_xdr_to_cli_set_vol_req (req->msg[0], &cli_req)) {
-                //failed to decode msg;
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                snprintf (err_str, sizeof (err_str), "Failed to decode "
+                          "request received from cli");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
@@ -1692,528 +1762,382 @@ glusterd_handle_set_volume (rpcsvc_request_t *req)
                                         cli_req.dict.dict_len,
                                         &dict);
                 if (ret < 0) {
-                        gf_log ("glusterd", GF_LOG_ERROR,
+                        gf_log (this->name, GF_LOG_ERROR,
                                 "failed to "
                                 "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
                         goto out;
+                } else {
+                        dict->extra_stdfree = cli_req.dict.dict_val;
                 }
         }
 
-        ret = glusterd_set_volume (req, dict);
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Failed to get volume "
+                          "name while handling volume set command");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        if (strcmp (volname, "help") == 0 ||
+            strcmp (volname, "help-xml") == 0) {
+                ret = glusterd_volset_help (dict, &op_errstr);
+                help = _gf_true;
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "key1", &key);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Failed to get key while"
+                          " handling volume set for %s", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "value1", &value);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Failed to get value while"
+                          " handling volume set for %s", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+        gf_log (this->name, GF_LOG_DEBUG, "Received volume set request for "
+                "volume %s", volname);
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_SET_VOLUME, dict);
 
 out:
+        if (help)
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req, dict,
+                                                     (op_errstr)? op_errstr:"");
+        else if (ret) {
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
+        }
+        if (op_errstr)
+                GF_FREE (op_errstr);
+
         return ret;
 }
 
 int
-glusterd_handle_remove_brick (rpcsvc_request_t *req)
+glusterd_handle_set_volume (rpcsvc_request_t *req)
 {
-        int32_t                         ret = -1;
-        gf1_cli_remove_brick_req        cli_req = {0,};
-        dict_t                          *dict = NULL;
-        int32_t                         count = 0;
-        char                            *brick = NULL;
-        char                            key[256] = {0,};
-        char                            *brick_list = NULL;
-        int                             i = 1;
-        glusterd_volinfo_t              *volinfo = NULL;
-        glusterd_brickinfo_t            *brickinfo = NULL;
-        int32_t                         pos = 0;
-        int32_t                         sub_volume = 0;
-        int32_t                         sub_volume_start = 0;
-        int32_t                         sub_volume_end = 0;
-        glusterd_brickinfo_t            *tmp = NULL;
-        int32_t                         err_ret = 0;
-        char                            *err_str = NULL;
-        gf1_cli_remove_brick_rsp        rsp = {0,};
-        void                            *cli_rsp = NULL;
-        char                            vol_type[256] = {0,};
+        return glusterd_big_locked_handler (req, __glusterd_handle_set_volume);
+}
+
+int
+__glusterd_handle_sync_volume (rpcsvc_request_t *req)
+{
+        int32_t                          ret     = -1;
+        gf_cli_req                       cli_req = {{0,}};
+        dict_t                           *dict = NULL;
+        gf_cli_rsp                       cli_rsp = {0.};
+        char                             msg[2048] = {0,};
+        char                             *volname = NULL;
+        gf1_cli_sync_volume              flags = 0;
+        char                             *hostname = NULL;
+        xlator_t                         *this = NULL;
 
         GF_ASSERT (req);
+        this = THIS;
+        GF_ASSERT (this);
 
-        if (!gf_xdr_to_cli_remove_brick_req (req->msg[0], &cli_req)) {
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        gf_cmd_log ("Volume remove-brick","on volname: %s attempted",cli_req.volname);
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received rem brick req");
-
-        if (cli_req.bricks.bricks_len) {
+        if (cli_req.dict.dict_len) {
                 /* Unserialize the dictionary */
                 dict  = dict_new ();
 
-                ret = dict_unserialize (cli_req.bricks.bricks_val,
-                                        cli_req.bricks.bricks_len,
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
                                         &dict);
                 if (ret < 0) {
-                        gf_log ("glusterd", GF_LOG_ERROR,
+                        gf_log (this->name, GF_LOG_ERROR,
                                 "failed to "
                                 "unserialize req-buffer to dictionary");
+                        snprintf (msg, sizeof (msg), "Unable to decode the "
+                                  "command");
                         goto out;
                 } else {
-                        dict->extra_stdfree = cli_req.bricks.bricks_val;
+                        dict->extra_stdfree = cli_req.dict.dict_val;
                 }
         }
 
-        ret = dict_get_int32 (dict, "count", &count);
+        ret = dict_get_str (dict, "hostname", &hostname);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get count");
+                snprintf (msg, sizeof (msg), "Failed to get hostname");
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
                 goto out;
         }
 
-        err_str = GF_MALLOC (2048 * sizeof(*err_str),gf_common_mt_char);
-
-        if (!err_str) {
-                gf_log ("",GF_LOG_ERROR,"glusterd_handle_remove_brick: "
-                        "Unable to get memory");
-                ret = -1;
-                goto out;
-        }
-
-        ret = glusterd_volinfo_find (cli_req.volname, &volinfo);
+        ret = dict_get_str (dict, "volname", &volname);
         if (ret) {
-                 snprintf (err_str, 2048, "volname %s not found",
-                          cli_req.volname);
-                 gf_log ("", GF_LOG_ERROR, "%s", err_str);
-                 err_ret = 1;
-                 goto out;
-        }
-
-        if (volinfo->type == GF_CLUSTER_TYPE_REPLICATE)
-                strcpy (vol_type, "replica");
-        else if (volinfo->type == GF_CLUSTER_TYPE_STRIPE)
-                strcpy (vol_type, "stripe");
-        else
-                strcpy (vol_type, "distribute");
-
-        if ((volinfo->type == (GF_CLUSTER_TYPE_REPLICATE ||
-             GF_CLUSTER_TYPE_STRIPE)) &&
-            !(volinfo->brick_count <= volinfo->sub_count)) {
-                if (volinfo->sub_count && (count % volinfo->sub_count != 0)) {
-                        snprintf (err_str, 2048, "Remove brick incorrect"
-                                  " brick count of %d for %s %d",
-                                   count, vol_type, volinfo->sub_count);
-                        gf_log ("", GF_LOG_ERROR, "%s", err_str);
-                        err_ret = 1;
-                        ret = -1;
+                ret = dict_get_int32 (dict, "flags", (int32_t*)&flags);
+                if (ret) {
+                        snprintf (msg, sizeof (msg), "Failed to get volume name"
+                                  " or flags");
+                        gf_log (this->name, GF_LOG_ERROR, "%s", msg);
                         goto out;
                 }
-
         }
 
-        brick_list = GF_MALLOC (120000 * sizeof(*brick_list),gf_common_mt_char);
+        gf_log (this->name, GF_LOG_INFO, "Received volume sync req "
+                "for volume %s", (flags & GF_CLI_SYNC_ALL) ? "all" : volname);
 
-        if (!brick_list) {
-                gf_log ("",GF_LOG_ERROR,"glusterd_handle_remove_brick: "
-                        "Unable to get memory");
+        if (gf_is_local_addr (hostname)) {
                 ret = -1;
+                snprintf (msg, sizeof (msg), "sync from localhost"
+                          " not allowed");
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
                 goto out;
         }
 
-        strcpy (brick_list, " ");
-        while ( i <= count) {
-                snprintf (key, 256, "brick%d", i);
-                ret = dict_get_str (dict, key, &brick);
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to get %s", key);
-                        goto out;
-                }
-                gf_log ("", GF_LOG_DEBUG, "Remove brick count %d brick: %s",
-                        i, brick);
-
-                ret = glusterd_brickinfo_get(brick, volinfo, &brickinfo);
-                if (ret) {
-                        snprintf(err_str, 2048," Incorrect brick %s for volname"
-                                " %s", brick, cli_req.volname);
-                        gf_log ("", GF_LOG_ERROR, "%s", err_str);
-                        err_ret = 1;
-                        goto out;
-                }
-                strcat(brick_list, brick);
-                strcat(brick_list, " ");
-
-                i++;
-                if ((volinfo->type == GF_CLUSTER_TYPE_NONE) ||
-                    (volinfo->brick_count <= volinfo->sub_count))
-                        continue;
- 
-                pos = 0;
-                list_for_each_entry (tmp, &volinfo->bricks, brick_list) {
-
-                        if ((!strcmp (tmp->hostname,brickinfo->hostname)) &&
-                            !strcmp (tmp->path, brickinfo->path)) {
-                                gf_log ("", GF_LOG_NORMAL, "Found brick");
-                                if (!sub_volume && volinfo->sub_count) {
-                                        sub_volume = (pos / volinfo->
-                                                      sub_count) + 1;
-                                        sub_volume_start = volinfo->sub_count *
-                                                           (sub_volume - 1);
-                                        sub_volume_end = (volinfo->sub_count *
-                                                          sub_volume) -1 ;
-                                } else {
-                                        if (pos < sub_volume_start ||
-                                            pos >sub_volume_end) {
-                                                ret = -1;
-                                                snprintf(err_str, 2048,"Bricks"
-                                                         " not from same subvol"
-                                                         " for %s", vol_type);
-                                                gf_log ("",GF_LOG_ERROR,
-                                                        "%s", err_str);
-                                                err_ret = 1;
-                                                goto out;
-                                        }
-                                }
-                                break;
-                        }
-                        pos++;
-                }
-        }
-        gf_cmd_log ("Volume remove-brick","volname: %s count:%d bricks:%s",
-                    cli_req.volname, count, brick_list);
-
-        ret = glusterd_remove_brick (req, dict);
-
-        gf_cmd_log ("Volume remove-brick","on volname: %s %s",cli_req.volname,
-                    (ret) ? "FAILED" : "SUCCESS");
+        ret = glusterd_op_begin_synctask (req, GD_OP_SYNC_VOLUME, dict);
 
 out:
-        if ((ret || err_ret) && dict)
-                dict_unref (dict);
-        if (err_ret) {
-                rsp.op_ret = -1;
-                rsp.op_errno = 0;
-                rsp.volname = "";
-                rsp.op_errstr = err_str;
-                cli_rsp = &rsp;
-                glusterd_submit_reply(req, cli_rsp, NULL, 0, NULL,
-                                      gf_xdr_serialize_cli_remove_brick_rsp);
-                if (!glusterd_opinfo_unlock())
-                        gf_log ("glusterd", GF_LOG_ERROR, "Unlock on "
-                               "opinfo failed");
+        if (ret) {
+                cli_rsp.op_ret = -1;
+                cli_rsp.op_errstr = msg;
+                if (msg[0] == '\0')
+                        snprintf (msg, sizeof (msg), "Operation failed");
+                glusterd_to_cli (req, &cli_rsp, NULL, 0, NULL,
+                                 (xdrproc_t)xdr_gf_cli_rsp, dict);
 
                 ret = 0; //sent error to cli, prevent second reply
-
         }
-        if (brick_list)
-                GF_FREE (brick_list);
-        if (err_str)
-                GF_FREE (err_str);
-        if (cli_req.volname)
-                free (cli_req.volname); //its malloced by xdr
+
         return ret;
 }
 
 int
-glusterd_handle_log_filename (rpcsvc_request_t *req)
+glusterd_handle_sync_volume (rpcsvc_request_t *req)
 {
-        int32_t                   ret     = -1;
-        gf1_cli_log_filename_req  cli_req = {0,};
-        dict_t                   *dict    = NULL;
+        return glusterd_big_locked_handler (req, __glusterd_handle_sync_volume);
+}
 
-        GF_ASSERT (req);
+int
+glusterd_fsm_log_send_resp (rpcsvc_request_t *req, int op_ret,
+                            char *op_errstr, dict_t *dict)
+{
 
-        if (!gf_xdr_to_cli_log_filename_req (req->msg[0], &cli_req)) {
-                //failed to decode msg;
-                req->rpc_err = GARBAGE_ARGS;
-                goto out;
-        }
+        int                             ret = -1;
+        gf1_cli_fsm_log_rsp             rsp = {0};
 
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received log filename req "
-                "for volume %s", cli_req.volname);
+        GF_ASSERT (req);
+        GF_ASSERT (op_errstr);
 
-        dict = dict_new ();
-        if (!dict)
-                goto out;
+        rsp.op_ret = op_ret;
+        rsp.op_errstr = op_errstr;
+        if (rsp.op_ret == 0)
+                ret = dict_allocate_and_serialize (dict, &rsp.fsm_log.fsm_log_val,
+                                                &rsp.fsm_log.fsm_log_len);
 
-        ret = dict_set_dynmstr (dict, "volname", cli_req.volname);
-        if (ret)
-                goto out;
-        ret = dict_set_dynmstr (dict, "brick", cli_req.brick);
-        if (ret)
-                goto out;
-        ret = dict_set_dynmstr (dict, "path", cli_req.path);
-        if (ret)
-                goto out;
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gf1_cli_fsm_log_rsp);
+        GF_FREE (rsp.fsm_log.fsm_log_val);
 
-        ret = glusterd_log_filename (req, dict);
+        gf_log ("glusterd", GF_LOG_DEBUG, "Responded, ret: %d", ret);
 
-out:
-        if (ret && dict)
-                dict_unref (dict);
-        return ret;
+        return 0;
 }
 
 int
-glusterd_handle_log_locate (rpcsvc_request_t *req)
+__glusterd_handle_fsm_log (rpcsvc_request_t *req)
 {
-        int32_t                 ret     = -1;
-        gf1_cli_log_locate_req  cli_req = {0,};
-        gf1_cli_log_locate_rsp  rsp     = {0,};
-        glusterd_conf_t        *priv = NULL;
-        glusterd_volinfo_t     *volinfo = NULL;
-        glusterd_brickinfo_t   *brickinfo = NULL;
-        char                    tmp_str[PATH_MAX] = {0,};
+        int32_t                         ret = -1;
+        gf1_cli_fsm_log_req             cli_req = {0,};
+        dict_t                          *dict = NULL;
+        glusterd_sm_tr_log_t            *log = NULL;
+        xlator_t                        *this = NULL;
+        glusterd_conf_t                 *conf = NULL;
+        char                            msg[2048] = {0};
+        glusterd_peerinfo_t             *peerinfo = NULL;
 
         GF_ASSERT (req);
 
-        priv    = THIS->private;
-
-        if (!gf_xdr_to_cli_log_locate_req (req->msg[0], &cli_req)) {
+        ret = xdr_to_generic (req->msg[0], &cli_req,
+                              (xdrproc_t)xdr_gf1_cli_fsm_log_req);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
+                snprintf (msg, sizeof (msg), "Garbage request");
                 goto out;
         }
 
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received log locate req "
-                "for volume %s", cli_req.volname);
-
-        if (strchr (cli_req.brick, ':')) {
-                /* TODO: need to get info of only that brick and then
-                   tell what is the exact location */
-                gf_log ("", GF_LOG_DEBUG, "brick : %s", cli_req.brick);
+        if (strcmp ("", cli_req.name) == 0) {
+                this = THIS;
+                conf = this->private;
+                log = &conf->op_sm_log;
+        } else {
+                ret = glusterd_friend_find_by_hostname (cli_req.name,
+                                                        &peerinfo);
+                if (ret) {
+                        snprintf (msg, sizeof (msg), "%s is not a peer",
+                                  cli_req.name);
+                        goto out;
+                }
+                log = &peerinfo->sm_log;
         }
 
-        ret = glusterd_volinfo_find (cli_req.volname, &volinfo);
-        if (ret) {
-                rsp.path = "request sent on non-existent volume";
+        dict = dict_new ();
+        if (!dict) {
+                ret = -1;
                 goto out;
         }
 
-        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
-                if (brickinfo->logfile) {
-                        strcpy (tmp_str, brickinfo->logfile);
-                        rsp.path = dirname (tmp_str);
-                } else {
-                        snprintf (tmp_str, PATH_MAX, "%s/logs/bricks/",
-                                  priv->workdir);
-                        rsp.path = tmp_str;
-                }
-                break;
-        }
-
-        ret = 0;
+        ret = glusterd_sm_tr_log_add_to_dict (dict, log);
 out:
-        rsp.op_ret = ret;
-        if (!rsp.path)
-                rsp.path = "";
+        (void)glusterd_fsm_log_send_resp (req, ret, msg, dict);
+        free (cli_req.name);//malloced by xdr
+        if (dict)
+                dict_unref (dict);
 
-        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gf_xdr_serialize_cli_log_locate_rsp);
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
 
-        if (cli_req.brick)
-                free (cli_req.brick); //its malloced by xdr
-        if (cli_req.volname)
-                free (cli_req.volname); //its malloced by xdr
-        return ret;
+        return 0;//send 0 to avoid double reply
 }
 
 int
-glusterd_handle_log_rotate (rpcsvc_request_t *req)
+glusterd_handle_fsm_log (rpcsvc_request_t *req)
 {
-        int32_t                 ret     = -1;
-        gf1_cli_log_rotate_req  cli_req = {0,};
-        dict_t                 *dict    = NULL;
-
-        GF_ASSERT (req);
-
-        if (!gf_xdr_to_cli_log_rotate_req (req->msg[0], &cli_req)) {
-                //failed to decode msg;
-                req->rpc_err = GARBAGE_ARGS;
-                goto out;
-        }
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received log rotate req "
-                "for volume %s", cli_req.volname);
-
-        dict = dict_new ();
-        if (!dict)
-                goto out;
-
-        ret = dict_set_dynmstr (dict, "volname", cli_req.volname);
-        if (ret)
-                goto out;
-
-        ret = dict_set_dynmstr (dict, "brick", cli_req.brick);
-        if (ret)
-                goto out;
-
-        ret = dict_set_uint64 (dict, "rotate-key", (uint64_t)time (NULL));
-        if (ret)
-                goto out;
-
-        ret = glusterd_log_rotate (req, dict);
-
-out:
-        if (ret && dict)
-                dict_unref (dict);
-        return ret;
+        return glusterd_big_locked_handler (req, __glusterd_handle_fsm_log);
 }
 
 int
-glusterd_handle_sync_volume (rpcsvc_request_t *req)
+glusterd_op_lock_send_resp (rpcsvc_request_t *req, int32_t status)
 {
-        int32_t                          ret     = -1;
-        gf1_cli_sync_volume_req          cli_req = {0,};
-        dict_t                           *dict = NULL;
-        gf1_cli_sync_volume_rsp          cli_rsp = {0.};
-        char                             msg[2048] = {0,};
-        gf_boolean_t                     free_hostname = _gf_true;
-        gf_boolean_t                     free_volname = _gf_true;
-        glusterd_volinfo_t               *volinfo = NULL;
 
-        GF_ASSERT (req);
-
-        if (!gf_xdr_to_cli_sync_volume_req (req->msg[0], &cli_req)) {
-                //failed to decode msg;
-                req->rpc_err = GARBAGE_ARGS;
-                goto out;
-        }
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received volume sync req "
-                "for volume %s",
-                (cli_req.flags & GF_CLI_SYNC_ALL) ? "all" : cli_req.volname);
+        gd1_mgmt_cluster_lock_rsp       rsp = {{0},};
+        int                             ret = -1;
 
-        dict = dict_new ();
-        if (!dict) {
-                gf_log ("", GF_LOG_ERROR, "Can't allocate sync vol dict");
-                goto out;
-        }
+        GF_ASSERT (req);
+        glusterd_get_uuid (&rsp.uuid);
+        rsp.op_ret = status;
 
-        if (!glusterd_is_local_addr (cli_req.hostname)) {
-                ret = -1;
-                snprintf (msg, sizeof (msg), "sync from localhost"
-                          " not allowed");
-                goto out;
-        }
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gd1_mgmt_cluster_lock_rsp);
 
-        ret = dict_set_dynmstr (dict, "hostname", cli_req.hostname);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "hostname set failed");
-                snprintf (msg, sizeof (msg), "hostname set failed");
-                goto out;
-        } else {
-                free_hostname = _gf_false;
-        }
+        gf_log (THIS->name, GF_LOG_DEBUG, "Responded to lock, ret: %d", ret);
 
-        ret = dict_set_int32 (dict, "flags", cli_req.flags);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "volume flags set failed");
-                snprintf (msg, sizeof (msg), "volume flags set failed");
-                goto out;
-        }
+        return 0;
+}
 
-        if (!cli_req.flags) {
-                ret = glusterd_volinfo_find (cli_req.volname, &volinfo);
-                if (!ret) {
-                        snprintf (msg, sizeof (msg), "please delete the "
-                                 "volume: %s before sync", cli_req.volname);
-                        ret = -1;
-                        goto out;
-                }
+int
+glusterd_op_unlock_send_resp (rpcsvc_request_t *req, int32_t status)
+{
 
-                ret = dict_set_dynmstr (dict, "volname", cli_req.volname);
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "volume name set failed");
-                        snprintf (msg, sizeof (msg), "volume name set failed");
-                        goto out;
-                } else {
-                        free_volname = _gf_false;
-                }
-        } else {
-                free_volname = _gf_false;
-                if (glusterd_volume_count_get ()) {
-                        snprintf (msg, sizeof (msg), "please delete all the "
-                                 "volumes before full sync");
-                        ret = -1;
-                        goto out;
-                }
-        }
+        gd1_mgmt_cluster_unlock_rsp     rsp = {{0},};
+        int                             ret = -1;
 
-        ret = glusterd_sync_volume (req, dict);
+        GF_ASSERT (req);
+        rsp.op_ret = status;
+        glusterd_get_uuid (&rsp.uuid);
 
-out:
-        if (ret) {
-                cli_rsp.op_ret = -1;
-                cli_rsp.op_errstr = msg;
-                glusterd_submit_reply(req, &cli_rsp, NULL, 0, NULL,
-                                      gf_xdr_from_cli_sync_volume_rsp);
-                if (free_hostname && cli_req.hostname)
-                        free (cli_req.hostname);
-                if (free_volname && cli_req.volname)
-                        free (cli_req.volname);
-                if (dict)
-                        dict_unref (dict);
-                if (!glusterd_opinfo_unlock())
-                        gf_log ("glusterd", GF_LOG_ERROR, "Unlock on "
-                               "opinfo failed");
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_rsp);
 
-                ret = 0; //sent error to cli, prevent second reply
-        }
+        gf_log (THIS->name, GF_LOG_DEBUG, "Responded to unlock, ret: %d", ret);
 
         return ret;
 }
 
 int
-glusterd_op_lock_send_resp (rpcsvc_request_t *req, int32_t status)
+glusterd_op_mgmt_v3_lock_send_resp (rpcsvc_request_t *req, uuid_t *txn_id,
+                                    int32_t status)
 {
 
-        gd1_mgmt_cluster_lock_rsp       rsp = {{0},};
-        int                             ret = -1;
+        gd1_mgmt_v3_lock_rsp           rsp = {{0},};
+        int                            ret = -1;
 
         GF_ASSERT (req);
+        GF_ASSERT (txn_id);
         glusterd_get_uuid (&rsp.uuid);
         rsp.op_ret = status;
+        if (rsp.op_ret)
+               rsp.op_errno = errno;
+        uuid_copy (rsp.txn_id, *txn_id);
 
         ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gd_xdr_serialize_mgmt_cluster_lock_rsp);
+                                     (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp);
 
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Responded, ret: %d", ret);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Responded to mgmt_v3 lock, ret: %d",
+                ret);
 
-        return 0;
+        return ret;
 }
 
 int
-glusterd_op_unlock_send_resp (rpcsvc_request_t *req, int32_t status)
+glusterd_op_mgmt_v3_unlock_send_resp (rpcsvc_request_t *req, uuid_t *txn_id,
+                                      int32_t status)
 {
 
-        gd1_mgmt_cluster_unlock_rsp     rsp = {{0},};
+        gd1_mgmt_v3_unlock_rsp          rsp = {{0},};
         int                             ret = -1;
 
         GF_ASSERT (req);
+        GF_ASSERT (txn_id);
         rsp.op_ret = status;
+        if (rsp.op_ret)
+               rsp.op_errno = errno;
         glusterd_get_uuid (&rsp.uuid);
+        uuid_copy (rsp.txn_id, *txn_id);
 
         ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gd_xdr_serialize_mgmt_cluster_unlock_rsp);
+                                     (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp);
 
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Responded to unlock, ret: %d", ret);
+        gf_log (THIS->name, GF_LOG_DEBUG,
+                "Responded to mgmt_v3 unlock, ret: %d",
+                ret);
 
         return ret;
 }
 
 int
-glusterd_handle_cluster_unlock (rpcsvc_request_t *req)
+__glusterd_handle_cluster_unlock (rpcsvc_request_t *req)
 {
         gd1_mgmt_cluster_unlock_req     unlock_req = {{0}, };
         int32_t                         ret = -1;
-        char                            str[50] = {0, };
         glusterd_op_lock_ctx_t          *ctx = NULL;
+        glusterd_peerinfo_t             *peerinfo = NULL;
+        xlator_t                        *this = NULL;
+        uuid_t                          *txn_id = &global_txn_id;
 
+        this = THIS;
+        GF_ASSERT (this);
         GF_ASSERT (req);
 
-        if (!gd_xdr_to_mgmt_cluster_unlock_req (req->msg[0], &unlock_req)) {
-                //failed to decode msg;
+        ret = xdr_to_generic (req->msg[0], &unlock_req,
+                              (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode unlock "
+                        "request received from peer");
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-        uuid_unparse (unlock_req.uuid, str);
 
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received UNLOCK from uuid: %s", str);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Received UNLOCK from uuid: %s", uuid_utoa (unlock_req.uuid));
+
+        if (glusterd_friend_find_by_uuid (unlock_req.uuid, &peerinfo)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s doesn't "
+                        "belong to the cluster. Ignoring request.",
+                        uuid_utoa (unlock_req.uuid));
+                ret = -1;
+                goto out;
+        }
 
         ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_op_lock_ctx_t);
 
@@ -2223,22 +2147,37 @@ glusterd_handle_cluster_unlock (rpcsvc_request_t *req)
         }
         uuid_copy (ctx->uuid, unlock_req.uuid);
         ctx->req = req;
+        ctx->dict = NULL;
 
-        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_UNLOCK, ctx);
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_UNLOCK, txn_id, ctx);
 
 out:
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
         return ret;
 }
 
 int
-glusterd_op_stage_send_resp (rpcsvc_request_t   *req,
-                             int32_t op, int32_t status, char *op_errstr)
+glusterd_handle_cluster_unlock (rpcsvc_request_t *req)
 {
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cluster_unlock);
+}
 
-        gd1_mgmt_stage_op_rsp           rsp = {{0},};
-        int                             ret = -1;
+int
+glusterd_op_stage_send_resp (rpcsvc_request_t   *req,
+                             int32_t op, int32_t status,
+                             char *op_errstr, dict_t *rsp_dict)
+{
+        gd1_mgmt_stage_op_rsp           rsp      = {{0},};
+        int                             ret      = -1;
+        xlator_t                       *this     = NULL;
 
+        this = THIS;
+        GF_ASSERT (this);
         GF_ASSERT (req);
+
         rsp.op_ret = status;
         glusterd_get_uuid (&rsp.uuid);
         rsp.op = op;
@@ -2247,50 +2186,20 @@ glusterd_op_stage_send_resp (rpcsvc_request_t   *req,
         else
                 rsp.op_errstr = "";
 
-        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gd_xdr_serialize_mgmt_stage_op_rsp);
-
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Responded to stage, ret: %d", ret);
-
-        return ret;
-}
-
-static int
-glusterd_fill_rb_commit_rsp (dict_t *rsp_dict)
-{
-        dict_t *dict    = NULL;
-        int32_t port_no = 0;
-        int     ret     = 0;
-
-        dict = glusterd_op_get_ctx (GD_OP_REPLACE_BRICK);
-        if (!dict) {
-                gf_log ("", GF_LOG_ERROR,
-                        "Operation Context is not present");
-                ret = 0;
-                goto out;
-        }
-
-        ret = dict_get_int32 (dict, "src-brick-port", &port_no);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not get src-brick-port => this must "
-                        "be a non-source glusterd process");
-                ret = 0;
-                goto out;
+        ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val,
+                                           &rsp.dict.dict_len);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get serialized length of dict");
+                return ret;
         }
 
-        gf_log ("", GF_LOG_DEBUG,
-                "This is the source glusterd => fill the src port");
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gd1_mgmt_stage_op_rsp);
 
-        ret = dict_set_int32 (rsp_dict, "src-brick-port", port_no);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not set commit rsp dict");
-                goto out;
-        }
+        gf_log (this->name, GF_LOG_DEBUG, "Responded to stage, ret: %d", ret);
+        GF_FREE (rsp.dict.dict_val);
 
-out:
         return ret;
 }
 
@@ -2301,7 +2210,10 @@ glusterd_op_commit_send_resp (rpcsvc_request_t *req,
 {
         gd1_mgmt_commit_op_rsp          rsp      = {{0}, };
         int                             ret      = -1;
+        xlator_t                        *this = NULL;
 
+        this = THIS;
+        GF_ASSERT (this);
         GF_ASSERT (req);
         rsp.op_ret = status;
         glusterd_get_uuid (&rsp.uuid);
@@ -2312,120 +2224,169 @@ glusterd_op_commit_send_resp (rpcsvc_request_t *req,
         else
                 rsp.op_errstr = "";
 
-        switch (op) {
-        case GD_OP_REPLACE_BRICK:
-                ret = glusterd_fill_rb_commit_rsp (rsp_dict);
-                if (ret)
+        if (rsp_dict) {
+                ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val,
+                                                   &rsp.dict.dict_len);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to get serialized length of dict");
                         goto out;
-                break;
-        default:
-                break;
-        }
-
-        ret = dict_allocate_and_serialize (rsp_dict,
-                                           &rsp.dict.dict_val,
-                                           (size_t *)&rsp.dict.dict_len);
-        if (ret < 0) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "failed to get serialized length of dict");
-                goto out;
+                }
         }
 
 
         ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gd_xdr_serialize_mgmt_commit_op_rsp);
+                                     (xdrproc_t)xdr_gd1_mgmt_commit_op_rsp);
 
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Responded to commit, ret: %d", ret);
+        gf_log (this->name, GF_LOG_DEBUG, "Responded to commit, ret: %d", ret);
 
 out:
-        if (rsp.dict.dict_val)
-                GF_FREE (rsp.dict.dict_val);
+        GF_FREE (rsp.dict.dict_val);
         return ret;
 }
 
 int
-glusterd_handle_incoming_friend_req (rpcsvc_request_t *req)
+__glusterd_handle_incoming_friend_req (rpcsvc_request_t *req)
 {
         int32_t                 ret = -1;
         gd1_mgmt_friend_req     friend_req = {{0},};
-        char                    str[50] = {0,};
-        dict_t                  *dict = NULL;
+        gf_boolean_t            run_fsm = _gf_true;
 
         GF_ASSERT (req);
-        if (!gd_xdr_to_mgmt_friend_req (req->msg[0], &friend_req)) {
+        ret = xdr_to_generic (req->msg[0], &friend_req,
+                              (xdrproc_t)xdr_gd1_mgmt_friend_req);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
-        uuid_unparse (friend_req.uuid, str);
-
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received probe from uuid: %s", str);
-
-        dict = dict_new ();
-        if (!dict) {
-                ret = -1;
-                goto out;
-        }
-
-        ret = dict_unserialize (friend_req.vols.vols_val,
-                                friend_req.vols.vols_len,
-                                &dict);
-
-        if (ret)
-                goto out;
-        else
-                dict->extra_stdfree = friend_req.vols.vols_val;
 
+        gf_log ("glusterd", GF_LOG_INFO,
+                "Received probe from uuid: %s", uuid_utoa (friend_req.uuid));
         ret = glusterd_handle_friend_req (req, friend_req.uuid,
                                           friend_req.hostname, friend_req.port,
-                                          dict);
+                                          &friend_req);
+
+        if (ret == GLUSTERD_CONNECTION_AWAITED) {
+                //fsm should be run after connection establishes
+                run_fsm = _gf_false;
+                ret = 0;
+        }
 
 out:
-        if (ret && dict)
-                dict_unref (dict);
-        if (friend_req.hostname)
-                free (friend_req.hostname);//malloced by xdr
+        free (friend_req.hostname);//malloced by xdr
+
+        if (run_fsm) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+        }
 
         return ret;
 }
 
 int
-glusterd_handle_incoming_unfriend_req (rpcsvc_request_t *req)
+glusterd_handle_incoming_friend_req (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_incoming_friend_req);
+}
+
+int
+__glusterd_handle_incoming_unfriend_req (rpcsvc_request_t *req)
 {
         int32_t                 ret = -1;
         gd1_mgmt_friend_req     friend_req = {{0},};
-        char                    str[50];
+        char               remote_hostname[UNIX_PATH_MAX + 1] = {0,};
 
         GF_ASSERT (req);
-        if (!gd_xdr_to_mgmt_friend_req (req->msg[0], &friend_req)) {
+        ret = xdr_to_generic (req->msg[0], &friend_req,
+                              (xdrproc_t)xdr_gd1_mgmt_friend_req);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
-        uuid_unparse (friend_req.uuid, str);
 
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received unfriend from uuid: %s", str);
+        gf_log ("glusterd", GF_LOG_INFO,
+                "Received unfriend from uuid: %s", uuid_utoa (friend_req.uuid));
 
+        ret = glusterd_remote_hostname_get (req, remote_hostname,
+                                            sizeof (remote_hostname));
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get the remote hostname");
+                goto out;
+        }
         ret = glusterd_handle_unfriend_req (req, friend_req.uuid,
-                                            friend_req.hostname, friend_req.port);
+                                            remote_hostname, friend_req.port);
 
 out:
-        if (friend_req.hostname)
-                free (friend_req.hostname);//malloced by xdr
-        if (friend_req.vols.vols_val)
-                free (friend_req.vols.vols_val);//malloced by xdr
+        free (friend_req.hostname);//malloced by xdr
+        free (friend_req.vols.vols_val);//malloced by xdr
+
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
         return ret;
 }
 
 int
-glusterd_handle_friend_update (rpcsvc_request_t *req)
+glusterd_handle_incoming_unfriend_req (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_incoming_unfriend_req);
+
+}
+
+int
+glusterd_handle_friend_update_delete (dict_t *dict)
+{
+        char                    *hostname = NULL;
+        int32_t                 ret = -1;
+
+        GF_ASSERT (dict);
+
+        ret = dict_get_str (dict, "hostname", &hostname);
+        if (ret)
+                goto out;
+
+        ret = glusterd_friend_remove (NULL, hostname);
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_friend_hostname_update (glusterd_peerinfo_t *peerinfo,
+                                char *hostname,
+                                gf_boolean_t store_update)
+{
+        char                    *new_hostname = NULL;
+        int                     ret = 0;
+
+        GF_ASSERT (peerinfo);
+        GF_ASSERT (hostname);
+
+        new_hostname = gf_strdup (hostname);
+        if (!new_hostname) {
+                ret = -1;
+                goto out;
+        }
+
+        GF_FREE (peerinfo->hostname);
+        peerinfo->hostname = new_hostname;
+        if (store_update)
+                ret = glusterd_store_peerinfo (peerinfo);
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+__glusterd_handle_friend_update (rpcsvc_request_t *req)
 {
         int32_t                 ret = -1;
         gd1_mgmt_friend_update     friend_req = {{0},};
-        char                    str[50] = {0,};
         glusterd_peerinfo_t     *peerinfo = NULL;
         glusterd_conf_t         *priv = NULL;
         xlator_t                *this = NULL;
@@ -2438,6 +2399,8 @@ glusterd_handle_friend_update (rpcsvc_request_t *req)
         int                     i = 1;
         int                     count = 0;
         uuid_t                  uuid = {0,};
+        glusterd_peerctx_args_t args = {0};
+        int32_t                 op = 0;
 
         GF_ASSERT (req);
 
@@ -2446,15 +2409,22 @@ glusterd_handle_friend_update (rpcsvc_request_t *req)
         priv = this->private;
         GF_ASSERT (priv);
 
-        if (!gd_xdr_to_mgmt_friend_update (req->msg[0], &friend_req)) {
+        ret = xdr_to_generic (req->msg[0], &friend_req,
+                              (xdrproc_t)xdr_gd1_mgmt_friend_update);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
-        uuid_unparse (friend_req.uuid, str);
 
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received friend update from uuid: %s", str);
+        ret = glusterd_friend_find (friend_req.uuid, NULL, &tmp);
+        if (ret) {
+                gf_log ("", GF_LOG_CRITICAL, "Received friend update request "
+                        "from unknown peer %s", uuid_utoa (friend_req.uuid));
+                goto out;
+        }
+        gf_log ("glusterd", GF_LOG_INFO,
+                "Received friend update from uuid: %s", uuid_utoa (friend_req.uuid));
 
         if (friend_req.friends.friends_len) {
                 /* Unserialize the dictionary */
@@ -2477,6 +2447,16 @@ glusterd_handle_friend_update (rpcsvc_request_t *req)
         if (ret)
                 goto out;
 
+        ret = dict_get_int32 (dict, "op", &op);
+        if (ret)
+                goto out;
+
+        if (GD_FRIEND_UPDATE_DEL == op) {
+                ret = glusterd_handle_friend_update_delete (dict);
+                goto out;
+        }
+
+        args.mode = GD_MODE_ON;
         while ( i <= count) {
                 snprintf (key, sizeof (key), "friend%d.uuid", i);
                 ret = dict_get_str (dict, key, &uuid_buf);
@@ -2488,11 +2468,17 @@ glusterd_handle_friend_update (rpcsvc_request_t *req)
                 if (ret)
                         goto out;
 
-                gf_log ("", GF_LOG_NORMAL, "Received uuid: %s, hostname:%s",
+                gf_log ("", GF_LOG_INFO, "Received uuid: %s, hostname:%s",
                                 uuid_buf, hostname);
 
-                if (!uuid_compare (uuid, priv->uuid)) {
-                        gf_log ("", GF_LOG_NORMAL, "Received my uuid as Friend");
+                if (uuid_is_null (uuid)) {
+                        gf_log (this->name, GF_LOG_WARNING, "Updates mustn't "
+                                "contain peer with 'null' uuid");
+                        continue;
+                }
+
+                if (!uuid_compare (uuid, MY_UUID)) {
+                        gf_log ("", GF_LOG_INFO, "Received my uuid as Friend");
                         i++;
                         continue;
                 }
@@ -2500,744 +2486,1025 @@ glusterd_handle_friend_update (rpcsvc_request_t *req)
                 ret = glusterd_friend_find (uuid, hostname, &tmp);
 
                 if (!ret) {
+                        if (strcmp (hostname, tmp->hostname) != 0) {
+                                glusterd_friend_hostname_update (tmp, hostname,
+                                                                 _gf_true);
+                        }
                         i++;
                         continue;
                 }
 
                 ret = glusterd_friend_add (hostname, friend_req.port,
                                            GD_FRIEND_STATE_BEFRIENDED,
-                                           &uuid, NULL, &peerinfo, 0);
+                                           &uuid, &peerinfo, 0, &args);
 
                 i++;
         }
 
 out:
-        uuid_copy (rsp.uuid, priv->uuid);
+        uuid_copy (rsp.uuid, MY_UUID);
         ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gd_xdr_serialize_mgmt_friend_update_rsp);
-        if (dict)
+                                     (xdrproc_t)xdr_gd1_mgmt_friend_update_rsp);
+        if (dict) {
+                if (!dict->extra_stdfree && friend_req.friends.friends_val)
+                        free (friend_req.friends.friends_val);//malloced by xdr
                 dict_unref (dict);
+        } else {
+                free (friend_req.friends.friends_val);//malloced by xdr
+        }
+
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
         return ret;
 }
 
 int
-glusterd_handle_probe_query (rpcsvc_request_t *req)
+glusterd_handle_friend_update (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_friend_update);
+}
+
+int
+__glusterd_handle_probe_query (rpcsvc_request_t *req)
 {
-        int32_t             ret = -1;
-        char                str[50];
-        xlator_t            *this = NULL;
-        glusterd_conf_t     *conf = NULL;
-        gd1_mgmt_probe_req  probe_req = {{0},};
-        gd1_mgmt_probe_rsp  rsp = {{0},};
-        glusterd_peer_hostname_t        *name = NULL;
+        int32_t                         ret = -1;
+        xlator_t                        *this = NULL;
+        glusterd_conf_t                 *conf = NULL;
+        gd1_mgmt_probe_req              probe_req = {{0},};
+        gd1_mgmt_probe_rsp              rsp = {{0},};
         glusterd_peerinfo_t             *peerinfo = NULL;
+        glusterd_peerctx_args_t         args = {0};
+        int                             port = 0;
         char               remote_hostname[UNIX_PATH_MAX + 1] = {0,};
 
         GF_ASSERT (req);
 
-        if (!gd_xdr_to_mgmt_probe_req (req->msg[0], &probe_req)) {
+        ret = xdr_to_generic (req->msg[0], &probe_req,
+                              (xdrproc_t)xdr_gd1_mgmt_probe_req);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
                 goto out;
         }
 
-
         this = THIS;
 
         conf = this->private;
-        uuid_unparse (probe_req.uuid, str);
-
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received probe from uuid: %s", str);
+        if (probe_req.port)
+                port = probe_req.port;
+        else
+                port = GF_DEFAULT_BASE_PORT;
+
+        gf_log ("glusterd", GF_LOG_INFO,
+                "Received probe from uuid: %s", uuid_utoa (probe_req.uuid));
+
+        /* Check for uuid collision and handle it in a user friendly way by
+         * sending the error.
+         */
+        if (!uuid_compare (probe_req.uuid, MY_UUID)) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Peer uuid %s is same as "
+                        "local uuid. Please check the uuid of both the peers "
+                        "from %s/%s", uuid_utoa (probe_req.uuid),
+                        GLUSTERD_DEFAULT_WORKDIR, GLUSTERD_INFO_FILE);
+                rsp.op_ret = -1;
+                rsp.op_errno = GF_PROBE_SAME_UUID;
+                rsp.port = port;
+                goto respond;
+        }
 
         ret = glusterd_remote_hostname_get (req, remote_hostname,
                                             sizeof (remote_hostname));
         if (ret) {
-                GF_ASSERT (0);
+                gf_log ("", GF_LOG_ERROR, "Unable to get the remote hostname");
                 goto out;
         }
         ret = glusterd_friend_find (probe_req.uuid, remote_hostname, &peerinfo);
-        if ((ret == 0 ) || list_empty (&conf->peers)) {
-                ret = glusterd_peer_hostname_new (probe_req.hostname, &name);
-
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to get new peer_hostname");
-                } else {
-                        list_add_tail (&name->hostname_list, &conf->hostnames);
-                }
-
-        } else {
+        if ((ret != 0 ) && (!list_empty (&conf->peers))) {
                 rsp.op_ret = -1;
                 rsp.op_errno = GF_PROBE_ANOTHER_CLUSTER;
+        } else if (ret) {
+                gf_log ("glusterd", GF_LOG_INFO, "Unable to find peerinfo"
+                        " for host: %s (%d)", remote_hostname, port);
+                args.mode = GD_MODE_ON;
+                ret = glusterd_friend_add (remote_hostname, port,
+                                           GD_FRIEND_STATE_PROBE_RCVD,
+                                           NULL, &peerinfo, 0, &args);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Failed to add peer %s",
+                                remote_hostname);
+                        rsp.op_errno = GF_PROBE_ADD_FAILED;
+                }
         }
 
-        uuid_copy (rsp.uuid, conf->uuid);
+respond:
+        uuid_copy (rsp.uuid, MY_UUID);
 
         rsp.hostname = probe_req.hostname;
+        rsp.op_errstr = "";
 
-        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gd_xdr_serialize_mgmt_probe_rsp);
+        glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                               (xdrproc_t)xdr_gd1_mgmt_probe_rsp);
+        ret = 0;
 
-        gf_log ("glusterd", GF_LOG_NORMAL, "Responded to %s, op_ret: %d, "
-                "op_errno: %d, ret: %d", probe_req.hostname,
+        gf_log ("glusterd", GF_LOG_INFO, "Responded to %s, op_ret: %d, "
+                "op_errno: %d, ret: %d", remote_hostname,
                 rsp.op_ret, rsp.op_errno, ret);
 
 out:
-        if (probe_req.hostname)
-                free (probe_req.hostname);//malloced by xdr
+        free (probe_req.hostname);//malloced by xdr
+
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
         return ret;
 }
 
-int
-glusterd_friend_remove (uuid_t uuid, char *hostname)
+int glusterd_handle_probe_query (rpcsvc_request_t *req)
 {
-        int                           ret = 0;
-        glusterd_peerinfo_t           *peerinfo = NULL;
-
-        ret = glusterd_friend_find (uuid, hostname, &peerinfo);
-        if (ret)
-                goto out;
-
-        ret = glusterd_friend_cleanup (peerinfo);
-out:
-        gf_log ("", GF_LOG_DEBUG, "returning %d");
-        return ret;
+        return glusterd_big_locked_handler (req, __glusterd_handle_probe_query);
 }
 
 int
-glusterd_friend_add (const char *hoststr, int port,
-                     glusterd_friend_sm_state_t state,
-                     uuid_t *uuid,
-                     struct rpc_clnt    *rpc,
-                     glusterd_peerinfo_t **friend,
-                     gf_boolean_t restore)
+__glusterd_handle_cli_profile_volume (rpcsvc_request_t *req)
 {
-        int                     ret = 0;
-        glusterd_conf_t         *priv = NULL;
-        glusterd_peerinfo_t     *peerinfo = NULL;
-        dict_t                  *options = NULL;
-        struct rpc_clnt_config  rpc_cfg = {0,};
-        glusterd_peer_hostname_t *name = NULL;
-        char                    *hostname = NULL;
-
-        priv = THIS->private;
-
-        peerinfo = GF_CALLOC (1, sizeof(*peerinfo), gf_gld_mt_peerinfo_t);
+        int32_t                         ret     = -1;
+        gf_cli_req                      cli_req = {{0,}};
+        dict_t                          *dict = NULL;
+        glusterd_op_t                   cli_op = GD_OP_PROFILE_VOLUME;
+        char                            *volname = NULL;
+        int32_t                         op = 0;
+        char                            err_str[2048] = {0,};
+        xlator_t                        *this = NULL;
 
-        if (!peerinfo)
-                return -1;
+        GF_ASSERT (req);
+        this = THIS;
+        GF_ASSERT (this);
 
-        if (friend)
-                *friend = peerinfo;
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
 
-        INIT_LIST_HEAD (&peerinfo->hostnames);
-        peerinfo->state.state = state;
-        if (hoststr) {
-                ret =  glusterd_peer_hostname_new ((char *)hoststr, &name);
-                if (ret)
+        if (cli_req.dict.dict_len > 0) {
+                dict = dict_new();
+                if (!dict)
                         goto out;
-                list_add_tail (&peerinfo->hostnames, &name->hostname_list);
-                rpc_cfg.remote_host = gf_strdup (hoststr);
-                peerinfo->hostname = gf_strdup (hoststr);
+                dict_unserialize (cli_req.dict.dict_val,
+                                  cli_req.dict.dict_len, &dict);
         }
-        INIT_LIST_HEAD (&peerinfo->uuid_list);
-
-        list_add_tail (&peerinfo->uuid_list, &priv->peers);
 
-        if (uuid) {
-                uuid_copy (peerinfo->uuid, *uuid);
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get volume "
+                          "name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
         }
 
+        gf_log (this->name, GF_LOG_INFO, "Received volume profile req "
+                "for volume %s", volname);
+        ret = dict_get_int32 (dict, "op", &op);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get operation");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
 
-        if (hoststr) {
-                options = dict_new ();
-                if (!options)
-                        return -1;
-
-                hostname = gf_strdup((char*)hoststr);
-                if (!hostname) {
-                        ret = -1;
-                        goto out;
-                }
-
-                ret = dict_set_dynstr (options, "remote-host", hostname);
-                if (ret)
-                        goto out;
+        ret = glusterd_op_begin (req, cli_op, dict, err_str, sizeof (err_str));
 
+out:
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
 
-                if (!port)
-                        port = GLUSTERD_DEFAULT_PORT;
+        free (cli_req.dict.dict_val);
 
-                rpc_cfg.remote_port = port;
+        if (ret) {
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
+        }
 
-                ret = dict_set_int32 (options, "remote-port", port);
-                if (ret)
-                        goto out;
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
 
-                ret = dict_set_str (options, "transport.address-family", "inet");
-                if (ret)
-                        goto out;
+int
+glusterd_handle_cli_profile_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_profile_volume);
+}
 
-                rpc = rpc_clnt_init (&rpc_cfg, options, THIS->ctx, THIS->name);
+int
+__glusterd_handle_getwd (rpcsvc_request_t *req)
+{
+        int32_t                 ret = -1;
+        gf1_cli_getwd_rsp     rsp = {0,};
+        glusterd_conf_t         *priv = NULL;
 
-                if (!rpc) {
-                        gf_log ("glusterd", GF_LOG_ERROR,
-                                "rpc init failed for peer: %s!", hoststr);
-                        ret = -1;
-                        goto out;
-                }
+        GF_ASSERT (req);
 
-                ret = rpc_clnt_register_notify (rpc, glusterd_rpc_notify,
-                                                peerinfo);
+        priv = THIS->private;
+        GF_ASSERT (priv);
 
-                peerinfo->rpc = rpc;
+        gf_log ("glusterd", GF_LOG_INFO, "Received getwd req");
 
-        }
+        rsp.wd = priv->workdir;
 
-        if (!restore)
-                ret = glusterd_store_update_peerinfo (peerinfo);
+        glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                               (xdrproc_t)xdr_gf1_cli_getwd_rsp);
+        ret = 0;
 
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
 
-out:
-        gf_log ("glusterd", GF_LOG_NORMAL, "connect returned %d", ret);
-        if (rpc_cfg.remote_host)
-                GF_FREE (rpc_cfg.remote_host);
         return ret;
 }
 
-
-
 int
-glusterd_probe_begin (rpcsvc_request_t *req, const char *hoststr, int port)
+glusterd_handle_getwd (rpcsvc_request_t *req)
 {
-        int                             ret = -1;
-        glusterd_peerinfo_t             *peerinfo = NULL;
-        glusterd_friend_sm_event_t      *event = NULL;
-        glusterd_probe_ctx_t            *ctx = NULL;
+        return glusterd_big_locked_handler (req, __glusterd_handle_getwd);
+}
 
-        GF_ASSERT (hoststr);
+int
+__glusterd_handle_mount (rpcsvc_request_t *req)
+{
+        gf1_cli_mount_req mnt_req = {0,};
+        gf1_cli_mount_rsp rsp     = {0,};
+        dict_t *dict              = NULL;
+        int ret                   = 0;
+        glusterd_conf_t     *priv   = NULL;
 
-        ret = glusterd_friend_find (NULL, (char *)hoststr, &peerinfo);
+        GF_ASSERT (req);
+	priv = THIS->private;
 
-        if (ret) {
-                gf_log ("glusterd", GF_LOG_NORMAL, "Unable to find peerinfo"
-                        " for host: %s (%d)", hoststr, port);
-                ret = glusterd_friend_add ((char *)hoststr, port,
-                                           GD_FRIEND_STATE_DEFAULT,
-                                           NULL, NULL, &peerinfo, 0);
+        ret = xdr_to_generic (req->msg[0], &mnt_req,
+                              (xdrproc_t)xdr_gf1_cli_mount_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                rsp.op_ret = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
         }
 
-        ret = glusterd_friend_sm_new_event
-                        (GD_FRIEND_EVENT_PROBE, &event);
+        gf_log ("glusterd", GF_LOG_INFO, "Received mount req");
 
-        if (ret) {
-                gf_log ("glusterd", GF_LOG_ERROR, "Unable to get new event");
-                return ret;
-        }
-
-        ctx = GF_CALLOC (1, sizeof(*ctx), gf_gld_mt_probe_ctx_t);
+        if (mnt_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
 
-        if (!ctx) {
-                return ret;
+                ret = dict_unserialize (mnt_req.dict.dict_val,
+                                        mnt_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log ("glusterd", GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        rsp.op_ret = -1;
+                        rsp.op_errno = -EINVAL;
+                        goto out;
+                } else {
+                        dict->extra_stdfree = mnt_req.dict.dict_val;
+                }
         }
 
-        ctx->hostname = gf_strdup (hoststr);
-        ctx->port = port;
-        ctx->req = req;
-
-        event->peerinfo = peerinfo;
-        event->ctx = ctx;
+	synclock_unlock (&priv->big_lock);
+        rsp.op_ret = glusterd_do_mount (mnt_req.label, dict,
+                                        &rsp.path, &rsp.op_errno);
+	synclock_lock (&priv->big_lock);
 
-        ret = glusterd_friend_sm_inject_event (event);
+ out:
+        if (!rsp.path)
+                rsp.path = "";
 
-        if (ret) {
-                gf_log ("glusterd", GF_LOG_ERROR, "Unable to inject event %d, "
-                        "ret = %d", event->event, ret);
-                return ret;
-        }
+        glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                               (xdrproc_t)xdr_gf1_cli_mount_rsp);
+        ret = 0;
 
-        if (!peerinfo->connected) {
-                return  GLUSTERD_CONNECTION_AWAITED;
-        }
+        if (dict)
+                dict_unref (dict);
+        if (*rsp.path)
+                GF_FREE (rsp.path);
 
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
 
         return ret;
 }
 
 int
-glusterd_deprobe_begin (rpcsvc_request_t *req, const char *hoststr, int port)
+glusterd_handle_mount (rpcsvc_request_t *req)
 {
-        int                             ret = -1;
-        glusterd_peerinfo_t             *peerinfo = NULL;
-        glusterd_friend_sm_event_t      *event = NULL;
-        glusterd_probe_ctx_t            *ctx = NULL;
+        return glusterd_big_locked_handler (req, __glusterd_handle_mount);
+}
 
-        GF_ASSERT (hoststr);
-        GF_ASSERT (req);
+int
+__glusterd_handle_umount (rpcsvc_request_t *req)
+{
+        gf1_cli_umount_req umnt_req = {0,};
+        gf1_cli_umount_rsp rsp      = {0,};
+        char *mountbroker_root      = NULL;
+        char mntp[PATH_MAX]         = {0,};
+        char *path                  = NULL;
+        runner_t runner             = {0,};
+        int ret                     = 0;
+        xlator_t *this              = THIS;
+        gf_boolean_t dir_ok         = _gf_false;
+        char *pdir                  = NULL;
+        char *t                     = NULL;
+        glusterd_conf_t     *priv   = NULL;
 
-        ret = glusterd_friend_find (NULL, (char *)hoststr, &peerinfo);
+        GF_ASSERT (req);
+        GF_ASSERT (this);
+        priv = this->private;
 
-        if (ret) {
-                gf_log ("glusterd", GF_LOG_NORMAL, "Unable to find peerinfo"
-                        " for host: %s %d", hoststr, port);
+        ret = xdr_to_generic (req->msg[0], &umnt_req,
+                              (xdrproc_t)xdr_gf1_cli_umount_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                rsp.op_ret = -1;
                 goto out;
         }
 
-        if (!peerinfo->rpc) {
-                //handle this case
+        gf_log ("glusterd", GF_LOG_INFO, "Received umount req");
+
+        if (dict_get_str (this->options, "mountbroker-root",
+                          &mountbroker_root) != 0) {
+                rsp.op_errno = ENOENT;
                 goto out;
         }
 
-        ret = glusterd_friend_sm_new_event
-                (GD_FRIEND_EVENT_INIT_REMOVE_FRIEND, &event);
-
-        if (ret) {
-                gf_log ("glusterd", GF_LOG_ERROR,
-                                "Unable to get new event");
-                return ret;
+        /* check if it is allowed to umount path */
+        path = gf_strdup (umnt_req.path);
+        if (!path) {
+                rsp.op_errno = ENOMEM;
+                goto out;
         }
-
-        ctx = GF_CALLOC (1, sizeof(*ctx), gf_gld_mt_probe_ctx_t);
-
-        if (!ctx) {
+        dir_ok = _gf_false;
+        pdir = dirname (path);
+        t = strtail (pdir, mountbroker_root);
+        if (t && *t == '/') {
+                t = strtail(++t, MB_HIVE);
+                if (t && !*t)
+                        dir_ok = _gf_true;
+        }
+        GF_FREE (path);
+        if (!dir_ok) {
+                rsp.op_errno = EACCES;
                 goto out;
         }
 
-        ctx->hostname = gf_strdup (hoststr);
-        ctx->port = port;
-        ctx->req = req;
-
-        event->ctx = ctx;
+        runinit (&runner);
+        runner_add_args (&runner, "umount", umnt_req.path, NULL);
+        if (umnt_req.lazy)
+                runner_add_arg (&runner, "-l");
+        synclock_unlock (&priv->big_lock);
+        rsp.op_ret = runner_run (&runner);
+        synclock_lock (&priv->big_lock);
+        if (rsp.op_ret == 0) {
+                if (realpath (umnt_req.path, mntp))
+                        rmdir (mntp);
+                else {
+                        rsp.op_ret = -1;
+                        rsp.op_errno = errno;
+                }
+                if (unlink (umnt_req.path) != 0) {
+                        rsp.op_ret = -1;
+                        rsp.op_errno = errno;
+                }
+        }
 
-        event->peerinfo = peerinfo;
+ out:
+        if (rsp.op_errno)
+                rsp.op_ret = -1;
 
-        ret = glusterd_friend_sm_inject_event (event);
+        glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                               (xdrproc_t)xdr_gf1_cli_umount_rsp);
+        ret = 0;
 
-        if (ret) {
-                gf_log ("glusterd", GF_LOG_ERROR, "Unable to inject event %d, "
-                        "ret = %d", event->event, ret);
-                goto out;
-        }
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
 
-out:
         return ret;
 }
 
-
 int
-glusterd_xfer_friend_remove_resp (rpcsvc_request_t *req, char *hostname, int port)
+glusterd_handle_umount (rpcsvc_request_t *req)
 {
-        gd1_mgmt_friend_rsp  rsp = {{0}, };
-        int32_t              ret = -1;
-        xlator_t             *this = NULL;
-        glusterd_conf_t      *conf = NULL;
-
-        GF_ASSERT (hostname);
-
-        rsp.op_ret = 0;
-        this = THIS;
-        GF_ASSERT (this);
-
-        conf = this->private;
+        return glusterd_big_locked_handler (req, __glusterd_handle_umount);
+}
 
-        uuid_copy (rsp.uuid, conf->uuid);
-        rsp.hostname = hostname;
-        rsp.port = port;
-        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gd_xdr_serialize_mgmt_friend_rsp);
+int
+glusterd_friend_remove (uuid_t uuid, char *hostname)
+{
+        int                           ret = 0;
+        glusterd_peerinfo_t           *peerinfo = NULL;
 
+        ret = glusterd_friend_find (uuid, hostname, &peerinfo);
+        if (ret)
+                goto out;
 
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Responded to %s (%d), ret: %d", hostname, port, ret);
+        ret = glusterd_friend_remove_cleanup_vols (peerinfo->uuid);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING, "Volumes cleanup failed");
+        ret = glusterd_friend_cleanup (peerinfo);
+out:
+        gf_log ("", GF_LOG_DEBUG, "returning %d", ret);
         return ret;
 }
 
 int
-glusterd_xfer_friend_add_resp (rpcsvc_request_t *req, char *hostname, int port,
-                               int32_t op_ret, int32_t op_errno)
+glusterd_rpc_create (struct rpc_clnt **rpc,
+                     dict_t *options,
+                     rpc_clnt_notify_t notify_fn,
+                     void *notify_data)
 {
-        gd1_mgmt_friend_rsp  rsp = {{0}, };
-        int32_t              ret = -1;
-        xlator_t             *this = NULL;
-        glusterd_conf_t      *conf = NULL;
-
-        GF_ASSERT (hostname);
+        struct rpc_clnt         *new_rpc = NULL;
+        int                     ret = -1;
+        xlator_t                *this = NULL;
 
         this = THIS;
         GF_ASSERT (this);
 
-        conf = this->private;
+        GF_ASSERT (options);
 
-        uuid_copy (rsp.uuid, conf->uuid);
-        rsp.op_ret = op_ret;
-        rsp.op_errno = op_errno;
-        rsp.hostname = gf_strdup (hostname);
-        rsp.port = port;
+        /* TODO: is 32 enough? or more ? */
+        new_rpc = rpc_clnt_new (options, this->ctx, this->name, 16);
+        if (!new_rpc)
+                goto out;
 
-        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gd_xdr_serialize_mgmt_friend_rsp);
+        ret = rpc_clnt_register_notify (new_rpc, notify_fn, notify_data);
+        *rpc = new_rpc;
+        if (ret)
+                goto out;
+        ret = rpc_clnt_start (new_rpc);
+out:
+        if (ret) {
+                if (new_rpc) {
+                        (void) rpc_clnt_unref (new_rpc);
+                }
+        }
 
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Responded to %s (%d), ret: %d", hostname, port, ret);
-        if (rsp.hostname)
-                GF_FREE (rsp.hostname)
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
         return ret;
 }
 
 int
-glusterd_xfer_cli_probe_resp (rpcsvc_request_t *req, int32_t op_ret,
-                              int32_t op_errno, char *hostname, int port)
+glusterd_transport_keepalive_options_get (int *interval, int *time)
 {
-        gf1_cli_probe_rsp    rsp = {0, };
-        int32_t              ret = -1;
-
-        GF_ASSERT (req);
+        int     ret = 0;
+        xlator_t *this = NULL;
 
-        rsp.op_ret = op_ret;
-        rsp.op_errno = op_errno;
-        rsp.hostname = hostname;
-        rsp.port = port;
-
-        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gf_xdr_serialize_cli_probe_rsp);
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Responded to CLI, ret: %d",ret);
+        this = THIS;
+        GF_ASSERT (this);
 
-        return ret;
+        ret = dict_get_int32 (this->options,
+                              "transport.socket.keepalive-interval",
+                              interval);
+        ret = dict_get_int32 (this->options,
+                              "transport.socket.keepalive-time",
+                              time);
+        return 0;
 }
 
 int
-glusterd_xfer_cli_deprobe_resp (rpcsvc_request_t *req, int32_t op_ret,
-                                int32_t op_errno, char *hostname)
+glusterd_transport_inet_options_build (dict_t **options, const char *hostname,
+                                       int port)
 {
-        gf1_cli_deprobe_rsp    rsp = {0, };
-        int32_t                ret = -1;
-
-        GF_ASSERT (req);
+        dict_t  *dict = NULL;
+        int32_t interval = -1;
+        int32_t time     = -1;
+        int     ret = 0;
 
-        rsp.op_ret = op_ret;
-        rsp.op_errno = op_errno;
-        rsp.hostname = hostname;
-
-        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gf_xdr_serialize_cli_deprobe_rsp);
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Responded to CLI, ret: %d",ret);
-
-        return ret;
-}
-int32_t
-glusterd_op_txn_begin ()
-{
-        int32_t                 ret = -1;
-        glusterd_conf_t         *priv = NULL;
-        int32_t                 locked = 0;
+        GF_ASSERT (options);
+        GF_ASSERT (hostname);
 
-        priv = THIS->private;
-        GF_ASSERT (priv);
+        if (!port)
+                port = GLUSTERD_DEFAULT_PORT;
 
-        ret = glusterd_lock (priv->uuid);
+        /* Build default transport options */
+        ret = rpc_transport_inet_options_build (&dict, hostname, port);
+        if (ret)
+                goto out;
 
+        /* Set frame-timeout to 10mins. Default timeout of 30 mins is too long
+         * when compared to 2 mins for cli timeout. This ensures users don't
+         * wait too long after cli timesout before being able to resume normal
+         * operations
+         */
+        ret = dict_set_int32 (dict, "frame-timeout", 600);
         if (ret) {
                 gf_log ("glusterd", GF_LOG_ERROR,
-                        "Unable to acquire local lock, ret: %d", ret);
+                        "Failed to set frame-timeout");
                 goto out;
         }
 
-        locked = 1;
-        gf_log ("glusterd", GF_LOG_NORMAL, "Acquired local lock");
-
-        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_START_LOCK, NULL);
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Returning %d", ret);
+        /* Set keepalive options */
+        glusterd_transport_keepalive_options_get (&interval, &time);
 
+        if ((interval > 0) || (time > 0))
+                ret = rpc_transport_keepalive_options_set (dict, interval, time);
+        *options = dict;
 out:
-        if (locked && ret)
-                glusterd_unlock (priv->uuid);
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
-int32_t
-glusterd_create_volume (rpcsvc_request_t *req, dict_t *dict)
+int
+glusterd_friend_rpc_create (xlator_t *this, glusterd_peerinfo_t *peerinfo,
+                            glusterd_peerctx_args_t *args)
 {
-        int32_t  ret       = -1;
-        data_t  *data = NULL;
+        dict_t                 *options = NULL;
+        int                    ret = -1;
+        glusterd_peerctx_t     *peerctx = NULL;
+        data_t                 *data = NULL;
 
-        GF_ASSERT (req);
-        GF_ASSERT (dict);
-
-        glusterd_op_set_op (GD_OP_CREATE_VOLUME);
-
-        glusterd_op_set_ctx (GD_OP_CREATE_VOLUME, dict);
+        peerctx = GF_CALLOC (1, sizeof (*peerctx), gf_gld_mt_peerctx_t);
+        if (!peerctx)
+                goto out;
 
-        glusterd_op_set_ctx_free (GD_OP_CREATE_VOLUME, _gf_true);
+        if (args)
+                peerctx->args = *args;
 
-        glusterd_op_set_req (req);
+        peerctx->peerinfo = peerinfo;
 
-        data = dict_get (dict, "volname");
-        if (!data)
+        ret = glusterd_transport_inet_options_build (&options,
+                                                     peerinfo->hostname,
+                                                     peerinfo->port);
+        if (ret)
                 goto out;
 
-        data = dict_get (dict, "type");
-        if (!data)
-                goto out;
+        /*
+         * For simulated multi-node testing, we need to make sure that we
+         * create our RPC endpoint with the same address that the peer would
+         * use to reach us.
+         */
+        if (this->options) {
+                data = dict_get(this->options,"transport.socket.bind-address");
+                if (data) {
+                        ret = dict_set(options,
+                                       "transport.socket.source-addr",data);
+                }
+        }
 
-        data = dict_get (dict, "count");
-        if (!data)
+        ret = glusterd_rpc_create (&peerinfo->rpc, options,
+                                   glusterd_peer_rpc_notify, peerctx);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to create rpc for"
+                        " peer %s", peerinfo->hostname);
                 goto out;
+        }
+        peerctx = NULL;
+        ret = 0;
+out:
+        GF_FREE (peerctx);
+        return ret;
+}
 
-        data = dict_get (dict, "bricks");
-        if (!data)
-                goto out;
+int
+glusterd_friend_add (const char *hoststr, int port,
+                     glusterd_friend_sm_state_t state,
+                     uuid_t *uuid,
+                     glusterd_peerinfo_t **friend,
+                     gf_boolean_t restore,
+                     glusterd_peerctx_args_t *args)
+{
+        int                     ret = 0;
+        xlator_t               *this = NULL;
+        glusterd_conf_t        *conf = NULL;
 
-        data = dict_get (dict, "transport");
-        if (!data)
-                goto out;
+        this = THIS;
+        conf = this->private;
+        GF_ASSERT (conf);
+        GF_ASSERT (hoststr);
 
-        data = dict_get (dict, "volume-id");
-        if (!data)
+        ret = glusterd_peerinfo_new (friend, state, uuid, hoststr, port);
+        if (ret) {
                 goto out;
+        }
+
+        /*
+         * We can't add to the list after calling glusterd_friend_rpc_create,
+         * even if it succeeds, because by then the callback to take it back
+         * off and free might have happened already (notably in the case of an
+         * invalid peer name).  That would mean we're adding something that had
+         * just been free, and we're likely to crash later.
+         */
+        list_add_tail (&(*friend)->uuid_list, &conf->peers);
+
+        //restore needs to first create the list of peers, then create rpcs
+        //to keep track of quorum in race-free manner. In restore for each peer
+        //rpc-create calls rpc_notify when the friend-list is partially
+        //constructed, leading to wrong quorum calculations.
+        if (!restore) {
+                ret = glusterd_store_peerinfo (*friend);
+                if (ret == 0) {
+                        synclock_unlock (&conf->big_lock);
+                        ret = glusterd_friend_rpc_create (this, *friend, args);
+                        synclock_lock (&conf->big_lock);
+                }
+                else {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to store peerinfo");
+                }
+        }
 
-        ret = glusterd_op_txn_begin ();
+        if (ret) {
+                (void) glusterd_friend_cleanup (*friend);
+                *friend = NULL;
+        }
 
 out:
+        gf_log (this->name, GF_LOG_INFO, "connect returned %d", ret);
         return ret;
 }
 
-int32_t
-glusterd_start_volume (rpcsvc_request_t *req, char *volname, int flags)
+int
+glusterd_probe_begin (rpcsvc_request_t *req, const char *hoststr, int port,
+                      dict_t *dict)
 {
-        int32_t      ret       = -1;
-        glusterd_op_start_volume_ctx_t  *ctx = NULL;
-
-        GF_ASSERT (req);
-        GF_ASSERT (volname);
-
-        ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_start_volume_ctx_t);
-
-        if (!ctx)
-                goto out;
+        int                             ret = -1;
+        glusterd_peerinfo_t             *peerinfo = NULL;
+        glusterd_peerctx_args_t         args = {0};
+        glusterd_friend_sm_event_t      *event = NULL;
 
-        strncpy (ctx->volume_name, volname, GD_VOLUME_NAME_MAX);
+        GF_ASSERT (hoststr);
 
-        glusterd_op_set_op (GD_OP_START_VOLUME);
+        ret = glusterd_friend_find (NULL, (char *)hoststr, &peerinfo);
 
-        glusterd_op_set_ctx (GD_OP_START_VOLUME, ctx);
-        glusterd_op_set_ctx_free (GD_OP_START_VOLUME, _gf_true);
-        glusterd_op_set_req (req);
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_INFO, "Unable to find peerinfo"
+                        " for host: %s (%d)", hoststr, port);
+                args.mode = GD_MODE_ON;
+                args.req  = req;
+                args.dict = dict;
+                ret = glusterd_friend_add ((char *)hoststr, port,
+                                           GD_FRIEND_STATE_DEFAULT,
+                                           NULL, &peerinfo, 0, &args);
+                if ((!ret) && (!peerinfo->connected)) {
+                        ret = GLUSTERD_CONNECTION_AWAITED;
+                }
 
-        ret = glusterd_op_txn_begin ();
+        } else if (peerinfo->connected &&
+                   (GD_FRIEND_STATE_BEFRIENDED == peerinfo->state.state)) {
+                ret = glusterd_friend_hostname_update (peerinfo, (char*)hoststr,
+                                                       _gf_false);
+                if (ret)
+                        goto out;
+                //this is just to rename so inject local acc for cluster update
+                ret = glusterd_friend_sm_new_event (GD_FRIEND_EVENT_LOCAL_ACC,
+                                                    &event);
+                if (!ret) {
+                        event->peerinfo = peerinfo;
+                        ret = glusterd_friend_sm_inject_event (event);
+                        glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_SUCCESS,
+                                                      NULL, (char*)hoststr,
+                                                      port, dict);
+                }
+        } else {
+                glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_FRIEND, NULL,
+                                              (char*)hoststr, port, dict);
+        }
 
 out:
+        gf_log ("", GF_LOG_DEBUG, "returning %d", ret);
         return ret;
 }
 
-int32_t
-glusterd_stop_volume (rpcsvc_request_t *req, char *volname, int flags)
+int
+glusterd_deprobe_begin (rpcsvc_request_t *req, const char *hoststr, int port,
+                        uuid_t uuid, dict_t *dict)
 {
-        int32_t      ret        = -1;
-        dict_t       *ctx       = NULL;
-        char         *dup_volname = NULL;
+        int                             ret = -1;
+        glusterd_peerinfo_t             *peerinfo = NULL;
+        glusterd_friend_sm_event_t      *event = NULL;
+        glusterd_probe_ctx_t            *ctx = NULL;
 
+        GF_ASSERT (hoststr);
         GF_ASSERT (req);
-        GF_ASSERT (volname);
 
-        ctx = dict_new ();
-
-        if (!ctx)
-                goto out;
+        ret = glusterd_friend_find (uuid, (char *)hoststr, &peerinfo);
 
-        dup_volname = gf_strdup(volname);
-        if (!dup_volname)
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_INFO, "Unable to find peerinfo"
+                        " for host: %s %d", hoststr, port);
                 goto out;
+        }
 
-        ret = dict_set_dynstr (ctx, "volname", dup_volname);
-        if (ret)
+        if (!peerinfo->rpc) {
+                //handle this case
                 goto out;
+        }
 
-        ret = dict_set_int32 (ctx, "flags", flags);
-        if (ret)
-                goto out;
+        ret = glusterd_friend_sm_new_event
+                (GD_FRIEND_EVENT_INIT_REMOVE_FRIEND, &event);
 
-        glusterd_op_set_op (GD_OP_STOP_VOLUME);
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR,
+                                "Unable to get new event");
+                return ret;
+        }
 
-        glusterd_op_set_ctx (GD_OP_STOP_VOLUME, ctx);
-        glusterd_op_set_ctx_free (GD_OP_STOP_VOLUME, _gf_true);
-        glusterd_op_set_req (req);
+        ctx = GF_CALLOC (1, sizeof(*ctx), gf_gld_mt_probe_ctx_t);
 
-        ret = glusterd_op_txn_begin ();
+        if (!ctx) {
+                goto out;
+        }
 
-out:
-        if (ret && ctx)
-                dict_unref (ctx);
-        return ret;
-}
+        ctx->hostname = gf_strdup (hoststr);
+        ctx->port = port;
+        ctx->req = req;
+        ctx->dict = dict;
 
-int32_t
-glusterd_delete_volume (rpcsvc_request_t *req, char *volname, int flags)
-{
-        int32_t      ret       = -1;
-        glusterd_op_delete_volume_ctx_t  *ctx = NULL;
+        event->ctx = ctx;
 
-        GF_ASSERT (req);
-        GF_ASSERT (volname);
+        event->peerinfo = peerinfo;
 
-        ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_delete_volume_ctx_t);
+        ret = glusterd_friend_sm_inject_event (event);
 
-        if (!ctx)
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Unable to inject event %d, "
+                        "ret = %d", event->event, ret);
                 goto out;
-
-        strncpy (ctx->volume_name, volname, GD_VOLUME_NAME_MAX);
-
-        glusterd_op_set_op (GD_OP_DELETE_VOLUME);
-
-        glusterd_op_set_ctx (GD_OP_DELETE_VOLUME, ctx);
-        glusterd_op_set_ctx_free (GD_OP_DELETE_VOLUME, _gf_true);
-        glusterd_op_set_req (req);
-
-        ret = glusterd_op_txn_begin ();
+        }
 
 out:
         return ret;
 }
 
-int32_t
-glusterd_add_brick (rpcsvc_request_t *req, dict_t *dict)
+
+int
+glusterd_xfer_friend_remove_resp (rpcsvc_request_t *req, char *hostname, int port)
 {
-        int32_t      ret       = -1;
+        gd1_mgmt_friend_rsp  rsp = {{0}, };
+        int32_t              ret = -1;
+        xlator_t             *this = NULL;
+        glusterd_conf_t      *conf = NULL;
 
-        GF_ASSERT (req);
-        GF_ASSERT (dict);
+        GF_ASSERT (hostname);
 
-        glusterd_op_set_op (GD_OP_ADD_BRICK);
+        rsp.op_ret = 0;
+        this = THIS;
+        GF_ASSERT (this);
 
-        glusterd_op_set_ctx (GD_OP_ADD_BRICK, dict);
-        glusterd_op_set_ctx_free (GD_OP_ADD_BRICK, _gf_true);
-        glusterd_op_set_req (req);
+        conf = this->private;
 
-        ret = glusterd_op_txn_begin ();
+        uuid_copy (rsp.uuid, MY_UUID);
+        rsp.hostname = hostname;
+        rsp.port = port;
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gd1_mgmt_friend_rsp);
 
+        gf_log ("glusterd", GF_LOG_INFO,
+                "Responded to %s (%d), ret: %d", hostname, port, ret);
         return ret;
 }
 
-int32_t
-glusterd_replace_brick (rpcsvc_request_t *req, dict_t *dict)
+
+int
+glusterd_xfer_friend_add_resp (rpcsvc_request_t *req, char *myhostname,
+                               char *remote_hostname, int port, int32_t op_ret,
+                               int32_t op_errno)
 {
-        int32_t      ret       = -1;
+        gd1_mgmt_friend_rsp  rsp = {{0}, };
+        int32_t              ret = -1;
+        xlator_t             *this = NULL;
+        glusterd_conf_t      *conf = NULL;
 
-        GF_ASSERT (req);
-        GF_ASSERT (dict);
+        GF_ASSERT (myhostname);
 
-        glusterd_op_set_op (GD_OP_REPLACE_BRICK);
+        this = THIS;
+        GF_ASSERT (this);
 
-        glusterd_op_set_ctx (GD_OP_REPLACE_BRICK, dict);
+        conf = this->private;
 
-        glusterd_op_set_ctx_free (GD_OP_REPLACE_BRICK, _gf_true);
-        glusterd_op_set_req (req);
+        uuid_copy (rsp.uuid, MY_UUID);
+        rsp.op_ret = op_ret;
+        rsp.op_errno = op_errno;
+        rsp.hostname = gf_strdup (myhostname);
+        rsp.port = port;
 
-        ret = glusterd_op_txn_begin ();
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gd1_mgmt_friend_rsp);
 
+        gf_log ("glusterd", GF_LOG_INFO,
+                "Responded to %s (%d), ret: %d", remote_hostname, port, ret);
+        GF_FREE (rsp.hostname);
         return ret;
 }
 
-int32_t
-glusterd_set_volume (rpcsvc_request_t *req, dict_t *dict)
+static void
+set_probe_error_str (int op_ret, int op_errno, char *op_errstr, char *errstr,
+                     size_t len, char *hostname, int port)
 {
-        int32_t      ret       = -1;
+        if ((op_errstr) && (strcmp (op_errstr, ""))) {
+                snprintf (errstr, len, "%s", op_errstr);
+                return;
+        }
 
-        GF_ASSERT (req);
-        GF_ASSERT (dict);
+        if (!op_ret) {
+                switch (op_errno) {
+                        case GF_PROBE_LOCALHOST:
+                                snprintf (errstr, len, "Probe on localhost not "
+                                          "needed");
+                                break;
 
-        glusterd_op_set_op (GD_OP_SET_VOLUME);
+                        case GF_PROBE_FRIEND:
+                                snprintf (errstr, len, "Host %s port %d already"
+                                          " in peer list", hostname, port);
+                                break;
 
-        glusterd_op_set_ctx (GD_OP_SET_VOLUME, dict);
+                        default:
+                                if (op_errno != 0)
+                                        snprintf (errstr, len, "Probe returned "
+                                                  "with unknown errno %d",
+                                                  op_errno);
+                                break;
+                }
+        } else {
+                switch (op_errno) {
+                        case GF_PROBE_ANOTHER_CLUSTER:
+                                snprintf (errstr, len, "%s is already part of "
+                                          "another cluster", hostname);
+                                break;
+
+                        case GF_PROBE_VOLUME_CONFLICT:
+                                snprintf (errstr, len, "Atleast one volume on "
+                                          "%s conflicts with existing volumes "
+                                          "in the cluster", hostname);
+                                break;
 
-        glusterd_op_set_ctx_free (GD_OP_SET_VOLUME, _gf_true);
+                        case GF_PROBE_UNKNOWN_PEER:
+                                snprintf (errstr, len, "%s responded with "
+                                          "'unknown peer' error, this could "
+                                          "happen if %s doesn't have localhost "
+                                          "in its peer database", hostname,
+                                          hostname);
+                                break;
 
-	glusterd_op_set_cli_op (GD_MGMT_CLI_SET_VOLUME);
+                        case GF_PROBE_ADD_FAILED:
+                                snprintf (errstr, len, "Failed to add peer "
+                                          "information on %s", hostname);
+                                break;
 
-	glusterd_op_set_req (req);
+                        case GF_PROBE_SAME_UUID:
+                                snprintf (errstr, len, "Peer uuid (host %s) is "
+                                          "same as local uuid", hostname);
+                                break;
 
-        ret = glusterd_op_txn_begin ();
+                        case GF_PROBE_QUORUM_NOT_MET:
+                                snprintf (errstr, len, "Cluster quorum is not "
+                                          "met. Changing peers is not allowed "
+                                          "in this state");
+                                break;
 
-        return ret;
+                        default:
+                                snprintf (errstr, len, "Probe returned with "
+                                          "unknown errno %d", op_errno);
+                                break;
+                }
+        }
 }
 
-int32_t
-glusterd_remove_brick (rpcsvc_request_t *req, dict_t *dict)
+int
+glusterd_xfer_cli_probe_resp (rpcsvc_request_t *req, int32_t op_ret,
+                              int32_t op_errno, char *op_errstr, char *hostname,
+                              int port, dict_t *dict)
 {
-        int32_t      ret       = -1;
+        gf_cli_rsp           rsp = {0,};
+        int32_t              ret = -1;
+        char        errstr[2048] = {0,};
+        char            *cmd_str = NULL;
+        xlator_t           *this = THIS;
 
         GF_ASSERT (req);
-        GF_ASSERT (dict);
+        GF_ASSERT (this);
+
+        (void) set_probe_error_str (op_ret, op_errno, op_errstr, errstr,
+                                    sizeof (errstr), hostname, port);
 
-        glusterd_op_set_op (GD_OP_REMOVE_BRICK);
+        if (dict) {
+                ret = dict_get_str (dict, "cmd-str", &cmd_str);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                "command string");
+        }
+
+        rsp.op_ret = op_ret;
+        rsp.op_errno = op_errno;
+        rsp.op_errstr = (errstr[0] != '\0') ? errstr : "";
+
+        gf_cmd_log ("", "%s : %s %s %s", cmd_str,
+                    (op_ret) ? "FAILED" : "SUCCESS",
+                    (errstr[0] != '\0') ? ":" : " ",
+                    (errstr[0] != '\0') ? errstr : " ");
 
-        glusterd_op_set_ctx (GD_OP_REMOVE_BRICK, dict);
-        glusterd_op_set_ctx_free (GD_OP_REMOVE_BRICK, _gf_true);
-        glusterd_op_set_req (req);
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gf_cli_rsp);
 
-        ret = glusterd_op_txn_begin ();
+        if (dict)
+                dict_unref (dict);
+        gf_log (this->name, GF_LOG_DEBUG, "Responded to CLI, ret: %d",ret);
 
         return ret;
 }
 
-int32_t
-glusterd_log_filename (rpcsvc_request_t *req, dict_t *dict)
+static void
+set_deprobe_error_str (int op_ret, int op_errno, char *op_errstr, char *errstr,
+                       size_t len, char *hostname)
 {
-        int32_t      ret       = -1;
+        if ((op_errstr) && (strcmp (op_errstr, ""))) {
+                snprintf (errstr, len, "%s", op_errstr);
+                return;
+        }
 
-        GF_ASSERT (req);
-        GF_ASSERT (dict);
+        if (op_ret) {
+                switch (op_errno) {
+                        case GF_DEPROBE_LOCALHOST:
+                                snprintf (errstr, len, "%s is localhost",
+                                hostname);
+                                break;
 
-        glusterd_op_set_op (GD_OP_LOG_FILENAME);
-        glusterd_op_set_ctx (GD_OP_LOG_FILENAME, dict);
-        glusterd_op_set_ctx_free (GD_OP_LOG_FILENAME, _gf_true);
-        glusterd_op_set_req (req);
+                        case GF_DEPROBE_NOT_FRIEND:
+                                snprintf (errstr, len, "%s is not part of "
+                                          "cluster", hostname);
+                                break;
 
-        ret = glusterd_op_txn_begin ();
+                        case GF_DEPROBE_BRICK_EXIST:
+                                snprintf (errstr, len, "Brick(s) with the peer "
+                                          "%s exist in cluster", hostname);
+                                break;
 
-        return ret;
+                        case GF_DEPROBE_FRIEND_DOWN:
+                                snprintf (errstr, len, "One of the peers is "
+                                          "probably down. Check with "
+                                          "'peer status'");
+                                break;
+
+                        case GF_DEPROBE_QUORUM_NOT_MET:
+                                snprintf (errstr, len, "Cluster quorum is not "
+                                          "met. Changing peers is not allowed "
+                                          "in this state");
+                                break;
+
+                        default:
+                                snprintf (errstr, len, "Detach returned with "
+                                          "unknown errno %d", op_errno);
+                                break;
+
+                }
+        }
 }
 
 
-int32_t
-glusterd_log_rotate (rpcsvc_request_t *req, dict_t *dict)
+int
+glusterd_xfer_cli_deprobe_resp (rpcsvc_request_t *req, int32_t op_ret,
+                                int32_t op_errno, char *op_errstr,
+                                char *hostname, dict_t *dict)
 {
-        int32_t      ret       = -1;
+        gf_cli_rsp             rsp = {0,};
+        int32_t                ret = -1;
+        char              *cmd_str = NULL;
+        char          errstr[2048] = {0,};
 
         GF_ASSERT (req);
-        GF_ASSERT (dict);
-
-        glusterd_op_set_op (GD_OP_LOG_ROTATE);
-        glusterd_op_set_ctx (GD_OP_LOG_ROTATE, dict);
-        glusterd_op_set_ctx_free (GD_OP_LOG_ROTATE, _gf_true);
-        glusterd_op_set_req (req);
 
-        ret = glusterd_op_txn_begin ();
+        (void) set_deprobe_error_str (op_ret, op_errno, op_errstr, errstr,
+                                      sizeof (errstr), hostname);
 
-        return ret;
-}
+        if (dict) {
+                ret = dict_get_str (dict, "cmd-str", &cmd_str);
+                if (ret)
+                        gf_log (THIS->name, GF_LOG_ERROR, "Failed to get "
+                                "command string");
+        }
 
-int32_t
-glusterd_sync_volume (rpcsvc_request_t *req, dict_t *ctx)
-{
-        int32_t      ret       = -1;
+        rsp.op_ret = op_ret;
+        rsp.op_errno = op_errno;
+        rsp.op_errstr = (errstr[0] != '\0') ? errstr : "";
 
-        GF_ASSERT (req);
-        GF_ASSERT (ctx);
+        gf_cmd_log ("", "%s : %s %s %s", cmd_str,
+                    (op_ret) ? "FAILED" : "SUCCESS",
+                    (errstr[0] != '\0') ? ":" : " ",
+                    (errstr[0] != '\0') ? errstr : " ");
 
-        glusterd_op_set_op (GD_OP_SYNC_VOLUME);
-        glusterd_op_set_ctx (GD_OP_SYNC_VOLUME, ctx);
-        glusterd_op_set_ctx_free (GD_OP_SYNC_VOLUME, _gf_true);
-        glusterd_op_set_req (req);
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gf_cli_rsp);
 
-        ret = glusterd_op_txn_begin ();
+        gf_log (THIS->name, GF_LOG_DEBUG, "Responded to CLI, ret: %d",ret);
 
         return ret;
 }
 
-
 int32_t
 glusterd_list_friends (rpcsvc_request_t *req, dict_t *dict, int32_t flags)
 {
@@ -3247,39 +3514,52 @@ glusterd_list_friends (rpcsvc_request_t *req, dict_t *dict, int32_t flags)
         int32_t                 count = 0;
         dict_t                  *friends = NULL;
         gf1_cli_peer_list_rsp   rsp = {0,};
+        char                    my_uuid_str[64] = {0,};
+        char                    key[256] = {0,};
 
         priv = THIS->private;
         GF_ASSERT (priv);
 
-        if (!list_empty (&priv->peers)) {
-                friends = dict_new ();
-                if (!friends) {
-                        gf_log ("", GF_LOG_WARNING, "Out of Memory");
-                        goto out;
-                }
-        } else {
-                ret = 0;
+        friends = dict_new ();
+        if (!friends) {
+                gf_log ("", GF_LOG_WARNING, "Out of Memory");
                 goto out;
         }
-
-        if (flags == GF_CLI_LIST_ALL) {
-                        list_for_each_entry (entry, &priv->peers, uuid_list) {
-                                count++;
-                                ret = glusterd_add_peer_detail_to_dict (entry,
+        if (!list_empty (&priv->peers)) {
+                list_for_each_entry (entry, &priv->peers, uuid_list) {
+                        count++;
+                        ret = glusterd_add_peer_detail_to_dict (entry,
                                                                 friends, count);
-                                if (ret)
-                                        goto out;
+                        if (ret)
+                                goto out;
+                }
+        }
 
-                        }
+        if (flags == GF_CLI_LIST_POOL_NODES) {
+                count++;
+                snprintf (key, 256, "friend%d.uuid", count);
+                uuid_utoa_r (MY_UUID, my_uuid_str);
+                ret = dict_set_str (friends, key, my_uuid_str);
+                if (ret)
+                        goto out;
 
-                        ret = dict_set_int32 (friends, "count", count);
+                snprintf (key, 256, "friend%d.hostname", count);
+                ret = dict_set_str (friends, key, "localhost");
+                if (ret)
+                        goto out;
 
-                        if (ret)
-                                goto out;
+                snprintf (key, 256, "friend%d.connected", count);
+                ret = dict_set_int32 (friends, key, 1);
+                if (ret)
+                        goto out;
         }
 
+        ret = dict_set_int32 (friends, "count", count);
+        if (ret)
+                goto out;
+
         ret = dict_allocate_and_serialize (friends, &rsp.friends.friends_val,
-                                           (size_t *)&rsp.friends.friends_len);
+                                           &rsp.friends.friends_len);
 
         if (ret)
                 goto out;
@@ -3292,10 +3572,10 @@ out:
 
         rsp.op_ret = ret;
 
-        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gf_xdr_serialize_cli_peer_list_rsp);
-        if (rsp.friends.friends_val)
-                GF_FREE (rsp.friends.friends_val);
+        glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                               (xdrproc_t)xdr_gf1_cli_peer_list_rsp);
+        ret = 0;
+        GF_FREE (rsp.friends.friends_val);
 
         return ret;
 }
@@ -3308,7 +3588,7 @@ glusterd_get_volumes (rpcsvc_request_t *req, dict_t *dict, int32_t flags)
         glusterd_volinfo_t      *entry = NULL;
         int32_t                 count = 0;
         dict_t                  *volumes = NULL;
-        gf1_cli_get_vol_rsp     rsp = {0,};
+        gf_cli_rsp              rsp = {0,};
         char                    *volname = NULL;
 
         priv = THIS->private;
@@ -3385,9 +3665,8 @@ respond:
         ret = dict_set_int32 (volumes, "count", count);
         if (ret)
                 goto out;
-
-        ret = dict_allocate_and_serialize (volumes, &rsp.volumes.volumes_val,
-                                           (size_t *)&rsp.volumes.volumes_len);
+        ret = dict_allocate_and_serialize (volumes, &rsp.dict.dict_val,
+                                           &rsp.dict.dict_len);
 
         if (ret)
                 goto out;
@@ -3396,61 +3675,435 @@ respond:
 out:
         rsp.op_ret = ret;
 
-        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                                     gf_xdr_serialize_cli_peer_list_rsp);
+        rsp.op_errstr = "";
+        glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                               (xdrproc_t)xdr_gf_cli_rsp);
+        ret = 0;
 
         if (volumes)
                 dict_unref (volumes);
 
-        if (rsp.volumes.volumes_val)
-                GF_FREE (rsp.volumes.volumes_val);
+        GF_FREE (rsp.dict.dict_val);
+        return ret;
+}
+
+int
+__glusterd_handle_status_volume (rpcsvc_request_t *req)
+{
+        int32_t                         ret     = -1;
+        uint32_t                        cmd     = 0;
+        dict_t                         *dict    = NULL;
+        char                           *volname = 0;
+        gf_cli_req                      cli_req = {{0,}};
+        glusterd_op_t                   cli_op  = GD_OP_STATUS_VOLUME;
+        char                            err_str[2048] = {0,};
+        xlator_t                       *this = NULL;
+
+        GF_ASSERT (req);
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (cli_req.dict.dict_len > 0) {
+                dict = dict_new();
+                if (!dict)
+                        goto out;
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len, &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                "unserialize buffer");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
+                        goto out;
+                }
+
+        }
+
+        ret = dict_get_uint32 (dict, "cmd", &cmd);
+        if (ret)
+                goto out;
+
+        if (!(cmd & GF_CLI_STATUS_ALL)) {
+                ret = dict_get_str (dict, "volname", &volname);
+                if (ret) {
+                        snprintf (err_str, sizeof (err_str), "Unable to get "
+                                  "volume name");
+                        gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                        goto out;
+                }
+                gf_log (this->name, GF_LOG_INFO,
+                        "Received status volume req for volume %s", volname);
+
+        }
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_STATUS_VOLUME, dict);
+
+out:
+
+        if (ret) {
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
+        }
+        free (cli_req.dict.dict_val);
+
+        return ret;
+}
+
+int
+glusterd_handle_status_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_status_volume);
+}
+
+int
+__glusterd_handle_cli_clearlocks_volume (rpcsvc_request_t *req)
+{
+        int32_t                         ret = -1;
+        gf_cli_req                      cli_req = {{0,}};
+        glusterd_op_t                   cli_op = GD_OP_CLEARLOCKS_VOLUME;
+        char                            *volname = NULL;
+        dict_t                          *dict = NULL;
+        char                            err_str[2048] = {0,};
+        xlator_t                        *this = NULL;
+
+        GF_ASSERT (req);
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = -1;
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (cli_req.dict.dict_len) {
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to unserialize req-buffer to"
+                                " dictionary");
+                        snprintf (err_str, sizeof (err_str), "unable to decode "
+                                  "the command");
+                        goto out;
+                }
+
+        } else {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR, "Empty cli request.");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get volume "
+                          "name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_INFO, "Received clear-locks volume req "
+                "for volume %s", volname);
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_CLEARLOCKS_VOLUME, dict);
+
+out:
+        if (ret) {
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
+        }
+        free (cli_req.dict.dict_val);
+
+        return ret;
+}
+
+int
+glusterd_handle_cli_clearlocks_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_clearlocks_volume);
+}
+
+static int
+get_brickinfo_from_brickid (char *brickid, glusterd_brickinfo_t **brickinfo)
+{
+        glusterd_volinfo_t      *volinfo    = NULL;
+        char                    *volid_str  = NULL;
+        char                    *brick      = NULL;
+        char                    *brickid_dup = NULL;
+        uuid_t                  volid       = {0};
+        int                     ret         = -1;
+
+        brickid_dup = gf_strdup (brickid);
+        if (!brickid_dup)
+                goto out;
+
+        volid_str = brickid_dup;
+        brick = strchr (brickid_dup, ':');
+        *brick = '\0';
+        brick++;
+        if (!volid_str || !brick)
+                goto out;
+
+        uuid_parse (volid_str, volid);
+        ret = glusterd_volinfo_find_by_volume_id (volid, &volinfo);
+        if (ret) {
+                /* Check if it a snapshot volume */
+                ret = glusterd_snap_volinfo_find_by_volume_id (volid, &volinfo);
+                if (ret)
+                        goto out;
+        }
+
+        ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo,
+                                                      brickinfo);
+        if (ret)
+                goto out;
+
+        ret = 0;
+out:
+        GF_FREE (brickid_dup);
         return ret;
 }
 
 int
-glusterd_rpc_notify (struct rpc_clnt *rpc, void *mydata, rpc_clnt_event_t event,
-                     void *data)
+__glusterd_brick_rpc_notify (struct rpc_clnt *rpc, void *mydata,
+                          rpc_clnt_event_t event, void *data)
 {
         xlator_t                *this = NULL;
-        char                    *handshake = "on";
         glusterd_conf_t         *conf = NULL;
         int                     ret = 0;
-        glusterd_peerinfo_t     *peerinfo = NULL;
+        char                    *brickid = NULL;
+        glusterd_brickinfo_t    *brickinfo = NULL;
+
+        brickid = mydata;
+        if (!brickid)
+                return 0;
+
+        ret = get_brickinfo_from_brickid (brickid, &brickinfo);
+        if (ret)
+                return 0;
 
-        peerinfo = mydata;
         this = THIS;
+        GF_ASSERT (this);
         conf = this->private;
-
+        GF_ASSERT (conf);
 
         switch (event) {
         case RPC_CLNT_CONNECT:
-        {
+                gf_log (this->name, GF_LOG_DEBUG, "Connected to %s:%s",
+                        brickinfo->hostname, brickinfo->path);
+                glusterd_set_brick_status (brickinfo, GF_BRICK_STARTED);
+                ret = default_notify (this, GF_EVENT_CHILD_UP, NULL);
+
+                break;
+
+        case RPC_CLNT_DISCONNECT:
+                if (GF_BRICK_STARTED == brickinfo->status)
+                        gf_log (this->name, GF_LOG_INFO, "Disconnected from "
+                                "%s:%s", brickinfo->hostname, brickinfo->path);
+
+                glusterd_set_brick_status (brickinfo, GF_BRICK_STOPPED);
+                if (rpc_clnt_is_disabled (rpc))
+                        GF_FREE (brickid);
+                break;
+
+        default:
+                gf_log (this->name, GF_LOG_TRACE,
+                        "got some other RPC event %d", event);
+                break;
+        }
+
+        return ret;
+}
+
+int
+glusterd_brick_rpc_notify (struct rpc_clnt *rpc, void *mydata,
+                           rpc_clnt_event_t event, void *data)
+{
+        return glusterd_big_locked_notify (rpc, mydata, event, data,
+                                           __glusterd_brick_rpc_notify);
+}
 
+int
+__glusterd_nodesvc_rpc_notify (struct rpc_clnt *rpc, void *mydata,
+                               rpc_clnt_event_t event, void *data)
+{
+        xlator_t                *this = NULL;
+        glusterd_conf_t         *conf = NULL;
+        char                    *server = NULL;
+        int                     ret = 0;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        server = mydata;
+        if (!server)
+                return 0;
+
+        switch (event) {
+        case RPC_CLNT_CONNECT:
                 gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_CONNECT");
-                peerinfo->connected = 1;
-                glusterd_friend_sm ();
-                glusterd_op_sm ();
+                (void) glusterd_nodesvc_set_online_status (server, _gf_true);
+                ret = default_notify (this, GF_EVENT_CHILD_UP, NULL);
 
-                if ((ret < 0) || (strcasecmp (handshake, "on"))) {
-                        //ret = glusterd_handshake (this, peerinfo->rpc);
+                break;
 
-                } else {
-                        //conf->rpc->connected = 1;
-                        ret = default_notify (this, GF_EVENT_CHILD_UP, NULL);
+        case RPC_CLNT_DISCONNECT:
+                gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_DISCONNECT");
+                (void) glusterd_nodesvc_set_online_status (server, _gf_false);
+                break;
+
+        default:
+                gf_log (this->name, GF_LOG_TRACE,
+                        "got some other RPC event %d", event);
+                break;
+        }
+
+        return ret;
+}
+
+int
+glusterd_nodesvc_rpc_notify (struct rpc_clnt *rpc, void *mydata,
+                             rpc_clnt_event_t event, void *data)
+{
+        return glusterd_big_locked_notify (rpc, mydata, event, data,
+                                           __glusterd_nodesvc_rpc_notify);
+}
+
+int
+glusterd_friend_remove_notify (glusterd_peerctx_t *peerctx)
+{
+        int                             ret = -1;
+        glusterd_friend_sm_event_t      *new_event = NULL;
+        glusterd_peerinfo_t             *peerinfo = peerctx->peerinfo;
+        rpcsvc_request_t                *req = peerctx->args.req;
+        char                            *errstr = peerctx->errstr;
+        dict_t                          *dict = NULL;
+
+        GF_ASSERT (peerctx);
+
+        peerinfo = peerctx->peerinfo;
+        req = peerctx->args.req;
+        dict = peerctx->args.dict;
+        errstr = peerctx->errstr;
+
+        ret = glusterd_friend_sm_new_event (GD_FRIEND_EVENT_REMOVE_FRIEND,
+                                            &new_event);
+        if (!ret) {
+                if (!req) {
+                        gf_log (THIS->name, GF_LOG_WARNING,
+                                "Unable to find the request for responding "
+                                "to User (%s)", peerinfo->hostname);
+                        goto out;
                 }
+
+                glusterd_xfer_cli_probe_resp (req, -1, ENOTCONN, errstr,
+                                              peerinfo->hostname,
+                                              peerinfo->port, dict);
+
+                new_event->peerinfo = peerinfo;
+                ret = glusterd_friend_sm_inject_event (new_event);
+
+        } else {
+                gf_log ("glusterd", GF_LOG_ERROR,
+                        "Unable to create event for removing peer %s",
+                        peerinfo->hostname);
+        }
+
+out:
+        return ret;
+}
+
+int
+__glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata,
+                            rpc_clnt_event_t event, void *data)
+{
+        xlator_t             *this        = NULL;
+        glusterd_conf_t      *conf        = NULL;
+        int                   ret         = 0;
+        glusterd_peerinfo_t  *peerinfo    = NULL;
+        glusterd_peerctx_t   *peerctx     = NULL;
+        gf_boolean_t         quorum_action = _gf_false;
+        glusterd_volinfo_t   *volinfo     = NULL;
+
+        peerctx = mydata;
+        if (!peerctx)
+                return 0;
+
+        peerinfo = peerctx->peerinfo;
+        this = THIS;
+        conf = this->private;
+
+        switch (event) {
+        case RPC_CLNT_CONNECT:
+        {
+                gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_CONNECT");
+                peerinfo->connected = 1;
+                peerinfo->quorum_action = _gf_true;
+
+                ret = glusterd_peer_dump_version (this, rpc, peerctx);
+                if (ret)
+                        gf_log ("", GF_LOG_ERROR, "glusterd handshake failed");
                 break;
         }
 
         case RPC_CLNT_DISCONNECT:
+        {
+                gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_DISCONNECT %d",
+                        peerinfo->state.state);
+
+                if (peerinfo->connected) {
+                        list_for_each_entry (volinfo, &conf->volumes, vol_list) {
+                                ret = glusterd_mgmt_v3_unlock (volinfo->volname,
+                                                               peerinfo->uuid,
+                                                               "vol");
+                                if (ret)
+                                        gf_log (this->name, GF_LOG_TRACE,
+                                                "Lock not released for %s",
+                                                volinfo->volname);
+                        }
 
-                //Inject friend disconnected here
+                        ret = 0;
+                }
 
-                gf_log (this->name, GF_LOG_DEBUG, "got RPC_CLNT_DISCONNECT");
-                peerinfo->connected = 0;
+                if ((peerinfo->quorum_contrib != QUORUM_DOWN) &&
+                    (peerinfo->state.state == GD_FRIEND_STATE_BEFRIENDED)) {
+                        peerinfo->quorum_contrib = QUORUM_DOWN;
+                        quorum_action = _gf_true;
+                        peerinfo->quorum_action = _gf_false;
+                }
 
-                //default_notify (this, GF_EVENT_CHILD_DOWN, NULL);
-                break;
+                /* Remove peer if it is not a friend and connection/handshake
+                *  fails, and notify cli. Happens only during probe.
+                */
+                if (peerinfo->state.state == GD_FRIEND_STATE_DEFAULT) {
+                        glusterd_friend_remove_notify (peerctx);
+                        goto out;
+                }
 
+                peerinfo->connected = 0;
+                break;
+        }
         default:
                 gf_log (this->name, GF_LOG_TRACE,
                         "got some other RPC event %d", event);
@@ -3458,5 +4111,127 @@ glusterd_rpc_notify (struct rpc_clnt *rpc, void *mydata, rpc_clnt_event_t event,
                 break;
         }
 
+out:
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+        if (quorum_action)
+                glusterd_do_quorum_action ();
         return ret;
 }
+
+int
+glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata,
+                          rpc_clnt_event_t event, void *data)
+{
+        return glusterd_big_locked_notify (rpc, mydata, event, data,
+                                           __glusterd_peer_rpc_notify);
+}
+
+int
+glusterd_null (rpcsvc_request_t *req)
+{
+
+        return 0;
+}
+
+rpcsvc_actor_t gd_svc_mgmt_actors[] = {
+        [GLUSTERD_MGMT_NULL]           = { "NULL",           GLUSTERD_MGMT_NULL,           glusterd_null,                  NULL, 0, DRC_NA},
+        [GLUSTERD_MGMT_CLUSTER_LOCK]   = { "CLUSTER_LOCK",   GLUSTERD_MGMT_CLUSTER_LOCK,   glusterd_handle_cluster_lock,   NULL, 0, DRC_NA},
+        [GLUSTERD_MGMT_CLUSTER_UNLOCK] = { "CLUSTER_UNLOCK", GLUSTERD_MGMT_CLUSTER_UNLOCK, glusterd_handle_cluster_unlock, NULL, 0, DRC_NA},
+        [GLUSTERD_MGMT_STAGE_OP]       = { "STAGE_OP",       GLUSTERD_MGMT_STAGE_OP,       glusterd_handle_stage_op,       NULL, 0, DRC_NA},
+        [GLUSTERD_MGMT_COMMIT_OP]      = { "COMMIT_OP",      GLUSTERD_MGMT_COMMIT_OP,      glusterd_handle_commit_op,      NULL, 0, DRC_NA},
+};
+
+struct rpcsvc_program gd_svc_mgmt_prog = {
+        .progname  = "GlusterD svc mgmt",
+        .prognum   = GD_MGMT_PROGRAM,
+        .progver   = GD_MGMT_VERSION,
+        .numactors = GLUSTERD_MGMT_MAXVALUE,
+        .actors    = gd_svc_mgmt_actors,
+	.synctask  = _gf_true,
+};
+
+rpcsvc_actor_t gd_svc_peer_actors[] = {
+        [GLUSTERD_FRIEND_NULL]    = { "NULL",          GLUSTERD_MGMT_NULL,     glusterd_null,                         NULL, 0, DRC_NA},
+        [GLUSTERD_PROBE_QUERY]    = { "PROBE_QUERY",   GLUSTERD_PROBE_QUERY,   glusterd_handle_probe_query,           NULL, 0, DRC_NA},
+        [GLUSTERD_FRIEND_ADD]     = { "FRIEND_ADD",    GLUSTERD_FRIEND_ADD,    glusterd_handle_incoming_friend_req,   NULL, 0, DRC_NA},
+        [GLUSTERD_FRIEND_REMOVE]  = { "FRIEND_REMOVE", GLUSTERD_FRIEND_REMOVE, glusterd_handle_incoming_unfriend_req, NULL, 0, DRC_NA},
+        [GLUSTERD_FRIEND_UPDATE]  = { "FRIEND_UPDATE", GLUSTERD_FRIEND_UPDATE, glusterd_handle_friend_update,         NULL, 0, DRC_NA},
+};
+
+struct rpcsvc_program gd_svc_peer_prog = {
+        .progname  = "GlusterD svc peer",
+        .prognum   = GD_FRIEND_PROGRAM,
+        .progver   = GD_FRIEND_VERSION,
+        .numactors = GLUSTERD_FRIEND_MAXVALUE,
+        .actors    = gd_svc_peer_actors,
+	.synctask  = _gf_false,
+};
+
+
+
+rpcsvc_actor_t gd_svc_cli_actors[] = {
+        [GLUSTER_CLI_PROBE]              = { "CLI_PROBE",         GLUSTER_CLI_PROBE,            glusterd_handle_cli_probe,             NULL, 0, DRC_NA},
+        [GLUSTER_CLI_CREATE_VOLUME]      = { "CLI_CREATE_VOLUME", GLUSTER_CLI_CREATE_VOLUME,    glusterd_handle_create_volume,         NULL, 0, DRC_NA},
+        [GLUSTER_CLI_DEFRAG_VOLUME]      = { "CLI_DEFRAG_VOLUME", GLUSTER_CLI_DEFRAG_VOLUME,    glusterd_handle_defrag_volume,         NULL, 0, DRC_NA},
+        [GLUSTER_CLI_DEPROBE]            = { "FRIEND_REMOVE",     GLUSTER_CLI_DEPROBE,          glusterd_handle_cli_deprobe,           NULL, 0, DRC_NA},
+        [GLUSTER_CLI_LIST_FRIENDS]       = { "LIST_FRIENDS",      GLUSTER_CLI_LIST_FRIENDS,     glusterd_handle_cli_list_friends,      NULL, 0, DRC_NA},
+        [GLUSTER_CLI_UUID_RESET]         = { "UUID_RESET",        GLUSTER_CLI_UUID_RESET,       glusterd_handle_cli_uuid_reset,        NULL, 0, DRC_NA},
+        [GLUSTER_CLI_UUID_GET]           = { "UUID_GET",          GLUSTER_CLI_UUID_GET,         glusterd_handle_cli_uuid_get,          NULL, 0, DRC_NA},
+        [GLUSTER_CLI_START_VOLUME]       = { "START_VOLUME",      GLUSTER_CLI_START_VOLUME,     glusterd_handle_cli_start_volume,      NULL, 0, DRC_NA},
+        [GLUSTER_CLI_STOP_VOLUME]        = { "STOP_VOLUME",       GLUSTER_CLI_STOP_VOLUME,      glusterd_handle_cli_stop_volume,       NULL, 0, DRC_NA},
+        [GLUSTER_CLI_DELETE_VOLUME]      = { "DELETE_VOLUME",     GLUSTER_CLI_DELETE_VOLUME,    glusterd_handle_cli_delete_volume,     NULL, 0, DRC_NA},
+        [GLUSTER_CLI_GET_VOLUME]         = { "GET_VOLUME",        GLUSTER_CLI_GET_VOLUME,       glusterd_handle_cli_get_volume,        NULL, 0, DRC_NA},
+        [GLUSTER_CLI_ADD_BRICK]          = { "ADD_BRICK",         GLUSTER_CLI_ADD_BRICK,        glusterd_handle_add_brick,             NULL, 0, DRC_NA},
+        [GLUSTER_CLI_REPLACE_BRICK]      = { "REPLACE_BRICK",     GLUSTER_CLI_REPLACE_BRICK,    glusterd_handle_replace_brick,         NULL, 0, DRC_NA},
+        [GLUSTER_CLI_REMOVE_BRICK]       = { "REMOVE_BRICK",      GLUSTER_CLI_REMOVE_BRICK,     glusterd_handle_remove_brick,          NULL, 0, DRC_NA},
+        [GLUSTER_CLI_LOG_ROTATE]         = { "LOG FILENAME",      GLUSTER_CLI_LOG_ROTATE,       glusterd_handle_log_rotate,            NULL, 0, DRC_NA},
+        [GLUSTER_CLI_SET_VOLUME]         = { "SET_VOLUME",        GLUSTER_CLI_SET_VOLUME,       glusterd_handle_set_volume,            NULL, 0, DRC_NA},
+        [GLUSTER_CLI_SYNC_VOLUME]        = { "SYNC_VOLUME",       GLUSTER_CLI_SYNC_VOLUME,      glusterd_handle_sync_volume,           NULL, 0, DRC_NA},
+        [GLUSTER_CLI_RESET_VOLUME]       = { "RESET_VOLUME",      GLUSTER_CLI_RESET_VOLUME,     glusterd_handle_reset_volume,          NULL, 0, DRC_NA},
+        [GLUSTER_CLI_FSM_LOG]            = { "FSM_LOG",           GLUSTER_CLI_FSM_LOG,          glusterd_handle_fsm_log,               NULL, 0, DRC_NA},
+        [GLUSTER_CLI_GSYNC_SET]          = { "GSYNC_SET",         GLUSTER_CLI_GSYNC_SET,        glusterd_handle_gsync_set,             NULL, 0, DRC_NA},
+        [GLUSTER_CLI_PROFILE_VOLUME]     = { "STATS_VOLUME",      GLUSTER_CLI_PROFILE_VOLUME,   glusterd_handle_cli_profile_volume,    NULL, 0, DRC_NA},
+        [GLUSTER_CLI_QUOTA]              = { "QUOTA",             GLUSTER_CLI_QUOTA,            glusterd_handle_quota,                 NULL, 0, DRC_NA},
+        [GLUSTER_CLI_GETWD]              = { "GETWD",             GLUSTER_CLI_GETWD,            glusterd_handle_getwd,                 NULL, 1, DRC_NA},
+        [GLUSTER_CLI_STATUS_VOLUME]      = {"STATUS_VOLUME",      GLUSTER_CLI_STATUS_VOLUME,    glusterd_handle_status_volume,         NULL, 0, DRC_NA},
+        [GLUSTER_CLI_MOUNT]              = { "MOUNT",             GLUSTER_CLI_MOUNT,            glusterd_handle_mount,                 NULL, 1, DRC_NA},
+        [GLUSTER_CLI_UMOUNT]             = { "UMOUNT",            GLUSTER_CLI_UMOUNT,           glusterd_handle_umount,                NULL, 1, DRC_NA},
+        [GLUSTER_CLI_HEAL_VOLUME]        = { "HEAL_VOLUME",       GLUSTER_CLI_HEAL_VOLUME,      glusterd_handle_cli_heal_volume,       NULL, 0, DRC_NA},
+        [GLUSTER_CLI_STATEDUMP_VOLUME]   = {"STATEDUMP_VOLUME",   GLUSTER_CLI_STATEDUMP_VOLUME, glusterd_handle_cli_statedump_volume,  NULL, 0, DRC_NA},
+        [GLUSTER_CLI_LIST_VOLUME]        = {"LIST_VOLUME",        GLUSTER_CLI_LIST_VOLUME,      glusterd_handle_cli_list_volume,       NULL, 0, DRC_NA},
+        [GLUSTER_CLI_CLRLOCKS_VOLUME]    = {"CLEARLOCKS_VOLUME",  GLUSTER_CLI_CLRLOCKS_VOLUME,  glusterd_handle_cli_clearlocks_volume, NULL, 0, DRC_NA},
+        [GLUSTER_CLI_COPY_FILE]          = {"COPY_FILE",          GLUSTER_CLI_COPY_FILE,        glusterd_handle_copy_file,             NULL, 0, DRC_NA},
+        [GLUSTER_CLI_SYS_EXEC]           = {"SYS_EXEC",           GLUSTER_CLI_SYS_EXEC,         glusterd_handle_sys_exec,              NULL, 0, DRC_NA},
+        [GLUSTER_CLI_SNAP]               = {"SNAP",               GLUSTER_CLI_SNAP,             glusterd_handle_snapshot,              NULL, 0, DRC_NA},
+};
+
+struct rpcsvc_program gd_svc_cli_prog = {
+        .progname  = "GlusterD svc cli",
+        .prognum   = GLUSTER_CLI_PROGRAM,
+        .progver   = GLUSTER_CLI_VERSION,
+        .numactors = GLUSTER_CLI_MAXVALUE,
+        .actors    = gd_svc_cli_actors,
+	.synctask  = _gf_true,
+};
+
+/* This is a minimal RPC prog, which contains only the readonly RPC procs from
+ * the cli rpcsvc
+ */
+rpcsvc_actor_t gd_svc_cli_actors_ro[] = {
+        [GLUSTER_CLI_LIST_FRIENDS]       = { "LIST_FRIENDS",      GLUSTER_CLI_LIST_FRIENDS,     glusterd_handle_cli_list_friends,      NULL, 0, DRC_NA},
+        [GLUSTER_CLI_UUID_GET]           = { "UUID_GET",          GLUSTER_CLI_UUID_GET,         glusterd_handle_cli_uuid_get,          NULL, 0, DRC_NA},
+        [GLUSTER_CLI_GET_VOLUME]         = { "GET_VOLUME",        GLUSTER_CLI_GET_VOLUME,       glusterd_handle_cli_get_volume,        NULL, 0, DRC_NA},
+        [GLUSTER_CLI_GETWD]              = { "GETWD",             GLUSTER_CLI_GETWD,            glusterd_handle_getwd,                 NULL, 1, DRC_NA},
+        [GLUSTER_CLI_STATUS_VOLUME]      = {"STATUS_VOLUME",      GLUSTER_CLI_STATUS_VOLUME,    glusterd_handle_status_volume,         NULL, 0, DRC_NA},
+        [GLUSTER_CLI_LIST_VOLUME]        = {"LIST_VOLUME",        GLUSTER_CLI_LIST_VOLUME,      glusterd_handle_cli_list_volume,       NULL, 0, DRC_NA},
+};
+
+struct rpcsvc_program gd_svc_cli_prog_ro = {
+        .progname  = "GlusterD svc cli read-only",
+        .prognum   = GLUSTER_CLI_PROGRAM,
+        .progver   = GLUSTER_CLI_VERSION,
+        .numactors = GLUSTER_CLI_MAXVALUE,
+        .actors    = gd_svc_cli_actors_ro,
+	.synctask  = _gf_true,
+};
diff --git a/xlators/mgmt/glusterd/src/glusterd-handshake.c b/xlators/mgmt/glusterd/src/glusterd-handshake.c
index 88549d727..0f0357c4c 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handshake.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handshake.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
@@ -24,112 +14,411 @@
 #endif
 
 #include "xlator.h"
+#include "defaults.h"
 #include "glusterfs.h"
 #include "compat-errno.h"
 
 #include "glusterd.h"
 #include "glusterd-utils.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-store.h"
 
 #include "glusterfs3.h"
 #include "protocol-common.h"
 #include "rpcsvc.h"
+#include "rpc-common-xdr.h"
+
+extern struct rpc_clnt_program gd_peer_prog;
+extern struct rpc_clnt_program gd_mgmt_prog;
+extern struct rpc_clnt_program gd_mgmt_v3_prog;
+extern struct rpc_clnt_program gd_mgmt_v3_prog;
+
 
+#define TRUSTED_PREFIX         "trusted-"
 
 typedef ssize_t (*gfs_serialize_t) (struct iovec outmsg, void *data);
 
+static int
+get_snap_volname_and_volinfo (const char *volpath, char **volname,
+                              glusterd_volinfo_t **volinfo)
+{
+        int              ret            = -1;
+        char            *save_ptr       = NULL;
+        char            *str_token      = NULL;
+        char            *snapname       = NULL;
+        char            *volname_token  = NULL;
+        char            *vol            = NULL;
+        glusterd_snap_t *snap           = NULL;
+        xlator_t        *this           = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (volpath);
+        GF_ASSERT (volinfo);
+
+        str_token = gf_strdup (volpath);
+        if (NULL == str_token) {
+                goto out;
+        }
+
+        /* Input volname will have below formats:
+         * /snaps/<snapname>/<volname>.<hostname>
+         * or
+         * /snaps/<snapname>/<parent-volname>
+         * We need to extract snapname and parent_volname */
+
+        /*split string by "/" */
+        strtok_r (str_token, "/",  &save_ptr);
+        snapname  = strtok_r(NULL, "/", &save_ptr);
+        if (!snapname) {
+                gf_log(this->name, GF_LOG_ERROR, "Invalid path: %s", volpath);
+                goto out;
+        }
+
+        volname_token = strtok_r(NULL, "/", &save_ptr);
+        if (!volname_token) {
+                gf_log(this->name, GF_LOG_ERROR, "Invalid path: %s", volpath);
+                goto out;
+        }
+
+        snap = glusterd_find_snap_by_name (snapname);
+        if (!snap) {
+                gf_log(this->name, GF_LOG_ERROR, "Failed to "
+                                "fetch snap %s", snapname);
+                goto out;
+        }
+
+        /* Find if its a parent volume name or snap volume
+         * name. This function will succeed if volname_token
+         * is a parent volname
+         */
+        ret = glusterd_volinfo_find (volname_token, volinfo);
+        if (ret) {
+                *volname = gf_strdup (volname_token);
+                if (NULL == *volname) {
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = glusterd_snap_volinfo_find (volname_token, snap,
+                                                  volinfo);
+                if (ret) {
+                        /* Split the volume name */
+                        vol = strtok_r (volname_token, ".", &save_ptr);
+                        if (!vol) {
+                                gf_log(this->name, GF_LOG_ERROR, "Invalid "
+                                                "volname (%s)", volname_token);
+                                goto out;
+                        }
+
+                        ret = glusterd_snap_volinfo_find (vol, snap, volinfo);
+                        if (ret) {
+                                gf_log(this->name, GF_LOG_ERROR, "Failed to "
+                                       "fetch snap volume from volname (%s)",
+                                       vol);
+                                goto out;
+                        }
+                }
+        } else {
+                /*volname_token is parent volname*/
+                ret = glusterd_snap_volinfo_find_from_parent_volname (
+                                volname_token, snap, volinfo);
+                if (ret) {
+                        gf_log(this->name, GF_LOG_ERROR, "Failed to "
+                                        "fetch snap volume from parent "
+                                        "volname (%s)", volname_token);
+                        goto out;
+                }
+
+                /* Since volname_token is a parent volname we should
+                 * get the snap volname here*/
+                *volname = gf_strdup ((*volinfo)->volname);
+                if (NULL == *volname) {
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+out:
+        if (ret && NULL != *volname) {
+                GF_FREE (*volname);
+                *volname = NULL;
+        }
+        return ret;
+}
 
 static size_t
 build_volfile_path (const char *volname, char *path,
-                    size_t path_len)
-{
-        struct stat         stbuf       = {0,};
-        int32_t             ret         = -1;
-        glusterd_conf_t    *priv        = NULL;
-        char               *vol         = NULL;
-        char               *dup_volname = NULL;
-        char               *free_ptr    = NULL;
-        char               *tmp         = NULL;
-        glusterd_volinfo_t *volinfo     = NULL;
-
-        priv    = THIS->private;
-        dup_volname = gf_strdup (volname);
+                    size_t path_len, char *trusted_str)
+{
+        struct stat              stbuf                  = {0,};
+        int32_t                  ret                    = -1;
+        glusterd_conf_t         *priv                   = NULL;
+        char                    *vol                    = NULL;
+        char                    *dup_volname            = NULL;
+        char                    *free_ptr               = NULL;
+        char                    *save_ptr               = NULL;
+        char                    *str_token              = NULL;
+        glusterd_volinfo_t      *volinfo                = NULL;
+        char                    *server                 = NULL;
+        const char              *volname_ptr            = NULL;
+        char                     path_prefix [PATH_MAX] = {0,};
+        xlator_t                *this                   = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (volname);
+        GF_ASSERT (path);
+
+        if (strstr (volname, "gluster/")) {
+                server = strchr (volname, '/') + 1;
+                glusterd_get_nodesvc_volfile (server, priv->workdir,
+                                              path, path_len);
+                ret = 1;
+                goto out;
+        } else if ((str_token = strstr (volname, "/snaps/"))) {
+                ret = get_snap_volname_and_volinfo (str_token, &dup_volname,
+                                                    &volinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get snap"
+                                " volinfo from path (%s)", volname);
+                        ret = -1;
+                        goto out;
+                }
+
+                snprintf (path_prefix, sizeof (path_prefix), "%s/snaps/%s",
+                          priv->workdir, volinfo->snapshot->snapname);
+
+                free_ptr = dup_volname;
+                volname_ptr = dup_volname;
+                goto gotvolinfo;
+        } else if (volname[0] != '/') {
+                /* Normal behavior */
+                dup_volname = gf_strdup (volname);
+        } else {
+                /* Bringing in NFS like behavior for mount command,    */
+                /* With this, one can mount a volume with below cmd    */
+                /* bash# mount -t glusterfs server:/volume /mnt/pnt    */
+                dup_volname = gf_strdup (&volname[1]);
+        }
+
+        if (!dup_volname) {
+                gf_log(THIS->name, GF_LOG_ERROR, "strdup failed");
+                ret = -1;
+                goto out;
+        }
         free_ptr = dup_volname;
+        volname_ptr = volname;
+
+        snprintf (path_prefix, sizeof (path_prefix), "%s/vols",
+                  priv->workdir);
 
         ret = glusterd_volinfo_find (dup_volname, &volinfo);
+
         if (ret) {
                 /* Split the volume name */
-                vol = strtok_r (dup_volname, ".", &tmp);
+                vol = strtok_r (dup_volname, ".", &save_ptr);
                 if (!vol)
                         goto out;
+
                 ret = glusterd_volinfo_find (vol, &volinfo);
                 if (ret)
                         goto out;
         }
-        ret = snprintf (path, path_len, "%s/vols/%s/%s.vol",
-                        priv->workdir, volinfo->volname, volname);
+
+gotvolinfo:
+        if (!glusterd_auth_get_username (volinfo))
+                trusted_str = NULL;
+
+        ret = snprintf (path, path_len, "%s/%s/%s.vol", path_prefix,
+                        volinfo->volname, volname_ptr);
         if (ret == -1)
                 goto out;
 
         ret = stat (path, &stbuf);
+
         if ((ret == -1) && (errno == ENOENT)) {
-                ret = snprintf (path, path_len, "%s/vols/%s/%s-fuse.vol",
-                                priv->workdir, volinfo->volname, volname);
+                snprintf (path, path_len, "%s/%s/%s%s-fuse.vol",
+                          path_prefix, volinfo->volname,
+                          (trusted_str ? trusted_str : ""),
+                          dup_volname);
+
                 ret = stat (path, &stbuf);
         }
+
         if ((ret == -1) && (errno == ENOENT)) {
-                ret = snprintf (path, path_len, "%s/vols/%s/%s-tcp.vol",
-                                priv->workdir, volinfo->volname, volname);
+                snprintf (path, path_len, "%s/%s/%s-tcp.vol",
+                          path_prefix, volinfo->volname, volname_ptr);
         }
 
         ret = 1;
 out:
-        if (free_ptr)
-                GF_FREE (free_ptr);
+        GF_FREE (free_ptr);
         return ret;
 }
 
+/* Get and store op-versions of the clients sending the getspec request
+ * Clients of versions <= 3.3, don't send op-versions, their op-versions are
+ * defaulted to 1
+ */
 static int
-xdr_to_glusterfs_req (rpcsvc_request_t *req, void *arg, gfs_serialize_t sfunc)
+_get_client_op_versions (gf_getspec_req *args, peer_info_t *peerinfo)
 {
-        int                     ret = -1;
+        int    ret                   = 0;
+        int    client_max_op_version = 1;
+        int    client_min_op_version = 1;
+        dict_t *dict                 = NULL;
 
-        if (!req)
-                return -1;
+        GF_ASSERT (args);
+        GF_ASSERT (peerinfo);
 
-        ret = sfunc (req->msg[0], arg);
+        if (args->xdata.xdata_len) {
+                dict = dict_new ();
+                if (!dict) {
+                        ret = -1;
+                        goto out;
+                }
 
-        if (ret > 0)
-                ret = 0;
+                ret = dict_unserialize (args->xdata.xdata_val,
+                                        args->xdata.xdata_len, &dict);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR,
+                                "Failed to unserialize request dictionary");
+                        goto out;
+                }
 
+                ret = dict_get_int32 (dict, "min-op-version",
+                                      &client_min_op_version);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR,
+                                "Failed to get client-min-op-version");
+                        goto out;
+                }
+
+                ret = dict_get_int32 (dict, "max-op-version",
+                                      &client_max_op_version);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR,
+                                "Failed to get client-max-op-version");
+                        goto out;
+                }
+        }
+
+        peerinfo->max_op_version = client_max_op_version;
+        peerinfo->min_op_version = client_min_op_version;
+
+out:
         return ret;
 }
 
-
-int
-server_getspec (rpcsvc_request_t *req)
+/* Checks if the client supports the volume, ie. client can understand all the
+ * options in the volfile
+ */
+static gf_boolean_t
+_client_supports_volume (peer_info_t *peerinfo, int32_t *op_errno)
 {
-        int32_t               ret = -1;
-        int32_t               op_errno = 0;
-        int32_t               spec_fd = -1;
-        size_t                file_len = 0;
-        char                  filename[ZR_PATH_MAX] = {0,};
-        struct stat           stbuf = {0,};
-        char                 *volume = NULL;
-        int                   cookie = 0;
+        gf_boolean_t       ret       = _gf_true;
+        glusterd_volinfo_t *volinfo  = NULL;
 
-        gf_getspec_req    args = {0,};
-        gf_getspec_rsp    rsp  = {0,};
+        GF_ASSERT (peerinfo);
+        GF_ASSERT (op_errno);
 
 
-        if (xdr_to_glusterfs_req (req, &args, xdr_to_getspec_req)) {
+        /* Only check when the volfile being requested is a volume. Not finding
+         * a volinfo implies that the volfile requested for is not of a gluster
+         * volume. A non volume volfile is requested by the local gluster
+         * services like shd and nfs-server. These need not be checked as they
+         * will be running at the same op-version as glusterd and will be able
+         * to support all the features
+         */
+        if ((glusterd_volinfo_find (peerinfo->volname, &volinfo) == 0) &&
+            ((peerinfo->min_op_version > volinfo->client_op_version) ||
+             (peerinfo->max_op_version < volinfo->client_op_version))) {
+                ret = _gf_false;
+                *op_errno = ENOTSUP;
+                gf_log ("glusterd", GF_LOG_INFO,
+                        "Client %s (%d -> %d) doesn't support required "
+                        "op-version (%d). Rejecting volfile request.",
+                        peerinfo->identifier, peerinfo->min_op_version,
+                        peerinfo->max_op_version, volinfo->client_op_version);
+        }
+
+        return ret;
+}
+
+int
+__server_getspec (rpcsvc_request_t *req)
+{
+        int32_t               ret                    = -1;
+        int32_t               op_errno               = 0;
+        int32_t               spec_fd                = -1;
+        size_t                file_len               = 0;
+        char                  filename[PATH_MAX]  = {0,};
+        struct stat           stbuf                  = {0,};
+        char                 *volume                 = NULL;
+        char                 *tmp                    = NULL;
+        int                   cookie                 = 0;
+        rpc_transport_t      *trans                  = NULL;
+        gf_getspec_req        args                   = {0,};
+        gf_getspec_rsp        rsp                    = {0,};
+        char                  addrstr[RPCSVC_PEER_STRLEN] = {0};
+        peer_info_t          *peerinfo               = NULL;
+
+        ret = xdr_to_generic (req->msg[0], &args,
+                              (xdrproc_t)xdr_gf_getspec_req);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
                 goto fail;
         }
 
+        peerinfo = &req->trans->peerinfo;
+
         volume = args.key;
+        /* Need to strip leading '/' from volnames. This was introduced to
+         * support nfs style mount parameters for native gluster mount
+         */
+        if (volume[0] == '/')
+                strncpy (peerinfo->volname, &volume[1], strlen(&volume[1]));
+        else
+                strncpy (peerinfo->volname, volume, strlen(volume));
+
+        ret = _get_client_op_versions (&args, peerinfo);
+        if (ret)
+                goto fail;
+
+        if (!_client_supports_volume (peerinfo, &op_errno)) {
+                ret = -1;
+                goto fail;
+        }
+
+        trans = req->trans;
+        /* addrstr will be empty for cli socket connections */
+        ret = rpcsvc_transport_peername (trans, (char *)&addrstr,
+                                         sizeof (addrstr));
+        if (ret)
+                goto fail;
+
+        tmp  = strrchr (addrstr, ':');
+        if (tmp)
+                *tmp = '\0';
 
-        ret = build_volfile_path (volume, filename, sizeof (filename));
+        /* The trusted volfiles are given to the glusterd owned process like NFS
+         * server, self-heal daemon etc., so that they are not inadvertently
+         * blocked by a auth.{allow,reject} setting. The trusted volfile is not
+         * meant for external users.
+         */
+        if (strlen (addrstr) && gf_is_local_addr (addrstr)) {
+
+                ret = build_volfile_path (volume, filename,
+                                          sizeof (filename),
+                                          TRUSTED_PREFIX);
+        } else {
+                ret = build_volfile_path (volume, filename,
+                                          sizeof (filename), NULL);
+        }
 
         if (ret > 0) {
                 /* to allocate the proper buffer to hold the file data */
@@ -154,7 +443,7 @@ server_getspec (rpcsvc_request_t *req)
         }
 
         if (file_len) {
-                rsp.spec = CALLOC (file_len, sizeof (char));
+                rsp.spec = CALLOC (file_len+1, sizeof (char));
                 if (!rsp.spec) {
                         ret = -1;
                         op_errno = ENOMEM;
@@ -175,29 +464,903 @@ fail:
                 rsp.op_errno = cookie;
 
         if (!rsp.spec)
-                rsp.spec = "";
+                rsp.spec = strdup ("");
 
         glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                               (gd_serialize_t)xdr_serialize_getspec_rsp);
-        if (args.key)
-                free (args.key);//malloced by xdr
-        if (rsp.spec && (strcmp (rsp.spec, "")))
-                free (rsp.spec);
+                               (xdrproc_t)xdr_gf_getspec_rsp);
+        free (args.key);//malloced by xdr
+        free (rsp.spec);
+
+        return 0;
+}
+
+int
+server_getspec (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __server_getspec);
+}
+
+int32_t
+__server_event_notify (rpcsvc_request_t *req)
+{
+        int32_t                 ret             = -1;
+        int32_t                 op_errno        =  0;
+        gf_event_notify_req     args            = {0,};
+        gf_event_notify_rsp     rsp             = {0,};
+        dict_t                 *dict            = NULL;
+        gf_boolean_t            need_rsp        = _gf_true;
+
+        ret = xdr_to_generic (req->msg[0], &args,
+                              (xdrproc_t)xdr_gf_event_notify_req);
+        if (ret < 0) {
+                req->rpc_err = GARBAGE_ARGS;
+                goto fail;
+        }
+
+        if (args.dict.dict_len) {
+                dict = dict_new ();
+                if (!dict)
+                        return ret;
+                ret = dict_unserialize (args.dict.dict_val,
+                                        args.dict.dict_len, &dict);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Failed to unserialize req");
+                        goto fail;
+                }
+        }
+
+        switch (args.op) {
+        case GF_EN_DEFRAG_STATUS:
+                gf_log ("", GF_LOG_INFO,
+                        "received defrag status updated");
+                if (dict) {
+                        glusterd_defrag_event_notify_handle (dict);
+                        need_rsp = _gf_false;
+                }
+                break;
+        default:
+                gf_log ("", GF_LOG_ERROR, "Unknown op received in event "
+                        "notify");
+                ret = -1;
+                break;
+        }
+
+fail:
+        rsp.op_ret   = ret;
+
+        if (op_errno)
+                rsp.op_errno = gf_errno_to_error (op_errno);
+
+        if (need_rsp)
+                glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                       (xdrproc_t)xdr_gf_event_notify_rsp);
+        if (dict)
+                dict_unref (dict);
+        free (args.dict.dict_val);//malloced by xdr
 
         return 0;
 }
 
+int32_t
+server_event_notify (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __server_event_notify);
+}
+
+int
+gd_validate_cluster_op_version (xlator_t *this, int cluster_op_version,
+                                char *peerid)
+{
+        int              ret  = -1;
+        glusterd_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        if (cluster_op_version > GD_OP_VERSION_MAX) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "operating version %d is more than the maximum "
+                        "supported (%d) on the machine (as per peer request "
+                        "from %s)", cluster_op_version, GD_OP_VERSION_MAX,
+                        peerid);
+                goto out;
+        }
+
+        /* The peer can only reduce its op-version when it doesn't have any
+         * volumes. Reducing op-version when it already contains volumes can
+         * lead to inconsistencies in the cluster
+         */
+        if ((cluster_op_version < conf->op_version) &&
+            !list_empty (&conf->volumes)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "cannot reduce operating version to %d from current "
+                        "version %d as volumes exist (as per peer request from "
+                        "%s)", cluster_op_version, conf->op_version, peerid);
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+__glusterd_mgmt_hndsk_versions (rpcsvc_request_t *req)
+{
+        dict_t            *dict            = NULL;
+        xlator_t          *this            = NULL;
+        glusterd_conf_t   *conf            = NULL;
+        int                ret             = -1;
+        int                op_errno        = EINVAL;
+        gf_mgmt_hndsk_req  args            = {{0,},};
+        gf_mgmt_hndsk_rsp  rsp             = {0,};
+
+        this = THIS;
+        conf = this->private;
+
+        ret = xdr_to_generic (req->msg[0], &args,
+                              (xdrproc_t)xdr_gf_mgmt_hndsk_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        ret = dict_set_int32 (dict, GD_OP_VERSION_KEY, conf->op_version);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set operating version");
+                rsp.op_ret = ret;
+                goto out;
+        }
+
+        ret = dict_set_int32 (dict, GD_MIN_OP_VERSION_KEY, GD_OP_VERSION_MIN);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set %s", GD_MIN_OP_VERSION_KEY);
+                rsp.op_ret = ret;
+                goto out;
+        }
+
+        ret = dict_set_int32 (dict, GD_MAX_OP_VERSION_KEY, GD_OP_VERSION_MAX);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set %s", GD_MAX_OP_VERSION_KEY);
+                rsp.op_ret = ret;
+                goto out;
+        }
+
+        ret = 0;
+
+        GF_PROTOCOL_DICT_SERIALIZE (this, dict, (&rsp.hndsk.hndsk_val),
+                                    rsp.hndsk.hndsk_len, op_errno, out);
+out:
+
+        rsp.op_ret = ret;
+        rsp.op_errno = op_errno;
+
+        glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                               (xdrproc_t)xdr_gf_mgmt_hndsk_rsp);
+
+        ret = 0;
+
+        if (dict)
+                dict_unref (dict);
+
+        if (args.hndsk.hndsk_val)
+                free (args.hndsk.hndsk_val);
+
+        if (rsp.hndsk.hndsk_val)
+                GF_FREE (rsp.hndsk.hndsk_val);
+
+        return ret;
+}
+
+int
+glusterd_mgmt_hndsk_versions (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_mgmt_hndsk_versions);
+}
+
+int
+__glusterd_mgmt_hndsk_versions_ack (rpcsvc_request_t *req)
+{
+        dict_t            *clnt_dict       = NULL;
+        xlator_t          *this            = NULL;
+        glusterd_conf_t   *conf            = NULL;
+        int                ret             = -1;
+        int                op_errno        = EINVAL;
+        int                peer_op_version = 0;
+        gf_mgmt_hndsk_req  args            = {{0,},};
+        gf_mgmt_hndsk_rsp  rsp             = {0,};
+
+        this = THIS;
+        conf = this->private;
+
+        ret = xdr_to_generic (req->msg[0], &args,
+                              (xdrproc_t)xdr_gf_mgmt_hndsk_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        GF_PROTOCOL_DICT_UNSERIALIZE (this, clnt_dict, args.hndsk.hndsk_val,
+                                      (args.hndsk.hndsk_len), ret, op_errno,
+                                      out);
+
+        ret = dict_get_int32 (clnt_dict, GD_OP_VERSION_KEY, &peer_op_version);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to get the op-version key peer=%s",
+                        req->trans->peerinfo.identifier);
+                goto out;
+        }
+
+        ret = gd_validate_cluster_op_version (this, peer_op_version,
+                                              req->trans->peerinfo.identifier);
+        if (ret)
+                goto out;
+
+
+        /* As this is ACK from the Cluster for the versions supported,
+           can set the op-version of 'this' glusterd to the one
+           received. */
+        gf_log (this->name, GF_LOG_INFO, "using the op-version %d",
+                peer_op_version);
+        conf->op_version = peer_op_version;
+        ret = glusterd_store_global_info (this);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "Failed to store op-version");
+
+out:
+        rsp.op_ret = ret;
+        rsp.op_errno = op_errno;
+
+        glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                               (xdrproc_t)xdr_gf_mgmt_hndsk_rsp);
+
+        ret = 0;
+
+        if (clnt_dict)
+                dict_unref (clnt_dict);
+
+        if (args.hndsk.hndsk_val)
+                free (args.hndsk.hndsk_val);
+
+        return ret;
+}
+
+int
+glusterd_mgmt_hndsk_versions_ack (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_mgmt_hndsk_versions_ack);
+}
 
 rpcsvc_actor_t gluster_handshake_actors[] = {
-        [GF_HNDSK_NULL]      = {"NULL",      GF_HNDSK_NULL,      NULL, NULL, NULL },
-        [GF_HNDSK_GETSPEC]   = {"GETSPEC",   GF_HNDSK_GETSPEC,   server_getspec, NULL, NULL },
+        [GF_HNDSK_NULL]         = {"NULL",        GF_HNDSK_NULL,         NULL,                NULL, 0, DRC_NA},
+        [GF_HNDSK_GETSPEC]      = {"GETSPEC",     GF_HNDSK_GETSPEC,      server_getspec,      NULL, 0, DRC_NA},
+        [GF_HNDSK_EVENT_NOTIFY] = {"EVENTNOTIFY", GF_HNDSK_EVENT_NOTIFY, server_event_notify, NULL, 0, DRC_NA},
 };
 
 
 struct rpcsvc_program gluster_handshake_prog = {
-        .progname  = "GlusterFS Handshake",
+        .progname  = "Gluster Handshake",
         .prognum   = GLUSTER_HNDSK_PROGRAM,
         .progver   = GLUSTER_HNDSK_VERSION,
         .actors    = gluster_handshake_actors,
         .numactors = GF_HNDSK_MAXVALUE,
 };
+
+/* A minimal RPC program just for the cli getspec command */
+rpcsvc_actor_t gluster_cli_getspec_actors[] = {
+        [GF_HNDSK_GETSPEC]      = {"GETSPEC",     GF_HNDSK_GETSPEC,      server_getspec,      NULL, 0, DRC_NA},
+};
+
+struct rpcsvc_program gluster_cli_getspec_prog = {
+        .progname  = "Gluster Handshake (CLI Getspec)",
+        .prognum   = GLUSTER_HNDSK_PROGRAM,
+        .progver   = GLUSTER_HNDSK_VERSION,
+        .actors    = gluster_cli_getspec_actors,
+        .numactors = GF_HNDSK_MAXVALUE,
+};
+
+
+char *glusterd_dump_proc[GF_DUMP_MAXVALUE] = {
+        [GF_DUMP_NULL] = "NULL",
+        [GF_DUMP_DUMP] = "DUMP",
+};
+
+rpc_clnt_prog_t glusterd_dump_prog = {
+        .progname       = "GLUSTERD-DUMP",
+        .prognum        = GLUSTER_DUMP_PROGRAM,
+        .progver        = GLUSTER_DUMP_VERSION,
+        .procnames      = glusterd_dump_proc,
+};
+
+
+rpcsvc_actor_t glusterd_mgmt_hndsk_actors[] = {
+        [GD_MGMT_HNDSK_NULL]            = {"NULL", GD_MGMT_HNDSK_NULL, NULL,
+                                           NULL, 0},
+        [GD_MGMT_HNDSK_VERSIONS]        = {"MGMT-VERS", GD_MGMT_HNDSK_VERSIONS,
+                                           glusterd_mgmt_hndsk_versions, NULL,
+                                           0},
+        [GD_MGMT_HNDSK_VERSIONS_ACK]    = {"MGMT-VERS-ACK",
+                                           GD_MGMT_HNDSK_VERSIONS_ACK,
+                                           glusterd_mgmt_hndsk_versions_ack,
+                                           NULL, 0},
+};
+
+struct rpcsvc_program glusterd_mgmt_hndsk_prog = {
+        .progname       = "Gluster MGMT Handshake",
+        .prognum        = GD_MGMT_HNDSK_PROGRAM,
+        .progver        = GD_MGMT_HNDSK_VERSION,
+        .actors         = glusterd_mgmt_hndsk_actors,
+        .numactors      = GD_MGMT_HNDSK_MAXVALUE,
+};
+
+char *glusterd_mgmt_hndsk_proc[GD_MGMT_HNDSK_MAXVALUE] = {
+        [GD_MGMT_HNDSK_NULL]          = "NULL",
+        [GD_MGMT_HNDSK_VERSIONS] = "MGMT-VERS",
+        [GD_MGMT_HNDSK_VERSIONS_ACK] = "MGMT-VERS-ACK",
+};
+
+rpc_clnt_prog_t gd_clnt_mgmt_hndsk_prog = {
+        .progname  = "Gluster MGMT Handshake",
+        .prognum   = GD_MGMT_HNDSK_PROGRAM,
+        .progver   = GD_MGMT_HNDSK_VERSION,
+        .procnames = glusterd_mgmt_hndsk_proc,
+};
+
+
+static int
+glusterd_event_connected_inject (glusterd_peerctx_t *peerctx)
+{
+        GF_ASSERT (peerctx);
+
+        glusterd_friend_sm_event_t      *event = NULL;
+        glusterd_probe_ctx_t            *ctx = NULL;
+        int                             ret = -1;
+        glusterd_peerinfo_t             *peerinfo = NULL;
+
+
+        ret = glusterd_friend_sm_new_event
+                        (GD_FRIEND_EVENT_CONNECTED, &event);
+
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get new event");
+                goto out;
+        }
+
+        ctx = GF_CALLOC (1, sizeof(*ctx), gf_gld_mt_probe_ctx_t);
+
+        if (!ctx) {
+                ret = -1;
+                gf_log ("", GF_LOG_ERROR, "Memory not available");
+                goto out;
+        }
+
+        peerinfo = peerctx->peerinfo;
+        ctx->hostname = gf_strdup (peerinfo->hostname);
+        ctx->port = peerinfo->port;
+        ctx->req = peerctx->args.req;
+        ctx->dict = peerctx->args.dict;
+
+        event->peerinfo = peerinfo;
+        event->ctx = ctx;
+
+        ret = glusterd_friend_sm_inject_event (event);
+
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Unable to inject "
+                        "EVENT_CONNECTED ret = %d", ret);
+                goto out;
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+
+int
+gd_validate_peer_op_version (xlator_t *this, glusterd_peerinfo_t *peerinfo,
+                             dict_t *dict, char **errstr)
+{
+        int              ret  = -1;
+        glusterd_conf_t *conf = NULL;
+        int32_t          peer_op_version = 0;
+        int32_t          peer_min_op_version = 0;
+        int32_t          peer_max_op_version = 0;
+
+        if (!dict && !this && !peerinfo)
+                goto out;
+
+        conf = this->private;
+
+        ret = dict_get_int32 (dict, GD_OP_VERSION_KEY, &peer_op_version);
+        if (ret)
+                goto out;
+
+        ret = dict_get_int32 (dict, GD_MAX_OP_VERSION_KEY,
+                              &peer_max_op_version);
+        if (ret)
+                goto out;
+
+        ret = dict_get_int32 (dict, GD_MIN_OP_VERSION_KEY,
+                              &peer_min_op_version);
+        if (ret)
+                goto out;
+
+        ret = -1;
+        /* Check if peer can support our op_version */
+        if ((peer_max_op_version < conf->op_version) ||
+            (peer_min_op_version > conf->op_version)) {
+                ret = gf_asprintf (errstr, "Peer %s does not support required "
+                                   "op-version", peerinfo->hostname);
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name , GF_LOG_DEBUG, "Peer %s %s", peerinfo->hostname,
+                ((ret < 0) ? "rejected" : "accepted"));
+        return ret;
+}
+
+int
+__glusterd_mgmt_hndsk_version_ack_cbk (struct rpc_req *req, struct iovec *iov,
+                                     int count, void *myframe)
+{
+        int                  ret      = -1;
+        int                  op_errno = EINVAL;
+        gf_mgmt_hndsk_rsp    rsp      = {0,};
+        xlator_t            *this     = NULL;
+        call_frame_t        *frame    = NULL;
+        glusterd_peerinfo_t *peerinfo = NULL;
+        glusterd_peerctx_t  *peerctx  = NULL;
+        char                msg[1024] = {0,};
+
+        this = THIS;
+        frame = myframe;
+        peerctx = frame->local;
+        peerinfo = peerctx->peerinfo;
+
+        if (-1 == req->rpc_status) {
+                snprintf (msg, sizeof (msg),
+                          "Error through RPC layer, retry again later");
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                peerctx->errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_mgmt_hndsk_rsp);
+        if (ret < 0) {
+                snprintf (msg, sizeof (msg), "Failed to decode XDR");
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                peerctx->errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        op_errno = rsp.op_errno;
+        if (-1 == rsp.op_ret) {
+                ret = -1;
+                snprintf (msg, sizeof (msg),
+                          "Failed to get handshake ack from remote server");
+                gf_log (frame->this->name, GF_LOG_ERROR, "%s", msg);
+                peerctx->errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        /* TODO: this is hardcoded as of now, but I don't forsee any problems
+         * with this as long as we are properly handshaking operating versions
+         */
+        peerinfo->mgmt = &gd_mgmt_prog;
+        peerinfo->peer = &gd_peer_prog;
+        peerinfo->mgmt_v3 = &gd_mgmt_v3_prog;
+
+        ret = default_notify (this, GF_EVENT_CHILD_UP, NULL);
+
+        if (GD_MODE_ON == peerctx->args.mode) {
+                ret = glusterd_event_connected_inject (peerctx);
+                peerctx->args.req = NULL;
+        } else if (GD_MODE_SWITCH_ON == peerctx->args.mode) {
+                peerctx->args.mode = GD_MODE_ON;
+        } else {
+                gf_log (this->name, GF_LOG_WARNING, "unknown mode %d",
+                        peerctx->args.mode);
+        }
+
+        glusterd_friend_sm ();
+
+        ret = 0;
+out:
+
+        frame->local = NULL;
+        STACK_DESTROY (frame->root);
+
+        if (ret != 0)
+                rpc_transport_disconnect (peerinfo->rpc->conn.trans);
+
+        if (rsp.hndsk.hndsk_val)
+                free (rsp.hndsk.hndsk_val);
+
+        return 0;
+}
+
+int
+glusterd_mgmt_hndsk_version_ack_cbk (struct rpc_req *req, struct iovec *iov,
+                                     int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_mgmt_hndsk_version_ack_cbk);
+}
+
+int
+__glusterd_mgmt_hndsk_version_cbk (struct rpc_req *req, struct iovec *iov,
+                                 int count, void *myframe)
+{
+        int                  ret       = -1;
+        int                  op_errno  = EINVAL;
+        gf_mgmt_hndsk_rsp    rsp       = {0,};
+        gf_mgmt_hndsk_req    arg       = {{0,}};
+        xlator_t            *this      = NULL;
+        call_frame_t        *frame     = NULL;
+        glusterd_peerinfo_t *peerinfo  = NULL;
+        glusterd_peerctx_t  *peerctx   = NULL;
+        dict_t              *dict      = NULL;
+        dict_t              *rsp_dict  = NULL;
+        glusterd_conf_t     *conf      = NULL;
+        char                 msg[1024] = {0,};
+
+        this = THIS;
+        conf = this->private;
+        frame = myframe;
+        peerctx = frame->local;
+        peerinfo = peerctx->peerinfo;
+
+        if (-1 == req->rpc_status) {
+                ret = -1;
+                snprintf (msg, sizeof (msg),
+                          "Error through RPC layer, retry again later");
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                peerctx->errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_mgmt_hndsk_rsp);
+        if (ret < 0) {
+                snprintf (msg, sizeof (msg), "Failed to decode management "
+                          "handshake response");
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                peerctx->errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        GF_PROTOCOL_DICT_UNSERIALIZE (this, dict, rsp.hndsk.hndsk_val,
+                                      rsp.hndsk.hndsk_len, ret, op_errno,
+                                      out);
+
+        op_errno = rsp.op_errno;
+        if (-1 == rsp.op_ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get the 'versions' from peer (%s)",
+                        req->conn->trans->peerinfo.identifier);
+                goto out;
+        }
+
+        /* Check if peer can be part of cluster */
+        ret = gd_validate_peer_op_version (this, peerinfo, dict,
+                                           &peerctx->errstr);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to validate the operating version of peer (%s)",
+                        peerinfo->hostname);
+                goto out;
+        }
+
+        rsp_dict = dict_new ();
+        if (!rsp_dict)
+                goto out;
+
+        ret = dict_set_int32 (rsp_dict, GD_OP_VERSION_KEY, conf->op_version);
+        if (ret) {
+                gf_log(this->name, GF_LOG_ERROR,
+                       "failed to set operating version in dict");
+                goto out;
+        }
+
+        GF_PROTOCOL_DICT_SERIALIZE (this, rsp_dict, (&arg.hndsk.hndsk_val),
+                                    arg.hndsk.hndsk_len, op_errno, out);
+
+        ret = glusterd_submit_request (peerctx->peerinfo->rpc, &arg, frame,
+                                       &gd_clnt_mgmt_hndsk_prog,
+                                       GD_MGMT_HNDSK_VERSIONS_ACK, NULL, this,
+                                       glusterd_mgmt_hndsk_version_ack_cbk,
+                                       (xdrproc_t)xdr_gf_mgmt_hndsk_req);
+
+out:
+        if (ret) {
+                frame->local = NULL;
+                STACK_DESTROY (frame->root);
+                rpc_transport_disconnect (peerinfo->rpc->conn.trans);
+        }
+
+        if (rsp.hndsk.hndsk_val)
+                free (rsp.hndsk.hndsk_val);
+
+        if (arg.hndsk.hndsk_val)
+                GF_FREE (arg.hndsk.hndsk_val);
+
+        if (dict)
+                dict_unref (dict);
+
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        return 0;
+}
+
+int
+glusterd_mgmt_hndsk_version_cbk (struct rpc_req *req, struct iovec *iov,
+                                     int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_mgmt_hndsk_version_cbk);
+}
+
+int
+glusterd_mgmt_handshake (xlator_t *this, glusterd_peerctx_t *peerctx)
+{
+        call_frame_t        *frame    = NULL;
+        gf_mgmt_hndsk_req    req      = {{0,},};
+        int                  ret      = -1;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame)
+                goto out;
+
+        frame->local = peerctx;
+
+        ret = glusterd_submit_request (peerctx->peerinfo->rpc, &req, frame,
+                                       &gd_clnt_mgmt_hndsk_prog,
+                                       GD_MGMT_HNDSK_VERSIONS, NULL, this,
+                                       glusterd_mgmt_hndsk_version_cbk,
+                                       (xdrproc_t)xdr_gf_mgmt_hndsk_req);
+        ret = 0;
+out:
+        if (ret && frame)
+                STACK_DESTROY (frame->root);
+
+        return ret;
+}
+
+int
+glusterd_set_clnt_mgmt_program (glusterd_peerinfo_t *peerinfo,
+                                gf_prog_detail *prog)
+{
+        gf_prog_detail  *trav   = NULL;
+        int             ret     = -1;
+
+        if (!peerinfo || !prog)
+                goto out;
+
+        trav = prog;
+
+        while (trav) {
+                ret = -1;
+                if ((gd_mgmt_prog.prognum == trav->prognum) &&
+                    (gd_mgmt_prog.progver == trav->progver)) {
+                        peerinfo->mgmt = &gd_mgmt_prog;
+                        ret = 0;
+                }
+
+                if ((gd_peer_prog.prognum == trav->prognum) &&
+                    (gd_peer_prog.progver == trav->progver)) {
+                        peerinfo->peer = &gd_peer_prog;
+                        ret = 0;
+                }
+
+                if (ret) {
+                        gf_log ("", GF_LOG_DEBUG,
+                                "%s (%"PRId64":%"PRId64") not supported",
+                                trav->progname, trav->prognum,
+                                trav->progver);
+                }
+
+                trav = trav->next;
+        }
+
+        if (peerinfo->mgmt) {
+                 gf_log ("", GF_LOG_INFO,
+                         "Using Program %s, Num (%d), Version (%d)",
+                         peerinfo->mgmt->progname, peerinfo->mgmt->prognum,
+                         peerinfo->mgmt->progver);
+        }
+
+        if (peerinfo->peer) {
+                 gf_log ("", GF_LOG_INFO,
+                         "Using Program %s, Num (%d), Version (%d)",
+                         peerinfo->peer->progname, peerinfo->peer->prognum,
+                         peerinfo->peer->progver);
+        }
+
+        if (peerinfo->mgmt_v3) {
+                 gf_log ("", GF_LOG_INFO,
+                         "Using Program %s, Num (%d), Version (%d)",
+                         peerinfo->mgmt_v3->progname, peerinfo->mgmt_v3->prognum,
+                         peerinfo->mgmt_v3->progver);
+        }
+
+        ret = 0;
+out:
+        return ret;
+
+}
+
+static gf_boolean_t
+_mgmt_hndsk_prog_present (gf_prog_detail *prog) {
+        gf_boolean_t    ret = _gf_false;
+        gf_prog_detail  *trav = NULL;
+
+        GF_ASSERT (prog);
+
+        trav = prog;
+
+        while (trav) {
+                if ((trav->prognum == GD_MGMT_HNDSK_PROGRAM) &&
+                    (trav->progver == GD_MGMT_HNDSK_VERSION)) {
+                        ret = _gf_true;
+                        goto out;
+                }
+                trav = trav->next;
+        }
+out:
+        return ret;
+}
+
+int
+__glusterd_peer_dump_version_cbk (struct rpc_req *req, struct iovec *iov,
+                                int count, void *myframe)
+{
+        int                  ret      = -1;
+        gf_dump_rsp          rsp      = {0,};
+        xlator_t            *this     = NULL;
+        gf_prog_detail      *trav     = NULL;
+        gf_prog_detail      *next     = NULL;
+        call_frame_t        *frame    = NULL;
+        glusterd_peerinfo_t *peerinfo = NULL;
+        glusterd_peerctx_t  *peerctx  = NULL;
+        glusterd_conf_t     *conf     = NULL;
+        char                msg[1024] = {0,};
+
+        this = THIS;
+        conf = this->private;
+        frame = myframe;
+        peerctx = frame->local;
+        peerinfo = peerctx->peerinfo;
+
+        if (-1 == req->rpc_status) {
+                snprintf (msg, sizeof (msg),
+                          "Error through RPC layer, retry again later");
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                peerctx->errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gf_dump_rsp);
+        if (ret < 0) {
+                snprintf (msg, sizeof (msg), "Failed to decode XDR");
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                peerctx->errstr = gf_strdup (msg);
+                goto out;
+        }
+        if (-1 == rsp.op_ret) {
+                snprintf (msg, sizeof (msg),
+                          "Failed to get the 'versions' from remote server");
+                gf_log (frame->this->name, GF_LOG_ERROR, "%s", msg);
+                peerctx->errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        if (_mgmt_hndsk_prog_present (rsp.prog)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Proceeding to op-version handshake with peer %s",
+                        peerinfo->hostname);
+                ret = glusterd_mgmt_handshake (this, peerctx);
+                goto out;
+        } else if (conf->op_version > 1) {
+                ret = -1;
+                snprintf (msg, sizeof (msg),
+                          "Peer %s does not support required op-version",
+                          peerinfo->hostname);
+                peerctx->errstr = gf_strdup (msg);
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        /* Make sure we assign the proper program to peer */
+        ret = glusterd_set_clnt_mgmt_program (peerinfo, rsp.prog);
+        if (ret) {
+                gf_log ("", GF_LOG_WARNING, "failed to set the mgmt program");
+                goto out;
+        }
+
+        ret = default_notify (this, GF_EVENT_CHILD_UP, NULL);
+
+        if (GD_MODE_ON == peerctx->args.mode) {
+                ret = glusterd_event_connected_inject (peerctx);
+                peerctx->args.req = NULL;
+        } else if (GD_MODE_SWITCH_ON == peerctx->args.mode) {
+                peerctx->args.mode = GD_MODE_ON;
+        } else {
+                gf_log ("", GF_LOG_WARNING, "unknown mode %d",
+                        peerctx->args.mode);
+        }
+
+        glusterd_friend_sm();
+        glusterd_op_sm();
+
+        ret = 0;
+
+out:
+
+        /* don't use GF_FREE, buffer was allocated by libc */
+        if (rsp.prog) {
+                trav = rsp.prog;
+                while (trav) {
+                        next = trav->next;
+                        free (trav->progname);
+                        free (trav);
+                        trav = next;
+                }
+        }
+
+        frame->local = NULL;
+        STACK_DESTROY (frame->root);
+
+        if (ret != 0)
+                rpc_transport_disconnect (peerinfo->rpc->conn.trans);
+
+        return 0;
+}
+
+
+int
+glusterd_peer_dump_version_cbk (struct rpc_req *req, struct iovec *iov,
+                                int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_peer_dump_version_cbk);
+}
+
+int
+glusterd_peer_dump_version (xlator_t *this, struct rpc_clnt *rpc,
+                            glusterd_peerctx_t *peerctx)
+{
+        call_frame_t        *frame    = NULL;
+        gf_dump_req          req      = {0,};
+        int                  ret      = -1;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame)
+                goto out;
+
+        frame->local = peerctx;
+
+        req.gfs_id = 0xcafe;
+
+        ret = glusterd_submit_request (peerctx->peerinfo->rpc, &req, frame,
+                                       &glusterd_dump_prog, GF_DUMP_DUMP,
+                                       NULL, this,
+                                       glusterd_peer_dump_version_cbk,
+                                       (xdrproc_t)xdr_gf_dump_req);
+out:
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-hooks.c b/xlators/mgmt/glusterd/src/glusterd-hooks.c
new file mode 100644
index 000000000..2b43a452e
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-hooks.c
@@ -0,0 +1,531 @@
+/*
+   Copyright (c) 2007-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "globals.h"
+#include "glusterfs.h"
+#include "dict.h"
+#include "xlator.h"
+#include "logging.h"
+#include "run.h"
+#include "defaults.h"
+#include "compat.h"
+#include "compat-errno.h"
+#include "glusterd.h"
+#include "glusterd-sm.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-utils.h"
+#include "glusterd-store.h"
+#include "glusterd-hooks.h"
+
+#include <fnmatch.h>
+
+#define EMPTY ""
+char glusterd_hook_dirnames[GD_OP_MAX][256] =
+{
+        [GD_OP_NONE]                    = EMPTY,
+        [GD_OP_CREATE_VOLUME]           = "create",
+        [GD_OP_START_BRICK]             = EMPTY,
+        [GD_OP_STOP_BRICK]              = EMPTY,
+        [GD_OP_DELETE_VOLUME]           = "delete",
+        [GD_OP_START_VOLUME]            = "start",
+        [GD_OP_STOP_VOLUME]             = "stop",
+        [GD_OP_DEFRAG_VOLUME]           = EMPTY,
+        [GD_OP_ADD_BRICK]               = "add-brick",
+        [GD_OP_REMOVE_BRICK]            = "remove-brick",
+        [GD_OP_REPLACE_BRICK]           = EMPTY,
+        [GD_OP_SET_VOLUME]              = "set",
+        [GD_OP_RESET_VOLUME]            = EMPTY,
+        [GD_OP_SYNC_VOLUME]             = EMPTY,
+        [GD_OP_LOG_ROTATE]              = EMPTY,
+        [GD_OP_GSYNC_CREATE]            = "gsync-create",
+        [GD_OP_GSYNC_SET]               = EMPTY,
+        [GD_OP_PROFILE_VOLUME]          = EMPTY,
+        [GD_OP_QUOTA]                   = EMPTY,
+        [GD_OP_STATUS_VOLUME]           = EMPTY,
+        [GD_OP_REBALANCE]               = EMPTY,
+        [GD_OP_HEAL_VOLUME]             = EMPTY,
+        [GD_OP_STATEDUMP_VOLUME]        = EMPTY,
+        [GD_OP_LIST_VOLUME]             = EMPTY,
+        [GD_OP_CLEARLOCKS_VOLUME]       = EMPTY,
+        [GD_OP_DEFRAG_BRICK_VOLUME]     = EMPTY,
+};
+#undef EMPTY
+
+static inline gf_boolean_t
+glusterd_is_hook_enabled (char *script)
+{
+        return (script[0] == 'S');
+}
+
+int
+glusterd_hooks_create_hooks_directory (char *basedir)
+{
+        int  ret                                        = -1;
+        int  op                                         = GD_OP_NONE;
+        int  type                                       = GD_COMMIT_HOOK_NONE;
+        char version_dir[PATH_MAX]                      = {0, };
+        char path[PATH_MAX]                             = {0, };
+        char *cmd_subdir                                = NULL;
+        char type_subdir[GD_COMMIT_HOOK_MAX][256]       = {{0, },
+                                                           "pre",
+                                                           "post"};
+        glusterd_conf_t *priv                           = NULL;
+
+        priv = THIS->private;
+
+        snprintf (path, sizeof (path), "%s/hooks", basedir);
+        ret = mkdir_p (path, 0777, _gf_true);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_CRITICAL, "Unable to create %s due"
+                         "to %s", path, strerror (errno));
+                goto out;
+        }
+
+        GLUSTERD_GET_HOOKS_DIR (version_dir, GLUSTERD_HOOK_VER, priv);
+        ret = mkdir_p (version_dir, 0777, _gf_true);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_CRITICAL, "Unable to create %s due "
+                        "to %s", version_dir, strerror (errno));
+                goto out;
+        }
+
+        for (op = GD_OP_NONE+1; op < GD_OP_MAX; op++) {
+                cmd_subdir = glusterd_hooks_get_hooks_cmd_subdir (op);
+                if (strlen (cmd_subdir) == 0)
+                        continue;
+
+                snprintf (path, sizeof (path), "%s/%s", version_dir,
+                          cmd_subdir);
+                ret = mkdir_p (path, 0777, _gf_true);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_CRITICAL,
+                                "Unable to create %s due to %s",
+                                path, strerror (errno));
+                        goto out;
+                }
+
+                for (type = GD_COMMIT_HOOK_PRE; type < GD_COMMIT_HOOK_MAX;
+                     type++) {
+                        snprintf (path, sizeof (path), "%s/%s/%s",
+                                  version_dir, cmd_subdir, type_subdir[type]);
+                        ret = mkdir_p (path, 0777, _gf_true);
+                        if (ret) {
+                                gf_log (THIS->name, GF_LOG_CRITICAL,
+                                        "Unable to create %s due to %s",
+                                        path, strerror (errno));
+                                goto out;
+                        }
+                }
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+char*
+glusterd_hooks_get_hooks_cmd_subdir (glusterd_op_t op)
+{
+        GF_ASSERT ((op > GD_OP_NONE) && (op < GD_OP_MAX));
+
+        return glusterd_hook_dirnames[op];
+}
+
+int
+glusterd_hooks_set_volume_args (dict_t *dict, runner_t *runner)
+{
+        int     i               = 0;
+        int     count           = 0;
+        int     ret             = -1;
+        char    query[1024]     = {0,};
+        char    *key            = NULL;
+        char    *value          = NULL;
+
+        ret = dict_get_int32 (dict, "count", &count);
+        if (ret)
+                goto out;
+
+        /* This will not happen unless op_ctx
+         * is corrupted*/
+        if (!count)
+                goto out;
+
+        runner_add_arg (runner, "-o");
+        for (i = 1; (ret == 0); i++) {
+                snprintf (query, sizeof (query), "key%d", i);
+                ret = dict_get_str (dict, query, &key);
+                if (ret)
+                       continue;
+
+                snprintf (query, sizeof (query), "value%d", i);
+                ret = dict_get_str (dict, query, &value);
+                if (ret)
+                       continue;
+
+                runner_argprintf (runner, "%s=%s", key, value);
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+static int
+glusterd_hooks_add_op_args (runner_t *runner, glusterd_op_t op,
+                            dict_t *op_ctx, glusterd_commit_hook_type_t type)
+{
+        char                   *hooks_args       = NULL;
+        int                     vol_count        = 0;
+        gf_boolean_t            truth            = _gf_false;
+        glusterd_volinfo_t      *voliter         = NULL;
+        glusterd_conf_t         *priv            = NULL;
+        int                     ret              = -1;
+
+        priv = THIS->private;
+        list_for_each_entry (voliter, &priv->volumes,
+                             vol_list) {
+                if (glusterd_is_volume_started (voliter))
+                        vol_count++;
+        }
+
+        ret = 0;
+        switch (op) {
+                case GD_OP_START_VOLUME:
+                        if (type == GD_COMMIT_HOOK_PRE &&
+                            vol_count == 0)
+                                truth = _gf_true;
+
+                        else if (type == GD_COMMIT_HOOK_POST &&
+                                 vol_count == 1)
+                                truth = _gf_true;
+
+                        else
+                                truth = _gf_false;
+
+                        runner_argprintf (runner, "--first=%s",
+                                          truth? "yes":"no");
+                        break;
+
+                case GD_OP_STOP_VOLUME:
+                        if (type == GD_COMMIT_HOOK_PRE &&
+                            vol_count == 1)
+                                truth = _gf_true;
+
+                        else if (type == GD_COMMIT_HOOK_POST &&
+                                 vol_count == 0)
+                                truth = _gf_true;
+
+                        else
+                                truth = _gf_false;
+
+                        runner_argprintf (runner, "--last=%s",
+                                          truth? "yes":"no");
+                        break;
+
+                case GD_OP_SET_VOLUME:
+                        ret = glusterd_hooks_set_volume_args (op_ctx, runner);
+                        break;
+
+                case GD_OP_GSYNC_CREATE:
+                        ret = dict_get_str (op_ctx, "hooks_args", &hooks_args);
+                        if (ret)
+                                gf_log ("", GF_LOG_DEBUG,
+                                        "No Hooks Arguments.");
+                        else
+                                gf_log ("", GF_LOG_DEBUG,
+                                        "Hooks Args = %s", hooks_args);
+                        if (hooks_args)
+                                runner_argprintf (runner, "%s", hooks_args);
+                        break;
+
+                default:
+                        break;
+
+        }
+
+        return ret;
+}
+
+int
+glusterd_hooks_run_hooks (char *hooks_path, glusterd_op_t op, dict_t *op_ctx,
+                          glusterd_commit_hook_type_t type)
+{
+        xlator_t        *this                   = NULL;
+        glusterd_conf_t *priv                   = NULL;
+        runner_t        runner                  = {0, };
+        struct dirent   *entry                  = NULL;
+        DIR             *hookdir                = NULL;
+        char            *volname                = NULL;
+        char            **lines                 = NULL;
+        int             N                       = 8; /*arbitrary*/
+        int             lineno                  = 0;
+        int             line_count              = 0;
+        int             ret                     = -1;
+
+        this = THIS;
+        priv = this->private;
+
+        ret = dict_get_str (op_ctx, "volname", &volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL, "Failed to get volname "
+                        "from operation context");
+                goto out;
+        }
+
+        hookdir = opendir (hooks_path);
+        if (!hookdir) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR, "Failed to open dir %s, due "
+                        "to %s", hooks_path, strerror (errno));
+                goto out;
+        }
+
+        lines = GF_CALLOC (1, N * sizeof (*lines), gf_gld_mt_charptr);
+        if (!lines) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = -1;
+        line_count = 0;
+        glusterd_for_each_entry (entry, hookdir);
+        while (entry) {
+                if (line_count == N-1) {
+                        N *= 2;
+                        lines = GF_REALLOC (lines, N * sizeof (char *));
+                        if (!lines)
+                                goto out;
+                }
+
+                if (glusterd_is_hook_enabled (entry->d_name)) {
+                        lines[line_count] = gf_strdup (entry->d_name);
+                        line_count++;
+                }
+
+                glusterd_for_each_entry (entry, hookdir);
+        }
+
+        lines[line_count] = NULL;
+        lines = GF_REALLOC (lines, (line_count + 1) * sizeof (char *));
+        if (!lines)
+                goto out;
+
+        qsort (lines, line_count, sizeof (*lines), glusterd_compare_lines);
+
+        for (lineno = 0; lineno < line_count; lineno++) {
+
+                runinit (&runner);
+                runner_argprintf (&runner, "%s/%s", hooks_path, lines[lineno]);
+                /*Add future command line arguments to hook scripts below*/
+                runner_argprintf (&runner, "--volname=%s", volname);
+                ret = glusterd_hooks_add_op_args (&runner, op, op_ctx, type);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to add "
+                                "command specific arguments");
+                        goto out;
+                }
+
+                ret = runner_run_reuse (&runner);
+                if (ret) {
+                        runner_log (&runner, this->name, GF_LOG_ERROR,
+                                    "Failed to execute script");
+                } else {
+                        runner_log (&runner, this->name, GF_LOG_INFO,
+                                    "Ran script");
+                }
+                runner_end (&runner);
+        }
+
+        ret = 0;
+out:
+        if (lines) {
+                for (lineno = 0; lineno < line_count+1; lineno++)
+                        GF_FREE (lines[lineno]);
+
+                GF_FREE (lines);
+        }
+
+        if (hookdir)
+                closedir (hookdir);
+
+        return ret;
+}
+
+int
+glusterd_hooks_post_stub_enqueue (char *scriptdir, glusterd_op_t op,
+                                  dict_t *op_ctx)
+{
+        int                     ret     = -1;
+        glusterd_hooks_stub_t   *stub   = NULL;
+        glusterd_hooks_private_t *hooks_priv = NULL;
+        glusterd_conf_t         *conf   = NULL;
+
+        conf = THIS->private;
+        hooks_priv = conf->hooks_priv;
+
+        ret = glusterd_hooks_stub_init (&stub, scriptdir, op, op_ctx);
+        if (ret)
+                goto out;
+
+        pthread_mutex_lock (&hooks_priv->mutex);
+        {
+                hooks_priv->waitcount++;
+                list_add_tail (&stub->all_hooks, &hooks_priv->list);
+                pthread_cond_signal (&hooks_priv->cond);
+        }
+        pthread_mutex_unlock (&hooks_priv->mutex);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_hooks_stub_init (glusterd_hooks_stub_t **stub, char *scriptdir,
+                          glusterd_op_t op, dict_t *op_ctx)
+{
+        int                     ret             = -1;
+        glusterd_hooks_stub_t   *hooks_stub     = NULL;
+
+        GF_ASSERT (stub);
+        if (!stub)
+                goto out;
+
+        hooks_stub = GF_CALLOC (1, sizeof (*hooks_stub),
+                                gf_gld_mt_hooks_stub_t);
+        if (!hooks_stub)
+                goto out;
+
+        INIT_LIST_HEAD (&hooks_stub->all_hooks);
+        hooks_stub->op = op;
+        hooks_stub->scriptdir = gf_strdup (scriptdir);
+        if (!hooks_stub->scriptdir)
+                goto out;
+
+        hooks_stub->op_ctx = dict_copy_with_ref (op_ctx, hooks_stub->op_ctx);
+        if (!hooks_stub->op_ctx)
+                goto out;
+
+        *stub = hooks_stub;
+        ret = 0;
+out:
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to initialize "
+                        "post hooks stub");
+                glusterd_hooks_stub_cleanup (hooks_stub);
+        }
+
+        return ret;
+}
+
+void
+glusterd_hooks_stub_cleanup (glusterd_hooks_stub_t *stub)
+{
+        if (!stub) {
+                gf_log_callingfn (THIS->name, GF_LOG_WARNING,
+                                  "hooks_stub is NULL");
+                return;
+        }
+
+        if (stub->op_ctx)
+                dict_unref (stub->op_ctx);
+
+        GF_FREE (stub->scriptdir);
+
+        GF_FREE (stub);
+}
+
+static void*
+hooks_worker (void *args)
+{
+        glusterd_conf_t *conf = NULL;
+        glusterd_hooks_private_t *hooks_priv = NULL;
+        glusterd_hooks_stub_t *stub = NULL;
+
+        THIS = args;
+        conf = THIS->private;
+        hooks_priv = conf->hooks_priv;
+
+        for (;;) {
+                pthread_mutex_lock (&hooks_priv->mutex);
+                {
+                        while (list_empty (&hooks_priv->list)) {
+                                pthread_cond_wait (&hooks_priv->cond,
+                                                   &hooks_priv->mutex);
+                        }
+                        stub = list_entry (hooks_priv->list.next,
+                                           glusterd_hooks_stub_t,
+                                           all_hooks);
+                        list_del_init (&stub->all_hooks);
+                        hooks_priv->waitcount--;
+
+                }
+                pthread_mutex_unlock (&hooks_priv->mutex);
+
+                glusterd_hooks_run_hooks (stub->scriptdir, stub->op,
+                                          stub->op_ctx, GD_COMMIT_HOOK_POST);
+                glusterd_hooks_stub_cleanup (stub);
+        }
+
+        return NULL;
+}
+
+int
+glusterd_hooks_priv_init (glusterd_hooks_private_t **new)
+{
+        int                      ret            = -1;
+        glusterd_hooks_private_t *hooks_priv    = NULL;
+
+        if (!new)
+                goto out;
+
+        hooks_priv = GF_CALLOC (1, sizeof (*hooks_priv),
+                                gf_gld_mt_hooks_priv_t);
+        if (!hooks_priv)
+                goto out;
+
+        pthread_mutex_init (&hooks_priv->mutex, NULL);
+        pthread_cond_init (&hooks_priv->cond, NULL);
+        INIT_LIST_HEAD (&hooks_priv->list);
+        hooks_priv->waitcount = 0;
+
+        *new = hooks_priv;
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_hooks_spawn_worker (xlator_t *this)
+{
+        int                       ret          = -1;
+        glusterd_conf_t           *conf        = NULL;
+        glusterd_hooks_private_t  *hooks_priv  = NULL;
+
+
+        ret = glusterd_hooks_priv_init (&hooks_priv);
+        if (ret)
+                goto out;
+
+        conf = this->private;
+        conf->hooks_priv = hooks_priv;
+        ret = pthread_create (&hooks_priv->worker, NULL, hooks_worker,
+                              (void *)this);
+        if (ret)
+                gf_log (this->name, GF_LOG_CRITICAL, "Failed to spawn post "
+                        "hooks worker thread");
+out:
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-hooks.h b/xlators/mgmt/glusterd/src/glusterd-hooks.h
new file mode 100644
index 000000000..c597ddd2a
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-hooks.h
@@ -0,0 +1,89 @@
+/*
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _GLUSTERD_HOOKS_H_
+#define _GLUSTERD_HOOKS_H_
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include <fnmatch.h>
+
+#define GLUSTERD_GET_HOOKS_DIR(path, version, priv) \
+        snprintf (path, PATH_MAX, "%s/hooks/%d", priv->workdir,\
+                  version);
+
+#define GLUSTERD_HOOK_VER       1
+
+#define GD_HOOKS_SPECIFIC_KEY   "user.*"
+
+typedef enum glusterd_commit_hook_type {
+        GD_COMMIT_HOOK_NONE = 0,
+        GD_COMMIT_HOOK_PRE,
+        GD_COMMIT_HOOK_POST,
+        GD_COMMIT_HOOK_MAX
+} glusterd_commit_hook_type_t;
+
+typedef struct hooks_private {
+        struct list_head        list;
+        int                     waitcount; //debug purposes
+        pthread_mutex_t         mutex;
+        pthread_cond_t          cond;
+        pthread_t               worker;
+} glusterd_hooks_private_t;
+
+typedef struct hooks_stub {
+        struct list_head        all_hooks;
+        char                    *scriptdir;
+        glusterd_op_t           op;
+        dict_t                  *op_ctx;
+
+} glusterd_hooks_stub_t;
+
+
+static inline gf_boolean_t
+is_key_glusterd_hooks_friendly (char *key)
+{
+        gf_boolean_t is_friendly = _gf_false;
+
+        /* This is very specific to hooks friendly behavior */
+        if (fnmatch (GD_HOOKS_SPECIFIC_KEY, key, FNM_NOESCAPE) == 0) {
+                gf_log (THIS->name, GF_LOG_DEBUG, "user namespace key %s", key);
+                is_friendly = _gf_true;
+        }
+
+        return is_friendly;
+}
+
+int
+glusterd_hooks_create_hooks_directory (char *basedir);
+
+char *
+glusterd_hooks_get_hooks_cmd_subdir (glusterd_op_t op);
+
+int
+glusterd_hooks_run_hooks (char *hooks_path, glusterd_op_t op, dict_t *op_ctx,
+                          glusterd_commit_hook_type_t type);
+int
+glusterd_hooks_spawn_worker (xlator_t *this);
+
+int
+glusterd_hooks_stub_init (glusterd_hooks_stub_t **stub, char *scriptdir,
+                          glusterd_op_t op, dict_t *op_ctx);
+void
+glusterd_hooks_stub_cleanup (glusterd_hooks_stub_t *stub);
+
+int
+glusterd_hooks_post_stub_enqueue (char *scriptdir, glusterd_op_t op,
+                                      dict_t *op_ctx);
+int
+glusterd_hooks_priv_init (glusterd_hooks_private_t **new);
+#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-locks.c b/xlators/mgmt/glusterd/src/glusterd-locks.c
new file mode 100644
index 000000000..0af2a186f
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-locks.c
@@ -0,0 +1,637 @@
+/*
+   Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "common-utils.h"
+#include "cli1-xdr.h"
+#include "xdr-generic.h"
+#include "glusterd.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-store.h"
+#include "glusterd-utils.h"
+#include "glusterd-volgen.h"
+#include "glusterd-locks.h"
+#include "run.h"
+#include "syscall.h"
+
+#include <signal.h>
+
+#define GF_MAX_LOCKING_ENTITIES 2
+
+/* Valid entities that the mgmt_v3 lock can hold locks upon    *
+ * To add newer entities to be locked, we can just add more    *
+ * entries to this table along with the type and default value */
+valid_entities   valid_types[] = {
+        { "vol",  _gf_true  },
+        { "snap", _gf_false },
+        { NULL              },
+};
+
+static dict_t *mgmt_v3_lock;
+
+/* Checks if the lock request is for a valid entity */
+gf_boolean_t
+glusterd_mgmt_v3_is_type_valid (char *type)
+{
+        int32_t         i   = 0;
+        gf_boolean_t    ret = _gf_false;
+
+        GF_ASSERT (type);
+
+        for (i = 0; valid_types[i].type; i++) {
+                if (!strcmp (type, valid_types[i].type)) {
+                        ret = _gf_true;
+                        break;
+                }
+        }
+
+        return ret;
+}
+
+/* Initialize the global mgmt_v3 lock list(dict) when
+ * glusterd is spawned */
+int32_t
+glusterd_mgmt_v3_lock_init ()
+{
+        int32_t ret = -1;
+
+        mgmt_v3_lock = dict_new ();
+        if (!mgmt_v3_lock) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+/* Destroy the global mgmt_v3 lock list(dict) when
+ * glusterd cleanup is performed */
+void
+glusterd_mgmt_v3_lock_fini ()
+{
+        if (mgmt_v3_lock)
+                dict_destroy (mgmt_v3_lock);
+}
+
+int32_t
+glusterd_get_mgmt_v3_lock_owner (char *key, uuid_t *uuid)
+{
+        int32_t             ret      = -1;
+        mgmt_v3_lock_obj   *lock_obj = NULL;
+        uuid_t              no_owner = {"\0"};
+        xlator_t           *this = NULL;
+
+        GF_ASSERT(THIS);
+        this = THIS;
+
+        if (!key || !uuid) {
+                gf_log (this->name, GF_LOG_ERROR, "key or uuid is null.");
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_bin(mgmt_v3_lock, key, (void **) &lock_obj);
+        if (!ret)
+                uuid_copy (*uuid, lock_obj->lock_owner);
+        else
+                uuid_copy (*uuid, no_owner);
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+/* This function is called with the locked_count and type, to   *
+ * release all the acquired locks. */
+static int32_t
+glusterd_release_multiple_locks_per_entity (dict_t *dict, uuid_t uuid,
+                                            int32_t locked_count,
+                                            char *type)
+{
+        char           name_buf[PATH_MAX]    = "";
+        char          *name                  = NULL;
+        int32_t        i                     = -1;
+        int32_t        op_ret                = 0;
+        int32_t        ret                   = -1;
+        xlator_t      *this                  = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+        GF_ASSERT (dict);
+        GF_ASSERT (type);
+
+        if (locked_count == 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "No %s locked as part of this transaction",
+                        type);
+                goto out;
+        }
+
+        /* Release all the locks held */
+        for (i = 0; i < locked_count; i++) {
+                snprintf (name_buf, sizeof(name_buf),
+                          "%sname%d", type, i+1);
+
+                /* Looking for volname1, volname2 or snapname1, *
+                 * as key in the dict snapname2 */
+                ret = dict_get_str (dict, name_buf, &name);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to get %s locked_count = %d",
+                                name_buf, locked_count);
+                        op_ret = ret;
+                        continue;
+                }
+
+                ret = glusterd_mgmt_v3_unlock (name, uuid, type);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to release lock for %s.",
+                                name);
+                        op_ret = ret;
+               }
+        }
+
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", op_ret);
+        return op_ret;
+}
+
+/* Given the count and type of the entity this function acquires     *
+ * locks on multiple elements of the same entity. For example:       *
+ * If type is "vol" this function tries to acquire locks on multiple *
+ * volumes */
+static int32_t
+glusterd_acquire_multiple_locks_per_entity (dict_t *dict, uuid_t uuid,
+                                            int32_t count, char *type)
+{
+        char           name_buf[PATH_MAX]    = "";
+        char          *name                  = NULL;
+        int32_t        i                     = -1;
+        int32_t        ret                   = -1;
+        int32_t        locked_count          = 0;
+        xlator_t      *this                  = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+        GF_ASSERT (dict);
+        GF_ASSERT (type);
+
+        /* Locking one element after other */
+        for (i = 0; i < count; i++) {
+                snprintf (name_buf, sizeof(name_buf),
+                          "%sname%d", type, i+1);
+
+                /* Looking for volname1, volname2 or snapname1, *
+                 * as key in the dict snapname2 */
+                ret = dict_get_str (dict, name_buf, &name);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to get %s count = %d",
+                                name_buf, count);
+                        break;
+                }
+
+                ret = glusterd_mgmt_v3_lock (name, uuid, type);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to acquire lock for %s %s "
+                                "on behalf of %s. Reversing "
+                                "this transaction", type, name,
+                                uuid_utoa(uuid));
+                        break;
+                }
+                locked_count++;
+        }
+
+        if (count == locked_count) {
+                /* If all locking ops went successfuly, return as success */
+                ret = 0;
+                goto out;
+        }
+
+        /* If we failed to lock one element, unlock others and return failure */
+        ret = glusterd_release_multiple_locks_per_entity (dict, uuid,
+                                                          locked_count,
+                                                          type);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to release multiple %s locks",
+                        type);
+        }
+        ret = -1;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+/* Given the type of entity, this function figures out if it should unlock a   *
+ * single element of multiple elements of the said entity. For example:        *
+ * if the type is "vol", this function will accordingly unlock a single volume *
+ * or multiple volumes */
+static int32_t
+glusterd_mgmt_v3_unlock_entity (dict_t *dict, uuid_t uuid, char *type,
+                                gf_boolean_t default_value)
+{
+        char           name_buf[PATH_MAX]    = "";
+        char          *name                  = NULL;
+        int32_t        count                 = -1;
+        int32_t        ret                   = -1;
+        gf_boolean_t   hold_locks            = _gf_false;
+        xlator_t      *this                  = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+        GF_ASSERT (dict);
+        GF_ASSERT (type);
+
+        snprintf (name_buf, sizeof(name_buf), "hold_%s_locks", type);
+        hold_locks = dict_get_str_boolean (dict, name_buf, default_value);
+
+        if (hold_locks == _gf_false) {
+                /* Locks were not held for this particular entity *
+                 * Hence nothing to release */
+                ret = 0;
+                goto out;
+        }
+
+        /* Looking for volcount or snapcount in the dict */
+        snprintf (name_buf, sizeof(name_buf), "%scount", type);
+        ret = dict_get_int32 (dict, name_buf, &count);
+        if (ret) {
+                /* count is not present. Only one *
+                 * element name needs to be unlocked */
+                snprintf (name_buf, sizeof(name_buf), "%sname",
+                          type);
+                ret = dict_get_str (dict, name_buf, &name);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to fetch %sname", type);
+                        goto out;
+                }
+
+                ret = glusterd_mgmt_v3_unlock (name, uuid, type);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to release lock for %s %s "
+                                "on behalf of %s.", type, name,
+                                uuid_utoa(uuid));
+                        goto out;
+                }
+        } else {
+                /* Unlocking one element name after another */
+                ret = glusterd_release_multiple_locks_per_entity (dict,
+                                                                  uuid,
+                                                                  count,
+                                                                  type);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to release all %s locks", type);
+                        goto out;
+                }
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+/* Given the type of entity, this function figures out if it should lock a   *
+ * single element or multiple elements of the said entity. For example:      *
+ * if the type is "vol", this function will accordingly lock a single volume *
+ * or multiple volumes */
+static int32_t
+glusterd_mgmt_v3_lock_entity (dict_t *dict, uuid_t uuid, char *type,
+                              gf_boolean_t default_value)
+{
+        char           name_buf[PATH_MAX]    = "";
+        char          *name                  = NULL;
+        int32_t        count                 = -1;
+        int32_t        ret                   = -1;
+        gf_boolean_t   hold_locks            = _gf_false;
+        xlator_t      *this                  = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+        GF_ASSERT (dict);
+        GF_ASSERT (type);
+
+        snprintf (name_buf, sizeof(name_buf), "hold_%s_locks", type);
+        hold_locks = dict_get_str_boolean (dict, name_buf, default_value);
+
+        if (hold_locks == _gf_false) {
+                /* Not holding locks for this particular entity */
+                ret = 0;
+                goto out;
+        }
+
+        /* Looking for volcount or snapcount in the dict */
+        snprintf (name_buf, sizeof(name_buf), "%scount", type);
+        ret = dict_get_int32 (dict, name_buf, &count);
+        if (ret) {
+                /* count is not present. Only one *
+                 * element name needs to be locked */
+                snprintf (name_buf, sizeof(name_buf), "%sname",
+                          type);
+                ret = dict_get_str (dict, name_buf, &name);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to fetch %sname", type);
+                        goto out;
+                }
+
+                ret = glusterd_mgmt_v3_lock (name, uuid, type);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to acquire lock for %s %s "
+                                "on behalf of %s.", type, name,
+                                uuid_utoa(uuid));
+                        goto out;
+                }
+        } else {
+                /* Locking one element name after another */
+                ret = glusterd_acquire_multiple_locks_per_entity (dict,
+                                                                  uuid,
+                                                                  count,
+                                                                  type);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to acquire all %s locks", type);
+                        goto out;
+                }
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+/* Try to release locks of multiple entities like *
+ * volume, snaps etc. */
+int32_t
+glusterd_multiple_mgmt_v3_unlock (dict_t *dict, uuid_t uuid)
+{
+        int32_t        i                     = -1;
+        int32_t        ret                   = -1;
+        int32_t        op_ret                = 0;
+        xlator_t      *this                  = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+
+        if (!dict) {
+                gf_log (this->name, GF_LOG_ERROR, "dict is null.");
+                ret = -1;
+                goto out;
+        }
+
+        for (i = 0; valid_types[i].type; i++) {
+                ret = glusterd_mgmt_v3_unlock_entity
+                                            (dict, uuid,
+                                             valid_types[i].type,
+                                             valid_types[i].default_value);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to unlock all %s",
+                                valid_types[i].type);
+                        op_ret = ret;
+                }
+        }
+
+        ret = op_ret;
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+/* Try to acquire locks on multiple entities like *
+ * volume, snaps etc. */
+int32_t
+glusterd_multiple_mgmt_v3_lock (dict_t *dict, uuid_t uuid)
+{
+        int32_t        i                     = -1;
+        int32_t        ret                   = -1;
+        int32_t        locked_count          = 0;
+        xlator_t      *this                  = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+
+        if (!dict) {
+                gf_log (this->name, GF_LOG_ERROR, "dict is null.");
+                ret = -1;
+                goto out;
+        }
+
+        /* Locking one entity after other */
+        for (i = 0; valid_types[i].type; i++) {
+                ret = glusterd_mgmt_v3_lock_entity
+                                            (dict, uuid,
+                                             valid_types[i].type,
+                                             valid_types[i].default_value);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to lock all %s",
+                                valid_types[i].type);
+                        break;
+                }
+                locked_count++;
+        }
+
+        if (locked_count == GF_MAX_LOCKING_ENTITIES) {
+                /* If all locking ops went successfuly, return as success */
+                ret = 0;
+                goto out;
+        }
+
+        /* If we failed to lock one entity, unlock others and return failure */
+        for (i = 0; i < locked_count; i++) {
+                ret = glusterd_mgmt_v3_unlock_entity
+                                              (dict, uuid,
+                                               valid_types[i].type,
+                                               valid_types[i].default_value);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to unlock all %s",
+                                valid_types[i].type);
+                }
+        }
+        ret = -1;
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_mgmt_v3_lock (const char *name, uuid_t uuid, char *type)
+{
+        char                 key[PATH_MAX] = "";
+        int32_t              ret           = -1;
+        mgmt_v3_lock_obj    *lock_obj      = NULL;
+        gf_boolean_t         is_valid      = _gf_true;
+        uuid_t               owner         = {0};
+        xlator_t            *this          = NULL;
+
+        GF_ASSERT(THIS);
+        this = THIS;
+
+        if (!name || !type) {
+                gf_log (this->name, GF_LOG_ERROR, "name or type is null.");
+                ret = -1;
+                goto out;
+        }
+
+        is_valid = glusterd_mgmt_v3_is_type_valid (type);
+        if (is_valid != _gf_true) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Invalid entity. Cannot perform locking "
+                        "operation on %s types", type);
+                ret = -1;
+                goto out;
+        }
+
+        ret = snprintf (key, sizeof(key), "%s_%s", name, type);
+        if (ret != strlen(name) + 1 + strlen(type)) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR, "Unable to create key");
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Trying to acquire lock of %s %s for %s as %s",
+                type, name, uuid_utoa (uuid), key);
+
+        ret = glusterd_get_mgmt_v3_lock_owner (key, &owner);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Unable to get mgmt_v3 lock owner");
+                goto out;
+        }
+
+        /* If the lock has already been held for the given volume
+         * we fail */
+        if (!uuid_is_null (owner)) {
+                gf_log (this->name, GF_LOG_ERROR, "Lock for %s held by %s",
+                        name, uuid_utoa (owner));
+                ret = -1;
+                goto out;
+        }
+
+        lock_obj = GF_CALLOC (1, sizeof(mgmt_v3_lock_obj),
+                              gf_common_mt_mgmt_v3_lock_obj_t);
+        if (!lock_obj) {
+                ret = -1;
+                goto out;
+        }
+
+        uuid_copy (lock_obj->lock_owner, uuid);
+
+        ret = dict_set_bin (mgmt_v3_lock, key, lock_obj,
+                            sizeof(*lock_obj));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to set lock owner in mgmt_v3 lock");
+                if (lock_obj)
+                        GF_FREE (lock_obj);
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Lock for %s %s successfully held by %s",
+                type, name, uuid_utoa (uuid));
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_mgmt_v3_unlock (const char *name, uuid_t uuid, char *type)
+{
+        char           key[PATH_MAX] = "";
+        int32_t        ret           = -1;
+        gf_boolean_t   is_valid      = _gf_true;
+        uuid_t         owner         = {0};
+        xlator_t            *this          = NULL;
+
+        GF_ASSERT(THIS);
+        this = THIS;
+
+        if (!name || !type) {
+                gf_log (this->name, GF_LOG_ERROR, "name is null.");
+                ret = -1;
+                goto out;
+        }
+
+        is_valid = glusterd_mgmt_v3_is_type_valid (type);
+        if (is_valid != _gf_true) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Invalid entity. Cannot perform unlocking "
+                        "operation on %s types", type);
+                ret = -1;
+                goto out;
+        }
+
+        ret = snprintf (key, sizeof(key), "%s_%s",
+                        name, type);
+        if (ret != strlen(name) + 1 + strlen(type)) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to create key");
+                ret = -1;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Trying to release lock of %s %s for %s as %s",
+                type, name, uuid_utoa (uuid), key);
+
+        ret = glusterd_get_mgmt_v3_lock_owner (key, &owner);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Unable to get mgmt_v3 lock owner");
+                goto out;
+        }
+
+        if (uuid_is_null (owner)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Lock for %s %s not held", type, name);
+                ret = -1;
+                goto out;
+        }
+
+        ret = uuid_compare (uuid, owner);
+        if (ret) {
+
+                gf_log (this->name, GF_LOG_ERROR, "Lock owner mismatch. "
+                        "Lock for %s %s held by %s",
+                        type, name, uuid_utoa (owner));
+                goto out;
+        }
+
+        /* Removing the mgmt_v3 lock from the global list */
+        dict_del (mgmt_v3_lock, key);
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Lock for %s %s successfully released",
+                type, name);
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-locks.h b/xlators/mgmt/glusterd/src/glusterd-locks.h
new file mode 100644
index 000000000..83eb8c997
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-locks.h
@@ -0,0 +1,51 @@
+/*
+   Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _GLUSTERD_LOCKS_H_
+#define _GLUSTERD_LOCKS_H_
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+typedef struct mgmt_v3_lock_object_ {
+        uuid_t              lock_owner;
+} mgmt_v3_lock_obj;
+
+typedef struct mgmt_v3_lock_valid_entities {
+        char          *type;          /* Entity type like vol, snap */
+        gf_boolean_t   default_value; /* The default value that  *
+                                       * determines if the locks *
+                                       * should be held for that *
+                                       * entity */
+} valid_entities;
+
+int32_t
+glusterd_mgmt_v3_lock_init ();
+
+void
+glusterd_mgmt_v3_lock_fini ();
+
+int32_t
+glusterd_get_mgmt_v3_lock_owner (char *volname, uuid_t *uuid);
+
+int32_t
+glusterd_mgmt_v3_lock (const char *key, uuid_t uuid, char *type);
+
+int32_t
+glusterd_mgmt_v3_unlock (const char *key, uuid_t uuid, char *type);
+
+int32_t
+glusterd_multiple_mgmt_v3_lock (dict_t *dict, uuid_t uuid);
+
+int32_t
+glusterd_multiple_mgmt_v3_unlock (dict_t *dict, uuid_t uuid);
+
+#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-log-ops.c b/xlators/mgmt/glusterd/src/glusterd-log-ops.c
new file mode 100644
index 000000000..33bd95c03
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-log-ops.c
@@ -0,0 +1,271 @@
+/*
+   Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "common-utils.h"
+#include "cli1-xdr.h"
+#include "xdr-generic.h"
+#include "glusterd.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-store.h"
+#include "glusterd-utils.h"
+#include "glusterd-volgen.h"
+
+#include <signal.h>
+
+int
+__glusterd_handle_log_rotate (rpcsvc_request_t *req)
+{
+        int32_t                 ret     = -1;
+        gf_cli_req              cli_req = {{0,}};
+        dict_t                 *dict    = NULL;
+        glusterd_op_t           cli_op = GD_OP_LOG_ROTATE;
+        char                   *volname = NULL;
+        char                   msg[2048] = {0,};
+        xlator_t               *this = NULL;
+
+        GF_ASSERT (req);
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (msg, sizeof (msg), "Unable to decode the "
+                                  "command");
+                        goto out;
+                }
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Failed to get volume name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_INFO, "Received log rotate req "
+                "for volume %s", volname);
+
+        ret = dict_set_uint64 (dict, "rotate-key", (uint64_t)time (NULL));
+        if (ret)
+                goto out;
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_LOG_ROTATE, dict);
+
+out:
+        if (ret) {
+                if (msg[0] == '\0')
+                        snprintf (msg, sizeof (msg), "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, msg);
+        }
+
+        free (cli_req.dict.dict_val);
+        return ret;
+}
+
+int
+glusterd_handle_log_rotate (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_log_rotate);
+}
+
+/* op-sm */
+int
+glusterd_op_stage_log_rotate (dict_t *dict, char **op_errstr)
+{
+        int                                     ret = -1;
+        char                                    *volname = NULL;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        gf_boolean_t                            exists = _gf_false;
+        char                                    msg[2048] = {0};
+        char                                    *brick = NULL;
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        exists = glusterd_check_volume_exists (volname);
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (!exists) {
+                snprintf (msg, sizeof (msg), "Volume %s does not exist",
+                          volname);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                ret = -1;
+                goto out;
+        }
+
+        if (_gf_false == glusterd_is_volume_started (volinfo)) {
+                snprintf (msg, sizeof (msg), "Volume %s needs to be started before"
+                          " log rotate.", volname);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "brick", &brick);
+        /* If no brick is specified, do log-rotate for
+           all the bricks in the volume */
+        if (ret) {
+                ret = 0;
+                goto out;
+        }
+
+        ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo, NULL);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Incorrect brick %s "
+                          "for volume %s", brick, volname);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+
+int
+glusterd_op_log_rotate (dict_t *dict)
+{
+        int                   ret                = -1;
+        glusterd_conf_t      *priv               = NULL;
+        glusterd_volinfo_t   *volinfo            = NULL;
+        glusterd_brickinfo_t *brickinfo          = NULL;
+        xlator_t             *this               = NULL;
+        char                 *volname            = NULL;
+        char                 *brick              = NULL;
+        char                  logfile[PATH_MAX]  = {0,};
+        char                  pidfile[PATH_MAX]  = {0,};
+        FILE                 *file               = NULL;
+        pid_t                 pid                = 0;
+        uint64_t              key                = 0;
+        int                   valid_brick        = 0;
+        glusterd_brickinfo_t *tmpbrkinfo         = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "volname not found");
+                goto out;
+        }
+
+        ret = dict_get_uint64 (dict, "rotate-key", &key);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "rotate key not found");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "brick", &brick);
+        /* If no brick is specified, do log-rotate for
+           all the bricks in the volume */
+        if (ret)
+                goto cont;
+
+        ret = glusterd_brickinfo_new_from_brick (brick, &tmpbrkinfo);
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR,
+                        "cannot get brickinfo from brick");
+                goto out;
+        }
+
+cont:
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret)
+                goto out;
+
+        ret = -1;
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                if (uuid_compare (brickinfo->uuid, MY_UUID))
+                        continue;
+
+                if (brick &&
+                    (strcmp (tmpbrkinfo->hostname, brickinfo->hostname) ||
+                     strcmp (tmpbrkinfo->path,brickinfo->path)))
+                        continue;
+
+                valid_brick = 1;
+
+                GLUSTERD_GET_BRICK_PIDFILE (pidfile, volinfo, brickinfo, priv);
+                file = fopen (pidfile, "r+");
+                if (!file) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to open pidfile: %s",
+                                pidfile);
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = fscanf (file, "%d", &pid);
+                if (ret <= 0) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to read pidfile: %s",
+                                pidfile);
+                        ret = -1;
+                        goto out;
+                }
+                fclose (file);
+                file = NULL;
+
+                snprintf (logfile, PATH_MAX, "%s.%"PRIu64,
+                          brickinfo->logfile, key);
+
+                ret = rename (brickinfo->logfile, logfile);
+                if (ret)
+                        gf_log ("", GF_LOG_WARNING, "rename failed");
+
+                ret = kill (pid, SIGHUP);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to SIGHUP to %d", pid);
+                        goto out;
+                }
+                ret = 0;
+
+                /* If request was for brick, only one iteration is enough */
+                if (brick)
+                        break;
+        }
+
+        if (ret && !valid_brick)
+                ret = 0;
+
+out:
+        if (tmpbrkinfo)
+                glusterd_brickinfo_delete (tmpbrkinfo);
+
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-mem-types.h b/xlators/mgmt/glusterd/src/glusterd-mem-types.h
index 8b99e9e4c..e6f6a0333 100644
--- a/xlators/mgmt/glusterd/src/glusterd-mem-types.h
+++ b/xlators/mgmt/glusterd/src/glusterd-mem-types.h
@@ -1,29 +1,19 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
 
-
 #ifndef __GLUSTERD_MEM_TYPES_H__
 #define __GLUSTERD_MEM_TYPES_H__
 
 #include "mem-types.h"
 
-enum gf_gld_mem_types_ {
+typedef enum gf_gld_mem_types_ {
         gf_gld_mt_dir_entry_t                   = gf_common_mt_end + 1,
         gf_gld_mt_volfile_ctx                   = gf_common_mt_end + 2,
         gf_gld_mt_glusterd_state_t              = gf_common_mt_end + 3,
@@ -58,7 +48,28 @@ enum gf_gld_mem_types_ {
         gf_gld_mt_log_filename_ctx_t            = gf_common_mt_end + 32,
         gf_gld_mt_log_locate_ctx_t              = gf_common_mt_end + 33,
         gf_gld_mt_log_rotate_ctx_t              = gf_common_mt_end + 34,
-        gf_gld_mt_end                           = gf_common_mt_end + 35
-};
+        gf_gld_mt_peerctx_t                     = gf_common_mt_end + 35,
+        gf_gld_mt_sm_tr_log_t                   = gf_common_mt_end + 36,
+        gf_gld_mt_pending_node_t                = gf_common_mt_end + 37,
+        gf_gld_mt_brick_rsp_ctx_t               = gf_common_mt_end + 38,
+        gf_gld_mt_mop_brick_req_t               = gf_common_mt_end + 39,
+        gf_gld_mt_op_allack_ctx_t               = gf_common_mt_end + 40,
+        gf_gld_mt_linearr                       = gf_common_mt_end + 41,
+        gf_gld_mt_linebuf                       = gf_common_mt_end + 42,
+        gf_gld_mt_mount_pattern                 = gf_common_mt_end + 43,
+        gf_gld_mt_mount_comp_container          = gf_common_mt_end + 44,
+        gf_gld_mt_mount_component               = gf_common_mt_end + 45,
+        gf_gld_mt_mount_spec                    = gf_common_mt_end + 46,
+        gf_gld_mt_georep_meet_spec              = gf_common_mt_end + 47,
+        gf_gld_mt_nodesrv_t                     = gf_common_mt_end + 48,
+        gf_gld_mt_charptr                       = gf_common_mt_end + 49,
+        gf_gld_mt_hooks_stub_t                  = gf_common_mt_end + 50,
+        gf_gld_mt_hooks_priv_t                  = gf_common_mt_end + 51,
+        gf_gld_mt_mop_commit_req_t              = gf_common_mt_end + 52,
+        gf_gld_mt_int                           = gf_common_mt_end + 53,
+        gf_gld_mt_snap_t                        = gf_common_mt_end + 54,
+        gf_gld_mt_missed_snapinfo_t             = gf_common_mt_end + 55,
+        gf_gld_mt_end                           = gf_common_mt_end + 56,
+} gf_gld_mem_types_t;
 #endif
 
diff --git a/xlators/mgmt/glusterd/src/glusterd-mgmt-handler.c b/xlators/mgmt/glusterd/src/glusterd-mgmt-handler.c
new file mode 100644
index 000000000..a5f38ce9c
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-mgmt-handler.c
@@ -0,0 +1,924 @@
+/*
+   Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+/* rpc related syncops */
+#include "rpc-clnt.h"
+#include "protocol-common.h"
+#include "xdr-generic.h"
+#include "glusterd1-xdr.h"
+#include "glusterd-syncop.h"
+
+#include "glusterd.h"
+#include "glusterd-utils.h"
+#include "glusterd-locks.h"
+#include "glusterd-mgmt.h"
+#include "glusterd-op-sm.h"
+
+static int
+glusterd_mgmt_v3_null (rpcsvc_request_t *req)
+{
+        return 0;
+}
+
+static int
+glusterd_mgmt_v3_lock_send_resp (rpcsvc_request_t *req, int32_t status)
+{
+
+        gd1_mgmt_v3_lock_rsp          rsp   = {{0},};
+        int                           ret   = -1;
+        xlator_t                     *this  = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        rsp.op_ret = status;
+        if (rsp.op_ret)
+               rsp.op_errno = errno;
+
+        glusterd_get_uuid (&rsp.uuid);
+
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp);
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Responded to mgmt_v3 lock, ret: %d", ret);
+
+        return ret;
+}
+
+static int
+glusterd_synctasked_mgmt_v3_lock (rpcsvc_request_t *req,
+                                  gd1_mgmt_v3_lock_req *lock_req,
+                                  glusterd_op_lock_ctx_t *ctx)
+{
+        int32_t                         ret         = -1;
+        xlator_t                       *this        = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (ctx);
+        GF_ASSERT (ctx->dict);
+
+        /* Trying to acquire multiple mgmt_v3 locks */
+        ret = glusterd_multiple_mgmt_v3_lock (ctx->dict, ctx->uuid);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to acquire mgmt_v3 locks for %s",
+                         uuid_utoa (ctx->uuid));
+
+        ret = glusterd_mgmt_v3_lock_send_resp (req, ret);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_op_state_machine_mgmt_v3_lock (rpcsvc_request_t *req,
+                                        gd1_mgmt_v3_lock_req *lock_req,
+                                        glusterd_op_lock_ctx_t *ctx)
+{
+        int32_t                         ret         = -1;
+        xlator_t                       *this        = NULL;
+        glusterd_op_info_t              txn_op_info = {{0},};
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        glusterd_txn_opinfo_init (&txn_op_info, NULL, &lock_req->op,
+                                  ctx->dict, req);
+
+        ret = glusterd_set_txn_opinfo (&lock_req->txn_id, &txn_op_info);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to set transaction's opinfo");
+                goto out;
+        }
+
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_LOCK,
+                                           &lock_req->txn_id, ctx);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to inject event GD_OP_EVENT_LOCK");
+
+out:
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_handle_mgmt_v3_lock_fn (rpcsvc_request_t *req)
+{
+        gd1_mgmt_v3_lock_req            lock_req    = {{0},};
+        int32_t                         ret         = -1;
+        glusterd_peerinfo_t            *peerinfo    = NULL;
+        glusterd_op_lock_ctx_t         *ctx         = NULL;
+        xlator_t                       *this        = NULL;
+        gf_boolean_t                    is_synctasked = _gf_false;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        ret = xdr_to_generic (req->msg[0], &lock_req,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_lock_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode lock "
+                        "request received from peer");
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "Received mgmt_v3 lock req "
+                "from uuid: %s", uuid_utoa (lock_req.uuid));
+
+        if (glusterd_friend_find_by_uuid (lock_req.uuid, &peerinfo)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s doesn't "
+                        "belong to the cluster. Ignoring request.",
+                        uuid_utoa (lock_req.uuid));
+                ret = -1;
+                goto out;
+        }
+
+        ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_op_lock_ctx_t);
+        if (!ctx) {
+                ret = -1;
+                goto out;
+        }
+
+        uuid_copy (ctx->uuid, lock_req.uuid);
+        ctx->req = req;
+
+        ctx->dict = dict_new ();
+        if (!ctx->dict) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_unserialize (lock_req.dict.dict_val,
+                                lock_req.dict.dict_len, &ctx->dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to unserialize the dictionary");
+                goto out;
+        }
+
+        is_synctasked = dict_get_str_boolean (ctx->dict,
+                                              "is_synctasked", _gf_false);
+        if (is_synctasked)
+                ret = glusterd_synctasked_mgmt_v3_lock (req, &lock_req, ctx);
+        else
+                ret = glusterd_op_state_machine_mgmt_v3_lock (req, &lock_req,
+                                                              ctx);
+
+out:
+
+        if (ret) {
+                if (ctx->dict)
+                        dict_unref (ctx->dict);
+                if (ctx)
+                        GF_FREE (ctx);
+        }
+
+        free (lock_req.dict.dict_val);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_mgmt_v3_pre_validate_send_resp (rpcsvc_request_t   *req,
+                                        int32_t op, int32_t status,
+                                        char *op_errstr, dict_t *rsp_dict)
+{
+        gd1_mgmt_v3_pre_val_rsp          rsp      = {{0},};
+        int                             ret      = -1;
+        xlator_t                       *this     = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        rsp.op_ret = status;
+        glusterd_get_uuid (&rsp.uuid);
+        rsp.op = op;
+        if (op_errstr)
+                rsp.op_errstr = op_errstr;
+        else
+                rsp.op_errstr = "";
+
+        ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val,
+                                           &rsp.dict.dict_len);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get serialized length of dict");
+                goto out;
+        }
+
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gd1_mgmt_v3_pre_val_rsp);
+
+        GF_FREE (rsp.dict.dict_val);
+out:
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Responded to pre validation, ret: %d", ret);
+        return ret;
+}
+
+static int
+glusterd_handle_pre_validate_fn (rpcsvc_request_t *req)
+{
+        int32_t                         ret       = -1;
+        gd1_mgmt_v3_pre_val_req          op_req    = {{0},};
+        glusterd_peerinfo_t            *peerinfo  = NULL;
+        xlator_t                       *this      = NULL;
+        char                           *op_errstr = NULL;
+        dict_t                         *dict      = NULL;
+        dict_t                         *rsp_dict  = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        ret = xdr_to_generic (req->msg[0], &op_req,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_pre_val_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to decode pre validation "
+                        "request received from peer");
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s doesn't "
+                        "belong to the cluster. Ignoring request.",
+                        uuid_utoa (op_req.uuid));
+                ret = -1;
+                goto out;
+        }
+
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        ret = dict_unserialize (op_req.dict.dict_val,
+                                op_req.dict.dict_len, &dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to unserialize the dictionary");
+                goto out;
+        }
+
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get new dictionary");
+                return -1;
+        }
+
+        ret = gd_mgmt_v3_pre_validate_fn (op_req.op, dict, &op_errstr,
+                                         rsp_dict);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Pre Validation failed on operation %s",
+                        gd_op_list[op_req.op]);
+        }
+
+        ret = glusterd_mgmt_v3_pre_validate_send_resp (req, op_req.op,
+                                                      ret, op_errstr,
+                                                      rsp_dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to send Pre Validation "
+                        "response for operation %s",
+                        gd_op_list[op_req.op]);
+                goto out;
+        }
+
+out:
+        if (op_errstr && (strcmp (op_errstr, "")))
+                GF_FREE (op_errstr);
+
+        free (op_req.dict.dict_val);
+
+        if (dict)
+                dict_unref (dict);
+
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_mgmt_v3_brick_op_send_resp (rpcsvc_request_t   *req,
+                                    int32_t op, int32_t status,
+                                    char *op_errstr, dict_t *rsp_dict)
+{
+        gd1_mgmt_v3_brick_op_rsp         rsp      = {{0},};
+        int                             ret      = -1;
+        xlator_t                       *this     = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        rsp.op_ret = status;
+        glusterd_get_uuid (&rsp.uuid);
+        rsp.op = op;
+        if (op_errstr)
+                rsp.op_errstr = op_errstr;
+        else
+                rsp.op_errstr = "";
+
+        ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val,
+                                           &rsp.dict.dict_len);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get serialized length of dict");
+                goto out;
+        }
+
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gd1_mgmt_v3_brick_op_rsp);
+
+        GF_FREE (rsp.dict.dict_val);
+out:
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Responded to brick op, ret: %d", ret);
+        return ret;
+}
+
+static int
+glusterd_handle_brick_op_fn (rpcsvc_request_t *req)
+{
+        int32_t                         ret       = -1;
+        gd1_mgmt_v3_brick_op_req         op_req    = {{0},};
+        glusterd_peerinfo_t            *peerinfo  = NULL;
+        xlator_t                       *this      = NULL;
+        char                           *op_errstr = NULL;
+        dict_t                         *dict      = NULL;
+        dict_t                         *rsp_dict  = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        ret = xdr_to_generic (req->msg[0], &op_req,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_brick_op_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode brick op "
+                        "request received from peer");
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s doesn't "
+                        "belong to the cluster. Ignoring request.",
+                        uuid_utoa (op_req.uuid));
+                ret = -1;
+                goto out;
+        }
+
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        ret = dict_unserialize (op_req.dict.dict_val,
+                                op_req.dict.dict_len, &dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to unserialize the dictionary");
+                goto out;
+        }
+
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get new dictionary");
+                return -1;
+        }
+
+        ret = gd_mgmt_v3_brick_op_fn (op_req.op, dict, &op_errstr,
+                                     rsp_dict);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Brick Op failed on operation %s",
+                        gd_op_list[op_req.op]);
+        }
+
+        ret = glusterd_mgmt_v3_brick_op_send_resp (req, op_req.op,
+                                                  ret, op_errstr,
+                                                  rsp_dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to send brick op "
+                        "response for operation %s",
+                        gd_op_list[op_req.op]);
+                goto out;
+        }
+
+out:
+        if (op_errstr && (strcmp (op_errstr, "")))
+                GF_FREE (op_errstr);
+
+        free (op_req.dict.dict_val);
+
+        if (dict)
+                dict_unref (dict);
+
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_mgmt_v3_commit_send_resp (rpcsvc_request_t   *req,
+                                  int32_t op, int32_t status,
+                                  char *op_errstr, dict_t *rsp_dict)
+{
+        gd1_mgmt_v3_commit_rsp           rsp      = {{0},};
+        int                             ret      = -1;
+        xlator_t                       *this     = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        rsp.op_ret = status;
+        glusterd_get_uuid (&rsp.uuid);
+        rsp.op = op;
+        if (op_errstr)
+                rsp.op_errstr = op_errstr;
+        else
+                rsp.op_errstr = "";
+
+        ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val,
+                                           &rsp.dict.dict_len);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get serialized length of dict");
+                goto out;
+        }
+
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gd1_mgmt_v3_commit_rsp);
+
+        GF_FREE (rsp.dict.dict_val);
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Responded to commit, ret: %d", ret);
+        return ret;
+}
+
+static int
+glusterd_handle_commit_fn (rpcsvc_request_t *req)
+{
+        int32_t                         ret       = -1;
+        gd1_mgmt_v3_commit_req           op_req    = {{0},};
+        glusterd_peerinfo_t            *peerinfo  = NULL;
+        xlator_t                       *this      = NULL;
+        char                           *op_errstr = NULL;
+        dict_t                         *dict      = NULL;
+        dict_t                         *rsp_dict  = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        ret = xdr_to_generic (req->msg[0], &op_req,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_commit_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode commit "
+                        "request received from peer");
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s doesn't "
+                        "belong to the cluster. Ignoring request.",
+                        uuid_utoa (op_req.uuid));
+                ret = -1;
+                goto out;
+        }
+
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        ret = dict_unserialize (op_req.dict.dict_val,
+                                op_req.dict.dict_len, &dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to unserialize the dictionary");
+                goto out;
+        }
+
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get new dictionary");
+                return -1;
+        }
+
+        ret = gd_mgmt_v3_commit_fn (op_req.op, dict, &op_errstr,
+                                   rsp_dict);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "commit failed on operation %s",
+                        gd_op_list[op_req.op]);
+        }
+
+        ret = glusterd_mgmt_v3_commit_send_resp (req, op_req.op,
+                                                ret, op_errstr,
+                                                rsp_dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to send commit "
+                        "response for operation %s",
+                        gd_op_list[op_req.op]);
+                goto out;
+        }
+
+out:
+        if (op_errstr && (strcmp (op_errstr, "")))
+                GF_FREE (op_errstr);
+
+        free (op_req.dict.dict_val);
+
+        if (dict)
+                dict_unref (dict);
+
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_mgmt_v3_post_validate_send_resp (rpcsvc_request_t   *req,
+                                         int32_t op, int32_t status,
+                                         char *op_errstr, dict_t *rsp_dict)
+{
+        gd1_mgmt_v3_post_val_rsp         rsp      = {{0},};
+        int                             ret      = -1;
+        xlator_t                       *this     = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        rsp.op_ret = status;
+        glusterd_get_uuid (&rsp.uuid);
+        rsp.op = op;
+        if (op_errstr)
+                rsp.op_errstr = op_errstr;
+        else
+                rsp.op_errstr = "";
+
+        ret = dict_allocate_and_serialize (rsp_dict, &rsp.dict.dict_val,
+                                           &rsp.dict.dict_len);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get serialized length of dict");
+                goto out;
+        }
+
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gd1_mgmt_v3_post_val_rsp);
+
+        GF_FREE (rsp.dict.dict_val);
+out:
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Responded to post validation, ret: %d", ret);
+        return ret;
+}
+
+static int
+glusterd_handle_post_validate_fn (rpcsvc_request_t *req)
+{
+        int32_t                         ret       = -1;
+        gd1_mgmt_v3_post_val_req         op_req    = {{0},};
+        glusterd_peerinfo_t            *peerinfo  = NULL;
+        xlator_t                       *this      = NULL;
+        char                           *op_errstr = NULL;
+        dict_t                         *dict      = NULL;
+        dict_t                         *rsp_dict  = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        ret = xdr_to_generic (req->msg[0], &op_req,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_post_val_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to decode post validation "
+                        "request received from peer");
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (glusterd_friend_find_by_uuid (op_req.uuid, &peerinfo)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s doesn't "
+                        "belong to the cluster. Ignoring request.",
+                        uuid_utoa (op_req.uuid));
+                ret = -1;
+                goto out;
+        }
+
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        ret = dict_unserialize (op_req.dict.dict_val,
+                                op_req.dict.dict_len, &dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to unserialize the dictionary");
+                goto out;
+        }
+
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get new dictionary");
+                return -1;
+        }
+
+        ret = gd_mgmt_v3_post_validate_fn (op_req.op, op_req.op_ret, dict,
+                                           &op_errstr, rsp_dict);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Post Validation failed on operation %s",
+                        gd_op_list[op_req.op]);
+        }
+
+        ret = glusterd_mgmt_v3_post_validate_send_resp (req, op_req.op,
+                                                       ret, op_errstr,
+                                                       rsp_dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to send Post Validation "
+                        "response for operation %s",
+                        gd_op_list[op_req.op]);
+                goto out;
+        }
+
+out:
+        if (op_errstr && (strcmp (op_errstr, "")))
+                GF_FREE (op_errstr);
+
+        free (op_req.dict.dict_val);
+
+        if (dict)
+                dict_unref (dict);
+
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_mgmt_v3_unlock_send_resp (rpcsvc_request_t *req, int32_t status)
+{
+
+        gd1_mgmt_v3_unlock_rsp          rsp   = {{0},};
+        int                             ret   = -1;
+        xlator_t                       *this  = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        rsp.op_ret = status;
+        if (rsp.op_ret)
+               rsp.op_errno = errno;
+
+        glusterd_get_uuid (&rsp.uuid);
+
+        ret = glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
+                                     (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp);
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Responded to mgmt_v3 unlock, ret: %d", ret);
+
+        return ret;
+}
+
+static int
+glusterd_synctasked_mgmt_v3_unlock (rpcsvc_request_t *req,
+                                    gd1_mgmt_v3_unlock_req *unlock_req,
+                                    glusterd_op_lock_ctx_t *ctx)
+{
+        int32_t                         ret         = -1;
+        xlator_t                       *this        = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (ctx);
+
+        /* Trying to release multiple mgmt_v3 locks */
+        ret = glusterd_multiple_mgmt_v3_unlock (ctx->dict, ctx->uuid);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to release mgmt_v3 locks for %s",
+                        uuid_utoa(ctx->uuid));
+        }
+
+        ret = glusterd_mgmt_v3_unlock_send_resp (req, ret);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+
+static int
+glusterd_op_state_machine_mgmt_v3_unlock (rpcsvc_request_t *req,
+                                          gd1_mgmt_v3_unlock_req *lock_req,
+                                          glusterd_op_lock_ctx_t *ctx)
+{
+        int32_t                           ret      = -1;
+        xlator_t                         *this     = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_UNLOCK,
+                                           &lock_req->txn_id, ctx);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to inject event GD_OP_EVENT_UNLOCK");
+
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_handle_mgmt_v3_unlock_fn (rpcsvc_request_t *req)
+{
+        gd1_mgmt_v3_unlock_req            lock_req = {{0},};
+        int32_t                           ret      = -1;
+        glusterd_op_lock_ctx_t           *ctx      = NULL;
+        glusterd_peerinfo_t              *peerinfo = NULL;
+        xlator_t                         *this     = NULL;
+        gf_boolean_t                      is_synctasked = _gf_false;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        ret = xdr_to_generic (req->msg[0], &lock_req,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_unlock_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode unlock "
+                        "request received from peer");
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "Received mgmt_v3 unlock req "
+                "from uuid: %s", uuid_utoa (lock_req.uuid));
+
+        if (glusterd_friend_find_by_uuid (lock_req.uuid, &peerinfo)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s doesn't "
+                        "belong to the cluster. Ignoring request.",
+                        uuid_utoa (lock_req.uuid));
+                ret = -1;
+                goto out;
+        }
+
+        ctx = GF_CALLOC (1, sizeof (*ctx), gf_gld_mt_op_lock_ctx_t);
+        if (!ctx) {
+                ret = -1;
+                goto out;
+        }
+
+        uuid_copy (ctx->uuid, lock_req.uuid);
+        ctx->req = req;
+
+        ctx->dict = dict_new ();
+        if (!ctx->dict) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_unserialize (lock_req.dict.dict_val,
+                                lock_req.dict.dict_len, &ctx->dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to unserialize the dictionary");
+                goto out;
+        }
+
+        is_synctasked = dict_get_str_boolean (ctx->dict,
+                                              "is_synctasked", _gf_false);
+        if (is_synctasked)
+                ret = glusterd_synctasked_mgmt_v3_unlock (req, &lock_req, ctx);
+        else
+                ret = glusterd_op_state_machine_mgmt_v3_unlock (req, &lock_req,
+                                                                ctx);
+
+out:
+
+        if (ret) {
+                if (ctx->dict)
+                        dict_unref (ctx->dict);
+                if (ctx)
+                        GF_FREE (ctx);
+        }
+
+        free (lock_req.dict.dict_val);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_handle_mgmt_v3_lock (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            glusterd_handle_mgmt_v3_lock_fn);
+}
+
+static int
+glusterd_handle_pre_validate (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            glusterd_handle_pre_validate_fn);
+}
+
+static int
+glusterd_handle_brick_op (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            glusterd_handle_brick_op_fn);
+}
+
+static int
+glusterd_handle_commit (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            glusterd_handle_commit_fn);
+}
+
+static int
+glusterd_handle_post_validate (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            glusterd_handle_post_validate_fn);
+}
+
+int
+glusterd_handle_mgmt_v3_unlock (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            glusterd_handle_mgmt_v3_unlock_fn);
+}
+
+rpcsvc_actor_t gd_svc_mgmt_v3_actors[] = {
+        [GLUSTERD_MGMT_V3_NULL]          = { "NULL",           GLUSTERD_MGMT_V3_NULL,          glusterd_mgmt_v3_null,          NULL, 0, DRC_NA},
+        [GLUSTERD_MGMT_V3_LOCK]          = { "MGMT_V3_LOCK",   GLUSTERD_MGMT_V3_LOCK,          glusterd_handle_mgmt_v3_lock,   NULL, 0, DRC_NA},
+        [GLUSTERD_MGMT_V3_PRE_VALIDATE]  = { "PRE_VAL",        GLUSTERD_MGMT_V3_PRE_VALIDATE,  glusterd_handle_pre_validate,   NULL, 0, DRC_NA},
+        [GLUSTERD_MGMT_V3_BRICK_OP]      = { "BRCK_OP",        GLUSTERD_MGMT_V3_BRICK_OP,      glusterd_handle_brick_op,       NULL, 0, DRC_NA},
+        [GLUSTERD_MGMT_V3_COMMIT]        = { "COMMIT",         GLUSTERD_MGMT_V3_COMMIT,        glusterd_handle_commit,         NULL, 0, DRC_NA},
+        [GLUSTERD_MGMT_V3_POST_VALIDATE] = { "POST_VAL",       GLUSTERD_MGMT_V3_POST_VALIDATE, glusterd_handle_post_validate,  NULL, 0, DRC_NA},
+        [GLUSTERD_MGMT_V3_UNLOCK]        = { "MGMT_V3_UNLOCK", GLUSTERD_MGMT_V3_UNLOCK,        glusterd_handle_mgmt_v3_unlock, NULL, 0, DRC_NA},
+};
+
+struct rpcsvc_program gd_svc_mgmt_v3_prog = {
+        .progname  = "GlusterD svc mgmt v3",
+        .prognum   = GD_MGMT_V3_PROGRAM,
+        .progver   = GD_MGMT_V3_VERSION,
+        .numactors = GLUSTERD_MGMT_V3_MAXVALUE,
+        .actors    = gd_svc_mgmt_v3_actors,
+        .synctask  = _gf_true,
+};
diff --git a/xlators/mgmt/glusterd/src/glusterd-mgmt.c b/xlators/mgmt/glusterd/src/glusterd-mgmt.c
new file mode 100644
index 000000000..d52532e54
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-mgmt.c
@@ -0,0 +1,1893 @@
+/*
+   Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+/* rpc related syncops */
+#include "rpc-clnt.h"
+#include "protocol-common.h"
+#include "xdr-generic.h"
+#include "glusterd1-xdr.h"
+#include "glusterd-syncop.h"
+
+#include "glusterd.h"
+#include "glusterd-utils.h"
+#include "glusterd-locks.h"
+#include "glusterd-mgmt.h"
+#include "glusterd-op-sm.h"
+
+extern struct rpc_clnt_program gd_mgmt_v3_prog;
+
+
+static void
+gd_mgmt_v3_collate_errors (struct syncargs *args, int op_ret, int op_errno,
+                           char *op_errstr, int op_code,
+                           glusterd_peerinfo_t *peerinfo, u_char *uuid)
+{
+        char      *peer_str          = NULL;
+        char       err_str[PATH_MAX] = "Please check log file for details.";
+        char       op_err[PATH_MAX]  = "";
+        int32_t    len               = -1;
+        xlator_t  *this              = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (args);
+        GF_ASSERT (uuid);
+
+        if (op_ret) {
+                args->op_ret = op_ret;
+                args->op_errno = op_errno;
+
+                if (peerinfo)
+                        peer_str = peerinfo->hostname;
+                else
+                        peer_str = uuid_utoa (uuid);
+
+                if (op_errstr && strcmp (op_errstr, "")) {
+                        len = snprintf (err_str, sizeof(err_str) - 1,
+                                        "Error: %s", op_errstr);
+                        err_str[len] = '\0';
+                }
+
+                switch (op_code){
+                case GLUSTERD_MGMT_V3_LOCK:
+                        {
+                                len = snprintf (op_err, sizeof(op_err) - 1,
+                                                "Locking failed "
+                                                "on %s. %s", peer_str, err_str);
+                                break;
+                        }
+                case GLUSTERD_MGMT_V3_PRE_VALIDATE:
+                        {
+                                len = snprintf (op_err, sizeof(op_err) - 1,
+                                                "Pre Validation failed "
+                                                "on %s. %s", peer_str, err_str);
+                                break;
+                        }
+                case GLUSTERD_MGMT_V3_BRICK_OP:
+                        {
+                                len = snprintf (op_err, sizeof(op_err) - 1,
+                                                "Brick ops failed "
+                                                "on %s. %s", peer_str, err_str);
+                                break;
+                        }
+                case GLUSTERD_MGMT_V3_COMMIT:
+                        {
+                                len = snprintf (op_err, sizeof(op_err) - 1,
+                                                "Commit failed on %s. %s",
+                                                peer_str, err_str);
+                                break;
+                        }
+                case GLUSTERD_MGMT_V3_POST_VALIDATE:
+                        {
+                                len = snprintf (op_err, sizeof(op_err) - 1,
+                                                "Post Validation failed "
+                                                "on %s. %s", peer_str, err_str);
+                                break;
+                        }
+                case GLUSTERD_MGMT_V3_UNLOCK:
+                        {
+                                len = snprintf (op_err, sizeof(op_err) - 1,
+                                                "Unlocking failed "
+                                                "on %s. %s", peer_str, err_str);
+                                break;
+                        }
+                }
+                op_err[len] = '\0';
+
+                if (args->errstr) {
+                        len = snprintf (err_str, sizeof(err_str) - 1,
+                                        "%s\n%s", args->errstr,
+                                        op_err);
+                        GF_FREE (args->errstr);
+                        args->errstr = NULL;
+                } else
+                        len = snprintf (err_str, sizeof(err_str) - 1,
+                                        "%s", op_err);
+                err_str[len] = '\0';
+
+                gf_log (this->name, GF_LOG_ERROR, "%s", op_err);
+                args->errstr = gf_strdup (err_str);
+        }
+
+        return;
+}
+
+int32_t
+gd_mgmt_v3_pre_validate_fn (glusterd_op_t op, dict_t *dict,
+                            char **op_errstr, dict_t *rsp_dict)
+{
+        int32_t       ret = -1;
+        xlator_t     *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (rsp_dict);
+
+        switch (op) {
+               case GD_OP_SNAP:
+               {
+                       ret = glusterd_snapshot_prevalidate (dict, op_errstr,
+                                                            rsp_dict);
+                       if (ret) {
+                               gf_log (this->name, GF_LOG_WARNING,
+                                       "Snapshot Prevalidate Failed");
+                               goto out;
+                       }
+
+                       break;
+               }
+               default:
+                       break;
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "OP = %d. Returning %d", op, ret);
+        return ret;
+}
+
+int32_t
+gd_mgmt_v3_brick_op_fn (glusterd_op_t op, dict_t *dict,
+                        char **op_errstr, dict_t *rsp_dict)
+{
+        int32_t       ret = -1;
+        xlator_t     *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (rsp_dict);
+
+        switch (op) {
+        case GD_OP_SNAP:
+        {
+                ret = glusterd_snapshot_brickop (dict, op_errstr, rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "snapshot brickop "
+                                "failed");
+                        goto out;
+                }
+                break;
+        }
+        default:
+                break;
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "OP = %d. Returning %d", op, ret);
+        return ret;
+}
+
+int32_t
+gd_mgmt_v3_commit_fn (glusterd_op_t op, dict_t *dict,
+                      char **op_errstr, dict_t *rsp_dict)
+{
+        int32_t       ret = -1;
+        xlator_t     *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (rsp_dict);
+
+        switch (op) {
+               case GD_OP_SNAP:
+               {
+                       ret = glusterd_snapshot (dict, op_errstr, rsp_dict);
+                       if (ret) {
+                               gf_log (this->name, GF_LOG_WARNING,
+                                       "Snapshot Commit Failed");
+                               goto out;
+                       }
+                       break;
+               }
+               default:
+                       break;
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "OP = %d. Returning %d", op, ret);
+        return ret;
+}
+
+int32_t
+gd_mgmt_v3_post_validate_fn (glusterd_op_t op, int32_t op_ret, dict_t *dict,
+                             char **op_errstr, dict_t *rsp_dict)
+{
+        int32_t       ret = -1;
+        xlator_t     *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (rsp_dict);
+
+        switch (op) {
+               case GD_OP_SNAP:
+               {
+                       ret = glusterd_snapshot_postvalidate (dict, op_ret,
+                                                             op_errstr,
+                                                             rsp_dict);
+                       if (ret) {
+                               gf_log (this->name, GF_LOG_WARNING,
+                                       "postvalidate operation failed");
+                               goto out;
+                       }
+                       break;
+               }
+               default:
+                       break;
+        }
+
+        ret = 0;
+
+out:
+        gf_log (this->name, GF_LOG_TRACE, "OP = %d. Returning %d", op, ret);
+        return ret;
+}
+
+int32_t
+gd_mgmt_v3_lock_cbk_fn (struct rpc_req *req, struct iovec *iov,
+                        int count, void *myframe)
+{
+        int32_t                     ret        = -1;
+        struct syncargs            *args       = NULL;
+        glusterd_peerinfo_t        *peerinfo   = NULL;
+        gd1_mgmt_v3_lock_rsp        rsp        = {{0},};
+        call_frame_t               *frame      = NULL;
+        int32_t                     op_ret     = -1;
+        int32_t                     op_errno   = -1;
+        xlator_t                   *this       = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (myframe);
+
+        /* Even though the lock command has failed, while collating the errors
+           (gd_mgmt_v3_collate_errors), args->op_ret and args->op_errno will be
+           used. @args is obtained from frame->local. So before checking the
+           status of the request and going out if its a failure, args should be
+           set to frame->local. Otherwise, while collating args will be NULL.
+           This applies to other phases such as prevalidate, brickop, commit and
+           postvalidate also.
+        */
+        frame  = myframe;
+        args   = frame->local;
+        peerinfo = frame->cookie;
+        frame->local = NULL;
+        frame->cookie = NULL;
+
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        if (!iov) {
+                gf_log (this->name, GF_LOG_ERROR, "iov is NULL");
+                op_errno = EINVAL;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp);
+        if (ret < 0)
+                goto out;
+
+        uuid_copy (args->uuid, rsp.uuid);
+
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+
+out:
+        gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL,
+                                   GLUSTERD_MGMT_V3_LOCK,
+                                   peerinfo, rsp.uuid);
+        free (rsp.dict.dict_val);
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+        return 0;
+}
+
+int32_t
+gd_mgmt_v3_lock_cbk (struct rpc_req *req, struct iovec *iov,
+                     int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        gd_mgmt_v3_lock_cbk_fn);
+}
+
+int
+gd_mgmt_v3_lock (glusterd_op_t op, dict_t *op_ctx,
+                 glusterd_peerinfo_t *peerinfo,
+                 struct syncargs *args, uuid_t my_uuid,
+                 uuid_t recv_uuid)
+{
+        gd1_mgmt_v3_lock_req     req  = {{0},};
+        glusterd_conf_t         *conf = THIS->private;
+        int32_t                  ret  = -1;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (op_ctx);
+        GF_ASSERT (peerinfo);
+        GF_ASSERT (args);
+
+        ret = dict_allocate_and_serialize (op_ctx,
+                                           &req.dict.dict_val,
+                                           &req.dict.dict_len);
+        if (ret)
+                goto out;
+
+        uuid_copy (req.uuid, my_uuid);
+        req.op = op;
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo,
+                                        &gd_mgmt_v3_prog,
+                                        GLUSTERD_MGMT_V3_LOCK,
+                                        gd_mgmt_v3_lock_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_v3_lock_req);
+        synclock_lock (&conf->big_lock);
+out:
+        GF_FREE (req.dict.dict_val);
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_mgmt_v3_initiate_lockdown (glusterd_conf_t  *conf, glusterd_op_t op,
+                                    dict_t *dict, char **op_errstr, int npeers,
+                                    gf_boolean_t  *is_acquired)
+{
+        char                *volname    = NULL;
+        glusterd_peerinfo_t *peerinfo   = NULL;
+        int32_t              ret        = -1;
+        int32_t              peer_cnt   = 0;
+        struct syncargs      args       = {0};
+        struct list_head    *peers      = NULL;
+        uuid_t               peer_uuid  = {0};
+        xlator_t            *this       = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (conf);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (is_acquired);
+
+        peers = &conf->xaction_peers;
+
+        /* Trying to acquire multiple mgmt_v3 locks on local node */
+        ret = glusterd_multiple_mgmt_v3_lock (dict, MY_UUID);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to acquire mgmt_v3 locks on localhost");
+                goto out;
+        }
+
+        *is_acquired = _gf_true;
+
+        if (!npeers) {
+                ret = 0;
+                goto out;
+        }
+
+        /* Sending mgmt_v3 lock req to other nodes in the cluster */
+        gd_syncargs_init (&args, NULL);
+        synctask_barrier_init((&args));
+        peer_cnt = 0;
+        list_for_each_entry (peerinfo, peers, op_peers_list) {
+                gd_mgmt_v3_lock (op, dict, peerinfo, &args,
+                                 MY_UUID, peer_uuid);
+                peer_cnt++;
+        }
+        gd_synctask_barrier_wait((&args), peer_cnt);
+
+        if (args.errstr)
+                *op_errstr = gf_strdup (args.errstr);
+
+        ret = args.op_ret;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Sent lock op req for %s "
+                "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret);
+out:
+        if (ret) {
+                if (*op_errstr)
+                        gf_log (this->name, GF_LOG_ERROR, "%s",
+                                *op_errstr);
+
+                if (volname)
+                        ret = gf_asprintf (op_errstr,
+                                           "Another transaction is in progress "
+                                           "for %s. Please try again after "
+                                           "sometime.", volname);
+                else
+                        ret = gf_asprintf (op_errstr,
+                                           "Another transaction is in progress "
+                                           "Please try again after sometime.");
+
+                if (ret == -1)
+                        *op_errstr = NULL;
+
+                ret = -1;
+        }
+
+        return ret;
+}
+
+int
+glusterd_pre_validate_aggr_rsp_dict (glusterd_op_t op, dict_t *aggr, dict_t *rsp)
+{
+        int32_t              ret      = 0;
+        xlator_t            *this     = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (aggr);
+        GF_ASSERT (rsp);
+
+        switch (op) {
+        case GD_OP_SNAP:
+                ret = glusterd_snap_pre_validate_use_rsp_dict (aggr, rsp);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to aggregate prevalidate "
+                                "response dictionaries.");
+                        goto out;
+                }
+                break;
+        default:
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR, "Invalid op (%s)", gd_op_list[op]);
+
+                break;
+        }
+out:
+        return ret;
+}
+
+int32_t
+gd_mgmt_v3_pre_validate_cbk_fn (struct rpc_req *req, struct iovec *iov,
+                                int count, void *myframe)
+{
+        int32_t                     ret        = -1;
+        struct syncargs            *args       = NULL;
+        glusterd_peerinfo_t        *peerinfo   = NULL;
+        gd1_mgmt_v3_pre_val_rsp      rsp        = {{0},};
+        call_frame_t               *frame      = NULL;
+        int32_t                     op_ret     = -1;
+        int32_t                     op_errno   = -1;
+        dict_t                     *rsp_dict   = NULL;
+        xlator_t                   *this       = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (myframe);
+
+        frame  = myframe;
+        args   = frame->local;
+        peerinfo = frame->cookie;
+        frame->local = NULL;
+        frame->cookie = NULL;
+
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        if (!iov) {
+                gf_log (this->name, GF_LOG_ERROR, "iov is NULL");
+                op_errno = EINVAL;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_pre_val_rsp);
+        if (ret < 0)
+                goto out;
+
+        if (rsp.dict.dict_len) {
+                /* Unserialize the dictionary */
+                rsp_dict  = dict_new ();
+
+                ret = dict_unserialize (rsp.dict.dict_val,
+                                        rsp.dict.dict_len,
+                                        &rsp_dict);
+                if (ret < 0) {
+                        GF_FREE (rsp.dict.dict_val);
+                        goto out;
+                } else {
+                        rsp_dict->extra_stdfree = rsp.dict.dict_val;
+                }
+        }
+
+        uuid_copy (args->uuid, rsp.uuid);
+        pthread_mutex_lock (&args->lock_dict);
+        {
+                ret = glusterd_pre_validate_aggr_rsp_dict (rsp.op, args->dict,
+                                                           rsp_dict);
+        }
+        pthread_mutex_unlock (&args->lock_dict);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s",
+                        "Failed to aggregate response from "
+                        " node/brick");
+                if (!rsp.op_ret)
+                        op_ret = ret;
+                else {
+                        op_ret = rsp.op_ret;
+                        op_errno = rsp.op_errno;
+                }
+        } else {
+                op_ret = rsp.op_ret;
+                op_errno = rsp.op_errno;
+        }
+
+out:
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL,
+                                  GLUSTERD_MGMT_V3_PRE_VALIDATE,
+                                  peerinfo, rsp.uuid);
+
+        if (rsp.op_errstr)
+                free (rsp.op_errstr);
+
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+        return 0;
+}
+
+int32_t
+gd_mgmt_v3_pre_validate_cbk (struct rpc_req *req, struct iovec *iov,
+                             int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        gd_mgmt_v3_pre_validate_cbk_fn);
+}
+
+int
+gd_mgmt_v3_pre_validate_req (glusterd_op_t op, dict_t *op_ctx,
+                             glusterd_peerinfo_t *peerinfo,
+                             struct syncargs *args, uuid_t my_uuid,
+                             uuid_t recv_uuid)
+{
+        int32_t                  ret   = -1;
+        gd1_mgmt_v3_pre_val_req  req   = {{0},};
+        glusterd_conf_t         *conf  = THIS->private;
+        xlator_t                *this  = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (op_ctx);
+        GF_ASSERT (peerinfo);
+        GF_ASSERT (args);
+
+        ret = dict_allocate_and_serialize (op_ctx,
+                                           &req.dict.dict_val,
+                                           &req.dict.dict_len);
+        if (ret)
+                goto out;
+
+        uuid_copy (req.uuid, my_uuid);
+        req.op = op;
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo,
+                                        &gd_mgmt_v3_prog,
+                                        GLUSTERD_MGMT_V3_PRE_VALIDATE,
+                                        gd_mgmt_v3_pre_validate_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_v3_pre_val_req);
+        synclock_lock (&conf->big_lock);
+out:
+        GF_FREE (req.dict.dict_val);
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_mgmt_v3_pre_validate (glusterd_conf_t  *conf, glusterd_op_t op,
+                               dict_t *req_dict, char **op_errstr, int npeers)
+{
+        int32_t              ret        = -1;
+        int32_t              peer_cnt   = 0;
+        dict_t              *rsp_dict   = NULL;
+        glusterd_peerinfo_t *peerinfo   = NULL;
+        struct syncargs      args       = {0};
+        struct list_head    *peers      = NULL;
+        uuid_t               peer_uuid  = {0};
+        xlator_t            *this       = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (conf);
+        GF_ASSERT (req_dict);
+        GF_ASSERT (op_errstr);
+
+        peers = &conf->xaction_peers;
+
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to create response dictionary");
+                goto out;
+        }
+
+        /* Pre Validation on local node */
+        ret = gd_mgmt_v3_pre_validate_fn (op, req_dict, op_errstr,
+                                          rsp_dict);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Pre Validation failed for "
+                        "operation %s on local node",
+                        gd_op_list[op]);
+
+                if (*op_errstr == NULL) {
+                        ret = gf_asprintf (op_errstr,
+                                           "Pre-validation failed "
+                                           "on localhost. Please "
+                                           "check log file for details");
+                        if (ret == -1)
+                                *op_errstr = NULL;
+
+                        ret = -1;
+                }
+                goto out;
+        }
+
+        ret = glusterd_pre_validate_aggr_rsp_dict (op, req_dict,
+                                                   rsp_dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s",
+                        "Failed to aggregate response from "
+                        " node/brick");
+                goto out;
+        }
+
+        dict_unref (rsp_dict);
+        rsp_dict = NULL;
+
+        if (!npeers) {
+                ret = 0;
+                goto out;
+        }
+
+        /* Sending Pre Validation req to other nodes in the cluster */
+        gd_syncargs_init (&args, req_dict);
+        synctask_barrier_init((&args));
+        peer_cnt = 0;
+        list_for_each_entry (peerinfo, peers, op_peers_list) {
+                gd_mgmt_v3_pre_validate_req (op, req_dict, peerinfo, &args,
+                                             MY_UUID, peer_uuid);
+                peer_cnt++;
+        }
+        gd_synctask_barrier_wait((&args), peer_cnt);
+
+        if (args.op_ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Pre Validation failed on peers");
+
+                if (args.errstr)
+                         *op_errstr = gf_strdup (args.errstr);
+        }
+
+        ret = args.op_ret;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Sent pre valaidation req for %s "
+                "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret);
+out:
+        return ret;
+}
+
+int
+glusterd_mgmt_v3_build_payload (dict_t **req, char **op_errstr, dict_t *dict,
+                                glusterd_op_t op)
+{
+        int32_t                 ret      = -1;
+        dict_t                 *req_dict = NULL;
+        xlator_t               *this     = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (dict);
+
+        req_dict = dict_new ();
+        if (!req_dict)
+                goto out;
+
+        switch (op) {
+                case GD_OP_SNAP:
+                        dict_copy (dict, req_dict);
+                        break;
+                default:
+                        break;
+        }
+
+        *req = req_dict;
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+gd_mgmt_v3_brick_op_cbk_fn (struct rpc_req *req, struct iovec *iov,
+                            int count, void *myframe)
+{
+        int32_t                     ret        = -1;
+        struct syncargs            *args       = NULL;
+        glusterd_peerinfo_t        *peerinfo   = NULL;
+        gd1_mgmt_v3_brick_op_rsp     rsp        = {{0},};
+        call_frame_t               *frame      = NULL;
+        int32_t                     op_ret     = -1;
+        int32_t                     op_errno   = -1;
+        xlator_t                   *this       = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (myframe);
+
+        frame  = myframe;
+        args   = frame->local;
+        peerinfo = frame->cookie;
+        frame->local = NULL;
+        frame->cookie = NULL;
+
+        /* If the operation failed, then iov can be NULL. So better check the
+           status of the operation and then worry about iov (if the status of
+           the command is success)
+        */
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        if (!iov) {
+                gf_log (this->name, GF_LOG_ERROR, "iov is NULL");
+                op_errno = EINVAL;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_brick_op_rsp);
+        if (ret < 0)
+                goto out;
+
+        uuid_copy (args->uuid, rsp.uuid);
+
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+
+out:
+        gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL,
+                                   GLUSTERD_MGMT_V3_BRICK_OP,
+                                   peerinfo, rsp.uuid);
+
+        if (rsp.op_errstr)
+                free (rsp.op_errstr);
+
+        free (rsp.dict.dict_val);
+
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+        return 0;
+}
+
+int32_t
+gd_mgmt_v3_brick_op_cbk (struct rpc_req *req, struct iovec *iov,
+                         int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        gd_mgmt_v3_brick_op_cbk_fn);
+}
+
+int
+gd_mgmt_v3_brick_op_req (glusterd_op_t op, dict_t *op_ctx,
+                         glusterd_peerinfo_t *peerinfo,
+                         struct syncargs *args, uuid_t my_uuid,
+                         uuid_t recv_uuid)
+{
+        int32_t                   ret  = -1;
+        gd1_mgmt_v3_brick_op_req  req  = {{0},};
+        glusterd_conf_t          *conf = THIS->private;
+        xlator_t                 *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (op_ctx);
+        GF_ASSERT (peerinfo);
+        GF_ASSERT (args);
+
+        ret = dict_allocate_and_serialize (op_ctx,
+                                           &req.dict.dict_val,
+                                           &req.dict.dict_len);
+        if (ret)
+                goto out;
+
+        uuid_copy (req.uuid, my_uuid);
+        req.op = op;
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo,
+                                        &gd_mgmt_v3_prog,
+                                        GLUSTERD_MGMT_V3_BRICK_OP,
+                                        gd_mgmt_v3_brick_op_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_v3_brick_op_req);
+        synclock_lock (&conf->big_lock);
+out:
+        GF_FREE (req.dict.dict_val);
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_mgmt_v3_brick_op (glusterd_conf_t  *conf, glusterd_op_t op,
+                           dict_t *req_dict, char **op_errstr, int npeers)
+{
+        int32_t              ret        = -1;
+        int32_t              peer_cnt   = 0;
+        dict_t              *rsp_dict   = NULL;
+        glusterd_peerinfo_t *peerinfo   = NULL;
+        struct syncargs      args       = {0};
+        struct list_head    *peers      = NULL;
+        uuid_t               peer_uuid  = {0};
+        xlator_t            *this       = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (conf);
+        GF_ASSERT (req_dict);
+        GF_ASSERT (op_errstr);
+
+        peers = &conf->xaction_peers;
+
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to create response dictionary");
+                goto out;
+        }
+
+        /* Perform brick op on local node */
+        ret = gd_mgmt_v3_brick_op_fn (op, req_dict, op_errstr,
+                                     rsp_dict);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Brick ops failed for "
+                        "operation %s on local node",
+                        gd_op_list[op]);
+
+                if (*op_errstr == NULL) {
+                        ret = gf_asprintf (op_errstr,
+                                           "Brick ops failed "
+                                           "on localhost. Please "
+                                           "check log file for details");
+                        if (ret == -1)
+                                *op_errstr = NULL;
+
+                        ret = -1;
+                }
+                goto out;
+        }
+
+        dict_unref (rsp_dict);
+        rsp_dict = NULL;
+
+        if (!npeers) {
+                ret = 0;
+                goto out;
+        }
+
+        /* Sending brick op req to other nodes in the cluster */
+        gd_syncargs_init (&args, NULL);
+        synctask_barrier_init((&args));
+        peer_cnt = 0;
+        list_for_each_entry (peerinfo, peers, op_peers_list) {
+                gd_mgmt_v3_brick_op_req (op, req_dict, peerinfo, &args,
+                                         MY_UUID, peer_uuid);
+                peer_cnt++;
+        }
+        gd_synctask_barrier_wait((&args), peer_cnt);
+
+        if (args.op_ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Brick ops failed on peers");
+
+                if (args.errstr)
+                         *op_errstr = gf_strdup (args.errstr);
+        }
+
+        ret = args.op_ret;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Sent brick op req for %s "
+                "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret);
+out:
+        return ret;
+}
+
+int32_t
+gd_mgmt_v3_commit_cbk_fn (struct rpc_req *req, struct iovec *iov,
+                          int count, void *myframe)
+{
+        int32_t                     ret        = -1;
+        struct syncargs            *args       = NULL;
+        glusterd_peerinfo_t        *peerinfo   = NULL;
+        gd1_mgmt_v3_commit_rsp       rsp        = {{0},};
+        call_frame_t               *frame      = NULL;
+        int32_t                     op_ret     = -1;
+        int32_t                     op_errno   = -1;
+        dict_t                     *rsp_dict   = NULL;
+        xlator_t                   *this       = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (myframe);
+
+        frame  = myframe;
+        args   = frame->local;
+        peerinfo = frame->cookie;
+        frame->local = NULL;
+        frame->cookie = NULL;
+
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        if (!iov) {
+                gf_log (this->name, GF_LOG_ERROR, "iov is NULL");
+                op_errno = EINVAL;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_commit_rsp);
+        if (ret < 0)
+                goto out;
+
+        if (rsp.dict.dict_len) {
+                /* Unserialize the dictionary */
+                rsp_dict  = dict_new ();
+
+                ret = dict_unserialize (rsp.dict.dict_val,
+                                        rsp.dict.dict_len,
+                                        &rsp_dict);
+                if (ret < 0) {
+                        GF_FREE (rsp.dict.dict_val);
+                        goto out;
+                } else {
+                        rsp_dict->extra_stdfree = rsp.dict.dict_val;
+                }
+        }
+
+        uuid_copy (args->uuid, rsp.uuid);
+        pthread_mutex_lock (&args->lock_dict);
+        {
+                ret = glusterd_syncop_aggr_rsp_dict (rsp.op, args->dict,
+                                                     rsp_dict);
+        }
+        pthread_mutex_unlock (&args->lock_dict);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s",
+                        "Failed to aggregate response from "
+                        " node/brick");
+                if (!rsp.op_ret)
+                        op_ret = ret;
+                else {
+                        op_ret = rsp.op_ret;
+                        op_errno = rsp.op_errno;
+                }
+        } else {
+                op_ret = rsp.op_ret;
+                op_errno = rsp.op_errno;
+        }
+
+out:
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL,
+                                  GLUSTERD_MGMT_V3_COMMIT,
+                                  peerinfo, rsp.uuid);
+
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+        return 0;
+}
+
+int32_t
+gd_mgmt_v3_commit_cbk (struct rpc_req *req, struct iovec *iov,
+                       int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        gd_mgmt_v3_commit_cbk_fn);
+}
+
+int
+gd_mgmt_v3_commit_req (glusterd_op_t op, dict_t *op_ctx,
+                       glusterd_peerinfo_t *peerinfo,
+                       struct syncargs *args, uuid_t my_uuid,
+                       uuid_t recv_uuid)
+{
+        int32_t                  ret  = -1;
+        gd1_mgmt_v3_commit_req   req  = {{0},};
+        glusterd_conf_t         *conf = THIS->private;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (op_ctx);
+        GF_ASSERT (peerinfo);
+        GF_ASSERT (args);
+
+        ret = dict_allocate_and_serialize (op_ctx,
+                                           &req.dict.dict_val,
+                                           &req.dict.dict_len);
+        if (ret)
+                goto out;
+
+        uuid_copy (req.uuid, my_uuid);
+        req.op = op;
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo,
+                                        &gd_mgmt_v3_prog,
+                                        GLUSTERD_MGMT_V3_COMMIT,
+                                        gd_mgmt_v3_commit_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_v3_commit_req);
+        synclock_lock (&conf->big_lock);
+out:
+        GF_FREE (req.dict.dict_val);
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_mgmt_v3_commit (glusterd_conf_t  *conf, glusterd_op_t op,
+                         dict_t *op_ctx, dict_t *req_dict,
+                         char **op_errstr, int npeers)
+{
+        int32_t              ret        = -1;
+        int32_t              peer_cnt   = 0;
+        dict_t              *rsp_dict   = NULL;
+        glusterd_peerinfo_t *peerinfo   = NULL;
+        struct syncargs      args       = {0};
+        struct list_head    *peers      = NULL;
+        uuid_t               peer_uuid  = {0};
+        xlator_t            *this       = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (conf);
+        GF_ASSERT (op_ctx);
+        GF_ASSERT (req_dict);
+        GF_ASSERT (op_errstr);
+
+        peers = &conf->xaction_peers;
+
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to create response dictionary");
+                goto out;
+        }
+
+        /* Commit on local node */
+        ret = gd_mgmt_v3_commit_fn (op, req_dict, op_errstr,
+                                   rsp_dict);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Commit failed for "
+                        "operation %s on local node",
+                        gd_op_list[op]);
+
+                if (*op_errstr == NULL) {
+                        ret = gf_asprintf (op_errstr,
+                                           "Commit failed "
+                                           "on localhost. Please "
+                                           "check log file for details.");
+                        if (ret == -1)
+                                *op_errstr = NULL;
+
+                        ret = -1;
+                }
+                goto out;
+        }
+
+        ret = glusterd_syncop_aggr_rsp_dict (op, op_ctx,
+                                             rsp_dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s",
+                        "Failed to aggregate response from "
+                        " node/brick");
+                goto out;
+        }
+
+        dict_unref (rsp_dict);
+        rsp_dict = NULL;
+
+        if (!npeers) {
+                ret = 0;
+                goto out;
+        }
+
+        /* Sending commit req to other nodes in the cluster */
+        gd_syncargs_init (&args, op_ctx);
+        synctask_barrier_init((&args));
+        peer_cnt = 0;
+        list_for_each_entry (peerinfo, peers, op_peers_list) {
+                gd_mgmt_v3_commit_req (op, req_dict, peerinfo, &args,
+                                       MY_UUID, peer_uuid);
+                peer_cnt++;
+        }
+        gd_synctask_barrier_wait((&args), peer_cnt);
+
+        if (args.op_ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Commit failed on peers");
+
+                if (args.errstr)
+                         *op_errstr = gf_strdup (args.errstr);
+        }
+
+        ret = args.op_ret;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Sent commit req for %s to %d "
+                "peers. Returning %d", gd_op_list[op], peer_cnt, ret);
+out:
+        return ret;
+}
+
+int32_t
+gd_mgmt_v3_post_validate_cbk_fn (struct rpc_req *req, struct iovec *iov,
+                                 int count, void *myframe)
+{
+        int32_t                     ret        = -1;
+        struct syncargs            *args       = NULL;
+        glusterd_peerinfo_t        *peerinfo   = NULL;
+        gd1_mgmt_v3_post_val_rsp     rsp        = {{0},};
+        call_frame_t               *frame      = NULL;
+        int32_t                     op_ret     = -1;
+        int32_t                     op_errno   = -1;
+        xlator_t                   *this       = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (myframe);
+
+        frame  = myframe;
+        args   = frame->local;
+        peerinfo = frame->cookie;
+        frame->local = NULL;
+        frame->cookie = NULL;
+
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        if (!iov) {
+                gf_log (this->name, GF_LOG_ERROR, "iov is NULL");
+                op_errno = EINVAL;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_post_val_rsp);
+        if (ret < 0)
+                goto out;
+
+        uuid_copy (args->uuid, rsp.uuid);
+
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+
+out:
+        gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL,
+                                  GLUSTERD_MGMT_V3_POST_VALIDATE,
+                                  peerinfo, rsp.uuid);
+        if (rsp.op_errstr)
+                free (rsp.op_errstr);
+
+        free (rsp.dict.dict_val);
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+        return 0;
+}
+
+int32_t
+gd_mgmt_v3_post_validate_cbk (struct rpc_req *req, struct iovec *iov,
+                              int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        gd_mgmt_v3_post_validate_cbk_fn);
+}
+
+int
+gd_mgmt_v3_post_validate_req (glusterd_op_t op, int32_t op_ret, dict_t *op_ctx,
+                              glusterd_peerinfo_t *peerinfo,
+                              struct syncargs *args, uuid_t my_uuid,
+                              uuid_t recv_uuid)
+{
+        int32_t                   ret  = -1;
+        gd1_mgmt_v3_post_val_req  req  = {{0},};
+        glusterd_conf_t          *conf = THIS->private;
+        xlator_t                 *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (op_ctx);
+        GF_ASSERT (peerinfo);
+        GF_ASSERT (args);
+
+        ret = dict_allocate_and_serialize (op_ctx,
+                                           &req.dict.dict_val,
+                                           &req.dict.dict_len);
+        if (ret)
+                goto out;
+
+        uuid_copy (req.uuid, my_uuid);
+        req.op = op;
+        req.op_ret = op_ret;
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo,
+                                        &gd_mgmt_v3_prog,
+                                        GLUSTERD_MGMT_V3_POST_VALIDATE,
+                                        gd_mgmt_v3_post_validate_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_v3_post_val_req);
+        synclock_lock (&conf->big_lock);
+out:
+        GF_FREE (req.dict.dict_val);
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_mgmt_v3_post_validate (glusterd_conf_t  *conf, glusterd_op_t op,
+                                int32_t op_ret, dict_t *dict, dict_t *req_dict,
+                                char **op_errstr, int npeers)
+{
+        int32_t              ret        = -1;
+        int32_t              peer_cnt   = 0;
+        dict_t              *rsp_dict   = NULL;
+        glusterd_peerinfo_t *peerinfo   = NULL;
+        struct syncargs      args       = {0};
+        struct list_head    *peers      = NULL;
+        uuid_t               peer_uuid  = {0};
+        xlator_t            *this       = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (conf);
+        GF_ASSERT (dict);
+        GF_ASSERT (req_dict);
+        GF_ASSERT (op_errstr);
+
+        peers = &conf->xaction_peers;
+        GF_ASSERT (peers);
+
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to create response dictionary");
+                goto out;
+        }
+
+        /* Copy the contents of dict like missed snaps info to req_dict */
+        dict_copy (dict, req_dict);
+
+        /* Post Validation on local node */
+        ret = gd_mgmt_v3_post_validate_fn (op, op_ret, req_dict, op_errstr,
+                                           rsp_dict);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Post Validation failed for "
+                        "operation %s on local node",
+                        gd_op_list[op]);
+
+                if (*op_errstr == NULL) {
+                        ret = gf_asprintf (op_errstr,
+                                           "Post-validation failed "
+                                           "on localhost. Please check "
+                                           "log file for details");
+                        if (ret == -1)
+                                *op_errstr = NULL;
+
+                        ret = -1;
+                }
+                goto out;
+        }
+
+        dict_unref (rsp_dict);
+        rsp_dict = NULL;
+
+        if (!npeers) {
+                ret = 0;
+                goto out;
+        }
+
+        /* Sending Post Validation req to other nodes in the cluster */
+        gd_syncargs_init (&args, req_dict);
+        synctask_barrier_init((&args));
+        peer_cnt = 0;
+        list_for_each_entry (peerinfo, peers, op_peers_list) {
+                gd_mgmt_v3_post_validate_req (op, op_ret, req_dict, peerinfo,
+                                              &args, MY_UUID, peer_uuid);
+                peer_cnt++;
+        }
+        gd_synctask_barrier_wait((&args), peer_cnt);
+
+        if (args.op_ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Post Validation failed on peers");
+
+                if (args.errstr)
+                         *op_errstr = gf_strdup (args.errstr);
+        }
+
+        ret = args.op_ret;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Sent post valaidation req for %s "
+                "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret);
+out:
+        return ret;
+}
+
+int32_t
+gd_mgmt_v3_unlock_cbk_fn (struct rpc_req *req, struct iovec *iov,
+                          int count, void *myframe)
+{
+        int32_t                     ret        = -1;
+        struct syncargs            *args       = NULL;
+        glusterd_peerinfo_t        *peerinfo   = NULL;
+        gd1_mgmt_v3_unlock_rsp      rsp        = {{0},};
+        call_frame_t               *frame      = NULL;
+        int32_t                     op_ret     = -1;
+        int32_t                     op_errno   = -1;
+        xlator_t                   *this       = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (myframe);
+
+        frame  = myframe;
+        args   = frame->local;
+        peerinfo = frame->cookie;
+        frame->local = NULL;
+        frame->cookie = NULL;
+
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        if (!iov) {
+                gf_log (this->name, GF_LOG_ERROR, "iov is NULL");
+                op_errno = EINVAL;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp);
+        if (ret < 0)
+                goto out;
+
+        uuid_copy (args->uuid, rsp.uuid);
+
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+
+out:
+        gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL,
+                                   GLUSTERD_MGMT_V3_UNLOCK,
+                                   peerinfo, rsp.uuid);
+        free (rsp.dict.dict_val);
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+        return 0;
+}
+
+int32_t
+gd_mgmt_v3_unlock_cbk (struct rpc_req *req, struct iovec *iov,
+                       int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        gd_mgmt_v3_unlock_cbk_fn);
+}
+
+int
+gd_mgmt_v3_unlock (glusterd_op_t op, dict_t *op_ctx,
+                   glusterd_peerinfo_t *peerinfo,
+                   struct syncargs *args, uuid_t my_uuid,
+                   uuid_t recv_uuid)
+{
+        int32_t                  ret  = -1;
+        gd1_mgmt_v3_unlock_req   req  = {{0},};
+        glusterd_conf_t         *conf = THIS->private;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (op_ctx);
+        GF_ASSERT (peerinfo);
+        GF_ASSERT (args);
+
+        ret = dict_allocate_and_serialize (op_ctx,
+                                           &req.dict.dict_val,
+                                           &req.dict.dict_len);
+        if (ret)
+                goto out;
+
+        uuid_copy (req.uuid, my_uuid);
+        req.op = op;
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo,
+                                        &gd_mgmt_v3_prog,
+                                        GLUSTERD_MGMT_V3_UNLOCK,
+                                        gd_mgmt_v3_unlock_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_v3_unlock_req);
+        synclock_lock (&conf->big_lock);
+out:
+        GF_FREE (req.dict.dict_val);
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_mgmt_v3_release_peer_locks (glusterd_conf_t  *conf, glusterd_op_t op,
+                                     dict_t *dict, int32_t op_ret,
+                                     char **op_errstr, int npeers,
+                                     gf_boolean_t  is_acquired)
+{
+        int32_t              ret        = -1;
+        int32_t              peer_cnt   = 0;
+        uuid_t               peer_uuid  = {0};
+        xlator_t            *this       = NULL;
+        glusterd_peerinfo_t *peerinfo   = NULL;
+        struct syncargs      args       = {0};
+        struct list_head    *peers      = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (conf);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+
+        peers = &conf->xaction_peers;
+
+        /* If the lock has not been held during this
+         * transaction, do not send unlock requests */
+        if (!is_acquired)
+                goto out;
+
+        if (!npeers) {
+                ret = 0;
+                goto out;
+        }
+
+        /* Sending mgmt_v3 unlock req to other nodes in the cluster */
+        gd_syncargs_init (&args, NULL);
+        synctask_barrier_init((&args));
+        peer_cnt = 0;
+        list_for_each_entry (peerinfo, peers, op_peers_list) {
+                gd_mgmt_v3_unlock (op, dict, peerinfo, &args,
+                                   MY_UUID, peer_uuid);
+                peer_cnt++;
+        }
+        gd_synctask_barrier_wait((&args), peer_cnt);
+
+        if (args.op_ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unlock failed on peers");
+
+                if (!op_ret && args.errstr)
+                         *op_errstr = gf_strdup (args.errstr);
+        }
+
+        ret = args.op_ret;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Sent unlock op req for %s "
+                "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret);
+
+out:
+        return ret;
+}
+
+int32_t
+glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op,
+                                      dict_t *dict)
+{
+        int32_t                     ret              = -1;
+        int32_t                     op_ret           = -1;
+        int32_t                     npeers           = 0;
+        dict_t                      *req_dict        = NULL;
+        dict_t                      *tmp_dict        = NULL;
+        glusterd_conf_t             *conf            = NULL;
+        char                        *op_errstr       = NULL;
+        xlator_t                    *this            = NULL;
+        gf_boolean_t                is_acquired      = _gf_false;
+        uuid_t                      *originator_uuid = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (dict);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        /* Save the MY_UUID as the originator_uuid. This originator_uuid
+         * will be used by is_origin_glusterd() to determine if a node
+         * is the originator node for a command. */
+        originator_uuid = GF_CALLOC (1, sizeof(uuid_t),
+                                     gf_common_mt_uuid_t);
+        if (!originator_uuid) {
+                ret = -1;
+                goto out;
+        }
+
+        uuid_copy (*originator_uuid, MY_UUID);
+        ret = dict_set_bin (dict, "originator_uuid",
+                            originator_uuid, sizeof (uuid_t));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set originator_uuid.");
+                goto out;
+        }
+
+        /* Marking the operation as complete synctasked */
+        ret = dict_set_int32 (dict, "is_synctasked", _gf_true);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set synctasked flag.");
+                goto out;
+        }
+
+        /* Use a copy at local unlock as cli response will be sent before
+         * the unlock and the volname in the dict might be removed */
+        tmp_dict = dict_new();
+        if (!tmp_dict) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to create dict");
+                goto out;
+        }
+        dict_copy (dict, tmp_dict);
+
+        /* BUILD PEERS LIST */
+        INIT_LIST_HEAD (&conf->xaction_peers);
+        npeers = gd_build_peers_list  (&conf->peers, &conf->xaction_peers, op);
+
+        /* LOCKDOWN PHASE - Acquire mgmt_v3 locks */
+        ret = glusterd_mgmt_v3_initiate_lockdown (conf, op, dict, &op_errstr,
+                                                  npeers, &is_acquired);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "mgmt_v3 lockdown failed.");
+                goto out;
+        }
+
+        /* BUILD PAYLOAD */
+        ret = glusterd_mgmt_v3_build_payload (&req_dict, &op_errstr, dict, op);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD,
+                        gd_op_list[op]);
+                if (op_errstr == NULL)
+                        gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD);
+                goto out;
+        }
+
+        /* PRE-COMMIT VALIDATE PHASE */
+        ret = glusterd_mgmt_v3_pre_validate (conf, op, req_dict,
+                                             &op_errstr, npeers);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Pre Validation Failed");
+                goto out;
+        }
+
+        /* COMMIT OP PHASE */
+        ret = glusterd_mgmt_v3_commit (conf, op, dict, req_dict,
+                                       &op_errstr, npeers);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Commit Op Failed");
+                goto out;
+        }
+
+        /* POST-COMMIT VALIDATE PHASE */
+        /* As of now, post_validate is not handling any other
+           commands other than snapshot. So as of now, I am
+           sending 0 (op_ret as 0).
+        */
+        ret = glusterd_mgmt_v3_post_validate (conf, op, 0, dict, req_dict,
+                                              &op_errstr, npeers);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Post Validation Failed");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        op_ret = ret;
+        /* UNLOCK PHASE FOR PEERS*/
+        (void) glusterd_mgmt_v3_release_peer_locks (conf, op, dict,
+                                                    op_ret, &op_errstr,
+                                                    npeers, is_acquired);
+
+        /* LOCAL VOLUME(S) UNLOCK */
+        if (is_acquired) {
+                /* Trying to release multiple mgmt_v3 locks */
+                ret = glusterd_multiple_mgmt_v3_unlock (tmp_dict, MY_UUID);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to release mgmt_v3 locks on localhost");
+                        op_ret = ret;
+                }
+        }
+
+        /* SEND CLI RESPONSE */
+        glusterd_op_send_cli_response (op, op_ret, 0, req, dict, op_errstr);
+
+        if (req_dict)
+                dict_unref (req_dict);
+
+        if (tmp_dict)
+                dict_unref (tmp_dict);
+
+        if (op_errstr) {
+                GF_FREE (op_errstr);
+                op_errstr = NULL;
+        }
+
+        return 0;
+}
+
+int32_t
+glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op,
+                                       dict_t *dict)
+{
+        int32_t                     ret              = -1;
+        int32_t                     op_ret           = -1;
+        int32_t                     npeers           = 0;
+        dict_t                      *req_dict        = NULL;
+        dict_t                      *tmp_dict        = NULL;
+        glusterd_conf_t             *conf            = NULL;
+        char                        *op_errstr       = NULL;
+        xlator_t                    *this            = NULL;
+        gf_boolean_t                is_acquired      = _gf_false;
+        uuid_t                      *originator_uuid = NULL;
+        gf_boolean_t                success          = _gf_false;
+        char                        *tmp_errstr      = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (dict);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        /* Save the MY_UUID as the originator_uuid. This originator_uuid
+         * will be used by is_origin_glusterd() to determine if a node
+         * is the originator node for a command. */
+        originator_uuid = GF_CALLOC (1, sizeof(uuid_t),
+                                     gf_common_mt_uuid_t);
+        if (!originator_uuid) {
+                ret = -1;
+                goto out;
+        }
+
+        uuid_copy (*originator_uuid, MY_UUID);
+        ret = dict_set_bin (dict, "originator_uuid",
+                            originator_uuid, sizeof (uuid_t));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set originator_uuid.");
+                goto out;
+        }
+
+        /* Marking the operation as complete synctasked */
+        ret = dict_set_int32 (dict, "is_synctasked", _gf_true);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set synctasked flag.");
+                goto out;
+        }
+
+        /* Use a copy at local unlock as cli response will be sent before
+         * the unlock and the volname in the dict might be removed */
+        tmp_dict = dict_new();
+        if (!tmp_dict) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to create dict");
+                goto out;
+        }
+        dict_copy (dict, tmp_dict);
+
+        /* BUILD PEERS LIST */
+        INIT_LIST_HEAD (&conf->xaction_peers);
+        npeers = gd_build_peers_list  (&conf->peers, &conf->xaction_peers, op);
+
+        /* LOCKDOWN PHASE - Acquire mgmt_v3 locks */
+        ret = glusterd_mgmt_v3_initiate_lockdown (conf, op, dict, &op_errstr,
+                                                  npeers, &is_acquired);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "mgmt_v3 lockdown failed.");
+                goto out;
+        }
+
+        /* BUILD PAYLOAD */
+        ret = glusterd_mgmt_v3_build_payload (&req_dict, &op_errstr, dict, op);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD,
+                        gd_op_list[op]);
+                if (op_errstr == NULL)
+                        gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD);
+                goto out;
+        }
+
+        /* PRE-COMMIT VALIDATE PHASE */
+        ret = glusterd_mgmt_v3_pre_validate (conf, op, req_dict,
+                                            &op_errstr, npeers);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Pre Validation Failed");
+                goto out;
+        }
+
+        /* BRICK OP PHASE for initiating barrier*/
+        ret = dict_set_int32 (req_dict, "barrier", 1);
+        if (ret)
+                goto out;
+        ret = glusterd_mgmt_v3_brick_op (conf, op, req_dict,
+                                        &op_errstr, npeers);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Brick Ops Failed");
+                goto unbarrier;
+        }
+
+        /* COMMIT OP PHASE */
+        /* TODO: As of now, the plan is to do quorum check before sending the
+           commit fop and if the quorum succeeds, then commit is sent to all
+           the other glusterds.
+           snap create functionality now creates the in memory and on disk
+           objects for the snapshot (marking them as incomplete), takes the lvm
+           snapshot and then updates the status of the in memory and on disk
+           snap objects as complete. Suppose one of the glusterds goes down
+           after taking the lvm snapshot, but before updating the snap object,
+           then treat it as a snapshot create failure and trigger cleanup.
+           i.e the number of commit responses received by the originator
+           glusterd shold be the same as the number of peers it has sent the
+           request to (i.e npeers variable). If not, then originator glusterd
+           will initiate cleanup in post-validate fop.
+           Question: What if one of the other glusterds goes down as explained
+           above and along with it the originator glusterd also goes down?
+           Who will initiate the cleanup?
+        */
+        ret = glusterd_mgmt_v3_commit (conf, op, dict, req_dict,
+                                       &op_errstr, npeers);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Commit Op Failed");
+                /* If the main op fails, we should save the error string.
+                   Because, op_errstr will be used for unbarrier and
+                   unlock ops also. We might lose the actual error that
+                   caused the failure.
+                */
+                tmp_errstr = op_errstr;
+                op_errstr = NULL;
+                goto unbarrier;
+        }
+
+        success = _gf_true;
+unbarrier:
+        /* BRICK OP PHASE for removing the barrier*/
+        ret = dict_set_int32 (req_dict, "barrier", 0);
+        if (ret)
+                goto out;
+        ret = glusterd_mgmt_v3_brick_op (conf, op, req_dict,
+                                         &op_errstr, npeers);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Brick Ops Failed");
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        op_ret = ret;
+
+        if (success == _gf_false)
+                op_ret = -1;
+
+        /* POST-COMMIT VALIDATE PHASE */
+        ret = glusterd_mgmt_v3_post_validate (conf, op, op_ret, dict, req_dict,
+                                              &op_errstr, npeers);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Post Validation Failed");
+                op_ret = -1;
+        }
+
+        /* UNLOCK PHASE FOR PEERS*/
+        (void) glusterd_mgmt_v3_release_peer_locks (conf, op, dict,
+                                                    op_ret, &op_errstr,
+                                                    npeers, is_acquired);
+
+        /* If the commit op (snapshot taking) failed, then the error is stored
+           in tmp_errstr and unbarrier is called. Suppose, if unbarrier also
+           fails, then the error happened in unbarrier is logged and freed.
+           The error happened in commit op, which is stored in tmp_errstr
+           is sent to cli.
+        */
+        if (tmp_errstr) {
+                if (op_errstr) {
+                        gf_log (this->name, GF_LOG_ERROR, "unbarrier brick op"
+                                "failed with the error %s", op_errstr);
+                        GF_FREE (op_errstr);
+                        op_errstr = NULL;
+                }
+                op_errstr = tmp_errstr;
+        }
+
+        /* LOCAL VOLUME(S) UNLOCK */
+        if (is_acquired) {
+                /* Trying to release multiple mgmt_v3 locks */
+                ret = glusterd_multiple_mgmt_v3_unlock (tmp_dict, MY_UUID);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to release mgmt_v3 locks on localhost");
+                        op_ret = ret;
+                }
+        }
+
+        /* SEND CLI RESPONSE */
+        glusterd_op_send_cli_response (op, op_ret, 0, req, dict, op_errstr);
+
+        if (req_dict)
+                dict_unref (req_dict);
+
+        if (tmp_dict)
+                dict_unref (tmp_dict);
+
+        if (op_errstr) {
+                GF_FREE (op_errstr);
+                op_errstr = NULL;
+        }
+
+        return 0;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-mgmt.h b/xlators/mgmt/glusterd/src/glusterd-mgmt.h
new file mode 100644
index 000000000..b185a9bec
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-mgmt.h
@@ -0,0 +1,45 @@
+/*
+   Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _GLUSTERD_MGMT_H_
+#define _GLUSTERD_MGMT_H_
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+int32_t
+gd_mgmt_v3_pre_validate_fn (glusterd_op_t op, dict_t *dict,
+                           char **op_errstr, dict_t *rsp_dict);
+
+int32_t
+gd_mgmt_v3_brick_op_fn (glusterd_op_t op, dict_t *dict,
+                       char **op_errstr, dict_t *rsp_dict);
+
+int32_t
+gd_mgmt_v3_commit_fn (glusterd_op_t op, dict_t *dict,
+                     char **op_errstr, dict_t *rsp_dict);
+
+int32_t
+gd_mgmt_v3_post_validate_fn (glusterd_op_t op, int32_t op_ret, dict_t *dict,
+                            char **op_errstr, dict_t *rsp_dict);
+
+int32_t
+glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op,
+                                     dict_t *dict);
+
+int32_t
+glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op,
+                                      dict_t *dict);
+
+int
+glusterd_snap_pre_validate_use_rsp_dict (dict_t *dst, dict_t *src);
+
+#endif /* _GLUSTERD_MGMT_H_ */
diff --git a/xlators/mgmt/glusterd/src/glusterd-mountbroker.c b/xlators/mgmt/glusterd/src/glusterd-mountbroker.c
new file mode 100644
index 000000000..0d67d1303
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-mountbroker.c
@@ -0,0 +1,693 @@
+/*
+   Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+#include <inttypes.h>
+#include <fnmatch.h>
+#include <pwd.h>
+
+#include "globals.h"
+#include "glusterfs.h"
+#include "compat.h"
+#include "dict.h"
+#include "list.h"
+#include "logging.h"
+#include "defaults.h"
+#include "compat.h"
+#include "compat-errno.h"
+#include "run.h"
+#include "glusterd-mem-types.h"
+#include "glusterd.h"
+#include "glusterd-utils.h"
+#include "common-utils.h"
+#include "glusterd-mountbroker.h"
+#include "glusterd-op-sm.h"
+
+static int
+seq_dict_foreach (dict_t *dict,
+                  int (*fn)(char *str, void *data),
+                  void *data)
+{
+        char index[] = "4294967296"; // 1<<32
+        int        i = 0;
+        char    *val = NULL;
+        int      ret = 0;
+
+        for (;;i++) {
+                snprintf(index, sizeof(index), "%d", i);
+                ret = dict_get_str (dict, index, &val);
+                if (ret != 0)
+                        return ret == -ENOENT ? 0 : ret;
+                ret = fn (val, data);
+                if (ret != 0)
+                        return ret;
+        }
+}
+
+int
+parse_mount_pattern_desc (gf_mount_spec_t *mspec, char *pdesc)
+#define SYNTAX_ERR -2
+{
+        char *curs              = NULL;
+        char *c2                = NULL;
+        char sc                 = '\0';
+        char **cc               = NULL;
+        gf_mount_pattern_t *pat = NULL;
+        int pnum                = 0;
+        int ret                 = 0;
+        int lastsup             = -1;
+        int incl                = -1;
+        char **pcc              = NULL;
+        int pnc                 = 0;
+
+        skipwhite (&pdesc);
+
+        /* a bow to theory */
+        if (!*pdesc)
+                return 0;
+
+        /* count number of components, separated by '&' */
+        mspec->len = 0;
+        for (curs = pdesc; *curs; curs++) {
+                if (*curs == ')')
+                        mspec->len++;
+        }
+
+        mspec->patterns = GF_CALLOC (mspec->len, sizeof (*mspec->patterns),
+                                     gf_gld_mt_mount_pattern);
+        if (!mspec->patterns) {
+                ret = -1;
+                goto out;
+        }
+
+        pat = mspec->patterns;
+        curs = pdesc;
+        skipwhite (&curs);
+        for (;;) {
+                incl = -1;
+
+                /* check for pattern signedness modifier */
+                if (*curs == '-') {
+                        pat->negative = _gf_true;
+                        curs++;
+                }
+
+                /* now should come condition specifier,
+                 * then opening paren
+                 */
+                c2 = nwstrtail (curs, "SUB(");
+                if (c2) {
+                        pat->condition = SET_SUB;
+                        goto got_cond;
+                }
+                c2 = nwstrtail (curs, "SUP(");
+                if (c2) {
+                        pat->condition = SET_SUPER;
+                        lastsup = pat - mspec->patterns;
+                        goto got_cond;
+                }
+                c2 = nwstrtail (curs, "EQL(");
+                if (c2) {
+                        pat->condition = SET_EQUAL;
+                        goto got_cond;
+                }
+                c2 = nwstrtail (curs, "MEET(");
+                if (c2) {
+                        pat->condition = SET_INTERSECT;
+                        goto got_cond;
+                }
+                c2 = nwstrtail (curs, "SUB+(");
+                if (c2) {
+                        pat->condition = SET_SUB;
+                        incl = lastsup;
+                        goto got_cond;
+                }
+
+                ret = SYNTAX_ERR;
+                goto out;
+
+ got_cond:
+                curs = c2;
+                skipwhite (&curs);
+                /* count the number of components for pattern */
+                pnum = *curs == ')' ? 0 : 1;
+                for (c2 = curs ;*c2 != ')';) {
+                        if (strchr ("&|", *c2)) {
+                                ret = SYNTAX_ERR;
+                                goto out;
+                        }
+                        while (!strchr ("|&)", *c2) && !isspace (*c2))
+                                c2++;
+                        skipwhite (&c2);
+                        switch (*c2) {
+                        case ')':
+                                break;
+                        case '\0':
+                        case '&':
+                                ret = SYNTAX_ERR;
+                                goto out;
+                        case '|':
+                                *c2 = ' ';
+                                skipwhite (&c2);
+                                /* fall through */
+                        default:
+                                pnum++;
+                        }
+                }
+                if (incl >= 0) {
+                        pnc = 0;
+                        for (pcc = mspec->patterns[incl].components; *pcc; pcc++)
+                                pnc++;
+                        pnum += pnc;
+                }
+                pat->components = GF_CALLOC (pnum + 1, sizeof (*pat->components),
+                                             gf_gld_mt_mount_comp_container);
+                if (!pat->components) {
+                        ret = -1;
+                        goto out;
+                }
+
+                cc = pat->components;
+                /* copy over included component set */
+                if (incl >= 0) {
+                        memcpy (pat->components,
+                                mspec->patterns[incl].components,
+                                pnc * sizeof (*pat->components));
+                        cc += pnc;
+                }
+                /* parse and add components */
+                c2 = ""; /* reset c2 */
+                while (*c2 != ')') {
+                        c2 = curs;
+                        while (!isspace (*c2) && *c2 != ')')
+                                c2++;
+                        sc = *c2;
+                        *c2 = '\0';;
+                        *cc = gf_strdup (curs);
+                        if (!*cc) {
+                                ret = -1;
+                                goto out;
+                        }
+                        *c2 = sc;
+                        skipwhite (&c2);
+                        curs = c2;
+                        cc++;
+                }
+
+                curs++;
+                skipwhite (&curs);
+                if (*curs == '&') {
+                        curs++;
+                        skipwhite (&curs);
+                }
+
+                if (!*curs)
+                        break;
+                pat++;
+        }
+
+ out:
+        if (ret == SYNTAX_ERR) {
+                gf_log ("", GF_LOG_ERROR, "cannot parse mount patterns %s",
+                        pdesc);
+        }
+
+        /* We've allocted a lotta stuff here but don't bother with freeing
+         * on error, in that case we'll terminate anyway
+         */
+        return ret ? -1 : 0;
+}
+#undef SYNTAX_ERR
+
+
+const char *georep_mnt_desc_template =
+        "SUP("
+                "xlator-option=\\*-dht.assert-no-child-down=true "
+                "volfile-server=localhost "
+                "client-pid=%d "
+                "user-map-root=%s "
+        ")"
+        "SUB+("
+                "log-file="DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"*/* "
+                "log-level=* "
+                "volfile-id=* "
+        ")"
+        "MEET("
+                "%s"
+        ")";
+
+const char *hadoop_mnt_desc_template =
+        "SUP("
+                "volfile-server=%s "
+                "client-pid=%d "
+                "volfile-id=%s "
+                "user-map-root=%s "
+        ")"
+        "SUB+("
+                "log-file="DEFAULT_LOG_FILE_DIRECTORY"/"GHADOOP"*/* "
+                "log-level=* "
+        ")";
+
+int
+make_georep_mountspec (gf_mount_spec_t *mspec, const char *volnames,
+                       char *user)
+{
+        char *georep_mnt_desc = NULL;
+        char *meetspec        = NULL;
+        char *vols            = NULL;
+        char *vol             = NULL;
+        char *p               = NULL;
+        char *savetok         = NULL;
+        char *fa[3]           = {0,};
+        size_t siz            = 0;
+        int vc                = 0;
+        int i                 = 0;
+        int ret               = 0;
+
+        vols = gf_strdup ((char *)volnames);
+        if (!vols)
+                goto out;
+
+        for (vc = 1, p = vols; *p; p++) {
+                if (*p == ',')
+                        vc++;
+        }
+        siz = strlen (volnames) + vc * strlen("volfile-id=");
+        meetspec = GF_CALLOC (1, siz + 1, gf_gld_mt_georep_meet_spec);
+        if (!meetspec)
+                goto out;
+
+        for (p = vols;;) {
+                vol = strtok_r (p, ",", &savetok);
+                if (!vol) {
+                        GF_ASSERT (vc == 0);
+                        break;
+                }
+                p = NULL;
+                strcat (meetspec, "volfile-id=");
+                strcat (meetspec, vol);
+                if (--vc > 0)
+                        strcat (meetspec, " ");
+        }
+
+        ret = gf_asprintf (&georep_mnt_desc, georep_mnt_desc_template,
+                           GF_CLIENT_PID_GSYNCD, user, meetspec);
+        if (ret == -1) {
+                georep_mnt_desc = NULL;
+                goto out;
+        }
+
+        ret = parse_mount_pattern_desc (mspec, georep_mnt_desc);
+
+ out:
+        fa[0] = meetspec;
+        fa[1] = vols;
+        fa[2] = georep_mnt_desc;
+
+        for (i = 0; i < 3; i++) {
+                if (fa[i] == NULL)
+                        ret = -1;
+                else
+                        GF_FREE (fa[i]);
+        }
+
+        return ret;
+}
+
+int
+make_ghadoop_mountspec (gf_mount_spec_t *mspec, const char *volname,
+                        char *user, char *server)
+{
+        char *hadoop_mnt_desc = NULL;
+        int   ret             = 0;
+
+        ret = gf_asprintf (&hadoop_mnt_desc, hadoop_mnt_desc_template,
+                           server, GF_CLIENT_PID_HADOOP, volname, user);
+        if (ret == -1)
+                return ret;
+
+        return parse_mount_pattern_desc (mspec, hadoop_mnt_desc);
+}
+
+static gf_boolean_t
+match_comp (char *str, char *patcomp)
+{
+        char *c1 = patcomp;
+        char *c2 = str;
+
+        GF_ASSERT (c1);
+        GF_ASSERT (c2);
+
+        while (*c1 == *c2) {
+                if (!*c1)
+                        return _gf_true;
+                c1++;
+                c2++;
+                if (c1[-1] == '=')
+                        break;
+        }
+
+        return fnmatch (c1, c2, 0) == 0 ? _gf_true : _gf_false;
+}
+
+struct gf_set_descriptor {
+        gf_boolean_t priv[2];
+        gf_boolean_t common;
+};
+
+static int
+_gf_set_dict_iter1 (char *val, void *data)
+{
+        void **dataa                 = data;
+        struct gf_set_descriptor *sd = dataa[0];
+        char **curs                  = dataa[1];
+        gf_boolean_t priv            = _gf_true;
+
+        while (*curs) {
+                if (match_comp (val, *curs)) {
+                        priv = _gf_false;
+                        sd->common = _gf_true;
+                }
+                curs++;
+        }
+
+        if (priv)
+                sd->priv[0] = _gf_true;
+
+        return 0;
+}
+
+static int
+_gf_set_dict_iter2 (char *val, void *data)
+{
+        void **dataa      = data;
+        gf_boolean_t *boo = dataa[0];
+        char *comp        = dataa[1];
+
+        if (match_comp (val, comp))
+                *boo = _gf_true;
+
+        return 0;
+}
+
+static void
+relate_sets (struct gf_set_descriptor *sd, dict_t *argdict, char **complist)
+{
+        void *dataa[] = {NULL, NULL};
+        gf_boolean_t boo = _gf_false;
+
+        memset (sd, 0, sizeof (*sd));
+
+        dataa[0] = sd;
+        dataa[1] = complist;
+        seq_dict_foreach (argdict, _gf_set_dict_iter1, dataa);
+
+        while (*complist) {
+                boo = _gf_false;
+                dataa[0] = &boo;
+                dataa[1] = *complist;
+                seq_dict_foreach (argdict, _gf_set_dict_iter2, dataa);
+
+                if (boo)
+                        sd->common = _gf_true;
+                else
+                        sd->priv[1] = _gf_true;
+
+                complist++;
+        }
+}
+
+static int
+_arg_parse_uid (char *val, void *data)
+{
+        char *user        = strtail (val, "user-map-root=");
+        struct passwd *pw = NULL;
+
+        if (!user)
+                return 0;
+        pw = getpwnam (user);
+        if (!pw)
+                return -EINVAL;
+
+        if (*(int *)data >= 0)
+                /* uid ambiguity, already found */
+                return -EINVAL;
+
+        *(int *)data = pw->pw_uid;
+        return 0;
+}
+
+static int
+evaluate_mount_request (gf_mount_spec_t *mspec, dict_t *argdict)
+{
+        struct gf_set_descriptor sd = {{0,},};
+        int i                       = 0;
+        int uid                     = -1;
+        int ret                     = 0;
+        gf_boolean_t match          = _gf_false;
+
+        for (i = 0; i < mspec->len; i++) {
+                relate_sets (&sd, argdict, mspec->patterns[i].components);
+                switch (mspec->patterns[i].condition) {
+                case SET_SUB:
+                        match = !sd.priv[0];
+                        break;
+                case SET_SUPER:
+                        match = !sd.priv[1];
+                        break;
+                case SET_EQUAL:
+                        match = (!sd.priv[0] && !sd.priv[1]);
+                        break;
+                case SET_INTERSECT:
+                        match = sd.common;
+                        break;
+                default:
+                        GF_ASSERT(!"unreached");
+                }
+                if (mspec->patterns[i].negative)
+                        match = !match;
+
+                if (!match)
+                        return -EPERM;
+        }
+
+        ret = seq_dict_foreach (argdict, _arg_parse_uid, &uid);
+        if (ret != 0)
+                return ret;
+
+        return uid;
+}
+
+static int
+_volname_get (char *val, void *data)
+{
+        char **volname = data;
+
+        *volname = strtail (val, "volfile-id=");
+
+        return *volname ? 1 : 0;
+}
+
+static int
+_runner_add (char *val, void *data)
+{
+        runner_t *runner = data;
+
+        runner_argprintf (runner, "--%s", val);
+
+        return 0;
+}
+
+int
+glusterd_do_mount (char *label, dict_t *argdict, char **path, int *op_errno)
+{
+        glusterd_conf_t *priv      = NULL;
+        char *mountbroker_root     = NULL;
+        gf_mount_spec_t *mspec     = NULL;
+        int uid                    = -ENOENT;
+        char *volname              = NULL;
+        glusterd_volinfo_t *vol    = NULL;
+        char *mtptemp              = NULL;
+        char *mntlink              = NULL;
+        char *cookieswitch         = NULL;
+        char *cookie               = NULL;
+        char *sla                  = NULL;
+        struct stat st             = {0,};
+        runner_t runner            = {0,};
+        int ret                    = 0;
+        xlator_t *this             = THIS;
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        GF_ASSERT (op_errno);
+        *op_errno = 0;
+
+        if (dict_get_str (this->options, "mountbroker-root",
+                          &mountbroker_root) != 0) {
+                *op_errno = ENOENT;
+                goto out;
+        }
+
+        GF_ASSERT (label);
+        if (!*label) {
+                *op_errno = EINVAL;
+                goto out;
+        }
+
+        /* look up spec for label */
+        list_for_each_entry (mspec, &priv->mount_specs,
+                             speclist) {
+                if (strcmp (mspec->label, label) != 0)
+                        continue;
+                uid = evaluate_mount_request (mspec, argdict);
+                break;
+        }
+        if (uid < 0) {
+                *op_errno = -uid;
+                goto out;
+        }
+
+        /* some sanity check on arguments */
+        seq_dict_foreach (argdict, _volname_get, &volname);
+        if (!volname) {
+                *op_errno = EINVAL;
+                goto out;
+        }
+        if (glusterd_volinfo_find (volname, &vol) != 0 ||
+            !glusterd_is_volume_started (vol)) {
+                *op_errno = ENOENT;
+                goto out;
+        }
+
+        /* go do mount */
+
+        /** create actual mount dir */
+
+        /*** "overload" string name to be possible to used for cookie
+             creation, see below */
+        ret = gf_asprintf (&mtptemp, "%s/user%d/mtpt-%s-XXXXXX/cookie",
+                           mountbroker_root, uid, label);
+        if (ret == -1) {
+                mtptemp = NULL;
+                *op_errno = ENOMEM;
+                goto out;
+        }
+        /*** hide cookie part */
+        cookieswitch = strrchr (mtptemp, '/');
+        *cookieswitch = '\0';
+
+        sla = strrchr (mtptemp, '/');
+        *sla = '\0';
+        ret = mkdir (mtptemp, 0700);
+        if (ret == 0)
+                ret = chown (mtptemp, uid, 0);
+        else if (errno == EEXIST)
+                ret = 0;
+        if (ret == -1) {
+                *op_errno = errno;
+                goto out;
+        }
+        ret = lstat (mtptemp, &st);
+        if (ret == -1) {
+                *op_errno = errno;
+                goto out;
+        }
+        if (!(S_ISDIR (st.st_mode) && (st.st_mode & ~S_IFMT) == 0700 &&
+              st.st_uid == uid && st.st_gid == 0)) {
+                *op_errno = EACCES;
+                goto out;
+        }
+        *sla = '/';
+
+        if (!mkdtemp (mtptemp)) {
+                *op_errno = errno;
+                goto out;
+        }
+
+        /** create private "cookie" symlink */
+
+        /*** occupy an entry in the hive dir via mkstemp */
+        ret = gf_asprintf (&cookie, "%s/"MB_HIVE"/mntXXXXXX",
+                           mountbroker_root);
+        if (ret == -1) {
+                cookie = NULL;
+                *op_errno = ENOMEM;
+                goto out;
+        }
+        ret = mkstemp (cookie);
+        if (ret == -1) {
+                *op_errno = errno;
+                goto out;
+        }
+        close (ret);
+
+        /*** assembly the path from cookie to mountpoint */
+        sla = strchr (sla - 1, '/');
+        GF_ASSERT (sla);
+        ret = gf_asprintf (&mntlink, "../user%d%s", uid, sla);
+        if (ret == -1) {
+                *op_errno = ENOMEM;
+                goto out;
+        }
+
+        /*** create cookie link in (to-be) mountpoint,
+             move it over to the final place */
+        *cookieswitch = '/';
+        ret = symlink (mntlink, mtptemp);
+        if (ret != -1)
+                ret = rename (mtptemp, cookie);
+        *cookieswitch = '\0';
+        if (ret == -1) {
+                *op_errno = errno;
+                goto out;
+        }
+
+        /** invoke glusterfs on the mountpoint */
+
+        runinit (&runner);
+        runner_add_arg (&runner, SBIN_DIR"/glusterfs");
+        seq_dict_foreach (argdict, _runner_add, &runner);
+        runner_add_arg (&runner, mtptemp);
+        ret = runner_run_reuse (&runner);
+        if (ret == -1) {
+                *op_errno = EIO; /* XXX hacky fake */
+                runner_log (&runner, "", GF_LOG_ERROR, "command failed");
+        }
+        runner_end (&runner);
+
+ out:
+
+        if (*op_errno) {
+                ret = -1;
+                gf_log ("", GF_LOG_WARNING, "unsuccessful mount request (%s)",
+                        strerror (*op_errno));
+                if (mtptemp) {
+                        *cookieswitch = '/';
+                        unlink (mtptemp);
+                        *cookieswitch = '\0';
+                        rmdir (mtptemp);
+                }
+                if (cookie) {
+                        unlink (cookie);
+                        GF_FREE (cookie);
+                }
+
+        } else {
+                ret = 0;
+                *path = cookie;
+        }
+
+        GF_FREE (mtptemp);
+
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-mountbroker.h b/xlators/mgmt/glusterd/src/glusterd-mountbroker.h
new file mode 100644
index 000000000..426252ebe
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-mountbroker.h
@@ -0,0 +1,42 @@
+/*
+   Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#define MB_HIVE "mb_hive"
+
+typedef enum {
+        SET_SUB = 1,
+        SET_SUPER,
+        SET_EQUAL,
+        SET_INTERSECT
+} gf_setrel_t;
+
+struct gf_mount_pattern {
+        char **components;
+        gf_setrel_t condition;
+        gf_boolean_t negative;
+};
+typedef struct gf_mount_pattern gf_mount_pattern_t;
+
+struct gf_mount_spec {
+        struct list_head speclist;
+        char *label;
+        gf_mount_pattern_t *patterns;
+        size_t len;
+};
+typedef struct gf_mount_spec gf_mount_spec_t;
+
+
+int parse_mount_pattern_desc (gf_mount_spec_t *mspec, char *pdesc);
+
+int make_georep_mountspec (gf_mount_spec_t *mspec, const char *volname,
+                           char *user);
+int make_ghadoop_mountspec (gf_mount_spec_t *mspec, const char *volname,
+                            char *user, char *server);
+
+int glusterd_do_mount (char *label, dict_t *argdict, char **path, int *op_errno);
diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
index b8d2778e3..1666f5e4d 100644
--- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c
+++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is GF_FREE software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
@@ -25,6 +15,7 @@
 #include <time.h>
 #include <sys/uio.h>
 #include <sys/resource.h>
+#include <sys/mount.h>
 
 #include <libgen.h>
 #include "uuid.h"
@@ -44,43 +35,259 @@
 #include "glusterd-op-sm.h"
 #include "glusterd-utils.h"
 #include "glusterd-store.h"
-#include "cli1.h"
+#include "glusterd-hooks.h"
 #include "glusterd-volgen.h"
+#include "glusterd-locks.h"
+#include "syscall.h"
+#include "cli1-xdr.h"
+#include "common-utils.h"
+#include "run.h"
 
 #include <sys/types.h>
 #include <signal.h>
+#include <sys/wait.h>
+
+#define ALL_VOLUME_OPTION_CHECK(volname, key, ret, op_errstr, label)           \
+        do {                                                                   \
+                gf_boolean_t    _all = !strcmp ("all", volname);               \
+                gf_boolean_t    _ratio = !strcmp (key,                         \
+                                                  GLUSTERD_QUORUM_RATIO_KEY);  \
+                if (_all && !_ratio) {                                         \
+                        ret = -1;                                              \
+                        *op_errstr = gf_strdup ("Not a valid option for all "  \
+                                                "volumes");                    \
+                        goto label;                                            \
+                } else if (!_all && _ratio) {                                  \
+                        ret = -1;                                              \
+                        *op_errstr = gf_strdup ("Not a valid option for "      \
+                                                "single volume");              \
+                        goto label;                                            \
+                }                                                              \
+         } while (0)
 
 static struct list_head gd_op_sm_queue;
+pthread_mutex_t       gd_op_sm_lock;
 glusterd_op_info_t    opinfo = {{0},};
-static int glusterfs_port = GLUSTERD_DEFAULT_PORT;
+uuid_t                global_txn_id = {"\0"}; /* To be used in
+                                               * heterogeneous
+                                               * cluster with no
+                                               * transaction ids */
+
+static dict_t *txn_opinfo;
+
+struct txn_opinfo_object_ {
+        glusterd_op_info_t    opinfo;
+};
+typedef struct txn_opinfo_object_ txn_opinfo_obj;
+
+int32_t
+glusterd_txn_opinfo_dict_init ()
+{
+        int32_t ret = -1;
+
+        txn_opinfo = dict_new ();
+        if (!txn_opinfo) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
 
 void
-glusterd_destroy_lock_ctx (glusterd_op_lock_ctx_t *ctx)
+glusterd_txn_opinfo_dict_fini ()
 {
-        if (!ctx)
-                return;
-        GF_FREE (ctx);
+        if (txn_opinfo)
+                dict_destroy (txn_opinfo);
 }
 
 void
-glusterd_destroy_stage_ctx (glusterd_op_stage_ctx_t *ctx)
+glusterd_txn_opinfo_init (glusterd_op_info_t  *opinfo,
+                          glusterd_op_sm_state_info_t *state,
+                          glusterd_op_t *op,
+                          dict_t *op_ctx,
+                          rpcsvc_request_t *req)
 {
-        if (!ctx)
-                return;
+        GF_ASSERT (opinfo);
 
-        if (ctx->stage_req.buf.buf_val)
-                GF_FREE (ctx->stage_req.buf.buf_val);
-        GF_FREE (ctx);
+        if (state)
+                opinfo->state = *state;
+
+        if (op)
+                opinfo->op = *op;
+
+        opinfo->op_ctx = dict_ref(op_ctx);
+
+        if (req)
+                opinfo->req = req;
+
+        return;
+}
+
+int32_t
+glusterd_get_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t  *opinfo)
+{
+        int32_t          ret         = -1;
+        txn_opinfo_obj  *opinfo_obj  = NULL;
+
+        if (!txn_id || !opinfo) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Empty transaction id or opinfo received.");
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_bin(txn_opinfo, uuid_utoa (*txn_id),
+                           (void **) &opinfo_obj);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to get transaction opinfo");
+                goto out;
+        }
+
+        (*opinfo) = opinfo_obj->opinfo;
+
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_set_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t  *opinfo)
+{
+        int32_t          ret        = -1;
+        txn_opinfo_obj  *opinfo_obj = NULL;
+
+        if (!txn_id) {
+                gf_log ("", GF_LOG_ERROR, "Empty transaction id received.");
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_bin(txn_opinfo, uuid_utoa (*txn_id),
+                           (void **) &opinfo_obj);
+        if (ret) {
+                opinfo_obj = GF_CALLOC (1, sizeof(txn_opinfo_obj),
+                                        gf_common_mt_txn_opinfo_obj_t);
+                if (!opinfo_obj) {
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = dict_set_bin(txn_opinfo, uuid_utoa (*txn_id), opinfo_obj,
+                                   sizeof(txn_opinfo_obj));
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Unable to set opinfo for transaction ID : %s",
+                                uuid_utoa (*txn_id));
+                        goto out;
+                }
+        }
+
+        opinfo_obj->opinfo = (*opinfo);
+
+        ret = 0;
+out:
+        if (ret)
+                if (opinfo_obj)
+                        GF_FREE (opinfo_obj);
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_clear_txn_opinfo (uuid_t *txn_id)
+{
+        int32_t               ret         = -1;
+        glusterd_op_info_t    txn_op_info = {{0},};
+
+        if (!txn_id) {
+                gf_log ("", GF_LOG_ERROR, "Empty transaction id received.");
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_get_txn_opinfo (txn_id, &txn_op_info);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Transaction opinfo not found");
+                goto out;
+        }
+
+        dict_unref (txn_op_info.op_ctx);
+
+        dict_del(txn_opinfo, uuid_utoa (*txn_id));
+
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+static int glusterfs_port = GLUSTERD_DEFAULT_PORT;
+static char *glusterd_op_sm_state_names[] = {
+        "Default",
+        "Lock sent",
+        "Locked",
+        "Stage op sent",
+        "Staged",
+        "Commit op sent",
+        "Committed",
+        "Unlock sent",
+        "Stage op failed",
+        "Commit op failed",
+        "Brick op sent",
+        "Brick op failed",
+        "Brick op Committed",
+        "Brick op Commit failed",
+        "Ack drain",
+        "Invalid",
+};
+
+static char *glusterd_op_sm_event_names[] = {
+        "GD_OP_EVENT_NONE",
+        "GD_OP_EVENT_START_LOCK",
+        "GD_OP_EVENT_LOCK",
+        "GD_OP_EVENT_RCVD_ACC",
+        "GD_OP_EVENT_ALL_ACC",
+        "GD_OP_EVENT_STAGE_ACC",
+        "GD_OP_EVENT_COMMIT_ACC",
+        "GD_OP_EVENT_RCVD_RJT",
+        "GD_OP_EVENT_STAGE_OP",
+        "GD_OP_EVENT_COMMIT_OP",
+        "GD_OP_EVENT_UNLOCK",
+        "GD_OP_EVENT_START_UNLOCK",
+        "GD_OP_EVENT_ALL_ACK",
+        "GD_OP_EVENT_LOCAL_UNLOCK_NO_RESP",
+        "GD_OP_EVENT_INVALID"
+};
+
+extern struct volopt_map_entry glusterd_volopt_map[];
+
+char*
+glusterd_op_sm_state_name_get (int state)
+{
+        if (state < 0 || state >= GD_OP_STATE_MAX)
+                return glusterd_op_sm_state_names[GD_OP_STATE_MAX];
+        return glusterd_op_sm_state_names[state];
+}
+
+char*
+glusterd_op_sm_event_name_get (int event)
+{
+        if (event < 0 || event >= GD_OP_EVENT_MAX)
+                return glusterd_op_sm_event_names[GD_OP_EVENT_MAX];
+        return glusterd_op_sm_event_names[event];
 }
 
 void
-glusterd_destroy_commit_ctx (glusterd_op_commit_ctx_t *ctx)
+glusterd_destroy_lock_ctx (glusterd_op_lock_ctx_t *ctx)
 {
         if (!ctx)
                 return;
-
-        if (ctx->stage_req.buf.buf_val)
-                GF_FREE (ctx->stage_req.buf.buf_val);
         GF_FREE (ctx);
 }
 
@@ -92,1145 +299,1619 @@ glusterd_set_volume_status (glusterd_volinfo_t  *volinfo,
         volinfo->status = status;
 }
 
-static int
+gf_boolean_t
 glusterd_is_volume_started (glusterd_volinfo_t  *volinfo)
 {
         GF_ASSERT (volinfo);
-        return (!(volinfo->status == GLUSTERD_STATUS_STARTED));
+        return (volinfo->status == GLUSTERD_STATUS_STARTED);
 }
 
-gf_boolean_t
-glusterd_are_all_volumes_stopped ()
+static int
+glusterd_op_sm_inject_all_acc (uuid_t *txn_id)
 {
-        glusterd_conf_t                         *priv = NULL;
-        xlator_t                                *this = NULL;
-        glusterd_volinfo_t                      *voliter = NULL;
+        int32_t                 ret = -1;
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, txn_id, NULL);
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_brick_op_build_payload (glusterd_op_t op, glusterd_brickinfo_t *brickinfo,
+                                 gd1_mgmt_brick_op_req **req, dict_t *dict)
+{
+        int                     ret = -1;
+        gd1_mgmt_brick_op_req   *brick_req = NULL;
+        char                    *volname = NULL;
+        char                    name[1024] = {0,};
+        gf_xl_afr_op_t          heal_op = GF_AFR_OP_INVALID;
+        xlator_t                *this = NULL;
 
         this = THIS;
         GF_ASSERT (this);
-        priv = this->private;
-        GF_ASSERT (priv);
-
-        list_for_each_entry (voliter, &priv->volumes, vol_list) {
-                if (voliter->status == GLUSTERD_STATUS_STARTED)
-                        return _gf_false;
-        }
-
-        return _gf_true;
-
-}
 
-static int
-glusterd_op_get_len (glusterd_op_t op)
-{
         GF_ASSERT (op < GD_OP_MAX);
         GF_ASSERT (op > GD_OP_NONE);
-        int             ret = -1;
+        GF_ASSERT (req);
+
 
         switch (op) {
-                case GD_OP_CREATE_VOLUME:
-                        {
-                                dict_t *dict = glusterd_op_get_ctx (op);
-                                ret = dict_serialized_length (dict);
-                                return ret;
-                        }
-                        break;
+        case GD_OP_REMOVE_BRICK:
+        case GD_OP_STOP_VOLUME:
+                brick_req = GF_CALLOC (1, sizeof (*brick_req),
+                                       gf_gld_mt_mop_brick_req_t);
+                if (!brick_req)
+                        goto out;
+                brick_req->op = GLUSTERD_BRICK_TERMINATE;
+                brick_req->name = "";
+        break;
+        case GD_OP_PROFILE_VOLUME:
+                brick_req = GF_CALLOC (1, sizeof (*brick_req),
+                                       gf_gld_mt_mop_brick_req_t);
+
+                if (!brick_req)
+                        goto out;
 
-                case GD_OP_START_BRICK:
-                        break;
+                brick_req->op = GLUSTERD_BRICK_XLATOR_INFO;
+                brick_req->name = brickinfo->path;
 
-                case GD_OP_SET_VOLUME:
-                case GD_OP_REPLACE_BRICK:
-                case GD_OP_ADD_BRICK:
-                        {
-                                dict_t *dict = glusterd_op_get_ctx (op);
-                                ret = dict_serialized_length (dict);
-                                return ret;
-                        }
-                 case GD_OP_REMOVE_BRICK:
-                        {
-                                dict_t *dict = glusterd_op_get_ctx (op);
-                                ret = dict_serialized_length (dict);
-                                return ret;
-                        }
-                        break;
-                       break;
+                break;
+        case GD_OP_HEAL_VOLUME:
+        {
+                brick_req = GF_CALLOC (1, sizeof (*brick_req),
+                                       gf_gld_mt_mop_brick_req_t);
+                if (!brick_req)
+                        goto out;
 
-                default:
-                        GF_ASSERT (op);
+                brick_req->op = GLUSTERD_BRICK_XLATOR_OP;
+                brick_req->name = "";
+                ret = dict_get_int32 (dict, "heal-op", (int32_t*)&heal_op);
+                if (ret)
+                        goto out;
+                ret = dict_set_int32 (dict, "xl-op", heal_op);
+        }
+                break;
+        case GD_OP_STATUS_VOLUME:
+        {
+                brick_req = GF_CALLOC (1, sizeof (*brick_req),
+                                       gf_gld_mt_mop_brick_req_t);
+                if (!brick_req)
+                        goto out;
+                brick_req->op = GLUSTERD_BRICK_STATUS;
+                brick_req->name = "";
+        }
+                break;
+        case GD_OP_REBALANCE:
+        case GD_OP_DEFRAG_BRICK_VOLUME:
+                brick_req = GF_CALLOC (1, sizeof (*brick_req),
+                                       gf_gld_mt_mop_brick_req_t);
+                if (!brick_req)
+                        goto out;
 
+                brick_req->op = GLUSTERD_BRICK_XLATOR_DEFRAG;
+                ret = dict_get_str (dict, "volname", &volname);
+                if (ret)
+                        goto out;
+                snprintf (name, 1024, "%s-dht",volname);
+                brick_req->name = gf_strdup (name);
+
+                break;
+        case GD_OP_SNAP:
+                brick_req = GF_CALLOC (1, sizeof (*brick_req),
+                                       gf_gld_mt_mop_brick_req_t);
+                if (!brick_req)
+                        goto out;
+
+                brick_req->op = GLUSTERD_VOLUME_BARRIER_OP;
+                ret = dict_get_str (dict, "volname", &volname);
+                if (ret)
+                        goto out;
+                snprintf (name, 1024, "%s-server",volname);
+                brick_req->name = gf_strdup (name);
+
+                break;
+        default:
+                goto out;
+        break;
         }
 
-        return 0;
-}
+        ret = dict_allocate_and_serialize (dict, &brick_req->input.input_val,
+                                           &brick_req->input.input_len);
+        if (ret)
+                goto out;
+        *req = brick_req;
+        ret = 0;
 
-static int
-glusterd_op_sm_inject_all_acc ()
-{
-        int32_t                 ret = -1;
-        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL);
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+out:
+        if (ret && brick_req)
+                GF_FREE (brick_req);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
 int
-glusterd_op_build_payload (glusterd_op_t op, gd1_mgmt_stage_op_req **req)
+glusterd_node_op_build_payload (glusterd_op_t op, gd1_mgmt_brick_op_req **req,
+                                dict_t *dict)
 {
-        int                     len = 0;
         int                     ret = -1;
-        gd1_mgmt_stage_op_req   *stage_req = NULL;
-        void                    *ctx = NULL;
+        gd1_mgmt_brick_op_req   *brick_req = NULL;
 
         GF_ASSERT (op < GD_OP_MAX);
         GF_ASSERT (op > GD_OP_NONE);
         GF_ASSERT (req);
 
-        len = glusterd_op_get_len (op);
+        switch (op) {
+        case GD_OP_PROFILE_VOLUME:
+                brick_req = GF_CALLOC (1, sizeof (*brick_req),
+                                       gf_gld_mt_mop_brick_req_t);
+                if (!brick_req)
+                        goto out;
 
-        stage_req = GF_CALLOC (1, sizeof (*stage_req),
-                               gf_gld_mt_mop_stage_req_t);
+                brick_req->op = GLUSTERD_NODE_PROFILE;
+                brick_req->name = "";
 
-        if (!stage_req) {
-                gf_log ("", GF_LOG_ERROR, "Out of Memory");
-                goto out;
-        }
+                break;
+
+        case GD_OP_STATUS_VOLUME:
+                brick_req = GF_CALLOC (1, sizeof (*brick_req),
+                                       gf_gld_mt_mop_brick_req_t);
+                if (!brick_req)
+                        goto out;
 
+                brick_req->op = GLUSTERD_NODE_STATUS;
+                brick_req->name = "";
 
-        glusterd_get_uuid (&stage_req->uuid);
-        stage_req->op = op;
-        //stage_req->buf.buf_len = len;
+                break;
 
-        ctx = (void*)glusterd_op_get_ctx (op);
-        if (!ctx) {
-                gf_log ("", GF_LOG_ERROR, "Null Context for "
-                        "op %d", op);
-                ret = -1;
+        default:
                 goto out;
         }
 
-        switch (op) {
-                case GD_OP_CREATE_VOLUME:
-                        {
-                                dict_t  *dict = ctx;
-                                ++glusterfs_port;
-                                ret = dict_set_int32 (dict, "port", glusterfs_port);
-                                ret = dict_allocate_and_serialize (dict,
-                                                &stage_req->buf.buf_val,
-                                        (size_t *)&stage_req->buf.buf_len);
-                                if (ret) {
-                                        goto out;
-                                }
-                        }
-                        break;
-
-                case GD_OP_START_VOLUME:
-                        {
-                                glusterd_op_start_volume_ctx_t *ctx1 = ctx;
-                                stage_req->buf.buf_len  =
-                                        strlen (ctx1->volume_name);
-                                stage_req->buf.buf_val =
-                                        gf_strdup (ctx1->volume_name);
-                        }
-                        break;
-
-                case GD_OP_DELETE_VOLUME:
-                        {
-                                glusterd_op_delete_volume_ctx_t *ctx1 = ctx;
-                                stage_req->buf.buf_len  =
-                                        strlen (ctx1->volume_name);
-                                stage_req->buf.buf_val =
-                                        gf_strdup (ctx1->volume_name);
-                        }
-                        break;
-
-                case GD_OP_STOP_VOLUME:
-                case GD_OP_ADD_BRICK:
-                case GD_OP_REPLACE_BRICK:
-                case GD_OP_SET_VOLUME:
-                case GD_OP_REMOVE_BRICK:
-                case GD_OP_LOG_FILENAME:
-                case GD_OP_LOG_ROTATE:
-                case GD_OP_SYNC_VOLUME:
-                        {
-                                dict_t  *dict = ctx;
-                                ret = dict_allocate_and_serialize (dict,
-                                                &stage_req->buf.buf_val,
-                                        (size_t *)&stage_req->buf.buf_len);
-                                if (ret) {
-                                        goto out;
-                                }
-                        }
-                        break;
+        ret = dict_allocate_and_serialize (dict, &brick_req->input.input_val,
+                                           &brick_req->input.input_len);
 
-                default:
-                        break;
-        }
+        if (ret)
+                goto out;
 
-        *req = stage_req;
+        *req = brick_req;
         ret = 0;
 
 out:
+        if (ret && brick_req)
+                GF_FREE (brick_req);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
 static int
-glusterd_op_stage_create_volume (gd1_mgmt_stage_op_req *req, char **op_errstr)
+glusterd_validate_quorum_options (xlator_t *this, char *fullkey, char *value,
+                                  char **op_errstr)
 {
-        int                                     ret = 0;
-        dict_t                                  *dict = NULL;
-        char                                    *volname = NULL;
-        gf_boolean_t                            exists = _gf_false;
-        char                                    *bricks = NULL;
-        char                                    *brick_list = NULL;
-        char                                    *free_ptr = NULL;
-        glusterd_brickinfo_t                    *brick_info = NULL;
-        int32_t                                 brick_count = 0;
-        int32_t                                 i = 0;
-        struct stat                             st_buf = {0,};
-        char                                    *brick = NULL;
-        char                                    *tmpptr = NULL;
-        char                                    cmd_str[1024];
-        xlator_t                                *this = NULL;
-        glusterd_conf_t                         *priv = NULL;
-        char                                    msg[2048] = {0,};
+        int             ret = 0;
+        char            *key = NULL;
+        volume_option_t *opt = NULL;
 
-        GF_ASSERT (req);
+        if (!glusterd_is_quorum_option (fullkey))
+                goto out;
+        key = strchr (fullkey, '.');
+        key++;
+        opt = xlator_volume_option_get (this, key);
+        ret = xlator_option_validate (this, key, value, opt, op_errstr);
+out:
+        return ret;
+}
+
+static int
+glusterd_check_client_op_version_support (char *volname, uint32_t op_version,
+                                          char **op_errstr)
+{
+        int                     ret = 0;
+        xlator_t                *this = NULL;
+        glusterd_conf_t         *priv = NULL;
+        rpc_transport_t         *xprt = NULL;
 
         this = THIS;
-        if (!this) {
-                gf_log ("glusterd", GF_LOG_ERROR,
-                        "this is NULL");
-                goto out;
+        GF_ASSERT(this);
+        priv = this->private;
+        GF_ASSERT(priv);
+
+        pthread_mutex_lock (&priv->xprt_lock);
+        list_for_each_entry (xprt, &priv->xprt_list, list) {
+                if ((!strcmp(volname, xprt->peerinfo.volname)) &&
+                    ((op_version > xprt->peerinfo.max_op_version) ||
+                     (op_version < xprt->peerinfo.min_op_version))) {
+                        ret = -1;
+                        break;
+                }
         }
+        pthread_mutex_unlock (&priv->xprt_lock);
 
-        priv = this->private;
-        if (!priv) {
-                gf_log ("glusterd", GF_LOG_ERROR,
-                        "priv is NULL");
-                goto out;
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "One or more clients "
+                        "don't support the required op-version");
+                ret = gf_asprintf (op_errstr, "One or more connected clients "
+                                   "cannot support the feature being set. "
+                                   "These clients need to be upgraded or "
+                                   "disconnected before running this command"
+                                   " again");
+                return -1;
         }
+        return 0;
+}
 
-        dict = dict_new ();
-        if (!dict)
-                goto out;
+static int
+glusterd_op_stage_set_volume (dict_t *dict, char **op_errstr)
+{
+        int                             ret                     = -1;
+        char                            *volname                = NULL;
+        int                             exists                  = 0;
+        char                            *key                    = NULL;
+        char                            *key_fixed              = NULL;
+        char                            *value                  = NULL;
+        char                            str[100]                = {0, };
+        int                             count                   = 0;
+        int                             dict_count              = 0;
+        char                            errstr[2048]            = {0, };
+        glusterd_volinfo_t              *volinfo                = NULL;
+        dict_t                          *val_dict               = NULL;
+        gf_boolean_t                    global_opt              = _gf_false;
+        glusterd_volinfo_t              *voliter                = NULL;
+        glusterd_conf_t                 *priv                   = NULL;
+        xlator_t                        *this                   = NULL;
+        uint32_t                        new_op_version          = 0;
+        uint32_t                        local_new_op_version    = 0;
+        uint32_t                        key_op_version          = 0;
+        uint32_t                        local_key_op_version    = 0;
+        gf_boolean_t                    origin_glusterd         = _gf_true;
+        gf_boolean_t                    check_op_version        = _gf_true;
+        gf_boolean_t                    all_vol                 = _gf_false;
+        struct volopt_map_entry         *vme                    = NULL;
+
+        GF_ASSERT (dict);
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
+        val_dict = dict_new();
+        if (!val_dict)
+                goto out;
 
+        ret = dict_get_str (dict, "volname", &volname);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
                 goto out;
         }
 
-        ret = dict_get_str (dict, "volname", &volname);
+        /* Check if we can support the required op-version
+         * This check is not done on the originator glusterd. The originator
+         * glusterd sets this value.
+         */
+        origin_glusterd = is_origin_glusterd (dict);
+
+        if (!origin_glusterd) {
+                /* Check for v3.3.x origin glusterd */
+                check_op_version = dict_get_str_boolean (dict,
+                                                         "check-op-version",
+                                                         _gf_false);
 
+                if (check_op_version) {
+                        ret = dict_get_uint32 (dict, "new-op-version",
+                                               &new_op_version);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to get new_op_version");
+                                goto out;
+                        }
+
+                        if ((new_op_version > GD_OP_VERSION_MAX) ||
+                            (new_op_version < GD_OP_VERSION_MIN)) {
+                                ret = -1;
+                                snprintf (errstr, sizeof (errstr),
+                                          "Required op_version (%d) is not "
+                                          "supported", new_op_version);
+                                gf_log (this->name, GF_LOG_ERROR, "%s", errstr);
+                                goto out;
+                        }
+                }
+        }
+
+        ret = dict_get_int32 (dict, "count", &dict_count);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Count(dict),not set in Volume-Set");
                 goto out;
         }
 
-        exists = glusterd_check_volume_exists (volname);
+        if (dict_count == 0) {
+        /*No options would be specified of volume set help */
+                if (dict_get (dict, "help" ))  {
+                        ret = 0;
+                        goto out;
+                }
 
-        if (exists) {
-                gf_log ("", GF_LOG_ERROR, "Volume with name: %s exists",
-                        volname);
+                if (dict_get (dict, "help-xml" )) {
+#if (HAVE_LIB_XML)
+                        ret = 0;
+                        goto out;
+#else
+                        ret  = -1;
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "libxml not present in the system");
+                        *op_errstr = gf_strdup ("Error: xml libraries not "
+                                                "present to produce xml-output");
+                        goto out;
+#endif
+                }
+                gf_log (this->name, GF_LOG_ERROR, "No options received ");
+                *op_errstr = gf_strdup ("Options not specified");
                 ret = -1;
-        } else {
-                ret = 0;
-        }
-        ret = dict_get_int32 (dict, "count", &brick_count);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get count");
                 goto out;
         }
 
-        ret = dict_get_str (dict, "bricks", &bricks);
+        ret = dict_get_str (dict, "volname", &volname);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get bricks");
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
                 goto out;
         }
 
-        if (bricks) {
-                brick_list = gf_strdup (bricks);
-                if (!brick_list) {
+        if (strcasecmp (volname, "all") != 0) {
+                exists = glusterd_check_volume_exists (volname);
+                if (!exists) {
+                        snprintf (errstr, sizeof (errstr),
+                                  FMTSTR_CHECK_VOL_EXISTS, volname);
+                        gf_log (this->name, GF_LOG_ERROR, "%s", errstr);
                         ret = -1;
-                        gf_log ("", GF_LOG_ERROR, "Out of memory");
                         goto out;
-                } else {
-                        free_ptr = brick_list;
                 }
-        }
 
-        while ( i < brick_count) {
-                i++;
-                brick= strtok_r (brick_list, " \n", &tmpptr);
-                brick_list = tmpptr;
-                ret = glusterd_brickinfo_from_brick (brick, &brick_info);
+                ret = glusterd_volinfo_find (volname, &volinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                FMTSTR_CHECK_VOL_EXISTS, volname);
+                        goto out;
+                }
+
+                ret = glusterd_validate_volume_id (dict, volinfo);
                 if (ret)
                         goto out;
-                snprintf (cmd_str, 1024, "%s", brick_info->path);
-                ret = glusterd_resolve_brick (brick_info);
+        } else {
+                all_vol = _gf_true;
+        }
+
+        local_new_op_version = priv->op_version;
+
+        for ( count = 1; ret != 1 ; count++ ) {
+                global_opt = _gf_false;
+                sprintf (str, "key%d", count);
+                ret = dict_get_str (dict, str, &key);
+                if (ret)
+                        break;
+
+                sprintf (str, "value%d", count);
+                ret = dict_get_str (dict, str, &value);
                 if (ret) {
-                        gf_log ("glusterd", GF_LOG_ERROR, "cannot resolve "
-                                "brick: %s:%s", brick_info->hostname,
-                                brick_info->path);
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "invalid key,value pair in 'volume set'");
+                        ret = -1;
                         goto out;
                 }
 
-                if (!uuid_compare (brick_info->uuid, priv->uuid)) {
-                        ret = stat (cmd_str, &st_buf);
-                        if (ret == -1) {
-                                snprintf (msg, 2048,"Volume name %s, brick"
-                                        ": %s:%s, path %s not present", volname,
-                                          brick_info->hostname, brick_info->path, brick_info->path);
-                                gf_log ("glusterd",GF_LOG_ERROR, "%s", msg);
-                                *op_errstr = gf_strdup (msg);
-                                goto out;
-                        }
+                if (strcmp (key, "config.memory-accounting") == 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "enabling memory accounting for volume %s",
+                                volname);
+                        ret = 0;
+                }
 
-                        if (!S_ISDIR (st_buf.st_mode)) {
-                                snprintf (msg, 2048, "Volume name %s, brick"
-                                          ": %s, path %s is not a directory", volname,
-                                          brick, brick_info->path);
-                                gf_log ("glusterd", GF_LOG_ERROR,
-                                        "%s", msg);
-                                *op_errstr = gf_strdup (msg);
+                if (strcmp (key, "config.transport") == 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "changing transport-type for volume %s",
+                                volname);
+                        ret = 0;
+                        /* if value is none of 'tcp/rdma/tcp,rdma' error out */
+                        if (!((strcasecmp (value, "rdma") == 0) ||
+                              (strcasecmp (value, "tcp") == 0) ||
+                              (strcasecmp (value, "tcp,rdma") == 0) ||
+                              (strcasecmp (value, "rdma,tcp") == 0))) {
+                                ret = snprintf (errstr, sizeof (errstr),
+                                                "transport-type %s does "
+                                                "not exist", value);
+                                /* lets not bother about above return value,
+                                   its a failure anyways */
                                 ret = -1;
                                 goto out;
                         }
-                        brick_list = tmpptr;
                 }
-                glusterd_brickinfo_delete (brick_info);
-        }
-out:
-        if (dict)
-                dict_unref (dict);
-        if (free_ptr)
-                GF_FREE (free_ptr);
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
 
-        return ret;
-}
+                if (is_key_glusterd_hooks_friendly (key))
+                        continue;
 
-static int
-glusterd_op_stage_start_volume (gd1_mgmt_stage_op_req *req, char **op_errstr)
-{
-        int                                     ret = 0;
-        char                                    volname [1024] = {0,};
-        gf_boolean_t                            exists = _gf_false;
-        glusterd_volinfo_t                      *volinfo = NULL;
-        glusterd_brickinfo_t                    *brickinfo = NULL;
-        struct stat                             statbuf = {0,};
-        char                                    msg[2048];
-        glusterd_conf_t                         *priv = NULL;
+                for (vme = &glusterd_volopt_map[0]; vme->key; vme++) {
+                        if ((vme->validate_fn) &&
+                            ((!strcmp (key, vme->key)) ||
+                             (!strcmp (key, strchr (vme->key, '.') + 1)))) {
+                                ret = vme->validate_fn (dict, key, value,
+                                                        op_errstr);
+                                if (ret)
+                                        goto out;
+                                break;
+                        }
+                }
 
-        GF_ASSERT (req);
+                exists = glusterd_check_option_exists (key, &key_fixed);
+                if (exists == -1) {
+                        ret = -1;
+                        goto out;
+                }
 
-        priv = THIS->private;
-        if (!priv) {
-                gf_log ("glusterd", GF_LOG_ERROR,
-                        "priv is NULL");
-                ret = -1;
-                goto out;
-        }
+                if (!exists) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Option with name: %s does not exist", key);
+                        ret = snprintf (errstr, sizeof (errstr),
+                                       "option : %s does not exist",
+                                       key);
+                        if (key_fixed)
+                                snprintf (errstr + ret, sizeof (errstr) - ret,
+                                          "\nDid you mean %s?", key_fixed);
+                        ret = -1;
+                        goto out;
+                }
 
-        strncpy (volname, req->buf.buf_val, req->buf.buf_len);
-        //volname = req->buf.buf_val;
+                if (key_fixed)
+                        key = key_fixed;
+                ALL_VOLUME_OPTION_CHECK (volname, key, ret, op_errstr, out);
+                ret = glusterd_validate_quorum_options (this, key, value,
+                                                        op_errstr);
+                if (ret)
+                        goto out;
 
-        exists = glusterd_check_volume_exists (volname);
+                local_key_op_version = glusterd_get_op_version_for_key (key);
+                if (local_key_op_version > local_new_op_version)
+                        local_new_op_version = local_key_op_version;
 
-        if (!exists) {
-                snprintf (msg, 2048, "Volume with name %s does not exist", volname);
-                gf_log ("", GF_LOG_ERROR, "%s",
-                        msg);
-                *op_errstr = gf_strdup (msg);
-                ret = -1;
-        } else {
-                ret = 0;
-        }
+                sprintf (str, "op-version%d", count);
+                if (origin_glusterd) {
+                        ret = dict_set_uint32 (dict, str, local_key_op_version);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set key-op-version in dict");
+                                goto out;
+                        }
+                } else if (check_op_version) {
+                        ret = dict_get_uint32 (dict, str, &key_op_version);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to get key-op-version from"
+                                        " dict");
+                                goto out;
+                        }
+                        if (local_key_op_version != key_op_version) {
+                                ret = -1;
+                                snprintf (errstr, sizeof (errstr),
+                                          "option: %s op-version mismatch",
+                                          key);
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "%s, required op-version = %"PRIu32", "
+                                        "available op-version = %"PRIu32,
+                                        errstr, key_op_version,
+                                        local_key_op_version);
+                                goto out;
+                        }
+                }
 
-        ret  = glusterd_volinfo_find (volname, &volinfo);
+                if (glusterd_check_globaloption (key))
+                        global_opt = _gf_true;
 
-        if (ret)
-                goto out;
+                ret = dict_set_str (val_dict, key, value);
 
-        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
-                ret = glusterd_resolve_brick (brickinfo);
                 if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to resolve brick"
-                                " with hostname: %s, export: %s",
-                                brickinfo->hostname,brickinfo->path);
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to set the options in 'volume set'");
+                        ret = -1;
                         goto out;
                 }
 
-                if (!uuid_compare (brickinfo->uuid, priv->uuid)) {
-                        ret = stat (brickinfo->path, &statbuf);
-                        if (ret == -1) {
-                                snprintf (msg, 2048, "Volume name %s, brick"
-                                          ": %s:%s, path %s is not present",
-                                          volname, brickinfo->hostname, brickinfo->path, brickinfo->path);
-                                gf_log ("glusterd", GF_LOG_ERROR,
-                                        "%s", msg);
-                                *op_errstr = gf_strdup (msg);
-                                goto out;
+                *op_errstr = NULL;
+                if (!global_opt && !all_vol)
+                        ret = glusterd_validate_reconfopts (volinfo, val_dict, op_errstr);
+                else if (!all_vol) {
+                        voliter = NULL;
+                        list_for_each_entry (voliter, &priv->volumes, vol_list) {
+                                ret = glusterd_validate_globalopts (voliter, val_dict, op_errstr);
+                                if (ret)
+                                        break;
                         }
                 }
 
-                if (GLUSTERD_STATUS_STARTED == volinfo->status) {
-                        snprintf (msg, 2048, "Volume %s already started",
-                                  volname);
-                        gf_log ("glusterd", GF_LOG_ERROR,
-                                "%s", msg);
-                        *op_errstr = gf_strdup (msg);
-                        ret = -1;
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Could not create "
+                                "temp volfile, some option failed: %s",
+                                *op_errstr);
+                        goto out;
+                }
+                dict_del (val_dict, key);
+
+                if (key_fixed) {
+                        GF_FREE (key_fixed);
+                        key_fixed = NULL;
                 }
         }
+
+        // Check if all the connected clients support the new op-version
+        ret = glusterd_check_client_op_version_support (volname,
+                                                        local_new_op_version,
+                                                        op_errstr);
+        if (ret)
+                goto out;
+
+        if (origin_glusterd) {
+                ret = dict_set_uint32 (dict, "new-op-version",
+                                       local_new_op_version);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to set new-op-version in dict");
+                        goto out;
+                }
+                /* Set this value in dict so other peers know to check for
+                 * op-version. This is a hack for 3.3.x compatibility
+                 *
+                 * TODO: Remove this and the other places this is referred once
+                 * 3.3.x compatibility is not required
+                 */
+                ret = dict_set_uint32 (dict, "check-op-version",
+                                       _gf_true);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to set check-op-version in dict");
+                        goto out;
+                }
+        }
+
+        ret = 0;
+
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        if (val_dict)
+                dict_unref (val_dict);
 
+        GF_FREE (key_fixed);
+        if (errstr[0] != '\0')
+                *op_errstr = gf_strdup (errstr);
+
+        if (ret) {
+                if (!(*op_errstr)) {
+                        *op_errstr = gf_strdup ("Error, Validation Failed");
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Error, Cannot Validate option :%s",
+                                *op_errstr);
+                } else {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Error, Cannot Validate option");
+                }
+        }
         return ret;
 }
 
 static int
-glusterd_op_stop_volume_args_get (gd1_mgmt_stage_op_req *req,
-                                  dict_t *dict, char** volname,
-                                  int *flags)
+glusterd_op_stage_reset_volume (dict_t *dict, char **op_errstr)
 {
-        int ret = -1;
+        int                                      ret           = 0;
+        char                                    *volname       = NULL;
+        gf_boolean_t                             exists        = _gf_false;
+        char                                    msg[2048]      = {0};
+        char                                    *key = NULL;
+        char                                    *key_fixed = NULL;
+        glusterd_volinfo_t                      *volinfo       = NULL;
+        xlator_t                                *this = NULL;
 
-        if (!req || !dict || !volname || !flags)
-                goto out;
+        this = THIS;
+        GF_ASSERT (this);
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
+        ret = dict_get_str (dict, "volname", &volname);
 
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
                 goto out;
         }
 
-        ret = dict_get_str (dict, "volname", volname);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
-                goto out;
+        if (strcasecmp (volname, "all") != 0) {
+                exists = glusterd_check_volume_exists (volname);
+                if (!exists) {
+                        snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS,
+                                  volname);
+                        ret = -1;
+                        goto out;
+                }
+                ret = glusterd_volinfo_find (volname, &volinfo);
+                if (ret) {
+                        snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS,
+                                  volname);
+                        goto out;
+                }
+
+                ret = glusterd_validate_volume_id (dict, volinfo);
+                if (ret)
+                        goto out;
         }
 
-        ret = dict_get_int32 (dict, "flags", flags);
+        ret = dict_get_str (dict, "key", &key);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get flags");
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get option key");
                 goto out;
         }
+        if (strcmp(key, "all")) {
+                exists = glusterd_check_option_exists (key, &key_fixed);
+                if (exists == -1) {
+                        ret = -1;
+                        goto out;
+                }
+                if (!exists) {
+                        ret = snprintf (msg, sizeof (msg),
+                                        "Option %s does not exist", key);
+                        if (key_fixed)
+                                snprintf (msg + ret, sizeof (msg) - ret,
+                                          "\nDid you mean %s?", key_fixed);
+                        ret = -1;
+                        goto out;
+                } else if (exists > 0) {
+                        if (key_fixed)
+                                key = key_fixed;
+                        ALL_VOLUME_OPTION_CHECK (volname, key, ret,
+                                                 op_errstr, out);
+                }
+        }
+
 out:
+        GF_FREE (key_fixed);
+
+        if (msg[0] != '\0') {
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
         return ret;
 }
+
+
+
 static int
-glusterd_op_stage_stop_volume (gd1_mgmt_stage_op_req *req)
+glusterd_op_stage_sync_volume (dict_t *dict, char **op_errstr)
 {
         int                                     ret = -1;
-        dict_t                                  *dict = NULL;
         char                                    *volname = NULL;
-        int                                     flags = 0;
+        char                                    *hostname = NULL;
         gf_boolean_t                            exists = _gf_false;
-        glusterd_volinfo_t                      *volinfo = NULL;
-
-        dict = dict_new ();
-        if (!dict)
-                goto out;
+        glusterd_peerinfo_t                     *peerinfo = NULL;
+        char                                    msg[2048] = {0,};
+        glusterd_volinfo_t                      *volinfo  = NULL;
 
-        ret = glusterd_op_stop_volume_args_get (req, dict, &volname, &flags);
-        if (ret)
+        ret = dict_get_str (dict, "hostname", &hostname);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "hostname couldn't be "
+                          "retrieved from msg");
+                *op_errstr = gf_strdup (msg);
                 goto out;
-
-        exists = glusterd_check_volume_exists (volname);
-
-        if (!exists) {
-                gf_log ("", GF_LOG_ERROR, "Volume with name %s does not exist",
-                        volname);
-                ret = -1;
-        } else {
-                ret = 0;
         }
 
-        ret  = glusterd_volinfo_find (volname, &volinfo);
-
-        if (ret)
-                goto out;
-
-        if (!(flags & GF_CLI_FLAG_OP_FORCE)) {
-                ret = glusterd_is_volume_started (volinfo);
+        if (gf_is_local_addr (hostname)) {
+                //volname is not present in case of sync all
+                ret = dict_get_str (dict, "volname", &volname);
+                if (!ret) {
+                        exists = glusterd_check_volume_exists (volname);
+                        if (!exists) {
+                                snprintf (msg, sizeof (msg), "Volume %s "
+                                         "does not exist", volname);
+                                *op_errstr = gf_strdup (msg);
+                                ret = -1;
+                                goto out;
+                        }
+                        ret = glusterd_volinfo_find (volname, &volinfo);
+                        if (ret)
+                                goto out;
 
+                } else {
+                        ret = 0;
+                }
+         } else {
+                ret = glusterd_friend_find (NULL, hostname, &peerinfo);
                 if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Volume %s "
-                                "has not been started", volname);
+                        snprintf (msg, sizeof (msg), "%s, is not a friend",
+                                  hostname);
+                        *op_errstr = gf_strdup (msg);
                         goto out;
                 }
-        }
 
+                if (!peerinfo->connected) {
+                        snprintf (msg, sizeof (msg), "%s, is not connected at "
+                                  "the moment", hostname);
+                        *op_errstr = gf_strdup (msg);
+                        ret = -1;
+                        goto out;
+                }
 
-out:
-        if (dict)
-                dict_unref (dict);
+        }
 
+out:
         gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
 
         return ret;
 }
 
 static int
-glusterd_op_stage_delete_volume (gd1_mgmt_stage_op_req *req)
+glusterd_op_stage_status_volume (dict_t *dict, char **op_errstr)
 {
-        int                                     ret = 0;
-        char                                    volname [1024] = {0,};
-        gf_boolean_t                            exists = _gf_false;
-        glusterd_volinfo_t                      *volinfo = NULL;
+        int                    ret            = -1;
+        uint32_t               cmd            = 0;
+        char                   msg[2048]      = {0,};
+        char                  *volname        = NULL;
+        char                  *brick          = NULL;
+        xlator_t              *this           = NULL;
+        glusterd_conf_t       *priv           = NULL;
+        glusterd_brickinfo_t  *brickinfo      = NULL;
+        glusterd_volinfo_t    *volinfo        = NULL;
+        dict_t                *vol_opts       = NULL;
+        gf_boolean_t           nfs_disabled   = _gf_false;
+        gf_boolean_t           shd_enabled    = _gf_true;
+
+        GF_ASSERT (dict);
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT(priv);
 
-        GF_ASSERT (req);
+        ret = dict_get_uint32 (dict, "cmd", &cmd);
+        if (ret)
+                goto out;
 
-        strncpy (volname, req->buf.buf_val, req->buf.buf_len);
+        if (cmd & GF_CLI_STATUS_ALL)
+                goto out;
 
-        exists = glusterd_check_volume_exists (volname);
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "Unable to get volume name");
+                goto out;
+        }
 
-        if (!exists) {
-                gf_log ("", GF_LOG_ERROR, "Volume with name %s does not exist",
-                        volname);
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (msg, sizeof(msg), "Volume %s does not exist",
+                          volname);
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
                 ret = -1;
                 goto out;
-        } else {
-                ret = 0;
         }
 
-        ret  = glusterd_volinfo_find (volname, &volinfo);
-
+        ret = glusterd_validate_volume_id (dict, volinfo);
         if (ret)
                 goto out;
 
         ret = glusterd_is_volume_started (volinfo);
-
         if (!ret) {
-                gf_log ("", GF_LOG_ERROR, "Volume %s has been started."
-                        "Volume needs to be stopped before deletion.",
-                        volname);
+                snprintf (msg, sizeof (msg), "Volume %s is not started",
+                          volname);
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
                 ret = -1;
                 goto out;
         }
 
+        vol_opts = volinfo->dict;
+
+        if ((cmd & GF_CLI_STATUS_NFS) != 0) {
+                nfs_disabled = dict_get_str_boolean (vol_opts, "nfs.disable",
+                                                     _gf_false);
+                if (nfs_disabled) {
+                        ret = -1;
+                        snprintf (msg, sizeof (msg),
+                                  "NFS server is disabled for volume %s",
+                                  volname);
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                        goto out;
+                }
+        } else if ((cmd & GF_CLI_STATUS_SHD) != 0) {
+                if (!glusterd_is_volume_replicate (volinfo)) {
+                        ret = -1;
+                        snprintf (msg, sizeof (msg),
+                                  "Volume %s is not of type replicate",
+                                  volname);
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                        goto out;
+                }
+
+                shd_enabled = dict_get_str_boolean (vol_opts,
+                                                    "cluster.self-heal-daemon",
+                                                    _gf_true);
+                if (!shd_enabled) {
+                        ret = -1;
+                        snprintf (msg, sizeof (msg),
+                                  "Self-heal Daemon is disabled for volume %s",
+                                  volname);
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                        goto out;
+                }
+
+        } else if ((cmd & GF_CLI_STATUS_BRICK) != 0) {
+                ret = dict_get_str (dict, "brick", &brick);
+                if (ret)
+                        goto out;
+
+                ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo,
+                                                              &brickinfo);
+                if (ret) {
+                        snprintf (msg, sizeof(msg), "No brick %s in"
+                                  " volume %s", brick, volname);
+                        gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+
+                        ret = -1;
+                        goto out;
+                }
+        }
+
         ret = 0;
 
-out:
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+ out:
+        if (ret) {
+                if (msg[0] != '\0')
+                        *op_errstr = gf_strdup (msg);
+                else
+                        *op_errstr = gf_strdup ("Validation Failed for Status");
+        }
 
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning: %d", ret);
         return ret;
 }
 
-static int
-glusterd_op_stage_add_brick (gd1_mgmt_stage_op_req *req, char **op_errstr)
-{
-        int                                     ret = 0;
-        dict_t                                  *dict = NULL;
-        char                                    *volname = NULL;
-        int                                     count = 0;
-        int                                     i = 0;
-        char                                    *bricks    = NULL;
-        char                                    *brick_list = NULL;
-        char                                    *saveptr = NULL;
-        char                                    *free_ptr = NULL;
-        char                                    *brick = NULL;
-        glusterd_brickinfo_t                    *brickinfo = NULL;
-        glusterd_volinfo_t                      *volinfo = NULL;
-        struct stat                             st_buf = {0,};
-        char                                    cmd_str[1024];
-        glusterd_conf_t                         *priv = NULL;
-        char                                    msg[2048] = {0,};
-
-        GF_ASSERT (req);
 
-        priv = THIS->private;
-        if (!priv)
-                goto out;
+static gf_boolean_t
+glusterd_is_profile_on (glusterd_volinfo_t *volinfo)
+{
+        int                                     ret = -1;
+        gf_boolean_t                            is_latency_on = _gf_false;
+        gf_boolean_t                            is_fd_stats_on = _gf_false;
 
-        dict = dict_new ();
-        if (!dict)
-                goto out;
+        GF_ASSERT (volinfo);
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
+        ret = glusterd_volinfo_get_boolean (volinfo, VKEY_DIAG_CNT_FOP_HITS);
+        if (ret != -1)
+                is_fd_stats_on = ret;
+        ret = glusterd_volinfo_get_boolean (volinfo, VKEY_DIAG_LAT_MEASUREMENT);
+        if (ret != -1)
+                is_latency_on = ret;
+        if ((_gf_true == is_latency_on) &&
+            (_gf_true == is_fd_stats_on))
+                return _gf_true;
+        return _gf_false;
+}
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
-                goto out;
-        }
+static int
+glusterd_op_stage_stats_volume (dict_t *dict, char **op_errstr)
+{
+        int                                     ret = -1;
+        char                                    *volname = NULL;
+        gf_boolean_t                            exists = _gf_false;
+        char                                    msg[2048] = {0,};
+        int32_t                                 stats_op = GF_CLI_STATS_NONE;
+        glusterd_volinfo_t                      *volinfo = NULL;
 
         ret = dict_get_str (dict, "volname", &volname);
-
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                snprintf (msg, sizeof (msg), "Volume name get failed");
                 goto out;
         }
 
+        exists = glusterd_check_volume_exists (volname);
         ret = glusterd_volinfo_find (volname, &volinfo);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to find volume: %s", volname);
+        if ((!exists) || (ret < 0)) {
+                snprintf (msg, sizeof (msg), "Volume %s, "
+                         "doesn't exist", volname);
+                ret = -1;
                 goto out;
         }
 
-        ret = dict_get_int32 (dict, "count", &count);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get count");
+        ret = glusterd_validate_volume_id (dict, volinfo);
+        if (ret)
                 goto out;
-        }
 
-        ret = dict_get_str (dict, "bricks", &bricks);
+        ret = dict_get_int32 (dict, "op", &stats_op);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get bricks");
+                snprintf (msg, sizeof (msg), "Volume profile op get failed");
                 goto out;
         }
 
-        if (bricks) {
-                brick_list = gf_strdup (bricks);
-                free_ptr = brick_list;
-        }
-
-        if (count)
-                brick = strtok_r (brick_list+1, " \n", &saveptr);
+        if (GF_CLI_STATS_START == stats_op) {
+                if (_gf_true == glusterd_is_profile_on (volinfo)) {
+                        snprintf (msg, sizeof (msg), "Profile on Volume %s is"
+                                  " already started", volinfo->volname);
+                        ret = -1;
+                        goto out;
+                }
 
-        while ( i < count) {
-                ret = glusterd_brickinfo_get (brick, volinfo, &brickinfo);
-                if (!ret) {
-                        gf_log ("", GF_LOG_ERROR, "Adding duplicate brick: %s",
-                                brick);
+        }
+        if ((GF_CLI_STATS_STOP == stats_op) ||
+            (GF_CLI_STATS_INFO == stats_op)) {
+                if (_gf_false == glusterd_is_profile_on (volinfo)) {
+                        snprintf (msg, sizeof (msg), "Profile on Volume %s is"
+                                  " not started", volinfo->volname);
                         ret = -1;
+
                         goto out;
-                } else {
-                        ret = glusterd_brickinfo_from_brick(brick, &brickinfo);
-                        if (ret) {
-                                gf_log ("", GF_LOG_ERROR, "Add-brick: Unable"
-                                       " to get brickinfo");
-                                goto out;
-                        }
                 }
-                snprintf (cmd_str, 1024, "%s", brickinfo->path);
-                ret = glusterd_resolve_brick (brickinfo);
-                if (ret) {
-                        gf_log ("glusterd", GF_LOG_ERROR,
-                                "resolve brick failed");
+        }
+        if ((GF_CLI_STATS_TOP == stats_op) ||
+            (GF_CLI_STATS_INFO == stats_op)) {
+                if (_gf_false == glusterd_is_volume_started (volinfo)) {
+                        snprintf (msg, sizeof (msg), "Volume %s is not started.",
+                                  volinfo->volname);
+                        gf_log ("glusterd", GF_LOG_ERROR, "%s", msg);
+                        ret = -1;
                         goto out;
                 }
+        }
+        ret = 0;
+out:
+        if (msg[0] != '\0') {
+                gf_log ("glusterd", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+        }
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
 
-                if (!uuid_compare (brickinfo->uuid, priv->uuid)) {
-                        ret = stat (cmd_str, &st_buf);
-                        if (ret == -1) {
-                                snprintf (msg, 2048, "Volume name %s, brick"
-                                          ": %s, path %s not present", volname,
-                                          brick, brickinfo->path);
-                                gf_log ("glusterd", GF_LOG_ERROR,
-                                        "%s", msg);
-                                *op_errstr = gf_strdup (msg);
-                                goto out;
-                        }
 
-                        if (!S_ISDIR (st_buf.st_mode)) {
-                                snprintf (msg, 2048, "Volume name %s, brick"
-                                          ": %s, path %s is not a directory", volname,
-                                          brick, brickinfo->path);
-                                gf_log ("glusterd", GF_LOG_ERROR,
-                                        "%s", msg);
-                                *op_errstr = gf_strdup (msg);
-                                ret = -1;
-                                goto out;
-                        }
+static int
+_delete_reconfig_opt (dict_t *this, char *key, data_t *value, void *data)
+{
+        int32_t        *is_force = 0;
+
+        GF_ASSERT (data);
+        is_force = (int32_t*)data;
+
+        if (*is_force != 1) {
+                if (_gf_true == glusterd_check_voloption_flags (key,
+                                                         OPT_FLAG_FORCE)) {
+                /* indicate to caller that we don't set the option
+                 * due to being protected
+                 */
+                        *is_force = *is_force | GD_OP_PROTECTED;
+                        goto out;
+                } else {
+                        *is_force = *is_force | GD_OP_UNPROTECTED;
                 }
-
-                glusterd_brickinfo_delete (brickinfo);
-                brick = strtok_r (NULL, " \n", &saveptr);
-                i++;
         }
 
+        gf_log ("", GF_LOG_DEBUG, "deleting dict with key=%s,value=%s",
+                key, value->data);
+        dict_del (this, key);
 out:
-        if (dict)
-                dict_unref (dict);
-        if (free_ptr)
-                GF_FREE (free_ptr);
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
-
-        return ret;
+        return 0;
 }
 
 static int
-glusterd_op_stage_replace_brick (gd1_mgmt_stage_op_req *req, char **op_errstr)
+_delete_reconfig_global_opt (dict_t *this, char *key, data_t *value, void *data)
 {
-        int                                      ret           = 0;
-        dict_t                                  *dict          = NULL;
-        char                                    *src_brick     = NULL;
-        char                                    *dst_brick     = NULL;
-        char                                    *volname       = NULL;
-        gf1_cli_replace_op                       replace_op    = 0;
-        glusterd_volinfo_t                      *volinfo       = NULL;
-        glusterd_brickinfo_t                    *src_brickinfo = NULL;
-        char                                    *host          = NULL;
-        char                                    *path          = NULL;
-        char                                    msg[2048]      = {0};
-        char                                    *dup_dstbrick  = NULL;
-        glusterd_peerinfo_t                     *peerinfo = NULL;
-        struct stat                             st_buf = {0,};
+        int32_t        *is_force = 0;
 
-        GF_ASSERT (req);
+        GF_ASSERT (data);
+        is_force = (int32_t*)data;
 
-        dict = dict_new ();
-        if (!dict)
+        if (strcmp (GLUSTERD_GLOBAL_OPT_VERSION, key) == 0)
                 goto out;
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
+        _delete_reconfig_opt (this, key, value, data);
+out:
+        return 0;
+}
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
-                goto out;
+static int
+glusterd_options_reset (glusterd_volinfo_t *volinfo, char *key,
+                        int32_t *is_force)
+{
+        int                      ret = 0;
+        data_t                  *value = NULL;
+        char                    *key_fixed = NULL;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (volinfo->dict);
+        GF_ASSERT (key);
+
+        if (!strncmp(key, "all", 3))
+                dict_foreach (volinfo->dict, _delete_reconfig_opt, is_force);
+        else {
+                value = dict_get (volinfo->dict, key);
+                if (!value) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "no value set for option %s", key);
+                        goto out;
+                }
+                _delete_reconfig_opt (volinfo->dict, key, value, is_force);
         }
 
-        ret = dict_get_str (dict, "src-brick", &src_brick);
+        gd_update_volume_op_versions (volinfo);
 
+        ret = glusterd_create_volfiles_and_notify_services (volinfo);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get src brick");
+                gf_log (this->name, GF_LOG_ERROR, "Unable to create volfile for"
+                        " 'volume reset'");
+                ret = -1;
                 goto out;
         }
 
-        gf_log ("", GF_LOG_DEBUG,
-                "src brick=%s", src_brick);
+        ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret)
+                goto out;
 
-        ret = dict_get_str (dict, "dst-brick", &dst_brick);
+        if (GLUSTERD_STATUS_STARTED == volinfo->status) {
+                ret = glusterd_nodesvcs_handle_reconfigure (volinfo);
+                if (ret)
+                        goto out;
+        }
+
+        ret = 0;
+
+out:
+        GF_FREE (key_fixed);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
 
+static int
+glusterd_op_reset_all_volume_options (xlator_t *this, dict_t *dict)
+{
+        char            *key            = NULL;
+        char            *key_fixed      = NULL;
+        int             ret             = -1;
+        int32_t         is_force        = 0;
+        glusterd_conf_t *conf           = NULL;
+        dict_t          *dup_opt        = NULL;
+        gf_boolean_t    all             = _gf_false;
+        char            *next_version   = NULL;
+        gf_boolean_t    quorum_action   = _gf_false;
+
+        conf = this->private;
+        ret = dict_get_str (dict, "key", &key);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get dest brick");
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get key");
                 goto out;
         }
 
-        gf_log ("", GF_LOG_DEBUG,
-                "dst brick=%s", dst_brick);
+        ret = dict_get_int32 (dict, "force", &is_force);
+        if (ret)
+                is_force = 0;
 
-        ret = dict_get_str (dict, "volname", &volname);
+        if (strcmp (key, "all")) {
+                ret = glusterd_check_option_exists (key, &key_fixed);
+                if (ret <= 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "Option %s does not "
+                                "exist", key);
+                        ret = -1;
+                        goto out;
+                }
+        } else {
+                all = _gf_true;
+        }
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+        if (key_fixed)
+                key = key_fixed;
+
+        ret = -1;
+        dup_opt = dict_new ();
+        if (!dup_opt)
                 goto out;
+        if (!all) {
+                dict_copy (conf->opts, dup_opt);
+                dict_del (dup_opt, key);
         }
+        ret = glusterd_get_next_global_opt_version_str (conf->opts,
+                                                        &next_version);
+        if (ret)
+                goto out;
 
-        ret = dict_get_int32 (dict, "operation", (int32_t *)&replace_op);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "dict get on replace-brick operation failed");
+        ret = dict_set_str (dup_opt, GLUSTERD_GLOBAL_OPT_VERSION, next_version);
+        if (ret)
                 goto out;
-        }
 
-        ret = glusterd_volinfo_find (volname, &volinfo);
-        if (ret) {
-                snprintf (msg, sizeof (msg), "volume: %s does not exist",
-                          volname);
-                *op_errstr = gf_strdup (msg);
+        ret = glusterd_store_options (this, dup_opt);
+        if (ret)
                 goto out;
-        }
 
-        if (GLUSTERD_STATUS_STARTED != volinfo->status) {
-                ret = -1;
-                snprintf (msg, sizeof (msg), "volume: %s is not started",
-                          volname);
-                *op_errstr = gf_strdup (msg);
+        if (glusterd_is_quorum_changed (conf->opts, key, NULL))
+                quorum_action = _gf_true;
+
+        ret = dict_set_dynstr (conf->opts, GLUSTERD_GLOBAL_OPT_VERSION,
+                               next_version);
+        if (ret)
                 goto out;
+        else
+                next_version = NULL;
+
+        if (!all) {
+                dict_del (conf->opts, key);
+        } else {
+                dict_foreach (conf->opts, _delete_reconfig_global_opt,
+                              &is_force);
         }
+out:
+        GF_FREE (key_fixed);
+        if (dup_opt)
+                dict_unref (dup_opt);
+
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
+        if (quorum_action)
+                glusterd_do_quorum_action ();
+        GF_FREE (next_version);
+        return ret;
+}
 
-        ret = glusterd_brickinfo_get (src_brick, volinfo,
-                                      &src_brickinfo);
+static int
+glusterd_op_reset_volume (dict_t *dict, char **op_rspstr)
+{
+        glusterd_volinfo_t      *volinfo    = NULL;
+        int                     ret         = -1;
+        char                    *volname    = NULL;
+        char                    *key        = NULL;
+        char                    *key_fixed  = NULL;
+        int32_t                 is_force    = 0;
+        gf_boolean_t            quorum_action = _gf_false;
+        xlator_t                *this         = NULL;
+
+        this = THIS;
+        ret = dict_get_str (dict, "volname", &volname);
         if (ret) {
-                snprintf (msg, sizeof (msg), "brick: %s does not exist in "
-                          "volume: %s", src_brick, volname);
-                *op_errstr = gf_strdup (msg);
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name" );
                 goto out;
         }
 
-        if (!glusterd_is_local_addr (src_brickinfo->hostname)) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "I AM THE SOURCE HOST");
+        if (strcasecmp (volname, "all") == 0) {
+                ret = glusterd_op_reset_all_volume_options (this, dict);
+                goto out;
         }
 
-        dup_dstbrick = gf_strdup (dst_brick);
-        if (!dup_dstbrick) {
-                ret = -1;
-                gf_log ("", GF_LOG_ERROR, "Memory allocation failed");
+        ret = dict_get_int32 (dict, "force", &is_force);
+        if (ret)
+                is_force = 0;
+
+        ret = dict_get_str (dict, "key", &key);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get option key");
                 goto out;
         }
-        host = strtok (dup_dstbrick, ":");
-        path = strtok (NULL, ":");
 
-        if (!host || !path) {
-                gf_log ("", GF_LOG_ERROR,
-                        "dst brick %s is not of form <HOSTNAME>:<export-dir>",
-                        dst_brick);
-                ret = -1;
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS,
+                        volname);
                 goto out;
         }
-        if (glusterd_is_exisiting_brick (host, path)) {
-                snprintf(msg, sizeof(msg), "Brick: %s:%s already in use",
-                         host, path);
-                *op_errstr = gf_strdup (msg);
+
+        if (strcmp (key, "all") &&
+            glusterd_check_option_exists (key, &key_fixed) != 1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "volinfo dict inconsistency: option %s not found",
+                        key);
                 ret = -1;
                 goto out;
         }
+        if (key_fixed)
+                key = key_fixed;
 
-        if (!glusterd_is_local_addr (host)) {
-                ret = stat (path, &st_buf);
-                if (ret == -1) {
-                        snprintf (msg, sizeof (msg) ,"path: %s for brick: %s"
-                        " does not exist", path, dst_brick);
-                        gf_log ("glusterd", GF_LOG_ERROR, "%s", msg);
-                        *op_errstr = gf_strdup (msg);
-                        goto out;
-                }
-                if (!S_ISDIR (st_buf.st_mode)) {
-                        snprintf (msg, sizeof (msg), "Volume name %s, brick"
-                                  ": %s, path %s is not a directory", volname,
-                                  dst_brick, path);
-                        gf_log ("glusterd", GF_LOG_ERROR,
-                                "%s", msg);
-                        *op_errstr = gf_strdup (msg);
-                        ret = -1;
-                        goto out;
-                }
-        } else {
-                ret = glusterd_friend_find (NULL, host, &peerinfo);
-                if (ret) {
-                        snprintf (msg, sizeof (msg), "%s, is not a friend",
-                                  host);
-                        *op_errstr = gf_strdup (msg);
-                        goto out;
-                }
+        if (glusterd_is_quorum_changed (volinfo->dict, key, NULL))
+                quorum_action = _gf_true;
 
-                if (!peerinfo->connected) {
-                        snprintf (msg, sizeof (msg), "%s, is not connected at "
-                                  "the moment", host);
-                        *op_errstr = gf_strdup (msg);
-                        ret = -1;
-                        goto out;
-                }
-
-                if (GD_FRIEND_STATE_BEFRIENDED != peerinfo->state.state) {
-                        snprintf (msg, sizeof (msg), "%s, is not befriended "
-                                  "at the moment", host);
-                        *op_errstr = gf_strdup (msg);
+        ret = glusterd_options_reset (volinfo, key, &is_force);
+        if (ret == -1) {
+                gf_asprintf(op_rspstr, "Volume reset : failed");
+        } else if (is_force & GD_OP_PROTECTED) {
+                if (is_force & GD_OP_UNPROTECTED) {
+                        gf_asprintf (op_rspstr, "All unprotected fields were"
+                                     " reset. To reset the protected fields,"
+                                     " use 'force'.");
+                } else {
                         ret = -1;
-                        goto out;
+                        gf_asprintf (op_rspstr, "'%s' is protected. To reset"
+                                     " use 'force'.", key);
                 }
         }
-        ret = 0;
 
 out:
-        if (dup_dstbrick)
-                GF_FREE (dup_dstbrick);
-        if (dict)
-                dict_unref (dict);
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        GF_FREE (key_fixed);
+        if (quorum_action)
+                glusterd_do_quorum_action ();
 
+        gf_log (this->name, GF_LOG_DEBUG, "'volume reset' returning %d", ret);
         return ret;
 }
 
-static int
-glusterd_op_stage_log_filename (gd1_mgmt_stage_op_req *req)
+int
+glusterd_stop_bricks (glusterd_volinfo_t *volinfo)
 {
-        int                                     ret = -1;
-        dict_t                                  *dict = NULL;
-        char                                    *volname = NULL;
-        gf_boolean_t                            exists = _gf_false;
+        glusterd_brickinfo_t                    *brickinfo = NULL;
 
-        GF_ASSERT (req);
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+            /*TODO: Need to change @del_brick in brick_stop to _gf_true
+             * once we enable synctask in peer rpc prog */
+                if (glusterd_brick_stop (volinfo, brickinfo, _gf_false))
+                        return -1;
+        }
 
-        dict = dict_new ();
-        if (!dict)
-                goto out;
+        return 0;
+}
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
-                goto out;
-        }
+int
+glusterd_start_bricks (glusterd_volinfo_t *volinfo)
+{
+        int                      ret            = -1;
+        glusterd_brickinfo_t    *brickinfo      = NULL;
 
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
-                goto out;
-        }
+        GF_ASSERT (volinfo);
 
-        exists = glusterd_check_volume_exists (volname);
-        if (!exists) {
-                gf_log ("", GF_LOG_ERROR, "Volume with name: %s not exists",
-                        volname);
-                ret = -1;
-                goto out;
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                ret = glusterd_brick_start (volinfo, brickinfo, _gf_false);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_ERROR,
+                                "Failed to start %s:%s for %s",
+                                brickinfo->hostname, brickinfo->path,
+                                volinfo->volname);
+                        goto out;
+                }
         }
 
+        ret = 0;
 out:
-        if (dict)
-                dict_unref (dict);
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
-
         return ret;
 }
 
 static int
-glusterd_op_stage_log_rotate (gd1_mgmt_stage_op_req *req)
+glusterd_op_set_all_volume_options (xlator_t *this, dict_t *dict)
 {
-        int                                     ret = -1;
-        dict_t                                  *dict = NULL;
-        char                                    *volname = NULL;
-        gf_boolean_t                            exists = _gf_false;
-
-        GF_ASSERT (req);
-
-        dict = dict_new ();
-        if (!dict)
+        char            *key            = NULL;
+        char            *key_fixed      = NULL;
+        char            *value          = NULL;
+        char            *dup_value      = NULL;
+        int             ret             = -1;
+        glusterd_conf_t *conf           = NULL;
+        dict_t          *dup_opt        = NULL;
+        char            *next_version   = NULL;
+        gf_boolean_t    quorum_action   = _gf_false;
+
+        conf = this->private;
+        ret = dict_get_str (dict, "key1", &key);
+        if (ret)
                 goto out;
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
+        ret = dict_get_str (dict, "value1", &value);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
+                gf_log (this->name, GF_LOG_ERROR,
+                        "invalid key,value pair in 'volume set'");
                 goto out;
         }
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+        ret = glusterd_check_option_exists (key, &key_fixed);
+        if (ret <= 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Invalid key %s", key);
+                ret = -1;
                 goto out;
         }
 
-        exists = glusterd_check_volume_exists (volname);
-        if (!exists) {
-                gf_log ("", GF_LOG_ERROR, "Volume with name: %s not exists",
-                        volname);
-                ret = -1;
+        if (key_fixed)
+                key = key_fixed;
+
+        ret = -1;
+        dup_opt = dict_new ();
+        if (!dup_opt)
+                goto out;
+        dict_copy (conf->opts, dup_opt);
+        ret = dict_set_str (dup_opt, key, value);
+        if (ret)
                 goto out;
-        }
 
-out:
-        if (dict)
-                dict_unref (dict);
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        ret = glusterd_get_next_global_opt_version_str (conf->opts,
+                                                        &next_version);
+        if (ret)
+                goto out;
+
+        ret = dict_set_str (dup_opt, GLUSTERD_GLOBAL_OPT_VERSION, next_version);
+        if (ret)
+                goto out;
+
+        dup_value = gf_strdup (value);
+        if (!dup_value)
+                goto out;
+
+        ret = glusterd_store_options (this, dup_opt);
+        if (ret)
+                goto out;
+
+        if (glusterd_is_quorum_changed (conf->opts, key, value))
+                quorum_action = _gf_true;
 
+        ret = dict_set_dynstr (conf->opts, GLUSTERD_GLOBAL_OPT_VERSION,
+                               next_version);
+        if (ret)
+                goto out;
+        else
+                next_version = NULL;
+
+        ret = dict_set_dynstr (conf->opts, key, dup_value);
+        if (ret)
+                goto out;
+out:
+        GF_FREE (key_fixed);
+        if (dup_opt)
+                dict_unref (dup_opt);
+
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
+        if (quorum_action)
+                glusterd_do_quorum_action ();
+        GF_FREE (next_version);
         return ret;
 }
 
 static int
-glusterd_op_stage_set_volume (gd1_mgmt_stage_op_req *req)
+glusterd_op_set_volume (dict_t *dict)
 {
-        int                                      ret           = 0;
-        dict_t                                  *dict          = NULL;
-        char                                    *volname       = NULL;
-	gf_boolean_t                             exists        = _gf_false;
-	char					*key	       = NULL;
-	char					 str[100]      = {0, };
-	int					 count	       = 0;
-
-        GF_ASSERT (req);
+        int                                      ret = 0;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        char                                    *volname = NULL;
+        xlator_t                                *this = NULL;
+        glusterd_conf_t                         *priv = NULL;
+        int                                      count = 1;
+        char                                    *key = NULL;
+        char                                    *key_fixed = NULL;
+        char                                    *value = NULL;
+        char                                     str[50] = {0, };
+        char                                    *op_errstr = NULL;
+        gf_boolean_t                             global_opt    = _gf_false;
+        gf_boolean_t                             global_opts_set = _gf_false;
+        glusterd_volinfo_t                      *voliter = NULL;
+        int32_t                                  dict_count = 0;
+        gf_boolean_t                             check_op_version = _gf_false;
+        uint32_t                                 new_op_version = 0;
+        gf_boolean_t                            quorum_action  = _gf_false;
 
-        dict = dict_new ();
-        if (!dict)
-                goto out;
+        this = THIS;
+        GF_ASSERT (this);
 
-	ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
+        priv = this->private;
+        GF_ASSERT (priv);
 
+        ret = dict_get_int32 (dict, "count", &dict_count);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
+                gf_log (this->name, GF_LOG_ERROR, "Count(dict),not set in Volume-Set");
                 goto out;
         }
 
-        ret = dict_get_str (dict, "volname", &volname);
+        if (dict_count == 0) {
+                ret = glusterd_volset_help (NULL, &op_errstr);
+                if (ret) {
+                        op_errstr = (op_errstr)? op_errstr:
+                                     "Volume set help internal error";
+                        gf_log (this->name, GF_LOG_ERROR, "%s", op_errstr);
+                }
+                goto out;
+         }
 
+        ret = dict_get_str (dict, "volname", &volname);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
                 goto out;
         }
 
-        exists = glusterd_check_volume_exists (volname);
+        if (strcasecmp (volname, "all") == 0) {
+                ret = glusterd_op_set_all_volume_options (this, dict);
+                goto out;
+        }
 
-        if (!exists) {
-                gf_log ("", GF_LOG_ERROR, "Volume with name: %s "
-                        "does not exist",
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS,
                         volname);
-                ret = -1;
                 goto out;
         }
-	
-
-	for ( count = 1; ret != -1 ; count++ ) {
-
-		sprintf (str, "key%d", count);
-		ret = dict_get_str (dict, str, &key);
-
-		
-		if (ret) 
-			break;
-		
-		exists = glusterd_check_option_exists(key);
-
-		if (!exists) {
-                	gf_log ("", GF_LOG_ERROR, "Option with name: %s "
-                        	"does not exist", key);
-			ret = -1;
-			goto out;
-        	}
-
 
-	}
-
-	if ( count == 1 ) {
-		gf_log ("", GF_LOG_ERROR, "No options received ");
-		ret = -1;
-		goto out;
-	}
-
-        ret = 0;
-
-out:
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
-
-        return ret;
-}
+        // TODO: Remove this once v3.3 compatability is not required
+        check_op_version = dict_get_str_boolean (dict, "check-op-version",
+                                                 _gf_false);
 
+        if (check_op_version) {
+                ret = dict_get_uint32 (dict, "new-op-version", &new_op_version);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to get new op-version from dict");
+                        goto out;
+                }
+        }
 
-char *set_option_list[] = {
-	"max-file-size",
-	"min-file-size",
-	"cache-timeout",
-	"priority",
-	"entry-change-log",
-	"read-subvolume",
-	"background-self-heal-count",
-	"metadata-self-heal",
-	"data-self-heal",
-	"entry-self-heal",
-	"strict-readdir",
-	"data-self-heal-window-size",
-	"data-change-log",
-	"metadata-change-log",
-	"frame-timeout",
-	"ping-timeout",
-	"cache-size",
-	"disk-usage-limit",
-	"min-free-disk-limit",
-	"block-size",
-	"inode-lru-limit",
-	"thread-count"	,
-	"lookup-unhashed",
-	"min-free-disk",
-        "write-behind",
-        "read-ahead",
-        "io-cache",
-        "quick-read",
-        "stat-prefetch",
-};
+        for (count = 1; ret != -1 ; count++) {
 
+                sprintf (str, "key%d", count);
+                ret = dict_get_str (dict, str, &key);
+                if (ret)
+                        break;
 
-gf_boolean_t
-glusterd_check_option_exists(char *optstring)
-{
-	//struct set_option_list		       *list;
-	char 				      **list = NULL; 
-	
+                sprintf (str, "value%d", count);
+                ret = dict_get_str (dict, str, &value);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "invalid key,value pair in 'volume set'");
+                        ret = -1;
+                        goto out;
+                }
 
-	for (list = &set_option_list[0]; *list ;list++) 
-		if (!strcmp (optstring, *list))
-			return _gf_true;
-	
+                if (strcmp (key, "config.memory-accounting") == 0) {
+                        ret = gf_string2boolean (value,
+                                                 &volinfo->memory_accounting);
+                }
 
-	return _gf_false;
+                if (strcmp (key, "config.transport") == 0) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "changing transport-type for volume %s to %s",
+                                volname, value);
+                        ret = 0;
+                        if (strcasecmp (value, "rdma") == 0) {
+                                volinfo->transport_type = GF_TRANSPORT_RDMA;
+                        } else if (strcasecmp (value, "tcp") == 0) {
+                                volinfo->transport_type = GF_TRANSPORT_TCP;
+                        } else if ((strcasecmp (value, "tcp,rdma") == 0) ||
+                                   (strcasecmp (value, "rdma,tcp") == 0)) {
+                                volinfo->transport_type =
+                                        GF_TRANSPORT_BOTH_TCP_RDMA;
+                        } else {
+                                ret = -1;
+                                goto out;
+                        }
+                }
 
-}
+                if (!is_key_glusterd_hooks_friendly (key)) {
+                        ret = glusterd_check_option_exists (key, &key_fixed);
+                        GF_ASSERT (ret);
+                        if (ret <= 0) {
+                                key_fixed = NULL;
+                                goto out;
+                        }
+                }
 
-static int
-glusterd_op_stage_remove_brick (gd1_mgmt_stage_op_req *req)
-{
-        int                                     ret = -1;
-        dict_t                                  *dict = NULL;
-        char                                    *volname = NULL;
-        glusterd_volinfo_t                      *volinfo = NULL;
-        dict_t                                  *ctx     = NULL;
-        char                                    *errstr  = NULL;
+                global_opt = _gf_false;
+                if (glusterd_check_globaloption (key)) {
+                        global_opt = _gf_true;
+                        global_opts_set = _gf_true;
+                }
 
-        GF_ASSERT (req);
+                if (!global_opt)
+                        value = gf_strdup (value);
 
-        dict = dict_new ();
-        if (!dict)
-                goto out;
+                if (!value) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to set the options in 'volume set'");
+                        ret = -1;
+                        goto out;
+                }
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
+                if (key_fixed)
+                        key = key_fixed;
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
-                goto out;
-        }
+                if (glusterd_is_quorum_changed (volinfo->dict, key, value))
+                        quorum_action = _gf_true;
 
-        ret = dict_get_str (dict, "volname", &volname);
+                if (global_opt) {
+                       list_for_each_entry (voliter, &priv->volumes, vol_list) {
+                               value = gf_strdup (value);
+                               ret = dict_set_dynstr (voliter->dict, key, value);
+                               if (ret)
+                                       goto out;
+                       }
+                } else {
+                        ret = dict_set_dynstr (volinfo->dict, key, value);
+                        if (ret)
+                                goto out;
+                }
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
-                goto out;
+                if (key_fixed) {
+                        GF_FREE (key_fixed);
+                        key_fixed = NULL;
+                }
         }
 
-        ret = glusterd_volinfo_find (volname, &volinfo);
-
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Volume %s does not exist", volname);
+        if (count == 1) {
+                gf_log (this->name, GF_LOG_ERROR, "No options received ");
+                ret = -1;
                 goto out;
         }
 
-        if (volinfo->brick_count == 1) {
-                ctx = glusterd_op_get_ctx (GD_OP_REMOVE_BRICK);
-                if (!ctx) {
-                        gf_log ("", GF_LOG_ERROR,
-                                "Operation Context is not present");
-                        ret = -1;
+        /* Update the cluster op-version before regenerating volfiles so that
+         * correct volfiles are generated
+         */
+        if (new_op_version > priv->op_version) {
+                priv->op_version = new_op_version;
+                ret = glusterd_store_global_info (this);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to store op-version");
                         goto out;
                 }
-                errstr = gf_strdup ("Deleting the last brick of the "
-                                    "volume is not allowed");
-                if (!errstr) {
-                        gf_log ("", GF_LOG_ERROR, "Out of memory");
+        }
+
+        if (!global_opts_set) {
+                gd_update_volume_op_versions (volinfo);
+                ret = glusterd_create_volfiles_and_notify_services (volinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to create volfile for"
+                                " 'volume set'");
                         ret = -1;
                         goto out;
                 }
 
-                ret = dict_set_dynstr (ctx, "errstr", errstr);
-                if (ret) {
-                        GF_FREE (errstr);
-                        gf_log ("", GF_LOG_DEBUG,
-                                "failed to set pump status in ctx");
+                ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+                if (ret)
                         goto out;
+
+                if (GLUSTERD_STATUS_STARTED == volinfo->status) {
+                        ret = glusterd_nodesvcs_handle_reconfigure (volinfo);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                         "Unable to restart NFS-Server");
+                                goto out;
+                        }
                 }
 
-                ret = -1;
-                goto out;
-        }
+        } else {
+                list_for_each_entry (voliter, &priv->volumes, vol_list) {
+                        volinfo = voliter;
+                        gd_update_volume_op_versions (volinfo);
+                        ret = glusterd_create_volfiles_and_notify_services (volinfo);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to create volfile for"
+                                        " 'volume set'");
+                                ret = -1;
+                                goto out;
+                        }
 
-out:
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+                        ret = glusterd_store_volinfo (volinfo,
+                                      GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+                        if (ret)
+                                goto out;
+
+                        if (GLUSTERD_STATUS_STARTED == volinfo->status) {
+                                ret = glusterd_nodesvcs_handle_reconfigure (volinfo);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_WARNING,
+                                                "Unable to restart NFS-Server");
+                                        goto out;
+                                }
+                        }
+                }
+        }
 
+ out:
+        GF_FREE (key_fixed);
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
+        if (quorum_action)
+                glusterd_do_quorum_action ();
         return ret;
 }
 
+
 static int
-glusterd_op_stage_sync_volume (gd1_mgmt_stage_op_req *req, char **op_errstr)
+glusterd_op_sync_volume (dict_t *dict, char **op_errstr,
+                         dict_t *rsp_dict)
 {
         int                                     ret = -1;
-        dict_t                                  *dict = NULL;
         char                                    *volname = NULL;
         char                                    *hostname = NULL;
-        gf_boolean_t                            exists = _gf_false;
-        glusterd_peerinfo_t                     *peerinfo = NULL;
         char                                    msg[2048] = {0,};
+        int                                     count = 1;
+        int                                     vol_count = 0;
+        glusterd_conf_t                         *priv = NULL;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        xlator_t                                *this = NULL;
 
-        GF_ASSERT (req);
-
-        dict = dict_new ();
-        if (!dict)
-                goto out;
-
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
-                goto out;
-        }
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
 
         ret = dict_get_str (dict, "hostname", &hostname);
         if (ret) {
@@ -1240,940 +1921,1433 @@ glusterd_op_stage_sync_volume (gd1_mgmt_stage_op_req *req, char **op_errstr)
                 goto out;
         }
 
-        ret = glusterd_is_local_addr (hostname);
-        if (ret) {
-                ret = glusterd_friend_find (NULL, hostname, &peerinfo);
+        if (!gf_is_local_addr (hostname)) {
+                ret = 0;
+                goto out;
+        }
+
+        //volname is not present in case of sync all
+        ret = dict_get_str (dict, "volname", &volname);
+        if (!ret) {
+                ret = glusterd_volinfo_find (volname, &volinfo);
                 if (ret) {
-                        snprintf (msg, sizeof (msg), "%s, is not a friend",
-                                  hostname);
-                        *op_errstr = gf_strdup (msg);
+                        gf_log ("", GF_LOG_ERROR, "Volume with name: %s "
+                                "not exists", volname);
                         goto out;
                 }
+        }
 
-                if (!peerinfo->connected) {
-                        snprintf (msg, sizeof (msg), "%s, is not connected at "
-                                  "the moment", hostname);
-                        *op_errstr = gf_strdup (msg);
-                        ret = -1;
-                        goto out;
-                }
-        } else {
+        if (!rsp_dict) {
+                //this should happen only on source
+                ret = 0;
+                goto out;
+        }
 
-                //volname is not present in case of sync all
-                ret = dict_get_str (dict, "volname", &volname);
-                if (!ret) {
-                        exists = glusterd_check_volume_exists (volname);
-                        if (!exists) {
-                                snprintf (msg, sizeof (msg), "volume: %s, "
-                                         "doesn't exist", volname);
-                                *op_errstr = gf_strdup (msg);
-                                ret = -1;
+        if (volname) {
+                ret = glusterd_add_volume_to_dict (volinfo, rsp_dict,
+                                                   1);
+                vol_count = 1;
+        } else {
+                list_for_each_entry (volinfo, &priv->volumes, vol_list) {
+                        ret = glusterd_add_volume_to_dict (volinfo,
+                                                           rsp_dict, count);
+                        if (ret)
                                 goto out;
-                        }
-                } else {
-                        ret = 0;
+
+                        vol_count = count++;
                 }
         }
+        ret = dict_set_int32 (rsp_dict, "count", vol_count);
 
 out:
-        if (dict)
-                dict_unref (dict);
         gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
 
         return ret;
 }
 
 static int
-glusterd_op_create_volume (gd1_mgmt_stage_op_req *req, char **op_errstr)
+glusterd_add_profile_volume_options (glusterd_volinfo_t *volinfo)
 {
-        int                   ret        = 0;
-        dict_t               *dict       = NULL;
-        char                 *volname    = NULL;
-        glusterd_conf_t      *priv       = NULL;
-        glusterd_volinfo_t   *volinfo    = NULL;
-        glusterd_brickinfo_t *brickinfo  = NULL;
-        xlator_t             *this       = NULL;
-        char                 *brick      = NULL;
-        int32_t               count      = 0;
-        int32_t               i          = 1;
-        char                 *bricks     = NULL;
-        char                 *brick_list = NULL;
-        char                 *free_ptr   = NULL;
-        char                 *saveptr    = NULL;
-        int32_t               sub_count  = 0;
-        char                 *trans_type = NULL;
-        char                 *str        = NULL;
-
-        GF_ASSERT (req);
-
-        this = THIS;
-        GF_ASSERT (this);
-
-        priv = this->private;
-        GF_ASSERT (priv);
+        int                                     ret = -1;
+        char                                    *latency_key = NULL;
+        char                                    *fd_stats_key = NULL;
 
-        dict = dict_new ();
-        if (!dict)
-                goto out;
+        GF_ASSERT (volinfo);
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
+        latency_key = VKEY_DIAG_LAT_MEASUREMENT;
+        fd_stats_key = VKEY_DIAG_CNT_FOP_HITS;
 
+        ret = dict_set_str (volinfo->dict, latency_key, "on");
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
+                gf_log ("glusterd", GF_LOG_ERROR, "failed to set the volume %s "
+                        "option %s value %s",
+                        volinfo->volname, latency_key, "on");
                 goto out;
         }
 
-        ret = glusterd_volinfo_new (&volinfo);
-
+        ret = dict_set_str (volinfo->dict, fd_stats_key, "on");
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
+                gf_log ("glusterd", GF_LOG_ERROR, "failed to set the volume %s "
+                        "option %s value %s",
+                        volinfo->volname, fd_stats_key, "on");
                 goto out;
         }
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
 
-        ret = dict_get_str (dict, "volname", &volname);
+static void
+glusterd_remove_profile_volume_options (glusterd_volinfo_t *volinfo)
+{
+        char                                    *latency_key = NULL;
+        char                                    *fd_stats_key = NULL;
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
-                goto out;
-        }
+        GF_ASSERT (volinfo);
 
-        strncpy (volinfo->volname, volname, GLUSTERD_MAX_VOLUME_NAME);
-        GF_ASSERT (volinfo->volname);
+        latency_key = VKEY_DIAG_LAT_MEASUREMENT;
+        fd_stats_key = VKEY_DIAG_CNT_FOP_HITS;
+        dict_del (volinfo->dict, latency_key);
+        dict_del (volinfo->dict, fd_stats_key);
+}
 
-        ret = dict_get_int32 (dict, "type", &volinfo->type);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get type");
-                goto out;
-        }
+static int
+glusterd_op_stats_volume (dict_t *dict, char **op_errstr,
+                          dict_t *rsp_dict)
+{
+        int                                     ret = -1;
+        char                                    *volname = NULL;
+        char                                    msg[2048] = {0,};
+        glusterd_volinfo_t                      *volinfo = NULL;
+        int32_t                                 stats_op = GF_CLI_STATS_NONE;
 
-        ret = dict_get_int32 (dict, "count", &volinfo->brick_count);
+        ret = dict_get_str (dict, "volname", &volname);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get count");
+                gf_log ("glusterd", GF_LOG_ERROR, "volume name get failed");
                 goto out;
         }
 
-        ret = dict_get_int32 (dict, "port", &volinfo->port);
+        ret = glusterd_volinfo_find (volname, &volinfo);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get port");
+                snprintf (msg, sizeof (msg), "Volume %s does not exists",
+                          volname);
+
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
                 goto out;
         }
 
-        count = volinfo->brick_count;
-
-        ret = dict_get_str (dict, "bricks", &bricks);
+        ret = dict_get_int32 (dict, "op", &stats_op);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get bricks");
+                gf_log ("glusterd", GF_LOG_ERROR, "volume profile op get failed");
                 goto out;
         }
 
-        if (GF_CLUSTER_TYPE_REPLICATE == volinfo->type) {
-                ret = dict_get_int32 (dict, "replica-count",
-                                      &sub_count);
+        switch (stats_op) {
+        case GF_CLI_STATS_START:
+                ret = glusterd_add_profile_volume_options (volinfo);
                 if (ret)
                         goto out;
-        } else if (GF_CLUSTER_TYPE_STRIPE == volinfo->type) {
-                ret = dict_get_int32 (dict, "stripe-count",
-                                      &sub_count);
-                if (ret)
-                        goto out;
-        }
-
-        ret = dict_get_str (dict, "transport", &trans_type);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get transport");
+                break;
+        case GF_CLI_STATS_STOP:
+                glusterd_remove_profile_volume_options (volinfo);
+                break;
+        case GF_CLI_STATS_INFO:
+        case GF_CLI_STATS_TOP:
+                //info is already collected in brick op.
+                //just goto out;
+                ret = 0;
                 goto out;
-        }
-
-        ret = dict_get_str (dict, "volume-id", &str);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume-id");
+                break;
+        default:
+                GF_ASSERT (0);
+                gf_log ("glusterd", GF_LOG_ERROR, "Invalid profile op: %d",
+                        stats_op);
+                ret = -1;
                 goto out;
+                break;
         }
-        ret = uuid_parse (str, volinfo->volume_id);
+        ret = glusterd_create_volfiles_and_notify_services (volinfo);
+
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "unable to parse uuid %s", str);
+                gf_log ("", GF_LOG_ERROR, "Unable to create volfile for"
+                                          " 'volume set'");
+                ret = -1;
                 goto out;
         }
 
-        if (strcasecmp (trans_type, "rdma") == 0) {
-                volinfo->transport_type = GF_TRANSPORT_RDMA;
-        } else {
-                volinfo->transport_type = GF_TRANSPORT_TCP;
-        }
-        volinfo->sub_count = sub_count;
-
-        if (bricks) {
-                brick_list = gf_strdup (bricks);
-                free_ptr = brick_list;
-        }
-
-        if (count)
-                brick = strtok_r (brick_list+1, " \n", &saveptr);
-
-        while ( i <= count) {
-                ret = glusterd_brickinfo_from_brick (brick, &brickinfo);
-                if (ret)
-                        goto out;
-
-                list_add_tail (&brickinfo->brick_list, &volinfo->bricks);
-                brick = strtok_r (NULL, " \n", &saveptr);
-                i++;
-        }
-        list_add_tail (&volinfo->vol_list, &priv->volumes);
-        volinfo->version++;
-        volinfo->defrag_status = 0;
-
-        ret = glusterd_store_create_volume (volinfo);
-
+        ret = glusterd_store_volinfo (volinfo,
+                                      GLUSTERD_VOLINFO_VER_AC_INCREMENT);
         if (ret)
                 goto out;
 
-        ret = glusterd_create_volfiles (volinfo);
+        if (GLUSTERD_STATUS_STARTED == volinfo->status)
+                ret = glusterd_nodesvcs_handle_reconfigure (volinfo);
 
-        ret = glusterd_volume_compute_cksum (volinfo);
-        if (ret)
-                goto out;
+        ret = 0;
 
 out:
-        if (dict)
-                dict_unref (dict);
-
-        if (free_ptr)
-                GF_FREE(free_ptr);
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
 
         return ret;
 }
 
 static int
-glusterd_op_add_brick (gd1_mgmt_stage_op_req *req, char **op_errstr)
+_add_brick_name_to_dict (dict_t *dict, char *key, glusterd_brickinfo_t *brick)
 {
-        int                                     ret = 0;
-        dict_t                                  *dict = NULL;
-        char                                    *volname = NULL;
-        glusterd_conf_t                         *priv = NULL;
-        glusterd_volinfo_t                      *volinfo = NULL;
-        glusterd_brickinfo_t                    *brickinfo = NULL;
-        xlator_t                                *this = NULL;
-        char                                    *brick = NULL;
-        int32_t                                 count = 0;
-        int32_t                                 i = 1;
-        char                                    *bricks    = NULL;
-        char                                    *brick_list = NULL;
-        char                                    *free_ptr1  = NULL;
-        char                                    *free_ptr2  = NULL;
-        char                                    *saveptr = NULL;
-        gf_boolean_t                            glfs_started = _gf_false;
-        int32_t                                 mybrick = 0;
+        int     ret = -1;
+        char    tmp[1024] = {0,};
+        char    *brickname = NULL;
+        xlator_t *this = NULL;
 
-        GF_ASSERT (req);
+        GF_ASSERT (dict);
+        GF_ASSERT (key);
+        GF_ASSERT (brick);
 
         this = THIS;
         GF_ASSERT (this);
 
-        priv = this->private;
-        GF_ASSERT (priv);
-
-        dict = dict_new ();
-        if (!dict)
-                goto out;
-
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
-
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
+        snprintf (tmp, sizeof (tmp), "%s:%s", brick->hostname, brick->path);
+        brickname = gf_strdup (tmp);
+        if (!brickname) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to dup brick name");
                 goto out;
         }
 
-        ret = dict_get_str (dict, "volname", &volname);
-
+        ret = dict_set_dynstr (dict, key, brickname);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to add brick name to dict");
                 goto out;
         }
+        brickname = NULL;
+out:
+        if (brickname)
+                GF_FREE (brickname);
+        return ret;
+}
 
-        ret = glusterd_volinfo_find (volname, &volinfo);
+static int
+_add_remove_bricks_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo,
+                            char *prefix)
+{
+        int             ret = -1;
+        int             count = 0;
+        int             i = 0;
+        char            brick_key[1024] = {0,};
+        char            dict_key[1024] ={0,};
+        char            *brick = NULL;
+        xlator_t        *this = NULL;
+
+        GF_ASSERT (dict);
+        GF_ASSERT (volinfo);
+        GF_ASSERT (prefix);
 
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = dict_get_int32 (volinfo->rebal.dict, "count", &count);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get brick count");
                 goto out;
         }
 
-
-        ret = dict_get_int32 (dict, "count", &count);
+        snprintf (dict_key, sizeof (dict_key), "%s.count", prefix);
+        ret = dict_set_int32 (dict, dict_key, count);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get count");
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set brick count in dict");
                 goto out;
         }
 
-        volinfo->brick_count += count;
+        for (i = 1; i <= count; i++) {
+                memset (brick_key, 0, sizeof (brick_key));
+                snprintf (brick_key, sizeof (brick_key), "brick%d", i);
 
-        ret = dict_get_str (dict, "bricks", &bricks);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get bricks");
-                goto out;
-        }
+                ret = dict_get_str (volinfo->rebal.dict, brick_key, &brick);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to get %s", brick_key);
+                        goto out;
+                }
 
-        if (bricks) {
-                brick_list = gf_strdup (bricks);
-                free_ptr1 = brick_list;
+                memset (dict_key, 0, sizeof (dict_key));
+                snprintf (dict_key, sizeof (dict_key), "%s.%s", prefix,
+                          brick_key);
+                ret = dict_set_str (dict, dict_key, brick);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to add brick to dict");
+                        goto out;
+                }
+                brick = NULL;
         }
 
-        if (count)
-                brick = strtok_r (brick_list+1, " \n", &saveptr);
+out:
+        return ret;
+}
 
-        while ( i <= count) {
-                ret = glusterd_brickinfo_from_brick (brick, &brickinfo);
-                if (ret)
-                        goto out;
+/* This adds the respective task-id and all available parameters of a task into
+ * a dictionary
+ */
+static int
+_add_task_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, int op, int index)
+{
 
-                list_add_tail (&brickinfo->brick_list, &volinfo->bricks);
-                brick = strtok_r (NULL, " \n", &saveptr);
-                i++;
-        }
+        int             ret = -1;
+        char            key[128] = {0,};
+        char            *uuid_str = NULL;
+        int             status = 0;
+        xlator_t        *this = NULL;
 
-        brick_list = gf_strdup (bricks);
-        free_ptr2 = brick_list;
-        i = 1;
+        GF_ASSERT (dict);
+        GF_ASSERT (volinfo);
 
-        if (count)
-                brick = strtok_r (brick_list+1, " \n", &saveptr);
+        this = THIS;
+        GF_ASSERT (this);
 
-        while (i <= count) {
+        switch (op) {
+        case GD_OP_REMOVE_BRICK:
+                snprintf (key, sizeof (key), "task%d", index);
+                ret = _add_remove_bricks_to_dict (dict, volinfo, key);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to add remove bricks to dict");
+                        goto out;
+                }
+        case GD_OP_REBALANCE:
+                uuid_str = gf_strdup (uuid_utoa (volinfo->rebal.rebalance_id));
+                status = volinfo->rebal.defrag_status;
+                break;
 
-                ret = glusterd_brickinfo_get (brick, volinfo, &brickinfo);
+        case GD_OP_REPLACE_BRICK:
+                snprintf (key, sizeof (key), "task%d.src-brick", index);
+                ret = _add_brick_name_to_dict (dict, key,
+                                               volinfo->rep_brick.src_brick);
                 if (ret)
                         goto out;
+                memset (key, 0, sizeof (key));
 
-                ret = glusterd_resolve_brick (brickinfo);
+                snprintf (key, sizeof (key), "task%d.dst-brick", index);
+                ret = _add_brick_name_to_dict (dict, key,
+                                               volinfo->rep_brick.dst_brick);
+                if (ret)
+                        goto out;
+                memset (key, 0, sizeof (key));
 
-                if (!ret && (!uuid_compare (brickinfo->uuid, priv->uuid)) &&
-                                (GLUSTERD_STATUS_STARTED == volinfo->status)) {
-                        ret = glusterd_create_volfiles (volinfo);
-                        if (ret)
-                                goto out;
+                uuid_str = gf_strdup (uuid_utoa (volinfo->rep_brick.rb_id));
+                status = volinfo->rep_brick.rb_status;
+                break;
 
-                        gf_log ("", GF_LOG_NORMAL, "About to start glusterfs"
-                                " for brick %s:%s", brickinfo->hostname,
-                                brickinfo->path);
-                        ret = glusterd_volume_start_glusterfs
-                                                (volinfo, brickinfo, mybrick);
-                        if (ret) {
-                                gf_log ("", GF_LOG_ERROR, "Unable to start "
-                                        "glusterfs, ret: %d", ret);
-                                goto out;
-                        }
-                        glfs_started = _gf_true;
-                        mybrick++;
-                }
-                i++;
-                brick = strtok_r (NULL, " \n", &saveptr);
+        default:
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR, "%s operation doesn't have a"
+                        " task_id", gd_op_list[op]);
+                goto out;
         }
 
-        if (!glfs_started) {
-                ret = glusterd_create_volfiles (volinfo);
-                if (ret)
-                        goto out;
+        snprintf (key, sizeof (key), "task%d.type", index);
+        ret = dict_set_str (dict, key, (char *)gd_op_list[op]);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Error setting task type in dict");
+                goto out;
         }
 
-        volinfo->version++;
-        volinfo->defrag_status = 0;
-
-        ret = glusterd_store_update_volume (volinfo);
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "task%d.id", index);
 
-        if (ret)
+        if (!uuid_str)
                 goto out;
-
-        ret = glusterd_volume_compute_cksum (volinfo);
-        if (ret)
+        ret = dict_set_dynstr (dict, key, uuid_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Error setting task id in dict");
                 goto out;
+        }
+        uuid_str = NULL;
 
-        if (GLUSTERD_STATUS_STARTED == volinfo->status)
-                ret = glusterd_check_generate_start_nfs (volinfo);
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "task%d.status", index);
+        ret = dict_set_int32 (dict, key, status);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Error setting task status in dict");
+                goto out;
+        }
 
 out:
-        if (dict)
-                dict_unref (dict);
-        if (free_ptr1)
-                GF_FREE (free_ptr1);
-        if (free_ptr2)
-                GF_FREE (free_ptr2);
+        if (uuid_str)
+                GF_FREE (uuid_str);
         return ret;
 }
 
 static int
-rb_regenerate_volfiles (glusterd_volinfo_t *volinfo, int32_t pump_needed)
+glusterd_aggregate_task_status (dict_t *rsp_dict, glusterd_volinfo_t *volinfo)
 {
-        dict_t *dict = NULL;
-        int ret = 0;
+        int        ret   = -1;
+        int        tasks = 0;
+        xlator_t  *this  = NULL;
 
-        dict = volinfo->dict;
+        this = THIS;
+        GF_ASSERT (this);
 
-        gf_log ("", GF_LOG_DEBUG,
-                "attempting to set pump value=%d", pump_needed);
+        if (!uuid_is_null (volinfo->rebal.rebalance_id)) {
+                ret = _add_task_to_dict (rsp_dict, volinfo, volinfo->rebal.op,
+                                         tasks);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to add task details to dict");
+                        goto out;
+                }
+                tasks++;
+        }
 
-        ret = dict_set_int32 (dict, "enable-pump", pump_needed);
+        if (!uuid_is_null (volinfo->rep_brick.rb_id)) {
+                ret = _add_task_to_dict (rsp_dict, volinfo, GD_OP_REPLACE_BRICK,
+                                         tasks);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to add task details to dict");
+                        goto out;
+                }
+                tasks++;
+        }
+
+        ret = dict_set_int32 (rsp_dict, "tasks", tasks);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "could not dict_set enable-pump");
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Error setting tasks count in dict");
                 goto out;
         }
-
-        ret = glusterd_create_volfiles (volinfo);
+        ret = 0;
 
 out:
         return ret;
 }
 
 static int
-rb_src_brick_restart (glusterd_volinfo_t *volinfo,
-                      glusterd_brickinfo_t *src_brickinfo,
-                      int activate_pump)
+glusterd_op_status_volume (dict_t *dict, char **op_errstr,
+                           dict_t *rsp_dict)
 {
-        int ret = 0;
+        int                     ret             = -1;
+        int                     node_count      = 0;
+        int                     brick_index     = -1;
+        int                     other_count     = 0;
+        int                     other_index     = 0;
+        uint32_t                cmd             = 0;
+        char                   *volname         = NULL;
+        char                   *brick           = NULL;
+        xlator_t               *this            = NULL;
+        glusterd_volinfo_t     *volinfo         = NULL;
+        glusterd_brickinfo_t   *brickinfo       = NULL;
+        glusterd_conf_t        *priv            = NULL;
+        dict_t                 *vol_opts        = NULL;
+        gf_boolean_t            nfs_disabled    = _gf_false;
+        gf_boolean_t            shd_enabled     = _gf_true;
+        gf_boolean_t            origin_glusterd = _gf_false;
 
-        gf_log ("", GF_LOG_DEBUG,
-                "Attempting to kill src");
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
 
-        ret = glusterd_volume_stop_glusterfs
-                (volinfo, src_brickinfo, 0);
+        GF_ASSERT (priv);
+
+        GF_ASSERT (dict);
+
+        origin_glusterd = is_origin_glusterd (dict);
+
+        ret = dict_get_uint32 (dict, "cmd", &cmd);
+        if (ret)
+                goto out;
+
+        if (origin_glusterd) {
+                ret = 0;
+                if ((cmd & GF_CLI_STATUS_ALL)) {
+                        ret = glusterd_get_all_volnames (rsp_dict);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to get all volume "
+                                        "names for status");
+                }
+        }
+
+        ret = dict_set_uint32 (rsp_dict, "cmd", cmd);
+        if (ret)
+                goto out;
+
+        if (cmd & GF_CLI_STATUS_ALL)
+                goto out;
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret)
+                goto out;
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to stop "
-                        "glusterfs, ret: %d", ret);
+                gf_log (this->name, GF_LOG_ERROR, "Volume with name: %s "
+                        "does not exist", volname);
                 goto out;
         }
+        vol_opts = volinfo->dict;
 
-        glusterd_delete_volfile (volinfo, src_brickinfo);
-        glusterd_store_delete_brick (volinfo, src_brickinfo);
+        if ((cmd & GF_CLI_STATUS_NFS) != 0) {
+                ret = glusterd_add_node_to_dict ("nfs", rsp_dict, 0, vol_opts);
+                if (ret)
+                        goto out;
+                other_count++;
+                node_count++;
 
-        if (activate_pump) {
-                ret = rb_regenerate_volfiles (volinfo, 1);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Could not regenerate volfiles with pump");
+        } else if ((cmd & GF_CLI_STATUS_SHD) != 0) {
+                ret = glusterd_add_node_to_dict ("glustershd", rsp_dict, 0,
+                                                 vol_opts);
+                if (ret)
                         goto out;
-                }
-        } else {
-                ret = rb_regenerate_volfiles (volinfo, 0);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Could not regenerate volfiles without pump");
+                other_count++;
+                node_count++;
+
+        } else if ((cmd & GF_CLI_STATUS_BRICK) != 0) {
+                ret = dict_get_str (dict, "brick", &brick);
+                if (ret)
                         goto out;
-                }
 
-        }
+                ret = glusterd_volume_brickinfo_get_by_brick (brick,
+                                                              volinfo,
+                                                              &brickinfo);
+                if (ret)
+                        goto out;
 
-        ret = glusterd_volume_start_glusterfs
-                (volinfo, src_brickinfo, 0);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to start "
-                        "glusterfs, ret: %d", ret);
+                if (uuid_compare (brickinfo->uuid, MY_UUID))
+                        goto out;
+
+                glusterd_add_brick_to_dict (volinfo, brickinfo, rsp_dict,
+                                            ++brick_index);
+                if (cmd & GF_CLI_STATUS_DETAIL)
+                        glusterd_add_brick_detail_to_dict (volinfo, brickinfo,
+                                                           rsp_dict,
+                                                           brick_index);
+                node_count++;
+
+        } else if ((cmd & GF_CLI_STATUS_TASKS) != 0) {
+                ret = glusterd_aggregate_task_status (rsp_dict, volinfo);
                 goto out;
-        }
 
+        } else {
+                list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                        brick_index++;
+                        if (uuid_compare (brickinfo->uuid, MY_UUID))
+                                continue;
 
-out:
-        return ret;
-}
+                        glusterd_add_brick_to_dict (volinfo, brickinfo,
+                                                    rsp_dict, brick_index);
 
-static int
-rb_send_xattr_command (glusterd_volinfo_t *volinfo,
-                       glusterd_brickinfo_t *src_brickinfo,
-                       glusterd_brickinfo_t *dst_brickinfo,
-                       const char *xattr_key,
-                       const char *value)
-{
-        glusterd_conf_t  *priv                        = NULL;
-        char              mount_point_path[PATH_MAX]  = {0,};
-        struct stat       buf;
-        int               ret                         = -1;
+                        if (cmd & GF_CLI_STATUS_DETAIL) {
+                                glusterd_add_brick_detail_to_dict (volinfo,
+                                                                   brickinfo,
+                                                                   rsp_dict,
+                                                                   brick_index);
+                        }
+                        node_count++;
+                }
 
-        priv = THIS->private;
+                if ((cmd & GF_CLI_STATUS_MASK) == GF_CLI_STATUS_NONE) {
+                        other_index = brick_index + 1;
+
+                        nfs_disabled = dict_get_str_boolean (vol_opts,
+                                                             "nfs.disable",
+                                                             _gf_false);
+                        if (!nfs_disabled) {
+                                ret = glusterd_add_node_to_dict ("nfs",
+                                                                 rsp_dict,
+                                                                 other_index,
+                                                                 vol_opts);
+                                if (ret)
+                                        goto out;
+                                other_index++;
+                                other_count++;
+                                node_count++;
+                        }
 
-        snprintf (mount_point_path, PATH_MAX, "%s/vols/%s/%s",
-                  priv->workdir, volinfo->volname,
-                  RB_CLIENT_MOUNTPOINT);
+                        shd_enabled = dict_get_str_boolean
+                                        (vol_opts, "cluster.self-heal-daemon",
+                                         _gf_true);
+                        if (glusterd_is_volume_replicate (volinfo)
+                            && shd_enabled) {
+                                ret = glusterd_add_node_to_dict ("glustershd",
+                                                                 rsp_dict,
+                                                                 other_index,
+                                                                 vol_opts);
+                                if (ret)
+                                        goto out;
+                                other_count++;
+                                node_count++;
+                        }
+                }
+        }
 
-        ret = stat (mount_point_path, &buf);
+        ret = dict_set_int32 (rsp_dict, "brick-index-max", brick_index);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "stat failed. Could not send "
-                        " %s command", xattr_key);
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Error setting brick-index-max to dict");
                 goto out;
         }
-
-        ret = lsetxattr (mount_point_path, xattr_key,
-                         value,
-                         strlen (value),
-                         0);
-
+        ret = dict_set_int32 (rsp_dict, "other-count", other_count);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "setxattr failed");
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Error setting other-count to dict");
                 goto out;
         }
+        ret = dict_set_int32 (rsp_dict, "count", node_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Error setting node count to dict");
+                goto out;
+        }
+
+        /* Active tasks */
+        /* Tasks are added only for normal volume status request for either a
+         * single volume or all volumes, and only by the origin glusterd
+         */
+        if (((cmd & GF_CLI_STATUS_MASK) != GF_CLI_STATUS_NONE) ||
+            !(cmd & (GF_CLI_STATUS_VOL | GF_CLI_STATUS_ALL)) ||
+            !origin_glusterd)
+                goto out;
 
+        ret = glusterd_aggregate_task_status (rsp_dict, volinfo);
+        if (ret)
+                goto out;
         ret = 0;
 
 out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
         return ret;
 }
 
 static int
-rb_spawn_dst_brick (glusterd_volinfo_t *volinfo,
-                    glusterd_brickinfo_t *brickinfo)
+glusterd_op_ac_none (glusterd_op_sm_event_t *event, void *ctx)
 {
-        glusterd_conf_t    *priv           = NULL;
-        char                cmd_str[8192]  = {0,};
-        int                 ret            = -1;
+        int ret = 0;
 
-        priv = THIS->private;
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret);
 
-        snprintf (cmd_str, 8192,
-                  "%s/sbin/glusterfs -f %s/vols/%s/%s -p %s/vols/%s/%s",
-                  GFS_PREFIX, priv->workdir, volinfo->volname,
-                  RB_DSTBRICKVOL_FILENAME,
-                  priv->workdir, volinfo->volname,
-                  RB_DSTBRICK_PIDFILE);
+        return ret;
+}
 
-        ret = gf_system (cmd_str);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not start glusterfs");
-                goto out;
-        }
+static int
+glusterd_op_ac_send_lock (glusterd_op_sm_event_t *event, void *ctx)
+{
+        int                   ret      = 0;
+        rpc_clnt_procedure_t *proc     = NULL;
+        glusterd_conf_t      *priv     = NULL;
+        xlator_t             *this     = NULL;
+        glusterd_peerinfo_t  *peerinfo = NULL;
+        uint32_t             pending_count = 0;
+        dict_t               *dict     = NULL;
 
-        gf_log ("", GF_LOG_DEBUG,
-                "Successfully started glusterfs: brick=%s:%s",
-                brickinfo->hostname, brickinfo->path);
+        this = THIS;
+        priv = this->private;
+        GF_ASSERT (priv);
 
-        ret = 0;
+        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
+                GF_ASSERT (peerinfo);
+
+                if (!peerinfo->connected || !peerinfo->mgmt)
+                        continue;
+                if ((peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED) &&
+                    (glusterd_op_get_op() != GD_OP_SYNC_VOLUME))
+                        continue;
+
+                /* Based on the op_version, acquire a cluster or mgmt_v3 lock */
+                if (priv->op_version < 3) {
+                        proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_LOCK];
+                        if (proc->fn) {
+                                ret = proc->fn (NULL, this, peerinfo);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_WARNING,
+                                                "Failed to send lock request "
+                                                "for operation 'Volume %s' to "
+                                                "peer %s",
+                                                gd_op_list[opinfo.op],
+                                                peerinfo->hostname);
+                                        continue;
+                                }
+                                pending_count++;
+                        }
+                } else {
+                        dict = glusterd_op_get_ctx ();
+                        dict_ref (dict);
+
+                        proc = &peerinfo->mgmt_v3->proctable
+                                                   [GLUSTERD_MGMT_V3_LOCK];
+                        if (proc->fn) {
+                                ret = dict_set_static_ptr (dict, "peerinfo",
+                                                           peerinfo);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "failed to set peerinfo");
+                                        dict_unref (dict);
+                                        goto out;
+                                }
+
+                                ret = proc->fn (NULL, this, dict);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_WARNING,
+                                                "Failed to send mgmt_v3 lock "
+                                                "request for operation "
+                                                "'Volume %s' to peer %s",
+                                                gd_op_list[opinfo.op],
+                                                peerinfo->hostname);
+                                        dict_unref (dict);
+                                        continue;
+                                }
+                                pending_count++;
+                        }
+                }
+        }
+
+        opinfo.pending_count = pending_count;
+        if (!opinfo.pending_count)
+                ret = glusterd_op_sm_inject_all_acc (&event->txn_id);
 
 out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
         return ret;
 }
 
 static int
-rb_spawn_glusterfs_client (glusterd_volinfo_t *volinfo,
-                           glusterd_brickinfo_t *brickinfo)
+glusterd_op_ac_send_unlock (glusterd_op_sm_event_t *event, void *ctx)
 {
-        glusterd_conf_t    *priv           = NULL;
-        char                cmd_str[8192]  = {0,};
-        struct stat         buf;
-        int                 ret            = -1;
+        int                   ret      = 0;
+        rpc_clnt_procedure_t *proc     = NULL;
+        glusterd_conf_t      *priv     = NULL;
+        xlator_t             *this     = NULL;
+        glusterd_peerinfo_t  *peerinfo = NULL;
+        uint32_t             pending_count = 0;
+        dict_t               *dict     = NULL;
 
-        priv = THIS->private;
+        this = THIS;
+        priv = this->private;
+        GF_ASSERT (priv);
 
-        snprintf (cmd_str, 4096,
-                  "%s/sbin/glusterfs -f %s/vols/%s/%s %s/vols/%s/%s",
-                  GFS_PREFIX, priv->workdir, volinfo->volname,
-                  RB_CLIENTVOL_FILENAME,
-                  priv->workdir, volinfo->volname,
-                  RB_CLIENT_MOUNTPOINT);
+        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
+                GF_ASSERT (peerinfo);
 
-        ret = gf_system (cmd_str);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not start glusterfs");
-                goto out;
+                if (!peerinfo->connected || !peerinfo->mgmt)
+                        continue;
+                if ((peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED) &&
+                    (glusterd_op_get_op() != GD_OP_SYNC_VOLUME))
+                        continue;
+
+                /* Based on the op_version, release the *
+                 * cluster or mgmt_v3 lock */
+                if (priv->op_version < 3) {
+                        proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_UNLOCK];
+                        if (proc->fn) {
+                                ret = proc->fn (NULL, this, peerinfo);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_WARNING,
+                                                "Failed to send unlock request "
+                                                "for operation 'Volume %s' to "
+                                                "peer %s",
+                                                gd_op_list[opinfo.op],
+                                                peerinfo->hostname);
+                                        continue;
+                                }
+                                pending_count++;
+                        }
+                } else {
+                        dict = glusterd_op_get_ctx ();
+                        dict_ref (dict);
+
+                        proc = &peerinfo->mgmt_v3->proctable
+                                                   [GLUSTERD_MGMT_V3_UNLOCK];
+                        if (proc->fn) {
+                                ret = dict_set_static_ptr (dict, "peerinfo",
+                                                           peerinfo);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "failed to set peerinfo");
+                                        dict_unref (dict);
+                                        goto out;
+                                }
+
+                                ret = proc->fn (NULL, this, dict);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_WARNING,
+                                                "Failed to send mgmt_v3 unlock "
+                                                "request for operation "
+                                                "'Volume %s' to peer %s",
+                                                gd_op_list[opinfo.op],
+                                                peerinfo->hostname);
+                                        dict_unref (dict);
+                                        continue;
+                                }
+                                pending_count++;
+                        }
+                }
         }
 
-        gf_log ("", GF_LOG_DEBUG,
-                "Successfully started glusterfs: brick=%s:%s",
-                brickinfo->hostname, brickinfo->path);
+        opinfo.pending_count = pending_count;
+        if (!opinfo.pending_count)
+                ret = glusterd_op_sm_inject_all_acc (&event->txn_id);
 
-        memset (cmd_str, 0, sizeof (cmd_str));
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
 
-        snprintf (cmd_str, 4096, "%s/vols/%s/%s",
-                  priv->workdir, volinfo->volname,
-                  RB_CLIENT_MOUNTPOINT);
+static int
+glusterd_op_ac_ack_drain (glusterd_op_sm_event_t *event, void *ctx)
+{
+        int ret = 0;
 
-        ret = stat (cmd_str, &buf);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "stat on mountpoint failed");
-                goto out;
-        }
+        if (opinfo.pending_count > 0)
+                opinfo.pending_count--;
 
-        gf_log ("", GF_LOG_DEBUG,
-                "stat on mountpoint succeeded");
+        if (!opinfo.pending_count)
+                ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK,
+                                                   &event->txn_id, NULL);
 
-        ret = 0;
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret);
 
-out:
         return ret;
 }
 
-static const char *client_volfile_str =  "volume mnt-client\n"
-        " type protocol/client\n"
-        " option remote-host %s\n"
-        " option remote-subvolume %s\n"
-        " option remote-port %d\n"
-        "end-volume\n"
-        "volume mnt-wb\n"
-        " type performance/write-behind\n"
-        " subvolumes mnt-client\n"
-        "end-volume\n";
-
 static int
-rb_generate_client_volfile (glusterd_volinfo_t *volinfo,
-                            glusterd_brickinfo_t *src_brickinfo)
+glusterd_op_ac_send_unlock_drain (glusterd_op_sm_event_t *event, void *ctx)
 {
-        glusterd_conf_t    *priv = NULL;
-        FILE *file = NULL;
-        char filename[PATH_MAX];
-        int ret = -1;
-
-        priv = THIS->private;
+        return glusterd_op_ac_ack_drain (event, ctx);
+}
 
-        gf_log ("", GF_LOG_DEBUG,
-                "Creating volfile");
+static int
+glusterd_op_ac_lock (glusterd_op_sm_event_t *event, void *ctx)
+{
+        int32_t                         ret             = 0;
+        char                           *volname         = NULL;
+        glusterd_op_lock_ctx_t         *lock_ctx        = NULL;
+        glusterd_conf_t                *priv            = NULL;
+        xlator_t                       *this            = NULL;
 
-        snprintf (filename, PATH_MAX, "%s/vols/%s/%s",
-                  priv->workdir, volinfo->volname,
-                  RB_CLIENTVOL_FILENAME);
+        GF_ASSERT (event);
+        GF_ASSERT (ctx);
 
-        file = fopen (filename, "w+");
-        if (!file) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Open of volfile failed");
-                ret = -1;
-                goto out;
-        }
+        this = THIS;
+        priv = this->private;
 
-        GF_ASSERT (src_brickinfo->port);
+        lock_ctx = (glusterd_op_lock_ctx_t *)ctx;
 
-        fprintf (file, client_volfile_str, src_brickinfo->hostname,
-                 src_brickinfo->path, src_brickinfo->port);
+        /* If the req came from a node running on older op_version
+         * the dict won't be present. Based on it acquiring a cluster
+         * or mgmt_v3 lock */
+        if (lock_ctx->dict == NULL) {
+                ret = glusterd_lock (lock_ctx->uuid);
+                glusterd_op_lock_send_resp (lock_ctx->req, ret);
+        } else {
+                ret = dict_get_str (lock_ctx->dict, "volname", &volname);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to acquire volname");
+                else {
+                        ret = glusterd_mgmt_v3_lock (volname, lock_ctx->uuid,
+                                                     "vol");
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to acquire lock for %s",
+                                        volname);
+                }
 
-        fclose (file);
+                glusterd_op_mgmt_v3_lock_send_resp (lock_ctx->req,
+                                                    &event->txn_id, ret);
 
-        ret = 0;
+                dict_unref (lock_ctx->dict);
+        }
 
-out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Lock Returned %d", ret);
         return ret;
 }
 
-static const char *dst_brick_volfile_str = "volume src-posix\n"
-        "type storage/posix\n"
-        "option directory %s\n"
-        "end-volume\n"
-        "volume %s\n"
-        "type features/locks\n"
-        "subvolumes src-posix\n"
-        "end-volume\n"
-        "volume src-server\n"
-        "type protocol/server\n"
-        "option auth.addr.%s.allow *\n"
-        "option listen-port 34034\n"
-        "option transport-type tcp\n"
-        "subvolumes %s\n"
-        "end-volume\n";
-
 static int
-rb_generate_dst_brick_volfile (glusterd_volinfo_t *volinfo,
-                               glusterd_brickinfo_t *dst_brickinfo)
+glusterd_op_ac_unlock (glusterd_op_sm_event_t *event, void *ctx)
 {
-        glusterd_conf_t    *priv = NULL;
-        FILE *file = NULL;
-        char filename[PATH_MAX];
-        int ret = -1;
+        int32_t                         ret             = 0;
+        char                           *volname         = NULL;
+        glusterd_op_lock_ctx_t         *lock_ctx        = NULL;
+        glusterd_conf_t                *priv            = NULL;
+        xlator_t                       *this            = NULL;
 
-        priv = THIS->private;
 
-        gf_log ("", GF_LOG_DEBUG,
-                "Creating volfile");
+        GF_ASSERT (event);
+        GF_ASSERT (ctx);
 
-        snprintf (filename, PATH_MAX, "%s/vols/%s/%s",
-                  priv->workdir, volinfo->volname,
-                  RB_DSTBRICKVOL_FILENAME);
+        this = THIS;
+        priv = this->private;
 
-        file = fopen (filename, "w+");
-        if (!file) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Open of volfile failed");
-                ret = -1;
-                goto out;
-        }
+        lock_ctx = (glusterd_op_lock_ctx_t *)ctx;
 
-        fprintf (file, dst_brick_volfile_str, dst_brickinfo->path,
-                 dst_brickinfo->path, dst_brickinfo->path,
-                 dst_brickinfo->path);
+        /* If the req came from a node running on older op_version
+         * the dict won't be present. Based on it releasing the cluster
+         * or mgmt_v3 lock */
+        if (lock_ctx->dict == NULL) {
+                ret = glusterd_unlock (lock_ctx->uuid);
+                glusterd_op_unlock_send_resp (lock_ctx->req, ret);
+        } else {
+                ret = dict_get_str (lock_ctx->dict, "volname", &volname);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to acquire volname");
+                else {
+                        ret = glusterd_mgmt_v3_unlock (volname, lock_ctx->uuid,
+                                                       "vol");
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to release lock for %s", volname);
+                }
 
-        fclose (file);
+                glusterd_op_mgmt_v3_unlock_send_resp (lock_ctx->req,
+                                                      &event->txn_id, ret);
 
-        ret = 0;
+                dict_unref (lock_ctx->dict);
+        }
 
-out:
+        gf_log (this->name, GF_LOG_DEBUG, "Unlock Returned %d", ret);
+
+        if (priv->pending_quorum_action)
+                glusterd_do_quorum_action ();
         return ret;
 }
 
 static int
-rb_mountpoint_mkdir (glusterd_volinfo_t *volinfo,
-                     glusterd_brickinfo_t *src_brickinfo)
+glusterd_op_ac_local_unlock (glusterd_op_sm_event_t *event, void *ctx)
 {
-        glusterd_conf_t *priv                       = NULL;
-        char             mount_point_path[PATH_MAX] = {0,};
-        int              ret                        = -1;
+        int              ret          = 0;
+        uuid_t          *originator   = NULL;
 
-        priv = THIS->private;
+        GF_ASSERT (event);
+        GF_ASSERT (ctx);
 
-        snprintf (mount_point_path, PATH_MAX, "%s/vols/%s/%s",
-                  priv->workdir, volinfo->volname,
-                  RB_CLIENT_MOUNTPOINT);
+        originator = (uuid_t *) ctx;
 
-        ret = mkdir (mount_point_path, 0777);
-        if (ret && (errno != EEXIST)) {
-                gf_log ("", GF_LOG_DEBUG, "mkdir failed, errno: %d", errno);
-                goto out;
-        }
+        ret = glusterd_unlock (*originator);
 
-        ret = 0;
+        gf_log (THIS->name, GF_LOG_DEBUG, "Unlock Returned %d", ret);
 
-out:
         return ret;
 }
 
 static int
-rb_mountpoint_rmdir (glusterd_volinfo_t *volinfo,
-                     glusterd_brickinfo_t *src_brickinfo)
+glusterd_op_ac_rcvd_lock_acc (glusterd_op_sm_event_t *event, void *ctx)
 {
-        glusterd_conf_t *priv                       = NULL;
-        char             mount_point_path[PATH_MAX] = {0,};
-        int              ret                        = -1;
+        int                     ret = 0;
 
-        priv = THIS->private;
+        GF_ASSERT (event);
 
-        snprintf (mount_point_path, PATH_MAX, "%s/vols/%s/%s",
-                  priv->workdir, volinfo->volname,
-                  RB_CLIENT_MOUNTPOINT);
+        if (opinfo.pending_count > 0)
+                opinfo.pending_count--;
 
-        ret = rmdir (mount_point_path);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG, "rmdir failed");
+        if (opinfo.pending_count > 0)
                 goto out;
-        }
 
-        ret = 0;
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC,
+                                           &event->txn_id, NULL);
+
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
 
 out:
         return ret;
 }
 
 static int
-rb_destroy_maintainence_client (glusterd_volinfo_t *volinfo,
-                                glusterd_brickinfo_t *src_brickinfo)
+glusterd_dict_set_volid (dict_t *dict, char *volname, char **op_errstr)
 {
-        glusterd_conf_t  *priv                        = NULL;
-        char              cmd_str[8192]               = {0,};
-        char              filename[PATH_MAX]          = {0,};
-        struct stat       buf;
-        char              mount_point_path[PATH_MAX]  = {0,};
-        int               ret                         = -1;
+        int                     ret = -1;
+        glusterd_volinfo_t      *volinfo = NULL;
+        char                    *volid = NULL;
+        char                    msg[1024] = {0,};
+        xlator_t                *this = NULL;
 
-        priv = THIS->private;
+        this = THIS;
+        GF_ASSERT (this);
 
-        snprintf (mount_point_path, PATH_MAX, "%s/vols/%s/%s",
-                  priv->workdir, volinfo->volname,
-                  RB_CLIENT_MOUNTPOINT);
+        if (!dict || !volname)
+                goto out;
 
-        ret = stat (mount_point_path, &buf);
+        ret = glusterd_volinfo_find (volname, &volinfo);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "stat failed. Cannot destroy maintainence "
-                        "client");
+                snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname);
                 goto out;
         }
-
-        snprintf (cmd_str, 8192, "/bin/umount -f %s/vols/%s/%s",
-                  priv->workdir, volinfo->volname,
-                  RB_CLIENT_MOUNTPOINT);
-
-        ret = gf_system (cmd_str);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "umount failed on maintainence client");
+        volid = gf_strdup (uuid_utoa (volinfo->volume_id));
+        if (!volid) {
+                ret = -1;
                 goto out;
         }
-
-        ret = rb_mountpoint_rmdir (volinfo, src_brickinfo);
+        ret = dict_set_dynstr (dict, "vol-id", volid);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "rmdir of mountpoint failed");
+                snprintf (msg, sizeof (msg), "Failed to set volume id of volume"
+                          " %s", volname);
                 goto out;
         }
+out:
+        if (msg[0] != '\0') {
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+        }
+        return ret;
+}
 
-        snprintf (filename, PATH_MAX, "%s/vols/%s/%s",
-                  priv->workdir, volinfo->volname,
-                  RB_CLIENTVOL_FILENAME);
+int
+glusterd_op_build_payload (dict_t **req, char **op_errstr, dict_t *op_ctx)
+{
+        int                     ret = -1;
+        void                    *ctx = NULL;
+        dict_t                  *dict = NULL;
+        dict_t                  *req_dict = NULL;
+        glusterd_op_t           op = GD_OP_NONE;
+        char                    *volname = NULL;
+        uint32_t                status_cmd = GF_CLI_STATUS_NONE;
+        char                    *errstr = NULL;
+        xlator_t                *this = NULL;
 
-        ret = unlink (filename);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "unlink failed");
+        GF_ASSERT (req);
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        req_dict = dict_new ();
+        if (!req_dict)
                 goto out;
+
+        if (!op_ctx) {
+                op  = glusterd_op_get_op ();
+                ctx = (void*)glusterd_op_get_ctx ();
+                if (!ctx) {
+                        gf_log (this->name, GF_LOG_ERROR, "Null Context for "
+                                "op %d", op);
+                        ret = -1;
+                        goto out;
+                }
+
+        } else {
+#define GD_SYNC_OPCODE_KEY "sync-mgmt-operation"
+                ret = dict_get_int32 (op_ctx, GD_SYNC_OPCODE_KEY, (int32_t*)&op);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get volume"
+                                " operation");
+                        goto out;
+                }
+                ctx = op_ctx;
+#undef GD_SYNC_OPCODE_KEY
         }
 
+        dict = ctx;
+        switch (op) {
+                case GD_OP_CREATE_VOLUME:
+                        {
+                                ++glusterfs_port;
+                                ret = dict_set_int32 (dict, "port",
+                                                      glusterfs_port);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Failed to set port in "
+                                                "dictionary");
+                                        goto out;
+                                }
+                                dict_copy (dict, req_dict);
+                        }
+                        break;
+
+                case GD_OP_GSYNC_CREATE:
+                case GD_OP_GSYNC_SET:
+                        {
+                                ret = glusterd_op_gsync_args_get (dict,
+                                                                  &errstr,
+                                                                  &volname,
+                                                                  NULL, NULL);
+                                if (ret == 0) {
+                                        ret = glusterd_dict_set_volid
+                                                (dict, volname, op_errstr);
+                                        if (ret)
+                                                goto out;
+                                }
+                                dict_copy (dict, req_dict);
+                        }
+                        break;
+
+                case GD_OP_SET_VOLUME:
+                        {
+                                ret = dict_get_str (dict, "volname", &volname);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_CRITICAL,
+                                                "volname is not present in "
+                                                "operation ctx");
+                                        goto out;
+                                }
+                                if (strcmp (volname, "help") &&
+                                    strcmp (volname, "help-xml") &&
+                                    strcasecmp (volname, "all")) {
+                                        ret = glusterd_dict_set_volid
+                                                 (dict, volname, op_errstr);
+                                        if (ret)
+                                                goto out;
+                                }
+                                dict_destroy (req_dict);
+                                req_dict = dict_ref (dict);
+                        }
+                        break;
+
+                case GD_OP_SYNC_VOLUME:
+                        {
+                                dict_copy (dict, req_dict);
+                                break;
+                        }
+
+                case GD_OP_REMOVE_BRICK:
+                        {
+                                dict_t *dict = ctx;
+                                ret = dict_get_str (dict, "volname", &volname);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_CRITICAL,
+                                                "volname is not present in "
+                                                "operation ctx");
+                                        goto out;
+                                }
+
+                                ret = glusterd_dict_set_volid (dict, volname,
+                                                               op_errstr);
+                                if (ret)
+                                        goto out;
+
+                                dict_destroy (req_dict);
+                                req_dict = dict_ref (dict);
+                        }
+                        break;
+
+                case GD_OP_STATUS_VOLUME:
+                        {
+                                ret = dict_get_uint32 (dict, "cmd",
+                                                       &status_cmd);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Status command not present "
+                                                "in op ctx");
+                                        goto out;
+                                }
+                                if (GF_CLI_STATUS_ALL & status_cmd) {
+                                        dict_copy (dict, req_dict);
+                                        break;
+                                }
+                        }
+                        /*fall-through*/
+                case GD_OP_DELETE_VOLUME:
+                case GD_OP_START_VOLUME:
+                case GD_OP_STOP_VOLUME:
+                case GD_OP_ADD_BRICK:
+                case GD_OP_REPLACE_BRICK:
+                case GD_OP_RESET_VOLUME:
+                case GD_OP_LOG_ROTATE:
+                case GD_OP_QUOTA:
+                case GD_OP_PROFILE_VOLUME:
+                case GD_OP_REBALANCE:
+                case GD_OP_HEAL_VOLUME:
+                case GD_OP_STATEDUMP_VOLUME:
+                case GD_OP_CLEARLOCKS_VOLUME:
+                case GD_OP_DEFRAG_BRICK_VOLUME:
+                        {
+                                ret = dict_get_str (dict, "volname", &volname);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_CRITICAL,
+                                                "volname is not present in "
+                                                "operation ctx");
+                                        goto out;
+                                }
+
+                                if (strcasecmp (volname, "all")) {
+                                        ret = glusterd_dict_set_volid (dict,
+                                                                       volname,
+                                                                     op_errstr);
+                                        if (ret)
+                                                goto out;
+                                }
+                                dict_copy (dict, req_dict);
+                        }
+                        break;
+
+                case GD_OP_COPY_FILE:
+                        {
+                                dict_copy (dict, req_dict);
+                                break;
+                        }
+
+                case GD_OP_SYS_EXEC:
+                        {
+                                dict_copy (dict, req_dict);
+                                break;
+                        }
+
+                default:
+                        break;
+        }
+
+        *req = req_dict;
         ret = 0;
 
 out:
         return ret;
 }
 
-static int
-rb_spawn_maintainence_client (glusterd_volinfo_t *volinfo,
-                              glusterd_brickinfo_t *src_brickinfo)
+gf_boolean_t
+glusterd_is_get_op (xlator_t *this, glusterd_op_t op, dict_t *dict)
 {
-        int ret = -1;
+        char            *key = NULL;
+        char            *volname = NULL;
+        int             ret = 0;
 
-        ret = rb_generate_client_volfile (volinfo, src_brickinfo);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG, "Unable to generate client "
-                        "volfile");
-                goto out;
-        }
+        if (op == GD_OP_STATUS_VOLUME)
+                return _gf_true;
 
-        ret = rb_mountpoint_mkdir (volinfo, src_brickinfo);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG, "Unable to mkdir "
-                        "mountpoint");
-                goto out;
+        if ((op == GD_OP_SET_VOLUME)) {
+                //check for set volume help
+                ret = dict_get_str (dict, "volname", &volname);
+                if (volname &&
+                    ((strcmp (volname, "help") == 0) ||
+                     (strcmp (volname, "help-xml") == 0))) {
+                        ret = dict_get_str (dict, "key1", &key);
+                        if (ret < 0)
+                                return _gf_true;
+                }
         }
 
-        ret = rb_spawn_glusterfs_client (volinfo, src_brickinfo);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG, "Unable to start glusterfs");
+        return _gf_false;
+}
+
+gf_boolean_t
+glusterd_is_op_quorum_validation_required (xlator_t *this, glusterd_op_t op,
+                                           dict_t *dict)
+{
+        gf_boolean_t    required = _gf_true;
+        char            *key = NULL;
+        char            *key_fixed = NULL;
+        int             ret = -1;
+
+        if (glusterd_is_get_op (this, op, dict)) {
+                required = _gf_false;
                 goto out;
         }
-
-        ret = 0;
+        if ((op != GD_OP_SET_VOLUME) && (op != GD_OP_RESET_VOLUME))
+                goto out;
+        if (op == GD_OP_SET_VOLUME)
+                ret = dict_get_str (dict, "key1", &key);
+        else if (op == GD_OP_RESET_VOLUME)
+                ret = dict_get_str (dict, "key", &key);
+        if (ret)
+                goto out;
+        ret = glusterd_check_option_exists (key, &key_fixed);
+        if (ret <= 0)
+                goto out;
+        if (key_fixed)
+                key = key_fixed;
+        if (glusterd_is_quorum_option (key))
+                required = _gf_false;
 out:
-        return ret;
+        GF_FREE (key_fixed);
+        return required;
 }
 
 static int
-rb_spawn_destination_brick (glusterd_volinfo_t *volinfo,
-                            glusterd_brickinfo_t *dst_brickinfo)
-
+glusterd_op_validate_quorum (xlator_t *this, glusterd_op_t op,
+                             dict_t *dict, char **op_errstr)
 {
-        int ret = -1;
+        int                     ret = 0;
+        char                    *volname = NULL;
+        glusterd_volinfo_t      *volinfo = NULL;
+        char                    *errstr = NULL;
 
-        ret = rb_generate_dst_brick_volfile (volinfo, dst_brickinfo);
+
+        errstr = "Quorum not met. Volume operation not allowed.";
+        if (!glusterd_is_op_quorum_validation_required (this, op, dict))
+                goto out;
+
+        ret = dict_get_str (dict, "volname", &volname);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG, "Unable to generate client "
-                        "volfile");
+                ret = 0;
                 goto out;
         }
 
-        ret = rb_spawn_dst_brick (volinfo, dst_brickinfo);
+        ret = glusterd_volinfo_find (volname, &volinfo);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG, "Unable to start glusterfs");
+                ret = 0;
+                goto out;
+        }
+
+        if (does_gd_meet_server_quorum (this)) {
+                ret = 0;
                 goto out;
         }
 
+        if (glusterd_is_volume_in_server_quorum (volinfo)) {
+                ret = -1;
+                *op_errstr = gf_strdup (errstr);
+                goto out;
+        }
         ret = 0;
 out:
         return ret;
 }
 
 static int
-rb_do_operation_start (glusterd_volinfo_t *volinfo,
-                       glusterd_brickinfo_t *src_brickinfo,
-                       glusterd_brickinfo_t *dst_brickinfo)
+glusterd_op_ac_send_stage_op (glusterd_op_sm_event_t *event, void *ctx)
 {
-        char start_value[8192] = {0,};
-        int ret = -1;
+        int                     ret = 0;
+        rpc_clnt_procedure_t    *proc = NULL;
+        glusterd_conf_t         *priv = NULL;
+        xlator_t                *this = NULL;
+        glusterd_peerinfo_t     *peerinfo = NULL;
+        dict_t                  *dict = NULL;
+        char                    *op_errstr  = NULL;
+        glusterd_op_t           op = GD_OP_NONE;
+        uint32_t                pending_count = 0;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
 
-        ret = rb_src_brick_restart (volinfo, src_brickinfo,
-                                    1);
+        op = glusterd_op_get_op ();
+
+        ret = glusterd_op_build_payload (&dict, &op_errstr, NULL);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not restart src-brick");
+                gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD,
+                        gd_op_list[op]);
+                if (op_errstr == NULL)
+                        gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD);
+                opinfo.op_errstr = op_errstr;
                 goto out;
         }
 
-        gf_log ("", GF_LOG_DEBUG,
-                "replace-brick sending start xattr");
+        ret = glusterd_op_validate_quorum (this, op, dict, &op_errstr);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s", op_errstr);
+                opinfo.op_errstr = op_errstr;
+                goto out;
+        }
 
-        ret = rb_spawn_maintainence_client (volinfo, src_brickinfo);
+        /* rsp_dict NULL from source */
+        ret = glusterd_op_stage_validate (op, dict, &op_errstr, NULL);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not spawn maintainence "
-                        "client");
+                gf_log (this->name, GF_LOG_ERROR, LOGSTR_STAGE_FAIL,
+                        gd_op_list[op], "localhost",
+                        (op_errstr) ? ":" : " ", (op_errstr) ? op_errstr : " ");
+                if (op_errstr == NULL)
+                        gf_asprintf (&op_errstr, OPERRSTR_STAGE_FAIL,
+                                     "localhost");
+                opinfo.op_errstr = op_errstr;
                 goto out;
         }
 
-	gf_log ("", GF_LOG_DEBUG,
-		"mounted the replace brick client");
+        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
+                GF_ASSERT (peerinfo);
 
-        snprintf (start_value, 8192, "%s:%s:",
-                  dst_brickinfo->hostname,
-                  dst_brickinfo->path);
+                if (!peerinfo->connected || !peerinfo->mgmt)
+                        continue;
+                if ((peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED) &&
+                    (glusterd_op_get_op() != GD_OP_SYNC_VOLUME))
+                        continue;
 
+                proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_STAGE_OP];
+                GF_ASSERT (proc);
+                if (proc->fn) {
+                        ret = dict_set_static_ptr (dict, "peerinfo", peerinfo);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                        "set peerinfo");
+                                goto out;
+                        }
 
-        ret = rb_send_xattr_command (volinfo, src_brickinfo,
-                                     dst_brickinfo, RB_PUMP_START_CMD,
-                                     start_value);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Failed to send command to pump");
+                        ret = proc->fn (NULL, this, dict);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_WARNING, "Failed to "
+                                        "send stage request for operation "
+                                        "'Volume %s' to peer %s",
+                                        gd_op_list[op], peerinfo->hostname);
+                                continue;
+                        }
+                        pending_count++;
+                }
         }
 
-        ret = rb_destroy_maintainence_client (volinfo, src_brickinfo);
+        opinfo.pending_count = pending_count;
+out:
+        if (dict)
+                dict_unref (dict);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Failed to destroy maintainence "
-                        "client");
-                goto out;
+                glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT,
+                                             &event->txn_id, NULL);
+                opinfo.op_ret = ret;
         }
 
-        gf_log ("", GF_LOG_DEBUG,
-		"unmounted the replace brick client");
-        ret = 0;
+        gf_log (this->name, GF_LOG_DEBUG, "Sent stage op request for "
+                "'Volume %s' to %d peers", gd_op_list[op],
+                opinfo.pending_count);
+
+        if (!opinfo.pending_count)
+                ret = glusterd_op_sm_inject_all_acc (&event->txn_id);
+
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
 
-out:
         return ret;
+
 }
 
-static int
-rb_do_operation_pause (glusterd_volinfo_t *volinfo,
-                       glusterd_brickinfo_t *src_brickinfo,
-                       glusterd_brickinfo_t *dst_brickinfo)
+static int32_t
+glusterd_op_start_rb_timer (dict_t *dict, uuid_t *txn_id)
 {
-        int ret = -1;
+        int32_t         op = 0;
+        struct timespec timeout = {0, };
+        glusterd_conf_t *priv = NULL;
+        int32_t         ret = -1;
+        dict_t          *rb_ctx = NULL;
 
-        gf_log ("", GF_LOG_NORMAL,
-                "replace-brick send pause xattr");
+        GF_ASSERT (dict);
+        priv = THIS->private;
 
-        ret = rb_spawn_maintainence_client (volinfo, src_brickinfo);
+        ret = dict_get_int32 (dict, "operation", &op);
         if (ret) {
                 gf_log ("", GF_LOG_DEBUG,
-                        "Could not spawn maintainence "
-                        "client");
+                        "dict_get on operation failed");
                 goto out;
         }
 
-	gf_log ("", GF_LOG_DEBUG,
-		"mounted the replace brick client");
+        if (op != GF_REPLACE_OP_START) {
+                ret = glusterd_op_sm_inject_all_acc (txn_id);
+                goto out;
+        }
 
-        ret = rb_send_xattr_command (volinfo, src_brickinfo,
-                                     dst_brickinfo, RB_PUMP_PAUSE_CMD,
-                                     "jargon");
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Failed to send command to pump");
+        timeout.tv_sec  = 5;
+        timeout.tv_nsec = 0;
 
+
+        rb_ctx = dict_copy (dict, rb_ctx);
+        if (!rb_ctx) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Couldn't copy "
+                        "replace brick context. Can't start replace brick");
+                ret = -1;
+                goto out;
         }
 
-        ret = rb_destroy_maintainence_client (volinfo, src_brickinfo);
+        ret = dict_set_bin (rb_ctx, "transaction_id",
+                            txn_id, sizeof(*txn_id));
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Failed to destroy maintainence "
-                        "client");
+                gf_log ("", GF_LOG_ERROR,
+                       "Failed to set transaction id.");
                 goto out;
-        }
+        } else
+               gf_log ("", GF_LOG_DEBUG,
+                        "transaction_id = %s", uuid_utoa (*txn_id));
 
-	gf_log ("", GF_LOG_DEBUG,
-		"unmounted the replace brick client");
+        priv->timer = gf_timer_call_after (THIS->ctx, timeout,
+                                           glusterd_do_replace_brick,
+                                           (void *) rb_ctx);
 
         ret = 0;
 
@@ -2181,1957 +3355,2306 @@ out:
         return ret;
 }
 
+/* This function takes a dict and converts the uuid values of key specified
+ * into hostnames
+ */
 static int
-rb_kill_destination_brick (glusterd_volinfo_t *volinfo,
-                           glusterd_brickinfo_t *dst_brickinfo)
+glusterd_op_volume_dict_uuid_to_hostname (dict_t *dict, const char *key_fmt,
+                                          int idx_min, int idx_max)
 {
-        glusterd_conf_t  *priv               = NULL;
-        char              pidfile[PATH_MAX]  = {0,};
-
-        priv = THIS->private;
-
-        snprintf (pidfile, PATH_MAX, "%s/vols/%s/%s",
-                  priv->workdir, volinfo->volname,
-                  RB_DSTBRICK_PIDFILE);
-
-        return glusterd_service_stop ("brick", pidfile, SIGQUIT, _gf_true);
-}
-
-static int
-rb_do_operation_abort (glusterd_volinfo_t *volinfo,
-                       glusterd_brickinfo_t *src_brickinfo,
-                       glusterd_brickinfo_t *dst_brickinfo)
-{
-        int ret = -1;
+        int             ret = -1;
+        int             i = 0;
+        char            key[1024];
+        char            *uuid_str = NULL;
+        uuid_t          uuid = {0,};
+        char            *hostname = NULL;
+        xlator_t        *this = NULL;
 
-        ret = rb_src_brick_restart (volinfo, src_brickinfo,
-                                    0);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not restart src-brick");
-                goto out;
-        }
+        this = THIS;
+        GF_ASSERT (this);
 
-        gf_log ("", GF_LOG_DEBUG,
-                "replace-brick sending abort xattr");
+        GF_ASSERT (dict);
+        GF_ASSERT (key_fmt);
 
-        ret = rb_spawn_maintainence_client (volinfo, src_brickinfo);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not spawn maintainence "
-                        "client");
-                goto out;
-        }
+        for (i = idx_min; i < idx_max; i++) {
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), key_fmt, i);
+                ret = dict_get_str (dict, key, &uuid_str);
+                if (ret)
+                        continue;
 
-	gf_log ("", GF_LOG_DEBUG,
-		"mounted the replace brick client");
+                gf_log (this->name, GF_LOG_DEBUG, "Got uuid %s",
+                        uuid_str);
 
-        ret = rb_send_xattr_command (volinfo, src_brickinfo,
-                                     dst_brickinfo, RB_PUMP_ABORT_CMD,
-                                     "jargon");
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Failed to send command to pump");
-	}
+                ret = uuid_parse (uuid_str, uuid);
+                /* if parsing fails don't error out
+                 * let the original value be retained
+                 */
+                if (ret)
+                        continue;
 
-        ret = rb_destroy_maintainence_client (volinfo, src_brickinfo);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Failed to destroy maintainence "
-                        "client");
-                goto out;
+                hostname = glusterd_uuid_to_hostname (uuid);
+                if (hostname) {
+                        gf_log (this->name, GF_LOG_DEBUG, "%s -> %s",
+                                uuid_str, hostname);
+                        ret = dict_set_dynstr (dict, key, hostname);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Error setting hostname %s to dict",
+                                        hostname);
+                                GF_FREE (hostname);
+                                goto out;
+                        }
+                }
         }
 
-	gf_log ("", GF_LOG_DEBUG,
-		"unmounted the replace brick client");
-
-        ret = 0;
-
 out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
 static int
-rb_do_operation_commit (glusterd_volinfo_t *volinfo,
-                        glusterd_brickinfo_t *src_brickinfo,
-                        glusterd_brickinfo_t *dst_brickinfo)
+reassign_defrag_status (dict_t *dict, char *key, gf_defrag_status_t *status)
 {
         int ret = 0;
 
-        ret = rb_src_brick_restart (volinfo, src_brickinfo,
-                                    0);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not restart src-brick");
-                goto out;
-        }
+        if (!*status)
+                return ret;
+
+        switch (*status) {
+        case GF_DEFRAG_STATUS_STARTED:
+                *status = GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED;
+                break;
+
+        case GF_DEFRAG_STATUS_STOPPED:
+                *status = GF_DEFRAG_STATUS_LAYOUT_FIX_STOPPED;
+                break;
+
+        case GF_DEFRAG_STATUS_COMPLETE:
+                *status = GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE;
+                break;
+
+        case GF_DEFRAG_STATUS_FAILED:
+                *status = GF_DEFRAG_STATUS_LAYOUT_FIX_FAILED;
+                break;
+        default:
+                break;
+         }
+
+        ret = dict_set_int32(dict, key, *status);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to reset defrag %s in dict", key);
 
-        gf_log ("", GF_LOG_DEBUG,
-                "received commit on %s:%s to %s:%s "
-                "on volume %s",
-                src_brickinfo->hostname,
-                src_brickinfo->path,
-                dst_brickinfo->hostname,
-                dst_brickinfo->path,
-                volinfo->volname);
-out:
         return ret;
 }
 
+/* Check and reassign the defrag_status enum got from the rebalance process
+ * of all peers so that the rebalance-status CLI command can display if a
+ * full-rebalance or just a fix-layout was carried out.
+ */
 static int
-rb_get_xattr_command (glusterd_volinfo_t *volinfo,
-                      glusterd_brickinfo_t *src_brickinfo,
-                      glusterd_brickinfo_t *dst_brickinfo,
-                      const char *xattr_key,
-                      const char **value)
+glusterd_op_check_peer_defrag_status (dict_t *dict, int count)
 {
-        glusterd_conf_t  *priv                        = NULL;
-        char              mount_point_path[PATH_MAX]  = {0,};
-        struct stat       buf;
-        int               ret                         = -1;
-
-        priv = THIS->private;
-
-        snprintf (mount_point_path, PATH_MAX, "%s/vols/%s/%s",
-                  priv->workdir, volinfo->volname,
-                  RB_CLIENT_MOUNTPOINT);
+        glusterd_volinfo_t *volinfo  = NULL;
+        gf_defrag_status_t status    = GF_DEFRAG_STATUS_NOT_STARTED;
+        char               key[256]  = {0,};
+        char               *volname  = NULL;
+        int                ret       = -1;
+        int                i         = 1;
 
-        ret = stat (mount_point_path, &buf);
+        ret = dict_get_str (dict, "volname", &volname);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "stat failed. Could not send "
-                        " %s command", xattr_key);
+                gf_log (THIS->name, GF_LOG_WARNING, "Unable to get volume name");
                 goto out;
         }
 
-        ret = lgetxattr (mount_point_path, xattr_key,
-                         (char *)(*value),
-                         8192);
-
+        ret = glusterd_volinfo_find (volname, &volinfo);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "getxattr failed");
+                gf_log (THIS->name, GF_LOG_WARNING, FMTSTR_CHECK_VOL_EXISTS,
+                        volname);
                 goto out;
         }
 
-        ret = 0;
+        if (volinfo->rebal.defrag_cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) {
+                /* Fix layout was not issued; we don't need to reassign
+                   the status */
+                ret = 0;
+                goto out;
+        }
 
+        do {
+                memset (key, 0, 256);
+                snprintf (key, 256, "status-%d", i);
+                ret = dict_get_int32 (dict, key, (int32_t *)&status);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_WARNING,
+                                "failed to get defrag %s", key);
+                        goto out;
+                }
+                ret = reassign_defrag_status (dict, key, &status);
+                if (ret)
+                        goto out;
+                i++;
+        } while (i <= count);
+
+        ret = 0;
 out:
         return ret;
+
 }
 
-static int
-rb_do_operation_status (glusterd_volinfo_t *volinfo,
-                        glusterd_brickinfo_t *src_brickinfo,
-                        glusterd_brickinfo_t *dst_brickinfo)
+/* This function is used to modify the op_ctx dict before sending it back
+ * to cli. This is useful in situations like changing the peer uuids to
+ * hostnames etc.
+ */
+void
+glusterd_op_modify_op_ctx (glusterd_op_t op, void *ctx)
 {
-        const char *status       = NULL;
-        char       *status_reply = NULL;
-        dict_t     *ctx          = NULL;
-        int ret = -1;
+        int             ret = -1;
+        dict_t          *op_ctx = NULL;
+        int             brick_index_max = -1;
+        int             other_count = 0;
+        int             count = 0;
+        uint32_t        cmd = GF_CLI_STATUS_NONE;
+        xlator_t        *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
 
-        if (!glusterd_is_local_addr (src_brickinfo->hostname)) {
-                gf_log ("", GF_LOG_NORMAL,
-                        "I AM THE SOURCE HOST");
-                ret = rb_spawn_maintainence_client (volinfo, src_brickinfo);
+        if (ctx)
+                op_ctx = ctx;
+        else
+                op_ctx = glusterd_op_get_ctx();
+
+        if (!op_ctx) {
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "Operation context is not present.");
+                goto out;
+        }
+
+        switch (op) {
+        case GD_OP_STATUS_VOLUME:
+                ret = dict_get_uint32 (op_ctx, "cmd", &cmd);
                 if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Could not spawn maintainence "
-                                "client");
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Failed to get status cmd");
+                        goto out;
+                }
+                if (!(cmd & GF_CLI_STATUS_NFS || cmd & GF_CLI_STATUS_SHD ||
+                    (cmd & GF_CLI_STATUS_MASK) == GF_CLI_STATUS_NONE)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "op_ctx modification not required for status "
+                                "operation being performed");
                         goto out;
                 }
 
-		gf_log ("", GF_LOG_DEBUG,
-			"mounted the replace brick client");
+                ret = dict_get_int32 (op_ctx, "brick-index-max",
+                                      &brick_index_max);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Failed to get brick-index-max");
+                        goto out;
+                }
 
-                ret = rb_get_xattr_command (volinfo, src_brickinfo,
-                                            dst_brickinfo, RB_PUMP_STATUS_CMD,
-                                            &status);
+                ret = dict_get_int32 (op_ctx, "other-count", &other_count);
                 if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Failed to get status from pump");
-                        goto umount;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Failed to get other-count");
+                        goto out;
                 }
 
-                gf_log ("", GF_LOG_DEBUG,
-                        "pump status is %s", status);
+                count = brick_index_max + other_count + 1;
 
-                ctx = glusterd_op_get_ctx (GD_OP_REPLACE_BRICK);
-                if (!ctx) {
-                        gf_log ("", GF_LOG_ERROR,
-                                "Operation Context is not present");
-                        ret = -1;
-                        goto umount;
-                }
-                status_reply = gf_strdup (status);
-                if (!status_reply) {
-                        gf_log ("", GF_LOG_ERROR, "Out of memory");
-                        ret = -1;
-                        goto umount;
+                ret = glusterd_op_volume_dict_uuid_to_hostname (op_ctx,
+                                                                "brick%d.path",
+                                                                0, count);
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Failed uuid to hostname conversion");
+
+                break;
+
+        case GD_OP_PROFILE_VOLUME:
+                ret = dict_get_str_boolean (op_ctx, "nfs", _gf_false);
+                if (!ret)
+                        goto out;
+
+                ret = dict_get_int32 (op_ctx, "count", &count);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Failed to get brick count");
+                        goto out;
                 }
 
-                ret = dict_set_dynstr (ctx, "status-reply",
-                                       status_reply);
+                ret = glusterd_op_volume_dict_uuid_to_hostname (op_ctx,
+                                                                "%d-brick",
+                                                                1, (count + 1));
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Failed uuid to hostname conversion");
+
+                break;
+
+        /* For both rebalance and remove-brick status, the glusterd op is the
+         * same
+         */
+        case GD_OP_DEFRAG_BRICK_VOLUME:
+                ret = dict_get_int32 (op_ctx, "count", &count);
                 if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "failed to set pump status in ctx");
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Failed to get count");
+                        goto out;
+                }
 
+                /* add 'node-name-%d' into op_ctx with value uuid_str.
+                   this will be used to convert to hostname later */
+                {
+                        char  key[1024];
+                        char *uuid_str = NULL;
+                        int   i;
+
+                        for (i = 1; i <= count; i++) {
+                                memset (key, 0, sizeof (key));
+                                snprintf (key, sizeof (key), "node-uuid-%d", i);
+                                ret = dict_get_str (op_ctx, key, &uuid_str);
+                                if (!ret) {
+                                        memset (key, 0, sizeof (key));
+                                        snprintf (key, sizeof (key),
+                                                  "node-name-%d", i);
+                                        ret = dict_set_str (op_ctx, key,
+                                                            uuid_str);
+                                }
+                        }
                 }
 
-	umount:
-                ret = rb_destroy_maintainence_client (volinfo, src_brickinfo);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Failed to destroy maintainence "
-                                "client");
-			goto out;
-		}
+                ret = glusterd_op_volume_dict_uuid_to_hostname (op_ctx,
+                                                                "node-name-%d",
+                                                                1, (count + 1));
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Failed uuid to hostname conversion");
+
+                ret = glusterd_op_check_peer_defrag_status (op_ctx, count);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to reset defrag status for fix-layout");
+                break;
+
+        default:
+                ret = 0;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "op_ctx modification not required");
+                break;
+
         }
 
-	gf_log ("", GF_LOG_DEBUG,
-		"unmounted the replace brick client");
 out:
-        return ret;
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "op_ctx modification failed");
+        return;
 }
 
 static int
-glusterd_op_replace_brick (gd1_mgmt_stage_op_req *req)
+glusterd_op_commit_hook (glusterd_op_t op, dict_t *op_ctx,
+                         glusterd_commit_hook_type_t type)
 {
-        int                                      ret = 0;
-        dict_t                                  *dict = NULL;
-        dict_t                                  *ctx  = NULL;
-        gf1_cli_replace_op                       replace_op;
-        glusterd_volinfo_t                      *volinfo = NULL;
-        char                                    *volname = NULL;
-        xlator_t                                *this = NULL;
-        glusterd_conf_t                         *priv = NULL;
-        char                                    *src_brick = NULL;
-        char                                    *dst_brick = NULL;
-        glusterd_brickinfo_t                    *src_brickinfo = NULL;
-        glusterd_brickinfo_t                    *dst_brickinfo = NULL;
-
-        GF_ASSERT (req);
-
-        this = THIS;
-        GF_ASSERT (this);
-
-        priv = this->private;
-        GF_ASSERT (priv);
+        glusterd_conf_t *priv                   = NULL;
+        char            hookdir[PATH_MAX]       = {0, };
+        char            scriptdir[PATH_MAX]     = {0, };
+        char            type_subdir[256]        = {0, };
+        char            *cmd_subdir             = NULL;
+        int             ret                     = -1;
 
-        dict = dict_new ();
-        if (!dict)
-                goto out;
+        priv = THIS->private;
+        switch (type) {
+                case GD_COMMIT_HOOK_NONE:
+                case GD_COMMIT_HOOK_MAX:
+                        /*Won't be called*/
+                        break;
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
-                goto out;
+                case GD_COMMIT_HOOK_PRE:
+                        strcpy (type_subdir, "pre");
+                        break;
+                case GD_COMMIT_HOOK_POST:
+                        strcpy (type_subdir, "post");
+                        break;
         }
 
-        ret = dict_get_str (dict, "src-brick", &src_brick);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get src brick");
-                goto out;
-        }
+        cmd_subdir = glusterd_hooks_get_hooks_cmd_subdir (op);
+        if (strlen (cmd_subdir) == 0)
+                return -1;
 
-        gf_log (this->name, GF_LOG_DEBUG,
-                "src brick=%s", src_brick);
+        GLUSTERD_GET_HOOKS_DIR (hookdir, GLUSTERD_HOOK_VER, priv);
+        snprintf (scriptdir, sizeof (scriptdir), "%s/%s/%s",
+                  hookdir, cmd_subdir, type_subdir);
 
-        ret = dict_get_str (dict, "dst-brick", &dst_brick);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get dst brick");
-                goto out;
-        }
+        switch (type) {
+                case GD_COMMIT_HOOK_NONE:
+                case GD_COMMIT_HOOK_MAX:
+                        /*Won't be called*/
+                        break;
 
-        gf_log (this->name, GF_LOG_DEBUG,
-                "dst brick=%s", dst_brick);
+                case GD_COMMIT_HOOK_PRE:
+                        ret = glusterd_hooks_run_hooks (scriptdir, op, op_ctx,
+                                                        type);
+                        break;
+                case GD_COMMIT_HOOK_POST:
+                        ret = glusterd_hooks_post_stub_enqueue (scriptdir, op,
+                                                                op_ctx);
+                        break;
+        }
 
-        ret = dict_get_str (dict, "volname", &volname);
+        return ret;
+}
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
-                goto out;
-        }
+static int
+glusterd_op_ac_send_commit_op (glusterd_op_sm_event_t *event, void *ctx)
+{
+        int                     ret = 0;
+        rpc_clnt_procedure_t    *proc = NULL;
+        glusterd_conf_t         *priv = NULL;
+        xlator_t                *this = NULL;
+        dict_t                  *dict = NULL;
+        dict_t                  *op_dict = NULL;
+        glusterd_peerinfo_t     *peerinfo = NULL;
+        char                    *op_errstr  = NULL;
+        glusterd_op_t           op = GD_OP_NONE;
+        uint32_t                pending_count = 0;
 
-        ret = dict_get_int32 (dict, "operation", (int32_t *)&replace_op);
-        if (ret) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "dict_get on operation failed");
-                goto out;
-        }
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
 
-        ret = glusterd_volinfo_find (volname, &volinfo);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
-                goto out;
-        }
+        op      = glusterd_op_get_op ();
+        op_dict = glusterd_op_get_ctx ();
 
-        ret = glusterd_brickinfo_get (src_brick, volinfo, &src_brickinfo);
+        ret = glusterd_op_build_payload (&dict, &op_errstr, NULL);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG, "Unable to get src-brickinfo");
+                gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD,
+                        gd_op_list[op]);
+                if (op_errstr == NULL)
+                        gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD);
+                opinfo.op_errstr = op_errstr;
                 goto out;
         }
 
-        ret = glusterd_brickinfo_from_brick (dst_brick, &dst_brickinfo);
+        ret = glusterd_op_commit_perform (op, dict, &op_errstr, NULL); //rsp_dict invalid for source
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG, "Unable to get dst-brickinfo");
+                gf_log (this->name, GF_LOG_ERROR, LOGSTR_COMMIT_FAIL,
+                        gd_op_list[op], "localhost", (op_errstr) ? ":" : " ",
+                        (op_errstr) ? op_errstr : " ");
+                if (op_errstr == NULL)
+                        gf_asprintf (&op_errstr, OPERRSTR_COMMIT_FAIL,
+                                     "localhost");
+                opinfo.op_errstr = op_errstr;
                 goto out;
         }
 
-        /* Set src-brick's port number to be used in the maintainance mount
-         * after all commit acks are received.
-         */
-        if (!glusterd_is_local_addr (src_brickinfo->hostname)) {
-                gf_log ("", GF_LOG_NORMAL,
-                        "adding src-brick port no");
 
-                ctx = glusterd_op_get_ctx (GD_OP_REPLACE_BRICK);
-                if (!ctx) {
-                        gf_log ("", GF_LOG_ERROR,
-                                "Operation Context is not present");
-                        ret = -1;
-                        goto out;
-                }
+        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
+                GF_ASSERT (peerinfo);
 
-                ret = dict_set_int32 (ctx, "src-brick-port",
-                                      src_brickinfo->port);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Could not set src-brick port no");
-                        goto out;
-                }
-        }
+                if (!peerinfo->connected || !peerinfo->mgmt)
+                        continue;
+                if ((peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED) &&
+                    (glusterd_op_get_op() != GD_OP_SYNC_VOLUME))
+                        continue;
 
-        switch (replace_op) {
-        case GF_REPLACE_OP_START:
-        {
-                if (!glusterd_is_local_addr (dst_brickinfo->hostname)) {
-                        gf_log ("", GF_LOG_NORMAL,
-                                "I AM THE DESTINATION HOST");
-                        ret = rb_spawn_destination_brick (volinfo, dst_brickinfo);
+                proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_COMMIT_OP];
+                GF_ASSERT (proc);
+                if (proc->fn) {
+                        ret = dict_set_static_ptr (dict, "peerinfo", peerinfo);
                         if (ret) {
-                                gf_log ("", GF_LOG_DEBUG,
-                                        "Failed to spawn destination brick");
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to set peerinfo");
                                 goto out;
                         }
+                        ret = proc->fn (NULL, this, dict);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_WARNING, "Failed to "
+                                        "send commit request for operation "
+                                        "'Volume %s' to peer %s",
+                                        gd_op_list[op], peerinfo->hostname);
+                                continue;
+                        }
+                        pending_count++;
                 }
         }
-                break;
 
-        case GF_REPLACE_OP_COMMIT:
-        {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Received commit - doing nothing");
+        opinfo.pending_count = pending_count;
+        gf_log (this->name, GF_LOG_DEBUG, "Sent commit op req for 'Volume %s' "
+                "to %d peers", gd_op_list[op], opinfo.pending_count);
+out:
+        if (dict)
+                dict_unref (dict);
+        if (ret) {
+                glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT,
+                                             &event->txn_id, NULL);
+                opinfo.op_ret = ret;
         }
-                break;
 
-        case GF_REPLACE_OP_PAUSE:
-        {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Recieved pause - doing nothing");
-        }
-                break;
+        if (!opinfo.pending_count) {
+                if (op == GD_OP_REPLACE_BRICK) {
+                        ret = glusterd_op_start_rb_timer (op_dict,
+                                                          &event->txn_id);
 
-        case GF_REPLACE_OP_ABORT:
-        {
-                if (!glusterd_is_local_addr (dst_brickinfo->hostname)) {
-                        gf_log ("", GF_LOG_NORMAL,
-                                "I AM THE DESTINATION HOST");
-                        ret = rb_kill_destination_brick (volinfo, dst_brickinfo);
-                        if (ret) {
-                                gf_log ("", GF_LOG_DEBUG,
-                                        "Failed to kill destination brick");
-                                goto out;
-                        }
+                } else {
+                        glusterd_op_modify_op_ctx (op, NULL);
+                        ret = glusterd_op_sm_inject_all_acc (&event->txn_id);
                 }
+                goto err;
         }
-                break;
 
-        case GF_REPLACE_OP_STATUS:
-        {
-                gf_log ("", GF_LOG_DEBUG,
-                        "received status - doing nothing");
-        }
-                break;
+err:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
 
-        default:
-                ret = -1;
-                goto out;
-        }
+        return ret;
 
-        if (ret)
+}
+
+static int
+glusterd_op_ac_rcvd_stage_op_acc (glusterd_op_sm_event_t *event, void *ctx)
+{
+        int                     ret = 0;
+
+        GF_ASSERT (event);
+
+        if (opinfo.pending_count > 0)
+                opinfo.pending_count--;
+
+        if (opinfo.pending_count > 0)
                 goto out;
 
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_ACC,
+                                           &event->txn_id, NULL);
+
 out:
-        if (dict)
-                dict_unref (dict);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+
         return ret;
 }
 
 static int
-glusterd_op_set_volume (gd1_mgmt_stage_op_req *req)
+glusterd_op_ac_stage_op_failed (glusterd_op_sm_event_t *event, void *ctx)
 {
-        int                                      ret = 0;
-        dict_t                                  *dict = NULL;
-        glusterd_volinfo_t                      *volinfo = NULL;
-        char                                    *volname = NULL;
-        xlator_t                                *this = NULL;
-        glusterd_conf_t                         *priv = NULL;
-	int					 count = 1;
-	char					*key = NULL;
-	char					*value = NULL;
-	char					 str[50] = {0, };
-        GF_ASSERT (req);
+        int                     ret = 0;
 
-        this = THIS;
-        GF_ASSERT (this);
+        GF_ASSERT (event);
 
-        priv = this->private;
-        GF_ASSERT (priv);
+        if (opinfo.pending_count > 0)
+                opinfo.pending_count--;
 
-        dict = dict_new ();
-        if (!dict)
+        if (opinfo.pending_count > 0)
                 goto out;
 
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK,
+                                           &event->txn_id, NULL);
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
-                goto out;
-        }
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
 
-        ret = dict_get_str (dict, "volname", &volname);
+        return ret;
+}
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
-                goto out;
-        }
+static int
+glusterd_op_ac_commit_op_failed (glusterd_op_sm_event_t *event, void *ctx)
+{
+        int                     ret = 0;
 
-        ret = glusterd_volinfo_find (volname, &volinfo);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
-                goto out;
-        }
+        GF_ASSERT (event);
 
-	for ( count = 1; ret != -1 ; count++ ) {
+        if (opinfo.pending_count > 0)
+                opinfo.pending_count--;
 
-		sprintf (str, "key%d", count);
-		ret = dict_get_str (dict, str, &key);
+        if (opinfo.pending_count > 0)
+                goto out;
 
-		
-		if (ret) 
-			break;
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK,
+                                           &event->txn_id, NULL);
 
-		sprintf (str, "value%d", count);
-		ret = dict_get_str (dict, str, &value);
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
 
-		if (ret) {
-                	gf_log ("", GF_LOG_ERROR, "invalid key,value pair"
-						  "in 'volume set'");
-			ret = -1;
-			goto out;
-        	}
+        return ret;
+}
 
-		ret = set_xlator_option (volinfo->dict, key, value);
-		
-		if (ret) {
-                	gf_log ("", GF_LOG_ERROR, "Unable to set the options"
-						  "in 'volume set'");
-			ret = -1;
-			goto out;
-        	}
-	}
+static int
+glusterd_op_ac_brick_op_failed (glusterd_op_sm_event_t *event, void *ctx)
+{
+        int                     ret = 0;
+        glusterd_op_brick_rsp_ctx_t *ev_ctx = NULL;
+        gf_boolean_t                free_errstr = _gf_false;
+        xlator_t                    *this = NULL;
 
-	if ( count == 1 ) {
-		gf_log ("", GF_LOG_ERROR, "No options received ");
-		ret = -1;
-		goto out;
-	}
+        this = THIS;
+        GF_ASSERT (this);
 
-	ret = glusterd_create_volfiles (volinfo);
+        GF_ASSERT (event);
+        GF_ASSERT (ctx);
+        ev_ctx = ctx;
 
-	if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to create volfile for"
-					  " 'volume set'");
-		ret = -1;
-		goto out;
+        ret = glusterd_remove_pending_entry (&opinfo.pending_bricks, ev_ctx->pending_node->node);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "unknown response received ");
+                ret = -1;
+                free_errstr = _gf_true;
+                goto out;
         }
+        if (opinfo.brick_pending_count > 0)
+                opinfo.brick_pending_count--;
+        if (opinfo.op_ret == 0)
+                opinfo.op_ret = ev_ctx->op_ret;
 
+        if (opinfo.op_errstr == NULL)
+                opinfo.op_errstr = ev_ctx->op_errstr;
+        else
+                free_errstr = _gf_true;
 
+        if (opinfo.brick_pending_count > 0)
+                goto out;
 
-
-        gf_log ("", GF_LOG_DEBUG, "Received set volume command");
-
-        ret = 0;
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK,
+                                           &event->txn_id, ev_ctx->commit_ctx);
 
 out:
+        if (ev_ctx->rsp_dict)
+                dict_unref (ev_ctx->rsp_dict);
+        if (free_errstr && ev_ctx->op_errstr)
+                GF_FREE (ev_ctx->op_errstr);
+        GF_FREE (ctx);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
         return ret;
 }
 
 static int
-glusterd_op_remove_brick (gd1_mgmt_stage_op_req *req)
+glusterd_op_ac_rcvd_commit_op_acc (glusterd_op_sm_event_t *event, void *ctx)
 {
-        int                                     ret = -1;
-        dict_t                                  *dict = NULL;
-        char                                    *volname = NULL;
-        glusterd_conf_t                         *priv = NULL;
-        glusterd_volinfo_t                      *volinfo = NULL;
-        glusterd_brickinfo_t                    *brickinfo = NULL;
-        xlator_t                                *this = NULL;
-        char                                    *brick = NULL;
-        int32_t                                 count = 0;
-        int32_t                                 i = 1;
-        gf_boolean_t                            glfs_stopped = _gf_false;
-        int32_t                                 mybrick = 0;
-        char                                    key[256] = {0,};
-        char                                    *dup_brick = NULL;
-
-        GF_ASSERT (req);
+        dict_t                 *op_ctx            = NULL;
+        int                     ret               = 0;
+        gf_boolean_t            commit_ack_inject = _gf_true;
+        glusterd_op_t           op                = GD_OP_NONE;
+        xlator_t               *this = NULL;
 
         this = THIS;
         GF_ASSERT (this);
+        op = glusterd_op_get_op ();
+        GF_ASSERT (event);
 
-        priv = this->private;
-        GF_ASSERT (priv);
+        if (opinfo.pending_count > 0)
+                opinfo.pending_count--;
 
-        dict = dict_new ();
-        if (!dict)
+        if (opinfo.pending_count > 0)
                 goto out;
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
+        if (op == GD_OP_REPLACE_BRICK) {
+                op_ctx = glusterd_op_get_ctx ();
+                if (!op_ctx) {
+                        gf_log (this->name, GF_LOG_CRITICAL, "Operation "
+                                "context is not present.");
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = glusterd_op_start_rb_timer (op_ctx, &event->txn_id);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Couldn't start "
+                                "replace-brick operation.");
+                        goto out;
+                }
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
+                commit_ack_inject = _gf_false;
                 goto out;
         }
 
-        ret = dict_get_str (dict, "volname", &volname);
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
-                goto out;
+out:
+        if (commit_ack_inject) {
+                if (ret)
+                        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT,
+                                                           &event->txn_id, NULL);
+                else if (!opinfo.pending_count) {
+                        glusterd_op_modify_op_ctx (op, NULL);
+                        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC,
+                                                           &event->txn_id, NULL);
+                }
+                /*else do nothing*/
         }
 
-        ret = glusterd_volinfo_find (volname, &volinfo);
+        return ret;
+}
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
-                goto out;
-        }
+static int
+glusterd_op_ac_rcvd_unlock_acc (glusterd_op_sm_event_t *event, void *ctx)
+{
+        int                     ret = 0;
 
+        GF_ASSERT (event);
 
-        ret = dict_get_int32 (dict, "count", &count);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get count");
+        if (opinfo.pending_count > 0)
+                opinfo.pending_count--;
+
+        if (opinfo.pending_count > 0)
                 goto out;
-        }
 
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC,
+                                           &event->txn_id, NULL);
 
-        while ( i <= count) {
-                snprintf (key, 256, "brick%d", i);
-                ret = dict_get_str (dict, key, &brick);
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to get %s", key);
-                        goto out;
-                }
-                if (dup_brick)
-                        GF_FREE (dup_brick);
-                dup_brick = gf_strdup (brick);
-                if (!dup_brick)
-                        goto out;
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
 
-                ret = glusterd_brickinfo_get (dup_brick, volinfo,  &brickinfo);
-                if (ret)
-                        goto out;
+out:
+        return ret;
+}
+
+int32_t
+glusterd_op_clear_errstr() {
+        opinfo.op_errstr = NULL;
+        return 0;
+}
+
+int32_t
+glusterd_op_set_ctx (void *ctx)
+{
+
+        opinfo.op_ctx = ctx;
+
+        return 0;
 
+}
 
-                ret = glusterd_resolve_brick (brickinfo);
+int32_t
+glusterd_op_reset_ctx ()
+{
 
+        glusterd_op_set_ctx (NULL);
+
+        return 0;
+}
+
+int32_t
+glusterd_op_txn_complete (uuid_t *txn_id)
+{
+        int32_t                 ret = -1;
+        glusterd_conf_t         *priv = NULL;
+        int32_t                 op = -1;
+        int32_t                 op_ret = 0;
+        int32_t                 op_errno = 0;
+        rpcsvc_request_t        *req = NULL;
+        void                    *ctx = NULL;
+        char                    *op_errstr = NULL;
+        char                    *volname = NULL;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        op  = glusterd_op_get_op ();
+        ctx = glusterd_op_get_ctx ();
+        op_ret = opinfo.op_ret;
+        op_errno = opinfo.op_errno;
+        req = opinfo.req;
+        if (opinfo.op_errstr)
+                op_errstr = opinfo.op_errstr;
+
+        opinfo.op_ret = 0;
+        opinfo.op_errno = 0;
+        glusterd_op_clear_op ();
+        glusterd_op_reset_ctx ();
+        glusterd_op_clear_errstr ();
+
+        /* Based on the op-version, we release the cluster or mgmt_v3 lock */
+        if (priv->op_version < 3) {
+                ret = glusterd_unlock (MY_UUID);
+                /* unlock cant/shouldnt fail here!! */
                 if (ret)
-                        goto out;
+                        gf_log (this->name, GF_LOG_CRITICAL,
+                                "Unable to clear local lock, ret: %d", ret);
+                else
+                        gf_log (this->name, GF_LOG_DEBUG, "Cleared local lock");
+        } else {
+                ret = dict_get_str (ctx, "volname", &volname);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to acquire volname");
 
-                if ((!uuid_compare (brickinfo->uuid, priv->uuid)) &&
-                    (GLUSTERD_STATUS_STARTED == volinfo->status)) {
-                        gf_log ("", GF_LOG_NORMAL, "About to stop glusterfs"
-                                " for brick %s:%s", brickinfo->hostname,
-                                brickinfo->path);
-                        ret = glusterd_volume_stop_glusterfs
-                                                (volinfo, brickinfo, mybrick);
-                        if (ret) {
-                                gf_log ("", GF_LOG_ERROR, "Unable to stop "
-                                        "glusterfs, ret: %d", ret);
-                                goto out;
-                        }
-                        glfs_stopped = _gf_true;
-                        mybrick++;
+                if (volname) {
+                        ret = glusterd_mgmt_v3_unlock (volname, MY_UUID,
+                                                       "vol");
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to release lock for %s",
+                                        volname);
                 }
+        }
 
-                glusterd_delete_volfile (volinfo, brickinfo);
-                glusterd_store_delete_brick (volinfo, brickinfo);
-                glusterd_brickinfo_delete (brickinfo);
-                volinfo->brick_count--;
+        ret = glusterd_op_send_cli_response (op, op_ret,
+                                             op_errno, req, ctx, op_errstr);
 
-                i++;
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Responding to cli failed, "
+                        "ret: %d", ret);
+                //Ignore this error, else state machine blocks
+                ret = 0;
         }
 
-        ret = glusterd_create_volfiles (volinfo);
-        if (ret)
-                goto out;
+        if (op_errstr && (strcmp (op_errstr, "")))
+                GF_FREE (op_errstr);
 
-        volinfo->version++;
-        volinfo->defrag_status = 0;
 
-        ret = glusterd_store_update_volume (volinfo);
+        if (priv->pending_quorum_action)
+                glusterd_do_quorum_action ();
 
+        /* Clearing the transaction opinfo */
+        ret = glusterd_clear_txn_opinfo (txn_id);
         if (ret)
-                goto out;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to clear transaction's opinfo");
 
-        ret = glusterd_volume_compute_cksum (volinfo);
-        if (ret)
-                goto out;
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
 
-        if (GLUSTERD_STATUS_STARTED == volinfo->status)
-                ret = glusterd_check_generate_start_nfs (volinfo);
+static int
+glusterd_op_ac_unlocked_all (glusterd_op_sm_event_t *event, void *ctx)
+{
+        int                     ret = 0;
+
+        GF_ASSERT (event);
+
+        ret = glusterd_op_txn_complete (&event->txn_id);
+
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
 
-out:
-        if (dict)
-                dict_unref (dict);
-        if (dup_brick)
-                GF_FREE (dup_brick);
         return ret;
 }
 
-
 static int
-glusterd_op_delete_volume (gd1_mgmt_stage_op_req *req)
+glusterd_op_ac_stage_op (glusterd_op_sm_event_t *event, void *ctx)
 {
-        int                                     ret = 0;
-        char                                    volname[1024] = {0,};
-        glusterd_conf_t                         *priv = NULL;
-        glusterd_volinfo_t                      *volinfo = NULL;
-        xlator_t                                *this = NULL;
-
-        GF_ASSERT (req);
+        int                     ret = -1;
+        glusterd_req_ctx_t      *req_ctx = NULL;
+        int32_t                 status = 0;
+        dict_t                  *rsp_dict  = NULL;
+        char                    *op_errstr = NULL;
+        dict_t                  *dict = NULL;
+        xlator_t                *this = NULL;
+        uuid_t                  *txn_id = NULL;
 
         this = THIS;
         GF_ASSERT (this);
-        priv = this->private;
-        GF_ASSERT (priv);
+        GF_ASSERT (ctx);
 
-        strncpy (volname, req->buf.buf_val, req->buf.buf_len);
+        req_ctx = ctx;
 
-        ret  = glusterd_volinfo_find (volname, &volinfo);
+        dict = req_ctx->dict;
 
-        if (ret)
-                goto out;
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get new dictionary");
+                return -1;
+        }
 
-        ret = glusterd_store_delete_volume (volinfo);
+        status = glusterd_op_stage_validate (req_ctx->op, dict, &op_errstr,
+                                             rsp_dict);
 
-        if (ret)
-                goto out;
+        if (status) {
+                gf_log (this->name, GF_LOG_ERROR, "Stage failed on operation"
+                        " 'Volume %s', Status : %d", gd_op_list[req_ctx->op],
+                        status);
+        }
 
-        ret = glusterd_volinfo_delete (volinfo);
+        txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t);
 
+        if (txn_id)
+                uuid_copy (*txn_id, event->txn_id);
+        else
+                gf_log (this->name, GF_LOG_ERROR, "Out of Memory");
+
+        ret = dict_set_bin (rsp_dict, "transaction_id",
+                            txn_id, sizeof(*txn_id));
         if (ret)
-                goto out;
+                gf_log (this->name, GF_LOG_ERROR,
+                       "Failed to set transaction id.");
+
+        ret = glusterd_op_stage_send_resp (req_ctx->req, req_ctx->op,
+                                           status, op_errstr, rsp_dict);
+
+        if (op_errstr && (strcmp (op_errstr, "")))
+                GF_FREE (op_errstr);
+
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
+
+        if (rsp_dict)
+                dict_unref (rsp_dict);
 
-out:
         return ret;
 }
 
+static gf_boolean_t
+glusterd_need_brick_op (glusterd_op_t op)
+{
+        gf_boolean_t ret        = _gf_false;
+
+        GF_ASSERT (GD_OP_NONE < op && op < GD_OP_MAX);
+
+        switch (op) {
+        case GD_OP_PROFILE_VOLUME:
+        case GD_OP_STATUS_VOLUME:
+        case GD_OP_DEFRAG_BRICK_VOLUME:
+        case GD_OP_HEAL_VOLUME:
+                ret = _gf_true;
+                break;
+        default:
+                ret = _gf_false;
+        }
+
+        return ret;
+}
+
+dict_t*
+glusterd_op_init_commit_rsp_dict (glusterd_op_t op)
+{
+        dict_t                  *rsp_dict = NULL;
+        dict_t                  *op_ctx   = NULL;
+
+        GF_ASSERT (GD_OP_NONE < op && op < GD_OP_MAX);
+
+        if (glusterd_need_brick_op (op)) {
+                op_ctx = glusterd_op_get_ctx ();
+                GF_ASSERT (op_ctx);
+                rsp_dict = dict_ref (op_ctx);
+        } else {
+                rsp_dict = dict_new ();
+        }
+
+        return rsp_dict;
+}
+
 static int
-glusterd_op_start_volume (gd1_mgmt_stage_op_req *req, char **op_errstr)
+glusterd_op_ac_commit_op (glusterd_op_sm_event_t *event, void *ctx)
 {
-        int                                     ret = 0;
-        char                                    volname[1024] = {0,};
-        glusterd_volinfo_t                      *volinfo = NULL;
-        glusterd_conf_t                         *priv = NULL;
-        glusterd_brickinfo_t                    *brickinfo = NULL;
-        xlator_t                                *this = NULL;
-        int32_t                                 mybrick = 0;
+        int                       ret        = 0;
+        glusterd_req_ctx_t       *req_ctx    = NULL;
+        int32_t                   status     = 0;
+        char                     *op_errstr  = NULL;
+        dict_t                   *dict       = NULL;
+        dict_t                   *rsp_dict   = NULL;
+        xlator_t                 *this       = NULL;
+        uuid_t                   *txn_id     = NULL;
 
         this = THIS;
         GF_ASSERT (this);
-        priv = this->private;
-        GF_ASSERT (priv);
-        GF_ASSERT (req);
+        GF_ASSERT (ctx);
 
-        strncpy (volname, req->buf.buf_val, req->buf.buf_len);
+        req_ctx = ctx;
 
-        ret  = glusterd_volinfo_find (volname, &volinfo);
+        dict = req_ctx->dict;
 
-        if (ret)
-                goto out;
-        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
-                if (uuid_is_null (brickinfo->uuid)) {
-                        ret = glusterd_resolve_brick (brickinfo);
-                        if (ret) {
-                                gf_log ("glusterd", GF_LOG_ERROR,
-                                        "cannot resolve brick: %s:%s",
-                                        brickinfo->hostname, brickinfo->path);
-                                goto out;
-                        }
+        rsp_dict = glusterd_op_init_commit_rsp_dict (req_ctx->op);
+        if (NULL == rsp_dict)
+                return -1;
 
-                }
 
-                if (!uuid_compare (brickinfo->uuid, priv->uuid)) {
-                        gf_log ("", GF_LOG_NORMAL, "About to start glusterfs"
-                                " for brick %s:%s", brickinfo->hostname,
-                                brickinfo->path);
-                        ret = glusterd_volume_start_glusterfs
-                                                (volinfo, brickinfo, mybrick);
-                        if (ret) {
-                                gf_log ("", GF_LOG_ERROR, "Unable to start "
-                                        "glusterfs, ret: %d", ret);
-                                goto out;
-                        }
-                        mybrick++;
-                }
+        if (GD_OP_CLEARLOCKS_VOLUME == req_ctx->op) {
+                /*clear locks should be run only on
+                 * originator glusterd*/
+                status = 0;
+
+        } else {
+                status = glusterd_op_commit_perform (req_ctx->op, dict,
+                                                     &op_errstr, rsp_dict);
         }
 
-        glusterd_set_volume_status (volinfo, GLUSTERD_STATUS_STARTED);
+        if (status)
+                gf_log (this->name, GF_LOG_ERROR, "Commit of operation "
+                        "'Volume %s' failed: %d", gd_op_list[req_ctx->op],
+                        status);
 
-        ret = glusterd_store_update_volume (volinfo);
-        if (ret)
-                goto out;
+        txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t);
+
+        if (txn_id)
+                uuid_copy (*txn_id, event->txn_id);
+        else
+                gf_log (this->name, GF_LOG_ERROR, "Out of Memory");
 
-        ret = glusterd_volume_compute_cksum (volinfo);
+        ret = dict_set_bin (rsp_dict, "transaction_id",
+                            txn_id, sizeof(*txn_id));
         if (ret)
-                goto out;
+                gf_log (this->name, GF_LOG_ERROR,
+                       "Failed to set transaction id.");
 
-        ret = glusterd_check_generate_start_nfs (volinfo);
+        ret = glusterd_op_commit_send_resp (req_ctx->req, req_ctx->op,
+                                            status, op_errstr, rsp_dict);
+
+        if (op_errstr && (strcmp (op_errstr, "")))
+                GF_FREE (op_errstr);
+
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
 
-out:
         return ret;
 }
 
 static int
-glusterd_op_log_filename (gd1_mgmt_stage_op_req *req)
+glusterd_op_ac_send_commit_failed (glusterd_op_sm_event_t *event, void *ctx)
 {
-        int                   ret                = -1;
-        dict_t               *dict               = NULL;
-        glusterd_conf_t      *priv               = NULL;
-        glusterd_volinfo_t   *volinfo            = NULL;
-        glusterd_brickinfo_t *brickinfo          = NULL;
-        xlator_t             *this               = NULL;
-        char                 *volname            = NULL;
-        char                 *brick              = NULL;
-        char                 *path               = NULL;
-        char                  logfile[PATH_MAX]  = {0,};
-        char                  exp_path[PATH_MAX] = {0,};
-        struct stat           stbuf              = {0,};
+        int                             ret = 0;
+        glusterd_req_ctx_t              *req_ctx = NULL;
+        dict_t                          *op_ctx = NULL;
 
-        GF_ASSERT (req);
+        GF_ASSERT (ctx);
 
-        this = THIS;
-        GF_ASSERT (this);
-        priv = this->private;
-        GF_ASSERT (priv);
+        req_ctx = ctx;
 
-        dict = dict_new ();
-        if (!dict) {
-                gf_log ("", GF_LOG_ERROR, "ENOMEM, !dict");
-                goto out;
-        }
+        op_ctx = glusterd_op_get_ctx ();
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
-                goto out;
-        }
+        ret = glusterd_op_commit_send_resp (req_ctx->req, req_ctx->op,
+                                            opinfo.op_ret, opinfo.op_errstr,
+                                            op_ctx);
 
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "volname not found");
-                goto out;
-        }
-        ret = dict_get_str (dict, "path", &path);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "path not found");
-                goto out;
+        if (opinfo.op_errstr && (strcmp (opinfo.op_errstr, ""))) {
+                GF_FREE (opinfo.op_errstr);
+                opinfo.op_errstr = NULL;
         }
 
-        ret = stat (path, &stbuf);
-        if (!S_ISDIR (stbuf.st_mode)) {
-                ret = -1;
-                gf_log ("", GF_LOG_ERROR, "not a directory");
-                goto out;
-        }
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
 
-        ret = dict_get_str (dict, "brick", &brick);
-        if (ret)
-                goto out;
+static int
+glusterd_op_sm_transition_state (glusterd_op_info_t *opinfo,
+                                 glusterd_op_sm_t *state,
+                                 glusterd_op_sm_event_type_t event_type)
+{
+        glusterd_conf_t         *conf = NULL;
 
-        if (!strchr (brick, ':'))
-                brick = NULL;
+        GF_ASSERT (state);
+        GF_ASSERT (opinfo);
 
-        ret  = glusterd_volinfo_find (volname, &volinfo);
-        if (ret)
-                goto out;
+        conf = THIS->private;
+        GF_ASSERT (conf);
 
-        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
-                if (uuid_compare (brickinfo->uuid, priv->uuid))
-                        continue;
+        (void) glusterd_sm_tr_log_transition_add (&conf->op_sm_log,
+                                           opinfo->state.state,
+                                           state[event_type].next_state,
+                                           event_type);
 
-                if (brick && strcmp (brickinfo->path, brick))
-                        continue;
+        opinfo->state.state = state[event_type].next_state;
+        return 0;
+}
 
-                GLUSTERD_REMOVE_SLASH_FROM_PATH (brickinfo->path, exp_path);
+int32_t
+glusterd_op_stage_validate (glusterd_op_t op, dict_t *dict, char **op_errstr,
+                            dict_t *rsp_dict)
+{
+        int ret = -1;
+        xlator_t *this = THIS;
 
-                snprintf (logfile, PATH_MAX, "%s/%s.log", path, exp_path);
+        switch (op) {
+                case GD_OP_CREATE_VOLUME:
+                        ret = glusterd_op_stage_create_volume (dict, op_errstr);
+                        break;
 
-                if (brickinfo->logfile)
-                        GF_FREE (brickinfo->logfile);
-                brickinfo->logfile = gf_strdup (logfile);
-        }
+                case GD_OP_START_VOLUME:
+                        ret = glusterd_op_stage_start_volume (dict, op_errstr);
+                        break;
 
-        ret = 0;
+                case GD_OP_STOP_VOLUME:
+                        ret = glusterd_op_stage_stop_volume (dict, op_errstr);
+                        break;
 
-out:
-        if (dict)
-                dict_unref (dict);
+                case GD_OP_DELETE_VOLUME:
+                        ret = glusterd_op_stage_delete_volume (dict, op_errstr);
+                        break;
+
+                case GD_OP_ADD_BRICK:
+                        ret = glusterd_op_stage_add_brick (dict, op_errstr);
+                        break;
+
+                case GD_OP_REPLACE_BRICK:
+                        ret = glusterd_op_stage_replace_brick (dict, op_errstr,
+                                                               rsp_dict);
+                        break;
+
+                case GD_OP_SET_VOLUME:
+                        ret = glusterd_op_stage_set_volume (dict, op_errstr);
+                        break;
+
+                case GD_OP_RESET_VOLUME:
+                        ret = glusterd_op_stage_reset_volume (dict, op_errstr);
+                        break;
+
+                case GD_OP_REMOVE_BRICK:
+                        ret = glusterd_op_stage_remove_brick (dict, op_errstr);
+                        break;
+
+                case GD_OP_LOG_ROTATE:
+                        ret = glusterd_op_stage_log_rotate (dict, op_errstr);
+                        break;
+
+                case GD_OP_SYNC_VOLUME:
+                        ret = glusterd_op_stage_sync_volume (dict, op_errstr);
+                        break;
+
+                case GD_OP_GSYNC_CREATE:
+                        ret = glusterd_op_stage_gsync_create (dict, op_errstr);
+                        break;
+
+                case GD_OP_GSYNC_SET:
+                        ret = glusterd_op_stage_gsync_set (dict, op_errstr);
+                        break;
+
+                case GD_OP_PROFILE_VOLUME:
+                        ret = glusterd_op_stage_stats_volume (dict, op_errstr);
+                        break;
+
+                case GD_OP_QUOTA:
+                        ret = glusterd_op_stage_quota (dict, op_errstr);
+                        break;
+
+                case GD_OP_STATUS_VOLUME:
+                        ret = glusterd_op_stage_status_volume (dict, op_errstr);
+                        break;
+
+                case GD_OP_REBALANCE:
+                case GD_OP_DEFRAG_BRICK_VOLUME:
+                        ret = glusterd_op_stage_rebalance (dict, op_errstr);
+                        break;
+
+                case GD_OP_HEAL_VOLUME:
+                        ret = glusterd_op_stage_heal_volume (dict, op_errstr);
+                        break;
+
+                case GD_OP_STATEDUMP_VOLUME:
+                        ret = glusterd_op_stage_statedump_volume (dict,
+                                                                  op_errstr);
+                        break;
+                case GD_OP_CLEARLOCKS_VOLUME:
+                        ret = glusterd_op_stage_clearlocks_volume (dict,
+                                                                   op_errstr);
+                        break;
+
+                case GD_OP_COPY_FILE:
+                        ret = glusterd_op_stage_copy_file (dict, op_errstr);
+                        break;
+
+                case GD_OP_SYS_EXEC:
+                        ret = glusterd_op_stage_sys_exec (dict, op_errstr);
+                        break;
+
+                default:
+                        gf_log (this->name, GF_LOG_ERROR, "Unknown op %s",
+                                gd_op_list[op]);
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "OP = %d. Returning %d", op, ret);
         return ret;
 }
 
-static int
-glusterd_op_log_rotate (gd1_mgmt_stage_op_req *req)
+
+int32_t
+glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr,
+                            dict_t *rsp_dict)
 {
-        int                   ret                = -1;
-        dict_t               *dict               = NULL;
-        glusterd_conf_t      *priv               = NULL;
-        glusterd_volinfo_t   *volinfo            = NULL;
-        glusterd_brickinfo_t *brickinfo          = NULL;
-        xlator_t             *this               = NULL;
-        char                 *volname            = NULL;
-        char                 *brick              = NULL;
-        char                  path[PATH_MAX]     = {0,};
-        char                  logfile[PATH_MAX]  = {0,};
-        char                  pidfile[PATH_MAX]  = {0,};
-        FILE                 *file               = NULL;
-        pid_t                 pid                = 0;
-        uint64_t              key                = 0;
+        int ret = -1;
+        xlator_t *this = THIS;
 
-        GF_ASSERT (req);
+        glusterd_op_commit_hook (op, dict, GD_COMMIT_HOOK_PRE);
+        switch (op) {
+                case GD_OP_CREATE_VOLUME:
+                        ret = glusterd_op_create_volume (dict, op_errstr);
+                        break;
 
-        this = THIS;
-        GF_ASSERT (this);
-        priv = this->private;
-        GF_ASSERT (priv);
+                case GD_OP_START_VOLUME:
+                        ret = glusterd_op_start_volume (dict, op_errstr);
+                        break;
 
-        dict = dict_new ();
-        if (!dict) {
-                gf_log ("", GF_LOG_ERROR, "ENOMEM, !dict");
-                goto out;
-        }
+                case GD_OP_STOP_VOLUME:
+                        ret = glusterd_op_stop_volume (dict);
+                        break;
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
-                goto out;
-        }
+                case GD_OP_DELETE_VOLUME:
+                        ret = glusterd_op_delete_volume (dict);
+                        break;
 
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "volname not found");
-                goto out;
-        }
+                case GD_OP_ADD_BRICK:
+                        ret = glusterd_op_add_brick (dict, op_errstr);
+                        break;
 
-        ret = dict_get_uint64 (dict, "rotate-key", &key);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "rotate key not found");
-                goto out;
-        }
+                case GD_OP_REPLACE_BRICK:
+                        ret = glusterd_op_replace_brick (dict, rsp_dict);
+                        break;
 
-        ret = dict_get_str (dict, "brick", &brick);
-        if (ret)
-                goto out;
+                case GD_OP_SET_VOLUME:
+                        ret = glusterd_op_set_volume (dict);
+                        break;
 
-        if (!strchr (brick, ':'))
-                brick = NULL;
+                case GD_OP_RESET_VOLUME:
+                        ret = glusterd_op_reset_volume (dict, op_errstr);
+                        break;
 
-        ret = glusterd_volinfo_find (volname, &volinfo);
-        if (ret)
-                goto out;
+                case GD_OP_REMOVE_BRICK:
+                        ret = glusterd_op_remove_brick (dict, op_errstr);
+                        break;
 
-        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
-                if (uuid_compare (brickinfo->uuid, priv->uuid))
-                        continue;
+                case GD_OP_LOG_ROTATE:
+                        ret = glusterd_op_log_rotate (dict);
+                        break;
 
-                if (brick && strcmp (brickinfo->path, brick))
-                        continue;
+                case GD_OP_SYNC_VOLUME:
+                        ret = glusterd_op_sync_volume (dict, op_errstr, rsp_dict);
+                        break;
 
-                GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv);
-                GLUSTERD_GET_BRICK_PIDFILE (pidfile, path, brickinfo->hostname,
-                                            brickinfo->path);
+                case GD_OP_GSYNC_CREATE:
+                        ret = glusterd_op_gsync_create (dict, op_errstr,
+                                                        rsp_dict);
+                        break;
 
-                file = fopen (pidfile, "r+");
-                if (!file) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to open pidfile: %s",
-                                pidfile);
-                        ret = -1;
-                        goto out;
-                }
+                case GD_OP_GSYNC_SET:
+                        ret = glusterd_op_gsync_set (dict, op_errstr, rsp_dict);
+                        break;
 
-                ret = fscanf (file, "%d", &pid);
-                if (ret <= 0) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to read pidfile: %s",
-                                pidfile);
-                        ret = -1;
-                        goto out;
-                }
-                fclose (file);
-                file = NULL;
+                case GD_OP_PROFILE_VOLUME:
+                        ret = glusterd_op_stats_volume (dict, op_errstr,
+                                                        rsp_dict);
+                        break;
 
-                snprintf (logfile, PATH_MAX, "%s.%"PRIu64,
-                          brickinfo->logfile, key);
+                case GD_OP_QUOTA:
+                        ret = glusterd_op_quota (dict, op_errstr, rsp_dict);
+                        break;
 
-                ret = rename (brickinfo->logfile, logfile);
-                if (ret)
-                        gf_log ("", GF_LOG_WARNING, "rename failed");
+                case GD_OP_STATUS_VOLUME:
+                        ret = glusterd_op_status_volume (dict, op_errstr, rsp_dict);
+                        break;
 
-                ret = kill (pid, SIGHUP);
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to SIGHUP to %d", pid);
-                        goto out;
-                }
+                case GD_OP_REBALANCE:
+                case GD_OP_DEFRAG_BRICK_VOLUME:
+                        ret = glusterd_op_rebalance (dict, op_errstr, rsp_dict);
+                        break;
+
+                case GD_OP_HEAL_VOLUME:
+                        ret = glusterd_op_heal_volume (dict, op_errstr);
+                        break;
+
+                case GD_OP_STATEDUMP_VOLUME:
+                        ret = glusterd_op_statedump_volume (dict, op_errstr);
+                        break;
+
+                case GD_OP_CLEARLOCKS_VOLUME:
+                        ret = glusterd_op_clearlocks_volume (dict, op_errstr,
+                                                             rsp_dict);
+                        break;
+
+                case GD_OP_COPY_FILE:
+                        ret = glusterd_op_copy_file (dict, op_errstr);
+                        break;
+
+                case GD_OP_SYS_EXEC:
+                        ret = glusterd_op_sys_exec (dict, op_errstr, rsp_dict);
+                        break;
+
+                default:
+                        gf_log (this->name, GF_LOG_ERROR, "Unknown op %s",
+                                gd_op_list[op]);
+                        break;
         }
 
-        ret = 0;
+        if (ret == 0)
+            glusterd_op_commit_hook (op, dict, GD_COMMIT_HOOK_POST);
 
-out:
-        if (dict)
-                dict_unref (dict);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
+
 static int
-glusterd_op_stop_volume (gd1_mgmt_stage_op_req *req)
+glusterd_bricks_select_stop_volume (dict_t *dict, char **op_errstr,
+                                    struct list_head *selected)
 {
         int                                     ret = 0;
         int                                     flags = 0;
         char                                    *volname = NULL;
-        glusterd_conf_t                         *priv = NULL;
         glusterd_volinfo_t                      *volinfo = NULL;
-        xlator_t                                *this = NULL;
-        dict_t                                  *dict = NULL;
         glusterd_brickinfo_t                    *brickinfo = NULL;
-        int32_t                                 mybrick = 0;
-
-        this = THIS;
-        GF_ASSERT (this);
-        priv = this->private;
-        GF_ASSERT (priv);
-
-        dict = dict_new ();
-        if (!dict)
-                goto out;
+        glusterd_pending_node_t                 *pending_node = NULL;
 
-        ret = glusterd_op_stop_volume_args_get (req, dict, &volname, &flags);
+        ret = glusterd_op_stop_volume_args_get (dict, &volname, &flags);
         if (ret)
                 goto out;
 
         ret  = glusterd_volinfo_find (volname, &volinfo);
-
-        if (ret)
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS,
+                        volname);
+                gf_asprintf (op_errstr, FMTSTR_CHECK_VOL_EXISTS, volname);
                 goto out;
+        }
 
         list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
-                if (uuid_is_null (brickinfo->uuid)) {
-                        ret = glusterd_resolve_brick (brickinfo);
-                        if (ret) {
-                                gf_log ("glusterd", GF_LOG_ERROR,
-                                        "cannot resolve brick: %s:%s",
-                                        brickinfo->hostname, brickinfo->path);
-                                goto out;
-                        }
-
-                }
-                if (!uuid_compare (brickinfo->uuid, priv->uuid)) {
-                        gf_log ("", GF_LOG_NORMAL, "About to stop glusterfs"
-                                " for brick %s:%s", brickinfo->hostname,
-                                brickinfo->path);
-                        ret = glusterd_volume_stop_glusterfs
-                                (volinfo, brickinfo, mybrick);
-                        if (ret) {
-                                gf_log ("", GF_LOG_ERROR, "Unable to stop "
-                                        "glusterfs, ret: %d", ret);
+                if (glusterd_is_brick_started (brickinfo)) {
+                        pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                                  gf_gld_mt_pending_node_t);
+                        if (!pending_node) {
+                                ret = -1;
                                 goto out;
+                        } else {
+                                pending_node->node = brickinfo;
+                                pending_node->type = GD_NODE_BRICK;
+                                list_add_tail (&pending_node->list, selected);
+                                pending_node = NULL;
                         }
-                        mybrick++;
                 }
         }
 
-        glusterd_set_volume_status (volinfo, GLUSTERD_STATUS_STOPPED);
-
-        ret = glusterd_store_update_volume (volinfo);
-        if (ret)
-                goto out;
-
-        ret = glusterd_volume_compute_cksum (volinfo);
-
-        if (glusterd_are_all_volumes_stopped ()) {
-                if (glusterd_is_nfs_started ()) {
-                        ret = glusterd_nfs_server_stop ();
-                        if (ret)
-                                goto out;
-                }
-        } else {
-                ret = glusterd_check_generate_start_nfs (volinfo);
-        }
 out:
-        if (flags & GF_CLI_FLAG_OP_FORCE)
-                ret = 0;
-        if (dict)
-                dict_unref (dict);
         return ret;
 }
 
 static int
-glusterd_op_sync_volume (gd1_mgmt_stage_op_req *req, char **op_errstr,
-                         dict_t *rsp_dict)
+glusterd_bricks_select_remove_brick (dict_t *dict, char **op_errstr,
+                                     struct list_head *selected)
 {
         int                                     ret = -1;
-        dict_t                                  *dict = NULL;
         char                                    *volname = NULL;
-        char                                    *hostname = NULL;
-        char                                    msg[2048] = {0,};
-        int                                     count = 1;
-        int                                     vol_count = 0;
-        glusterd_conf_t                         *priv = NULL;
         glusterd_volinfo_t                      *volinfo = NULL;
-        xlator_t                                *this = NULL;
+        glusterd_brickinfo_t                    *brickinfo = NULL;
+        char                                    *brick = NULL;
+        int32_t                                 count = 0;
+        int32_t                                 i = 1;
+        char                                    key[256] = {0,};
+        glusterd_pending_node_t                 *pending_node = NULL;
+        int32_t                                 force = 0;
+
 
-        GF_ASSERT (req);
-        this = THIS;
-        GF_ASSERT (this);
-        priv = this->private;
-        GF_ASSERT (priv);
 
+        ret = dict_get_str (dict, "volname", &volname);
 
-        dict = dict_new ();
-        if (!dict)
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
                 goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
 
-        ret = dict_unserialize (req->buf.buf_val, req->buf.buf_len, &dict);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unserialize dict");
+                gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
                 goto out;
         }
 
-        ret = dict_get_str (dict, "hostname", &hostname);
+        ret = dict_get_int32 (dict, "count", &count);
         if (ret) {
-                snprintf (msg, sizeof (msg), "hostname couldn't be "
-                          "retrieved from msg");
-                *op_errstr = gf_strdup (msg);
+                gf_log ("", GF_LOG_ERROR, "Unable to get count");
                 goto out;
         }
 
-        if (glusterd_is_local_addr (hostname)) {
+        ret = dict_get_int32 (dict, "force", &force);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_INFO, "force flag is not set");
                 ret = 0;
                 goto out;
         }
 
-        //volname is not present in case of sync all
-        ret = dict_get_str (dict, "volname", &volname);
-        if (!ret) {
-                ret = glusterd_volinfo_find (volname, &volinfo);
+        while ( i <= count) {
+                snprintf (key, 256, "brick%d", i);
+                ret = dict_get_str (dict, key, &brick);
                 if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Volume with name: %s "
-                                "not exists", volname);
+                        gf_log ("glusterd", GF_LOG_ERROR, "Unable to get brick");
                         goto out;
                 }
-        }
-
-        if (!rsp_dict) {
-                //this should happen only on source
-                ret = 0;
-                goto out;
-        }
 
-        if (volname) {
-                ret = glusterd_add_volume_to_dict (volinfo, rsp_dict,
-                                                   1);
-                vol_count = 1;
-        } else {
-                list_for_each_entry (volinfo, &priv->volumes, vol_list) {
-                        ret = glusterd_add_volume_to_dict (volinfo,
-                                                           rsp_dict, count);
-                        if (ret)
+                ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo,
+                                                              &brickinfo);
+                if (ret)
+                        goto out;
+                if (glusterd_is_brick_started (brickinfo)) {
+                        pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                                  gf_gld_mt_pending_node_t);
+                        if (!pending_node) {
+                                ret = -1;
                                 goto out;
-
-                        vol_count = count++;
+                        } else {
+                                pending_node->node = brickinfo;
+                                pending_node->type = GD_NODE_BRICK;
+                                list_add_tail (&pending_node->list, selected);
+                                pending_node = NULL;
+                        }
                 }
+                i++;
         }
-        ret = dict_set_int32 (rsp_dict, "count", vol_count);
-out:
-        if (dict)
-                dict_unref (dict);
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
 
+out:
         return ret;
 }
 
 static int
-glusterd_op_ac_none (glusterd_op_sm_event_t *event, void *ctx)
+glusterd_bricks_select_profile_volume (dict_t *dict, char **op_errstr,
+                                       struct list_head *selected)
 {
-        int ret = 0;
-
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        int                                     ret = -1;
+        char                                    *volname = NULL;
+        char                                    msg[2048] = {0,};
+        glusterd_conf_t                         *priv = NULL;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        xlator_t                                *this = NULL;
+        int32_t                                 stats_op = GF_CLI_STATS_NONE;
+        glusterd_brickinfo_t                    *brickinfo = NULL;
+        glusterd_pending_node_t                 *pending_node = NULL;
+        char                                    *brick = NULL;
 
-        return ret;
-}
 
-static int
-glusterd_op_ac_send_lock (glusterd_op_sm_event_t *event, void *ctx)
-{
-        int                     ret = 0;
-        rpc_clnt_procedure_t    *proc = NULL;
-        glusterd_conf_t         *priv = NULL;
-        xlator_t                *this = NULL;
 
         this = THIS;
+        GF_ASSERT (this);
         priv = this->private;
+        GF_ASSERT (priv);
 
-        proc = &priv->mgmt->proctable[GD_MGMT_CLUSTER_LOCK];
-        if (proc->fn) {
-                ret = proc->fn (NULL, this, NULL);
-                if (ret)
-                        goto out;
-        }
-
-        if (!opinfo.pending_count)
-                ret = glusterd_op_sm_inject_all_acc ();
-
-out:
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
-
-        return ret;
-}
 
-static int
-glusterd_op_ac_send_unlock (glusterd_op_sm_event_t *event, void *ctx)
-{
-        int                     ret = 0;
-        rpc_clnt_procedure_t    *proc = NULL;
-        glusterd_conf_t         *priv = NULL;
-        xlator_t                *this = NULL;
-
-        this = THIS;
-        priv = this->private;
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR, "volume name get failed");
+                goto out;
+        }
 
-        /*ret = glusterd_unlock (priv->uuid);
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Volume %s does not exists",
+                          volname);
 
-        if (ret)
+                *op_errstr = gf_strdup (msg);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
                 goto out;
-        */
-
-        proc = &priv->mgmt->proctable[GD_MGMT_CLUSTER_UNLOCK];
-        if (proc->fn) {
-                ret = proc->fn (NULL, this, NULL);
-                if (ret)
-                        goto out;
         }
 
-        if (!opinfo.pending_count)
-                ret = glusterd_op_sm_inject_all_acc ();
+        ret = dict_get_int32 (dict, "op", &stats_op);
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR, "volume profile op get failed");
+                goto out;
+        }
 
-out:
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        switch (stats_op) {
+        case GF_CLI_STATS_START:
+        case GF_CLI_STATS_STOP:
+                goto out;
+                break;
+        case GF_CLI_STATS_INFO:
+                ret = dict_get_str_boolean (dict, "nfs", _gf_false);
+                if (ret) {
+                        if (!glusterd_is_nodesvc_online ("nfs")) {
+                                ret = -1;
+                                gf_log (this->name, GF_LOG_ERROR, "NFS server"
+                                        " is not running");
+                                goto out;
+                        }
+                        pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                                  gf_gld_mt_pending_node_t);
+                        if (!pending_node) {
+                                ret = -1;
+                                goto out;
+                        }
+                        pending_node->node = priv->nfs;
+                        pending_node->type = GD_NODE_NFS;
+                        list_add_tail (&pending_node->list, selected);
+                        pending_node = NULL;
 
-        return ret;
+                        ret = 0;
+                        goto out;
 
-}
+                }
+                list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                        if (glusterd_is_brick_started (brickinfo)) {
+                                pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                                          gf_gld_mt_pending_node_t);
+                                if (!pending_node) {
+                                        ret = -1;
+                                        goto out;
+                                } else {
+                                        pending_node->node = brickinfo;
+                                        pending_node->type = GD_NODE_BRICK;
+                                        list_add_tail (&pending_node->list,
+                                                       selected);
+                                        pending_node = NULL;
+                                }
+                        }
+                }
+                break;
 
-static int
-glusterd_op_ac_lock (glusterd_op_sm_event_t *event, void *ctx)
-{
-        int                      ret = 0;
-        glusterd_op_lock_ctx_t   *lock_ctx = NULL;
-        int32_t                  status = 0;
+        case GF_CLI_STATS_TOP:
+                ret = dict_get_str_boolean (dict, "nfs", _gf_false);
+                if (ret) {
+                        if (!glusterd_is_nodesvc_online ("nfs")) {
+                                ret = -1;
+                                gf_log (this->name, GF_LOG_ERROR, "NFS server"
+                                        " is not running");
+                                goto out;
+                        }
+                        pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                                  gf_gld_mt_pending_node_t);
+                        if (!pending_node) {
+                                ret = -1;
+                                goto out;
+                        }
+                        pending_node->node = priv->nfs;
+                        pending_node->type = GD_NODE_NFS;
+                        list_add_tail (&pending_node->list, selected);
+                        pending_node = NULL;
 
+                        ret = 0;
+                        goto out;
 
-        GF_ASSERT (event);
-        GF_ASSERT (ctx);
+                }
+                ret = dict_get_str (dict, "brick", &brick);
+                if (!ret) {
+                        ret = glusterd_volume_brickinfo_get_by_brick (brick, volinfo,
+                                                                      &brickinfo);
+                        if (ret)
+                                goto out;
 
-        lock_ctx = (glusterd_op_lock_ctx_t *)ctx;
+                        if (!glusterd_is_brick_started (brickinfo))
+                                goto out;
 
-        status = glusterd_lock (lock_ctx->uuid);
+                        pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                                  gf_gld_mt_pending_node_t);
+                        if (!pending_node) {
+                                ret = -1;
+                                goto out;
+                        } else {
+                                pending_node->node = brickinfo;
+                                pending_node->type = GD_NODE_BRICK;
+                                list_add_tail (&pending_node->list,
+                                               selected);
+                                pending_node = NULL;
+                                goto out;
+                        }
+                }
+                ret = 0;
+                list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                        if (glusterd_is_brick_started (brickinfo)) {
+                                pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                                          gf_gld_mt_pending_node_t);
+                                if (!pending_node) {
+                                        ret = -1;
+                                        goto out;
+                                } else {
+                                        pending_node->node = brickinfo;
+                                        pending_node->type = GD_NODE_BRICK;
+                                        list_add_tail (&pending_node->list,
+                                                       selected);
+                                        pending_node = NULL;
+                                }
+                        }
+                }
+                break;
 
-        gf_log ("", GF_LOG_DEBUG, "Lock Returned %d", status);
+        default:
+                GF_ASSERT (0);
+                gf_log ("glusterd", GF_LOG_ERROR, "Invalid profile op: %d",
+                        stats_op);
+                ret = -1;
+                goto out;
+                break;
+        }
 
-        ret = glusterd_op_lock_send_resp (lock_ctx->req, status);
 
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
 
         return ret;
 }
 
 static int
-glusterd_op_ac_unlock (glusterd_op_sm_event_t *event, void *ctx)
+_add_rxlator_to_dict (dict_t *dict, char *volname, int index, int count)
 {
-        int ret = 0;
-        glusterd_op_lock_ctx_t   *lock_ctx = NULL;
-
-        GF_ASSERT (event);
-        GF_ASSERT (ctx);
-
-        lock_ctx = (glusterd_op_lock_ctx_t *)ctx;
+        int     ret             = -1;
+        char    key[128]        = {0,};
+        char    *xname          = NULL;
 
-        ret = glusterd_unlock (lock_ctx->uuid);
-
-        gf_log ("", GF_LOG_DEBUG, "Unlock Returned %d", ret);
-
-        ret = glusterd_op_unlock_send_resp (lock_ctx->req, ret);
+        snprintf (key, sizeof (key), "xl-%d", count);
+        ret = gf_asprintf (&xname, "%s-replicate-%d", volname, index);
+        if (ret == -1)
+                goto out;
 
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        ret = dict_set_dynstr (dict, key, xname);
+        if (ret)
+                goto out;
 
+        ret = dict_set_int32 (dict, xname, index);
+out:
         return ret;
 }
 
-static int
-glusterd_op_ac_rcvd_lock_acc (glusterd_op_sm_event_t *event, void *ctx)
-{
+int
+get_replica_index_for_per_replica_cmd (glusterd_volinfo_t *volinfo,
+                                       dict_t *dict) {
         int                     ret = 0;
+        char                    *hostname = NULL;
+        char                    *path = NULL;
+        int                     index = 0;
+        glusterd_brickinfo_t   *brickinfo = NULL;
+        int                     cmd_replica_index = -1;
+        int                     replica_count = -1;
 
-        GF_ASSERT (event);
-
-        opinfo.pending_count--;
 
-        if (opinfo.pending_count)
+        if (!dict)  {
+                ret = -1;
                 goto out;
+        }
 
-        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL);
-
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
-
-out:
-        return ret;
-}
-
-static int
-glusterd_op_ac_send_stage_op (glusterd_op_sm_event_t *event, void *ctx)
-{
-        int                     ret = 0;
-        rpc_clnt_procedure_t    *proc = NULL;
-        glusterd_conf_t         *priv = NULL;
-        xlator_t                *this = NULL;
+        ret = dict_get_str (dict, "per-replica-cmd-hostname", &hostname);
+        if (ret)
+                goto out;
+        ret = dict_get_str (dict, "per-replica-cmd-path", &path);
+        if (ret)
+                goto out;
 
-        this = THIS;
-        GF_ASSERT (this);
-        priv = this->private;
-        GF_ASSERT (priv);
-        GF_ASSERT (priv->mgmt);
+        replica_count = volinfo->replica_count;
 
-        proc = &priv->mgmt->proctable[GD_MGMT_STAGE_OP];
-        GF_ASSERT (proc);
-        if (proc->fn) {
-                ret = proc->fn (NULL, this, NULL);
-                if (ret)
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                if (uuid_is_null (brickinfo->uuid))
+                        (void)glusterd_resolve_brick (brickinfo);
+                if (!strcmp (brickinfo->path,  path) &&
+                    !strcmp (brickinfo->hostname, hostname)) {
+                        cmd_replica_index = index/(replica_count);
                         goto out;
+                }
+                index++;
         }
 
-        if (!opinfo.pending_count)
-                ret = glusterd_op_sm_inject_all_acc ();
 
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
-
-        return ret;
+        if (ret)
+                cmd_replica_index = -1;
 
+        return cmd_replica_index;
 }
 
-static int
-glusterd_op_ac_send_commit_op (glusterd_op_sm_event_t *event, void *ctx)
+int
+_select_rxlators_with_local_bricks (xlator_t *this, glusterd_volinfo_t *volinfo,
+                                    dict_t *dict, cli_cmd_type type)
 {
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        glusterd_conf_t         *priv   = NULL;
+        int                     index = 0;
+        int                     rxlator_count = 0;
+        int                     replica_count = 0;
+        gf_boolean_t            add     = _gf_false;
         int                     ret = 0;
-        rpc_clnt_procedure_t    *proc = NULL;
-        glusterd_conf_t         *priv = NULL;
-        xlator_t                *this = NULL;
+        int                     cmd_replica_index = -1;
 
-        this = THIS;
-        GF_ASSERT (this);
         priv = this->private;
-        GF_ASSERT (priv);
-        GF_ASSERT (priv->mgmt);
+        replica_count = volinfo->replica_count;
 
-        proc = &priv->mgmt->proctable[GD_MGMT_COMMIT_OP];
-        GF_ASSERT (proc);
-        if (proc->fn) {
-                ret = proc->fn (NULL, this, NULL);
-                if (!ret)
-                        goto out;
+        if (type == PER_REPLICA) {
+
+                cmd_replica_index = get_replica_index_for_per_replica_cmd
+                                    (volinfo, dict);
+                if (cmd_replica_index == -1) {
+                        ret = -1;
+                        goto err;
+                }
         }
 
-        if (!opinfo.pending_count)
-                ret = glusterd_op_sm_inject_all_acc ();
+        index = 1;
 
-out:
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                if (uuid_is_null (brickinfo->uuid))
+                        (void)glusterd_resolve_brick (brickinfo);
 
-        return ret;
+                switch (type) {
+                        case ALL_REPLICA:
+                                if (!uuid_compare (MY_UUID, brickinfo->uuid))
+                                        add = _gf_true;
+                                break;
+                        case PER_REPLICA:
+                                if (!uuid_compare (MY_UUID, brickinfo->uuid) &&
+                                 ((index-1)/replica_count == cmd_replica_index))
 
-}
+                                                add = _gf_true;
+                                break;
+                }
 
-static int
-glusterd_op_ac_rcvd_stage_op_acc (glusterd_op_sm_event_t *event, void *ctx)
-{
-        int                     ret = 0;
+                if (index % replica_count == 0) {
+                        if (add) {
+                                _add_rxlator_to_dict (dict, volinfo->volname,
+                                                      (index-1)/replica_count,
+                                                      rxlator_count);
+                                rxlator_count++;
+                        }
+                        add = _gf_false;
+                }
 
-        GF_ASSERT (event);
+                index++;
+        }
+err:
+        if (ret)
+                rxlator_count = -1;
 
-        opinfo.pending_count--;
+        return rxlator_count;
+}
 
-        if (opinfo.pending_count)
-                goto out;
+int
+_select_rxlators_for_full_self_heal (xlator_t *this,
+                                     glusterd_volinfo_t *volinfo,
+                                     dict_t *dict)
+{
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        glusterd_conf_t         *priv   = NULL;
+        int                     index = 1;
+        int                     rxlator_count = 0;
+        int                     replica_count = 0;
+        uuid_t                  candidate = {0};
 
-        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_ACC, NULL);
+        priv = this->private;
+        replica_count = volinfo->replica_count;
 
-out:
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                if (uuid_is_null (brickinfo->uuid))
+                        (void)glusterd_resolve_brick (brickinfo);
+
+                if (uuid_compare (brickinfo->uuid, candidate) > 0)
+                        uuid_copy (candidate, brickinfo->uuid);
+
+                if (index % replica_count == 0) {
+                        if (!uuid_compare (MY_UUID, candidate)) {
+                                _add_rxlator_to_dict (dict, volinfo->volname,
+                                                      (index-1)/replica_count,
+                                                      rxlator_count);
+                                rxlator_count++;
+                        }
+                        uuid_clear (candidate);
+                }
 
-        return ret;
+                index++;
+        }
+        return rxlator_count;
 }
 
+
 static int
-glusterd_op_ac_rcvd_commit_op_acc (glusterd_op_sm_event_t *event, void *ctx)
+glusterd_bricks_select_snap (dict_t *dict, char **op_errstr,
+                             struct list_head *selected)
 {
-        glusterd_volinfo_t     *volinfo = NULL;
-        int32_t                 op      = 0;
-        int32_t                 src_port = 0;
-        dict_t                 *dict    = NULL;
-        char                   *src_brick = NULL;
-        char                   *dst_brick = NULL;
-        char                   *volname   = NULL;
-        glusterd_brickinfo_t   *src_brickinfo = NULL;
-        glusterd_brickinfo_t   *dst_brickinfo = NULL;
-        int                     ret     = 0;
+        int                        ret           = -1;
+        glusterd_conf_t            *priv         = NULL;
+        xlator_t                   *this         = NULL;
+        glusterd_pending_node_t    *pending_node = NULL;
+        glusterd_volinfo_t         *volinfo      = NULL;
+        char                       *volname      = NULL;
+        glusterd_brickinfo_t       *brickinfo    = NULL;
+        int                         brick_index  = -1;
 
-        GF_ASSERT (event);
-
-        opinfo.pending_count--;
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
 
-        if (opinfo.pending_count)
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get"
+                        " volname");
+                goto out;
+        }
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret)
                 goto out;
 
-        dict = glusterd_op_get_ctx (GD_OP_REPLACE_BRICK);
-        if (dict) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Replace brick operation detected");
-
-                ret = dict_get_int32 (dict, "operation", &op);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "dict_get on operation failed");
-                        goto out;
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                brick_index++;
+                if (uuid_compare (brickinfo->uuid, MY_UUID) ||
+                    !glusterd_is_brick_started (brickinfo)) {
+                        continue;
                 }
-                ret = dict_get_str (dict, "src-brick", &src_brick);
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to get src brick");
+                pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                          gf_gld_mt_pending_node_t);
+                if (!pending_node) {
+                        ret = -1;
                         goto out;
                 }
+                pending_node->node = brickinfo;
+                pending_node->type = GD_NODE_BRICK;
+                pending_node->index = brick_index;
+                list_add_tail (&pending_node->list,
+                               selected);
+                pending_node = NULL;
+        }
 
-                gf_log ("", GF_LOG_DEBUG,
-                        "src brick=%s", src_brick);
-
-                ret = dict_get_str (dict, "dst-brick", &dst_brick);
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to get dst brick");
-                        goto out;
-                }
+        ret = 0;
 
-                gf_log ("", GF_LOG_DEBUG,
-                        "dst brick=%s", dst_brick);
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning ret %d", ret);
+        return ret;
+}
 
-                ret = dict_get_str (dict, "volname", &volname);
+static int
+fill_shd_status_for_local_bricks (dict_t *dict, glusterd_volinfo_t *volinfo,
+                                  cli_cmd_type type, dict_t *req_dict)
+{
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        char                    msg[1024] = {0,};
+        char                    key[1024]  = {0,};
+        char                    value[1024] = {0,};
+        int                     index = 0;
+        int                     ret = 0;
+        xlator_t               *this = NULL;
+        int                     cmd_replica_index = -1;
 
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+        this = THIS;
+        snprintf (msg, sizeof (msg), "self-heal-daemon is not running on");
+
+        if (type == PER_REPLICA) {
+                cmd_replica_index = get_replica_index_for_per_replica_cmd
+                                    (volinfo, req_dict);
+                if (cmd_replica_index == -1) {
+                        gf_log (THIS->name, GF_LOG_ERROR, "Could not find the "
+                                "replica index for per replica type command");
+                        ret = -1;
                         goto out;
                 }
+        }
 
-                ret = glusterd_volinfo_find (volname, &volinfo);
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
-                        goto out;
-                }
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                if (uuid_is_null (brickinfo->uuid))
+                        (void)glusterd_resolve_brick (brickinfo);
 
-                ret = glusterd_brickinfo_get (src_brick, volinfo, &src_brickinfo);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG, "Unable to get src-brickinfo");
-                        goto out;
+                if (uuid_compare (MY_UUID, brickinfo->uuid)) {
+                        index++;
+                        continue;
                 }
 
-                ret = glusterd_brickinfo_from_brick (dst_brick, &dst_brickinfo);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG, "Unable to get dst-brickinfo");
-                        goto out;
-                }
+                if (type == PER_REPLICA) {
+                      if (cmd_replica_index != (index/volinfo->replica_count)) {
+                              index++;
+                              continue;
+                        }
 
-                ret = dict_get_int32 (dict, "src-brick-port", &src_port);
+                }
+                snprintf (key, sizeof (key), "%d-status",index);
+                snprintf (value, sizeof (value), "%s %s",msg,
+                          uuid_utoa(MY_UUID));
+                ret = dict_set_dynstr (dict, key, gf_strdup(value));
                 if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to get src-brick port");
+                        gf_log (this->name, GF_LOG_ERROR, "Unable to"
+                                "set the dictionary for shd status msg");
                         goto out;
                 }
-
-                src_brickinfo->port = src_port;
-
-                switch (op) {
-                case GF_REPLACE_OP_START:
-                        ret = rb_do_operation_start (volinfo, src_brickinfo, dst_brickinfo);
-                        break;
-                case GF_REPLACE_OP_COMMIT:
-                        ret = rb_do_operation_commit (volinfo, src_brickinfo, dst_brickinfo);
-                        break;
-                case GF_REPLACE_OP_PAUSE:
-                        ret = rb_do_operation_pause (volinfo, src_brickinfo, dst_brickinfo);
-                        break;
-                case GF_REPLACE_OP_ABORT:
-                        ret = rb_do_operation_abort (volinfo, src_brickinfo, dst_brickinfo);
-                        break;
-                case GF_REPLACE_OP_STATUS:
-                        ret = rb_do_operation_status (volinfo, src_brickinfo, dst_brickinfo);
-                        break;
-                default:
-                        ret = -1;
+                snprintf (key, sizeof (key), "%d-shd-status",index);
+                ret = dict_set_str (dict, key, "off");
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Unable to"
+                                " set dictionary for shd status msg");
                         goto out;
                 }
 
+                index++;
         }
 
-        if (ret)
-                ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL);
-        else
-                ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC, NULL);
-
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
-
 out:
         return ret;
+
 }
 
+
 static int
-glusterd_op_ac_rcvd_unlock_acc (glusterd_op_sm_event_t *event, void *ctx)
+glusterd_bricks_select_heal_volume (dict_t *dict, char **op_errstr,
+                                    struct list_head *selected,
+                                    dict_t *rsp_dict)
 {
-        int                     ret = 0;
-
-        GF_ASSERT (event);
+        int                                     ret = -1;
+        char                                    *volname = NULL;
+        glusterd_conf_t                         *priv = NULL;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        xlator_t                                *this = NULL;
+        char                                    msg[2048] = {0,};
+        glusterd_pending_node_t                 *pending_node = NULL;
+        gf_xl_afr_op_t                          heal_op = GF_AFR_OP_INVALID;
+        int                                     rxlator_count = 0;
 
-        opinfo.pending_count--;
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
 
-        if (opinfo.pending_count)
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR, "volume name get failed");
                 goto out;
+        }
 
-        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL);
-
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
-
-out:
-        return ret;
-}
-
-
-int32_t
-glusterd_op_send_cli_response (int32_t op, int32_t op_ret,
-                               int32_t op_errno, rpcsvc_request_t *req,
-                               void *op_ctx, char *op_errstr)
-{
-        int32_t         ret = -1;
-        gd_serialize_t  sfunc = NULL;
-        void            *cli_rsp = NULL;
-        dict_t          *ctx = NULL;
-
-        switch (op) {
-                case GD_MGMT_CLI_CREATE_VOLUME:
-                        {
-                                gf1_cli_create_vol_rsp rsp = {0,};
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                rsp.volname = "";
-                                if (op_errstr)
-                                        rsp.op_errstr = op_errstr;
-                                else
-                                        rsp.op_errstr = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_serialize_cli_create_vol_rsp;
-                                break;
-                        }
-
-                case GD_MGMT_CLI_START_VOLUME:
-                        {
-                                gf1_cli_start_vol_rsp rsp = {0,};
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                rsp.volname = "";
-                                if (op_errstr)
-                                        rsp.op_errstr = op_errstr;
-                                else
-                                        rsp.op_errstr = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_serialize_cli_start_vol_rsp;
-                                break;
-                        }
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Volume %s does not exist",
+                          volname);
 
-                case GD_MGMT_CLI_STOP_VOLUME:
-                        {
-                                gf1_cli_stop_vol_rsp rsp = {0,};
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                rsp.volname = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_serialize_cli_stop_vol_rsp;
-                                break;
-                        }
+                *op_errstr = gf_strdup (msg);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
 
-                case GD_MGMT_CLI_DELETE_VOLUME:
-                        {
-                                gf1_cli_delete_vol_rsp rsp = {0,};
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                rsp.volname = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_serialize_cli_delete_vol_rsp;
-                                break;
-                        }
+        ret = dict_get_int32 (dict, "heal-op", (int32_t*)&heal_op);
+        if (ret || (heal_op == GF_AFR_OP_INVALID)) {
+                gf_log ("glusterd", GF_LOG_ERROR, "heal op invalid");
+                goto out;
+        }
 
-                case GD_MGMT_CLI_DEFRAG_VOLUME:
-                        {
-                                gf1_cli_defrag_vol_rsp rsp = {0,};
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                //rsp.volname = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_serialize_cli_defrag_vol_rsp;
-                                break;
+        switch (heal_op) {
+                case GF_AFR_OP_INDEX_SUMMARY:
+                case GF_AFR_OP_STATISTICS_HEAL_COUNT:
+                if (!glusterd_is_nodesvc_online ("glustershd")) {
+                        if (!rsp_dict) {
+                                gf_log (this->name, GF_LOG_ERROR, "Received "
+                                        "empty ctx.");
+                                goto out;
                         }
 
-                case GD_MGMT_CLI_ADD_BRICK:
-                        {
-                                gf1_cli_add_brick_rsp rsp = {0,};
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                rsp.volname = "";
-                                if (op_errstr)
-                                        rsp.op_errstr = op_errstr;
-                                else
-                                        rsp.op_errstr = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_serialize_cli_add_brick_rsp;
-                                break;
-                        }
+                        ret = fill_shd_status_for_local_bricks (rsp_dict,
+                                                                volinfo,
+                                                                ALL_REPLICA,
+                                                                dict);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR, "Unable to "
+                                        "fill the shd status for the local "
+                                        "bricks");
+                        goto out;
 
-                case GD_MGMT_CLI_REMOVE_BRICK:
-                        {
-                                gf1_cli_remove_brick_rsp rsp = {0,};
-                                ctx = op_ctx;
-                                if (ctx &&
-                                    dict_get_str (ctx, "errstr", &rsp.op_errstr))
-                                        rsp.op_errstr = "";
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                rsp.volname = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_serialize_cli_remove_brick_rsp;
-                                break;
+                }
+                break;
+                case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA:
+                if (!glusterd_is_nodesvc_online ("glustershd")) {
+                        if (!rsp_dict) {
+                                gf_log (this->name, GF_LOG_ERROR, "Received "
+                                        "empty ctx.");
+                                goto out;
                         }
+                        ret = fill_shd_status_for_local_bricks (rsp_dict,
+                                                                volinfo,
+                                                                PER_REPLICA,
+                                                                dict);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR, "Unable to "
+                                        "fill the shd status for the local"
+                                        " bricks.");
+                        goto out;
 
-                case GD_MGMT_CLI_REPLACE_BRICK:
-                        {
-                                gf1_cli_replace_brick_rsp rsp = {0,};
-                                ctx = op_ctx;
-                                if (ctx &&
-                                    dict_get_str (ctx, "status-reply", &rsp.status))
-                                        rsp.status = "";
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                if (op_errstr)
-                                        rsp.op_errstr = op_errstr;
-                                else
-                                        rsp.op_errstr = "";
-                                rsp.volname = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_serialize_cli_replace_brick_rsp;
-                                break;
-                        }
+                }
+                break;
+                default:
+                        break;
+        }
 
-                case GD_MGMT_CLI_SET_VOLUME:
-                        {
-                                gf1_cli_set_vol_rsp rsp = {0,};
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                rsp.volname = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_serialize_cli_set_vol_rsp;
-                                break;
-                        }
 
-                case GD_MGMT_CLI_LOG_FILENAME:
-                        {
-                                gf1_cli_log_filename_rsp rsp = {0,};
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                rsp.errstr = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_serialize_cli_log_filename_rsp;
-                                break;
-                        }
-                case GD_MGMT_CLI_LOG_ROTATE:
-                        {
-                                gf1_cli_log_rotate_rsp rsp = {0,};
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                rsp.errstr = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_serialize_cli_log_rotate_rsp;
-                                break;
-                        }
-                case GD_MGMT_CLI_SYNC_VOLUME:
-                        {
-                                gf1_cli_sync_volume_rsp rsp = {0,};
-                                rsp.op_ret = op_ret;
-                                rsp.op_errno = op_errno;
-                                if (op_errstr)
-                                        rsp.op_errstr = op_errstr;
-                                else
-                                        rsp.op_errstr = "";
-                                cli_rsp = &rsp;
-                                sfunc = gf_xdr_from_cli_sync_volume_rsp;
-                                break;
-                        }
+        switch (heal_op) {
+        case GF_AFR_OP_HEAL_FULL:
+                rxlator_count = _select_rxlators_for_full_self_heal (this,
+                                                                     volinfo,
+                                                                     dict);
+                break;
+        case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA:
+                rxlator_count = _select_rxlators_with_local_bricks (this,
+                                                                   volinfo,
+                                                                   dict,
+                                                                   PER_REPLICA);
+                break;
+        default:
+                rxlator_count = _select_rxlators_with_local_bricks (this,
+                                                                    volinfo,
+                                                                    dict,
+                                                                   ALL_REPLICA);
+                break;
+        }
+        if (!rxlator_count)
+                goto out;
+        if (rxlator_count == -1){
+                gf_log (this->name, GF_LOG_ERROR, "Could not determine the"
+                        "translator count");
+                ret = -1;
+                goto out;
         }
 
-        ret = glusterd_submit_reply (req, cli_rsp, NULL, 0, NULL,
-                                     sfunc);
-
+        ret = dict_set_int32 (dict, "count", rxlator_count);
         if (ret)
                 goto out;
 
+        pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                  gf_gld_mt_pending_node_t);
+        if (!pending_node) {
+                ret = -1;
+                goto out;
+        } else {
+                pending_node->node = priv->shd;
+                pending_node->type = GD_NODE_SHD;
+                list_add_tail (&pending_node->list, selected);
+                pending_node = NULL;
+        }
+
 out:
-        gf_log ("", GF_LOG_NORMAL, "Returning %d", ret);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning ret %d", ret);
         return ret;
-}
 
-int32_t
-glusterd_op_clear_errstr() {
-        opinfo.op_errstr = NULL;
-        return 0;
 }
 
-int32_t
-glusterd_op_txn_complete ()
+static int
+glusterd_bricks_select_rebalance_volume (dict_t *dict, char **op_errstr,
+                                         struct list_head *selected)
 {
-        int32_t                 ret = -1;
-        glusterd_conf_t         *priv = NULL;
-        int32_t                 op = -1;
-        int32_t                 op_ret = 0;
-        int32_t                 op_errno = 0;
-        int32_t                 cli_op = 0;
-        rpcsvc_request_t        *req = NULL;
-        void                    *ctx = NULL;
-        gf_boolean_t            ctx_free = _gf_false;
-        char                    *op_errstr = NULL;
-
-
-        priv = THIS->private;
-        GF_ASSERT (priv);
+        int                                     ret = -1;
+        char                                    *volname = NULL;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        xlator_t                                *this = NULL;
+        char                                    msg[2048] = {0,};
+        glusterd_pending_node_t                 *pending_node = NULL;
 
-        ret = glusterd_unlock (priv->uuid);
+        this = THIS;
+        GF_ASSERT (this);
 
+        ret = dict_get_str (dict, "volname", &volname);
         if (ret) {
-                gf_log ("glusterd", GF_LOG_CRITICAL,
-                        "Unable to clear local lock, ret: %d", ret);
+                gf_log ("glusterd", GF_LOG_ERROR, "volume name get failed");
                 goto out;
         }
 
-        gf_log ("glusterd", GF_LOG_NORMAL, "Cleared local lock");
-
-        op_ret = opinfo.op_ret;
-        op_errno = opinfo.op_errno;
-        cli_op = opinfo.cli_op;
-        req = opinfo.req;
-        if (opinfo.op_errstr)
-                op_errstr = opinfo.op_errstr;
-
-
-        opinfo.op_ret = 0;
-        opinfo.op_errno = 0;
-
-        op = glusterd_op_get_op ();
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Volume %s does not exist",
+                          volname);
 
-        if (op != -1) {
-                glusterd_op_clear_pending_op (op);
-                glusterd_op_clear_commit_op (op);
-                glusterd_op_clear_op (op);
-                ctx = glusterd_op_get_ctx (op);
-                ctx_free = glusterd_op_get_ctx_free (op);
-                glusterd_op_set_ctx (op, NULL);
-                glusterd_op_clear_ctx_free (op);
-                glusterd_op_clear_errstr ();
+                *op_errstr = gf_strdup (msg);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                goto out;
         }
-
-out:
-        pthread_mutex_unlock (&opinfo.lock);
-        ret = glusterd_op_send_cli_response (cli_op, op_ret,
-                                             op_errno, req, ctx, op_errstr);
-
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Responding to cli failed, ret: %d",
-                        ret);
-                //Ignore this error, else state machine blocks
-                ret = 0;
+        pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                  gf_gld_mt_pending_node_t);
+        if (!pending_node) {
+                ret = -1;
+                goto out;
+        } else {
+                pending_node->node = volinfo;
+                pending_node->type = GD_NODE_REBALANCE;
+                list_add_tail (&pending_node->list,
+                               &opinfo.pending_bricks);
+                pending_node = NULL;
         }
 
-        if (ctx_free && ctx && (op != -1))
-                glusterd_op_free_ctx (op, ctx, ctx_free);
-        if (op_errstr && (strcmp (op_errstr, "")))
-                GF_FREE (op_errstr);
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Returning %d", ret);
-        return ret;
-}
-
-static int
-glusterd_op_ac_unlocked_all (glusterd_op_sm_event_t *event, void *ctx)
-{
-        int                     ret = 0;
-
-        GF_ASSERT (event);
-
-        ret = glusterd_op_txn_complete ();
-
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
-
+out:
         return ret;
 }
 
 
-static int
-glusterd_op_ac_commit_error (glusterd_op_sm_event_t *event, void *ctx)
-{
-        int                     ret = 0;
 
-        //Log here with who failed the commit
-        //
-
-        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_START_UNLOCK, NULL);
-
-        return ret;
-}
 
 static int
-glusterd_op_ac_stage_op (glusterd_op_sm_event_t *event, void *ctx)
+glusterd_bricks_select_status_volume (dict_t *dict, char **op_errstr,
+                                      struct list_head *selected)
 {
         int                     ret = -1;
-        gd1_mgmt_stage_op_req   *req = NULL;
-        glusterd_op_stage_ctx_t *stage_ctx = NULL;
-        int32_t                 status = 0;
-        char                    *op_errstr = NULL;
+        int                     cmd = 0;
+        int                     brick_index = -1;
+        char                    *volname = NULL;
+        char                    *brickname = NULL;
+        glusterd_volinfo_t      *volinfo = NULL;
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        glusterd_pending_node_t *pending_node = NULL;
+        xlator_t                *this = NULL;
+        glusterd_conf_t         *priv = NULL;
 
-        GF_ASSERT (ctx);
+        GF_ASSERT (dict);
 
-        stage_ctx = ctx;
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
 
-        req = &stage_ctx->stage_req;
+        ret = dict_get_int32 (dict, "cmd", &cmd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get status type");
+                goto out;
+        }
 
-        status = glusterd_op_stage_validate (req, &op_errstr);
+        if (cmd & GF_CLI_STATUS_ALL)
+                goto out;
 
-        if (status) {
-                gf_log ("", GF_LOG_ERROR, "Validate failed: %d", status);
+        switch (cmd & GF_CLI_STATUS_MASK) {
+        case GF_CLI_STATUS_MEM:
+        case GF_CLI_STATUS_CLIENTS:
+        case GF_CLI_STATUS_INODE:
+        case GF_CLI_STATUS_FD:
+        case GF_CLI_STATUS_CALLPOOL:
+        case GF_CLI_STATUS_NFS:
+        case GF_CLI_STATUS_SHD:
+                break;
+        default:
+                goto out;
+        }
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volname");
+                goto out;
+        }
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                goto out;
         }
 
-        ret = glusterd_op_stage_send_resp (stage_ctx->req, req->op, status, op_errstr);
+        if ( (cmd & GF_CLI_STATUS_BRICK) != 0) {
+                ret = dict_get_str (dict, "brick", &brickname);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to get brick");
+                        goto out;
+                }
+                ret = glusterd_volume_brickinfo_get_by_brick (brickname,
+                                                              volinfo,
+                                                              &brickinfo);
+                if (ret)
+                        goto out;
 
-        if (op_errstr && (strcmp (op_errstr, "")))
-                GF_FREE (op_errstr);
+                if (uuid_compare (brickinfo->uuid, MY_UUID)||
+                    !glusterd_is_brick_started (brickinfo))
+                        goto out;
+
+                pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                          gf_gld_mt_pending_node_t);
+                if (!pending_node) {
+                        ret = -1;
+                        goto out;
+                }
+                pending_node->node = brickinfo;
+                pending_node->type = GD_NODE_BRICK;
+                pending_node->index = 0;
+                list_add_tail (&pending_node->list, selected);
+
+                ret = 0;
+        } else if ((cmd & GF_CLI_STATUS_NFS) != 0) {
+                if (!glusterd_is_nodesvc_online ("nfs")) {
+                        ret = -1;
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "NFS server is not running");
+                        goto out;
+                }
+                pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                          gf_gld_mt_pending_node_t);
+                if (!pending_node) {
+                        ret = -1;
+                        goto out;
+                }
+                pending_node->node = priv->nfs;
+                pending_node->type = GD_NODE_NFS;
+                pending_node->index = 0;
+                list_add_tail (&pending_node->list, selected);
 
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+                ret = 0;
+        } else if ((cmd & GF_CLI_STATUS_SHD) != 0) {
+                if (!glusterd_is_nodesvc_online ("glustershd")) {
+                        ret = -1;
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Self-heal daemon is not running");
+                        goto out;
+                }
+                pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                          gf_gld_mt_pending_node_t);
+                if (!pending_node) {
+                        ret = -1;
+                        goto out;
+                }
+                pending_node->node = priv->shd;
+                pending_node->type = GD_NODE_SHD;
+                pending_node->index = 0;
+                list_add_tail (&pending_node->list, selected);
 
+                ret = 0;
+        } else {
+                list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                        brick_index++;
+                        if (uuid_compare (brickinfo->uuid, MY_UUID) ||
+                            !glusterd_is_brick_started (brickinfo)) {
+                                continue;
+                        }
+                        pending_node = GF_CALLOC (1, sizeof (*pending_node),
+                                                  gf_gld_mt_pending_node_t);
+                        if (!pending_node) {
+                                ret = -1;
+                                gf_log (THIS->name ,GF_LOG_ERROR,
+                                        "Unable to allocate memory");
+                                goto out;
+                        }
+                        pending_node->node = brickinfo;
+                        pending_node->type = GD_NODE_BRICK;
+                        pending_node->index = brick_index;
+                        list_add_tail (&pending_node->list, selected);
+                        pending_node = NULL;
+                }
+        }
+out:
         return ret;
 }
 
 static int
-glusterd_op_ac_commit_op (glusterd_op_sm_event_t *event, void *ctx)
+glusterd_op_ac_send_brick_op (glusterd_op_sm_event_t *event, void *ctx)
 {
         int                             ret = 0;
-        gd1_mgmt_stage_op_req           *req = NULL;
-        glusterd_op_commit_ctx_t        *commit_ctx = NULL;
-        int32_t                         status = 0;
+        rpc_clnt_procedure_t            *proc = NULL;
+        glusterd_conf_t                 *priv = NULL;
+        xlator_t                        *this = NULL;
+        glusterd_op_t                   op = GD_OP_NONE;
+        glusterd_req_ctx_t              *req_ctx = NULL;
         char                            *op_errstr = NULL;
-        dict_t                          *rsp_dict = NULL;
-
-        GF_ASSERT (ctx);
 
-        commit_ctx = ctx;
-
-        req = &commit_ctx->stage_req;
+        this = THIS;
+        priv = this->private;
 
-        rsp_dict = dict_new ();
-        if (!rsp_dict) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Out of memory");
-                ret = -1;
-                goto out;
+        if (ctx) {
+                req_ctx = ctx;
+        } else {
+                req_ctx = GF_CALLOC (1, sizeof (*req_ctx),
+                                     gf_gld_mt_op_allack_ctx_t);
+                op = glusterd_op_get_op ();
+                req_ctx->op = op;
+                uuid_copy (req_ctx->uuid, MY_UUID);
+                ret = glusterd_op_build_payload (&req_ctx->dict, &op_errstr,
+                                                 NULL);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD,
+                                gd_op_list[op]);
+                        if (op_errstr == NULL)
+                                gf_asprintf (&op_errstr,
+                                             OPERRSTR_BUILD_PAYLOAD);
+                        opinfo.op_errstr = op_errstr;
+                        goto out;
+                }
         }
 
-
-        status = glusterd_op_commit_perform (req, &op_errstr, rsp_dict);
-
-        if (status) {
-                gf_log ("", GF_LOG_ERROR, "Commit failed: %d", status);
+        proc = &priv->gfs_mgmt->proctable[GLUSTERD_BRICK_OP];
+        if (proc->fn) {
+                ret = proc->fn (NULL, this, req_ctx);
+                if (ret)
+                        goto out;
         }
 
-        ret = glusterd_op_commit_send_resp (commit_ctx->req, req->op, status,
-                                            op_errstr, rsp_dict);
+        if (!opinfo.pending_count && !opinfo.brick_pending_count) {
+                glusterd_clear_pending_nodes (&opinfo.pending_bricks);
+                ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK,
+                                                   &event->txn_id, req_ctx);
+        }
 
 out:
-        if (rsp_dict)
-                dict_unref (rsp_dict);
-        if (op_errstr && (strcmp (op_errstr, "")))
-                GF_FREE (op_errstr);
-
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
 
         return ret;
 }
 
 
 static int
-glusterd_op_sm_transition_state (glusterd_op_info_t *opinfo,
-                                 glusterd_op_sm_t *state,
-                                 glusterd_op_sm_event_type_t event_type)
+glusterd_op_ac_rcvd_brick_op_acc (glusterd_op_sm_event_t *event, void *ctx)
 {
+        int                     ret = 0;
+        glusterd_op_brick_rsp_ctx_t *ev_ctx = NULL;
+        char                        *op_errstr = NULL;
+        glusterd_op_t               op = GD_OP_NONE;
+        gd_node_type                type = GD_NODE_NONE;
+        dict_t                      *op_ctx = NULL;
+        glusterd_req_ctx_t          *req_ctx = NULL;
+        void                        *pending_entry = NULL;
+        xlator_t                    *this = NULL;
 
-        GF_ASSERT (state);
-        GF_ASSERT (opinfo);
-
-        gf_log ("", GF_LOG_NORMAL, "Transitioning from %d to %d",
-                     opinfo->state.state, state[event_type].next_state);
-        opinfo->state.state =
-                state[event_type].next_state;
-        return 0;
-}
-
-int32_t
-glusterd_op_stage_validate (gd1_mgmt_stage_op_req *req, char **op_errstr)
-{
-        int     ret = -1;
-
-        GF_ASSERT (req);
-
-        switch (req->op) {
-                case GD_OP_CREATE_VOLUME:
-                        ret = glusterd_op_stage_create_volume (req, op_errstr);
-                        break;
-
-                case GD_OP_START_VOLUME:
-                        ret = glusterd_op_stage_start_volume (req, op_errstr);
-                        break;
-
-                case GD_OP_STOP_VOLUME:
-                        ret = glusterd_op_stage_stop_volume (req);
-                        break;
-
-                case GD_OP_DELETE_VOLUME:
-                        ret = glusterd_op_stage_delete_volume (req);
-                        break;
-
-                case GD_OP_ADD_BRICK:
-                        ret = glusterd_op_stage_add_brick (req, op_errstr);
-                        break;
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (event);
+        GF_ASSERT (ctx);
+        ev_ctx = ctx;
 
-                case GD_OP_REPLACE_BRICK:
-                        ret = glusterd_op_stage_replace_brick (req, op_errstr);
-                        break;
+        req_ctx = ev_ctx->commit_ctx;
+        GF_ASSERT (req_ctx);
 
-                case GD_OP_SET_VOLUME:
-                        ret = glusterd_op_stage_set_volume (req);
-                        break;
+        op = req_ctx->op;
+        op_ctx = glusterd_op_get_ctx ();
+        pending_entry = ev_ctx->pending_node->node;
+        type = ev_ctx->pending_node->type;
 
-                case GD_OP_REMOVE_BRICK:
-                        ret = glusterd_op_stage_remove_brick (req);
-                        break;
+        ret = glusterd_remove_pending_entry (&opinfo.pending_bricks,
+                                             pending_entry);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "unknown response received ");
+                ret = -1;
+                goto out;
+        }
 
-                case GD_OP_LOG_FILENAME:
-                        ret = glusterd_op_stage_log_filename (req);
-                        break;
+        if (opinfo.brick_pending_count > 0)
+                opinfo.brick_pending_count--;
 
-                case GD_OP_LOG_ROTATE:
-                        ret = glusterd_op_stage_log_rotate (req);
-                        break;
+        glusterd_handle_node_rsp (req_ctx->dict, pending_entry, op, ev_ctx->rsp_dict,
+                                  op_ctx, &op_errstr, type);
 
-                case GD_OP_SYNC_VOLUME:
-                        ret = glusterd_op_stage_sync_volume (req, op_errstr);
-                        break;
+        if (opinfo.brick_pending_count > 0)
+                goto out;
 
-                default:
-                        gf_log ("", GF_LOG_ERROR, "Unknown op %d",
-                                req->op);
-        }
+        ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, &event->txn_id,
+                                           ev_ctx->commit_ctx);
 
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+out:
+        if (ev_ctx->rsp_dict)
+                dict_unref (ev_ctx->rsp_dict);
+        GF_FREE (ev_ctx);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
 
         return ret;
 }
 
-
 int32_t
-glusterd_op_commit_perform (gd1_mgmt_stage_op_req *req, char **op_errstr,
-                            dict_t *rsp_dict)
+glusterd_op_bricks_select (glusterd_op_t op, dict_t *dict, char **op_errstr,
+                           struct list_head *selected, dict_t *rsp_dict)
 {
-        int     ret = -1;
-
-        GF_ASSERT (req);
-
-        switch (req->op) {
-                case GD_OP_CREATE_VOLUME:
-                        ret = glusterd_op_create_volume (req, op_errstr);
-                        break;
-
-                case GD_OP_START_VOLUME:
-                        ret = glusterd_op_start_volume (req, op_errstr);
-                        break;
-
-                case GD_OP_STOP_VOLUME:
-                        ret = glusterd_op_stop_volume (req);
-                        break;
+        int     ret = 0;
 
-                case GD_OP_DELETE_VOLUME:
-                        ret = glusterd_op_delete_volume (req);
-                        break;
-
-                case GD_OP_ADD_BRICK:
-                        ret = glusterd_op_add_brick (req, op_errstr);
-                        break;
-
-                case GD_OP_REPLACE_BRICK:
-                        ret = glusterd_op_replace_brick (req);
-                        break;
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (op > GD_OP_NONE);
+        GF_ASSERT (op < GD_OP_MAX);
 
-                case GD_OP_SET_VOLUME:
-                        ret = glusterd_op_set_volume (req);
-                        break;
+        switch (op) {
+        case GD_OP_STOP_VOLUME:
+                ret = glusterd_bricks_select_stop_volume (dict, op_errstr,
+                                                          selected);
+                break;
 
-                case GD_OP_REMOVE_BRICK:
-                        ret = glusterd_op_remove_brick (req);
-                        break;
+        case GD_OP_REMOVE_BRICK:
+                ret = glusterd_bricks_select_remove_brick (dict, op_errstr,
+                                                           selected);
+                break;
 
-                case GD_OP_LOG_FILENAME:
-                        ret = glusterd_op_log_filename (req);
-                        break;
+        case GD_OP_PROFILE_VOLUME:
+                ret = glusterd_bricks_select_profile_volume (dict, op_errstr,
+                                                             selected);
+                break;
 
-                case GD_OP_LOG_ROTATE:
-                        ret = glusterd_op_log_rotate (req);
-                        break;
+        case GD_OP_HEAL_VOLUME:
+                ret = glusterd_bricks_select_heal_volume (dict, op_errstr,
+                                                          selected, rsp_dict);
+                break;
 
-                case GD_OP_SYNC_VOLUME:
-                        ret = glusterd_op_sync_volume (req, op_errstr, rsp_dict);
-                        break;
+        case GD_OP_STATUS_VOLUME:
+                ret = glusterd_bricks_select_status_volume (dict, op_errstr,
+                                                            selected);
+                break;
 
-                default:
-                        gf_log ("", GF_LOG_ERROR, "Unknown op %d",
-                                req->op);
-        }
+        case GD_OP_DEFRAG_BRICK_VOLUME:
+                ret = glusterd_bricks_select_rebalance_volume (dict, op_errstr,
+                                                               selected);
+                break;
+        case GD_OP_SNAP:
+                ret = glusterd_bricks_select_snap (dict, op_errstr, selected);
+                break;
+        default:
+                break;
+         }
 
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
 
         return ret;
 }
@@ -4149,29 +5672,33 @@ glusterd_op_sm_t glusterd_op_state_default [] = {
         {GD_OP_STATE_DEFAULT, glusterd_op_ac_none}, //EVENT_COMMIT_OP
         {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
         {GD_OP_STATE_DEFAULT, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_none}, //EVENT_ALL_ACK
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_none}, //EVENT_LOCAL_UNLOCK_NO_RESP
         {GD_OP_STATE_DEFAULT, glusterd_op_ac_none}, //EVENT_MAX
 };
 
 glusterd_op_sm_t glusterd_op_state_lock_sent [] = {
         {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_none}, //EVENT_NONE
         {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_none},//EVENT_START_LOCK
-        {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_none}, //EVENT_LOCK
+        {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_lock}, //EVENT_LOCK
         {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_rcvd_lock_acc}, //EVENT_RCVD_ACC
         {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_send_stage_op}, //EVENT_ALL_ACC
         {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_none}, //EVENT_STAGE_ACC
         {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_none}, //EVENT_COMMIT_ACC
-        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_send_unlock}, //EVENT_RCVD_RJT
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_send_unlock_drain}, //EVENT_RCVD_RJT
         {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_none}, //EVENT_STAGE_OP
         {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_none}, //EVENT_COMMIT_OP
         {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
-        {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_none}, //EVENT_ALL_ACK
+        {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_none}, //EVENT_LOCAL_UNLOCK_NO_RESP
         {GD_OP_STATE_LOCK_SENT, glusterd_op_ac_none}, //EVENT_MAX
 };
 
 glusterd_op_sm_t glusterd_op_state_locked [] = {
         {GD_OP_STATE_LOCKED, glusterd_op_ac_none}, //EVENT_NONE
         {GD_OP_STATE_LOCKED, glusterd_op_ac_none},//EVENT_START_LOCK
-        {GD_OP_STATE_LOCKED, glusterd_op_ac_none}, //EVENT_LOCK
+        {GD_OP_STATE_LOCKED, glusterd_op_ac_lock}, //EVENT_LOCK
         {GD_OP_STATE_LOCKED, glusterd_op_ac_none}, //EVENT_RCVD_ACC
         {GD_OP_STATE_LOCKED, glusterd_op_ac_none}, //EVENT_ALL_ACC
         {GD_OP_STATE_LOCKED, glusterd_op_ac_none}, //EVENT_STAGE_ACC
@@ -4181,61 +5708,177 @@ glusterd_op_sm_t glusterd_op_state_locked [] = {
         {GD_OP_STATE_LOCKED, glusterd_op_ac_none}, //EVENT_COMMIT_OP
         {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
         {GD_OP_STATE_LOCKED, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_LOCKED, glusterd_op_ac_none}, //EVENT_ALL_ACK
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_local_unlock}, //EVENT_LOCAL_UNLOCK_NO_RESP
         {GD_OP_STATE_LOCKED, glusterd_op_ac_none}, //EVENT_MAX
 };
 
 glusterd_op_sm_t glusterd_op_state_stage_op_sent [] = {
         {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_none}, //EVENT_NONE
         {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_none},//EVENT_START_LOCK
-        {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_none}, //EVENT_LOCK
+        {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_lock}, //EVENT_LOCK
         {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_rcvd_stage_op_acc}, //EVENT_RCVD_ACC
-        {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_send_stage_op}, //EVENT_ALL_ACC
-        {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_send_commit_op}, //EVENT_STAGE_ACC
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_send_brick_op}, //EVENT_ALL_ACC
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_send_brick_op}, //EVENT_STAGE_ACC
         {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_none}, //EVENT_COMMIT_ACC
-        {GD_OP_STATE_UNLOCK_SENT,   glusterd_op_ac_send_unlock}, //EVENT_RCVD_RJT
+        {GD_OP_STATE_STAGE_OP_FAILED,   glusterd_op_ac_stage_op_failed}, //EVENT_RCVD_RJT
         {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_none}, //EVENT_STAGE_OP
         {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_none}, //EVENT_COMMIT_OP
-        {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
-        {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_none}, //EVENT_ALL_ACK
+        {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_none}, //EVENT_LOCAL_UNLOCK_NO_RESP
         {GD_OP_STATE_STAGE_OP_SENT, glusterd_op_ac_none}, //EVENT_MAX
 };
 
+glusterd_op_sm_t glusterd_op_state_stage_op_failed [] = {
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_none}, //EVENT_NONE
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_none},//EVENT_START_LOCK
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_lock}, //EVENT_LOCK
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_stage_op_failed}, //EVENT_RCVD_ACC
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_none}, //EVENT_ALL_ACC
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_none}, //EVENT_STAGE_ACC
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_none}, //EVENT_COMMIT_ACC
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_stage_op_failed}, //EVENT_RCVD_RJT
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_none}, //EVENT_STAGE_OP
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_none}, //EVENT_COMMIT_OP
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_send_unlock}, //EVENT_ALL_ACK
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_none}, //EVENT_LOCAL_UNLOCK_NO_RESP
+        {GD_OP_STATE_STAGE_OP_FAILED, glusterd_op_ac_none}, //EVENT_MAX
+};
+
 glusterd_op_sm_t glusterd_op_state_staged [] = {
         {GD_OP_STATE_STAGED, glusterd_op_ac_none}, //EVENT_NONE
         {GD_OP_STATE_STAGED, glusterd_op_ac_none},//EVENT_START_LOCK
-        {GD_OP_STATE_STAGED, glusterd_op_ac_none}, //EVENT_LOCK
+        {GD_OP_STATE_STAGED, glusterd_op_ac_lock}, //EVENT_LOCK
         {GD_OP_STATE_STAGED, glusterd_op_ac_none}, //EVENT_RCVD_ACC
         {GD_OP_STATE_STAGED, glusterd_op_ac_none}, //EVENT_ALL_ACC
         {GD_OP_STATE_STAGED, glusterd_op_ac_none}, //EVENT_STAGE_ACC
         {GD_OP_STATE_STAGED, glusterd_op_ac_none}, //EVENT_COMMIT_ACC
         {GD_OP_STATE_STAGED, glusterd_op_ac_none}, //EVENT_RCVD_RJT
         {GD_OP_STATE_STAGED, glusterd_op_ac_none}, //EVENT_STAGE_OP
-        {GD_OP_STATE_COMMITED, glusterd_op_ac_commit_op}, //EVENT_COMMIT_OP
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_send_brick_op}, //EVENT_COMMIT_OP
         {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
         {GD_OP_STATE_STAGED, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_STAGED, glusterd_op_ac_none}, //EVENT_ALL_ACK
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_local_unlock}, //EVENT_LOCAL_UNLOCK_NO_RESP
         {GD_OP_STATE_STAGED, glusterd_op_ac_none}, //EVENT_MAX
 };
 
+glusterd_op_sm_t glusterd_op_state_brick_op_sent [] = {
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_none}, //EVENT_NONE
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_none},//EVENT_START_LOCK
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_lock}, //EVENT_LOCK
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_rcvd_brick_op_acc}, //EVENT_RCVD_ACC
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_none}, //EVENT_ALL_ACC
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_none}, //EVENT_STAGE_ACC
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_none}, //EVENT_COMMIT_ACC
+        {GD_OP_STATE_BRICK_OP_FAILED,   glusterd_op_ac_brick_op_failed}, //EVENT_RCVD_RJT
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_none}, //EVENT_BRICK_OP
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_none}, //EVENT_COMMIT_OP
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_send_commit_op}, //EVENT_ALL_ACK
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_none}, //EVENT_LOCAL_UNLOCK_NO_RESP
+        {GD_OP_STATE_BRICK_OP_SENT, glusterd_op_ac_none}, //EVENT_MAX
+};
+
+glusterd_op_sm_t glusterd_op_state_brick_op_failed [] = {
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_none}, //EVENT_NONE
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_none},//EVENT_START_LOCK
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_lock}, //EVENT_LOCK
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_brick_op_failed}, //EVENT_RCVD_ACC
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_none}, //EVENT_ALL_ACC
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_none}, //EVENT_STAGE_ACC
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_none}, //EVENT_COMMIT_ACC
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_brick_op_failed}, //EVENT_RCVD_RJT
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_none}, //EVENT_BRICK_OP
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_none}, //EVENT_COMMIT_OP
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_send_unlock}, //EVENT_ALL_ACK
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_none}, //EVENT_LOCAL_UNLOCK_NO_RESP
+        {GD_OP_STATE_BRICK_OP_FAILED, glusterd_op_ac_none}, //EVENT_MAX
+};
+
+glusterd_op_sm_t glusterd_op_state_brick_committed [] = {
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_none}, //EVENT_NONE
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_none},//EVENT_START_LOCK
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_lock}, //EVENT_LOCK
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_rcvd_brick_op_acc}, //EVENT_RCVD_ACC
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_none}, //EVENT_ALL_ACC
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_none}, //EVENT_STAGE_ACC
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_none}, //EVENT_COMMIT_ACC
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_brick_op_failed}, //EVENT_RCVD_RJT
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_none}, //EVENT_STAGE_OP
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_none}, //EVENT_COMMIT_OP
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_COMMITED, glusterd_op_ac_commit_op}, //EVENT_ALL_ACK
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_local_unlock}, //EVENT_LOCAL_UNLOCK_NO_RESP
+        {GD_OP_STATE_BRICK_COMMITTED, glusterd_op_ac_none}, //EVENT_MAX
+};
+
+glusterd_op_sm_t glusterd_op_state_brick_commit_failed [] = {
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_none}, //EVENT_NONE
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_none},//EVENT_START_LOCK
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_lock}, //EVENT_LOCK
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_brick_op_failed}, //EVENT_RCVD_ACC
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_none}, //EVENT_ALL_ACC
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_none}, //EVENT_STAGE_ACC
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_none}, //EVENT_COMMIT_ACC
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_brick_op_failed}, //EVENT_RCVD_RJT
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_none}, //EVENT_STAGE_OP
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_none}, //EVENT_COMMIT_OP
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_send_commit_failed}, //EVENT_ALL_ACK
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_local_unlock}, //EVENT_LOCAL_UNLOCK_NO_RESP
+        {GD_OP_STATE_BRICK_COMMIT_FAILED, glusterd_op_ac_none}, //EVENT_MAX
+};
+
+glusterd_op_sm_t glusterd_op_state_commit_op_failed [] = {
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_none}, //EVENT_NONE
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_none},//EVENT_START_LOCK
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_lock}, //EVENT_LOCK
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_commit_op_failed}, //EVENT_RCVD_ACC
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_none}, //EVENT_ALL_ACC
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_none}, //EVENT_STAGE_ACC
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_none}, //EVENT_COMMIT_ACC
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_commit_op_failed}, //EVENT_RCVD_RJT
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_none}, //EVENT_STAGE_OP
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_none}, //EVENT_COMMIT_OP
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_send_unlock}, //EVENT_ALL_ACK
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_none}, //EVENT_LOCAL_UNLOCK_NO_RESP
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_none}, //EVENT_MAX
+};
+
 glusterd_op_sm_t glusterd_op_state_commit_op_sent [] = {
         {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_none}, //EVENT_NONE
         {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_none},//EVENT_START_LOCK
-        {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_none}, //EVENT_LOCK
+        {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_lock}, //EVENT_LOCK
         {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_rcvd_commit_op_acc}, //EVENT_RCVD_ACC
         {GD_OP_STATE_UNLOCK_SENT,    glusterd_op_ac_send_unlock}, //EVENT_ALL_ACC
         {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_none}, //EVENT_STAGE_ACC
         {GD_OP_STATE_UNLOCK_SENT,    glusterd_op_ac_send_unlock}, //EVENT_COMMIT_ACC
-        {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_commit_error}, //EVENT_RCVD_RJT
+        {GD_OP_STATE_COMMIT_OP_FAILED, glusterd_op_ac_commit_op_failed}, //EVENT_RCVD_RJT
         {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_none}, //EVENT_STAGE_OP
         {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_none}, //EVENT_COMMIT_OP
-        {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
-        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_send_unlock}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_DEFAULT,        glusterd_op_ac_unlock}, //EVENT_UNLOCK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_none}, //EVENT_ALL_ACK
+        {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_none}, //EVENT_LOCAL_UNLOCK_NO_RESP
         {GD_OP_STATE_COMMIT_OP_SENT, glusterd_op_ac_none}, //EVENT_MAX
 };
 
-glusterd_op_sm_t glusterd_op_state_commited [] = {
+glusterd_op_sm_t glusterd_op_state_committed [] = {
         {GD_OP_STATE_COMMITED, glusterd_op_ac_none}, //EVENT_NONE
         {GD_OP_STATE_COMMITED, glusterd_op_ac_none},//EVENT_START_LOCK
-        {GD_OP_STATE_COMMITED, glusterd_op_ac_none}, //EVENT_LOCK
+        {GD_OP_STATE_COMMITED, glusterd_op_ac_lock}, //EVENT_LOCK
         {GD_OP_STATE_COMMITED, glusterd_op_ac_none}, //EVENT_RCVD_ACC
         {GD_OP_STATE_COMMITED, glusterd_op_ac_none}, //EVENT_ALL_ACC
         {GD_OP_STATE_COMMITED, glusterd_op_ac_none}, //EVENT_STAGE_ACC
@@ -4245,25 +5888,46 @@ glusterd_op_sm_t glusterd_op_state_commited [] = {
         {GD_OP_STATE_COMMITED, glusterd_op_ac_none}, //EVENT_COMMIT_OP
         {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
         {GD_OP_STATE_COMMITED, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_COMMITED, glusterd_op_ac_none}, //EVENT_ALL_ACK
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_local_unlock}, //EVENT_LOCAL_UNLOCK_NO_RESP
         {GD_OP_STATE_COMMITED, glusterd_op_ac_none}, //EVENT_MAX
 };
 
 glusterd_op_sm_t glusterd_op_state_unlock_sent [] = {
         {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_none}, //EVENT_NONE
         {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_none},//EVENT_START_LOCK
-        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_none}, //EVENT_LOCK
+        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_lock}, //EVENT_LOCK
         {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_rcvd_unlock_acc}, //EVENT_RCVD_ACC
         {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlocked_all}, //EVENT_ALL_ACC
         {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_none}, //EVENT_STAGE_ACC
         {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_none}, //EVENT_COMMIT_ACC
-        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_none}, //EVENT_RCVD_RJT
+        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_rcvd_unlock_acc}, //EVENT_RCVD_RJT
         {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_none}, //EVENT_STAGE_OP
         {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_none}, //EVENT_COMMIT_OP
-        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
-        {GD_OP_STATE_COMMITED, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_none}, //EVENT_ALL_ACK
+        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_none}, //EVENT_LOCAL_UNLOCK_NO_RESP
         {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_none}, //EVENT_MAX
 };
 
+glusterd_op_sm_t glusterd_op_state_ack_drain [] = {
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_NONE
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none},//EVENT_START_LOCK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_lock}, //EVENT_LOCK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_send_unlock_drain}, //EVENT_RCVD_ACC
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_ALL_ACC
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_STAGE_ACC
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_COMMIT_ACC
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_send_unlock_drain}, //EVENT_RCVD_RJT
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_STAGE_OP
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_COMMIT_OP
+        {GD_OP_STATE_DEFAULT, glusterd_op_ac_unlock}, //EVENT_UNLOCK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_START_UNLOCK
+        {GD_OP_STATE_UNLOCK_SENT, glusterd_op_ac_send_unlock}, //EVENT_ALL_ACK
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_LOCAL_UNLOCK_NO_RESP
+        {GD_OP_STATE_ACK_DRAIN, glusterd_op_ac_none}, //EVENT_MAX
+};
 
 glusterd_op_sm_t *glusterd_op_state_table [] = {
         glusterd_op_state_default,
@@ -4272,8 +5936,15 @@ glusterd_op_sm_t *glusterd_op_state_table [] = {
         glusterd_op_state_stage_op_sent,
         glusterd_op_state_staged,
         glusterd_op_state_commit_op_sent,
-        glusterd_op_state_commited,
-        glusterd_op_state_unlock_sent
+        glusterd_op_state_committed,
+        glusterd_op_state_unlock_sent,
+        glusterd_op_state_stage_op_failed,
+        glusterd_op_state_commit_op_failed,
+        glusterd_op_state_brick_op_sent,
+        glusterd_op_state_brick_op_failed,
+        glusterd_op_state_brick_committed,
+        glusterd_op_state_brick_commit_failed,
+        glusterd_op_state_ack_drain
 };
 
 int
@@ -4300,7 +5971,7 @@ glusterd_op_sm_new_event (glusterd_op_sm_event_type_t event_type,
 
 int
 glusterd_op_sm_inject_event (glusterd_op_sm_event_type_t event_type,
-                             void *ctx)
+                             uuid_t *txn_id, void *ctx)
 {
         int32_t                 ret = -1;
         glusterd_op_sm_event_t  *event = NULL;
@@ -4315,8 +5986,11 @@ glusterd_op_sm_inject_event (glusterd_op_sm_event_type_t event_type,
 
         event->ctx = ctx;
 
-        gf_log ("glusterd", GF_LOG_NORMAL, "Enqueuing event: %d",
-                        event->event);
+        if (txn_id)
+                uuid_copy (event->txn_id, *txn_id);
+
+        gf_log (THIS->name, GF_LOG_DEBUG, "Enqueue event: '%s'",
+                glusterd_op_sm_event_name_get (event->event));
         list_add_tail (&event->list, &gd_op_sm_queue);
 
 out:
@@ -4324,6 +5998,24 @@ out:
 }
 
 void
+glusterd_destroy_req_ctx (glusterd_req_ctx_t *ctx)
+{
+        if (!ctx)
+                return;
+        if (ctx->dict)
+                dict_unref (ctx->dict);
+        GF_FREE (ctx);
+}
+
+void
+glusterd_destroy_local_unlock_ctx (uuid_t *ctx)
+{
+        if (!ctx)
+                return;
+        GF_FREE (ctx);
+}
+
+void
 glusterd_destroy_op_event_ctx (glusterd_op_sm_event_t *event)
 {
         if (!event)
@@ -4335,10 +6027,11 @@ glusterd_destroy_op_event_ctx (glusterd_op_sm_event_t *event)
                 glusterd_destroy_lock_ctx (event->ctx);
                 break;
         case GD_OP_EVENT_STAGE_OP:
-                glusterd_destroy_stage_ctx (event->ctx);
+        case GD_OP_EVENT_ALL_ACK:
+                glusterd_destroy_req_ctx (event->ctx);
                 break;
-        case GD_OP_EVENT_COMMIT_OP:
-                glusterd_destroy_commit_ctx (event->ctx);
+        case GD_OP_EVENT_LOCAL_UNLOCK_NO_RESP:
+                glusterd_destroy_local_unlock_ctx (event->ctx);
                 break;
         default:
                 break;
@@ -4351,10 +6044,21 @@ glusterd_op_sm ()
         glusterd_op_sm_event_t          *event = NULL;
         glusterd_op_sm_event_t          *tmp = NULL;
         int                             ret = -1;
+        int                             lock_err = 0;
         glusterd_op_sm_ac_fn            handler = NULL;
         glusterd_op_sm_t                *state = NULL;
         glusterd_op_sm_event_type_t     event_type = GD_OP_EVENT_NONE;
+        xlator_t                        *this = NULL;
+        glusterd_op_info_t              txn_op_info;
+
+        this = THIS;
+        GF_ASSERT (this);
 
+        if ((lock_err = pthread_mutex_trylock (&gd_op_sm_lock))) {
+                gf_log (this->name, GF_LOG_ERROR, "lock failed due to %s",
+                        strerror (lock_err));
+                goto lock_failed;
+        }
 
         while (!list_empty (&gd_op_sm_queue)) {
 
@@ -4362,6 +6066,23 @@ glusterd_op_sm ()
 
                         list_del_init (&event->list);
                         event_type = event->event;
+                        gf_log (this->name, GF_LOG_DEBUG, "Dequeued event of "
+                                "type: '%s'",
+                                glusterd_op_sm_event_name_get(event_type));
+
+                        gf_log ("", GF_LOG_DEBUG, "transaction ID = %s",
+                                uuid_utoa (event->txn_id));
+
+                        ret = glusterd_get_txn_opinfo (&event->txn_id,
+                                                       &txn_op_info);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to get transaction's opinfo");
+                                glusterd_destroy_op_event_ctx (event);
+                                GF_FREE (event);
+                                continue;
+                        } else
+                                opinfo = txn_op_info;
 
                         state = glusterd_op_state_table[opinfo.state.state];
 
@@ -4373,7 +6094,7 @@ glusterd_op_sm ()
                         ret = handler (event, event->ctx);
 
                         if (ret) {
-                                gf_log ("glusterd", GF_LOG_ERROR,
+                                gf_log (this->name, GF_LOG_ERROR,
                                         "handler returned: %d", ret);
                                 glusterd_destroy_op_event_ctx (event);
                                 GF_FREE (event);
@@ -4384,22 +6105,45 @@ glusterd_op_sm ()
                                                                 event_type);
 
                         if (ret) {
-                                gf_log ("glusterd", GF_LOG_ERROR,
+                                gf_log (this->name, GF_LOG_ERROR,
                                         "Unable to transition"
-                                        "state from %d to %d",
-                                         opinfo.state.state,
-                                         state[event_type].next_state);
+                                        "state from '%s' to '%s'",
+                         glusterd_op_sm_state_name_get(opinfo.state.state),
+                         glusterd_op_sm_state_name_get(state[event_type].next_state));
+                                (void ) pthread_mutex_unlock (&gd_op_sm_lock);
                                 return ret;
                         }
 
+                        if ((state[event_type].next_state ==
+                             GD_OP_STATE_DEFAULT) &&
+                           (event_type == GD_OP_EVENT_UNLOCK)) {
+                                /* Clearing the transaction opinfo */
+                                ret = glusterd_clear_txn_opinfo(&event->txn_id);
+                                if (ret)
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Unable to clear "
+                                                "transaction's opinfo");
+                        } else {
+                                ret = glusterd_set_txn_opinfo (&event->txn_id,
+                                                               &opinfo);
+                                if (ret)
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Unable to set "
+                                                "transaction's opinfo");
+                        }
+
                         glusterd_destroy_op_event_ctx (event);
                         GF_FREE (event);
+
                 }
         }
 
 
+        (void ) pthread_mutex_unlock (&gd_op_sm_lock);
         ret = 0;
 
+lock_failed:
+
         return ret;
 }
 
@@ -4410,9 +6154,7 @@ glusterd_op_set_op (glusterd_op_t op)
         GF_ASSERT (op < GD_OP_MAX);
         GF_ASSERT (op > GD_OP_NONE);
 
-        opinfo.op[op] = 1;
-        opinfo.pending_op[op] = 1;
-        opinfo.commit_op[op] = 1;
+        opinfo.op = op;
 
         return 0;
 
@@ -4422,40 +6164,8 @@ int32_t
 glusterd_op_get_op ()
 {
 
-        int     i = 0;
-        int32_t ret = 0;
+        return opinfo.op;
 
-        for ( i = 0; i < GD_OP_MAX; i++) {
-                if (opinfo.op[i])
-                        break;
-        }
-
-        if ( i == GD_OP_MAX)
-                ret = -1;
-        else
-                ret = i;
-
-        return ret;
-
-}
-
-
-int32_t
-glusterd_op_set_cli_op (gf_mgmt_procnum op)
-{
-
-        int32_t         ret;
-
-        ret = pthread_mutex_trylock (&opinfo.lock);
-
-        if (ret)
-                goto out;
-
-        opinfo.cli_op = op;
-
-out:
-        gf_log ("", GF_LOG_NORMAL, "Returning %d", ret);
-        return ret;
 }
 
 int32_t
@@ -4468,131 +6178,59 @@ glusterd_op_set_req (rpcsvc_request_t *req)
 }
 
 int32_t
-glusterd_op_clear_pending_op (glusterd_op_t op)
-{
-
-        GF_ASSERT (op < GD_OP_MAX);
-        GF_ASSERT (op > GD_OP_NONE);
-
-        opinfo.pending_op[op] = 0;
-
-        return 0;
-
-}
-
-int32_t
-glusterd_op_clear_commit_op (glusterd_op_t op)
-{
-
-        GF_ASSERT (op < GD_OP_MAX);
-        GF_ASSERT (op > GD_OP_NONE);
-
-        opinfo.commit_op[op] = 0;
-
-        return 0;
-
-}
-
-int32_t
 glusterd_op_clear_op (glusterd_op_t op)
 {
 
-        GF_ASSERT (op < GD_OP_MAX);
-        GF_ASSERT (op > GD_OP_NONE);
-
-        opinfo.op[op] = 0;
-
-        return 0;
-
-}
-
-int32_t
-glusterd_op_set_ctx (glusterd_op_t op, void *ctx)
-{
-
-        GF_ASSERT (op < GD_OP_MAX);
-        GF_ASSERT (op > GD_OP_NONE);
-
-        opinfo.op_ctx[op] = ctx;
+        opinfo.op = GD_OP_NONE;
 
         return 0;
 
 }
 
 int32_t
-glusterd_op_free_ctx (glusterd_op_t op, void *ctx, gf_boolean_t ctx_free)
+glusterd_op_free_ctx (glusterd_op_t op, void *ctx)
 {
 
-        GF_ASSERT (op < GD_OP_MAX);
-        GF_ASSERT (op > GD_OP_NONE);
-
-        if (ctx && ctx_free) {
+        if (ctx) {
                 switch (op) {
                 case GD_OP_CREATE_VOLUME:
+                case GD_OP_DELETE_VOLUME:
                 case GD_OP_STOP_VOLUME:
                 case GD_OP_ADD_BRICK:
                 case GD_OP_REMOVE_BRICK:
                 case GD_OP_REPLACE_BRICK:
-                case GD_OP_LOG_FILENAME:
                 case GD_OP_LOG_ROTATE:
                 case GD_OP_SYNC_VOLUME:
-                        dict_unref (ctx);
-                        break;
-                case GD_OP_DELETE_VOLUME:
+                case GD_OP_SET_VOLUME:
                 case GD_OP_START_VOLUME:
-                        GF_FREE (ctx);
+                case GD_OP_RESET_VOLUME:
+                case GD_OP_GSYNC_SET:
+                case GD_OP_QUOTA:
+                case GD_OP_PROFILE_VOLUME:
+                case GD_OP_STATUS_VOLUME:
+                case GD_OP_REBALANCE:
+                case GD_OP_HEAL_VOLUME:
+                case GD_OP_STATEDUMP_VOLUME:
+                case GD_OP_CLEARLOCKS_VOLUME:
+                case GD_OP_DEFRAG_BRICK_VOLUME:
+                        dict_unref (ctx);
                         break;
                 default:
+                        GF_ASSERT (0);
                         break;
                 }
         }
-        return 0;
-
-}
-
-void *
-glusterd_op_get_ctx (glusterd_op_t op)
-{
-        GF_ASSERT (op < GD_OP_MAX);
-        GF_ASSERT (op > GD_OP_NONE);
-
-        return opinfo.op_ctx[op];
-
-}
-
-int32_t
-glusterd_op_set_ctx_free (glusterd_op_t op, gf_boolean_t ctx_free)
-{
-
-        GF_ASSERT (op < GD_OP_MAX);
-        GF_ASSERT (op > GD_OP_NONE);
-
-        opinfo.ctx_free[op] = ctx_free;
 
+        glusterd_op_reset_ctx ();
         return 0;
 
 }
 
-int32_t
-glusterd_op_clear_ctx_free (glusterd_op_t op)
-{
-
-        GF_ASSERT (op < GD_OP_MAX);
-        GF_ASSERT (op > GD_OP_NONE);
-
-        opinfo.ctx_free[op] = _gf_false;
-
-        return 0;
-
-}
-
-gf_boolean_t
-glusterd_op_get_ctx_free (glusterd_op_t op)
+void *
+glusterd_op_get_ctx ()
 {
-        GF_ASSERT (op < GD_OP_MAX);
-        GF_ASSERT (op > GD_OP_NONE);
 
-        return opinfo.ctx_free[op];
+        return opinfo.op_ctx;
 
 }
 
@@ -4600,11 +6238,6 @@ int
 glusterd_op_sm_init ()
 {
         INIT_LIST_HEAD (&gd_op_sm_queue);
+        pthread_mutex_init (&gd_op_sm_lock, NULL);
         return 0;
 }
-
-int32_t
-glusterd_opinfo_unlock(){
-        return (pthread_mutex_unlock(&opinfo.lock));
-}
-
diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.h b/xlators/mgmt/glusterd/src/glusterd-op-sm.h
index 2f39d24f7..cf57b78e0 100644
--- a/xlators/mgmt/glusterd/src/glusterd-op-sm.h
+++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _GLUSTERD_OP_SM_H_
 #define _GLUSTERD_OP_SM_H_
 
@@ -25,6 +15,10 @@
 #include "config.h"
 #endif
 
+#ifndef GSYNC_CONF_TEMPLATE
+#define GSYNC_CONF_TEMPLATE GEOREP"/gsyncd_template.conf"
+#endif
+
 #include <pthread.h>
 #include "uuid.h"
 
@@ -38,6 +32,8 @@
 #include "protocol-common.h"
 
 #define GD_VOLUME_NAME_MAX 256
+#define GD_OP_PROTECTED    (0x02)
+#define GD_OP_UNPROTECTED  (0x04)
 
 typedef enum glusterd_op_sm_state_ {
         GD_OP_STATE_DEFAULT = 0,
@@ -48,6 +44,13 @@ typedef enum glusterd_op_sm_state_ {
         GD_OP_STATE_COMMIT_OP_SENT,
         GD_OP_STATE_COMMITED,
         GD_OP_STATE_UNLOCK_SENT,
+        GD_OP_STATE_STAGE_OP_FAILED,
+        GD_OP_STATE_COMMIT_OP_FAILED,
+        GD_OP_STATE_BRICK_OP_SENT,
+        GD_OP_STATE_BRICK_OP_FAILED,
+        GD_OP_STATE_BRICK_COMMITTED,
+        GD_OP_STATE_BRICK_COMMIT_FAILED,
+        GD_OP_STATE_ACK_DRAIN,
         GD_OP_STATE_MAX,
 } glusterd_op_sm_state_t;
 
@@ -64,6 +67,8 @@ typedef enum glusterd_op_sm_event_type_ {
         GD_OP_EVENT_COMMIT_OP,
         GD_OP_EVENT_UNLOCK,
         GD_OP_EVENT_START_UNLOCK,
+        GD_OP_EVENT_ALL_ACK,
+        GD_OP_EVENT_LOCAL_UNLOCK_NO_RESP,
         GD_OP_EVENT_MAX
 } glusterd_op_sm_event_type_t;
 
@@ -72,6 +77,7 @@ struct glusterd_op_sm_event_ {
         struct list_head                list;
         void                            *ctx;
         glusterd_op_sm_event_type_t     event;
+        uuid_t                          txn_id;
 };
 
 typedef struct glusterd_op_sm_event_ glusterd_op_sm_event_t;
@@ -83,27 +89,6 @@ typedef struct glusterd_op_sm_ {
         glusterd_op_sm_ac_fn        handler;
 } glusterd_op_sm_t;
 
-typedef enum glusterd_op_ {
-        GD_OP_NONE = 0,
-        GD_OP_CREATE_VOLUME,
-        GD_OP_START_BRICK,
-        GD_OP_STOP_BRICK,
-        GD_OP_DELETE_VOLUME,
-        GD_OP_START_VOLUME,
-        GD_OP_STOP_VOLUME,
-        GD_OP_RENAME_VOLUME,
-        GD_OP_DEFRAG_VOLUME,
-        GD_OP_ADD_BRICK,
-        GD_OP_REMOVE_BRICK,
-        GD_OP_REPLACE_BRICK,
-        GD_OP_SET_VOLUME,
-        GD_OP_SYNC_VOLUME,
-        GD_OP_LOG_FILENAME,
-        GD_OP_LOG_LOCATE,
-        GD_OP_LOG_ROTATE,
-        GD_OP_MAX,
-} glusterd_op_t;
-
 typedef struct glusterd_op_sm_state_info_ {
         glusterd_op_sm_state_t          state;
         struct timeval                  time;
@@ -112,30 +97,20 @@ typedef struct glusterd_op_sm_state_info_ {
 struct glusterd_op_info_ {
         glusterd_op_sm_state_info_t     state;
         int32_t                         pending_count;
+        int32_t                         brick_pending_count;
         int32_t                         op_count;
-        glusterd_op_t                   op[GD_OP_MAX];
-        glusterd_op_t                   pending_op[GD_OP_MAX];
-        glusterd_op_t                   commit_op[GD_OP_MAX];
+        glusterd_op_t                   op;
         struct list_head                op_peers;
-        void                            *op_ctx[GD_OP_MAX];
+        void                            *op_ctx;
         rpcsvc_request_t                *req;
         int32_t                         op_ret;
         int32_t                         op_errno;
-        pthread_mutex_t                 lock;
-        int32_t                         cli_op;
-        gf_boolean_t                    ctx_free[GD_OP_MAX];
         char                            *op_errstr;
+        struct  list_head               pending_bricks;
 };
 
 typedef struct glusterd_op_info_ glusterd_op_info_t;
 
-struct glusterd_op_start_volume_ctx_ {
-        char                    volume_name[GD_VOLUME_NAME_MAX];
-};
-
-typedef struct glusterd_op_start_volume_ctx_ glusterd_op_start_volume_ctx_t;
-typedef struct glusterd_op_start_volume_ctx_ glusterd_op_delete_volume_ctx_t;
-
 struct glusterd_op_log_filename_ctx_ {
         char                    volume_name[GD_VOLUME_NAME_MAX];
         char                    brick[GD_VOLUME_NAME_MAX];
@@ -145,31 +120,64 @@ typedef struct glusterd_op_log_filename_ctx_ glusterd_op_log_filename_ctx_t;
 
 struct glusterd_op_lock_ctx_ {
         uuid_t                  uuid;
+        dict_t                 *dict;
         rpcsvc_request_t        *req;
 };
 
 typedef struct glusterd_op_lock_ctx_ glusterd_op_lock_ctx_t;
 
-struct glusterd_op_stage_ctx_ {
-        rpcsvc_request_t        *req;
-        gd1_mgmt_stage_op_req   stage_req;
+struct glusterd_req_ctx_ {
+        rpcsvc_request_t *req;
+	u_char            uuid[16];
+	int               op;
+        dict_t           *dict;
 };
 
-typedef struct glusterd_op_stage_ctx_ glusterd_op_stage_ctx_t;
-
-struct glusterd_op_commit_ctx_ {
-        rpcsvc_request_t        *req;
-        gd1_mgmt_stage_op_req   stage_req;
-};
-
-typedef struct glusterd_op_commit_ctx_ glusterd_op_commit_ctx_t;
+typedef struct glusterd_req_ctx_ glusterd_req_ctx_t;
+
+typedef struct glusterd_op_brick_rsp_ctx_ {
+        int  op_ret;
+        char *op_errstr;
+        dict_t *rsp_dict;
+        glusterd_req_ctx_t *commit_ctx;
+        glusterd_pending_node_t *pending_node;
+} glusterd_op_brick_rsp_ctx_t;
+
+typedef struct glusterd_pr_brick_rsp_conv_t {
+        int count;
+        dict_t *dict;
+} glusterd_pr_brick_rsp_conv_t;
+
+typedef struct glusterd_heal_rsp_conv_ {
+        dict_t *dict;
+        glusterd_volinfo_t *volinfo;
+        xlator_t *this;
+} glusterd_heal_rsp_conv_t;
+
+typedef struct glusterd_status_rsp_conv_ {
+        int count;
+        int brick_index_max;
+        int other_count;
+        dict_t *dict;
+} glusterd_status_rsp_conv_t;
+
+typedef struct glusterd_gsync_status_temp {
+        dict_t *rsp_dict;
+        glusterd_volinfo_t *volinfo;
+        char *node;
+}glusterd_gsync_status_temp_t;
+
+typedef enum cli_cmd_type_ {
+        PER_REPLICA,
+        ALL_REPLICA,
+ } cli_cmd_type;
 
 int
 glusterd_op_sm_new_event (glusterd_op_sm_event_type_t event_type,
                           glusterd_op_sm_event_t **new_event);
 int
 glusterd_op_sm_inject_event (glusterd_op_sm_event_type_t event_type,
-                             void *ctx);
+                             uuid_t *txn_id, void *ctx);
 
 int
 glusterd_op_sm_init ();
@@ -178,70 +186,114 @@ int
 glusterd_op_sm ();
 
 int32_t
-glusterd_op_set_ctx (glusterd_op_t op, void *ctx);
+glusterd_op_set_ctx (void *ctx);
 
 int32_t
 glusterd_op_set_op (glusterd_op_t op);
 
-int32_t
-glusterd_op_clear_pending_op (glusterd_op_t op);
+int
+glusterd_op_build_payload (dict_t **req, char **op_errstr, dict_t *op_ctx);
 
 int32_t
-glusterd_op_clear_commit_op (glusterd_op_t op);
+glusterd_op_stage_validate (glusterd_op_t op, dict_t *req, char **op_errstr,
+                            dict_t *rsp_dict);
 
-int
-glusterd_op_build_payload (glusterd_op_t op, gd1_mgmt_stage_op_req **req);
+int32_t
+glusterd_op_commit_perform (glusterd_op_t op, dict_t *req, char **op_errstr,
+                            dict_t* dict);
 
 int32_t
-glusterd_op_stage_validate (gd1_mgmt_stage_op_req *req, char **op_errstr);
+glusterd_op_txn_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx,
+                       char *err_str, size_t err_len);
 
 int32_t
-glusterd_op_commit_perform (gd1_mgmt_stage_op_req *req, char **op_errstr,
-                            dict_t* dict);
+glusterd_op_txn_complete ();
 
 void *
-glusterd_op_get_ctx (glusterd_op_t op);
+glusterd_op_get_ctx ();
 
 int32_t
 glusterd_op_set_req (rpcsvc_request_t *req);
 
 int32_t
-glusterd_op_set_cli_op (gf_mgmt_procnum op);
-
-int32_t
-glusterd_op_send_cli_response (int32_t op, int32_t op_ret,
+glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret,
                                int32_t op_errno, rpcsvc_request_t *req,
                                void *ctx, char *op_errstr);
 int32_t
 glusterd_op_get_op ();
 
 int32_t
-glusterd_op_clear_pending_op (glusterd_op_t op);
+glusterd_op_clear_op ();
 
 int32_t
-glusterd_op_clear_commit_op (glusterd_op_t op);
+glusterd_op_free_ctx (glusterd_op_t op, void *ctx);
 
-int32_t
-glusterd_op_clear_op (glusterd_op_t op);
+int
+glusterd_check_option_exists(char *optstring, char **completion);
 
-int32_t
-glusterd_op_free_ctx (glusterd_op_t op, void *ctx, gf_boolean_t ctx_free);
+int
+set_xlator_option (dict_t *dict, char *key, char *value);
 
-int32_t
-glusterd_opinfo_unlock();
+void
+glusterd_do_replace_brick (void *data);
 
-int32_t
-glusterd_op_set_ctx_free (glusterd_op_t op, gf_boolean_t ctx_free);
+char*
+glusterd_op_sm_state_name_get (int state);
 
+char*
+glusterd_op_sm_event_name_get (int event);
+int32_t
+glusterd_op_bricks_select (glusterd_op_t op, dict_t *dict, char **op_errstr,
+                           struct list_head *selected, dict_t *rsp_dict);
+int
+glusterd_brick_op_build_payload (glusterd_op_t op, glusterd_brickinfo_t *brickinfo,
+                                 gd1_mgmt_brick_op_req **req, dict_t *dict);
+int
+glusterd_node_op_build_payload (glusterd_op_t op, gd1_mgmt_brick_op_req **req,
+                               dict_t *dict);
 int32_t
-glusterd_op_clear_ctx_free (glusterd_op_t op);
+glusterd_handle_brick_rsp (void *pending_entry, glusterd_op_t op,
+                           dict_t *rsp_dict, dict_t *ctx_dict, char **op_errstr,
+                           gd_node_type type);
 
-gf_boolean_t
-glusterd_op_get_ctx_free (glusterd_op_t op);
+dict_t*
+glusterd_op_init_commit_rsp_dict (glusterd_op_t op);
+
+void
+glusterd_op_modify_op_ctx (glusterd_op_t op, void *op_ctx);
 
+int32_t
+glusterd_volume_stats_read_perf (char *brick_path, int32_t blk_size,
+                int32_t blk_count, double *throughput, double *time);
+int32_t
+glusterd_volume_stats_write_perf (char *brick_path, int32_t blk_size,
+                int32_t blk_count, double *throughput, double *time);
+gf_boolean_t
+glusterd_is_volume_started (glusterd_volinfo_t  *volinfo);
+int
+glusterd_start_bricks (glusterd_volinfo_t *volinfo);
 gf_boolean_t
-glusterd_check_option_exists(char *optstring);
+glusterd_are_all_volumes_stopped ();
+int
+glusterd_stop_bricks (glusterd_volinfo_t *volinfo);
+int
+gsync_status (char *master, char *slave, char *conf_path, int *status);
 
 int
-set_xlator_option (dict_t *dict, char *key, char *value);
+glusterd_check_gsync_running (glusterd_volinfo_t *volinfo, gf_boolean_t *flag);
+
+int
+glusterd_defrag_volume_node_rsp (dict_t *req_dict, dict_t *rsp_dict,
+                                 dict_t *op_ctx);
+int
+glusterd_is_valid_vg (glusterd_brickinfo_t *brick, int check_tag, char *msg);
+
+int32_t
+glusterd_get_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t  *opinfo);
+
+int32_t
+glusterd_set_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t  *opinfo);
+
+int32_t
+glusterd_clear_txn_opinfo (uuid_t *txn_id);
 #endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-pmap.c b/xlators/mgmt/glusterd/src/glusterd-pmap.c
index aaa6ed131..a153ca1a9 100644
--- a/xlators/mgmt/glusterd/src/glusterd-pmap.c
+++ b/xlators/mgmt/glusterd/src/glusterd-pmap.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2010-2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
@@ -30,7 +20,8 @@
 #include "glusterd.h"
 #include "glusterd-utils.h"
 
-#include "portmap.h"
+#include "portmap-xdr.h"
+#include "xdr-generic.h"
 #include "protocol-common.h"
 #include "rpcsvc.h"
 
@@ -61,8 +52,8 @@ pmap_port_isfree (int port)
 }
 
 
-struct pmap_registry *
-pmap_registry_new (void)
+static struct pmap_registry *
+pmap_registry_new (xlator_t *this)
 {
         struct pmap_registry *pmap = NULL;
         int                   i = 0;
@@ -78,8 +69,8 @@ pmap_registry_new (void)
                         pmap->ports[i].type = GF_PMAP_PORT_FOREIGN;
         }
 
-        pmap->base_port = 6971; /* 6969 default for tcp, 6970 for IB */
-        pmap->last_alloc = 6971;
+        pmap->base_port = pmap->last_alloc =
+                ((glusterd_conf_t *)(this->private))->base_port;
 
         return pmap;
 }
@@ -95,7 +86,7 @@ pmap_registry_get (xlator_t *this)
 
         pmap = priv->pmap;
         if (!pmap) {
-                pmap = pmap_registry_new ();
+                pmap = pmap_registry_new (this);
                 if (!pmap)
                         return NULL;
                 priv->pmap = pmap;
@@ -227,8 +218,7 @@ pmap_registry_bind (xlator_t *this, int port, const char *brickname,
 
         p = port;
         pmap->ports[p].type = type;
-        if (pmap->ports[p].brickname)
-                free (pmap->ports[p].brickname);
+        free (pmap->ports[p].brickname);
         pmap->ports[p].brickname = strdup (brickname);
         pmap->ports[p].type = type;
         pmap->ports[p].xprt = xprt;
@@ -280,8 +270,7 @@ remove:
         gf_log ("pmap", GF_LOG_INFO, "removing brick %s on port %d",
                 pmap->ports[p].brickname, p);
 
-        if (pmap->ports[p].brickname)
-                free (pmap->ports[p].brickname);
+        free (pmap->ports[p].brickname);
 
         pmap->ports[p].brickname = NULL;
         pmap->ports[p].xprt = NULL;
@@ -290,36 +279,18 @@ out:
         return 0;
 }
 
-
-typedef ssize_t (*gfs_serialize_t) (struct iovec outmsg, void *data);
-
-
-static int
-xdr_to_glusterfs_req (rpcsvc_request_t *req, void *arg, gfs_serialize_t sfunc)
-{
-        int                     ret = -1;
-
-        if (!req)
-                return -1;
-
-        ret = sfunc (req->msg[0], arg);
-
-        if (ret > 0)
-                ret = 0;
-
-        return ret;
-}
-
-
 int
-gluster_pmap_portbybrick (rpcsvc_request_t *req)
+__gluster_pmap_portbybrick (rpcsvc_request_t *req)
 {
         pmap_port_by_brick_req    args = {0,};
         pmap_port_by_brick_rsp    rsp  = {0,};
         char                     *brick = NULL;
         int                       port = 0;
+        int                       ret = -1;
 
-        if (xdr_to_glusterfs_req (req, &args, xdr_to_pmap_port_by_brick_req)) {
+        ret = xdr_to_generic (req->msg[0], &args,
+                              (xdrproc_t)xdr_pmap_port_by_brick_req);
+        if (ret < 0) {
                 req->rpc_err = GARBAGE_ARGS;
                 goto fail;
         }
@@ -335,21 +306,30 @@ gluster_pmap_portbybrick (rpcsvc_request_t *req)
 
 fail:
         glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                               (gd_serialize_t)xdr_from_pmap_port_by_brick_rsp);
-        if (args.brick)
-                free (args.brick);//malloced by xdr
+                               (xdrproc_t)xdr_pmap_port_by_brick_rsp);
+        free (args.brick);//malloced by xdr
 
         return 0;
 }
 
 
 int
-gluster_pmap_brickbyport (rpcsvc_request_t *req)
+gluster_pmap_portbybrick (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __gluster_pmap_portbybrick);
+}
+
+
+int
+__gluster_pmap_brickbyport (rpcsvc_request_t *req)
 {
         pmap_brick_by_port_req    args = {0,};
         pmap_brick_by_port_rsp    rsp  = {0,};
+        int                       ret = -1;
 
-        if (xdr_to_glusterfs_req (req, &args, xdr_to_pmap_brick_by_port_req)) {
+        ret = xdr_to_generic (req->msg[0], &args,
+                              (xdrproc_t)xdr_pmap_brick_by_port_req);
+        if (ret < 0) {
                 req->rpc_err = GARBAGE_ARGS;
                 goto fail;
         }
@@ -362,21 +342,39 @@ gluster_pmap_brickbyport (rpcsvc_request_t *req)
 fail:
 
         glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                               (gd_serialize_t)xdr_from_pmap_brick_by_port_rsp);
+                               (xdrproc_t)xdr_pmap_brick_by_port_rsp);
 
         return 0;
 }
 
 
+int
+gluster_pmap_brickbyport (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __gluster_pmap_brickbyport);
+}
+
+
+static int
+glusterd_brick_update_signin (glusterd_brickinfo_t *brickinfo,
+                              gf_boolean_t value)
+{
+        brickinfo->signed_in = value;
+
+        return 0;
+}
 
 int
-gluster_pmap_signup (rpcsvc_request_t *req)
+__gluster_pmap_signup (rpcsvc_request_t *req)
 {
         pmap_signup_req    args = {0,};
         pmap_signup_rsp    rsp  = {0,};
+        int                ret = -1;
 
 
-        if (xdr_to_glusterfs_req (req, &args, xdr_to_pmap_signup_req)) {
+        ret = xdr_to_generic (req->msg[0], &args,
+                              (xdrproc_t)xdr_pmap_signup_req);
+        if (ret < 0) {
                 req->rpc_err = GARBAGE_ARGS;
                 goto fail;
         }
@@ -386,21 +384,29 @@ gluster_pmap_signup (rpcsvc_request_t *req)
 
 fail:
         glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                               (gd_serialize_t)xdr_from_pmap_signup_rsp);
-        if (args.brick)
-                free (args.brick);//malloced by xdr
+                               (xdrproc_t)xdr_pmap_signup_rsp);
+        free (args.brick);//malloced by xdr
 
         return 0;
 }
 
+int
+gluster_pmap_signup (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __gluster_pmap_signup);
+}
 
 int
-gluster_pmap_signin (rpcsvc_request_t *req)
+__gluster_pmap_signin (rpcsvc_request_t *req)
 {
         pmap_signin_req    args = {0,};
         pmap_signin_rsp    rsp  = {0,};
+        glusterd_brickinfo_t *brickinfo = NULL;
+        int                ret = -1;
 
-        if (xdr_to_glusterfs_req (req, &args, xdr_to_pmap_signin_req)) {
+        ret = xdr_to_generic (req->msg[0], &args,
+                              (xdrproc_t)xdr_pmap_signin_req);
+        if (ret < 0) {
                 req->rpc_err = GARBAGE_ARGS;
                 goto fail;
         }
@@ -408,25 +414,38 @@ gluster_pmap_signin (rpcsvc_request_t *req)
         rsp.op_ret = pmap_registry_bind (THIS, args.port, args.brick,
                                          GF_PMAP_PORT_BRICKSERVER, req->trans);
 
+        ret = glusterd_get_brickinfo (THIS, args.brick, args.port, _gf_true,
+                                      &brickinfo);
 fail:
         glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                               (gd_serialize_t)xdr_from_pmap_signin_rsp);
-        if (args.brick)
-                free (args.brick);//malloced by xdr
+                               (xdrproc_t)xdr_pmap_signin_rsp);
+        free (args.brick);//malloced by xdr
+
+        if (!ret)
+                glusterd_brick_update_signin (brickinfo, _gf_true);
 
         return 0;
 }
 
 
+int
+gluster_pmap_signin (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __gluster_pmap_signin);
+}
+
 
 int
-gluster_pmap_signout (rpcsvc_request_t *req)
+__gluster_pmap_signout (rpcsvc_request_t *req)
 {
         pmap_signout_req    args = {0,};
         pmap_signout_rsp    rsp  = {0,};
+        int                 ret = -1;
+        glusterd_brickinfo_t *brickinfo = NULL;
 
-
-        if (xdr_to_glusterfs_req (req, &args, xdr_to_pmap_signout_req)) {
+        ret = xdr_to_generic (req->msg[0], &args,
+                              (xdrproc_t)xdr_pmap_signout_req);
+        if (ret < 0) {
                 //failed to decode msg;
                 req->rpc_err = GARBAGE_ARGS;
                 goto fail;
@@ -435,28 +454,32 @@ gluster_pmap_signout (rpcsvc_request_t *req)
         rsp.op_ret = pmap_registry_remove (THIS, args.port, args.brick,
                                            GF_PMAP_PORT_BRICKSERVER, req->trans);
 
+        ret = glusterd_get_brickinfo (THIS, args.brick, args.port, _gf_true,
+                                      &brickinfo);
 fail:
         glusterd_submit_reply (req, &rsp, NULL, 0, NULL,
-                               (gd_serialize_t)xdr_from_pmap_signout_rsp);
-        if (args.brick)
-                free (args.brick);//malloced by xdr
+                               (xdrproc_t)xdr_pmap_signout_rsp);
+        free (args.brick);//malloced by xdr
+
+        if (!ret)
+                glusterd_brick_update_signin (brickinfo, _gf_false);
 
         return 0;
 }
 
+int
+gluster_pmap_signout (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __gluster_pmap_signout);
+}
 
 rpcsvc_actor_t gluster_pmap_actors[] = {
-        [GF_PMAP_NULL] = {"NULL", GF_HNDSK_NULL, NULL, NULL, NULL },
-        [GF_PMAP_PORTBYBRICK] = {"PORTBYBRICK", GF_PMAP_PORTBYBRICK,
-                                 gluster_pmap_portbybrick, NULL, NULL },
-        [GF_PMAP_BRICKBYPORT] = {"BRICKBYPORT", GF_PMAP_BRICKBYPORT,
-                                 gluster_pmap_brickbyport, NULL, NULL },
-        [GF_PMAP_SIGNIN] = {"SIGNIN", GF_PMAP_SIGNIN,
-                              gluster_pmap_signin, NULL, NULL },
-        [GF_PMAP_SIGNOUT] = {"SIGNOUT", GF_PMAP_SIGNOUT,
-                              gluster_pmap_signout, NULL, NULL },
-        [GF_PMAP_SIGNUP] = {"SIGNUP", GF_PMAP_SIGNUP,
-                              gluster_pmap_signup, NULL, NULL },
+        [GF_PMAP_NULL]        = {"NULL",        GF_PMAP_NULL,        NULL,                     NULL, 0, DRC_NA},
+        [GF_PMAP_PORTBYBRICK] = {"PORTBYBRICK", GF_PMAP_PORTBYBRICK, gluster_pmap_portbybrick, NULL, 0, DRC_NA},
+        [GF_PMAP_BRICKBYPORT] = {"BRICKBYPORT", GF_PMAP_BRICKBYPORT, gluster_pmap_brickbyport, NULL, 0, DRC_NA},
+        [GF_PMAP_SIGNIN]      = {"SIGNIN",      GF_PMAP_SIGNIN,      gluster_pmap_signin,      NULL, 0, DRC_NA},
+        [GF_PMAP_SIGNOUT]     = {"SIGNOUT",     GF_PMAP_SIGNOUT,     gluster_pmap_signout,     NULL, 0, DRC_NA},
+        [GF_PMAP_SIGNUP]      = {"SIGNUP",      GF_PMAP_SIGNUP,      gluster_pmap_signup,      NULL, 0, DRC_NA},
 };
 
 
diff --git a/xlators/mgmt/glusterd/src/glusterd-pmap.h b/xlators/mgmt/glusterd/src/glusterd-pmap.h
index a021da030..6336ee998 100644
--- a/xlators/mgmt/glusterd/src/glusterd-pmap.h
+++ b/xlators/mgmt/glusterd/src/glusterd-pmap.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _GLUSTERD_PMAP_H_
 #define _GLUSTERD_PMAP_H_
 
@@ -38,6 +28,8 @@
 #include "rpcsvc.h"
 
 
+#define GF_IANA_PRIV_PORTS_START 49152 /* RFC 6335 */
+
 struct pmap_port_status {
         gf_pmap_port_type_t type;
         char  *brickname;
@@ -55,5 +47,8 @@ int pmap_registry_bind (xlator_t *this, int port, const char *brickname,
                         gf_pmap_port_type_t type, void *xprt);
 int pmap_registry_remove (xlator_t *this, int port, const char *brickname,
                           gf_pmap_port_type_t type, void *xprt);
+int pmap_registry_search (xlator_t *this, const char *brickname,
+                          gf_pmap_port_type_t type);
+struct pmap_registry *pmap_registry_get (xlator_t *this);
 
 #endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-quota.c b/xlators/mgmt/glusterd/src/glusterd-quota.c
new file mode 100644
index 000000000..318267199
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-quota.c
@@ -0,0 +1,839 @@
+/*
+   Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "common-utils.h"
+#include "cli1-xdr.h"
+#include "xdr-generic.h"
+#include "glusterd.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-store.h"
+#include "glusterd-utils.h"
+#include "glusterd-volgen.h"
+#include "run.h"
+
+#include <sys/wait.h>
+
+int
+__glusterd_handle_quota (rpcsvc_request_t *req)
+{
+        int32_t                         ret = -1;
+        gf_cli_req                      cli_req = {{0,}};
+        dict_t                         *dict = NULL;
+        glusterd_op_t                   cli_op = GD_OP_QUOTA;
+        char                            operation[256] = {0, };
+        char                           *volname = NULL;
+        int32_t                         type = 0;
+        char                            msg[2048] = {0,};
+        xlator_t                       *this = NULL;
+
+        GF_ASSERT (req);
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                    "unserialize req-buffer to dictionary");
+                        snprintf (msg, sizeof (msg), "Unable to decode the "
+                                  "command");
+                        goto out;
+                } else {
+                        dict->extra_stdfree = cli_req.dict.dict_val;
+                }
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Unable to get volume name");
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name, "
+                        "while handling quota command");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "type", &type);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Unable to get type of command");
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get type of cmd, "
+                        "while handling quota command");
+               goto out;
+        }
+
+        switch (type) {
+        case GF_QUOTA_OPTION_TYPE_ENABLE:
+                strncpy (operation, "enable", sizeof (operation));
+                break;
+
+        case GF_QUOTA_OPTION_TYPE_DISABLE:
+                strncpy (operation, "disable", sizeof (operation));
+                break;
+
+        case GF_QUOTA_OPTION_TYPE_LIMIT_USAGE:
+                strncpy (operation, "limit-usage", sizeof (operation));
+                break;
+
+        case GF_QUOTA_OPTION_TYPE_REMOVE:
+                strncpy (operation, "remove", sizeof (operation));
+                break;
+        }
+        ret = glusterd_op_begin_synctask (req, GD_OP_QUOTA, dict);
+
+out:
+        if (ret) {
+                if (msg[0] == '\0')
+                        snprintf (msg, sizeof (msg), "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, msg);
+        }
+
+        return ret;
+}
+
+int
+glusterd_handle_quota (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __glusterd_handle_quota);
+}
+
+int32_t
+glusterd_check_if_quota_trans_enabled (glusterd_volinfo_t *volinfo)
+{
+        int32_t  ret           = 0;
+        int      flag          = _gf_false;
+
+        flag = glusterd_volinfo_get_boolean (volinfo, VKEY_FEATURES_QUOTA);
+        if (flag == -1) {
+                gf_log ("", GF_LOG_ERROR, "failed to get the quota status");
+                ret = -1;
+                goto out;
+        }
+
+        if (flag == _gf_false) {
+                gf_log ("", GF_LOG_ERROR, "first enable the quota translator");
+                ret = -1;
+                goto out;
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+/* At the end of the function, the variable found will be set
+ * to true if the path to be removed was present in the limit-list,
+ * else will be false.
+ */
+int32_t
+_glusterd_quota_remove_limits (char **quota_limits, char *path,
+                               gf_boolean_t *found)
+{
+        int      ret      = 0;
+        int      i        = 0;
+        int      size     = 0;
+        int      len      = 0;
+        int      pathlen  = 0;
+        int      skiplen  = 0;
+        int      flag     = 0;
+        char    *limits   = NULL;
+        char    *qlimits  = NULL;
+
+        if (found != NULL)
+                *found = _gf_false;
+
+        if (*quota_limits == NULL)
+                return -1;
+
+        qlimits = *quota_limits;
+
+        pathlen = strlen (path);
+
+        len = strlen (qlimits);
+
+        limits = GF_CALLOC (len + 1, sizeof (char), gf_gld_mt_char);
+        if (!limits)
+                return -1;
+
+        while (i < len) {
+                if (!memcmp ((void *) &qlimits [i], (void *)path, pathlen))
+                        if (qlimits [i + pathlen] == ':') {
+                                flag = 1;
+                                if (found != NULL)
+                                        *found = _gf_true;
+                        }
+
+                while (qlimits [i + size] != ',' &&
+                       qlimits [i + size] != '\0')
+                        size++;
+
+                if (!flag) {
+                        memcpy ((void *) &limits [i], (void *) &qlimits [i], size + 1);
+                } else {
+                        skiplen = size + 1;
+                        size = len - i - size;
+                        memcpy ((void *) &limits [i], (void *) &qlimits [i + skiplen], size);
+                        break;
+                }
+
+                i += size + 1;
+                size = 0;
+        }
+
+        if (!flag) {
+                ret = 1;
+        } else {
+                len = strlen (limits);
+
+                if (len == 0) {
+                        GF_FREE (qlimits);
+
+                        *quota_limits = NULL;
+
+                        goto out;
+                }
+
+                if (limits[len - 1] == ',') {
+                        limits[len - 1] = '\0';
+                        len --;
+                }
+
+                GF_FREE (qlimits);
+
+                qlimits = GF_CALLOC (len + 1, sizeof (char), gf_gld_mt_char);
+
+                if (!qlimits) {
+                        ret = -1;
+                        goto out;
+                }
+
+                memcpy ((void *) qlimits, (void *) limits, len + 1);
+
+                *quota_limits = qlimits;
+
+                ret = 0;
+        }
+
+out:
+        GF_FREE (limits);
+
+        return ret;
+}
+
+int32_t
+glusterd_quota_initiate_fs_crawl (glusterd_conf_t *priv, char *volname)
+{
+        pid_t                      pid;
+        int32_t                    ret              = 0;
+        int                        status           = 0;
+        char                       mountdir[]       = "/tmp/mntXXXXXX";
+        runner_t                   runner           = {0};
+
+        if (mkdtemp (mountdir) == NULL) {
+                gf_log ("glusterd", GF_LOG_DEBUG,
+                        "failed to create a temporary mount directory");
+                ret = -1;
+                goto out;
+        }
+
+        runinit (&runner);
+        runner_add_args (&runner, SBIN_DIR"/glusterfs",
+                         "-s", "localhost",
+                         "--volfile-id", volname,
+                         "-l", DEFAULT_LOG_FILE_DIRECTORY"/quota-crawl.log",
+                         mountdir, NULL);
+
+        synclock_unlock (&priv->big_lock);
+        ret = runner_run_reuse (&runner);
+        synclock_lock (&priv->big_lock);
+        if (ret == -1) {
+                runner_log (&runner, "glusterd", GF_LOG_DEBUG, "command failed");
+                runner_end (&runner);
+                goto out;
+        }
+        runner_end (&runner);
+
+        if ((pid = fork ()) < 0) {
+                gf_log ("glusterd", GF_LOG_WARNING, "fork from parent failed");
+                ret = -1;
+                goto out;
+        } else if (pid == 0) {//first child
+                /* fork one more to not hold back main process on
+                 * blocking call below
+                 */
+                pid = fork ();
+                if (pid)
+                        _exit (pid > 0 ? EXIT_SUCCESS : EXIT_FAILURE);
+
+                ret = chdir (mountdir);
+                if (ret == -1) {
+                        gf_log ("glusterd", GF_LOG_WARNING, "chdir %s failed, "
+                                "reason: %s", mountdir, strerror (errno));
+                        exit (EXIT_FAILURE);
+                }
+                runinit (&runner);
+                runner_add_args (&runner, "/usr/bin/find", "find", ".", NULL);
+                if (runner_start (&runner) == -1)
+                        _exit (EXIT_FAILURE);
+
+#ifndef GF_LINUX_HOST_OS
+                runner_end (&runner); /* blocks in waitpid */
+                runcmd ("umount", mountdir, NULL);
+#else
+                runcmd ("umount", "-l", mountdir, NULL);
+#endif
+                rmdir (mountdir);
+                _exit (EXIT_SUCCESS);
+        }
+        ret = (waitpid (pid, &status, 0) == pid &&
+               WIFEXITED (status) && WEXITSTATUS (status) == EXIT_SUCCESS) ? 0 : -1;
+
+out:
+        return ret;
+}
+
+char *
+glusterd_quota_get_limit_value (char *quota_limits, char *path)
+{
+        int32_t i, j, k, l, len;
+        int32_t pat_len, diff;
+        char   *ret_str = NULL;
+
+        len = strlen (quota_limits);
+        pat_len = strlen (path);
+        i = 0;
+        j = 0;
+
+        while (i < len) {
+                j = i;
+                k = 0;
+                while (path [k] == quota_limits [j]) {
+                        j++;
+                        k++;
+                }
+
+                l = j;
+
+                while (quota_limits [j] != ',' &&
+                       quota_limits [j] != '\0')
+                        j++;
+
+                if (quota_limits [l] == ':' && pat_len == (l - i)) {
+                        diff = j - i;
+                        ret_str = GF_CALLOC (diff + 1, sizeof (char),
+                                             gf_gld_mt_char);
+
+                        strncpy (ret_str, &quota_limits [i], diff);
+
+                        break;
+                }
+                i = ++j; //skip ','
+        }
+
+        return ret_str;
+}
+
+char*
+_glusterd_quota_get_limit_usages (glusterd_volinfo_t *volinfo,
+                                  char *path, char **op_errstr)
+{
+        int32_t  ret          = 0;
+        char    *quota_limits = NULL;
+        char    *ret_str      = NULL;
+
+        if (volinfo == NULL)
+                return NULL;
+
+        ret = glusterd_volinfo_get (volinfo, VKEY_FEATURES_LIMIT_USAGE,
+                                    &quota_limits);
+        if (ret)
+                return NULL;
+        if (quota_limits == NULL) {
+                ret_str = NULL;
+                *op_errstr = gf_strdup ("Limit not set on any directory");
+        } else if (path == NULL)
+                ret_str = gf_strdup (quota_limits);
+        else
+                ret_str = glusterd_quota_get_limit_value (quota_limits, path);
+
+        return ret_str;
+}
+
+int32_t
+glusterd_quota_get_limit_usages (glusterd_conf_t *priv,
+                                 glusterd_volinfo_t *volinfo,
+                                 char *volname,
+                                 dict_t *dict,
+                                 char **op_errstr,
+                                 dict_t *rsp_dict)
+{
+        int32_t i               = 0;
+        int32_t ret             = 0;
+        int32_t count           = 0;
+        char    *path           = NULL;
+        char    cmd_str [1024]  = {0, };
+        char   *ret_str         = NULL;
+
+        if (rsp_dict == NULL)
+                return 0;
+
+        ret = dict_get_int32 (dict, "count", &count);
+        if (ret < 0)
+                goto out;
+
+        if (count == 0) {
+                ret_str = _glusterd_quota_get_limit_usages (volinfo, NULL,
+                                                            op_errstr);
+        } else {
+                i = 0;
+                while (count--) {
+                        snprintf (cmd_str, 1024, "path%d", i++);
+
+                        ret = dict_get_str (dict, cmd_str, &path);
+                        if (ret < 0)
+                                goto out;
+
+                        ret_str = _glusterd_quota_get_limit_usages (volinfo, path, op_errstr);
+                }
+        }
+
+        if (ret_str) {
+                ret = dict_set_dynstr (rsp_dict, "limit_list", ret_str);
+        }
+out:
+        return ret;
+}
+
+int32_t
+glusterd_quota_enable (glusterd_volinfo_t *volinfo, char **op_errstr,
+                       gf_boolean_t *crawl)
+{
+        int32_t         ret     = -1;
+        char            *quota_status = NULL;
+
+        GF_VALIDATE_OR_GOTO ("glusterd", volinfo, out);
+        GF_VALIDATE_OR_GOTO ("glusterd", crawl, out);
+        GF_VALIDATE_OR_GOTO ("glusterd", op_errstr, out);
+
+        if (glusterd_is_volume_started (volinfo) == 0) {
+                *op_errstr = gf_strdup ("Volume is stopped, start volume "
+                                        "to enable quota.");
+                goto out;
+        }
+
+        ret = glusterd_check_if_quota_trans_enabled (volinfo);
+        if (ret == 0) {
+                *op_errstr = gf_strdup ("Quota is already enabled");
+                goto out;
+        }
+
+        quota_status = gf_strdup ("on");
+        if (!quota_status) {
+                gf_log ("", GF_LOG_ERROR, "memory allocation failed");
+                *op_errstr = gf_strdup ("Enabling quota has been unsuccessful");
+                goto out;
+        }
+
+        ret = dict_set_dynstr (volinfo->dict, VKEY_FEATURES_QUOTA, quota_status);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "dict set failed");
+                *op_errstr = gf_strdup ("Enabling quota has been unsuccessful");
+                goto out;
+        }
+
+        *op_errstr = gf_strdup ("Enabling quota has been successful");
+
+        *crawl = _gf_true;
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+glusterd_quota_disable (glusterd_volinfo_t *volinfo, char **op_errstr)
+{
+        int32_t  ret            = -1;
+        char    *quota_status   = NULL, *quota_limits = NULL;
+
+        GF_VALIDATE_OR_GOTO ("glusterd", volinfo, out);
+        GF_VALIDATE_OR_GOTO ("glusterd", op_errstr, out);
+
+        ret = glusterd_check_if_quota_trans_enabled (volinfo);
+        if (ret == -1) {
+                *op_errstr = gf_strdup ("Quota is already disabled");
+                goto out;
+        }
+
+        quota_status = gf_strdup ("off");
+        if (!quota_status) {
+                gf_log ("", GF_LOG_ERROR, "memory allocation failed");
+                *op_errstr = gf_strdup ("Disabling quota has been unsuccessful");
+                goto out;
+        }
+
+        ret = dict_set_dynstr (volinfo->dict, VKEY_FEATURES_QUOTA, quota_status);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "dict set failed");
+                *op_errstr = gf_strdup ("Disabling quota has been unsuccessful");
+                goto out;
+        }
+
+        *op_errstr = gf_strdup ("Disabling quota has been successful");
+
+        ret = glusterd_volinfo_get (volinfo, VKEY_FEATURES_LIMIT_USAGE,
+                                    &quota_limits);
+        if (ret) {
+                gf_log ("", GF_LOG_WARNING, "failed to get the quota limits");
+        } else {
+                GF_FREE (quota_limits);
+        }
+
+        dict_del (volinfo->dict, VKEY_FEATURES_LIMIT_USAGE);
+
+out:
+        return ret;
+}
+
+int32_t
+glusterd_quota_limit_usage (glusterd_volinfo_t *volinfo, dict_t *dict, char **op_errstr)
+{
+        int32_t          ret    = -1;
+        char            *path   = NULL;
+        char            *limit  = NULL;
+        char            *value  = NULL;
+        char             msg [1024] = {0,};
+        char            *quota_limits = NULL;
+
+        GF_VALIDATE_OR_GOTO ("glusterd", dict, out);
+        GF_VALIDATE_OR_GOTO ("glusterd", volinfo, out);
+        GF_VALIDATE_OR_GOTO ("glusterd", op_errstr, out);
+
+        ret = glusterd_check_if_quota_trans_enabled (volinfo);
+        if (ret == -1) {
+                *op_errstr = gf_strdup ("Quota is disabled, please enable "
+                                        "quota");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_get (volinfo, VKEY_FEATURES_LIMIT_USAGE,
+                                    &quota_limits);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "failed to get the quota limits");
+                *op_errstr = gf_strdup ("failed to set limit");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "path", &path);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to fetch quota limits" );
+                *op_errstr = gf_strdup ("failed to set limit");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "limit", &limit);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to fetch quota limits" );
+                *op_errstr = gf_strdup ("failed to set limit");
+                goto out;
+        }
+
+        if (quota_limits) {
+                ret = _glusterd_quota_remove_limits (&quota_limits, path, NULL);
+                if (ret == -1) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
+                        *op_errstr = gf_strdup ("failed to set limit");
+                        goto out;
+                }
+        }
+
+        if (quota_limits == NULL) {
+                ret = gf_asprintf (&value, "%s:%s", path, limit);
+                if (ret == -1) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
+                        *op_errstr = gf_strdup ("failed to set limit");
+                        goto out;
+                }
+        } else {
+                ret = gf_asprintf (&value, "%s,%s:%s",
+                                   quota_limits, path, limit);
+                if (ret == -1) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
+                        *op_errstr = gf_strdup ("failed to set limit");
+                        goto out;
+                }
+
+                GF_FREE (quota_limits);
+        }
+
+        quota_limits = value;
+
+        ret = dict_set_str (volinfo->dict, VKEY_FEATURES_LIMIT_USAGE,
+                            quota_limits);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to set quota limits" );
+                *op_errstr = gf_strdup ("failed to set limit");
+                goto out;
+        }
+        snprintf (msg, 1024, "limit set on %s", path);
+        *op_errstr = gf_strdup (msg);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+glusterd_quota_remove_limits (glusterd_volinfo_t *volinfo, dict_t *dict, char **op_errstr)
+{
+        int32_t         ret                   = -1;
+        char            str [PATH_MAX + 1024] = {0,};
+        char            *quota_limits         = NULL;
+        char            *path                 = NULL;
+        gf_boolean_t     flag                 = _gf_false;
+
+        GF_VALIDATE_OR_GOTO ("glusterd", dict, out);
+        GF_VALIDATE_OR_GOTO ("glusterd", volinfo, out);
+        GF_VALIDATE_OR_GOTO ("glusterd", op_errstr, out);
+
+        ret = glusterd_check_if_quota_trans_enabled (volinfo);
+        if (ret == -1) {
+                *op_errstr = gf_strdup ("Quota is disabled, please enable quota");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_get (volinfo, VKEY_FEATURES_LIMIT_USAGE,
+                                    &quota_limits);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "failed to get the quota limits");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "path", &path);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to fetch quota limits" );
+                goto out;
+        }
+
+        ret = _glusterd_quota_remove_limits (&quota_limits, path, &flag);
+        if (ret == -1) {
+                if (flag == _gf_true)
+                        snprintf (str, sizeof (str), "Removing limit on %s has "
+                                  "been unsuccessful", path);
+                else
+                        snprintf (str, sizeof (str), "%s has no limit set", path);
+                *op_errstr = gf_strdup (str);
+                goto out;
+        } else {
+                if (flag == _gf_true)
+                        snprintf (str, sizeof (str), "Removed quota limit on "
+                                  "%s", path);
+                else
+                        snprintf (str, sizeof (str), "no limit set on %s",
+                                  path);
+                *op_errstr = gf_strdup (str);
+        }
+
+        if (quota_limits) {
+                ret = dict_set_str (volinfo->dict, VKEY_FEATURES_LIMIT_USAGE,
+                                    quota_limits);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to set quota limits" );
+                        goto out;
+                }
+        } else {
+               dict_del (volinfo->dict, VKEY_FEATURES_LIMIT_USAGE);
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+
+int
+glusterd_op_quota (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
+{
+        glusterd_volinfo_t     *volinfo      = NULL;
+        int32_t                 ret          = -1;
+        char                   *volname      = NULL;
+        int                     type         = -1;
+        gf_boolean_t            start_crawl  = _gf_false;
+        glusterd_conf_t        *priv         = NULL;
+
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+
+        priv = THIS->private;
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volume name " );
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "type", &type);
+
+        if (type == GF_QUOTA_OPTION_TYPE_ENABLE) {
+                ret = glusterd_quota_enable (volinfo, op_errstr, &start_crawl);
+                if (ret < 0)
+                        goto out;
+
+                goto create_vol;
+        }
+
+        if (type == GF_QUOTA_OPTION_TYPE_DISABLE) {
+                ret = glusterd_quota_disable (volinfo, op_errstr);
+                if (ret < 0)
+                        goto out;
+
+                goto create_vol;
+        }
+
+        if (type == GF_QUOTA_OPTION_TYPE_LIMIT_USAGE) {
+                ret = glusterd_quota_limit_usage (volinfo, dict, op_errstr);
+                if (ret < 0)
+                        goto out;
+
+                goto create_vol;
+        }
+
+        if (type == GF_QUOTA_OPTION_TYPE_REMOVE) {
+                ret = glusterd_quota_remove_limits (volinfo, dict, op_errstr);
+                if (ret < 0)
+                        goto out;
+
+                goto create_vol;
+        }
+
+        if (type == GF_QUOTA_OPTION_TYPE_LIST) {
+                ret = glusterd_check_if_quota_trans_enabled (volinfo);
+                if (ret == -1) {
+                        *op_errstr = gf_strdup ("cannot list the limits, "
+                                                "quota is disabled");
+                        goto out;
+                }
+
+                ret = glusterd_quota_get_limit_usages (priv, volinfo, volname,
+                                                       dict, op_errstr, rsp_dict);
+
+                goto out;
+        }
+create_vol:
+        ret = glusterd_create_volfiles_and_notify_services (volinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to re-create volfile for"
+                                          " 'quota'");
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret)
+                goto out;
+
+        if (GLUSTERD_STATUS_STARTED == volinfo->status)
+                ret = glusterd_check_generate_start_nfs ();
+
+        ret = 0;
+
+out:
+        if (rsp_dict && start_crawl == _gf_true)
+                glusterd_quota_initiate_fs_crawl (priv, volname);
+
+        if (rsp_dict && *op_errstr) {
+                ret = dict_set_dynstr (rsp_dict, "errstr", *op_errstr);
+                if (ret) {
+                        GF_FREE (*op_errstr);
+                        gf_log ("", GF_LOG_DEBUG,
+                                "failed to set error message in ctx");
+                }
+                *op_errstr = NULL;
+        }
+
+        return ret;
+}
+
+int
+glusterd_op_stage_quota (dict_t *dict, char **op_errstr)
+{
+        int                ret           = 0;
+        char              *volname       = NULL;
+        gf_boolean_t       exists        = _gf_false;
+        int                type          = 0;
+        dict_t            *ctx           = NULL;
+
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        exists = glusterd_check_volume_exists (volname);
+        if (!exists) {
+                gf_log ("", GF_LOG_ERROR, "Volume with name: %s "
+                                "does not exist",
+                                volname);
+                *op_errstr = gf_strdup ("Invalid volume name");
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "type", &type);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get 'type' for quota op");
+                *op_errstr = gf_strdup ("Volume quota failed, internal error "
+                                        ", unable to get type of operation");
+                goto out;
+        }
+
+
+        ctx = glusterd_op_get_ctx();
+        if (ctx && (type == GF_QUOTA_OPTION_TYPE_ENABLE
+                    || type == GF_QUOTA_OPTION_TYPE_LIST)) {
+                /* Fuse mount req. only for enable & list-usage options*/
+                if (!glusterd_is_fuse_available ()) {
+                        gf_log ("glusterd", GF_LOG_ERROR, "Unable to open /dev/"
+                                "fuse (%s), quota command failed",
+                                strerror (errno));
+                        *op_errstr = gf_strdup ("Fuse unavailable");
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-rebalance.c b/xlators/mgmt/glusterd/src/glusterd-rebalance.c
new file mode 100644
index 000000000..b7b974c68
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-rebalance.c
@@ -0,0 +1,749 @@
+/*
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include <inttypes.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <sys/resource.h>
+#include <sys/statvfs.h>
+
+#include "globals.h"
+#include "compat.h"
+#include "protocol-common.h"
+#include "xlator.h"
+#include "logging.h"
+#include "timer.h"
+#include "glusterd-mem-types.h"
+#include "glusterd.h"
+#include "glusterd-sm.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-utils.h"
+#include "glusterd-store.h"
+#include "run.h"
+#include "glusterd-volgen.h"
+
+#include "syscall.h"
+#include "cli1-xdr.h"
+#include "xdr-generic.h"
+
+int32_t
+glusterd_brick_op_cbk (struct rpc_req *req, struct iovec *iov,
+                          int count, void *myframe);
+int
+glusterd_defrag_start_validate (glusterd_volinfo_t *volinfo, char *op_errstr,
+                                size_t len, glusterd_op_t op)
+{
+        int      ret = -1;
+        xlator_t *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        /* Check only if operation is not remove-brick */
+        if ((GD_OP_REMOVE_BRICK != op) &&
+            !gd_is_remove_brick_committed (volinfo)) {
+                gf_log (this->name, GF_LOG_DEBUG, "A remove-brick task on "
+                        "volume %s is not yet committed", volinfo->volname);
+                snprintf (op_errstr, len, "A remove-brick task on volume %s is"
+                          " not yet committed. Either commit or stop the "
+                          "remove-brick task.", volinfo->volname);
+                goto out;
+        }
+
+        if (glusterd_is_defrag_on (volinfo)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "rebalance on volume %s already started",
+                        volinfo->volname);
+                snprintf (op_errstr, len, "Rebalance on %s is already started",
+                          volinfo->volname);
+                goto out;
+        }
+
+        if (glusterd_is_rb_started (volinfo) ||
+            glusterd_is_rb_paused (volinfo)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Rebalance failed as replace brick is in progress on volume %s",
+                        volinfo->volname);
+                snprintf (op_errstr, len, "Rebalance failed as replace brick is in progress on "
+                          "volume %s", volinfo->volname);
+                goto out;
+        }
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+
+int32_t
+__glusterd_defrag_notify (struct rpc_clnt *rpc, void *mydata,
+                          rpc_clnt_event_t event, void *data)
+{
+        glusterd_volinfo_t      *volinfo = NULL;
+        glusterd_defrag_info_t  *defrag  = NULL;
+        int                     ret      = 0;
+        char                    pidfile[PATH_MAX];
+        glusterd_conf_t        *priv    = NULL;
+
+        priv = THIS->private;
+        if (!priv)
+                return 0;
+
+        volinfo = mydata;
+        if (!volinfo)
+                return 0;
+
+        defrag = volinfo->rebal.defrag;
+        if (!defrag)
+                return 0;
+
+        if ((event == RPC_CLNT_DISCONNECT) && defrag->connected)
+                volinfo->rebal.defrag = NULL;
+
+        GLUSTERD_GET_DEFRAG_PID_FILE(pidfile, volinfo, priv);
+
+        switch (event) {
+        case RPC_CLNT_CONNECT:
+        {
+                if (defrag->connected)
+                        return 0;
+
+                LOCK (&defrag->lock);
+                {
+                        defrag->connected = 1;
+                }
+                UNLOCK (&defrag->lock);
+
+               gf_log ("", GF_LOG_DEBUG, "%s got RPC_CLNT_CONNECT",
+                        rpc->conn.trans->name);
+               break;
+        }
+
+        case RPC_CLNT_DISCONNECT:
+        {
+                if (!defrag->connected)
+                        return 0;
+
+                LOCK (&defrag->lock);
+                {
+                        defrag->connected = 0;
+                }
+                UNLOCK (&defrag->lock);
+
+                if (!glusterd_is_service_running (pidfile, NULL)) {
+                        if (volinfo->rebal.defrag_status ==
+                                                GF_DEFRAG_STATUS_STARTED) {
+                                volinfo->rebal.defrag_status =
+                                                   GF_DEFRAG_STATUS_FAILED;
+                        }
+                 }
+
+                glusterd_store_perform_node_state_store (volinfo);
+
+                if (defrag->rpc) {
+                        rpc_clnt_unref (defrag->rpc);
+                        defrag->rpc = NULL;
+                }
+                if (defrag->cbk_fn)
+                        defrag->cbk_fn (volinfo,
+                                        volinfo->rebal.defrag_status);
+
+                GF_FREE (defrag);
+                gf_log ("", GF_LOG_DEBUG, "%s got RPC_CLNT_DISCONNECT",
+                        rpc->conn.trans->name);
+                break;
+        }
+        default:
+                gf_log ("", GF_LOG_TRACE,
+                        "got some other RPC event %d", event);
+                ret = 0;
+                break;
+        }
+
+        return ret;
+}
+
+int32_t
+glusterd_defrag_notify (struct rpc_clnt *rpc, void *mydata,
+                        rpc_clnt_event_t event, void *data)
+{
+        return glusterd_big_locked_notify (rpc, mydata, event,
+                                           data, __glusterd_defrag_notify);
+}
+
+int
+glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr,
+                              size_t len, int cmd, defrag_cbk_fn_t cbk,
+                              glusterd_op_t op)
+{
+        int                    ret = -1;
+        glusterd_defrag_info_t *defrag =  NULL;
+        runner_t               runner = {0,};
+        glusterd_conf_t        *priv = NULL;
+        char                   defrag_path[PATH_MAX];
+        char                   sockfile[PATH_MAX] = {0,};
+        char                   pidfile[PATH_MAX] = {0,};
+        char                   logfile[PATH_MAX] = {0,};
+        dict_t                 *options = NULL;
+        char                   valgrind_logfile[PATH_MAX] = {0,};
+
+        priv    = THIS->private;
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (op_errstr);
+
+        ret = glusterd_defrag_start_validate (volinfo, op_errstr, len, op);
+        if (ret)
+                goto out;
+        if (!volinfo->rebal.defrag)
+                volinfo->rebal.defrag =
+                        GF_CALLOC (1, sizeof (*volinfo->rebal.defrag),
+                                   gf_gld_mt_defrag_info);
+        if (!volinfo->rebal.defrag)
+                goto out;
+
+        defrag = volinfo->rebal.defrag;
+
+        defrag->cmd = cmd;
+
+        volinfo->rebal.defrag_cmd = cmd;
+        volinfo->rebal.op = op;
+
+        LOCK_INIT (&defrag->lock);
+
+        volinfo->rebal.defrag_status = GF_DEFRAG_STATUS_STARTED;
+
+        glusterd_volinfo_reset_defrag_stats (volinfo);
+        glusterd_store_perform_node_state_store (volinfo);
+
+        GLUSTERD_GET_DEFRAG_DIR (defrag_path, volinfo, priv);
+        ret = mkdir_p (defrag_path, 0777, _gf_true);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to create "
+                        "directory %s", defrag_path);
+                goto out;
+        }
+
+        GLUSTERD_GET_DEFRAG_SOCK_FILE (sockfile, volinfo, priv);
+        GLUSTERD_GET_DEFRAG_PID_FILE (pidfile, volinfo, priv);
+        snprintf (logfile, PATH_MAX, "%s/%s-rebalance.log",
+                    DEFAULT_LOG_FILE_DIRECTORY, volinfo->volname);
+        runinit (&runner);
+
+        if (priv->valgrind) {
+                snprintf (valgrind_logfile, PATH_MAX,
+                          "%s/valgrind-%s-rebalance.log",
+                          DEFAULT_LOG_FILE_DIRECTORY,
+                          volinfo->volname);
+
+                runner_add_args (&runner, "valgrind", "--leak-check=full",
+                                 "--trace-children=yes", "--track-origins=yes",
+                                 NULL);
+                runner_argprintf (&runner, "--log-file=%s", valgrind_logfile);
+        }
+
+        runner_add_args (&runner, SBIN_DIR"/glusterfs",
+                         "-s", "localhost", "--volfile-id", volinfo->volname,
+                         "--xlator-option", "*dht.use-readdirp=yes",
+                         "--xlator-option", "*dht.lookup-unhashed=yes",
+                         "--xlator-option", "*dht.assert-no-child-down=yes",
+                         "--xlator-option", "*replicate*.data-self-heal=off",
+                         "--xlator-option",
+                         "*replicate*.metadata-self-heal=off",
+                         "--xlator-option", "*replicate*.entry-self-heal=off",
+                         "--xlator-option", "*replicate*.readdir-failover=off",
+                         "--xlator-option", "*dht.readdir-optimize=on",
+                         NULL);
+        runner_add_arg (&runner, "--xlator-option");
+        runner_argprintf ( &runner, "*dht.rebalance-cmd=%d",cmd);
+        runner_add_arg (&runner, "--xlator-option");
+        runner_argprintf (&runner, "*dht.node-uuid=%s", uuid_utoa(MY_UUID));
+        runner_add_arg (&runner, "--socket-file");
+        runner_argprintf (&runner, "%s",sockfile);
+        runner_add_arg (&runner, "--pid-file");
+        runner_argprintf (&runner, "%s",pidfile);
+        runner_add_arg (&runner, "-l");
+        runner_argprintf (&runner, logfile);
+        if (volinfo->memory_accounting)
+                runner_add_arg (&runner, "--mem-accounting");
+
+        ret = runner_run_nowait (&runner);
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_DEBUG, "rebalance command failed");
+                goto out;
+        }
+
+        sleep (5);
+
+        /* Setting frame-timeout to 10mins (600seconds).
+         * Unix domain sockets ensures that the connection is reliable. The
+         * default timeout of 30mins used for unreliable network connections is
+         * too long for unix domain socket connections.
+         */
+        ret = rpc_transport_unix_options_build (&options, sockfile, 600);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Unix options build failed");
+                goto out;
+        }
+
+        synclock_unlock (&priv->big_lock);
+        ret = glusterd_rpc_create (&defrag->rpc, options,
+                                   glusterd_defrag_notify, volinfo);
+        synclock_lock (&priv->big_lock);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "RPC create failed");
+                goto out;
+        }
+
+        if (cbk)
+                defrag->cbk_fn = cbk;
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+
+int
+glusterd_rebalance_rpc_create (glusterd_volinfo_t *volinfo,
+                               glusterd_conf_t *priv, int cmd)
+{
+        dict_t                  *options = NULL;
+        char                     sockfile[PATH_MAX] = {0,};
+        int                      ret = -1;
+        glusterd_defrag_info_t  *defrag =  NULL;
+
+        if (!volinfo->rebal.defrag)
+                volinfo->rebal.defrag =
+                        GF_CALLOC (1, sizeof (*volinfo->rebal.defrag),
+                                   gf_gld_mt_defrag_info);
+
+        if (!volinfo->rebal.defrag)
+                goto out;
+
+        defrag = volinfo->rebal.defrag;
+
+        defrag->cmd = cmd;
+
+        LOCK_INIT (&defrag->lock);
+
+        GLUSTERD_GET_DEFRAG_SOCK_FILE (sockfile, volinfo, priv);
+
+        /* Setting frame-timeout to 10mins (600seconds).
+         * Unix domain sockets ensures that the connection is reliable. The
+         * default timeout of 30mins used for unreliable network connections is
+         * too long for unix domain socket connections.
+         */
+        ret = rpc_transport_unix_options_build (&options, sockfile, 600);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Unix options build failed");
+                goto out;
+        }
+
+        synclock_unlock (&priv->big_lock);
+        ret = glusterd_rpc_create (&defrag->rpc, options,
+                                   glusterd_defrag_notify, volinfo);
+        synclock_lock (&priv->big_lock);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "RPC create failed");
+                goto out;
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_rebalance_cmd_validate (int cmd, char *volname,
+                                 glusterd_volinfo_t **volinfo,
+                                 char *op_errstr, size_t len)
+{
+        int ret = -1;
+
+        if (glusterd_volinfo_find(volname, volinfo)) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Received rebalance on invalid"
+                        " volname %s", volname);
+                snprintf (op_errstr, len, "Volume %s does not exist",
+                          volname);
+                goto out;
+        }
+        if ((*volinfo)->brick_count <= (*volinfo)->dist_leaf_count) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Volume %s is not a "
+                "distribute type or contains only 1 brick", volname);
+                snprintf (op_errstr, len, "Volume %s is not a distribute "
+                          "volume or contains only 1 brick.\n"
+                          "Not performing rebalance", volname);
+                goto out;
+        }
+
+        if ((*volinfo)->status != GLUSTERD_STATUS_STARTED) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Received rebalance on stopped"
+                        " volname %s", volname);
+                snprintf (op_errstr, len, "Volume %s needs to "
+                          "be started to perform rebalance", volname);
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+__glusterd_handle_defrag_volume (rpcsvc_request_t *req)
+{
+        int32_t                 ret       = -1;
+        gf_cli_req              cli_req   = {{0,}};
+        glusterd_conf_t        *priv      = NULL;
+        dict_t                 *dict      = NULL;
+        char                   *volname   = NULL;
+        gf_cli_defrag_type      cmd       = 0;
+        char                    msg[2048] = {0,};
+        xlator_t               *this      = NULL;
+
+        GF_ASSERT (req);
+        this = THIS;
+        GF_ASSERT (this);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (msg, sizeof (msg), "Unable to decode the "
+                                  "command");
+                        goto out;
+                }
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Failed to get volume name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "rebalance-command", (int32_t*)&cmd);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Failed to get command");
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        ret = dict_set_static_bin (dict, "node-uuid", MY_UUID, 16);
+        if (ret)
+                goto out;
+
+        if ((cmd == GF_DEFRAG_CMD_STATUS) ||
+              (cmd == GF_DEFRAG_CMD_STOP)) {
+                ret = glusterd_op_begin (req, GD_OP_DEFRAG_BRICK_VOLUME,
+                                         dict, msg, sizeof (msg));
+        } else
+                ret = glusterd_op_begin (req, GD_OP_REBALANCE, dict,
+                                         msg, sizeof (msg));
+
+out:
+
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
+        if (ret) {
+                if (msg[0] == '\0')
+                        snprintf (msg, sizeof (msg), "Operation failed");
+                ret = glusterd_op_send_cli_response (GD_OP_REBALANCE, ret, 0,
+                                                     req, dict, msg);
+
+        }
+
+        free (cli_req.dict.dict_val);//malloced by xdr
+
+        return 0;
+}
+
+int
+glusterd_handle_defrag_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, __glusterd_handle_defrag_volume);
+}
+
+
+int
+glusterd_op_stage_rebalance (dict_t *dict, char **op_errstr)
+{
+        char                    *volname = NULL;
+        int                     ret = 0;
+        int32_t                 cmd = 0;
+        char                    msg[2048] = {0};
+        glusterd_volinfo_t      *volinfo = NULL;
+        char                    *task_id_str = NULL;
+        dict_t                  *op_ctx = NULL;
+        xlator_t                *this = 0;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "volname not found");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "rebalance-command", &cmd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "cmd not found");
+                goto out;
+        }
+
+        ret = glusterd_rebalance_cmd_validate (cmd, volname, &volinfo,
+                                               msg, sizeof (msg));
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "failed to validate");
+                goto out;
+        }
+        switch (cmd) {
+        case GF_DEFRAG_CMD_START:
+        case GF_DEFRAG_CMD_START_LAYOUT_FIX:
+        case GF_DEFRAG_CMD_START_FORCE:
+                if (is_origin_glusterd (dict)) {
+                        op_ctx = glusterd_op_get_ctx ();
+                        if (!op_ctx) {
+                                ret = -1;
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to get op_ctx");
+                                goto out;
+                        }
+
+                        ret = glusterd_generate_and_set_task_id
+                                (op_ctx, GF_REBALANCE_TID_KEY);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to generate task-id");
+                                goto out;
+                        }
+                } else {
+                        ret = dict_get_str (dict, GF_REBALANCE_TID_KEY,
+                                            &task_id_str);
+                        if (ret) {
+                                snprintf (msg, sizeof (msg),
+                                          "Missing rebalance-id");
+                                gf_log (this->name, GF_LOG_WARNING, "%s", msg);
+                                ret = 0;
+                        }
+                }
+                ret = glusterd_defrag_start_validate (volinfo, msg,
+                                                      sizeof (msg),
+                                                      GD_OP_REBALANCE);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                        "start validate failed");
+                        goto out;
+                }
+                break;
+        case GF_DEFRAG_CMD_STATUS:
+        case GF_DEFRAG_CMD_STOP:
+                break;
+        default:
+                break;
+        }
+
+        ret = 0;
+out:
+        if (ret && op_errstr && msg[0])
+                *op_errstr = gf_strdup (msg);
+
+        return ret;
+}
+
+
+int
+glusterd_op_rebalance (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
+{
+        char                    *volname   = NULL;
+        int                     ret       = 0;
+        int32_t                 cmd       = 0;
+        char                    msg[2048] = {0};
+        glusterd_volinfo_t      *volinfo   = NULL;
+        glusterd_conf_t         *priv      = NULL;
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        glusterd_brickinfo_t    *tmp      = NULL;
+        gf_boolean_t            volfile_update = _gf_false;
+        char                    *task_id_str = NULL;
+        dict_t                  *ctx = NULL;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "volname not given");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "rebalance-command", &cmd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "command not given");
+                goto out;
+        }
+
+
+        ret = glusterd_rebalance_cmd_validate (cmd, volname, &volinfo,
+                                               msg, sizeof (msg));
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "cmd validate failed");
+                goto out;
+        }
+
+        /* Set task-id, if available, in op_ctx dict for operations other than
+         * start
+         */
+        if (cmd == GF_DEFRAG_CMD_STATUS || cmd == GF_DEFRAG_CMD_STOP) {
+                if (!uuid_is_null (volinfo->rebal.rebalance_id)) {
+                        ctx = glusterd_op_get_ctx ();
+                        if (!ctx) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to get op_ctx");
+                                ret = -1;
+                                goto out;
+                        }
+
+                        if (GD_OP_REMOVE_BRICK == volinfo->rebal.op)
+                                ret = glusterd_copy_uuid_to_dict
+                                        (volinfo->rebal.rebalance_id, ctx,
+                                         GF_REMOVE_BRICK_TID_KEY);
+                        else
+                                ret = glusterd_copy_uuid_to_dict
+                                        (volinfo->rebal.rebalance_id, ctx,
+                                         GF_REBALANCE_TID_KEY);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set task-id");
+                                goto out;
+                        }
+                }
+        }
+
+        switch (cmd) {
+        case GF_DEFRAG_CMD_START:
+        case GF_DEFRAG_CMD_START_LAYOUT_FIX:
+        case GF_DEFRAG_CMD_START_FORCE:
+                ret = dict_get_str (dict, GF_REBALANCE_TID_KEY, &task_id_str);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_DEBUG, "Missing rebalance "
+                                "id");
+                        ret = 0;
+                } else {
+                        uuid_parse (task_id_str, volinfo->rebal.rebalance_id) ;
+                        volinfo->rebal.op = GD_OP_REBALANCE;
+                }
+                ret = glusterd_handle_defrag_start (volinfo, msg, sizeof (msg),
+                                                    cmd, NULL, GD_OP_REBALANCE);
+                 break;
+        case GF_DEFRAG_CMD_STOP:
+                /* Clear task-id only on explicitly stopping rebalance.
+                 * Also clear the stored operation, so it doesn't cause trouble
+                 * with future rebalance/remove-brick starts
+                 */
+                uuid_clear (volinfo->rebal.rebalance_id);
+                volinfo->rebal.op = GD_OP_NONE;
+
+                /* Fall back to the old volume file in case of decommission*/
+                list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks,
+                                          brick_list) {
+                        if (!brickinfo->decommissioned)
+                                continue;
+                        brickinfo->decommissioned = 0;
+                        volfile_update = _gf_true;
+                }
+
+                if (volfile_update == _gf_false) {
+                        ret = 0;
+                        break;
+                }
+
+                ret = glusterd_create_volfiles_and_notify_services (volinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "failed to create volfiles");
+                        goto out;
+                }
+
+                ret = glusterd_store_volinfo (volinfo,
+                                             GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "failed to store volinfo");
+                        goto out;
+                }
+
+                ret = 0;
+                break;
+
+        case GF_DEFRAG_CMD_STATUS:
+                break;
+        default:
+                break;
+        }
+
+out:
+        if (ret && op_errstr && msg[0])
+                *op_errstr = gf_strdup (msg);
+
+        return ret;
+}
+
+int32_t
+glusterd_defrag_event_notify_handle (dict_t *dict)
+{
+        glusterd_volinfo_t      *volinfo = NULL;
+        char                    *volname = NULL;
+        int32_t                  ret     = -1;
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Failed to get volname");
+                return ret;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Failed to get volinfo for %s"
+                        , volname);
+                return ret;
+        }
+
+        ret = glusterd_defrag_volume_status_update (volinfo, dict);
+
+        if (ret)
+                gf_log ("", GF_LOG_ERROR, "Failed to update status");
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-replace-brick.c b/xlators/mgmt/glusterd/src/glusterd-replace-brick.c
new file mode 100644
index 000000000..54b830870
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-replace-brick.c
@@ -0,0 +1,2024 @@
+/*
+   Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "common-utils.h"
+#include "cli1-xdr.h"
+#include "xdr-generic.h"
+#include "glusterfs.h"
+#include "glusterd.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-store.h"
+#include "glusterd-utils.h"
+#include "glusterd-volgen.h"
+#include "run.h"
+#include "syscall.h"
+
+#include <signal.h>
+
+#define GLUSTERD_GET_RB_MNTPT(path, len, volinfo)                           \
+        snprintf (path, len,                                                \
+                  DEFAULT_VAR_RUN_DIRECTORY"/%s-"RB_CLIENT_MOUNTPOINT,      \
+                  volinfo->volname);
+
+extern uuid_t global_txn_id;
+
+int
+glusterd_get_replace_op_str (gf1_cli_replace_op op, char *op_str)
+{
+        int     ret = -1;
+
+        if (!op_str)
+                goto out;
+
+        switch (op) {
+                case GF_REPLACE_OP_START:
+                        strcpy (op_str, "start");
+                        break;
+                case GF_REPLACE_OP_COMMIT:
+                        strcpy (op_str, "commit");
+                        break;
+                case GF_REPLACE_OP_PAUSE:
+                        strcpy (op_str, "pause");
+                        break;
+                case GF_REPLACE_OP_ABORT:
+                        strcpy (op_str, "abort");
+                        break;
+                case GF_REPLACE_OP_STATUS:
+                        strcpy (op_str, "status");
+                        break;
+                case GF_REPLACE_OP_COMMIT_FORCE:
+                        strcpy (op_str, "commit-force");
+                        break;
+                default:
+                        strcpy (op_str, "unknown");
+                        break;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+__glusterd_handle_replace_brick (rpcsvc_request_t *req)
+{
+        int32_t                         ret = -1;
+        gf_cli_req                      cli_req = {{0,}};
+        dict_t                          *dict = NULL;
+        char                            *src_brick = NULL;
+        char                            *dst_brick = NULL;
+        int32_t                         op = 0;
+        char                            operation[256];
+        glusterd_op_t                   cli_op = GD_OP_REPLACE_BRICK;
+        char                            *volname = NULL;
+        char                            msg[2048] = {0,};
+        xlator_t                        *this = NULL;
+
+        GF_ASSERT (req);
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_INFO, "Received replace brick req");
+
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (msg, sizeof (msg), "Unable to decode the "
+                                  "command");
+                        goto out;
+                }
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Could not get volume name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "operation", &op);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "dict_get on operation failed");
+                snprintf (msg, sizeof (msg), "Could not get operation");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "src-brick", &src_brick);
+
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Failed to get src brick");
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+        gf_log (this->name, GF_LOG_DEBUG,
+                "src brick=%s", src_brick);
+
+        ret = dict_get_str (dict, "dst-brick", &dst_brick);
+
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Failed to get dest brick");
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        (void) glusterd_get_replace_op_str (op, operation);
+        gf_log (this->name, GF_LOG_DEBUG, "dst brick=%s", dst_brick);
+        gf_log (this->name, GF_LOG_INFO, "Received replace brick %s request",
+                operation);
+
+        ret = glusterd_op_begin (req, GD_OP_REPLACE_BRICK, dict,
+                                 msg, sizeof (msg));
+
+out:
+        free (cli_req.dict.dict_val);//malloced by xdr
+
+        glusterd_friend_sm ();
+        glusterd_op_sm ();
+
+        if (ret) {
+                if (msg[0] == '\0')
+                        snprintf (msg, sizeof (msg), "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, msg);
+        }
+
+        return ret;
+}
+
+int
+glusterd_handle_replace_brick (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_replace_brick);
+}
+
+static int
+glusterd_get_rb_dst_brickinfo (glusterd_volinfo_t *volinfo,
+                               glusterd_brickinfo_t **brickinfo)
+{
+        int32_t                 ret = -1;
+
+        if (!volinfo || !brickinfo)
+                goto out;
+
+        *brickinfo = volinfo->rep_brick.dst_brick;
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+int
+glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr,
+                                 dict_t *rsp_dict)
+{
+        int                                      ret           = 0;
+        int32_t                                  port          = 0;
+        char                                    *src_brick     = NULL;
+        char                                    *dst_brick     = NULL;
+        char                                    *volname       = NULL;
+        int                                      replace_op    = 0;
+        glusterd_volinfo_t                      *volinfo       = NULL;
+        glusterd_brickinfo_t                    *src_brickinfo = NULL;
+        char                                    *host          = NULL;
+        char                                    *path          = NULL;
+        char                                     msg[2048]     = {0};
+        char                                    *dup_dstbrick  = NULL;
+        glusterd_peerinfo_t                     *peerinfo = NULL;
+        glusterd_brickinfo_t                    *dst_brickinfo = NULL;
+        gf_boolean_t                             is_run        = _gf_false;
+        dict_t                                  *ctx           = NULL;
+        glusterd_conf_t                         *priv          = NULL;
+        char                                    *savetok       = NULL;
+        char                                     pidfile[PATH_MAX] = {0};
+        char                                    *task_id_str = NULL;
+        xlator_t                                *this = NULL;
+        gf_boolean_t                            is_force = _gf_false;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_str (dict, "src-brick", &src_brick);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get src brick");
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "src brick=%s", src_brick);
+
+        ret = dict_get_str (dict, "dst-brick", &dst_brick);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get dest brick");
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "dst brick=%s", dst_brick);
+
+        ret = dict_get_str (dict, "volname", &volname);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "operation", (int32_t *)&replace_op);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "dict get on replace-brick operation failed");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "volume: %s does not exist",
+                          volname);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        if (GLUSTERD_STATUS_STARTED != volinfo->status) {
+                ret = -1;
+                snprintf (msg, sizeof (msg), "volume: %s is not started",
+                          volname);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        if (!glusterd_store_is_valid_brickpath (volname, dst_brick) ||
+                !glusterd_is_valid_volfpath (volname, dst_brick)) {
+                snprintf (msg, sizeof (msg), "brick path %s is too "
+                          "long.", dst_brick);
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_check_gsync_running (volinfo, &is_run);
+        if (ret && (is_run == _gf_false))
+                gf_log (this->name, GF_LOG_WARNING, "Unable to get the status"
+                                " of active "GEOREP" session");
+        if (is_run) {
+                gf_log (this->name, GF_LOG_WARNING, GEOREP" sessions active"
+                        "for the volume %s ", volname);
+                snprintf (msg, sizeof(msg), GEOREP" sessions are active "
+                                "for the volume %s.\nStop "GEOREP " sessions "
+                                "involved in this volume. Use 'volume "GEOREP
+                                " status' command for more info.",
+                                volname);
+                *op_errstr = gf_strdup (msg);
+                ret = -1;
+                goto out;
+        }
+
+        if (glusterd_is_defrag_on(volinfo)) {
+                snprintf (msg, sizeof(msg), "Volume name %s rebalance is in "
+                          "progress. Please retry after completion", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                ret = -1;
+                goto out;
+        }
+
+        ctx = glusterd_op_get_ctx();
+
+        switch (replace_op) {
+        case GF_REPLACE_OP_START:
+                if (glusterd_is_rb_started (volinfo)) {
+                        snprintf (msg, sizeof (msg), "Replace brick is already "
+                                  "started for volume");
+                        gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                        *op_errstr = gf_strdup (msg);
+                        ret = -1;
+                        goto out;
+                }
+                if (is_origin_glusterd (dict)) {
+                        if (!ctx) {
+                                ret = -1;
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to get op_ctx");
+                                goto out;
+                        }
+
+                        ret = glusterd_generate_and_set_task_id
+                                (ctx, GF_REPLACE_BRICK_TID_KEY);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to generate task-id");
+                                goto out;
+                        }
+
+                } else {
+                        ret = dict_get_str (dict, GF_REPLACE_BRICK_TID_KEY,
+                                            &task_id_str);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "Missing replace-brick-id");
+                                ret = 0;
+                        }
+                }
+                is_force = dict_get_str_boolean (dict, "force", _gf_false);
+
+                break;
+
+        case GF_REPLACE_OP_PAUSE:
+                if (glusterd_is_rb_paused (volinfo)) {
+                        gf_log (this->name, GF_LOG_ERROR, "Replace brick is "
+                                "already paused for volume ");
+                        ret = -1;
+                        goto out;
+                } else if (!glusterd_is_rb_started(volinfo)) {
+                        gf_log (this->name, GF_LOG_ERROR, "Replace brick is not"
+                                " started for volume ");
+                        ret = -1;
+                        goto out;
+                }
+                break;
+
+        case GF_REPLACE_OP_ABORT:
+                if (!glusterd_is_rb_ongoing (volinfo)) {
+                        gf_log (this->name, GF_LOG_ERROR, "Replace brick is not"
+                                " started or paused for volume ");
+                        ret = -1;
+                        goto out;
+                }
+                break;
+
+        case GF_REPLACE_OP_COMMIT:
+                if (!glusterd_is_rb_ongoing (volinfo)) {
+                        gf_log (this->name, GF_LOG_ERROR, "Replace brick is not "
+                                "started for volume ");
+                        ret = -1;
+                        goto out;
+                }
+                break;
+
+        case GF_REPLACE_OP_COMMIT_FORCE:
+                is_force = _gf_true;
+                break;
+
+        case GF_REPLACE_OP_STATUS:
+
+                if (glusterd_is_rb_ongoing (volinfo) == _gf_false) {
+                        ret = gf_asprintf (op_errstr, "replace-brick not"
+                                           " started on volume %s",
+                                           volinfo->volname);
+                        if (ret < 0) {
+                                *op_errstr = NULL;
+                                goto out;
+                        }
+
+                        gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr);
+                        ret = -1;
+                        goto out;
+                }
+                break;
+
+        default:
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_volume_brickinfo_get_by_brick (src_brick, volinfo,
+                                                      &src_brickinfo);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "brick: %s does not exist in "
+                          "volume: %s", src_brick, volname);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        if (ctx) {
+                if (!glusterd_is_fuse_available ()) {
+                        gf_log (this->name, GF_LOG_ERROR, "Unable to open /dev/"
+                                "fuse (%s), replace-brick command failed",
+                                strerror (errno));
+                        snprintf (msg, sizeof(msg), "Fuse unavailable\n "
+                                "Replace-brick failed");
+                        *op_errstr = gf_strdup (msg);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        if (gf_is_local_addr (src_brickinfo->hostname)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "I AM THE SOURCE HOST");
+                if (src_brickinfo->port && rsp_dict) {
+                        ret = dict_set_int32 (rsp_dict, "src-brick-port",
+                                              src_brickinfo->port);
+                        if (ret) {
+                                gf_log ("", GF_LOG_DEBUG,
+                                        "Could not set src-brick-port=%d",
+                                        src_brickinfo->port);
+                        }
+                }
+
+                GLUSTERD_GET_BRICK_PIDFILE (pidfile, volinfo, src_brickinfo,
+                                            priv);
+                if ((replace_op != GF_REPLACE_OP_COMMIT_FORCE) &&
+                    !glusterd_is_service_running (pidfile, NULL)) {
+                        snprintf(msg, sizeof(msg), "Source brick %s:%s "
+                                 "is not online.", src_brickinfo->hostname,
+                                 src_brickinfo->path);
+                        *op_errstr = gf_strdup (msg);
+                        ret = -1;
+                        goto out;
+                }
+
+
+        }
+
+        dup_dstbrick = gf_strdup (dst_brick);
+        if (!dup_dstbrick) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR, "Memory allocation failed");
+                goto out;
+        }
+        host = strtok_r (dup_dstbrick, ":", &savetok);
+        path = strtok_r (NULL, ":", &savetok);
+
+        if (!host || !path) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "dst brick %s is not of form <HOSTNAME>:<export-dir>",
+                        dst_brick);
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_brickinfo_new_from_brick (dst_brick, &dst_brickinfo);
+        if (ret)
+                goto out;
+
+        ret = glusterd_new_brick_validate (dst_brick, dst_brickinfo,
+                                           msg, sizeof (msg));
+        if (ret) {
+                *op_errstr = gf_strdup (msg);
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr);
+                goto out;
+        }
+
+       if (!glusterd_is_rb_ongoing (volinfo) &&
+            (replace_op == GF_REPLACE_OP_START ||
+             replace_op == GF_REPLACE_OP_COMMIT_FORCE)) {
+
+                volinfo->rep_brick.src_brick = src_brickinfo;
+                volinfo->rep_brick.dst_brick = dst_brickinfo;
+        }
+
+        if (glusterd_rb_check_bricks (volinfo, src_brickinfo, dst_brickinfo)) {
+
+                ret = -1;
+                *op_errstr = gf_strdup ("Incorrect source or "
+                                        "destination brick");
+                if (*op_errstr)
+                        gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr);
+                goto out;
+       }
+
+        if (!glusterd_is_rb_ongoing (volinfo) &&
+            gf_is_local_addr (host)) {
+                ret = glusterd_validate_and_create_brickpath (dst_brickinfo,
+                                                  volinfo->volume_id,
+                                                  op_errstr, is_force);
+                if (ret)
+                        goto out;
+        }
+
+        if (!gf_is_local_addr (host)) {
+                ret = glusterd_friend_find (NULL, host, &peerinfo);
+                if (ret) {
+                        snprintf (msg, sizeof (msg), "%s, is not a friend",
+                                  host);
+                        *op_errstr = gf_strdup (msg);
+                        goto out;
+                }
+
+                if (!peerinfo->connected) {
+                        snprintf (msg, sizeof (msg), "%s, is not connected at "
+                                  "the moment", host);
+                        *op_errstr = gf_strdup (msg);
+                        ret = -1;
+                        goto out;
+                }
+
+                if (GD_FRIEND_STATE_BEFRIENDED != peerinfo->state.state) {
+                        snprintf (msg, sizeof (msg), "%s, is not befriended "
+                                  "at the moment", host);
+                        *op_errstr = gf_strdup (msg);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        if (replace_op == GF_REPLACE_OP_START &&
+            gf_is_local_addr (volinfo->rep_brick.dst_brick->hostname)) {
+                port = pmap_registry_alloc (THIS);
+                if (!port) {
+                        gf_log (THIS->name, GF_LOG_CRITICAL,
+                                "No free ports available");
+                        ret = -1;
+                        goto out;
+                }
+
+                ctx = glusterd_op_get_ctx();
+                ret = dict_set_int32 ((ctx)?ctx:rsp_dict, "dst-brick-port",
+                                      port);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_ERROR, "Failed to set dst "
+                                "brick port");
+                        goto out;
+                }
+                volinfo->rep_brick.dst_brick->port = port;
+        }
+
+        ret = 0;
+
+out:
+        GF_FREE (dup_dstbrick);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+static int
+rb_set_mntfd (int mntfd)
+{
+        int     ret = -1;
+        dict_t *ctx = NULL;
+
+        ctx = glusterd_op_get_ctx ();
+        if (!ctx) {
+                gf_log (THIS->name, GF_LOG_CRITICAL, "Failed to get op ctx");
+                goto out;
+        }
+        ret = dict_set_int32 (ctx, "mntfd", mntfd);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_DEBUG, "Failed to set mnt fd "
+                        "in op ctx");
+out:
+        return ret;
+}
+
+static int
+rb_get_mntfd (int *mntfd)
+{
+        int     ret = -1;
+        dict_t *ctx = NULL;
+
+        ctx = glusterd_op_get_ctx ();
+        if (!ctx) {
+                gf_log (THIS->name, GF_LOG_CRITICAL, "Failed to get op ctx");
+                goto out;
+        }
+        ret = dict_get_int32 (ctx, "mntfd", mntfd);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_DEBUG, "Failed to get mnt fd "
+                        "from op ctx");
+out:
+        return ret;
+}
+
+static int
+rb_regenerate_volfiles (glusterd_volinfo_t *volinfo,
+                        glusterd_brickinfo_t *brickinfo,
+                        int32_t pump_needed)
+{
+        dict_t *dict = NULL;
+        int ret = 0;
+
+        dict = volinfo->dict;
+
+        gf_log ("", GF_LOG_DEBUG,
+                "attempting to set pump value=%d", pump_needed);
+
+        ret = dict_set_int32 (dict, "enable-pump", pump_needed);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG,
+                        "could not dict_set enable-pump");
+                goto out;
+        }
+
+        ret = glusterd_create_rb_volfiles (volinfo, brickinfo);
+
+        dict_del (dict, "enable-pump");
+
+out:
+        return ret;
+}
+
+static int
+rb_src_brick_restart (glusterd_volinfo_t *volinfo,
+                      glusterd_brickinfo_t *src_brickinfo,
+                      int activate_pump)
+{
+        int                     ret = 0;
+
+        gf_log ("", GF_LOG_DEBUG,
+                "Attempting to kill src");
+
+        ret = glusterd_nfs_server_stop (volinfo);
+
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to stop nfs, ret: %d",
+                        ret);
+        }
+
+        ret = glusterd_volume_stop_glusterfs (volinfo, src_brickinfo,
+                                              _gf_false);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to stop "
+                        "glusterfs, ret: %d", ret);
+                goto out;
+        }
+
+        glusterd_delete_volfile (volinfo, src_brickinfo);
+
+        if (activate_pump) {
+                ret = rb_regenerate_volfiles (volinfo, src_brickinfo, 1);
+                if (ret) {
+                        gf_log ("", GF_LOG_DEBUG,
+                                "Could not regenerate volfiles with pump");
+                        goto out;
+                }
+        } else {
+                ret = rb_regenerate_volfiles (volinfo, src_brickinfo, 0);
+                if (ret) {
+                        gf_log ("", GF_LOG_DEBUG,
+                                "Could not regenerate volfiles without pump");
+                        goto out;
+                }
+
+        }
+
+        sleep (2);
+        ret = glusterd_volume_start_glusterfs (volinfo, src_brickinfo,
+                                               _gf_false);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to start "
+                        "glusterfs, ret: %d", ret);
+                goto out;
+        }
+
+out:
+        ret = glusterd_nfs_server_start (volinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to start nfs, ret: %d",
+                        ret);
+        }
+        return ret;
+}
+
+static int
+rb_send_xattr_command (glusterd_volinfo_t *volinfo,
+                       glusterd_brickinfo_t *src_brickinfo,
+                       glusterd_brickinfo_t *dst_brickinfo,
+                       const char *xattr_key, const char *value)
+{
+        int             ret   = -1;
+        int             mntfd = -1;
+
+        ret = rb_get_mntfd (&mntfd);
+        if (ret)
+                goto out;
+
+        ret = sys_fsetxattr (mntfd, xattr_key, value, strlen (value) + 1, 0);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_DEBUG, "setxattr on key: "
+                        "%s, reason: %s", xattr_key, strerror (errno));
+
+out:
+        return ret;
+}
+
+static int
+rb_spawn_dst_brick (glusterd_volinfo_t *volinfo,
+                    glusterd_brickinfo_t *brickinfo)
+{
+        glusterd_conf_t    *priv           = NULL;
+        runner_t            runner         = {0,};
+        int                 ret            = -1;
+        int32_t             port           = 0;
+
+        priv = THIS->private;
+
+        port = brickinfo->port;
+        GF_ASSERT (port);
+
+        runinit (&runner);
+        runner_add_arg (&runner, SBIN_DIR"/glusterfs");
+        runner_argprintf (&runner, "-f" "%s/vols/%s/"RB_DSTBRICKVOL_FILENAME,
+                          priv->workdir, volinfo->volname);
+        runner_argprintf (&runner, "-p" "%s/vols/%s/"RB_DSTBRICK_PIDFILE,
+                          priv->workdir, volinfo->volname);
+        runner_add_arg (&runner, "--xlator-option");
+        runner_argprintf (&runner, "src-server.listen-port=%d", port);
+        if (volinfo->memory_accounting)
+                runner_add_arg (&runner, "--mem-accounting");
+
+        ret = runner_run_nowait (&runner);
+        if (ret) {
+                pmap_registry_remove (THIS, 0, brickinfo->path,
+                                      GF_PMAP_PORT_BRICKSERVER, NULL);
+                gf_log ("", GF_LOG_DEBUG,
+                        "Could not start glusterfs");
+                goto out;
+        }
+
+        gf_log ("", GF_LOG_DEBUG,
+                "Successfully started glusterfs: brick=%s:%s",
+                brickinfo->hostname, brickinfo->path);
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+static int
+rb_spawn_glusterfs_client (glusterd_volinfo_t *volinfo,
+                           glusterd_brickinfo_t *brickinfo)
+{
+        xlator_t           *this            = NULL;
+        glusterd_conf_t    *priv            = NULL;
+        runner_t            runner          = {0,};
+        struct stat         buf             = {0,};
+        char                mntpt[PATH_MAX] = {0,};
+        int                 mntfd           = -1;
+        int                 ret             = -1;
+
+        this = THIS;
+        priv = this->private;
+
+        GLUSTERD_GET_RB_MNTPT (mntpt, sizeof (mntpt), volinfo);
+        runinit (&runner);
+        runner_add_arg (&runner, SBIN_DIR"/glusterfs");
+        runner_argprintf (&runner, "-f" "%s/vols/%s/"RB_CLIENTVOL_FILENAME,
+                          priv->workdir, volinfo->volname);
+        runner_add_arg (&runner, mntpt);
+        if (volinfo->memory_accounting)
+                runner_add_arg (&runner, "--mem-accounting");
+
+        ret = runner_run_reuse (&runner);
+        if (ret) {
+                runner_log (&runner, this->name, GF_LOG_DEBUG,
+                            "Could not start glusterfs");
+                runner_end (&runner);
+                goto out;
+        } else {
+                runner_log (&runner, this->name, GF_LOG_DEBUG,
+                            "Successfully started  glusterfs");
+                runner_end (&runner);
+        }
+
+        ret = stat (mntpt, &buf);
+        if (ret) {
+                 gf_log (this->name, GF_LOG_DEBUG, "stat on mount point %s "
+                         "failed", mntpt);
+                 goto out;
+        }
+
+        mntfd = open (mntpt, O_DIRECTORY);
+        if (mntfd == -1)
+                goto out;
+
+        ret = rb_set_mntfd (mntfd);
+        if (ret)
+                goto out;
+
+        runinit (&runner);
+        runner_add_args (&runner, "/bin/umount", "-l", mntpt, NULL);
+        ret = runner_run_reuse (&runner);
+        if (ret) {
+                runner_log (&runner, this->name, GF_LOG_DEBUG,
+                            "Lazy unmount failed on maintenance client");
+                runner_end (&runner);
+                goto out;
+        } else {
+                runner_log (&runner, this->name, GF_LOG_DEBUG,
+                            "Successfully unmounted  maintenance client");
+                runner_end (&runner);
+        }
+
+
+out:
+
+        return ret;
+}
+
+static const char *client_volfile_str =  "volume mnt-client\n"
+        " type protocol/client\n"
+        " option remote-host %s\n"
+        " option remote-subvolume %s\n"
+        " option remote-port %d\n"
+        " option transport-type %s\n"
+        " option username %s\n"
+        " option password %s\n"
+        "end-volume\n"
+        "volume mnt-wb\n"
+        " type performance/write-behind\n"
+        " subvolumes mnt-client\n"
+        "end-volume\n";
+
+static int
+rb_generate_client_volfile (glusterd_volinfo_t *volinfo,
+                            glusterd_brickinfo_t *src_brickinfo)
+{
+        glusterd_conf_t  *priv                  = NULL;
+        xlator_t         *this                  = NULL;
+        FILE             *file                  = NULL;
+        char              filename[PATH_MAX]    = {0, };
+        int               ret                   = -1;
+        int               fd                    = -1;
+        char             *ttype                 = NULL;
+
+        this = THIS;
+        priv = this->private;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Creating volfile");
+
+        snprintf (filename, PATH_MAX, "%s/vols/%s/%s",
+                  priv->workdir, volinfo->volname,
+                  RB_CLIENTVOL_FILENAME);
+
+        fd = open (filename, O_CREAT | O_RDONLY, S_IRUSR | S_IWUSR);
+        if (fd < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s", strerror (errno));
+                goto out;
+        }
+        close (fd);
+
+        file = fopen (filename, "w+");
+        if (!file) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Open of volfile failed");
+                ret = -1;
+                goto out;
+        }
+
+        GF_ASSERT (src_brickinfo->port);
+
+	ttype = glusterd_get_trans_type_rb (volinfo->transport_type);
+	if (NULL == ttype){
+		ret = -1;
+		goto out;
+	}
+
+        fprintf (file, client_volfile_str, src_brickinfo->hostname,
+                 src_brickinfo->path,
+                 src_brickinfo->port, ttype,
+                 glusterd_auth_get_username (volinfo),
+                 glusterd_auth_get_password (volinfo));
+
+        fclose (file);
+        GF_FREE (ttype);
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+static const char *dst_brick_volfile_str = "volume src-posix\n"
+        " type storage/posix\n"
+        " option directory %s\n"
+        " option volume-id %s\n"
+        "end-volume\n"
+        "volume %s\n"
+        " type features/locks\n"
+        " subvolumes src-posix\n"
+        "end-volume\n"
+        "volume src-server\n"
+        " type protocol/server\n"
+        " option auth.login.%s.allow %s\n"
+        " option auth.login.%s.password %s\n"
+        " option auth.addr.%s.allow *\n"
+        " option transport-type %s\n"
+        " subvolumes %s\n"
+        "end-volume\n";
+
+static int
+rb_generate_dst_brick_volfile (glusterd_volinfo_t *volinfo,
+                               glusterd_brickinfo_t *dst_brickinfo)
+{
+        glusterd_conf_t    *priv                = NULL;
+        xlator_t           *this                = NULL;
+        FILE               *file                = NULL;
+        char                filename[PATH_MAX]  = {0, };
+        int                 ret                 = -1;
+        int                 fd                  = -1;
+        char               *trans_type          = NULL;
+
+        this = THIS;
+        priv = this->private;
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Creating volfile");
+
+        snprintf (filename, PATH_MAX, "%s/vols/%s/%s",
+                  priv->workdir, volinfo->volname,
+                  RB_DSTBRICKVOL_FILENAME);
+
+        fd = creat (filename, S_IRUSR | S_IWUSR);
+        if (fd < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s", strerror (errno));
+                goto out;
+        }
+        close (fd);
+
+        file = fopen (filename, "w+");
+        if (!file) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Open of volfile failed");
+                ret = -1;
+                goto out;
+        }
+
+	trans_type = glusterd_get_trans_type_rb (volinfo->transport_type);
+	if (NULL == trans_type){
+		ret = -1;
+		goto out;
+	}
+
+        fprintf (file, dst_brick_volfile_str,
+                 dst_brickinfo->path,
+                 uuid_utoa (volinfo->volume_id),
+                 dst_brickinfo->path,
+                 dst_brickinfo->path,
+                 glusterd_auth_get_username (volinfo),
+                 glusterd_auth_get_username (volinfo),
+                 glusterd_auth_get_password (volinfo),
+                 dst_brickinfo->path,
+                 trans_type,
+                 dst_brickinfo->path);
+
+	GF_FREE (trans_type);
+
+        fclose (file);
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+
+static int
+rb_mountpoint_mkdir (glusterd_volinfo_t *volinfo,
+                     glusterd_brickinfo_t *src_brickinfo)
+{
+        char             mntpt[PATH_MAX]        = {0,};
+        int              ret                    = -1;
+
+        GLUSTERD_GET_RB_MNTPT (mntpt, sizeof (mntpt), volinfo);
+        ret = mkdir (mntpt, 0777);
+        if (ret && (errno != EEXIST)) {
+                gf_log ("", GF_LOG_DEBUG, "mkdir failed, due to %s",
+                        strerror (errno));
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+static int
+rb_mountpoint_rmdir (glusterd_volinfo_t *volinfo,
+                     glusterd_brickinfo_t *src_brickinfo)
+{
+        char             mntpt[PATH_MAX] = {0,};
+        int              ret                        = -1;
+
+        GLUSTERD_GET_RB_MNTPT (mntpt, sizeof (mntpt), volinfo);
+        ret = rmdir (mntpt);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "rmdir failed, reason: %s",
+                        strerror (errno));
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+static int
+rb_destroy_maintenance_client (glusterd_volinfo_t *volinfo,
+                               glusterd_brickinfo_t *src_brickinfo)
+{
+        xlator_t         *this                        = NULL;
+        glusterd_conf_t  *priv                        = NULL;
+        char              volfile[PATH_MAX]           = {0,};
+        int               ret                         = -1;
+        int               mntfd                       = -1;
+
+        this = THIS;
+        priv = this->private;
+
+        ret = rb_get_mntfd (&mntfd);
+        if (ret)
+                goto out;
+
+        ret = close (mntfd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to close mount "
+                        "point directory");
+                goto out;
+        }
+
+        ret = rb_mountpoint_rmdir (volinfo, src_brickinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "rmdir of mountpoint "
+                        "failed");
+                goto out;
+        }
+
+        snprintf (volfile, PATH_MAX, "%s/vols/%s/%s", priv->workdir,
+                  volinfo->volname, RB_CLIENTVOL_FILENAME);
+
+        ret = unlink (volfile);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "unlink of %s failed, reason: %s",
+                        volfile, strerror (errno));
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+static int
+rb_spawn_maintenance_client (glusterd_volinfo_t *volinfo,
+                             glusterd_brickinfo_t *src_brickinfo)
+{
+        int ret = -1;
+
+        ret = rb_generate_client_volfile (volinfo, src_brickinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Unable to generate client "
+                        "volfile");
+                goto out;
+        }
+
+        ret = rb_mountpoint_mkdir (volinfo, src_brickinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Unable to mkdir "
+                        "mountpoint");
+                goto out;
+        }
+
+        ret = rb_spawn_glusterfs_client (volinfo, src_brickinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Unable to start glusterfs");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+static int
+rb_spawn_destination_brick (glusterd_volinfo_t *volinfo,
+                            glusterd_brickinfo_t *dst_brickinfo)
+
+{
+        int ret = -1;
+
+        ret = rb_generate_dst_brick_volfile (volinfo, dst_brickinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Unable to generate client "
+                        "volfile");
+                goto out;
+        }
+
+        ret = rb_spawn_dst_brick (volinfo, dst_brickinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Unable to start glusterfs");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+static int
+rb_kill_destination_brick (glusterd_volinfo_t *volinfo,
+                           glusterd_brickinfo_t *dst_brickinfo)
+{
+        glusterd_conf_t  *priv               = NULL;
+        char              pidfile[PATH_MAX]  = {0,};
+
+        priv = THIS->private;
+
+        snprintf (pidfile, PATH_MAX, "%s/vols/%s/%s",
+                  priv->workdir, volinfo->volname,
+                  RB_DSTBRICK_PIDFILE);
+
+        return glusterd_service_stop ("brick", pidfile, SIGTERM, _gf_true);
+}
+
+static int
+rb_get_xattr_command (glusterd_volinfo_t *volinfo,
+                      glusterd_brickinfo_t *src_brickinfo,
+                      glusterd_brickinfo_t *dst_brickinfo,
+                      const char *xattr_key,
+                      char *value)
+{
+        int             ret    = -1;
+        int             mntfd  = -1;
+
+        ret = rb_get_mntfd (&mntfd);
+        if (ret)
+                goto out;
+
+        ret = sys_fgetxattr (mntfd, xattr_key, value, 8192);
+
+        if (ret < 0) {
+                gf_log (THIS->name, GF_LOG_DEBUG, "getxattr on key: %s "
+                        "failed, reason: %s", xattr_key, strerror (errno));
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+static int
+rb_send_cmd (glusterd_volinfo_t *volinfo,
+             glusterd_brickinfo_t *src,
+             glusterd_brickinfo_t *dst,
+             gf1_cli_replace_op op)
+{
+        char         start_value[8192]          = {0,};
+        char         status_str[8192]           = {0,};
+        char        *status_reply               = NULL;
+        char        *tmp                        = NULL;
+        char        *save_ptr                   = NULL;
+        char         filename[PATH_MAX]         = {0,};
+        char        *current_file               = NULL;
+        uint64_t     files                      = 0;
+        int          status                     = 0;
+        dict_t      *ctx                        = NULL;
+        int          ret                        = 0;
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (src);
+        GF_ASSERT (dst);
+        GF_ASSERT ((op > GF_REPLACE_OP_NONE)
+                   && (op <= GF_REPLACE_OP_COMMIT_FORCE));
+
+        switch (op) {
+                case GF_REPLACE_OP_START:
+                {
+                        snprintf (start_value, sizeof (start_value),
+                                  "%s:%s:%d", dst->hostname, dst->path,
+                                  dst->port);
+                        ret = rb_send_xattr_command (volinfo, src, dst,
+                                                     RB_PUMP_CMD_START,
+                                                     start_value);
+                }
+                break;
+                case GF_REPLACE_OP_PAUSE:
+                {
+                        ret = rb_send_xattr_command (volinfo, src, dst,
+                                                     RB_PUMP_CMD_PAUSE,
+                                                     RB_PUMP_DEF_ARG);
+                }
+                break;
+                case GF_REPLACE_OP_ABORT:
+                {
+                        ret = rb_send_xattr_command (volinfo, src, dst,
+                                                     RB_PUMP_CMD_ABORT,
+                                                     RB_PUMP_DEF_ARG);
+                }
+                break;
+                case GF_REPLACE_OP_COMMIT:
+                {
+                        ret = rb_send_xattr_command (volinfo, src, dst,
+                                                     RB_PUMP_CMD_COMMIT,
+                                                     RB_PUMP_DEF_ARG);
+                }
+                break;
+                case GF_REPLACE_OP_STATUS:
+                {
+                        ret = rb_get_xattr_command (volinfo, src, dst,
+                                                    RB_PUMP_CMD_STATUS,
+                                                    status_str);
+                        if (ret)
+                                goto out;
+
+                        ctx = glusterd_op_get_ctx ();
+                        GF_ASSERT (ctx);
+                        if (!ctx) {
+                                ret = -1;
+                                gf_log (THIS->name, GF_LOG_CRITICAL,
+                                        "ctx is not present.");
+                                goto out;
+                        }
+
+                        /* Split status reply into different parts */
+                        tmp = strtok_r (status_str, ":", &save_ptr);
+                        if (!tmp) {
+                                ret = -1;
+                                gf_log (THIS->name, GF_LOG_ERROR,
+                                        "Couldn't tokenize status string");
+                                goto out;
+                        }
+                        sscanf (tmp, "status=%d", &status);
+                        ret = dict_set_int32 (ctx, "status", status);
+                        if (ret) {
+                                gf_log (THIS->name, GF_LOG_ERROR, "Couldn't "
+                                        "set rb status in context");
+                                goto out;
+                        }
+
+                        tmp = NULL;
+                        tmp = strtok_r (NULL, ":", &save_ptr);
+                        if (!tmp) {
+                                ret = -1;
+                                gf_log (THIS->name, GF_LOG_ERROR,
+                                        "Couldn't tokenize status string");
+                                goto out;
+                        }
+                        sscanf (tmp, "no_of_files=%"SCNu64, &files);
+                        ret = dict_set_uint64 (ctx, "files", files);
+                        if (ret) {
+                                gf_log (THIS->name, GF_LOG_ERROR, "Couldn't "
+                                        "set rb files in context");
+                                goto out;
+                        }
+
+                        if (status == 0) {
+                                tmp = NULL;
+                                tmp = strtok_r (NULL, ":", &save_ptr);
+                                if (!tmp) {
+                                        ret = -1;
+                                        gf_log (THIS->name, GF_LOG_ERROR,
+                                                "Couldn't tokenize status "
+                                                "string");
+                                        goto out;
+                                }
+                                sscanf (tmp, "current_file=%s", filename);
+                                current_file = gf_strdup (filename);
+                                ret = dict_set_dynstr (ctx, "current_file",
+                                                       current_file);
+                                if (ret) {
+                                        GF_FREE (current_file);
+                                        gf_log (THIS->name, GF_LOG_ERROR,
+                                                "Couldn't set rb current file "
+                                                "in context");
+                                        goto out;
+                                }
+                        }
+                        if (status) {
+                                ret = gf_asprintf (&status_reply,
+                                                  "Number of files migrated = %"
+                                                  PRIu64"\tMigration complete",
+                                                  files);
+                        } else {
+                                ret = gf_asprintf (&status_reply,
+                                                  "Number of files migrated = %"
+                                                  PRIu64"\tCurrent file = %s",
+                                                  files, filename);
+                        }
+                        if (ret == -1) {
+                                gf_log (THIS->name, GF_LOG_ERROR,
+                                        "Failed to create status_reply string");
+                                goto out;
+                        }
+                        ret = dict_set_dynstr (ctx, "status-reply",
+                                               status_reply);
+                        if (ret) {
+                                GF_FREE (status_reply);
+                                gf_log (THIS->name, GF_LOG_ERROR, "Couldn't "
+                                        "set rb status response in context.");
+                                goto out;
+                        }
+                }
+                break;
+                default:
+                {
+                        GF_ASSERT (0);
+                        ret = -1;
+                        gf_log (THIS->name, GF_LOG_CRITICAL, "Invalid replace"
+                                " brick subcommand.");
+                }
+                break;
+        }
+out:
+        return ret;
+}
+
+static int
+rb_do_operation (glusterd_volinfo_t *volinfo,
+                 glusterd_brickinfo_t *src_brickinfo,
+                 glusterd_brickinfo_t *dst_brickinfo,
+                 gf1_cli_replace_op op)
+{
+
+        int             ret             = -1;
+        char            op_str[256]     = {0, };
+        xlator_t        *this           = NULL;
+
+        this = THIS;
+
+        ret = rb_spawn_maintenance_client (volinfo, src_brickinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "Could not spawn "
+                        "maintenance client");
+                goto umount;
+        }
+
+        ret = rb_send_cmd (volinfo, src_brickinfo, dst_brickinfo, op);
+        if (ret) {
+                (void) glusterd_get_replace_op_str (op, op_str);
+                gf_log (this->name, GF_LOG_DEBUG, "Sending replace-brick "
+                        "sub-command %s failed.", op_str);
+        }
+
+umount:
+        if (rb_destroy_maintenance_client (volinfo, src_brickinfo))
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to destroy "
+                        "maintenance client");
+
+         return ret;
+}
+
+/* Set src-brick's port number to be used in the maintenance mount
+ * after all commit acks are received.
+ */
+static int
+rb_update_srcbrick_port (glusterd_brickinfo_t *src_brickinfo, dict_t *rsp_dict,
+                         dict_t *req_dict, int32_t replace_op)
+{
+        xlator_t *this            = NULL;
+        dict_t   *ctx             = NULL;
+        int       ret             = 0;
+        int       dict_ret        = 0;
+        int       src_port        = 0;
+
+        this = THIS;
+
+        dict_ret = dict_get_int32 (req_dict, "src-brick-port", &src_port);
+        if (src_port)
+                src_brickinfo->port = src_port;
+
+        if (gf_is_local_addr (src_brickinfo->hostname)) {
+                gf_log ("", GF_LOG_INFO,
+                        "adding src-brick port no");
+
+                src_brickinfo->port = pmap_registry_search (this,
+                                      src_brickinfo->path, GF_PMAP_PORT_BRICKSERVER);
+                if (!src_brickinfo->port &&
+                    replace_op != GF_REPLACE_OP_COMMIT_FORCE ) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Src brick port not available");
+                        ret = -1;
+                        goto out;
+                }
+
+                if (rsp_dict) {
+                        ret = dict_set_int32 (rsp_dict, "src-brick-port", src_brickinfo->port);
+                        if (ret) {
+                                gf_log ("", GF_LOG_DEBUG,
+                                        "Could not set src-brick port no");
+                                goto out;
+                        }
+                }
+
+                ctx = glusterd_op_get_ctx ();
+                if (ctx) {
+                        ret = dict_set_int32 (ctx, "src-brick-port", src_brickinfo->port);
+                        if (ret) {
+                                gf_log ("", GF_LOG_DEBUG,
+                                        "Could not set src-brick port no");
+                                goto out;
+                        }
+                }
+
+        }
+
+out:
+        return ret;
+
+}
+
+static int
+rb_update_dstbrick_port (glusterd_brickinfo_t *dst_brickinfo, dict_t *rsp_dict,
+                         dict_t *req_dict, int32_t replace_op)
+{
+        dict_t *ctx           = NULL;
+        int     ret           = 0;
+        int     dict_ret      = 0;
+        int     dst_port      = 0;
+
+        dict_ret = dict_get_int32 (req_dict, "dst-brick-port", &dst_port);
+        if (!dict_ret)
+                dst_brickinfo->port = dst_port;
+
+
+        if (gf_is_local_addr (dst_brickinfo->hostname)) {
+                gf_log ("", GF_LOG_INFO,
+                        "adding dst-brick port no");
+
+                if (rsp_dict) {
+                        ret = dict_set_int32 (rsp_dict, "dst-brick-port",
+                                              dst_brickinfo->port);
+                        if (ret) {
+                                gf_log ("", GF_LOG_DEBUG,
+                                        "Could not set dst-brick port no in rsp dict");
+                                goto out;
+                        }
+                }
+
+                ctx = glusterd_op_get_ctx ();
+                if (ctx) {
+                        ret = dict_set_int32 (ctx, "dst-brick-port",
+                                              dst_brickinfo->port);
+                        if (ret) {
+                                gf_log ("", GF_LOG_DEBUG,
+                                        "Could not set dst-brick port no");
+                                goto out;
+                        }
+                }
+        }
+out:
+        return ret;
+}
+
+static int
+glusterd_op_perform_replace_brick (glusterd_volinfo_t  *volinfo,
+                                   char *old_brick, char  *new_brick)
+{
+        glusterd_brickinfo_t                    *old_brickinfo = NULL;
+        glusterd_brickinfo_t                    *new_brickinfo = NULL;
+        int32_t                                 ret = -1;
+
+        GF_ASSERT (volinfo);
+
+        ret = glusterd_brickinfo_new_from_brick (new_brick,
+                                                 &new_brickinfo);
+        if (ret)
+                goto out;
+
+        ret = glusterd_resolve_brick (new_brickinfo);
+
+        if (ret)
+                goto out;
+
+        ret = glusterd_volume_brickinfo_get_by_brick (old_brick,
+                                                      volinfo, &old_brickinfo);
+        if (ret)
+                goto out;
+
+        list_add_tail (&new_brickinfo->brick_list,
+                       &old_brickinfo->brick_list);
+
+        volinfo->brick_count++;
+
+        ret = glusterd_op_perform_remove_brick (volinfo, old_brick, 1, NULL);
+        if (ret)
+                goto out;
+
+        ret = glusterd_create_volfiles_and_notify_services (volinfo);
+        if (ret)
+                goto out;
+
+        if (GLUSTERD_STATUS_STARTED == volinfo->status) {
+                ret = glusterd_brick_start (volinfo, new_brickinfo, _gf_false);
+                if (ret)
+                        goto out;
+        }
+
+out:
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict)
+{
+        int                                      ret = 0;
+        dict_t                                  *ctx  = NULL;
+        int                                      replace_op = 0;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        char                                    *volname = NULL;
+        xlator_t                                *this = NULL;
+        glusterd_conf_t                         *priv = NULL;
+        char                                    *src_brick = NULL;
+        char                                    *dst_brick = NULL;
+        glusterd_brickinfo_t                    *src_brickinfo = NULL;
+        glusterd_brickinfo_t                    *dst_brickinfo = NULL;
+        char                                    *task_id_str = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_str (dict, "src-brick", &src_brick);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get src brick");
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "src brick=%s", src_brick);
+
+        ret = dict_get_str (dict, "dst-brick", &dst_brick);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get dst brick");
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "dst brick=%s", dst_brick);
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "operation", (int32_t *)&replace_op);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "dict_get on operation failed");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to allocate memory");
+                goto out;
+        }
+
+        ret = glusterd_volume_brickinfo_get_by_brick (src_brick, volinfo,
+                                                      &src_brickinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Unable to get src-brickinfo");
+                goto out;
+        }
+
+
+        ret = glusterd_get_rb_dst_brickinfo (volinfo, &dst_brickinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get "
+                         "replace brick destination brickinfo");
+                goto out;
+        }
+
+        ret = glusterd_resolve_brick (dst_brickinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Unable to resolve dst-brickinfo");
+                goto out;
+        }
+
+        ret = rb_update_srcbrick_port (src_brickinfo, rsp_dict,
+                                       dict, replace_op);
+        if (ret)
+                goto out;
+
+
+	if ((GF_REPLACE_OP_START != replace_op)) {
+
+                /* Set task-id, if available, in op_ctx dict for operations
+                 * other than start
+                 */
+                if  (is_origin_glusterd (dict)) {
+                        ctx = glusterd_op_get_ctx();
+                        if (!ctx) {
+                                gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                        "get op_ctx");
+                                ret = -1;
+                                goto out;
+                        }
+                        if (!uuid_is_null (volinfo->rep_brick.rb_id)) {
+                                ret = glusterd_copy_uuid_to_dict
+                                        (volinfo->rep_brick.rb_id, ctx,
+                                         GF_REPLACE_BRICK_TID_KEY);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Failed to set "
+                                                "replace-brick-id");
+                                        goto out;
+                                }
+                        }
+                }
+	}
+        ret = rb_update_dstbrick_port (dst_brickinfo, rsp_dict,
+                                       dict, replace_op);
+        if (ret)
+                goto out;
+
+        switch (replace_op) {
+        case GF_REPLACE_OP_START:
+        {
+                ret = dict_get_str (dict, GF_REPLACE_BRICK_TID_KEY, &task_id_str);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Missing replace-brick-id");
+                        ret = 0;
+                } else {
+                        uuid_parse (task_id_str, volinfo->rep_brick.rb_id);
+                }
+
+                if (gf_is_local_addr (dst_brickinfo->hostname)) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "I AM THE DESTINATION HOST");
+                        if (!glusterd_is_rb_paused (volinfo)) {
+                                ret = rb_spawn_destination_brick
+                                        (volinfo, dst_brickinfo);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_DEBUG,
+                                                "Failed to spawn destination "
+                                                "brick");
+                                        goto out;
+                                }
+                        } else {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Replace brick is already started=> no "
+                                        "need to restart dst brick ");
+                        }
+		}
+
+
+		if (gf_is_local_addr (src_brickinfo->hostname)) {
+		        ret = rb_src_brick_restart (volinfo, src_brickinfo,
+				                    1);
+		        if (ret) {
+				gf_log (this->name, GF_LOG_DEBUG,
+	                                "Could not restart src-brick");
+			        goto out;
+			}
+		}
+
+		if (gf_is_local_addr (dst_brickinfo->hostname)) {
+			gf_log (this->name, GF_LOG_INFO,
+				"adding dst-brick port no");
+
+                        ret = rb_update_dstbrick_port (dst_brickinfo, rsp_dict,
+                                                       dict, replace_op);
+                        if (ret)
+                                goto out;
+                }
+
+                glusterd_set_rb_status (volinfo, GF_RB_STATUS_STARTED);
+		break;
+	}
+
+        case GF_REPLACE_OP_COMMIT:
+        {
+                ctx = glusterd_op_get_ctx ();
+                if (ctx) {
+                        ret = rb_do_operation (volinfo, src_brickinfo,
+                                               dst_brickinfo,
+                                               GF_REPLACE_OP_COMMIT);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Commit operation failed");
+                                goto out;
+                        }
+                }
+        }
+                /* fall through */
+        case GF_REPLACE_OP_COMMIT_FORCE:
+        {
+                if (gf_is_local_addr (dst_brickinfo->hostname)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "I AM THE DESTINATION HOST");
+                        ret = rb_kill_destination_brick (volinfo,
+                                                         dst_brickinfo);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_CRITICAL,
+                                        "Unable to cleanup dst brick");
+                                goto out;
+                        }
+                }
+
+                ret = glusterd_nodesvcs_stop (volinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to stop nfs server, ret: %d", ret);
+                }
+
+		ret = glusterd_op_perform_replace_brick (volinfo, src_brick,
+							 dst_brick);
+		if (ret) {
+			gf_log (this->name, GF_LOG_CRITICAL, "Unable to add "
+				"dst-brick: %s to volume: %s", dst_brick,
+                                volinfo->volname);
+		        (void) glusterd_nodesvcs_handle_graph_change (volinfo);
+			goto out;
+		}
+
+		volinfo->rebal.defrag_status = 0;
+
+		ret = glusterd_nodesvcs_handle_graph_change (volinfo);
+		if (ret) {
+                        gf_log (this->name, GF_LOG_CRITICAL,
+                                "Failed to generate nfs volume file");
+		}
+
+
+		ret = glusterd_fetchspec_notify (THIS);
+                glusterd_set_rb_status (volinfo, GF_RB_STATUS_NONE);
+                glusterd_brickinfo_delete (volinfo->rep_brick.dst_brick);
+                volinfo->rep_brick.src_brick = NULL;
+                volinfo->rep_brick.dst_brick = NULL;
+                uuid_clear (volinfo->rep_brick.rb_id);
+        }
+        break;
+
+        case GF_REPLACE_OP_PAUSE:
+        {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Received pause - doing nothing");
+                ctx = glusterd_op_get_ctx ();
+                if (ctx) {
+                        ret = rb_do_operation (volinfo, src_brickinfo,
+                                               dst_brickinfo,
+                                               GF_REPLACE_OP_PAUSE);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Pause operation failed");
+                                goto out;
+                        }
+                }
+
+                glusterd_set_rb_status (volinfo, GF_RB_STATUS_PAUSED);
+        }
+                break;
+
+        case GF_REPLACE_OP_ABORT:
+        {
+
+                ctx = glusterd_op_get_ctx ();
+                if (ctx) {
+                        ret = rb_do_operation (volinfo, src_brickinfo,
+                                               dst_brickinfo,
+                                               GF_REPLACE_OP_ABORT);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Abort operation failed");
+                                goto out;
+                        }
+                }
+
+                if (gf_is_local_addr (src_brickinfo->hostname)) {
+                        ret = rb_src_brick_restart (volinfo, src_brickinfo,
+                                                    0);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Couldn't restart src brick "
+                                        "with pump xlator disabled.");
+                                goto out;
+                        }
+                }
+
+                if (gf_is_local_addr (dst_brickinfo->hostname)) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "I AM THE DESTINATION HOST");
+                        ret = rb_kill_destination_brick (volinfo, dst_brickinfo);
+                        if (ret) {
+                                gf_log ("", GF_LOG_DEBUG,
+                                        "Failed to kill destination brick");
+                                goto out;
+                        }
+                }
+                glusterd_set_rb_status (volinfo, GF_RB_STATUS_NONE);
+                glusterd_brickinfo_delete (volinfo->rep_brick.dst_brick);
+                volinfo->rep_brick.src_brick = NULL;
+                volinfo->rep_brick.dst_brick = NULL;
+        }
+        break;
+
+        case GF_REPLACE_OP_STATUS:
+        {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "received status - doing nothing");
+                ctx = glusterd_op_get_ctx ();
+                if (ctx) {
+                        if (glusterd_is_rb_paused (volinfo)) {
+                                ret = dict_set_str (ctx, "status-reply",
+                                                 "replace brick has been paused");
+                                if (ret)
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "failed to set pump status"
+                                                " in ctx");
+                                goto out;
+                        }
+
+                        ret = rb_do_operation (volinfo, src_brickinfo,
+                                               dst_brickinfo,
+                                               GF_REPLACE_OP_STATUS);
+                        if (ret)
+                                goto out;
+                }
+
+        }
+                break;
+
+        default:
+                ret = -1;
+                goto out;
+        }
+        if (!ret && replace_op != GF_REPLACE_OP_STATUS)
+		ret = glusterd_store_volinfo (volinfo,
+                                              GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "Couldn't store"
+                        " replace brick operation's state");
+
+out:
+        return ret;
+}
+
+void
+glusterd_do_replace_brick (void *data)
+{
+        glusterd_volinfo_t     *volinfo = NULL;
+        int32_t                 op      = 0;
+        int32_t                 src_port = 0;
+        int32_t                 dst_port = 0;
+        dict_t                 *dict    = NULL;
+        char                   *src_brick = NULL;
+        char                   *dst_brick = NULL;
+        char                   *volname   = NULL;
+        glusterd_brickinfo_t   *src_brickinfo = NULL;
+        glusterd_brickinfo_t   *dst_brickinfo = NULL;
+	glusterd_conf_t	       *priv = NULL;
+        uuid_t                 *txn_id = &global_txn_id;
+
+        int ret = 0;
+
+        dict = data;
+
+	GF_ASSERT (THIS);
+
+	priv = THIS->private;
+
+	if (priv->timer) {
+		gf_timer_call_cancel (THIS->ctx, priv->timer);
+		priv->timer = NULL;
+                gf_log ("", GF_LOG_DEBUG,
+                        "Cancelling timer thread");
+	}
+
+        gf_log ("", GF_LOG_DEBUG,
+                "Replace brick operation detected");
+
+        ret = dict_get_bin (dict, "transaction_id", (void **)&txn_id);
+
+        gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id));
+
+        ret = dict_get_int32 (dict, "operation", &op);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG,
+                        "dict_get on operation failed");
+                goto out;
+        }
+        ret = dict_get_str (dict, "src-brick", &src_brick);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get src brick");
+                goto out;
+        }
+
+        gf_log ("", GF_LOG_DEBUG,
+                "src brick=%s", src_brick);
+
+        ret = dict_get_str (dict, "dst-brick", &dst_brick);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get dst brick");
+                goto out;
+        }
+
+        gf_log ("", GF_LOG_DEBUG,
+                "dst brick=%s", dst_brick);
+
+        ret = dict_get_str (dict, "volname", &volname);
+
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to allocate memory");
+                goto out;
+        }
+
+        ret = glusterd_volume_brickinfo_get_by_brick (src_brick, volinfo,
+                                                      &src_brickinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Unable to get src-brickinfo");
+                goto out;
+        }
+
+        ret = glusterd_get_rb_dst_brickinfo (volinfo, &dst_brickinfo);
+        if (!dst_brickinfo) {
+                gf_log ("", GF_LOG_DEBUG, "Unable to get dst-brickinfo");
+                goto out;
+        }
+
+        ret = glusterd_resolve_brick (dst_brickinfo);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Unable to resolve dst-brickinfo");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "src-brick-port", &src_port);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get src-brick port");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "dst-brick-port", &dst_port);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get dst-brick port");
+        }
+
+        dst_brickinfo->port = dst_port;
+        src_brickinfo->port = src_port;
+
+        switch (op) {
+        case GF_REPLACE_OP_START:
+                if (!dst_port) {
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = rb_do_operation (volinfo, src_brickinfo, dst_brickinfo,
+                                       GF_REPLACE_OP_START);
+                if (ret)
+                        goto out;
+                break;
+        case GF_REPLACE_OP_PAUSE:
+        case GF_REPLACE_OP_ABORT:
+        case GF_REPLACE_OP_COMMIT:
+        case GF_REPLACE_OP_COMMIT_FORCE:
+        case GF_REPLACE_OP_STATUS:
+                break;
+        default:
+                ret = -1;
+                goto out;
+        }
+
+out:
+        if (ret)
+                ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT,
+                                                   txn_id, NULL);
+        else
+                ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC,
+                                                   txn_id, NULL);
+
+        glusterd_op_sm ();
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c b/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c
new file mode 100644
index 000000000..d5200a4ae
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c
@@ -0,0 +1,1974 @@
+/*
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "rpc-clnt.h"
+#include "glusterd1-xdr.h"
+#include "cli1-xdr.h"
+
+#include "xdr-generic.h"
+
+#include "compat-errno.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-sm.h"
+#include "glusterd.h"
+#include "protocol-common.h"
+#include "glusterd-utils.h"
+#include "common-utils.h"
+#include <sys/uio.h>
+
+
+#define SERVER_PATH_MAX  (16 * 1024)
+
+
+extern glusterd_op_info_t opinfo;
+extern uuid_t global_txn_id;
+
+int32_t
+glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret,
+                               int32_t op_errno, rpcsvc_request_t *req,
+                               void *op_ctx, char *op_errstr)
+{
+        int32_t         ret = -1;
+        void            *cli_rsp = NULL;
+        dict_t          *ctx = NULL;
+        char            *free_ptr = NULL;
+        glusterd_conf_t *conf = NULL;
+        xdrproc_t       xdrproc = NULL;
+        char            *errstr = NULL;
+        int32_t         status = 0;
+        int32_t         count = 0;
+        gf_cli_rsp      rsp = {0,};
+        xlator_t        *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+
+        GF_ASSERT (conf);
+
+        ctx = op_ctx;
+
+        switch (op) {
+        case GD_OP_REMOVE_BRICK:
+        {
+                if (ctx)
+                        ret = dict_get_str (ctx, "errstr", &errstr);
+                break;
+        }
+        case GD_OP_RESET_VOLUME:
+        {
+                if (op_ret && !op_errstr)
+                        errstr = "Error while resetting options";
+                break;
+        }
+        case GD_OP_REBALANCE:
+        case GD_OP_DEFRAG_BRICK_VOLUME:
+        {
+                if (ctx) {
+                        ret = dict_get_int32 (ctx, "status", &status);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_TRACE,
+                                        "failed to get status");
+                        }
+                }
+                break;
+        }
+        case GD_OP_GSYNC_CREATE:
+        case GD_OP_GSYNC_SET:
+        {
+               if (ctx) {
+                        ret = dict_get_str (ctx, "errstr", &errstr);
+                        ret = dict_set_str (ctx, "glusterd_workdir", conf->workdir);
+                        /* swallow error here, that will be re-triggered in cli */
+
+               }
+               break;
+
+        }
+        case GD_OP_QUOTA:
+        {
+                if (ctx && !op_errstr) {
+                        ret = dict_get_str (ctx, "errstr", &errstr);
+                }
+                break;
+        }
+        case GD_OP_PROFILE_VOLUME:
+        {
+                if (ctx && dict_get_int32 (ctx, "count", &count)) {
+                        ret = dict_set_int32 (ctx, "count", 0);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to set count in dictionary");
+                        }
+                }
+                break;
+        }
+        case GD_OP_START_BRICK:
+        case GD_OP_STOP_BRICK:
+        {
+                gf_log (this->name, GF_LOG_DEBUG, "op '%s' not supported",
+                        gd_op_list[op]);
+                break;
+        }
+        case GD_OP_NONE:
+        case GD_OP_MAX:
+        {
+                gf_log (this->name, GF_LOG_ERROR, "invalid operation");
+                break;
+        }
+        case GD_OP_CREATE_VOLUME:
+        case GD_OP_START_VOLUME:
+        case GD_OP_STOP_VOLUME:
+        case GD_OP_DELETE_VOLUME:
+        case GD_OP_DEFRAG_VOLUME:
+        case GD_OP_ADD_BRICK:
+        case GD_OP_LOG_ROTATE:
+        case GD_OP_SYNC_VOLUME:
+        case GD_OP_STATEDUMP_VOLUME:
+        case GD_OP_REPLACE_BRICK:
+        case GD_OP_STATUS_VOLUME:
+        case GD_OP_SET_VOLUME:
+        case GD_OP_LIST_VOLUME:
+        case GD_OP_CLEARLOCKS_VOLUME:
+        case GD_OP_HEAL_VOLUME:
+        case GD_OP_SNAP:
+        {
+                /*nothing specific to be done*/
+                break;
+        }
+        case GD_OP_COPY_FILE:
+        {
+               if (ctx)
+                        ret = dict_get_str (ctx, "errstr", &errstr);
+               break;
+        }
+        case GD_OP_SYS_EXEC:
+        {
+               if (ctx) {
+                        ret = dict_get_str (ctx, "errstr", &errstr);
+                        ret = dict_set_str (ctx, "glusterd_workdir",
+                                            conf->workdir);
+               }
+               break;
+        }
+        }
+
+        rsp.op_ret = op_ret;
+        rsp.op_errno = errno;
+        if (errstr)
+                rsp.op_errstr = errstr;
+        else if (op_errstr)
+                rsp.op_errstr = op_errstr;
+
+        if (!rsp.op_errstr)
+                rsp.op_errstr = "";
+
+        if (ctx) {
+                ret = dict_allocate_and_serialize (ctx, &rsp.dict.dict_val,
+                                                   &rsp.dict.dict_len);
+                if (ret < 0 )
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                "serialize buffer");
+                else
+                        free_ptr = rsp.dict.dict_val;
+        }
+
+        /* needed by 'rebalance status' */
+        if (status)
+                rsp.op_errno = status;
+
+        cli_rsp = &rsp;
+        xdrproc = (xdrproc_t) xdr_gf_cli_rsp;
+
+        glusterd_to_cli (req, cli_rsp, NULL, 0, NULL,
+                         xdrproc, ctx);
+        ret = 0;
+
+        GF_FREE (free_ptr);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_big_locked_cbk (struct rpc_req *req, struct iovec *iov,
+                         int count, void *myframe, fop_cbk_fn_t fn)
+{
+        glusterd_conf_t *priv = THIS->private;
+        int             ret   = -1;
+
+        synclock_lock (&priv->big_lock);
+        ret = fn (req, iov, count, myframe);
+        synclock_unlock (&priv->big_lock);
+
+        return ret;
+}
+
+int
+__glusterd_probe_cbk (struct rpc_req *req, struct iovec *iov,
+                        int count, void *myframe)
+{
+        gd1_mgmt_probe_rsp    rsp   = {{0},};
+        int                   ret   = 0;
+        glusterd_peerinfo_t           *peerinfo = NULL;
+        glusterd_friend_sm_event_t    *event = NULL;
+        glusterd_probe_ctx_t          *ctx = NULL;
+
+        if (-1 == req->rpc_status) {
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_probe_rsp);
+        if (ret < 0) {
+                gf_log ("", GF_LOG_ERROR, "error");
+                //rsp.op_ret   = -1;
+                //rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+        gf_log ("glusterd", GF_LOG_INFO,
+                "Received probe resp from uuid: %s, host: %s",
+                uuid_utoa (rsp.uuid), rsp.hostname);
+        if (rsp.op_ret != 0) {
+                ctx = ((call_frame_t *)myframe)->local;
+                ((call_frame_t *)myframe)->local = NULL;
+
+                GF_ASSERT (ctx);
+
+                if (ctx->req) {
+                        glusterd_xfer_cli_probe_resp (ctx->req, rsp.op_ret,
+                                                      rsp.op_errno,
+                                                      rsp.op_errstr,
+                                                      ctx->hostname, ctx->port,
+                                                      ctx->dict);
+                }
+
+                glusterd_destroy_probe_ctx (ctx);
+                (void) glusterd_friend_remove (rsp.uuid, rsp.hostname);
+                ret = rsp.op_ret;
+                goto out;
+        }
+        ret = glusterd_friend_find (rsp.uuid, rsp.hostname, &peerinfo);
+        if (ret) {
+                GF_ASSERT (0);
+        }
+
+        if (strncasecmp (rsp.hostname, peerinfo->hostname, 1024)) {
+                gf_log (THIS->name, GF_LOG_INFO, "Host: %s  with uuid: %s "
+                        "already present in cluster with alias hostname: %s",
+                        rsp.hostname, uuid_utoa (rsp.uuid), peerinfo->hostname);
+
+                ctx = ((call_frame_t *)myframe)->local;
+                ((call_frame_t *)myframe)->local = NULL;
+
+                GF_ASSERT (ctx);
+
+                rsp.op_errno = GF_PROBE_FRIEND;
+                if (ctx->req) {
+                        glusterd_xfer_cli_probe_resp (ctx->req, rsp.op_ret,
+                                                      rsp.op_errno,
+                                                      rsp.op_errstr,
+                                                      ctx->hostname, ctx->port,
+                                                      ctx->dict);
+                }
+
+                glusterd_destroy_probe_ctx (ctx);
+                (void) glusterd_friend_remove (NULL, rsp.hostname);
+                ret = rsp.op_ret;
+                goto out;
+        }
+
+        uuid_copy (peerinfo->uuid, rsp.uuid);
+
+        ret = glusterd_friend_sm_new_event
+                        (GD_FRIEND_EVENT_INIT_FRIEND_REQ, &event);
+
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR,
+                         "Unable to get event");
+                goto out;
+        }
+
+        event->peerinfo = peerinfo;
+        event->ctx      = ((call_frame_t *)myframe)->local;
+        ((call_frame_t *)myframe)->local = NULL;
+        ret = glusterd_friend_sm_inject_event (event);
+
+
+        if (!ret) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+        }
+
+        gf_log ("glusterd", GF_LOG_INFO, "Received resp to probe req");
+
+out:
+        free (rsp.hostname);//malloced by xdr
+        GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe));
+        return ret;
+}
+
+int
+glusterd_probe_cbk (struct rpc_req *req, struct iovec *iov,
+                    int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_probe_cbk);
+}
+
+
+int
+__glusterd_friend_add_cbk (struct rpc_req * req, struct iovec *iov,
+                            int count, void *myframe)
+{
+        gd1_mgmt_friend_rsp           rsp   = {{0},};
+        int                           ret   = -1;
+        glusterd_friend_sm_event_t        *event = NULL;
+        glusterd_friend_sm_event_type_t    event_type = GD_FRIEND_EVENT_NONE;
+        glusterd_peerinfo_t           *peerinfo = NULL;
+        int32_t                       op_ret = -1;
+        int32_t                       op_errno = -1;
+        glusterd_probe_ctx_t          *ctx = NULL;
+        glusterd_friend_update_ctx_t  *ev_ctx = NULL;
+
+        if (-1 == req->rpc_status) {
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_friend_rsp);
+        if (ret < 0) {
+                gf_log ("", GF_LOG_ERROR, "error");
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+
+        gf_log ("glusterd", GF_LOG_INFO,
+                "Received %s from uuid: %s, host: %s, port: %d",
+                (op_ret)?"RJT":"ACC", uuid_utoa (rsp.uuid), rsp.hostname, rsp.port);
+
+        ret = glusterd_friend_find (rsp.uuid, rsp.hostname, &peerinfo);
+
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "received friend add response from"
+                        " unknown peer uuid: %s", uuid_utoa (rsp.uuid));
+                goto out;
+        }
+
+        if (op_ret)
+                event_type = GD_FRIEND_EVENT_RCVD_RJT;
+        else
+                event_type = GD_FRIEND_EVENT_RCVD_ACC;
+
+        ret = glusterd_friend_sm_new_event (event_type, &event);
+
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR,
+                         "Unable to get event");
+                goto out;
+        }
+        event->peerinfo = peerinfo;
+        ev_ctx = GF_CALLOC (1, sizeof (*ev_ctx),
+                                gf_gld_mt_friend_update_ctx_t);
+        if (!ev_ctx) {
+                ret = -1;
+                goto out;
+        }
+
+        uuid_copy (ev_ctx->uuid, rsp.uuid);
+        ev_ctx->hostname = gf_strdup (rsp.hostname);
+
+        event->ctx = ev_ctx;
+        ret = glusterd_friend_sm_inject_event (event);
+
+        if (ret)
+                goto out;
+
+out:
+        ctx = ((call_frame_t *)myframe)->local;
+        ((call_frame_t *)myframe)->local = NULL;
+
+        GF_ASSERT (ctx);
+
+        if (ctx->req)//reverse probe doesn't have req
+                ret = glusterd_xfer_cli_probe_resp (ctx->req, op_ret, op_errno,
+                                                    NULL, ctx->hostname,
+                                                    ctx->port, ctx->dict);
+        if (!ret) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+        }
+        if (ctx)
+                glusterd_destroy_probe_ctx (ctx);
+        free (rsp.hostname);//malloced by xdr
+        GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe));
+        return ret;
+}
+
+int
+glusterd_friend_add_cbk (struct rpc_req *req, struct iovec *iov,
+                    int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_friend_add_cbk);
+}
+
+int
+__glusterd_friend_remove_cbk (struct rpc_req * req, struct iovec *iov,
+                               int count, void *myframe)
+{
+        gd1_mgmt_friend_rsp             rsp   = {{0},};
+        glusterd_conf_t                 *conf = NULL;
+        int                             ret   = -1;
+        glusterd_friend_sm_event_t      *event = NULL;
+        glusterd_friend_sm_event_type_t event_type = GD_FRIEND_EVENT_NONE;
+        glusterd_peerinfo_t             *peerinfo = NULL;
+        int32_t                         op_ret = -1;
+        int32_t                         op_errno = -1;
+        glusterd_probe_ctx_t            *ctx = NULL;
+        gf_boolean_t                    move_sm_now = _gf_true;
+
+        conf  = THIS->private;
+        GF_ASSERT (conf);
+
+        ctx = ((call_frame_t *)myframe)->local;
+        ((call_frame_t *)myframe)->local = NULL;
+        GF_ASSERT (ctx);
+
+        if (-1 == req->rpc_status) {
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                move_sm_now = _gf_false;
+                goto inject;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_friend_rsp);
+        if (ret < 0) {
+                gf_log ("", GF_LOG_ERROR, "error");
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto respond;
+        }
+
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+
+        gf_log ("glusterd", GF_LOG_INFO,
+                "Received %s from uuid: %s, host: %s, port: %d",
+                (op_ret)?"RJT":"ACC", uuid_utoa (rsp.uuid), rsp.hostname, rsp.port);
+
+inject:
+        ret = glusterd_friend_find (rsp.uuid, ctx->hostname, &peerinfo);
+
+        if (ret) {
+                //can happen as part of rpc clnt connection cleanup
+                //when the frame timeout happens after 30 minutes
+                goto respond;
+        }
+
+        event_type = GD_FRIEND_EVENT_REMOVE_FRIEND;
+
+        ret = glusterd_friend_sm_new_event (event_type, &event);
+
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR,
+                         "Unable to get event");
+                goto respond;
+        }
+        event->peerinfo = peerinfo;
+
+        ret = glusterd_friend_sm_inject_event (event);
+
+        if (ret)
+                goto respond;
+
+        /*friend_sm would be moved on CLNT_DISCONNECT, consequently
+          cleaning up peerinfo. Else, we run the risk of triggering
+          a clnt_destroy within saved_frames_unwind.
+        */
+        op_ret = 0;
+
+
+respond:
+        ret = glusterd_xfer_cli_deprobe_resp (ctx->req, op_ret, op_errno, NULL,
+                                              ctx->hostname, ctx->dict);
+        if (!ret && move_sm_now) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+        }
+
+        if (ctx) {
+                glusterd_broadcast_friend_delete (ctx->hostname, NULL);
+                glusterd_destroy_probe_ctx (ctx);
+        }
+
+        free (rsp.hostname);//malloced by xdr
+        GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe));
+        return ret;
+}
+
+int
+glusterd_friend_remove_cbk (struct rpc_req *req, struct iovec *iov,
+                    int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_friend_remove_cbk);
+}
+
+int32_t
+__glusterd_friend_update_cbk (struct rpc_req *req, struct iovec *iov,
+                              int count, void *myframe)
+{
+        int                           ret    = -1;
+        gd1_mgmt_friend_update_rsp    rsp    = {{0}, };
+        xlator_t                      *this  = NULL;
+
+        GF_ASSERT (req);
+        this = THIS;
+
+        if (-1 == req->rpc_status) {
+                gf_log (this->name, GF_LOG_ERROR, "RPC Error");
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_friend_update_rsp);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to serialize friend"
+                        " update repsonse");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_INFO, "Received %s from uuid: %s",
+                (ret)?"RJT":"ACC", uuid_utoa (rsp.uuid));
+
+        GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe));
+        return ret;
+}
+
+int
+glusterd_friend_update_cbk (struct rpc_req *req, struct iovec *iov,
+                    int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_friend_update_cbk);
+}
+
+int32_t
+__glusterd_cluster_lock_cbk (struct rpc_req *req, struct iovec *iov,
+                              int count, void *myframe)
+{
+        gd1_mgmt_cluster_lock_rsp     rsp   = {{0},};
+        int                           ret   = -1;
+        int32_t                       op_ret = -1;
+        glusterd_op_sm_event_type_t   event_type = GD_OP_EVENT_NONE;
+        glusterd_peerinfo_t           *peerinfo = NULL;
+        xlator_t                      *this = NULL;
+        uuid_t                        *txn_id = &global_txn_id;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        if (-1 == req->rpc_status) {
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_cluster_lock_rsp);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode lock "
+                        "response received from peer");
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+out:
+        op_ret = rsp.op_ret;
+
+        gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG,
+                "Received lock %s from uuid: %s", (op_ret) ? "RJT" : "ACC",
+                uuid_utoa (rsp.uuid));
+
+        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL, "Lock response received "
+                        "from unknown peer: %s", uuid_utoa (rsp.uuid));
+        }
+
+        if (op_ret) {
+                event_type = GD_OP_EVENT_RCVD_RJT;
+                opinfo.op_ret = op_ret;
+                opinfo.op_errstr = gf_strdup ("Another transaction could be in "
+                                              "progress. Please try again after"
+                                              " sometime.");
+        } else {
+                event_type = GD_OP_EVENT_RCVD_ACC;
+        }
+
+        ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL);
+
+        if (!ret) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+        }
+
+        GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe));
+        return ret;
+}
+
+int32_t
+glusterd_cluster_lock_cbk (struct rpc_req *req, struct iovec *iov,
+                              int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_cluster_lock_cbk);
+}
+
+static int32_t
+glusterd_mgmt_v3_lock_peers_cbk_fn (struct rpc_req *req, struct iovec *iov,
+                                    int count, void *myframe)
+{
+        gd1_mgmt_v3_lock_rsp          rsp   = {{0},};
+        int                           ret   = -1;
+        int32_t                       op_ret = -1;
+        glusterd_op_sm_event_type_t   event_type = GD_OP_EVENT_NONE;
+        glusterd_peerinfo_t           *peerinfo = NULL;
+        xlator_t                      *this = NULL;
+        uuid_t                        *txn_id = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        if (-1 == req->rpc_status) {
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to decode mgmt_v3 lock "
+                        "response received from peer");
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+out:
+        op_ret = rsp.op_ret;
+
+        txn_id = &rsp.txn_id;
+
+        gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG,
+                "Received mgmt_v3 lock %s from uuid: %s",
+                (op_ret) ? "RJT" : "ACC", uuid_utoa (rsp.uuid));
+
+        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "mgmt_v3 lock response received "
+                        "from unknown peer: %s", uuid_utoa (rsp.uuid));
+        }
+
+        if (op_ret) {
+                event_type = GD_OP_EVENT_RCVD_RJT;
+                opinfo.op_ret = op_ret;
+                opinfo.op_errstr = gf_strdup ("Another transaction could be in "
+                                              "progress. Please try again after"
+                                              " sometime.");
+        } else {
+                event_type = GD_OP_EVENT_RCVD_ACC;
+        }
+
+        ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL);
+
+        if (!ret) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+        }
+
+        GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe));
+        return ret;
+}
+
+int32_t
+glusterd_mgmt_v3_lock_peers_cbk (struct rpc_req *req, struct iovec *iov,
+                                 int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        glusterd_mgmt_v3_lock_peers_cbk_fn);
+}
+
+static int32_t
+glusterd_mgmt_v3_unlock_peers_cbk_fn (struct rpc_req *req, struct iovec *iov,
+                                      int count, void *myframe)
+{
+        gd1_mgmt_v3_unlock_rsp        rsp   = {{0},};
+        int                           ret   = -1;
+        int32_t                       op_ret = -1;
+        glusterd_op_sm_event_type_t   event_type = GD_OP_EVENT_NONE;
+        glusterd_peerinfo_t           *peerinfo = NULL;
+        xlator_t                      *this = NULL;
+        uuid_t                        *txn_id = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        if (-1 == req->rpc_status) {
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to decode mgmt_v3 unlock "
+                        "response received from peer");
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+out:
+        op_ret = rsp.op_ret;
+
+        txn_id = &rsp.txn_id;
+
+        gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG,
+                "Received mgmt_v3 unlock %s from uuid: %s",
+                (op_ret) ? "RJT" : "ACC",
+                uuid_utoa (rsp.uuid));
+
+        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "mgmt_v3 unlock response received "
+                        "from unknown peer: %s", uuid_utoa (rsp.uuid));
+        }
+
+        if (op_ret) {
+                event_type = GD_OP_EVENT_RCVD_RJT;
+                opinfo.op_ret = op_ret;
+                opinfo.op_errstr = gf_strdup ("Another transaction could be in "
+                                              "progress. Please try again after"
+                                              " sometime.");
+        } else {
+                event_type = GD_OP_EVENT_RCVD_ACC;
+        }
+
+        ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL);
+
+        if (!ret) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+        }
+
+        GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe));
+        return ret;
+}
+
+int32_t
+glusterd_mgmt_v3_unlock_peers_cbk (struct rpc_req *req, struct iovec *iov,
+                                   int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        glusterd_mgmt_v3_unlock_peers_cbk_fn);
+}
+
+int32_t
+__glusterd_cluster_unlock_cbk (struct rpc_req *req, struct iovec *iov,
+                                 int count, void *myframe)
+{
+        gd1_mgmt_cluster_lock_rsp     rsp   = {{0},};
+        int                           ret   = -1;
+        int32_t                       op_ret = -1;
+        glusterd_op_sm_event_type_t   event_type = GD_OP_EVENT_NONE;
+        glusterd_peerinfo_t           *peerinfo = NULL;
+        xlator_t                      *this = NULL;
+        uuid_t                        *txn_id = &global_txn_id;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        if (-1 == req->rpc_status) {
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_rsp);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode unlock "
+                        "response received from peer");
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                goto out;
+        }
+
+out:
+        op_ret = rsp.op_ret;
+
+        gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG,
+                "Received unlock %s from uuid: %s",
+                (op_ret)?"RJT":"ACC", uuid_utoa (rsp.uuid));
+
+        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL, "Unlock response received "
+                        "from unknown peer %s", uuid_utoa (rsp.uuid));
+        }
+
+        if (op_ret) {
+                event_type = GD_OP_EVENT_RCVD_RJT;
+                opinfo.op_ret = op_ret;
+        } else {
+                event_type = GD_OP_EVENT_RCVD_ACC;
+        }
+
+        ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL);
+
+        if (!ret) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+        }
+
+        GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe));
+        return ret;
+}
+
+int32_t
+glusterd_cluster_unlock_cbk (struct rpc_req *req, struct iovec *iov,
+                                 int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_cluster_unlock_cbk);
+}
+
+int32_t
+__glusterd_stage_op_cbk (struct rpc_req *req, struct iovec *iov,
+                          int count, void *myframe)
+{
+        gd1_mgmt_stage_op_rsp         rsp   = {{0},};
+        int                           ret   = -1;
+        int32_t                       op_ret = -1;
+        glusterd_op_sm_event_type_t   event_type = GD_OP_EVENT_NONE;
+        glusterd_peerinfo_t           *peerinfo = NULL;
+        dict_t                        *dict   = NULL;
+        char                          err_str[2048] = {0};
+        char                          *peer_str = NULL;
+        xlator_t                      *this = NULL;
+        uuid_t                        *txn_id = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        if (-1 == req->rpc_status) {
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                /* use standard allocation because to keep uniformity
+                   in freeing it */
+                rsp.op_errstr = strdup ("error");
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_stage_op_rsp);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode stage "
+                        "response received from peer");
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                /* use standard allocation because to keep uniformity
+                   in freeing it */
+                rsp.op_errstr = strdup ("Failed to decode stage response "
+                                        "received from peer.");
+                goto out;
+        }
+
+        if (rsp.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (rsp.dict.dict_val,
+                                        rsp.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize rsp-buffer to dictionary");
+			event_type = GD_OP_EVENT_RCVD_RJT;
+                        goto out;
+                } else {
+                        dict->extra_stdfree = rsp.dict.dict_val;
+                }
+        }
+
+out:
+        op_ret = rsp.op_ret;
+
+        gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG,
+                "Received stage %s from uuid: %s",
+                (op_ret) ? "RJT" : "ACC", uuid_utoa (rsp.uuid));
+
+        ret = dict_get_bin (dict, "transaction_id", (void **)&txn_id);
+
+        gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id));
+
+        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL, "Stage response received "
+                        "from unknown peer: %s", uuid_utoa (rsp.uuid));
+        }
+
+        if (op_ret) {
+                event_type = GD_OP_EVENT_RCVD_RJT;
+                opinfo.op_ret = op_ret;
+                if (strcmp ("", rsp.op_errstr)) {
+                        opinfo.op_errstr = gf_strdup (rsp.op_errstr);
+                } else {
+                        if (peerinfo)
+                                peer_str = peerinfo->hostname;
+                        else
+                                peer_str = uuid_utoa (rsp.uuid);
+                        snprintf (err_str, sizeof (err_str),
+                                  OPERRSTR_STAGE_FAIL, peer_str);
+                        opinfo.op_errstr = gf_strdup (err_str);
+                }
+                if (!opinfo.op_errstr) {
+                        ret = -1;
+                        goto out;
+                }
+        } else {
+                event_type = GD_OP_EVENT_RCVD_ACC;
+        }
+
+        switch (rsp.op) {
+        case GD_OP_REPLACE_BRICK:
+                glusterd_rb_use_rsp_dict (NULL, dict);
+                break;
+        }
+
+        ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL);
+
+        if (!ret) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+        }
+
+        free (rsp.op_errstr); //malloced by xdr
+        if (dict) {
+                if (!dict->extra_stdfree && rsp.dict.dict_val)
+                        free (rsp.dict.dict_val); //malloced by xdr
+                dict_unref (dict);
+        } else {
+                free (rsp.dict.dict_val); //malloced by xdr
+        }
+        GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe));
+        return ret;
+}
+
+int32_t
+glusterd_stage_op_cbk (struct rpc_req *req, struct iovec *iov,
+                          int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_stage_op_cbk);
+}
+
+int32_t
+__glusterd_commit_op_cbk (struct rpc_req *req, struct iovec *iov,
+                          int count, void *myframe)
+{
+        gd1_mgmt_commit_op_rsp         rsp   = {{0},};
+        int                           ret   = -1;
+        int32_t                       op_ret = -1;
+        glusterd_op_sm_event_type_t   event_type = GD_OP_EVENT_NONE;
+        glusterd_peerinfo_t           *peerinfo = NULL;
+        dict_t                        *dict = NULL;
+        char                          err_str[2048] = {0};
+        char                          *peer_str = NULL;
+        xlator_t                      *this = NULL;
+        uuid_t                        *txn_id = NULL;
+
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        if (-1 == req->rpc_status) {
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                /* use standard allocation because to keep uniformity
+                   in freeing it */
+                rsp.op_errstr = strdup ("error");
+		event_type = GD_OP_EVENT_RCVD_RJT;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_commit_op_rsp);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode commit "
+                        "response received from peer");
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                /* use standard allocation because to keep uniformity
+                   in freeing it */
+                rsp.op_errstr = strdup ("Failed to decode commit response "
+                                        "received from peer.");
+		event_type = GD_OP_EVENT_RCVD_RJT;
+                goto out;
+        }
+
+        if (rsp.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (rsp.dict.dict_val,
+                                        rsp.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize rsp-buffer to dictionary");
+			event_type = GD_OP_EVENT_RCVD_RJT;
+                        goto out;
+                } else {
+                        dict->extra_stdfree = rsp.dict.dict_val;
+                }
+        }
+
+        op_ret = rsp.op_ret;
+
+        gf_log (this->name, (op_ret) ? GF_LOG_ERROR : GF_LOG_DEBUG,
+                "Received commit %s from uuid: %s",
+                (op_ret)?"RJT":"ACC", uuid_utoa (rsp.uuid));
+
+        ret = dict_get_bin (dict, "transaction_id", (void **)&txn_id);
+
+        gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id));
+
+        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL, "Commit response for "
+                        "'Volume %s' received from unknown peer: %s",
+                        gd_op_list[opinfo.op], uuid_utoa (rsp.uuid));
+        }
+
+        if (op_ret) {
+                event_type = GD_OP_EVENT_RCVD_RJT;
+                opinfo.op_ret = op_ret;
+                if (strcmp ("", rsp.op_errstr)) {
+                        opinfo.op_errstr = gf_strdup(rsp.op_errstr);
+                } else {
+                        if (peerinfo)
+                                peer_str = peerinfo->hostname;
+                        else
+                                peer_str = uuid_utoa (rsp.uuid);
+                        snprintf (err_str, sizeof (err_str),
+                                  OPERRSTR_COMMIT_FAIL, peer_str);
+                        opinfo.op_errstr = gf_strdup (err_str);
+                }
+                if (!opinfo.op_errstr) {
+                        ret = -1;
+                        goto out;
+                }
+        } else {
+                event_type = GD_OP_EVENT_RCVD_ACC;
+                switch (rsp.op) {
+                case GD_OP_REPLACE_BRICK:
+                        ret = glusterd_rb_use_rsp_dict (NULL, dict);
+                        if (ret)
+                                goto out;
+                break;
+
+                case GD_OP_SYNC_VOLUME:
+                        ret = glusterd_sync_use_rsp_dict (NULL, dict);
+                        if (ret)
+                                goto out;
+                break;
+
+                case GD_OP_PROFILE_VOLUME:
+                        ret = glusterd_profile_volume_use_rsp_dict (NULL, dict);
+                        if (ret)
+                                goto out;
+                break;
+
+                case GD_OP_GSYNC_SET:
+                        ret = glusterd_gsync_use_rsp_dict (NULL, dict, rsp.op_errstr);
+                        if (ret)
+                                goto out;
+                break;
+
+                case GD_OP_STATUS_VOLUME:
+                        ret = glusterd_volume_status_copy_to_op_ctx_dict (NULL, dict);
+                        if (ret)
+                                goto out;
+                break;
+
+                case GD_OP_REBALANCE:
+                case GD_OP_DEFRAG_BRICK_VOLUME:
+                        ret = glusterd_volume_rebalance_use_rsp_dict (NULL, dict);
+                        if (ret)
+                                goto out;
+                break;
+
+                case GD_OP_HEAL_VOLUME:
+                        ret = glusterd_volume_heal_use_rsp_dict (NULL, dict);
+                        if (ret)
+                                goto out;
+
+                break;
+
+                default:
+                break;
+                }
+        }
+
+out:
+        ret = glusterd_op_sm_inject_event (event_type, txn_id, NULL);
+
+        if (!ret) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+        }
+
+        if (dict)
+                dict_unref (dict);
+        free (rsp.op_errstr); //malloced by xdr
+        GLUSTERD_STACK_DESTROY (((call_frame_t *)myframe));
+        return ret;
+}
+
+int32_t
+glusterd_commit_op_cbk (struct rpc_req *req, struct iovec *iov,
+                          int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_commit_op_cbk);
+}
+
+int32_t
+glusterd_rpc_probe (call_frame_t *frame, xlator_t *this,
+                    void *data)
+{
+        gd1_mgmt_probe_req      req = {{0},};
+        int                     ret = 0;
+        int                     port = 0;
+        char                    *hostname = NULL;
+        glusterd_peerinfo_t     *peerinfo = NULL;
+        glusterd_conf_t         *priv = NULL;
+        dict_t                  *dict = NULL;
+
+        if (!frame || !this ||  !data) {
+                ret = -1;
+                goto out;
+        }
+
+        dict = data;
+        priv = this->private;
+
+        GF_ASSERT (priv);
+        ret = dict_get_str (dict, "hostname", &hostname);
+        if (ret)
+                goto out;
+        ret = dict_get_int32 (dict, "port", &port);
+        if (ret)
+                port = GF_DEFAULT_BASE_PORT;
+
+        ret = dict_get_ptr (dict, "peerinfo", VOID (&peerinfo));
+        if (ret)
+                goto out;
+
+        uuid_copy (req.uuid, MY_UUID);
+        req.hostname = gf_strdup (hostname);
+        req.port = port;
+
+        ret = glusterd_submit_request (peerinfo->rpc, &req, frame, peerinfo->peer,
+                                       GLUSTERD_PROBE_QUERY,
+                                       NULL, this, glusterd_probe_cbk,
+                                       (xdrproc_t)xdr_gd1_mgmt_probe_req);
+
+out:
+        GF_FREE (req.hostname);
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+
+int32_t
+glusterd_rpc_friend_add (call_frame_t *frame, xlator_t *this,
+                         void *data)
+{
+        gd1_mgmt_friend_req         req      = {{0},};
+        int                         ret      = 0;
+        glusterd_peerinfo_t        *peerinfo = NULL;
+        glusterd_conf_t            *priv     = NULL;
+        glusterd_friend_sm_event_t *event    = NULL;
+        dict_t                     *vols     = NULL;
+
+
+        if (!frame || !this || !data) {
+                ret = -1;
+                goto out;
+        }
+
+        event = data;
+        priv = this->private;
+
+        GF_ASSERT (priv);
+
+        peerinfo = event->peerinfo;
+
+        ret = glusterd_build_volume_dict (&vols);
+        if (ret)
+                goto out;
+
+        uuid_copy (req.uuid, MY_UUID);
+        req.hostname = peerinfo->hostname;
+        req.port = peerinfo->port;
+
+        ret = dict_allocate_and_serialize (vols, &req.vols.vols_val,
+                                           &req.vols.vols_len);
+        if (ret)
+                goto out;
+
+        ret = glusterd_submit_request (peerinfo->rpc, &req, frame, peerinfo->peer,
+                                       GLUSTERD_FRIEND_ADD,
+                                       NULL, this, glusterd_friend_add_cbk,
+                                       (xdrproc_t)xdr_gd1_mgmt_friend_req);
+
+
+out:
+        GF_FREE (req.vols.vols_val);
+
+        if (vols)
+                dict_unref (vols);
+
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_rpc_friend_remove (call_frame_t *frame, xlator_t *this,
+                            void *data)
+{
+        gd1_mgmt_friend_req             req = {{0},};
+        int                             ret = 0;
+        glusterd_peerinfo_t             *peerinfo = NULL;
+        glusterd_conf_t                 *priv = NULL;
+        glusterd_friend_sm_event_t      *event = NULL;
+
+        if (!frame || !this || !data) {
+                ret = -1;
+                goto out;
+        }
+
+        event = data;
+        priv = this->private;
+
+        GF_ASSERT (priv);
+
+        peerinfo = event->peerinfo;
+
+        uuid_copy (req.uuid, MY_UUID);
+        req.hostname = peerinfo->hostname;
+        req.port = peerinfo->port;
+        ret = glusterd_submit_request (peerinfo->rpc, &req, frame, peerinfo->peer,
+                                       GLUSTERD_FRIEND_REMOVE, NULL,
+                                       this, glusterd_friend_remove_cbk,
+                                       (xdrproc_t)xdr_gd1_mgmt_friend_req);
+
+out:
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+
+int32_t
+glusterd_rpc_friend_update (call_frame_t *frame, xlator_t *this,
+                            void *data)
+{
+        gd1_mgmt_friend_update  req         = {{0},};
+        int                     ret         = 0;
+        glusterd_conf_t        *priv        = NULL;
+        dict_t                 *friends     = NULL;
+        call_frame_t           *dummy_frame = NULL;
+        glusterd_peerinfo_t    *peerinfo    = NULL;
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        friends = data;
+        if (!friends)
+                goto out;
+
+        ret = dict_get_ptr (friends, "peerinfo", VOID(&peerinfo));
+        if (ret)
+                goto out;
+
+        ret = dict_allocate_and_serialize (friends, &req.friends.friends_val,
+                                           &req.friends.friends_len);
+        if (ret)
+                goto out;
+
+        uuid_copy (req.uuid, MY_UUID);
+
+        dummy_frame = create_frame (this, this->ctx->pool);
+        ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame,
+                                       peerinfo->peer,
+                                       GLUSTERD_FRIEND_UPDATE, NULL,
+                                       this, glusterd_friend_update_cbk,
+                                       (xdrproc_t)xdr_gd1_mgmt_friend_update);
+
+out:
+        GF_FREE (req.friends.friends_val);
+
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_cluster_lock (call_frame_t *frame, xlator_t *this,
+                          void *data)
+{
+        gd1_mgmt_cluster_lock_req       req = {{0},};
+        int                             ret = -1;
+        glusterd_peerinfo_t             *peerinfo = NULL;
+        glusterd_conf_t                 *priv = NULL;
+        call_frame_t                    *dummy_frame = NULL;
+
+        if (!this)
+                goto out;
+
+        peerinfo = data;
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        glusterd_get_uuid (&req.uuid);
+
+        dummy_frame = create_frame (this, this->ctx->pool);
+        if (!dummy_frame)
+                goto out;
+
+        ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame,
+                                       peerinfo->mgmt, GLUSTERD_MGMT_CLUSTER_LOCK,
+                                       NULL,
+                                       this, glusterd_cluster_lock_cbk,
+                                       (xdrproc_t)xdr_gd1_mgmt_cluster_lock_req);
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_mgmt_v3_lock_peers (call_frame_t *frame, xlator_t *this,
+                             void *data)
+{
+        gd1_mgmt_v3_lock_req             req         = {{0},};
+        int                              ret         = -1;
+        glusterd_peerinfo_t             *peerinfo    = NULL;
+        glusterd_conf_t                 *priv        = NULL;
+        call_frame_t                    *dummy_frame = NULL;
+        dict_t                          *dict        = NULL;
+        uuid_t                          *txn_id      = NULL;
+
+        if (!this)
+                goto out;
+
+        dict = data;
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_ptr (dict, "peerinfo", VOID (&peerinfo));
+        if (ret)
+                goto out;
+
+        //peerinfo should not be in payload
+        dict_del (dict, "peerinfo");
+
+        glusterd_get_uuid (&req.uuid);
+
+        ret = dict_allocate_and_serialize (dict, &req.dict.dict_val,
+                                           &req.dict.dict_len);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to serialize dict "
+                        "to request buffer");
+                goto out;
+        }
+
+        /* Sending valid transaction ID to peers */
+        ret = dict_get_bin (dict, "transaction_id",
+                            (void **)&txn_id);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                       "Failed to get transaction id.");
+                goto out;
+        } else {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Transaction_id = %s", uuid_utoa (*txn_id));
+                uuid_copy (req.txn_id, *txn_id);
+        }
+
+        dummy_frame = create_frame (this, this->ctx->pool);
+        if (!dummy_frame)
+                goto out;
+
+        ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame,
+                                       peerinfo->mgmt_v3,
+                                       GLUSTERD_MGMT_V3_LOCK, NULL,
+                                       this, glusterd_mgmt_v3_lock_peers_cbk,
+                                       (xdrproc_t)xdr_gd1_mgmt_v3_lock_req);
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_mgmt_v3_unlock_peers (call_frame_t *frame, xlator_t *this,
+                               void *data)
+{
+        gd1_mgmt_v3_unlock_req           req         = {{0},};
+        int                              ret         = -1;
+        glusterd_peerinfo_t             *peerinfo    = NULL;
+        glusterd_conf_t                 *priv        = NULL;
+        call_frame_t                    *dummy_frame = NULL;
+        dict_t                          *dict        = NULL;
+        uuid_t                          *txn_id      = NULL;
+
+        if (!this)
+                goto out;
+
+        dict = data;
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_ptr (dict, "peerinfo", VOID (&peerinfo));
+        if (ret)
+                goto out;
+
+        //peerinfo should not be in payload
+        dict_del (dict, "peerinfo");
+
+        glusterd_get_uuid (&req.uuid);
+
+        ret = dict_allocate_and_serialize (dict, &req.dict.dict_val,
+                                           &req.dict.dict_len);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to serialize dict "
+                        "to request buffer");
+                goto out;
+        }
+
+        /* Sending valid transaction ID to peers */
+        ret = dict_get_bin (dict, "transaction_id",
+                            (void **)&txn_id);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                       "Failed to get transaction id.");
+                goto out;
+        } else {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Transaction_id = %s", uuid_utoa (*txn_id));
+                uuid_copy (req.txn_id, *txn_id);
+        }
+
+        dummy_frame = create_frame (this, this->ctx->pool);
+        if (!dummy_frame)
+                goto out;
+
+        ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame,
+                                       peerinfo->mgmt_v3,
+                                       GLUSTERD_MGMT_V3_UNLOCK, NULL,
+                                       this, glusterd_mgmt_v3_unlock_peers_cbk,
+                                       (xdrproc_t)xdr_gd1_mgmt_v3_unlock_req);
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_cluster_unlock (call_frame_t *frame, xlator_t *this,
+                            void *data)
+{
+        gd1_mgmt_cluster_lock_req       req = {{0},};
+        int                             ret = -1;
+        glusterd_peerinfo_t             *peerinfo = NULL;
+        glusterd_conf_t                 *priv = NULL;
+        call_frame_t                    *dummy_frame = NULL;
+
+        if (!this ) {
+                ret = -1;
+                goto out;
+        }
+        peerinfo = data;
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        glusterd_get_uuid (&req.uuid);
+
+        dummy_frame = create_frame (this, this->ctx->pool);
+        if (!dummy_frame)
+                goto out;
+
+        ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame,
+                                       peerinfo->mgmt, GLUSTERD_MGMT_CLUSTER_UNLOCK,
+                                       NULL,
+                                       this, glusterd_cluster_unlock_cbk,
+                                       (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_req);
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_stage_op (call_frame_t *frame, xlator_t *this,
+                      void *data)
+{
+        gd1_mgmt_stage_op_req           req = {{0,},};
+        int                             ret = -1;
+        glusterd_peerinfo_t             *peerinfo = NULL;
+        glusterd_conf_t                 *priv = NULL;
+        call_frame_t                    *dummy_frame = NULL;
+        dict_t                          *dict = NULL;
+        gf_boolean_t                    is_alloc = _gf_true;
+
+        if (!this) {
+                goto out;
+        }
+
+        dict = data;
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_ptr (dict, "peerinfo", VOID (&peerinfo));
+        if (ret)
+                goto out;
+
+        //peerinfo should not be in payload
+        dict_del (dict, "peerinfo");
+
+        glusterd_get_uuid (&req.uuid);
+        req.op = glusterd_op_get_op ();
+
+        ret = dict_allocate_and_serialize (dict, &req.buf.buf_val,
+                                           &req.buf.buf_len);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to serialize dict "
+                        "to request buffer");
+                goto out;
+        }
+
+
+        dummy_frame = create_frame (this, this->ctx->pool);
+        if (!dummy_frame)
+                goto out;
+
+        ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame,
+                                       peerinfo->mgmt, GLUSTERD_MGMT_STAGE_OP,
+                                       NULL,
+                                       this, glusterd_stage_op_cbk,
+                                       (xdrproc_t)xdr_gd1_mgmt_stage_op_req);
+
+out:
+        if ((_gf_true == is_alloc) && req.buf.buf_val)
+                GF_FREE (req.buf.buf_val);
+
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_commit_op (call_frame_t *frame, xlator_t *this,
+                      void *data)
+{
+        gd1_mgmt_commit_op_req  req         = {{0,},};
+        int                     ret         = -1;
+        glusterd_peerinfo_t    *peerinfo    = NULL;
+        glusterd_conf_t        *priv        = NULL;
+        call_frame_t           *dummy_frame = NULL;
+        dict_t                 *dict        = NULL;
+        gf_boolean_t            is_alloc    = _gf_true;
+
+        if (!this) {
+                goto out;
+        }
+
+        dict = data;
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_ptr (dict, "peerinfo", VOID (&peerinfo));
+        if (ret)
+                goto out;
+
+        //peerinfo should not be in payload
+        dict_del (dict, "peerinfo");
+
+        glusterd_get_uuid (&req.uuid);
+        req.op = glusterd_op_get_op ();
+
+        ret = dict_allocate_and_serialize (dict, &req.buf.buf_val,
+                                           &req.buf.buf_len);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to serialize dict to "
+                        "request buffer");
+                goto out;
+        }
+
+        dummy_frame = create_frame (this, this->ctx->pool);
+        if (!dummy_frame)
+                goto out;
+
+        ret = glusterd_submit_request (peerinfo->rpc, &req, dummy_frame,
+                                       peerinfo->mgmt, GLUSTERD_MGMT_COMMIT_OP,
+                                       NULL,
+                                       this, glusterd_commit_op_cbk,
+                                       (xdrproc_t)xdr_gd1_mgmt_commit_op_req);
+
+out:
+        if ((_gf_true == is_alloc) && req.buf.buf_val)
+                GF_FREE (req.buf.buf_val);
+
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+__glusterd_brick_op_cbk (struct rpc_req *req, struct iovec *iov,
+                          int count, void *myframe)
+{
+        gd1_mgmt_brick_op_rsp         rsp   = {0};
+        int                           ret   = -1;
+        int32_t                       op_ret = -1;
+        glusterd_op_sm_event_type_t   event_type = GD_OP_EVENT_NONE;
+        call_frame_t                  *frame = NULL;
+        glusterd_op_brick_rsp_ctx_t   *ev_ctx = NULL;
+        dict_t                        *dict = NULL;
+        int                           index = 0;
+        glusterd_req_ctx_t            *req_ctx = NULL;
+        glusterd_pending_node_t       *node = NULL;
+        xlator_t                      *this = NULL;
+        uuid_t                        *txn_id = &global_txn_id;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        GF_ASSERT (req);
+        frame = myframe;
+        req_ctx = frame->local;
+
+        if (-1 == req->rpc_status) {
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                /* use standard allocation because to keep uniformity
+                   in freeing it */
+                rsp.op_errstr = strdup ("error");
+		event_type = GD_OP_EVENT_RCVD_RJT;
+                goto out;
+        }
+
+        ret =  xdr_to_generic (*iov, &rsp, (xdrproc_t)xdr_gd1_mgmt_brick_op_rsp);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to decode brick op "
+                        "response received");
+                rsp.op_ret   = -1;
+                rsp.op_errno = EINVAL;
+                rsp.op_errstr = strdup ("Unable to decode brick op response");
+		event_type = GD_OP_EVENT_RCVD_RJT;
+                goto out;
+        }
+
+        if (rsp.output.output_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (rsp.output.output_val,
+                                        rsp.output.output_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                "unserialize rsp-buffer to dictionary");
+			event_type = GD_OP_EVENT_RCVD_RJT;
+                        goto out;
+                } else {
+                        dict->extra_stdfree = rsp.output.output_val;
+                }
+        }
+
+        op_ret = rsp.op_ret;
+
+        /* Add index to rsp_dict for GD_OP_STATUS_VOLUME */
+        if (GD_OP_STATUS_VOLUME == req_ctx->op) {
+                node = frame->cookie;
+                index = node->index;
+                ret = dict_set_int32 (dict, "index", index);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Error setting index on brick status rsp dict");
+                        rsp.op_ret = -1;
+                        event_type = GD_OP_EVENT_RCVD_RJT;
+                        goto out;
+                }
+        }
+out:
+
+        ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id);
+
+        gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id));
+
+        ev_ctx = GF_CALLOC (1, sizeof (*ev_ctx), gf_gld_mt_brick_rsp_ctx_t);
+        GF_ASSERT (ev_ctx);
+        if (op_ret) {
+                event_type = GD_OP_EVENT_RCVD_RJT;
+                ev_ctx->op_ret = op_ret;
+                ev_ctx->op_errstr = gf_strdup(rsp.op_errstr);
+        } else {
+                event_type = GD_OP_EVENT_RCVD_ACC;
+        }
+        ev_ctx->pending_node = frame->cookie;
+        ev_ctx->rsp_dict  = dict;
+        ev_ctx->commit_ctx = frame->local;
+        ret = glusterd_op_sm_inject_event (event_type, txn_id, ev_ctx);
+        if (!ret) {
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+        }
+
+        if (ret && dict)
+                dict_unref (dict);
+        free (rsp.op_errstr); //malloced by xdr
+        GLUSTERD_STACK_DESTROY (frame);
+        return ret;
+}
+
+int32_t
+glusterd_brick_op_cbk (struct rpc_req *req, struct iovec *iov,
+                          int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        __glusterd_brick_op_cbk);
+}
+
+int32_t
+glusterd_brick_op (call_frame_t *frame, xlator_t *this,
+                      void *data)
+{
+
+        gd1_mgmt_brick_op_req           *req = NULL;
+        int                             ret = 0;
+        glusterd_conf_t                 *priv = NULL;
+        call_frame_t                    *dummy_frame = NULL;
+        char                            *op_errstr = NULL;
+        int                             pending_bricks = 0;
+        glusterd_pending_node_t         *pending_node;
+        glusterd_req_ctx_t              *req_ctx = NULL;
+        struct rpc_clnt                 *rpc = NULL;
+        dict_t                          *op_ctx = NULL;
+        uuid_t                          *txn_id = &global_txn_id;
+
+        if (!this) {
+                ret = -1;
+                goto out;
+        }
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        req_ctx = data;
+        GF_ASSERT (req_ctx);
+        INIT_LIST_HEAD (&opinfo.pending_bricks);
+        ret = glusterd_op_bricks_select (req_ctx->op, req_ctx->dict, &op_errstr,
+                                         &opinfo.pending_bricks, NULL);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to select bricks "
+                        "while performing brick op during 'Volume %s'",
+                        gd_op_list[opinfo.op]);
+                opinfo.op_errstr = op_errstr;
+                goto out;
+        }
+
+
+        ret = dict_get_bin (req_ctx->dict, "transaction_id", (void **)&txn_id);
+
+        gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", uuid_utoa (*txn_id));
+
+        list_for_each_entry (pending_node, &opinfo.pending_bricks, list) {
+                dummy_frame = create_frame (this, this->ctx->pool);
+                if (!dummy_frame)
+                        continue;
+
+                if ((pending_node->type == GD_NODE_NFS) ||
+                    ((pending_node->type == GD_NODE_SHD) &&
+                    (req_ctx->op == GD_OP_STATUS_VOLUME)))
+                        ret = glusterd_node_op_build_payload
+                                (req_ctx->op,
+                                 (gd1_mgmt_brick_op_req **)&req,
+                                 req_ctx->dict);
+                else {
+                        ret = glusterd_brick_op_build_payload
+                                (req_ctx->op, pending_node->node,
+                                 (gd1_mgmt_brick_op_req **)&req,
+                                 req_ctx->dict);
+
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                        "build brick op payload during "
+                                        "'Volume %s'", gd_op_list[req_ctx->op]);
+                                goto out;
+                        }
+                }
+
+                dummy_frame->local = data;
+                dummy_frame->cookie = pending_node;
+
+                rpc = glusterd_pending_node_get_rpc (pending_node);
+                if (!rpc) {
+                        if (pending_node->type == GD_NODE_REBALANCE) {
+                                opinfo.brick_pending_count = 0;
+                                ret = 0;
+                                if (req) {
+                                        GF_FREE (req->input.input_val);
+                                        GF_FREE (req);
+                                        req = NULL;
+                                }
+                                GLUSTERD_STACK_DESTROY (dummy_frame);
+
+                                op_ctx = glusterd_op_get_ctx ();
+                                if (!op_ctx)
+                                        goto out;
+                                glusterd_defrag_volume_node_rsp (req_ctx->dict,
+                                                                 NULL, op_ctx);
+
+                                goto out;
+                        }
+
+                        ret = -1;
+                        gf_log (this->name, GF_LOG_ERROR, "Brick Op failed "
+                                "due to rpc failure.");
+                        goto out;
+                }
+
+                ret = glusterd_submit_request (rpc, req, dummy_frame,
+                                               priv->gfs_mgmt,
+                                               req->op, NULL,
+                                               this, glusterd_brick_op_cbk,
+                                               (xdrproc_t)xdr_gd1_mgmt_brick_op_req);
+                if (req) {
+                        GF_FREE (req->input.input_val);
+                        GF_FREE (req);
+                        req = NULL;
+                }
+                if (!ret)
+                        pending_bricks++;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "Sent brick op req for operation "
+                "'Volume %s' to %d bricks", gd_op_list[req_ctx->op],
+                pending_bricks);
+        opinfo.brick_pending_count = pending_bricks;
+
+out:
+        if (ret) {
+                glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT,
+                                             txn_id, data);
+                opinfo.op_ret = ret;
+        }
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+struct rpc_clnt_procedure gd_brick_actors[GLUSTERD_BRICK_MAXVALUE] = {
+        [GLUSTERD_BRICK_NULL] = {"NULL", NULL },
+        [GLUSTERD_BRICK_OP]   = {"BRICK_OP", glusterd_brick_op},
+};
+
+struct rpc_clnt_procedure gd_peer_actors[GLUSTERD_FRIEND_MAXVALUE] = {
+        [GLUSTERD_FRIEND_NULL]   = {"NULL", NULL },
+        [GLUSTERD_PROBE_QUERY]   = {"PROBE_QUERY", glusterd_rpc_probe},
+        [GLUSTERD_FRIEND_ADD]    = {"FRIEND_ADD", glusterd_rpc_friend_add},
+        [GLUSTERD_FRIEND_REMOVE] = {"FRIEND_REMOVE", glusterd_rpc_friend_remove},
+        [GLUSTERD_FRIEND_UPDATE] = {"FRIEND_UPDATE", glusterd_rpc_friend_update},
+};
+
+struct rpc_clnt_procedure gd_mgmt_actors[GLUSTERD_MGMT_MAXVALUE] = {
+        [GLUSTERD_MGMT_NULL]           = {"NULL", NULL },
+        [GLUSTERD_MGMT_CLUSTER_LOCK]   = {"CLUSTER_LOCK", glusterd_cluster_lock},
+        [GLUSTERD_MGMT_CLUSTER_UNLOCK] = {"CLUSTER_UNLOCK", glusterd_cluster_unlock},
+        [GLUSTERD_MGMT_STAGE_OP]       = {"STAGE_OP", glusterd_stage_op},
+        [GLUSTERD_MGMT_COMMIT_OP]      = {"COMMIT_OP", glusterd_commit_op},
+};
+
+struct rpc_clnt_procedure gd_mgmt_v3_actors[GLUSTERD_MGMT_V3_MAXVALUE] = {
+        [GLUSTERD_MGMT_V3_NULL]           = {"NULL", NULL },
+        [GLUSTERD_MGMT_V3_LOCK]           = {"MGMT_V3_LOCK", glusterd_mgmt_v3_lock_peers},
+        [GLUSTERD_MGMT_V3_UNLOCK]         = {"MGMT_V3_UNLOCK", glusterd_mgmt_v3_unlock_peers},
+};
+
+struct rpc_clnt_program gd_mgmt_prog = {
+        .progname  = "glusterd mgmt",
+        .prognum   = GD_MGMT_PROGRAM,
+        .progver   = GD_MGMT_VERSION,
+        .proctable = gd_mgmt_actors,
+        .numproc   = GLUSTERD_MGMT_MAXVALUE,
+};
+
+struct rpc_clnt_program gd_brick_prog = {
+        .progname  = "brick operations",
+        .prognum   = GD_BRICK_PROGRAM,
+        .progver   = GD_BRICK_VERSION,
+        .proctable = gd_brick_actors,
+        .numproc   = GLUSTERD_BRICK_MAXVALUE,
+};
+
+struct rpc_clnt_program gd_peer_prog = {
+        .progname  = "Peer mgmt",
+        .prognum   = GD_FRIEND_PROGRAM,
+        .progver   = GD_FRIEND_VERSION,
+        .proctable = gd_peer_actors,
+        .numproc   = GLUSTERD_FRIEND_MAXVALUE,
+};
+
+struct rpc_clnt_program gd_mgmt_v3_prog = {
+        .progname  = "glusterd mgmt v3",
+        .prognum   = GD_MGMT_V3_PROGRAM,
+        .progver   = GD_MGMT_V3_VERSION,
+        .proctable = gd_mgmt_v3_actors,
+        .numproc   = GLUSTERD_MGMT_V3_MAXVALUE,
+};
diff --git a/xlators/mgmt/glusterd/src/glusterd-sm.c b/xlators/mgmt/glusterd/src/glusterd-sm.c
index c5e1800cd..c671edf68 100644
--- a/xlators/mgmt/glusterd/src/glusterd-sm.c
+++ b/xlators/mgmt/glusterd/src/glusterd-sm.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is GF_FREE software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
@@ -41,19 +31,66 @@
 #include "compat-errno.h"
 #include "statedump.h"
 #include "glusterd-sm.h"
+#include "glusterd-op-sm.h"
 #include "glusterd-utils.h"
 #include "glusterd-store.h"
 
 static struct list_head gd_friend_sm_queue;
 
+static  char *glusterd_friend_sm_state_names[] = {
+        "Establishing Connection",
+        "Probe Sent to Peer",
+        "Probe Received from Peer",
+        "Peer in Cluster",
+        "Accepted peer request",
+        "Sent and Received peer request",
+        "Peer Rejected",
+        "Peer detach in progress",
+        "Probe Received from peer",
+        "Connected to Peer",
+        "Peer is connected and Accepted",
+        "Invalid State"
+};
+
+static char *glusterd_friend_sm_event_names[] = {
+        "GD_FRIEND_EVENT_NONE",
+        "GD_FRIEND_EVENT_PROBE",
+        "GD_FRIEND_EVENT_INIT_FRIEND_REQ",
+        "GD_FRIEND_EVENT_RCVD_ACC",
+        "GD_FRIEND_EVENT_LOCAL_ACC",
+        "GD_FRIEND_EVENT_RCVD_RJT",
+        "GD_FRIEND_EVENT_LOCAL_RJT",
+        "GD_FRIEND_EVENT_RCVD_FRIEND_REQ",
+        "GD_FRIEND_EVENT_INIT_REMOVE_FRIEND",
+        "GD_FRIEND_EVENT_RCVD_REMOVE_FRIEND",
+        "GD_FRIEND_EVENT_REMOVE_FRIEND",
+        "GD_FRIEND_EVENT_CONNECTED",
+        "GD_FRIEND_EVENT_MAX"
+};
+
+char*
+glusterd_friend_sm_state_name_get (int state)
+{
+        if (state < 0 || state >= GD_FRIEND_STATE_MAX)
+                return glusterd_friend_sm_state_names[GD_FRIEND_STATE_MAX];
+        return glusterd_friend_sm_state_names[state];
+}
+
+char*
+glusterd_friend_sm_event_name_get (int event)
+{
+        if (event < 0 || event >= GD_FRIEND_EVENT_MAX)
+                return glusterd_friend_sm_event_names[GD_FRIEND_EVENT_MAX];
+        return glusterd_friend_sm_event_names[event];
+}
+
 void
 glusterd_destroy_probe_ctx (glusterd_probe_ctx_t *ctx)
 {
         if (!ctx)
                 return;
 
-        if (ctx->hostname)
-                GF_FREE (ctx->hostname);
+        GF_FREE (ctx->hostname);
         GF_FREE (ctx);
 }
 
@@ -65,15 +102,83 @@ glusterd_destroy_friend_req_ctx (glusterd_friend_req_ctx_t *ctx)
 
         if (ctx->vols)
                 dict_unref (ctx->vols);
-        if (ctx->hostname)
-                GF_FREE (ctx->hostname);
-        if (ctx->remote_hostname)
-                GF_FREE (ctx->remote_hostname);
+        GF_FREE (ctx->hostname);
+        GF_FREE (ctx);
+}
+
+void
+glusterd_destroy_friend_update_ctx (glusterd_friend_update_ctx_t *ctx)
+{
+        if (!ctx)
+                return;
+        GF_FREE (ctx->hostname);
         GF_FREE (ctx);
 }
 
-#define glusterd_destroy_friend_update_ctx(ctx)\
-        glusterd_destroy_friend_req_ctx(ctx)
+int
+glusterd_broadcast_friend_delete (char *hostname, uuid_t uuid)
+{
+        int                             ret = 0;
+        rpc_clnt_procedure_t            *proc = NULL;
+        xlator_t                        *this = NULL;
+        glusterd_friend_update_ctx_t    ctx = {{0},};
+        glusterd_peerinfo_t             *peerinfo = NULL;
+        glusterd_conf_t                 *priv = NULL;
+        dict_t                          *friends = NULL;
+        char                            key[100] = {0,};
+        int32_t                         count = 0;
+
+        this = THIS;
+        priv = this->private;
+
+        GF_ASSERT (priv);
+
+        ctx.hostname = hostname;
+        ctx.op = GD_FRIEND_UPDATE_DEL;
+
+        friends = dict_new ();
+        if (!friends)
+                goto out;
+
+        snprintf (key, sizeof (key), "op");
+        ret = dict_set_int32 (friends, key, ctx.op);
+        if (ret)
+                goto out;
+
+        snprintf (key, sizeof (key), "hostname");
+        ret = dict_set_str (friends, key, hostname);
+        if (ret)
+                goto out;
+
+        ret = dict_set_int32 (friends, "count", count);
+        if (ret)
+                goto out;
+
+        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
+                if (!peerinfo->connected || !peerinfo->peer)
+                        continue;
+
+                ret = dict_set_static_ptr (friends, "peerinfo", peerinfo);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "failed to set peerinfo");
+                        goto out;
+                }
+
+                proc = &peerinfo->peer->proctable[GLUSTERD_FRIEND_UPDATE];
+                if (proc->fn) {
+                        ret = proc->fn (NULL, this, friends);
+                }
+        }
+
+        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+
+out:
+        if (friends)
+                dict_unref (friends);
+
+        return ret;
+}
+
 
 static int
 glusterd_ac_none (glusterd_friend_sm_event_t *event, void *ctx)
@@ -99,14 +204,50 @@ static int
 glusterd_ac_reverse_probe_begin (glusterd_friend_sm_event_t *event, void *ctx)
 {
         int ret = 0;
-        glusterd_friend_update_ctx_t    *ev_ctx = NULL;
+        glusterd_peerinfo_t             *peerinfo = NULL;
+        glusterd_friend_sm_event_t      *new_event = NULL;
+        glusterd_probe_ctx_t            *new_ev_ctx = NULL;
 
         GF_ASSERT (event);
         GF_ASSERT (ctx);
 
-        ev_ctx = ctx;
-        ret = glusterd_probe_begin (NULL, ev_ctx->remote_hostname, 0);
+        peerinfo = event->peerinfo;
+        ret = glusterd_friend_sm_new_event
+                (GD_FRIEND_EVENT_PROBE, &new_event);
 
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Unable to get new new_event");
+                ret = -1;
+                goto out;
+        }
+
+        new_ev_ctx = GF_CALLOC (1, sizeof(*new_ev_ctx), gf_gld_mt_probe_ctx_t);
+
+        if (!new_ev_ctx) {
+                ret = -1;
+                goto out;
+        }
+
+        new_ev_ctx->hostname = gf_strdup (peerinfo->hostname);
+        new_ev_ctx->port = peerinfo->port;
+        new_ev_ctx->req = NULL;
+
+        new_event->peerinfo = peerinfo;
+        new_event->ctx = new_ev_ctx;
+
+        ret = glusterd_friend_sm_inject_event (new_event);
+
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Unable to inject new_event %d, "
+                                "ret = %d", new_event->event, ret);
+        }
+
+out:
+        if (ret) {
+                GF_FREE (new_event);
+                GF_FREE (new_ev_ctx->hostname);
+                GF_FREE (new_ev_ctx);
+        }
         gf_log ("", GF_LOG_DEBUG, "returning with %d", ret);
         return ret;
 }
@@ -128,9 +269,10 @@ glusterd_ac_friend_add (glusterd_friend_sm_event_t *event, void *ctx)
         conf = this->private;
 
         GF_ASSERT (conf);
-        GF_ASSERT (conf->mgmt);
 
-        proc = &conf->mgmt->proctable[GD_MGMT_FRIEND_ADD];
+        if (!peerinfo->peer)
+                goto out;
+        proc = &peerinfo->peer->proctable[GLUSTERD_FRIEND_ADD];
         if (proc->fn) {
                 frame = create_frame (this, this->ctx->pool);
                 if (!frame) {
@@ -155,6 +297,7 @@ glusterd_ac_friend_probe (glusterd_friend_sm_event_t *event, void *ctx)
         glusterd_conf_t         *conf = NULL;
         xlator_t                *this = NULL;
         glusterd_probe_ctx_t    *probe_ctx = NULL;
+        glusterd_peerinfo_t     *peerinfo = NULL;
         dict_t                  *dict = NULL;
 
         GF_ASSERT (ctx);
@@ -168,11 +311,17 @@ glusterd_ac_friend_probe (glusterd_friend_sm_event_t *event, void *ctx)
         conf = this->private;
 
         GF_ASSERT (conf);
-        if (!conf->mgmt)
-                goto out;
 
+        ret = glusterd_friend_find (NULL, probe_ctx->hostname, &peerinfo);
+        if (ret) {
+                //We should not reach this state ideally
+                GF_ASSERT (0);
+                goto out;
+        }
 
-        proc = &conf->mgmt->proctable[GD_MGMT_PROBE_QUERY];
+        if (!peerinfo->peer)
+                goto out;
+        proc = &peerinfo->peer->proctable[GLUSTERD_PROBE_QUERY];
         if (proc->fn) {
                 frame = create_frame (this, this->ctx->pool);
                 if (!frame) {
@@ -189,6 +338,13 @@ glusterd_ac_friend_probe (glusterd_friend_sm_event_t *event, void *ctx)
                 ret = dict_set_int32 (dict, "port", probe_ctx->port);
                 if (ret)
                         goto out;
+
+                ret = dict_set_static_ptr (dict, "peerinfo", peerinfo);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "failed to set peerinfo");
+                        goto out;
+                }
+
                 ret = proc->fn (frame, this, dict);
                 if (ret)
                         goto out;
@@ -205,7 +361,8 @@ out:
 }
 
 static int
-glusterd_ac_send_friend_remove_req (glusterd_friend_sm_event_t *event, void *ctx)
+glusterd_ac_send_friend_remove_req (glusterd_friend_sm_event_t *event,
+                                    void *data)
 {
         int                     ret = 0;
         glusterd_peerinfo_t     *peerinfo = NULL;
@@ -213,7 +370,9 @@ glusterd_ac_send_friend_remove_req (glusterd_friend_sm_event_t *event, void *ctx
         call_frame_t            *frame = NULL;
         glusterd_conf_t         *conf = NULL;
         xlator_t                *this = NULL;
-
+        glusterd_friend_sm_event_type_t event_type = GD_FRIEND_EVENT_NONE;
+        glusterd_probe_ctx_t            *ctx = NULL;
+        glusterd_friend_sm_event_t      *new_event = NULL;
 
         GF_ASSERT (event);
         peerinfo = event->peerinfo;
@@ -222,15 +381,46 @@ glusterd_ac_send_friend_remove_req (glusterd_friend_sm_event_t *event, void *ctx
         conf = this->private;
 
         GF_ASSERT (conf);
-        GF_ASSERT (conf->mgmt);
 
-        proc = &conf->mgmt->proctable[GD_MGMT_FRIEND_REMOVE];
+        ctx = event->ctx;
+
+        if (!peerinfo->connected) {
+                event_type = GD_FRIEND_EVENT_REMOVE_FRIEND;
+
+                ret = glusterd_friend_sm_new_event (event_type, &new_event);
+
+                if (!ret) {
+                        new_event->peerinfo = peerinfo;
+                        ret = glusterd_friend_sm_inject_event (new_event);
+                } else {
+                        gf_log ("glusterd", GF_LOG_ERROR,
+                                 "Unable to get event");
+                }
+
+                if (ctx)
+                        ret = glusterd_xfer_cli_deprobe_resp (ctx->req, ret, 0,
+                                                              NULL,
+                                                              ctx->hostname,
+                                                              ctx->dict);
+                glusterd_friend_sm ();
+                glusterd_op_sm ();
+
+                if (ctx) {
+                        glusterd_broadcast_friend_delete (ctx->hostname, NULL);
+                        glusterd_destroy_probe_ctx (ctx);
+                }
+                goto out;
+        }
+
+        if (!peerinfo->peer)
+                goto out;
+        proc = &peerinfo->peer->proctable[GLUSTERD_FRIEND_REMOVE];
         if (proc->fn) {
                 frame = create_frame (this, this->ctx->pool);
                 if (!frame) {
                         goto out;
                 }
-                frame->local = ctx;
+                frame->local = data;
                 ret = proc->fn (frame, this, event);
         }
 
@@ -240,42 +430,131 @@ out:
         return ret;
 }
 
+static gf_boolean_t
+glusterd_should_update_peer (glusterd_peerinfo_t *peerinfo,
+                             glusterd_peerinfo_t *cur_peerinfo)
+{
+        gf_boolean_t is_valid = _gf_false;
+
+        if ((peerinfo == cur_peerinfo) ||
+            (peerinfo->state.state == GD_FRIEND_STATE_BEFRIENDED))
+                is_valid = _gf_true;
+
+        return is_valid;
+}
+
 static int
 glusterd_ac_send_friend_update (glusterd_friend_sm_event_t *event, void *ctx)
 {
-        int                     ret = 0;
-        glusterd_peerinfo_t     *peerinfo = NULL;
-        rpc_clnt_procedure_t    *proc = NULL;
-        call_frame_t            *frame = NULL;
-        glusterd_conf_t         *conf = NULL;
-        xlator_t                *this = NULL;
-
+        int                           ret               = 0;
+        glusterd_peerinfo_t          *cur_peerinfo      = NULL;
+        glusterd_peerinfo_t          *peerinfo          = NULL;
+        rpc_clnt_procedure_t         *proc              = NULL;
+        xlator_t                     *this              = NULL;
+        glusterd_friend_update_ctx_t  ev_ctx            = {{0}};
+        glusterd_conf_t              *priv              = NULL;
+        dict_t                       *friends           = NULL;
+        char                          key[100]          = {0,};
+        char                         *dup_buf           = NULL;
+        int32_t                       count             = 0;
 
         GF_ASSERT (event);
-        peerinfo = event->peerinfo;
+        cur_peerinfo = event->peerinfo;
 
         this = THIS;
-        conf = this->private;
+        priv = this->private;
 
-        GF_ASSERT (conf);
-        GF_ASSERT (conf->mgmt);
+        GF_ASSERT (priv);
 
-        proc = &conf->mgmt->proctable[GD_MGMT_FRIEND_UPDATE];
-        if (proc->fn) {
-                frame = create_frame (this, this->ctx->pool);
-                if (!frame) {
+        ev_ctx.op = GD_FRIEND_UPDATE_ADD;
+
+        friends = dict_new ();
+        if (!friends)
+                goto out;
+
+        snprintf (key, sizeof (key), "op");
+        ret = dict_set_int32 (friends, key, ev_ctx.op);
+        if (ret)
+                goto out;
+
+        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
+                if (!glusterd_should_update_peer (peerinfo, cur_peerinfo))
+                        continue;
+
+                count++;
+                snprintf (key, sizeof (key), "friend%d.uuid", count);
+                dup_buf = gf_strdup (uuid_utoa (peerinfo->uuid));
+                ret = dict_set_dynstr (friends, key, dup_buf);
+                if (ret)
+                        goto out;
+                snprintf (key, sizeof (key), "friend%d.hostname", count);
+                ret = dict_set_str (friends, key, peerinfo->hostname);
+                if (ret)
+                        goto out;
+                gf_log ("", GF_LOG_INFO, "Added uuid: %s, host: %s",
+                        dup_buf, peerinfo->hostname);
+        }
+
+        ret = dict_set_int32 (friends, "count", count);
+        if (ret)
+                goto out;
+
+        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
+                if (!peerinfo->connected || !peerinfo->peer)
+                        continue;
+
+                if (!glusterd_should_update_peer (peerinfo, cur_peerinfo))
+                        continue;
+
+                ret = dict_set_static_ptr (friends, "peerinfo", peerinfo);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "failed to set peerinfo");
                         goto out;
                 }
-                frame->local = ctx;
-                ret = proc->fn (frame, this, event);
+
+                proc = &peerinfo->peer->proctable[GLUSTERD_FRIEND_UPDATE];
+                if (proc->fn) {
+                        ret = proc->fn (NULL, this, friends);
+                }
         }
 
-out:
         gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
 
+out:
+        if (friends)
+                dict_unref (friends);
+
         return ret;
 }
 
+static int
+glusterd_peer_detach_cleanup (glusterd_conf_t *priv)
+{
+        int                     ret = -1;
+        glusterd_volinfo_t      *volinfo = NULL;
+        glusterd_volinfo_t      *tmp_volinfo = NULL;
+
+        GF_ASSERT (priv);
+
+        list_for_each_entry_safe (volinfo,tmp_volinfo,
+                                  &priv->volumes, vol_list) {
+                if (!glusterd_friend_contains_vol_bricks (volinfo,
+                                                          MY_UUID)) {
+                        gf_log (THIS->name, GF_LOG_INFO,
+                                "Deleting stale volume %s", volinfo->volname);
+                        ret = glusterd_delete_volume (volinfo);
+                        if (ret) {
+                                gf_log (THIS->name, GF_LOG_ERROR,
+                                        "Error deleting stale volume");
+                                goto out;
+                        }
+                }
+        }
+        ret = 0;
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
 
 static int
 glusterd_ac_handle_friend_remove_req (glusterd_friend_sm_event_t *event,
@@ -285,32 +564,40 @@ glusterd_ac_handle_friend_remove_req (glusterd_friend_sm_event_t *event,
         glusterd_peerinfo_t             *peerinfo = NULL;
         glusterd_friend_req_ctx_t       *ev_ctx = NULL;
         glusterd_friend_sm_event_t      *new_event = NULL;
+        glusterd_conf_t                 *priv = NULL;
 
         GF_ASSERT (ctx);
         ev_ctx = ctx;
         peerinfo = event->peerinfo;
         GF_ASSERT (peerinfo);
 
-        uuid_clear (peerinfo->uuid);
+        priv = THIS->private;
+        GF_ASSERT (priv);
 
         ret = glusterd_xfer_friend_remove_resp (ev_ctx->req, ev_ctx->hostname,
                                                 ev_ctx->port);
 
-        peerinfo->rpc = rpc_clnt_unref (peerinfo->rpc);
-        peerinfo->rpc = NULL;
-
-        ret = glusterd_friend_sm_new_event (GD_FRIEND_EVENT_REMOVE_FRIEND,
-                                            &new_event);
+        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
 
-        if (ret)
-                goto out;
-
-        new_event->peerinfo = peerinfo;
+                ret = glusterd_friend_sm_new_event (GD_FRIEND_EVENT_REMOVE_FRIEND,
+                                                    &new_event);
+                if (ret)
+                        goto out;
 
-        ret = glusterd_friend_sm_inject_event (new_event);
+                new_event->peerinfo = peerinfo;
 
+                ret = glusterd_friend_sm_inject_event (new_event);
+                if (ret)
+                        goto out;
+        }
+        ret = glusterd_peer_detach_cleanup (priv);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "Peer detach cleanup was not successful");
+                ret = 0;
+        }
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret);
 
         return ret;
 }
@@ -320,13 +607,15 @@ glusterd_ac_friend_remove (glusterd_friend_sm_event_t *event, void *ctx)
 {
         int                     ret = -1;
 
-        ret = glusterd_friend_cleanup (event->peerinfo);
+        ret = glusterd_friend_remove_cleanup_vols (event->peerinfo->uuid);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING, "Volumes cleanup failed");
 
+        ret = glusterd_friend_cleanup (event->peerinfo);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Cleanup returned: %d", ret);
+                gf_log (THIS->name, GF_LOG_ERROR, "Cleanup returned: %d", ret);
         }
 
-
         return 0;
 }
 
@@ -353,7 +642,6 @@ glusterd_ac_handle_friend_add_req (glusterd_friend_sm_event_t *event, void *ctx)
         int                             status = 0;
         int32_t                         op_ret = -1;
         int32_t                         op_errno = 0;
-        char               remote_hostname[UNIX_PATH_MAX + 1] = {0,};
 
         GF_ASSERT (ctx);
         ev_ctx = ctx;
@@ -363,15 +651,15 @@ glusterd_ac_handle_friend_add_req (glusterd_friend_sm_event_t *event, void *ctx)
         uuid_copy (peerinfo->uuid, ev_ctx->uuid);
 
         //Build comparison logic here.
-        ret = glusterd_compare_friend_data (ev_ctx->vols, &status);
+        ret = glusterd_compare_friend_data (ev_ctx->vols, &status,
+                                            peerinfo->hostname);
         if (ret)
                 goto out;
 
         if (GLUSTERD_VOL_COMP_RJT != status) {
                 event_type = GD_FRIEND_EVENT_LOCAL_ACC;
                 op_ret = 0;
-        }
-        else {
+        } else {
                 event_type = GD_FRIEND_EVENT_LOCAL_RJT;
                 op_errno = GF_PROBE_VOLUME_CONFLICT;
                 op_ret = -1;
@@ -394,21 +682,15 @@ glusterd_ac_handle_friend_add_req (glusterd_friend_sm_event_t *event, void *ctx)
 
         uuid_copy (new_ev_ctx->uuid, ev_ctx->uuid);
         new_ev_ctx->hostname = gf_strdup (ev_ctx->hostname);
+        new_ev_ctx->op = GD_FRIEND_UPDATE_ADD;
 
-        ret = glusterd_remote_hostname_get (ev_ctx->req, remote_hostname,
-                                            sizeof (remote_hostname));
-        if (ret) {
-                ret = -1;
-                goto out;
-        }
-
-        new_ev_ctx->remote_hostname = gf_strdup (remote_hostname);
         new_event->ctx = new_ev_ctx;
 
         glusterd_friend_sm_inject_event (new_event);
 
         ret = glusterd_xfer_friend_add_resp (ev_ctx->req, ev_ctx->hostname,
-                                             ev_ctx->port, op_ret, op_errno);
+                                             peerinfo->hostname, ev_ctx->port,
+                                             op_ret, op_errno);
 
 out:
         gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
@@ -425,10 +707,11 @@ glusterd_friend_sm_transition_state (glusterd_peerinfo_t *peerinfo,
         GF_ASSERT (state);
         GF_ASSERT (peerinfo);
 
-        //peerinfo->state.state = state;
+        (void) glusterd_sm_tr_log_transition_add (&peerinfo->sm_log,
+                                           peerinfo->state.state,
+                                           state[event_type].next_state,
+                                           event_type);
 
-        gf_log ("", GF_LOG_NORMAL, "Transitioning from %d to %d",
-                        peerinfo->state.state, state[event_type].next_state);
         peerinfo->state.state = state[event_type].next_state;
         return 0;
 }
@@ -446,9 +729,57 @@ glusterd_sm_t glusterd_state_default [] = {
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_send_friend_remove_req}, //EV_INIT_REMOVE_FRIEND
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_none}, //EVENT_RCVD_REMOVE_FRIEND
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_remove}, //EVENT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_probe}, //EVENT_CONNECTED
+        {GD_FRIEND_STATE_DEFAULT, glusterd_ac_none}, //EVENT_MAX
+};
+
+glusterd_sm_t glusterd_state_probe_rcvd [] = {
+        {GD_FRIEND_STATE_PROBE_RCVD, glusterd_ac_none},
+        {GD_FRIEND_STATE_PROBE_RCVD, glusterd_ac_none}, //EV_PROBE
+        {GD_FRIEND_STATE_PROBE_RCVD, glusterd_ac_none}, //EV_INIT_FRIEND_REQ
+        {GD_FRIEND_STATE_PROBE_RCVD, glusterd_ac_none}, //EVENT_RCVD_ACC
+        {GD_FRIEND_STATE_PROBE_RCVD, glusterd_ac_none}, //EVENT_RCVD_LOCAL_ACC
+        {GD_FRIEND_STATE_PROBE_RCVD, glusterd_ac_none}, //EVENT_RCVD_RJT
+        {GD_FRIEND_STATE_PROBE_RCVD, glusterd_ac_none}, //EVENT_RCVD_LOCAL_RJT
+        {GD_FRIEND_STATE_REQ_RCVD, glusterd_ac_handle_friend_add_req}, //EVENT_RCV_FRIEND_REQ
+        {GD_FRIEND_STATE_DEFAULT, glusterd_ac_send_friend_remove_req}, //EV_INIT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_DEFAULT, glusterd_ac_none}, //EVENT_RCVD_REMOVE_FRIEND
+        {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_remove}, //EVENT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_CONNECTED_RCVD, glusterd_ac_none}, //EVENT_CONNECTED
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_none}, //EVENT_MAX
 };
 
+glusterd_sm_t glusterd_state_connected_rcvd [] = {
+        {GD_FRIEND_STATE_CONNECTED_RCVD, glusterd_ac_none},
+        {GD_FRIEND_STATE_CONNECTED_RCVD, glusterd_ac_none}, //EV_PROBE
+        {GD_FRIEND_STATE_CONNECTED_RCVD, glusterd_ac_none}, //EV_INIT_FRIEND_REQ
+        {GD_FRIEND_STATE_CONNECTED_RCVD, glusterd_ac_none}, //EVENT_RCVD_ACC
+        {GD_FRIEND_STATE_CONNECTED_ACCEPTED, glusterd_ac_reverse_probe_begin}, //EVENT_RCVD_LOCAL_ACC
+        {GD_FRIEND_STATE_CONNECTED_RCVD, glusterd_ac_none}, //EVENT_RCVD_RJT
+        {GD_FRIEND_STATE_REJECTED, glusterd_ac_none}, //EVENT_RCVD_LOCAL_RJT
+        {GD_FRIEND_STATE_CONNECTED_RCVD, glusterd_ac_handle_friend_add_req}, //EVENT_RCV_FRIEND_REQ
+        {GD_FRIEND_STATE_DEFAULT, glusterd_ac_send_friend_remove_req}, //EV_INIT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_DEFAULT, glusterd_ac_none}, //EVENT_RCVD_REMOVE_FRIEND
+        {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_remove}, //EVENT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_CONNECTED_RCVD, glusterd_ac_none}, //EVENT_CONNECTED
+        {GD_FRIEND_STATE_CONNECTED_RCVD, glusterd_ac_none}, //EVENT_MAX
+};
+
+glusterd_sm_t glusterd_state_connected_accepted [] = {
+        {GD_FRIEND_STATE_CONNECTED_ACCEPTED, glusterd_ac_none},
+        {GD_FRIEND_STATE_CONNECTED_ACCEPTED, glusterd_ac_friend_probe}, //EV_PROBE
+        {GD_FRIEND_STATE_REQ_SENT_RCVD, glusterd_ac_friend_add}, //EV_INIT_FRIEND_REQ
+        {GD_FRIEND_STATE_CONNECTED_ACCEPTED, glusterd_ac_none}, //EVENT_RCVD_ACC
+        {GD_FRIEND_STATE_CONNECTED_ACCEPTED, glusterd_ac_none}, //EVENT_RCVD_LOCAL_ACC
+        {GD_FRIEND_STATE_CONNECTED_ACCEPTED, glusterd_ac_none}, //EVENT_RCVD_RJT
+        {GD_FRIEND_STATE_CONNECTED_ACCEPTED, glusterd_ac_none}, //EVENT_RCVD_LOCAL_RJT
+        {GD_FRIEND_STATE_CONNECTED_ACCEPTED, glusterd_ac_none}, //EVENT_RCV_FRIEND_REQ
+        {GD_FRIEND_STATE_DEFAULT, glusterd_ac_send_friend_remove_req}, //EV_INIT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_DEFAULT, glusterd_ac_none}, //EVENT_RCVD_REMOVE_FRIEND
+        {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_remove}, //EVENT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_CONNECTED_ACCEPTED, glusterd_ac_none}, //EVENT_CONNECTED
+        {GD_FRIEND_STATE_CONNECTED_ACCEPTED, glusterd_ac_none}, //EVENT_MAX
+};
 
 glusterd_sm_t  glusterd_state_req_sent [] = {
         {GD_FRIEND_STATE_REQ_SENT, glusterd_ac_none}, //EVENT_NONE,
@@ -462,21 +793,23 @@ glusterd_sm_t  glusterd_state_req_sent [] = {
         {GD_FRIEND_STATE_UNFRIEND_SENT, glusterd_ac_send_friend_remove_req}, //EVENT_INIT_REMOVE_FRIEND,
         {GD_FRIEND_STATE_REQ_SENT, glusterd_ac_none}, //EVENT_RCVD_REMOVE_FRIEND
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_remove}, //EVENT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_REQ_SENT, glusterd_ac_none},//EVENT_CONNECTED
         {GD_FRIEND_STATE_REQ_SENT, glusterd_ac_none},//EVENT_MAX
 };
 
 glusterd_sm_t  glusterd_state_req_rcvd [] = {
         {GD_FRIEND_STATE_REQ_RCVD, glusterd_ac_none}, //EVENT_NONE,
-        {GD_FRIEND_STATE_REQ_RCVD, glusterd_ac_friend_probe}, //EVENT_PROBE,
-        {GD_FRIEND_STATE_REQ_SENT_RCVD, glusterd_ac_friend_add}, //EVENT_INIT_FRIEND_REQ,
+        {GD_FRIEND_STATE_REQ_RCVD, glusterd_ac_none}, //EVENT_PROBE,
+        {GD_FRIEND_STATE_REQ_SENT_RCVD, glusterd_ac_none}, //EVENT_INIT_FRIEND_REQ,
         {GD_FRIEND_STATE_REQ_RCVD, glusterd_ac_none}, //EVENT_RCVD_ACC
-        {GD_FRIEND_STATE_REQ_RCVD, glusterd_ac_reverse_probe_begin}, //EVENT_RCVD_LOCAL_ACC
+        {GD_FRIEND_STATE_REQ_ACCEPTED, glusterd_ac_none}, //EVENT_RCVD_LOCAL_ACC
         {GD_FRIEND_STATE_REQ_RCVD, glusterd_ac_none}, //EVENT_RCVD_RJT
         {GD_FRIEND_STATE_REJECTED, glusterd_ac_none}, //EVENT_RCVD_LOCAL_RJT
         {GD_FRIEND_STATE_REQ_RCVD, glusterd_ac_none}, //EVENT_RCV_FRIEND_REQ
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_send_friend_remove_req}, //EVENT_INIT_REMOVE_FRIEND,
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_handle_friend_remove_req}, //EVENT_RCVD_REMOVE_FRIEND
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_remove}, //EVENT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_CONNECTED_RCVD, glusterd_ac_none},//EVENT_CONNECTED
         {GD_FRIEND_STATE_REQ_RCVD, glusterd_ac_none},//EVENT_MAX
 };
 
@@ -485,13 +818,14 @@ glusterd_sm_t  glusterd_state_befriended [] = {
         {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_none}, //EVENT_PROBE,
         {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_none}, //EVENT_INIT_FRIEND_REQ,
         {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_none}, //EVENT_RCVD_ACC
-        {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_none}, //EVENT_RCVD_LOCAL_ACC
-        {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_none}, //EVENT_RCVD_RJT
-        {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_none}, //EVENT_RCVD_LOCAL_RJT
+        {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_send_friend_update}, //EVENT_RCVD_LOCAL_ACC
+        {GD_FRIEND_STATE_REJECTED, glusterd_ac_none}, //EVENT_RCVD_RJT
+        {GD_FRIEND_STATE_REJECTED, glusterd_ac_none}, //EVENT_RCVD_LOCAL_RJT
         {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_handle_friend_add_req}, //EVENT_RCV_FRIEND_REQ
         {GD_FRIEND_STATE_UNFRIEND_SENT, glusterd_ac_send_friend_remove_req}, //EVENT_INIT_REMOVE_FRIEND,
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_handle_friend_remove_req}, //EVENT_RCVD_REMOVE_FRIEND
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_remove}, //EVENT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_friend_add},//EVENT_CONNECTED
         {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_none},//EVENT_MAX
 };
 
@@ -507,6 +841,7 @@ glusterd_sm_t  glusterd_state_req_sent_rcvd [] = {
         {GD_FRIEND_STATE_UNFRIEND_SENT, glusterd_ac_send_friend_remove_req}, //EVENT_INIT_REMOVE_FRIEND,
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_handle_friend_remove_req}, //EVENT_RCVD_REMOVE_FRIEND
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_remove}, //EVENT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_REQ_SENT_RCVD, glusterd_ac_none},//EVENT_CONNECTED
         {GD_FRIEND_STATE_REQ_SENT_RCVD, glusterd_ac_none},//EVENT_MAX
 };
 
@@ -514,14 +849,15 @@ glusterd_sm_t  glusterd_state_rejected [] = {
         {GD_FRIEND_STATE_REJECTED, glusterd_ac_none}, //EVENT_NONE,
         {GD_FRIEND_STATE_REJECTED, glusterd_ac_friend_probe}, //EVENT_PROBE,
         {GD_FRIEND_STATE_REQ_SENT, glusterd_ac_friend_add}, //EVENT_INIT_FRIEND_REQ,
-        {GD_FRIEND_STATE_REJECTED, glusterd_ac_none}, //EVENT_RCVD_ACC
-        {GD_FRIEND_STATE_REJECTED, glusterd_ac_none}, //EVENT_RCVD_LOCAL_ACC
+        {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_none}, //EVENT_RCVD_ACC
+        {GD_FRIEND_STATE_BEFRIENDED, glusterd_ac_none}, //EVENT_RCVD_LOCAL_ACC
         {GD_FRIEND_STATE_REJECTED, glusterd_ac_none}, //EVENT_RCVD_RJT
         {GD_FRIEND_STATE_REJECTED, glusterd_ac_none}, //EVENT_RCVD_LOCAL_RJT
         {GD_FRIEND_STATE_REQ_RCVD, glusterd_ac_handle_friend_add_req}, //EVENT_RCV_FRIEND_REQ
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_send_friend_remove_req}, //EVENT_INIT_REMOVE_FRIEND
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_handle_friend_remove_req}, //EVENT_RCVD_REMOVE_FRIEND
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_remove}, //EVENT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_REJECTED, glusterd_ac_friend_add},//EVENT_CONNECTED
         {GD_FRIEND_STATE_REQ_RCVD, glusterd_ac_none},//EVENT_MAX
 };
 
@@ -537,11 +873,12 @@ glusterd_sm_t  glusterd_state_req_accepted [] = {
         {GD_FRIEND_STATE_REQ_ACCEPTED, glusterd_ac_send_friend_remove_req}, //EVENT_INIT_REMOVE_FRIEND
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_handle_friend_remove_req}, //EVENT_RCVD_REMOVE_FRIEND
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_remove}, //EVENT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_CONNECTED_ACCEPTED, glusterd_ac_reverse_probe_begin},//EVENT_CONNECTED
         {GD_FRIEND_STATE_REQ_SENT, glusterd_ac_none},//EVENT_MAX
 };
 
 glusterd_sm_t  glusterd_state_unfriend_sent [] = {
-        {GD_FRIEND_STATE_REQ_ACCEPTED, glusterd_ac_none}, //EVENT_NONE,
+        {GD_FRIEND_STATE_UNFRIEND_SENT, glusterd_ac_none}, //EVENT_NONE,
         {GD_FRIEND_STATE_UNFRIEND_SENT, glusterd_ac_error}, //EVENT_PROBE,
         {GD_FRIEND_STATE_UNFRIEND_SENT, glusterd_ac_none}, //EVENT_INIT_FRIEND_REQ,
         {GD_FRIEND_STATE_UNFRIEND_SENT, glusterd_ac_none}, //EVENT_RCVD_ACC
@@ -552,6 +889,7 @@ glusterd_sm_t  glusterd_state_unfriend_sent [] = {
         {GD_FRIEND_STATE_UNFRIEND_SENT, glusterd_ac_none}, //EVENT_INIT_REMOVE_FRIEND
         {GD_FRIEND_STATE_UNFRIEND_SENT, glusterd_ac_none}, //EVENT_RCVD_REMOVE_FRIEND
         {GD_FRIEND_STATE_DEFAULT, glusterd_ac_friend_remove}, //EVENT_REMOVE_FRIEND
+        {GD_FRIEND_STATE_UNFRIEND_SENT, glusterd_ac_none},//EVENT_CONNECTED
         {GD_FRIEND_STATE_UNFRIEND_SENT, glusterd_ac_none},//EVENT_MAX
 };
 
@@ -564,6 +902,9 @@ glusterd_sm_t *glusterd_friend_state_table [] = {
         glusterd_state_req_sent_rcvd,
         glusterd_state_rejected,
         glusterd_state_unfriend_sent,
+        glusterd_state_probe_rcvd,
+        glusterd_state_connected_rcvd,
+        glusterd_state_connected_accepted
 };
 
 int
@@ -592,8 +933,8 @@ int
 glusterd_friend_sm_inject_event (glusterd_friend_sm_event_t *event)
 {
         GF_ASSERT (event);
-        gf_log ("glusterd", GF_LOG_NORMAL, "Enqueuing event: %d",
-                        event->event);
+        gf_log ("glusterd", GF_LOG_DEBUG, "Enqueue event: '%s'",
+                glusterd_friend_sm_event_name_get (event->event));
         list_add_tail (&event->list, &gd_friend_sm_queue);
 
         return 0;
@@ -614,65 +955,32 @@ glusterd_destroy_friend_event_context (glusterd_friend_sm_event_t *event)
         case GD_FRIEND_EVENT_LOCAL_RJT:
         case GD_FRIEND_EVENT_RCVD_ACC:
         case GD_FRIEND_EVENT_RCVD_RJT:
-                glusterd_destroy_friend_update_ctx(event->ctx);
+                glusterd_destroy_friend_update_ctx (event->ctx);
                 break;
         default:
                 break;
         }
 }
 
-int
-glusterd_check_and_add_friend (glusterd_friend_sm_event_t *event)
+gf_boolean_t
+gd_does_peer_affect_quorum (glusterd_friend_sm_state_t old_state,
+                            glusterd_friend_sm_event_type_t event_type,
+                            glusterd_peerinfo_t *peerinfo)
 {
-        rpcsvc_request_t                *req = NULL;
-        glusterd_peerinfo_t             *peerinfo   = NULL;
-        glusterd_friend_sm_event_type_t  event_type = 0;
-        glusterd_friend_req_ctx_t       *fr_ctx = NULL;
-        glusterd_probe_ctx_t            *pb_ctx = NULL;
-        gf_boolean_t                    add_friend = _gf_false;
-        char                            rhost[UNIX_PATH_MAX + 1] = {0};
-        char                            *host_str = NULL;
-        int                             port       = 6969; //TODO, use standard
-        int                             ret = 0;
+        gf_boolean_t    affects = _gf_false;
 
-        peerinfo = event->peerinfo;
-        event_type = event->event;
-
-        if (!peerinfo &&
-           (GD_FRIEND_EVENT_PROBE == event_type)) {
-                add_friend = _gf_true;
-                pb_ctx = event->ctx;
-                req = pb_ctx->req;
-        }
-        if (!peerinfo &&
-            (GD_FRIEND_EVENT_RCVD_FRIEND_REQ == event_type)) {
-                add_friend = _gf_true;
-                fr_ctx = event->ctx;
-                req = fr_ctx->req;
-        }
-        if (add_friend) {
-                if (req) {
-                        ret = glusterd_remote_hostname_get (req, rhost,
-                                                            sizeof (rhost));
-                        if (!ret)
-                                host_str = rhost;
-                }
-                ret = glusterd_friend_add ((const char*)host_str, port,
-                                          GD_FRIEND_STATE_DEFAULT,
-                                          NULL, NULL, &peerinfo, 0);
-                if (ret) {
-                        gf_log ("glusterd", GF_LOG_ERROR, "Unable to add peer, "
-                                "ret = %d", ret);
-                        ret = 1;
-                        goto out;
-                }
-                GF_ASSERT (peerinfo);
-                event->peerinfo = peerinfo;
+        //When glusterd comes up with friends in BEFRIENDED state in store,
+        //wait until compare-data happens.
+        if ((old_state == GD_FRIEND_STATE_BEFRIENDED) &&
+            (event_type != GD_FRIEND_EVENT_RCVD_ACC) &&
+            (event_type != GD_FRIEND_EVENT_LOCAL_ACC))
+                goto out;
+        if ((peerinfo->state.state == GD_FRIEND_STATE_BEFRIENDED)
+            && peerinfo->connected) {
+                affects = _gf_true;
         }
-        ret = 0;
-
 out:
-        return ret;
+        return affects;
 }
 
 int
@@ -686,23 +994,28 @@ glusterd_friend_sm ()
         glusterd_peerinfo_t             *peerinfo   = NULL;
         glusterd_friend_sm_event_type_t  event_type = 0;
         gf_boolean_t                     is_await_conn = _gf_false;
-        int                              loop       = 0;
+        gf_boolean_t                     quorum_action = _gf_false;
+        glusterd_friend_sm_state_t       old_state = GD_FRIEND_STATE_DEFAULT;
 
         while (!list_empty (&gd_friend_sm_queue)) {
                 list_for_each_entry_safe (event, tmp, &gd_friend_sm_queue, list) {
 
                         list_del_init (&event->list);
                         event_type = event->event;
+                        peerinfo = event->peerinfo;
+                        if (!peerinfo) {
+                                gf_log ("glusterd", GF_LOG_CRITICAL, "Received"
+                                        " event %s with empty peer info",
+                                glusterd_friend_sm_event_name_get (event_type));
 
-
-                        loop = glusterd_check_and_add_friend (event);
-                        if (loop)
+                                GF_FREE (event);
                                 continue;
+                        }
+                        gf_log ("", GF_LOG_DEBUG, "Dequeued event of type: '%s'",
+                                glusterd_friend_sm_event_name_get (event_type));
 
-                        peerinfo = event->peerinfo;
-                        if (!peerinfo)
-                                goto out;
 
+                        old_state = peerinfo->state.state;
                         state = glusterd_friend_state_table[peerinfo->state.state];
 
                         GF_ASSERT (state);
@@ -731,16 +1044,28 @@ glusterd_friend_sm ()
                                 continue;
                         }
 
-                        ret = glusterd_friend_sm_transition_state (peerinfo, state, event_type);
+                        ret = glusterd_friend_sm_transition_state (peerinfo,
+                                                            state, event_type);
 
                         if (ret) {
                                 gf_log ("glusterd", GF_LOG_ERROR, "Unable to transition"
-                                        "state from %d to %d", peerinfo->state.state,
-                                         state[event_type].next_state);
+                                        " state from '%s' to '%s' for event '%s'",
+                        glusterd_friend_sm_state_name_get(peerinfo->state.state),
+                        glusterd_friend_sm_state_name_get(state[event_type].next_state),
+                                glusterd_friend_sm_event_name_get(event_type));
                                 goto out;
                         }
 
-                        ret = glusterd_store_update_peerinfo (peerinfo);
+                        if (gd_does_peer_affect_quorum (old_state, event_type,
+                                                        peerinfo)) {
+                                peerinfo->quorum_contrib = QUORUM_UP;
+                                if (peerinfo->quorum_action) {
+                                        peerinfo->quorum_action = _gf_false;
+                                        quorum_action = _gf_true;
+                                }
+                        }
+
+                        ret = glusterd_store_peerinfo (peerinfo);
 
                         glusterd_destroy_friend_event_context (event);
                         GF_FREE (event);
@@ -751,9 +1076,27 @@ glusterd_friend_sm ()
                         break;
         }
 
-
         ret = 0;
 out:
+        if (quorum_action) {
+            /* When glusterd is restarted, it needs to wait until the 'friends' view
+             * of the volumes settle, before it starts any of the internal daemons.
+             *
+             * Every friend that was part of the cluster, would send its
+             * cluster-view, 'our' way. For every friend, who belongs to
+             * a partition which has a different cluster-view from our
+             * partition, we may update our cluster-view. For subsequent
+             * friends from that partition would agree with us, if the first
+             * friend wasn't rejected. For every first friend, whom we agreed with,
+             * we would need to start internal daemons/bricks belonging to the
+             * new volumes.
+             * glusterd_spawn_daemons calls functions that are idempotent. ie,
+             * the functions spawn process(es) only if they are not started yet.
+             *
+             * */
+                glusterd_spawn_daemons (NULL);
+                glusterd_do_quorum_action ();
+        }
         return ret;
 }
 
diff --git a/xlators/mgmt/glusterd/src/glusterd-sm.h b/xlators/mgmt/glusterd/src/glusterd-sm.h
index 1b845b902..b9bedbe69 100644
--- a/xlators/mgmt/glusterd/src/glusterd-sm.h
+++ b/xlators/mgmt/glusterd/src/glusterd-sm.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _GLUSTERD_SM_H_
 #define _GLUSTERD_SM_H_
 
@@ -37,14 +27,21 @@
 #include "byte-order.h"
 //#include "glusterd.h"
 #include "rpcsvc.h"
-
-struct glusterd_store_handle_ {
-        char    *path;
-        int     fd;
-        FILE    *read;
-};
-
-typedef struct glusterd_store_handle_  glusterd_store_handle_t;
+#include "store.h"
+
+typedef enum gd_quorum_contribution_ {
+        QUORUM_NONE,
+        QUORUM_WAITING,
+        QUORUM_DOWN,
+        QUORUM_UP
+} gd_quorum_contrib_t;
+
+typedef enum gd_quorum_status_ {
+        QUORUM_UNKNOWN,
+        QUORUM_NOT_APPLICABLE,
+        QUORUM_MEETS,
+        QUORUM_DOES_NOT_MEET
+} gd_quorum_status_t;
 
 typedef enum glusterd_friend_sm_state_ {
         GD_FRIEND_STATE_DEFAULT = 0,
@@ -55,6 +52,9 @@ typedef enum glusterd_friend_sm_state_ {
         GD_FRIEND_STATE_REQ_SENT_RCVD,
         GD_FRIEND_STATE_REJECTED,
         GD_FRIEND_STATE_UNFRIEND_SENT,
+        GD_FRIEND_STATE_PROBE_RCVD,
+        GD_FRIEND_STATE_CONNECTED_RCVD,
+        GD_FRIEND_STATE_CONNECTED_ACCEPTED,
         GD_FRIEND_STATE_MAX
 } glusterd_friend_sm_state_t;
 
@@ -68,23 +68,63 @@ typedef struct glusterd_peer_hostname_ {
         struct list_head        hostname_list;
 }glusterd_peer_hostname_t;
 
+typedef struct glusterd_sm_transition_ {
+        int             old_state;
+        int             event;
+        int             new_state;
+        time_t          time;
+} glusterd_sm_transition_t;
+
+typedef struct glusterd_sm_tr_log_ {
+        glusterd_sm_transition_t    *transitions;
+        size_t                      current;
+        size_t                      size;
+        size_t                      count;
+        char*                       (*state_name_get) (int);
+        char*                       (*event_name_get) (int);
+} glusterd_sm_tr_log_t;
+
 struct glusterd_peerinfo_ {
         uuid_t                          uuid;
-        char                            uuid_str[50];
+        char                            uuid_str[50]; /* Retrieve this using
+                                                       * gd_peer_uuid_str ()
+                                                       */
         glusterd_peer_state_info_t      state;
         char                            *hostname;
         int                             port;
         struct list_head                uuid_list;
         struct list_head                op_peers_list;
-        struct list_head                hostnames;
         struct rpc_clnt                 *rpc;
+        rpc_clnt_prog_t                 *mgmt;
+        rpc_clnt_prog_t                 *peer;
+        rpc_clnt_prog_t                 *mgmt_v3;
         int                             connected;
-        glusterd_store_handle_t         *shandle;
+        gf_store_handle_t         *shandle;
+        glusterd_sm_tr_log_t            sm_log;
+        gf_boolean_t                    quorum_action;
+        gd_quorum_contrib_t             quorum_contrib;
+        gf_boolean_t                    locked;
 };
 
 typedef struct glusterd_peerinfo_ glusterd_peerinfo_t;
 
+typedef enum glusterd_ev_gen_mode_ {
+        GD_MODE_OFF,
+        GD_MODE_ON,
+        GD_MODE_SWITCH_ON
+} glusterd_ev_gen_mode_t;
+
+typedef struct glusterd_peer_ctx_args_ {
+        rpcsvc_request_t        *req;
+        glusterd_ev_gen_mode_t  mode;
+        dict_t                  *dict;
+} glusterd_peerctx_args_t;
 
+typedef struct glusterd_peer_ctx_ {
+        glusterd_peerctx_args_t        args;
+        glusterd_peerinfo_t            *peerinfo;
+        char                           *errstr;
+} glusterd_peerctx_t;
 
 typedef enum glusterd_friend_sm_event_type_ {
         GD_FRIEND_EVENT_NONE = 0,
@@ -98,10 +138,18 @@ typedef enum glusterd_friend_sm_event_type_ {
         GD_FRIEND_EVENT_INIT_REMOVE_FRIEND,
         GD_FRIEND_EVENT_RCVD_REMOVE_FRIEND,
         GD_FRIEND_EVENT_REMOVE_FRIEND,
+        GD_FRIEND_EVENT_CONNECTED,
         GD_FRIEND_EVENT_MAX
 } glusterd_friend_sm_event_type_t;
 
 
+typedef enum glusterd_friend_update_op_ {
+        GD_FRIEND_UPDATE_NONE = 0,
+        GD_FRIEND_UPDATE_ADD,
+        GD_FRIEND_UPDATE_DEL,
+} glusterd_friend_update_op_t;
+
+
 struct glusterd_friend_sm_event_ {
         struct list_head        list;
         glusterd_peerinfo_t     *peerinfo;
@@ -121,18 +169,22 @@ typedef struct glusterd_sm_ {
 typedef struct glusterd_friend_req_ctx_ {
         uuid_t                  uuid;
         char                    *hostname;
-        char                    *remote_hostname;
         rpcsvc_request_t        *req;
         int                      port;
         dict_t                  *vols;
 } glusterd_friend_req_ctx_t;
 
-typedef glusterd_friend_req_ctx_t glusterd_friend_update_ctx_t;
+typedef struct glusterd_friend_update_ctx_ {
+        uuid_t                  uuid;
+        char                    *hostname;
+        int                     op;
+} glusterd_friend_update_ctx_t;
 
 typedef struct glusterd_probe_ctx_ {
         char                    *hostname;
         rpcsvc_request_t        *req;
         int                      port;
+        dict_t                  *dict;
 } glusterd_probe_ctx_t;
 int
 glusterd_friend_sm_new_event (glusterd_friend_sm_event_type_t event_type,
@@ -151,4 +203,15 @@ glusterd_destroy_probe_ctx (glusterd_probe_ctx_t *ctx);
 
 void
 glusterd_destroy_friend_req_ctx (glusterd_friend_req_ctx_t *ctx);
+
+char*
+glusterd_friend_sm_state_name_get (int state);
+
+char*
+glusterd_friend_sm_event_name_get (int event);
+
+int
+glusterd_broadcast_friend_delete (char *hostname, uuid_t uuid);
+void
+glusterd_destroy_friend_update_ctx (glusterd_friend_update_ctx_t *ctx);
 #endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-snapshot.c b/xlators/mgmt/glusterd/src/glusterd-snapshot.c
new file mode 100644
index 000000000..9b811cd05
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-snapshot.c
@@ -0,0 +1,5590 @@
+/*
+   Copyright (c) 2013-2014 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include <inttypes.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <sys/resource.h>
+#include <sys/statvfs.h>
+#include <sys/mount.h>
+
+#include "globals.h"
+#include "compat.h"
+#include "protocol-common.h"
+#include "xlator.h"
+#include "logging.h"
+#include "timer.h"
+#include "glusterd-mem-types.h"
+#include "glusterd.h"
+#include "glusterd-sm.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-utils.h"
+#include "glusterd-store.h"
+#include "run.h"
+#include "glusterd-volgen.h"
+#include "glusterd-mgmt.h"
+#include "glusterd-syncop.h"
+
+#include "syscall.h"
+#include "cli1-xdr.h"
+#include "xdr-generic.h"
+
+#ifdef GF_LINUX_HOST_OS
+#include <mntent.h>
+#endif
+
+char snap_mount_folder[PATH_MAX];
+
+static int32_t
+glusterd_find_missed_snap (dict_t *rsp_dict, glusterd_volinfo_t *vol,
+                           char *snap_uuid, struct list_head *peers,
+                           int32_t op);
+
+/* This function will restore a snapshot volumes
+ *
+ * @param dict          dictionary containing snapshot restore request
+ * @param op_errstr     In case of any failure error message will be returned
+ *                      in this variable
+ * @return              Negative value on Failure and 0 in success
+ */
+int
+glusterd_snapshot_restore (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
+{
+        int                     ret             = -1;
+        char                    *volname        = NULL;
+        char                    *snapname       = NULL;
+        xlator_t                *this           = NULL;
+        glusterd_volinfo_t      *snap_volinfo   = NULL;
+        glusterd_volinfo_t      *volinfo        = NULL;
+        glusterd_snap_t         *snap           = NULL;
+        glusterd_conf_t         *priv           = NULL;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (rsp_dict);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_str (dict, "snapname", &snapname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                        "snap name");
+                goto out;
+        }
+
+        snap = glusterd_find_snap_by_name (snapname);
+        if (NULL == snap) {
+                ret = gf_asprintf (op_errstr, "Snap (%s) not found",
+                                        snapname);
+                if (ret < 0) {
+                        goto out;
+                }
+                gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr);
+                ret = -1;
+                goto out;
+        }
+
+        /* TODO : As of now there is only volume in snapshot.
+         * Change this when multiple volume snapshot is introduced
+         */
+        snap_volinfo = list_entry (snap->volumes.next, glusterd_volinfo_t,
+                                   vol_list);
+
+        ret = glusterd_volinfo_find (snap_volinfo->parent_volname, &volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not get volinfo of "
+                        "%s", snap_volinfo->parent_volname);
+                goto out;
+        }
+
+        if (is_origin_glusterd (dict) == _gf_true) {
+                /* From origin glusterd check if      *
+                 * any peers with snap bricks is down */
+                ret = glusterd_find_missed_snap (rsp_dict, snap_volinfo,
+                                                 snap_volinfo->volname,
+                                                 &priv->peers,
+                                                 GF_SNAP_OPTION_TYPE_RESTORE);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to find missed snap restores");
+                        goto out;
+                }
+        }
+
+        ret = gd_restore_snap_volume (rsp_dict, volinfo, snap_volinfo);
+        if (ret) {
+                /* No need to update op_errstr because it is assumed
+                 * that the called function will do that in case of
+                 * failure.
+                 */
+                gf_log (this->name, GF_LOG_ERROR, "Failed to restore "
+                        "snap for %s volume", volname);
+                goto out;
+        }
+
+        ret = 0;
+
+        /* TODO: Need to check if we need to delete the snap after the
+         * operation is successful or not. Also need to persist the state
+         * of restore operation in the store.
+         */
+out:
+        return ret;
+}
+
+/* This function is called before actual restore is taken place. This function
+ * will validate whether the snapshot volumes are ready to be restored or not.
+ *
+ * @param dict          dictionary containing snapshot restore request
+ * @param op_errstr     In case of any failure error message will be returned
+ *                      in this variable
+ * @param rsp_dict      response dictionary
+ * @return              Negative value on Failure and 0 in success
+ */
+int
+glusterd_snapshot_restore_prevalidate (dict_t *dict, char **op_errstr,
+                                       dict_t *rsp_dict)
+{
+        int                     ret             = -1;
+        int32_t                 i               = 0;
+        int32_t                 volcount        = 0;
+        gf_boolean_t            snap_restored   = _gf_false;
+        char                    key[PATH_MAX]   = {0, };
+        char                    *volname        = NULL;
+        char                    *snapname       = NULL;
+        glusterd_volinfo_t      *volinfo        = NULL;
+        glusterd_snap_t         *snap           = NULL;
+        xlator_t                *this           = NULL;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (rsp_dict);
+
+        ret = dict_get_str (dict, "snapname", &snapname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                        "snap name");
+                goto out;
+        }
+
+        snap = glusterd_find_snap_by_name (snapname);
+        if (NULL == snap) {
+                ret = gf_asprintf (op_errstr, "Snap (%s) not found",
+                                snapname);
+                if (ret < 0) {
+                        goto out;
+                }
+                gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr);
+                ret = -1;
+                goto out;
+        }
+
+        snap_restored = snap->snap_restored;
+
+        if (snap_restored) {
+                ret = gf_asprintf (op_errstr, "Snap (%s) is already "
+                                  "restored", snapname);
+                if (ret < 0) {
+                        goto out;
+                }
+                gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr);
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_str (rsp_dict, "snapname", snapname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+                        "snap name");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "volcount", &volcount);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get volume count");
+                goto out;
+        }
+
+        /* Snapshot restore will only work if all the volumes,
+           that are part of the snapshot, are stopped. */
+        for (i = 1; i <= volcount; ++i) {
+                snprintf (key, sizeof (key), "volname%d", i);
+                ret = dict_get_str (dict, key, &volname);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                "get volume name");
+                        goto out;
+                }
+
+                ret = glusterd_volinfo_find (volname, &volinfo);
+                if (ret) {
+                        ret = gf_asprintf (op_errstr, "Volume (%s) not found",
+                                           volname);
+                        if (ret < 0) {
+                                goto out;
+                        }
+                        gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr);
+                        ret = -1;
+                        goto out;
+                }
+
+                if (glusterd_is_volume_started (volinfo)) {
+                        ret = gf_asprintf (op_errstr, "Volume (%s) has been "
+                        "started. Volume needs to be stopped before restoring "
+                        "a snapshot.", volname);
+                        if (ret < 0) {
+                                goto out;
+                        }
+                        gf_log (this->name, GF_LOG_ERROR, "%s", *op_errstr);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+snap_max_hard_limits_validate (dict_t *dict, char *volname,
+                          uint64_t value, char **op_errstr)
+{
+        char                err_str[PATH_MAX] = "";
+        glusterd_conf_t    *conf              = NULL;
+        glusterd_volinfo_t *volinfo           = NULL;
+        int                 ret               = -1;
+        uint64_t            max_limit         = GLUSTERD_SNAPS_MAX_HARD_LIMIT;
+        xlator_t           *this              = NULL;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+
+        conf = this->private;
+
+        GF_ASSERT (conf);
+
+        if (volname) {
+                ret = glusterd_volinfo_find (volname, &volinfo);
+                if (!ret) {
+                        if (volinfo->is_snap_volume) {
+                                ret = -1;
+                                snprintf (err_str, PATH_MAX,
+                                         "%s is a snap volume. Configuring "
+                                         "snap-max-hard-limit for a snap "
+                                         "volume is prohibited.", volname);
+                                goto out;
+                        }
+                }
+        }
+
+        if (value) {
+                /* Max limit for the system is GLUSTERD_SNAPS_MAX_HARD_LIMIT
+                 * but max limit for a volume is conf->snap_max_hard_limit.
+                 */
+                if (volname) {
+                        max_limit = conf->snap_max_hard_limit;
+                } else {
+                        max_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT;
+                }
+        }
+
+        if ((value < 0) || (value > max_limit)) {
+                ret = -1;
+                snprintf (err_str, PATH_MAX, "Invalid snap-max-hard-limit"
+                          "%"PRIu64 ". Expected range 0 - %"PRIu64,
+                          value, max_limit);
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if (ret) {
+                *op_errstr = gf_strdup (err_str);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+        }
+        return ret;
+}
+
+int
+glusterd_snapshot_config_prevalidate (dict_t *dict, char **op_errstr)
+{
+        char               *volname             = NULL;
+        glusterd_volinfo_t *volinfo             = NULL;
+        xlator_t           *this                = NULL;
+        int                 ret                 = -1;
+        int                 config_command      = 0;
+        char                err_str[PATH_MAX]   = {0,};
+        glusterd_conf_t    *conf                = NULL;
+        uint64_t            value               = 0;
+        uint64_t            hard_limit          = 0;
+        uint64_t            soft_limit          = 0;
+        gf_loglevel_t       loglevel            = GF_LOG_ERROR;
+        uint64_t            max_limit           = GLUSTERD_SNAPS_MAX_HARD_LIMIT;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+
+        conf = this->private;
+
+        GF_ASSERT (conf);
+
+        ret = dict_get_int32 (dict, "config-command", &config_command);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str),
+                          "failed to get config-command type");
+                goto out;
+        }
+
+        ret = dict_get_uint64 (dict, "snap-max-hard-limit", &hard_limit);
+
+        ret = dict_get_uint64 (dict, "snap-max-soft-limit", &soft_limit);
+
+        ret = dict_get_str (dict, "volname", &volname);
+
+        if (volname) {
+                ret = glusterd_volinfo_find (volname, &volinfo);
+                if (ret) {
+                        snprintf (err_str, sizeof (err_str),
+                                  "Volume %s does not exist.", volname);
+                        goto out;
+                }
+        }
+
+        switch (config_command) {
+        case GF_SNAP_CONFIG_TYPE_SET:
+                if (hard_limit) {
+                        /* Validations for snap-max-hard-limits */
+                        ret = snap_max_hard_limits_validate (dict, volname,
+                                                         hard_limit, op_errstr);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "snap-max-hard-limit validation "
+                                        "failed.");
+                                goto out;
+                        }
+                }
+
+                if (soft_limit) {
+                        max_limit = GLUSTERD_SNAPS_MAX_SOFT_LIMIT_PERCENT;
+                        if ((soft_limit < 0) || (soft_limit > max_limit)) {
+                                ret = -1;
+                                snprintf (err_str, PATH_MAX, "Invalid "
+                                         "snap-max-soft-limit ""%"
+                                         PRIu64 ". Expected range 0 - %"PRIu64,
+                                         value, max_limit);
+                                goto out;
+                        }
+                        break;
+                }
+        default:
+                break;
+        }
+
+        ret = 0;
+out:
+
+        if (ret && err_str[0] != '\0') {
+                gf_log (this->name, loglevel, "%s", err_str);
+                *op_errstr = gf_strdup (err_str);
+        }
+
+        return ret;
+}
+
+int
+glusterd_snap_create_pre_val_use_rsp_dict (dict_t *dst, dict_t *src)
+{
+        char                  *snap_brick_dir        = NULL;
+        char                  *snap_device           = NULL;
+        char                  *tmpstr                = NULL;
+        char                   key[PATH_MAX]         = "";
+        char                   snapbrckcnt[PATH_MAX] = "";
+        char                   snapbrckord[PATH_MAX] = "";
+        int                    ret                   = -1;
+        int64_t                i                     = -1;
+        int64_t                j                     = -1;
+        int64_t                volume_count          = 0;
+        int64_t                brick_count           = 0;
+        int64_t                brick_order           = 0;
+        xlator_t              *this                  = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (dst);
+        GF_ASSERT (src);
+
+        ret = dict_get_int64 (src, "volcount", &volume_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to "
+                        "get the volume count");
+                goto out;
+        }
+
+        for (i = 0; i < volume_count; i++) {
+                memset (snapbrckcnt, '\0', sizeof(snapbrckcnt));
+                ret = snprintf (snapbrckcnt, sizeof(snapbrckcnt) - 1,
+                                "vol%ld_brickcount", i+1);
+                ret = dict_get_int64 (src, snapbrckcnt, &brick_count);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "No bricks for this volume in this dict");
+                        continue;
+                }
+
+                for (j = 0; j < brick_count; j++) {
+                        /* Fetching data from source dict */
+                        snprintf (key, sizeof(key) - 1,
+                                  "vol%ld.brickdir%ld", i+1, j);
+
+                        ret = dict_get_ptr (src, key,
+                                            (void **)&snap_brick_dir);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "Unable to fetch %s", key);
+                                continue;
+                        }
+
+                        snprintf (key, sizeof(key) - 1,
+                                  "vol%ld.brick_snapdevice%ld", i+1, j);
+
+                        ret = dict_get_ptr (src, key,
+                                            (void **)&snap_device);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to fetch snap_device");
+                                goto out;
+                        }
+
+                        snprintf (snapbrckord, sizeof(snapbrckord) - 1,
+                                  "vol%ld.brick%ld.order", i+1, j);
+
+                        ret = dict_get_int64 (src, snapbrckord, &brick_order);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to get brick order");
+                                goto out;
+                        }
+
+                        /* Adding the data in the dst dict */
+                        snprintf (key, sizeof(key) - 1,
+                                  "vol%ld.brickdir%ld", i+1, brick_order);
+
+                        tmpstr = gf_strdup (snap_brick_dir);
+                        if (!tmpstr) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Out Of Memory");
+                                ret = -1;
+                                goto out;
+                        }
+                        ret = dict_set_dynstr (dst, key, tmpstr);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set %s", key);
+                                GF_FREE (tmpstr);
+                                goto out;
+                        }
+
+                        snprintf (key, sizeof(key) - 1,
+                                  "vol%ld.brick_snapdevice%ld",
+                                  i+1, brick_order);
+
+                        tmpstr = gf_strdup (snap_device);
+                        if (!tmpstr) {
+                                ret = -1;
+                                goto out;
+                        }
+                        ret = dict_set_dynstr (dst, key, tmpstr);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set %s", key);
+                                GF_FREE (tmpstr);
+                                goto out;
+                        }
+
+                }
+        }
+
+        ret = 0;
+out:
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_snap_pre_validate_use_rsp_dict (dict_t *dst, dict_t *src)
+{
+        int             ret             = -1;
+        int32_t         snap_command    = 0;
+        xlator_t       *this            = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        if (!dst || !src) {
+                gf_log (this->name, GF_LOG_ERROR, "Source or Destination "
+                        "dict is empty.");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dst, "type", &snap_command);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "unable to get the type of "
+                        "the snapshot command");
+                goto out;
+        }
+
+        switch (snap_command) {
+        case GF_SNAP_OPTION_TYPE_CREATE:
+                ret = glusterd_snap_create_pre_val_use_rsp_dict (dst, src);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Unable to use "
+                                "rsp dict");
+                        goto out;
+                }
+                break;
+        default:
+                break;
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_snapshot_create_prevalidate (dict_t *dict, char **op_errstr,
+                                      dict_t *rsp_dict)
+{
+        char                  *volname                   = NULL;
+        char                  *snapname                  = NULL;
+        char                  *device                    = NULL;
+        char                  *tmpstr                    = NULL;
+        char                  *brick_dir                 = NULL;
+        char                   snap_brick_dir[PATH_MAX]  = "";
+        char                  *mnt_pt                    = NULL;
+        char                   key[PATH_MAX]             = "";
+        char                   snap_mount[PATH_MAX]      = "";
+        char                   snap_volname[64]          = "";
+        char                   err_str[PATH_MAX]         = "";
+        int                    ret                       = -1;
+        int64_t                i                         = 0;
+        int64_t                volcount                  = 0;
+        int64_t                brick_count               = 0;
+        int64_t                brick_order               = 0;
+        glusterd_brickinfo_t  *brickinfo                 = NULL;
+        glusterd_volinfo_t    *volinfo                   = NULL;
+        xlator_t              *this                      = NULL;
+        uuid_t                *snap_volid                = NULL;
+        gf_loglevel_t          loglevel                  = GF_LOG_ERROR;
+        glusterd_conf_t       *conf                      = NULL;
+        int64_t                effective_max_limit       = 0;
+
+        this = THIS;
+        GF_ASSERT (op_errstr);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        ret = dict_get_int64 (dict, "volcount", &volcount);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Failed to "
+                          "get the volume count");
+                goto out;
+        }
+        if (volcount <= 0) {
+                snprintf (err_str, sizeof (err_str), "Invalid volume count %ld "
+                          "supplied", volcount);
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "snapname", &snapname);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Failed to get snapname");
+                goto out;
+        }
+
+        if (glusterd_find_snap_by_name (snapname)) {
+                ret = -1;
+                snprintf (err_str, sizeof (err_str), "Snap %s already exists",
+                          snapname);
+                goto out;
+        }
+
+        for (i = 1; i <= volcount; i++) {
+                snprintf (key, sizeof (key), "volname%ld", i);
+                ret = dict_get_str (dict, key, &volname);
+                if (ret) {
+                        snprintf (err_str, sizeof (err_str),
+                                  "failed to get volume name");
+                        goto out;
+                }
+                ret = glusterd_volinfo_find (volname, &volinfo);
+                if (ret) {
+                        snprintf (err_str, sizeof (err_str),
+                                  "Volume (%s) does not exist ", volname);
+                        goto out;
+                }
+
+                ret = -1;
+                if (!glusterd_is_volume_started (volinfo)) {
+                        snprintf (err_str, sizeof (err_str), "volume %s is "
+                                  "not started", volinfo->volname);
+                        loglevel = GF_LOG_WARNING;
+                        goto out;
+                }
+                if (glusterd_is_defrag_on (volinfo)) {
+                        snprintf (err_str, sizeof (err_str),
+                                  "rebalance process is running for the "
+                                  "volume %s", volname);
+                        loglevel = GF_LOG_WARNING;
+                        goto out;
+                }
+                /* TODO: Also check whether geo replication is running */
+
+                if (volinfo->is_snap_volume == _gf_true) {
+                        snprintf (err_str, sizeof (err_str),
+                                  "Volume %s is a snap volume", volname);
+                        loglevel = GF_LOG_WARNING;
+                        goto out;
+                }
+
+                if (volinfo->snap_max_hard_limit < conf->snap_max_hard_limit)
+                        effective_max_limit = volinfo->snap_max_hard_limit;
+                else
+                        effective_max_limit = conf->snap_max_hard_limit;
+
+                if (volinfo->snap_count >= effective_max_limit) {
+                        snprintf (err_str, sizeof (err_str),
+                                  "The number of existing snaps has reached "
+                                  "the effective maximum limit of %"PRIu64" ,"
+                                  "for the volume %s", effective_max_limit,
+                                  volname);
+                        loglevel = GF_LOG_WARNING;
+                        goto out;
+                }
+
+                snprintf (key, sizeof(key) - 1, "vol%ld_volid", i);
+                ret = dict_get_bin (dict, key, (void **)&snap_volid);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to fetch snap_volid");
+                        goto out;
+                }
+
+                /* snap volume uuid is used as lvm snapshot name.
+                   This will avoid restrictions on snapshot names
+                   provided by user */
+                GLUSTERD_GET_UUID_NOHYPHEN (snap_volname, *snap_volid);
+
+                brick_count = 0;
+                brick_order = 0;
+                /* Adding snap bricks mount paths to the dict */
+                list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                        if (uuid_compare (brickinfo->uuid, MY_UUID)) {
+                                brick_order++;
+                                continue;
+                        }
+
+                        if (!glusterd_is_brick_started (brickinfo)) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "brick %s:%s is not started",
+                                        brickinfo->hostname,
+                                        brickinfo->path);
+                                brick_order++;
+                                brick_count++;
+                                continue;
+                        }
+
+                        device = glusterd_get_brick_mount_details (brickinfo);
+                        if (!device) {
+                                snprintf (err_str, sizeof (err_str),
+                                          "getting device name for the brick "
+                                          "%s:%s failed", brickinfo->hostname,
+                                          brickinfo->path);
+                                ret = -1;
+                                goto out;
+                        }
+
+                        device = glusterd_build_snap_device_path (device,
+                                                                  snap_volname);
+                        if (!device) {
+                                snprintf (err_str, sizeof (err_str),
+                                          "cannot copy the snapshot device "
+                                          "name (volname: %s, snapname: %s)",
+                                          volinfo->volname, snapname);
+                                loglevel = GF_LOG_WARNING;
+                                ret = -1;
+                                goto out;
+                        }
+
+                        snprintf (key, sizeof(key),
+                                  "vol%ld.brick_snapdevice%ld", i,
+                                  brick_count);
+                        ret = dict_set_dynstr (rsp_dict, key, device);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set %s", key);
+                                GF_FREE (device);
+                                goto out;
+                        }
+
+                        ret = glusterd_get_brick_root (brickinfo->path,
+                                                       &mnt_pt);
+                        if (ret) {
+                                snprintf (err_str, sizeof (err_str),
+                                "could not get the root of the brick path %s",
+                                brickinfo->path);
+                                loglevel = GF_LOG_WARNING;
+                                goto out;
+                        }
+                        if (strncmp (brickinfo->path, mnt_pt, strlen(mnt_pt))) {
+                                snprintf (err_str, sizeof (err_str),
+                                "brick: %s brick mount: %s",
+                                 brickinfo->path, mnt_pt);
+                                loglevel = GF_LOG_WARNING;
+                                goto out;
+                        }
+
+                        brick_dir = &brickinfo->path[strlen (mnt_pt)];
+                        brick_dir++;
+
+                        snprintf (snap_brick_dir, sizeof (snap_brick_dir),
+                                  "/%s", brick_dir);
+
+                        tmpstr = gf_strdup (snap_brick_dir);
+                        if (!tmpstr) {
+                                ret = -1;
+                                goto out;
+                        }
+                        snprintf (key, sizeof(key), "vol%ld.brickdir%ld", i,
+                                  brick_count);
+                        ret = dict_set_dynstr (rsp_dict, key, tmpstr);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set %s", snap_mount);
+                                goto out;
+                        }
+                        tmpstr = NULL;
+
+                        snprintf (key, sizeof(key) - 1, "vol%ld.brick%ld.order",
+                                  i, brick_count);
+                        ret = dict_set_int64 (rsp_dict, key, brick_order);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set %s", key);
+                                goto out;
+                        }
+
+                        brick_count++;
+                        brick_order++;
+                }
+                snprintf (key, sizeof(key) - 1, "vol%ld_brickcount", i);
+                ret = dict_set_int64 (rsp_dict, key, brick_count);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to set %s",
+                                key);
+                        goto out;
+                }
+        }
+
+        ret = dict_set_int64 (rsp_dict, "volcount", volcount);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set volcount");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if (ret)
+                GF_FREE (tmpstr);
+
+        if (ret && err_str[0] != '\0') {
+                gf_log (this->name, loglevel, "%s", err_str);
+                *op_errstr = gf_strdup (err_str);
+        }
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+glusterd_snap_t*
+glusterd_new_snap_object()
+{
+        glusterd_snap_t        *snap    = NULL;
+
+        snap = GF_CALLOC (1, sizeof (*snap), gf_gld_mt_snap_t);
+
+        if (snap) {
+                if (LOCK_INIT (&snap->lock)) {
+                        gf_log (THIS->name, GF_LOG_ERROR, "Failed initiating"
+                                " snap lock");
+                        GF_FREE (snap);
+                        return NULL;
+                }
+
+                INIT_LIST_HEAD (&snap->snap_list);
+                INIT_LIST_HEAD (&snap->volumes);
+                snap->snapname[0] = 0;
+                snap->snap_status = GD_SNAP_STATUS_INIT;
+        }
+
+        return snap;
+
+};
+
+/* Function glusterd_list_add_snapvol adds the volinfo object (snapshot volume)
+   to the snapshot object list and to the parent volume list */
+int32_t
+glusterd_list_add_snapvol (glusterd_volinfo_t *origin_vol,
+                           glusterd_volinfo_t *snap_vol)
+{
+        int ret               = -1;
+        glusterd_snap_t *snap = NULL;
+
+        GF_VALIDATE_OR_GOTO ("glusterd", origin_vol, out);
+        GF_VALIDATE_OR_GOTO ("glusterd", snap_vol, out);
+
+        snap = snap_vol->snapshot;
+        GF_ASSERT (snap);
+
+        list_add_tail (&snap_vol->vol_list, &snap->volumes);
+        LOCK (&origin_vol->lock);
+        {
+                list_add_order (&snap_vol->snapvol_list,
+                               &origin_vol->snap_volumes,
+                               glusterd_compare_snap_vol_time);
+                origin_vol->snap_count++;
+        }
+        UNLOCK (&origin_vol->lock);
+
+        gf_log (THIS->name, GF_LOG_DEBUG, "Snap %s added to the list",
+                snap->snapname);
+        ret = 0;
+out:
+        return ret;
+}
+
+glusterd_snap_t*
+glusterd_find_snap_by_name (char *snapname)
+{
+        glusterd_snap_t *snap  = NULL;
+        glusterd_conf_t *priv  = NULL;
+
+        priv = THIS->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (snapname);
+
+
+        list_for_each_entry (snap, &priv->snapshots, snap_list) {
+                if (!strcmp (snap->snapname, snapname)) {
+                        gf_log (THIS->name, GF_LOG_DEBUG, "Found "
+                                "snap %s (%s)", snap->snapname,
+                                uuid_utoa (snap->snap_id));
+                        goto out;
+                }
+        }
+        snap = NULL;
+out:
+        return snap;
+}
+
+glusterd_snap_t*
+glusterd_find_snap_by_id (uuid_t snap_id)
+{
+        glusterd_snap_t *snap  = NULL;
+        glusterd_conf_t *priv  = NULL;
+
+        priv = THIS->private;
+        GF_ASSERT (priv);
+
+        if (uuid_is_null(snap_id))
+                goto out;
+
+        list_for_each_entry (snap, &priv->snapshots, snap_list) {
+                if (!uuid_compare (snap->snap_id, snap_id)) {
+                        gf_log (THIS->name, GF_LOG_DEBUG, "Found "
+                                "snap %s (%s)", snap->snapname,
+                                uuid_utoa (snap->snap_id));
+                        goto out;
+                }
+        }
+        snap = NULL;
+out:
+        return snap;
+}
+
+int
+glusterd_do_lvm_snapshot_remove (glusterd_volinfo_t *snap_vol,
+                                 glusterd_brickinfo_t *brickinfo,
+                                 const char *mount_pt, const char *snap_device)
+{
+        int                     ret               = -1;
+        xlator_t               *this              = NULL;
+        glusterd_conf_t        *priv              = NULL;
+        runner_t                runner            = {0,};
+        char                    msg[1024]         = {0, };
+        char                    pidfile[PATH_MAX] = {0, };
+        pid_t                   pid               = -1;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        if (!brickinfo) {
+                gf_log (this->name, GF_LOG_ERROR, "brickinfo NULL");
+                goto out;
+        }
+
+        GF_ASSERT (snap_vol);
+        GF_ASSERT (mount_pt);
+        GF_ASSERT (snap_device);
+
+        GLUSTERD_GET_BRICK_PIDFILE (pidfile, snap_vol, brickinfo, priv);
+        if (glusterd_is_service_running (pidfile, &pid)) {
+                ret = kill (pid, SIGKILL);
+                if (ret && errno != ESRCH) {
+                        gf_log (this->name, GF_LOG_ERROR, "Unable to kill pid "
+                                "%d reason : %s", pid, strerror(errno));
+                        goto out;
+                }
+        }
+
+        runinit (&runner);
+        snprintf (msg, sizeof (msg), "umount the snapshot mounted path %s",
+                  mount_pt);
+        runner_add_args (&runner, "umount", mount_pt, NULL);
+        runner_log (&runner, "", GF_LOG_DEBUG, msg);
+
+        /* We need not do synclock_unlock => runner_run => synclock_lock here.
+           Because it is needed if we are running a glusterfs process in
+           runner_run, so that when the glusterfs process started wants to
+           communicate to glusterd, glusterd wont be able to respond if it
+           has held the big lock. So we do unlock, run glusterfs process
+           (thus communicate to glusterd), lock. But since this is not a
+           glusterfs command that is being run, unlocking and then relocking
+           is not needed.
+        */
+        ret = runner_run (&runner);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "unmounting the "
+                        "path %s (brick: %s) failed (%s)", mount_pt,
+                        brickinfo->path, strerror (errno));
+                goto out;
+        }
+
+        runinit (&runner);
+        snprintf (msg, sizeof(msg), "remove snapshot of the brick %s:%s, "
+                  "device: %s", brickinfo->hostname, brickinfo->path,
+                  snap_device);
+        runner_add_args (&runner, "/sbin/lvremove", "-f", snap_device, NULL);
+        runner_log (&runner, "", GF_LOG_DEBUG, msg);
+
+        ret = runner_run (&runner);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "removing snapshot of the "
+                        "brick (%s:%s) of device %s failed",
+                        brickinfo->hostname, brickinfo->path, snap_device);
+                goto out;
+        }
+
+out:
+        return ret;
+}
+
+int32_t
+glusterd_lvm_snapshot_remove (dict_t *rsp_dict, glusterd_volinfo_t *snap_vol)
+{
+        char                 *mnt_pt        = NULL;
+        struct mntent        *entry         = NULL;
+        int32_t               brick_count   = -1;
+        int32_t               ret           = -1;
+        glusterd_brickinfo_t *brickinfo     = NULL;
+        xlator_t             *this          = NULL;
+        FILE                 *mtab          = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (snap_vol);
+
+        if (!snap_vol) {
+                gf_log (this->name, GF_LOG_ERROR, "snap volinfo is NULL");
+                goto out;
+        }
+
+        brick_count = -1;
+        list_for_each_entry (brickinfo, &snap_vol->bricks, brick_list) {
+                brick_count++;
+                if (uuid_compare (brickinfo->uuid, MY_UUID))
+                        continue;
+
+                if (brickinfo->snap_status == -1) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "snapshot was pending. lvm not present "
+                                "for brick %s:%s of the snap %s.",
+                                brickinfo->hostname, brickinfo->path,
+                                snap_vol->snapshot->snapname);
+
+                        /* Adding missed delete to the dict */
+                        ret = glusterd_add_missed_snaps_to_dict
+                                                   (rsp_dict,
+                                                    snap_vol->volname,
+                                                    brickinfo,
+                                                    brick_count + 1,
+                                                    GF_SNAP_OPTION_TYPE_DELETE);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to add missed snapshot info "
+                                        "for %s:%s in the rsp_dict",
+                                        brickinfo->hostname,
+                                        brickinfo->path);
+                                goto out;
+                        }
+
+                        continue;
+                }
+
+                ret = glusterd_get_brick_root (brickinfo->path, &mnt_pt);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "getting the root "
+                               "of the brick for volume %s (snap %s) failed ",
+                               snap_vol->volname, snap_vol->snapshot->snapname);
+                        goto out;
+                }
+
+                entry = glusterd_get_mnt_entry_info (mnt_pt, mtab);
+                if (!entry) {
+                        gf_log (this->name, GF_LOG_WARNING, "getting the mount"
+                                " entry for the brick %s:%s of the snap %s "
+                                "(volume: %s) failed", brickinfo->hostname,
+                                brickinfo->path, snap_vol->snapshot->snapname,
+                                snap_vol->volname);
+                        ret = -1;
+                        goto out;
+                }
+                ret = glusterd_do_lvm_snapshot_remove (snap_vol, brickinfo,
+                                                       mnt_pt,
+                                                       entry->mnt_fsname);
+                if (mtab)
+                        endmntent (mtab);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                "remove the snapshot %s (%s)",
+                                brickinfo->path, entry->mnt_fsname);
+                        goto out;
+                }
+
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+glusterd_snap_volume_remove (dict_t *rsp_dict,
+                             glusterd_volinfo_t *snap_vol,
+                             gf_boolean_t remove_lvm,
+                             gf_boolean_t force)
+{
+        int                   ret         = -1;
+        int                   save_ret    =  0;
+        glusterd_brickinfo_t *brickinfo   = NULL;
+        glusterd_volinfo_t   *origin_vol  = NULL;
+        xlator_t             *this        = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (snap_vol);
+
+        if (!snap_vol) {
+                gf_log(this->name, GF_LOG_WARNING, "snap_vol in NULL");
+                ret = -1;
+                goto out;
+        }
+
+        list_for_each_entry (brickinfo, &snap_vol->bricks, brick_list) {
+                if (uuid_compare (brickinfo->uuid, MY_UUID))
+                        continue;
+
+                ret = glusterd_brick_stop (snap_vol, brickinfo, _gf_false);
+                if (ret) {
+                        gf_log(this->name, GF_LOG_WARNING, "Failed to stop "
+                               "brick for volume %s", snap_vol->volname);
+                        save_ret = ret;
+
+                        /* Continue to cleaning up the snap in case of error
+                           if force flag is enabled */
+                        if (!force)
+                                goto out;
+                }
+        }
+
+        /* Only remove the backend lvm when required */
+        if (remove_lvm) {
+                ret = glusterd_lvm_snapshot_remove (rsp_dict, snap_vol);
+                if (ret) {
+                        gf_log(this->name, GF_LOG_WARNING, "Failed to remove "
+                               "lvm snapshot volume %s", snap_vol->volname);
+                        save_ret = ret;
+                        if (!force)
+                                goto out;
+                }
+        }
+
+        ret = glusterd_store_delete_volume (snap_vol);
+        if (ret) {
+                gf_log(this->name, GF_LOG_WARNING, "Failed to remove volume %s "
+                       "from store", snap_vol->volname);
+                save_ret = ret;
+                if (!force)
+                        goto out;
+        }
+
+        if (!list_empty(&snap_vol->snapvol_list)) {
+                ret = glusterd_volinfo_find (snap_vol->parent_volname,
+                                             &origin_vol);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                "parent volinfo %s  for volume  %s",
+                                snap_vol->parent_volname, snap_vol->volname);
+                        save_ret = ret;
+                        if (!force)
+                                goto out;
+                }
+                origin_vol->snap_count--;
+        }
+
+        ret = glusterd_volinfo_delete (snap_vol);
+        if (ret) {
+                gf_log(this->name, GF_LOG_WARNING, "Failed to remove volinfo "
+                       "%s ", snap_vol->volname);
+                save_ret = ret;
+                if (!force)
+                        goto out;
+        }
+
+        if (save_ret)
+                ret = save_ret;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_snapobject_delete (glusterd_snap_t *snap)
+{
+        if (snap == NULL) {
+                gf_log(THIS->name, GF_LOG_WARNING, "snap is NULL");
+                return -1;
+        }
+
+        list_del_init (&snap->snap_list);
+        list_del_init (&snap->volumes);
+        if (LOCK_DESTROY(&snap->lock))
+                gf_log (THIS->name, GF_LOG_WARNING, "Failed destroying lock"
+                        "of snap %s", snap->snapname);
+
+        GF_FREE (snap->description);
+        GF_FREE (snap);
+
+        return 0;
+}
+
+int32_t
+glusterd_snap_remove (dict_t *rsp_dict,
+                      glusterd_snap_t *snap,
+                      gf_boolean_t remove_lvm,
+                      gf_boolean_t force)
+{
+        int                 ret       = -1;
+        int                 save_ret  =  0;
+        glusterd_volinfo_t *snap_vol  = NULL;
+        glusterd_volinfo_t *tmp       = NULL;
+        xlator_t           *this      = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (snap);
+
+        if (!snap) {
+                gf_log(this->name, GF_LOG_WARNING, "snap is NULL");
+                ret = -1;
+                goto out;
+        }
+
+        list_for_each_entry_safe (snap_vol, tmp, &snap->volumes, vol_list) {
+                ret = glusterd_snap_volume_remove (rsp_dict, snap_vol,
+                                                   remove_lvm, force);
+                if (ret) {
+                        gf_log(this->name, GF_LOG_WARNING, "Failed to remove "
+                               "volinfo %s for snap %s", snap_vol->volname,
+                               snap->snapname);
+                        save_ret = ret;
+
+                        /* Continue to cleaning up the snap in case of error
+                           if force flag is enabled */
+                        if (!force)
+                                goto out;
+                }
+        }
+
+        ret = glusterd_store_delete_snap (snap);
+        if (ret) {
+                gf_log(this->name, GF_LOG_WARNING, "Failed to remove snap %s "
+                       "from store", snap->snapname);
+                save_ret = ret;
+                if (!force)
+                        goto out;
+        }
+
+        ret = glusterd_snapobject_delete (snap);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING, "Failed to delete "
+                        "snap object %s", snap->snapname);
+
+        if (save_ret)
+                ret = save_ret;
+out:
+        gf_log (THIS->name, GF_LOG_TRACE, "returning %d", ret);
+        return ret;
+}
+
+static int
+glusterd_snapshot_get_snapvol_detail (dict_t *dict,
+                                      glusterd_volinfo_t *snap_vol,
+                                      char *keyprefix, int detail)
+{
+        int                 ret           = -1;
+        int                 snap_limit    = 0;
+        char                key[PATH_MAX] = {0,};
+        char               *value         = NULL;
+        glusterd_volinfo_t *origin_vol    = NULL;
+        glusterd_conf_t    *conf          = NULL;
+        xlator_t           *this          = NULL;
+
+        this = THIS;
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        GF_ASSERT (dict);
+        GF_ASSERT (snap_vol);
+        GF_ASSERT (keyprefix);
+
+        /* Volume Name */
+        value = gf_strdup (snap_vol->volname);
+        if (!value)
+                goto out;
+
+        snprintf (key, sizeof (key), "%s.volname", keyprefix);
+        ret = dict_set_dynstr (dict, key, value);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+                        "volume name in dictionary: %s", key);
+                goto out;
+        }
+
+        /* Volume ID */
+        value = gf_strdup (uuid_utoa (snap_vol->volume_id));
+        if (NULL == value) {
+                ret = -1;
+                goto out;
+        }
+
+        snprintf (key, sizeof (key), "%s.vol-id", keyprefix);
+        ret = dict_set_dynstr (dict, key, value);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+                        "volume id in dictionary: %s", key);
+                goto out;
+        }
+        value = NULL;
+
+        /* volume status */
+        snprintf (key, sizeof (key), "%s.vol-status", keyprefix);
+        switch (snap_vol->status) {
+        case GLUSTERD_STATUS_STARTED:
+                ret = dict_set_str (dict, key, "Started");
+                break;
+        case GLUSTERD_STATUS_STOPPED:
+                ret = dict_set_str (dict, key, "Stopped");
+                break;
+        case GD_SNAP_STATUS_NONE:
+                ret = dict_set_str (dict, key, "None");
+                break;
+        default:
+                gf_log (this->name, GF_LOG_ERROR, "Invalid volume status");
+                ret = -1;
+                goto out;
+        }
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set volume status"
+                        " in dictionary: %s", key);
+                goto out;
+        }
+
+
+        ret = glusterd_volinfo_find (snap_vol->parent_volname, &origin_vol);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to get the parent "
+                        "volinfo for the volume %s", snap_vol->volname);
+                goto out;
+        }
+
+        /* Snaps available */
+        if (conf->snap_max_hard_limit < origin_vol->snap_max_hard_limit) {
+                snap_limit = conf->snap_max_hard_limit;
+                gf_log(this->name, GF_LOG_DEBUG, "system snap-max-hard-limit is"
+                       " lesser than volume snap-max-hard-limit, "
+                       "snap-max-hard-limit value is set to %d", snap_limit);
+        } else {
+                snap_limit = origin_vol->snap_max_hard_limit;
+                gf_log(this->name, GF_LOG_DEBUG, "volume snap-max-hard-limit is"
+                       " lesser than system snap-max-hard-limit, "
+                       "snap-max-hard-limit value is set to %d", snap_limit);
+        }
+
+        snprintf (key, sizeof (key), "%s.snaps-available", keyprefix);
+        if (snap_limit > origin_vol->snap_count)
+                ret = dict_set_int32 (dict, key,
+                        snap_limit - origin_vol->snap_count);
+        else
+                ret = dict_set_int32 (dict, key, 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set available snaps");
+                goto out;
+        }
+
+        snprintf (key, sizeof (key), "%s.snapcount", keyprefix);
+        ret = dict_set_int32 (dict, key, origin_vol->snap_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not save snapcount");
+                goto out;
+        }
+
+        if (!detail)
+                goto out;
+
+        /* Parent volume name */
+        value = gf_strdup (snap_vol->parent_volname);
+        if (!value)
+                goto out;
+
+        snprintf (key, sizeof (key), "%s.origin-volname", keyprefix);
+        ret = dict_set_dynstr (dict, key, value);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set parent "
+                        "volume name in dictionary: %s", key);
+                goto out;
+        }
+        value = NULL;
+
+        ret = 0;
+out:
+        if (value)
+                GF_FREE (value);
+
+        return ret;
+}
+
+static int
+glusterd_snapshot_get_snap_detail (dict_t *dict, glusterd_snap_t *snap,
+                                char *keyprefix, glusterd_volinfo_t *volinfo)
+{
+        int                 ret           = -1;
+        int                 volcount      = 0;
+        char                key[PATH_MAX] = {0,};
+        char               *value         = NULL;
+        char               *timestr       = NULL;
+        struct tm          *tmptr         = NULL;
+        glusterd_volinfo_t *snap_vol      = NULL;
+        glusterd_volinfo_t *tmp_vol       = NULL;
+        xlator_t           *this          = NULL;
+
+        this = THIS;
+
+        GF_ASSERT (dict);
+        GF_ASSERT (snap);
+        GF_ASSERT (keyprefix);
+
+        /* Snap Name */
+        value = gf_strdup (snap->snapname);
+        if (!value)
+                goto out;
+
+        snprintf (key, sizeof (key), "%s.snapname", keyprefix);
+        ret = dict_set_dynstr (dict, key, value);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+                        "snap name in dictionary");
+                goto out;
+        }
+
+        /* Snap ID */
+        value = gf_strdup (uuid_utoa (snap->snap_id));
+        if (NULL == value) {
+                ret = -1;
+                goto out;
+        }
+
+        snprintf (key, sizeof (key), "%s.snap-id", keyprefix);
+        ret = dict_set_dynstr (dict, key, value);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+                                "snap id in dictionary");
+                goto out;
+        }
+        value = NULL;
+
+        tmptr = localtime (&(snap->time_stamp));
+        if (NULL == tmptr) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to convert "
+                                "time_t to *tm");
+                ret = -1;
+                goto out;
+        }
+
+        timestr = GF_CALLOC (1, PATH_MAX, gf_gld_mt_char);
+        if (NULL == timestr) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = strftime (timestr, PATH_MAX, "%Y-%m-%d %H:%M:%S", tmptr);
+        if (0 == ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to convert time_t "
+                        "to string");
+                ret = -1;
+                goto out;
+        }
+
+        snprintf (key, sizeof (key), "%s.snap-time", keyprefix);
+        ret = dict_set_dynstr (dict, key, timestr);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+                                "snap time stamp in dictionary");
+                goto out;
+        }
+        timestr = NULL;
+
+        /* If snap description is provided then add that into dictionary */
+        if (NULL != snap->description) {
+                value = gf_strdup (snap->description);
+                if (NULL == value) {
+                        ret = -1;
+                        goto out;
+                }
+
+                snprintf (key, sizeof (key), "%s.snap-desc", keyprefix);
+                ret = dict_set_dynstr (dict, key, value);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+                                        "snap description in dictionary");
+                        goto out;
+                }
+                value = NULL;
+        }
+
+        snprintf (key, sizeof (key), "%s.snap-status", keyprefix);
+        switch (snap->snap_status) {
+        case GD_SNAP_STATUS_INIT:
+                ret = dict_set_str (dict, key, "Init");
+                break;
+        case GD_SNAP_STATUS_IN_USE:
+                ret = dict_set_str (dict, key, "In-use");
+                break;
+        case GD_SNAP_STATUS_DECOMMISSION:
+                ret = dict_set_str (dict, key, "Decommisioned");
+                break;
+        case GD_SNAP_STATUS_RESTORED:
+                ret = dict_set_str (dict, key, "Restored");
+                break;
+        case GD_SNAP_STATUS_NONE:
+                ret = dict_set_str (dict, key, "None");
+                break;
+        default:
+                gf_log (this->name, GF_LOG_ERROR, "Invalid snap status");
+                ret = -1;
+                goto out;
+        }
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set snap status "
+                        "in dictionary");
+                goto out;
+        }
+
+        if (volinfo) {
+                volcount = 1;
+                snprintf (key, sizeof (key), "%s.vol%d", keyprefix, volcount);
+                ret = glusterd_snapshot_get_snapvol_detail (dict,
+                                                        volinfo, key, 0);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                "get volume detail %s for snap %s",
+                                snap_vol->volname, snap->snapname);
+                        goto out;
+                }
+                goto done;
+        }
+
+        list_for_each_entry_safe (snap_vol, tmp_vol, &snap->volumes, vol_list) {
+                volcount++;
+                snprintf (key, sizeof (key), "%s.vol%d", keyprefix, volcount);
+                ret = glusterd_snapshot_get_snapvol_detail (dict,
+                                                        snap_vol, key, 1);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                "get volume detail %s for snap %s",
+                                snap_vol->volname, snap->snapname);
+                        goto out;
+                }
+        }
+
+done:
+        snprintf (key, sizeof (key), "%s.vol-count", keyprefix);
+        ret = dict_set_int32 (dict, key, volcount);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set %s",
+                        key);
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if (value)
+                GF_FREE (value);
+
+        if (timestr)
+                GF_FREE(timestr);
+
+        return ret;
+}
+
+static int
+glusterd_snapshot_get_all_snap_info (dict_t *dict)
+{
+        int                 ret           = -1;
+        int                 snapcount     = 0;
+        char                key[PATH_MAX] = {0,};
+        glusterd_snap_t    *snap          = NULL;
+        glusterd_snap_t    *tmp_snap      = NULL;
+        glusterd_conf_t    *priv          = NULL;
+        xlator_t           *this          = NULL;
+
+        this = THIS;
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        /* General parameter validation */
+        GF_ASSERT (dict);
+
+        list_for_each_entry_safe (snap, tmp_snap, &priv->snapshots, snap_list) {
+                snapcount++;
+                snprintf (key, sizeof (key), "snap%d", snapcount);
+                ret = glusterd_snapshot_get_snap_detail (dict, snap, key, NULL);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                "snapdetail for snap %s", snap->snapname);
+                        goto out;
+                }
+        }
+
+        ret = dict_set_int32 (dict, "snap-count", snapcount);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_snapshot_get_info_by_volume (dict_t *dict, char *volname,
+                                      char *err_str, size_t len)
+{
+        int                  ret           = -1;
+        int                  snapcount     = 0;
+        int                  snap_limit    = 0;
+        char                *value         = NULL;
+        char                 key[PATH_MAX] = "";
+        glusterd_volinfo_t  *volinfo       = NULL;
+        glusterd_volinfo_t  *snap_vol      = NULL;
+        glusterd_volinfo_t  *tmp_vol       = NULL;
+        glusterd_conf_t     *conf          = NULL;
+        xlator_t            *this          = NULL;
+
+        this = THIS;
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        GF_ASSERT (dict);
+        GF_ASSERT (volname);
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (err_str, len, "Volume (%s) does not exist", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        /* Snaps available */
+        if (conf->snap_max_hard_limit < volinfo->snap_max_hard_limit) {
+                snap_limit = conf->snap_max_hard_limit;
+                gf_log(this->name, GF_LOG_DEBUG, "system snap-max-hard-limit is"
+                       " lesser than volume snap-max-hard-limit, "
+                       "snap-max-hard-limit value is set to %d", snap_limit);
+        } else {
+                snap_limit = volinfo->snap_max_hard_limit;
+                gf_log(this->name, GF_LOG_DEBUG, "volume snap-max-hard-limit is"
+                       " lesser than system snap-max-hard-limit, "
+                       "snap-max-hard-limit value is set to %d", snap_limit);
+        }
+
+        if (snap_limit > volinfo->snap_count)
+                ret = dict_set_int32 (dict, "snaps-available",
+                        snap_limit - volinfo->snap_count);
+        else
+                ret = dict_set_int32 (dict, "snaps-available", 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set available snaps");
+                goto out;
+        }
+
+        /* Origin volume name */
+        value = gf_strdup (volinfo->volname);
+        if (!value)
+                goto out;
+
+        ret = dict_set_dynstr (dict, "origin-volname", value);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set parent "
+                        "volume name in dictionary: %s", key);
+                goto out;
+        }
+        value = NULL;
+
+        list_for_each_entry_safe (snap_vol, tmp_vol, &volinfo->snap_volumes,
+                                  snapvol_list) {
+                snapcount++;
+                snprintf (key, sizeof (key), "snap%d", snapcount);
+                ret = glusterd_snapshot_get_snap_detail (dict,
+                                                         snap_vol->snapshot,
+                                                         key, snap_vol);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                "snapdetail for snap %s",
+                                snap_vol->snapshot->snapname);
+                        goto out;
+                }
+        }
+        ret = dict_set_int32 (dict, "snap-count", snapcount);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if (value)
+                GF_FREE (value);
+
+        return ret;
+}
+
+/* This function will be called from RPC handler routine.
+ * This function is responsible for getting the requested
+ * snapshot info into the dictionary.
+ *
+ * @param req   RPC request object. Required for sending a response back.
+ * @param op    glusterd operation. Required for sending a response back.
+ * @param dict  pointer to dictionary which will contain both
+ *              request and response key-pair values.
+ * @return -1 on error and 0 on success
+ */
+int
+glusterd_handle_snapshot_info (rpcsvc_request_t *req, glusterd_op_t op,
+                               dict_t *dict, char *err_str, size_t len)
+{
+        int              ret             =       -1;
+        int8_t          snap_driven      =        1;
+        char            *volname         =      NULL;
+        char            *snapname        =      NULL;
+        glusterd_snap_t *snap            =      NULL;
+        xlator_t        *this            =      NULL;
+        int32_t         cmd              =      GF_SNAP_INFO_TYPE_ALL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        GF_VALIDATE_OR_GOTO (this->name, req, out);
+        GF_VALIDATE_OR_GOTO (this->name, dict, out);
+
+
+        ret = dict_get_int32 (dict, "cmd", &cmd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get type "
+                        "of snapshot info");
+                goto out;
+        }
+
+        switch (cmd) {
+                case GF_SNAP_INFO_TYPE_ALL:
+                {
+                        ret = glusterd_snapshot_get_all_snap_info (dict);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to get info of all snaps");
+                                goto out;
+                        }
+                        break;
+                }
+
+                case GF_SNAP_INFO_TYPE_SNAP:
+                {
+                        ret = dict_get_str (dict, "snapname", &snapname);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to get snap name");
+                                goto out;
+                        }
+
+                        ret = dict_set_int32 (dict, "snap-count", 1);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set snapcount");
+                                goto out;
+                        }
+
+                        snap = glusterd_find_snap_by_name (snapname);
+                        if (!snap) {
+                                snprintf (err_str, len,
+                                         "Snap (%s) does not exist", snapname);
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "%s", err_str);
+                                ret = -1;
+                                goto out;
+                        }
+                        ret = glusterd_snapshot_get_snap_detail (dict, snap,
+                                                                "snap1", NULL);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to get snap detail of snap "
+                                        "%s", snap->snapname);
+                                goto out;
+                        }
+                        break;
+                }
+
+                case GF_SNAP_INFO_TYPE_VOL:
+                {
+                        ret = dict_get_str (dict, "volname", &volname);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to get volname");
+                                goto out;
+                        }
+                        ret = glusterd_snapshot_get_info_by_volume (dict,
+                                                        volname, err_str, len);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to get volume info of volume "
+                                        "%s", volname);
+                                goto out;
+                        }
+                        snap_driven = 0;
+                        break;
+                }
+        }
+
+        ret = dict_set_int8 (dict, "snap-driven", snap_driven);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set snap-driven");
+                goto out;
+        }
+
+        /* If everything is successful then send the response back to cli.
+         * In case of failure the caller of this function will take care
+           of the response */
+        ret = glusterd_op_send_cli_response (op, 0, 0, req, dict, err_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to send cli "
+                                "response");
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+/* This function sets all the snapshot names in the dictionary */
+int
+glusterd_snapshot_get_all_snapnames (dict_t *dict)
+{
+        int              ret             = -1;
+        int              snapcount       = 0;
+        char            *snapname        = NULL;
+        char             key[PATH_MAX]   = {0,};
+        glusterd_snap_t *snap            = NULL;
+        glusterd_snap_t *tmp_snap        = NULL;
+        glusterd_conf_t *priv            = NULL;
+        xlator_t        *this            = NULL;
+
+        this = THIS;
+        priv = this->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (dict);
+
+        list_for_each_entry_safe (snap, tmp_snap, &priv->snapshots, snap_list) {
+                snapcount++;
+                snapname = gf_strdup (snap->snapname);
+                if (!snapname) {
+                        gf_log (this->name, GF_LOG_ERROR, "strdup failed");
+                        ret = -1;
+                        goto out;
+                }
+                snprintf (key, sizeof (key), "snapname%d", snapcount);
+                ret = dict_set_dynstr (dict, key, snapname);
+                if (ret) {
+                        GF_FREE (snapname);
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to set %s",
+                                key);
+                        goto out;
+                }
+        }
+
+        ret = dict_set_int32 (dict, "snap-count", snapcount);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount");
+                goto out;
+        }
+
+        ret = 0;
+out:
+
+        return ret;
+}
+
+/* This function sets all the snapshot names
+   under a given volume in the dictionary */
+int
+glusterd_snapshot_get_vol_snapnames (dict_t *dict, glusterd_volinfo_t *volinfo)
+{
+        int                 ret             = -1;
+        int                 snapcount       = 0;
+        char               *snapname        = NULL;
+        char                key[PATH_MAX]   = {0,};
+        glusterd_volinfo_t *snap_vol        = NULL;
+        glusterd_volinfo_t *tmp_vol         = NULL;
+        xlator_t           *this            = NULL;
+
+        this = THIS;
+        GF_ASSERT (dict);
+        GF_ASSERT (volinfo);
+
+        list_for_each_entry_safe (snap_vol, tmp_vol,
+                                  &volinfo->snap_volumes, snapvol_list) {
+                snapcount++;
+                snapname = gf_strdup (snap_vol->snapshot->snapname);
+                if (!snapname) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "strdup failed");
+                        ret = -1;
+                        goto out;
+                }
+                snprintf (key, sizeof (key), "snapname%d", snapcount);
+                ret = dict_set_dynstr (dict, key, snapname);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                "set %s", key);
+                        GF_FREE (snapname);
+                        goto out;
+                }
+        }
+
+        ret = dict_set_int32 (dict, "snap-count", snapcount);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set snapcount");
+                goto out;
+        }
+
+        ret = 0;
+out:
+
+        return ret;
+}
+
+int
+glusterd_handle_snapshot_list (rpcsvc_request_t *req, glusterd_op_t op,
+                               dict_t *dict, char *err_str, size_t len)
+{
+        int                 ret             = -1;
+        char               *volname         = NULL;
+        glusterd_volinfo_t *volinfo         = NULL;
+        xlator_t           *this            = NULL;
+
+        this = THIS;
+
+        GF_VALIDATE_OR_GOTO (this->name, req, out);
+        GF_VALIDATE_OR_GOTO (this->name, dict, out);
+
+        /* Ignore error for getting volname as it is optional */
+        ret = dict_get_str (dict, "volname", &volname);
+
+        if (NULL == volname) {
+                ret = glusterd_snapshot_get_all_snapnames (dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to get snapshot list");
+                        goto out;
+                }
+        } else {
+                ret = glusterd_volinfo_find (volname, &volinfo);
+                if (ret) {
+                        snprintf (err_str, len,
+                                 "Volume (%s) does not exist", volname);
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s", err_str);
+                        goto out;
+                }
+
+                ret = glusterd_snapshot_get_vol_snapnames (dict, volinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to get snapshot list for volume %s",
+                                volname);
+                        goto out;
+                }
+        }
+
+        /* If everything is successful then send the response back to cli.
+        In case of failure the caller of this function will take of response.*/
+        ret = glusterd_op_send_cli_response (op, 0, 0, req, dict, err_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to send cli "
+                                "response");
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+/* This is a snapshot create handler function. This function will be
+ * executed in the originator node. This function is responsible for
+ * calling mgmt_v3 framework to do the actual snap creation on all the bricks
+ *
+ * @param req           RPC request object
+ * @param op            gluster operation
+ * @param dict          dictionary containing snapshot restore request
+ * @param err_str       In case of an err this string should be populated
+ * @param len           length of err_str buffer
+ *
+ * @return              Negative value on Failure and 0 in success
+ */
+int
+glusterd_handle_snapshot_create (rpcsvc_request_t *req, glusterd_op_t op,
+                               dict_t *dict, char *err_str, size_t len)
+{
+        int           ret                     = -1;
+        char         *volname                 = NULL;
+        char         *snapname                = NULL;
+        int64_t       volcount                = 0;
+        xlator_t     *this                    = NULL;
+        char          key[PATH_MAX]           = "";
+        char         *username                = NULL;
+        char         *password                = NULL;
+        uuid_t       *uuid_ptr                = NULL;
+        uuid_t        tmp_uuid                = {0};
+        int           i                       = 0;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+        GF_ASSERT (dict);
+        GF_ASSERT (err_str);
+
+        ret = dict_get_int64 (dict, "volcount", &volcount);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to "
+                        "get the volume count");
+                goto out;
+        }
+        if (volcount <= 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Invalid volume count %ld "
+                        "supplied", volcount);
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "snapname", &snapname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to get the snapname");
+                goto out;
+        }
+
+        if (strlen(snapname) >= GLUSTERD_MAX_SNAP_NAME) {
+                snprintf (err_str, len, "snapname cannot exceed 255 "
+                          "characters");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                ret = -1;
+                goto out;
+        }
+
+        uuid_ptr = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t);
+        if (!uuid_ptr) {
+                gf_log (this->name, GF_LOG_ERROR, "Out Of Memory");
+                ret = -1;
+                goto out;
+        }
+
+        uuid_generate (*uuid_ptr);
+        ret = dict_set_bin (dict, "snap-id", uuid_ptr, sizeof(uuid_t));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to set snap-id");
+                GF_FREE (uuid_ptr);
+                goto out;
+        }
+        uuid_ptr = NULL;
+
+        ret = dict_set_int64 (dict, "snap-time", (int64_t)time(NULL));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to set snap-time");
+                goto out;
+        }
+
+        for (i = 1; i <= volcount; i++) {
+                snprintf (key, sizeof (key), "volname%d", i);
+                ret = dict_get_str (dict, key, &volname);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to get volume name");
+                        goto out;
+                }
+
+                /* generate internal username and password  for the snap*/
+                uuid_generate (tmp_uuid);
+                username = gf_strdup (uuid_utoa (tmp_uuid));
+                snprintf (key, sizeof(key), "volume%d_username", i);
+                ret = dict_set_dynstr (dict, key, username);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to set snap "
+                                "username for volume %s", volname);
+                        GF_FREE (username);
+                        goto out;
+                }
+
+                uuid_generate (tmp_uuid);
+                password = gf_strdup (uuid_utoa (tmp_uuid));
+                snprintf (key, sizeof(key), "volume%d_password", i);
+                ret = dict_set_dynstr (dict, key, password);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to set snap "
+                                "password for volume %s", volname);
+                        GF_FREE (password);
+                        goto out;
+                }
+
+                uuid_ptr = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t);
+                if (!uuid_ptr) {
+                        gf_log (this->name, GF_LOG_ERROR, "Out Of Memory");
+                        ret = -1;
+                        goto out;
+                }
+
+                snprintf (key, sizeof(key) - 1, "vol%d_volid", i);
+                uuid_generate (*uuid_ptr);
+                ret = dict_set_bin (dict, key, uuid_ptr, sizeof(uuid_t));
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to set snap_volid");
+                        GF_FREE (uuid_ptr);
+                        goto out;
+                }
+        }
+
+        ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to initiate snap "
+                                "phases");
+        }
+
+out:
+        return ret;
+}
+
+/* This is a snapshot status handler function. This function will be
+ * executed in a originator node. This function is responsible for
+ * calling mgmt v3 framework to get the actual snapshot status from
+ * all the bricks
+ *
+ * @param req           RPC request object
+ * @param op            gluster operation
+ * @param dict          dictionary containing snapshot status request
+ * @param err_str       In case of an err this string should be populated
+ * @param len           length of err_str buffer
+ *
+ * return :  0  in case of success.
+ *          -1  in case of failure.
+ *
+ */
+int
+glusterd_handle_snapshot_status (rpcsvc_request_t *req, glusterd_op_t op,
+                                 dict_t *dict, char *err_str, size_t len)
+{
+        int                     ret             =       -1;
+        char                    *volname        =       NULL;
+        char                    *snapname       =       NULL;
+        char                    *buf            =       NULL;
+        glusterd_conf_t         *conf           =       NULL;
+        xlator_t                *this           =       NULL;
+        int32_t                 cmd             =       -1;
+        int                     i               =       0;
+        dict_t                  *voldict        =       NULL;
+        char                    key[PATH_MAX]   =       "";
+        glusterd_volinfo_t     *volinfo         =       NULL;
+        glusterd_snap_t         *snap           =       NULL;
+        glusterd_volinfo_t      *snap_volinfo   =       NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+
+        GF_ASSERT (conf);
+        GF_ASSERT (req);
+        GF_ASSERT (dict);
+        GF_ASSERT (err_str);
+
+        ret = dict_get_int32 (dict, "cmd", &cmd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not get status type");
+                goto out;
+        }
+        switch (cmd) {
+                case GF_SNAP_STATUS_TYPE_ALL:
+                {
+                        /* IF we give "gluster snapshot status"
+                         * then lock is held on all snaps.
+                         * This is the place where necessary information
+                         * (snapname and snapcount)is populated in dictionary
+                         * for locking.
+                         */
+                        ++i;
+                        list_for_each_entry (snap, &conf->snapshots, snap_list)
+                        {
+                                snprintf (key, sizeof (key), "snapname%d", i);
+                                buf = gf_strdup (snap->snapname);
+                                if (!buf) {
+                                        ret = -1;
+                                        goto out;
+                                }
+                                ret = dict_set_dynstr (dict, key, buf);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Could not save snapname (%s) "
+                                                "in the dictionary",
+                                                snap->snapname);
+                                        GF_FREE (buf);
+                                        goto out;
+                                }
+
+                                buf = NULL;
+                                i++;
+                        }
+
+                        ret = dict_set_int32 (dict, "snapcount", i - 1);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Could not "
+                                        "save snapcount in the dictionary");
+                                goto out;
+                        }
+                        break;
+                }
+
+                case GF_SNAP_STATUS_TYPE_SNAP:
+                {
+                        /* IF we give "gluster snapshot status <snapname>"
+                         * then lock is held on single snap.
+                         * This is the place where necessary information
+                         * (snapname)is populated in dictionary
+                         * for locking.
+                         */
+                        ret = dict_get_str (dict, "snapname", &snapname);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to fetch snap name");
+                                goto out;
+                        }
+
+                        snap = glusterd_find_snap_by_name (snapname);
+                        if (!snap) {
+                                snprintf (err_str, len, "Snap (%s)"
+                                          "does not exist", snapname);
+                                gf_log(this->name, GF_LOG_ERROR,
+                                        "%s", err_str);
+                                ret = -1;
+                                goto out;
+                        }
+                        break;
+                }
+                case GF_SNAP_STATUS_TYPE_VOL:
+                        ret = dict_get_str (dict, "volname", &volname);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to fetch volname");
+                                goto out;
+                        }
+
+                        ret = glusterd_volinfo_find (volname, &volinfo);
+                        if (ret) {
+                                snprintf (err_str, len, "Volume (%s) "
+                                          "does not exist", volname);
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "%s", err_str);
+                                goto out;
+                        }
+
+                        i = 1;
+                        list_for_each_entry (snap_volinfo,
+                                        &volinfo->snap_volumes, snapvol_list) {
+                                snprintf (key, sizeof (key), "snapname%d", i);
+
+                                buf = gf_strdup
+                                             (snap_volinfo->snapshot->snapname);
+                                if (!buf) {
+                                        ret = -1;
+                                        goto out;
+                                }
+
+                                ret = dict_set_dynstr (dict, key, buf);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Could not save snapname");
+                                        GF_FREE (buf);
+                                        goto out;
+                                }
+
+                                buf = NULL;
+                                i++;
+                        }
+
+                        ret = dict_set_int32 (dict, "snapcount", i-1);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Could not save snapcount");
+                                goto out;
+                        }
+                        break;
+                default:
+                {
+                        gf_log (this->name, GF_LOG_ERROR, "Unknown type");
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        /* Volume lock is not necessary for snapshot status, hence
+         * turning it off
+         */
+        ret = dict_set_int8 (dict, "hold_vol_locks", 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Setting volume lock "
+                        "flag failed");
+                goto out;
+        }
+
+        ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to initiate "
+                        "snap phases");
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        if (voldict) {
+                dict_unref (voldict);
+        }
+        return ret;
+}
+
+
+/* This is a snapshot restore handler function. This function will be
+ * executed in the originator node. This function is responsible for
+ * calling mgmt_v3 framework to do the actual restore on all the bricks
+ *
+ * @param req           RPC request object
+ * @param op            gluster operation
+ * @param dict          dictionary containing snapshot restore request
+ * @param err_str       In case of an err this string should be populated
+ * @param len           length of err_str buffer
+ *
+ * @return              Negative value on Failure and 0 in success
+ */
+int
+glusterd_handle_snapshot_restore (rpcsvc_request_t *req, glusterd_op_t op,
+                               dict_t *dict, char *err_str, size_t len)
+{
+        int                     ret             = -1;
+        char                    *snapname       = NULL;
+        char                    *buf            = NULL;
+        glusterd_conf_t         *conf           = NULL;
+        xlator_t                *this           = NULL;
+        glusterd_snap_t         *snap           = NULL;
+        glusterd_volinfo_t      *snap_volinfo   = NULL;
+        int32_t                 i               =    0;
+        char                    key[PATH_MAX]   =    "";
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+
+        GF_ASSERT (conf);
+        GF_ASSERT (req);
+        GF_ASSERT (dict);
+        GF_ASSERT (err_str);
+
+        ret = dict_get_str (dict, "snapname", &snapname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                        "get snapname");
+                goto out;
+        }
+
+        snap = glusterd_find_snap_by_name (snapname);
+        if (!snap) {
+                snprintf (err_str, len, "Snap (%s) does not exist", snapname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                ret = -1;
+                goto out;
+        }
+
+        list_for_each_entry (snap_volinfo, &snap->volumes, vol_list) {
+                i++;
+                snprintf (key, sizeof (key), "volname%d", i);
+                buf = gf_strdup (snap_volinfo->parent_volname);
+                if (!buf) {
+                        ret = -1;
+                        goto out;
+                }
+                ret = dict_set_dynstr (dict, key, buf);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Could not set "
+                                "parent volume name %s in the dict",
+                                snap_volinfo->parent_volname);
+                        GF_FREE (buf);
+                        goto out;
+                }
+                buf = NULL;
+        }
+
+        ret = dict_set_int32 (dict, "volcount", i);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not save volume count");
+                goto out;
+        }
+
+        ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to initiate snap "
+                                "phases");
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+glusterd_snap_t*
+glusterd_create_snap_object (dict_t *dict, dict_t *rsp_dict)
+{
+        char              *snapname    = NULL;
+        uuid_t            *snap_id     = NULL;
+        char              *description = NULL;
+        glusterd_snap_t   *snap        = NULL;
+        xlator_t          *this        = NULL;
+        glusterd_conf_t   *priv        = NULL;
+        int                ret         = -1;
+        int64_t            time_stamp  = 0;
+
+        this = THIS;
+        priv = this->private;
+
+        GF_ASSERT (dict);
+        GF_ASSERT (rsp_dict);
+
+        /* Fetch snapname, description, id and time from dict */
+        ret = dict_get_str (dict, "snapname", &snapname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snapname");
+                goto out;
+        }
+
+        /* Ignore ret value for description*/
+        ret = dict_get_str (dict, "description", &description);
+
+        ret = dict_get_bin (dict, "snap-id", (void **)&snap_id);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snap_id");
+                goto out;
+        }
+
+        ret = dict_get_int64 (dict, "snap-time", &time_stamp);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snap-time");
+                goto out;
+        }
+        if (time_stamp <= 0) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR, "Invalid time-stamp: %ld",
+                        time_stamp);
+                goto out;
+        }
+
+        list_for_each_entry (snap, &priv->snapshots, snap_list) {
+                if (!strcmp (snap->snapname, snapname) ||
+                    !uuid_compare (snap->snap_id, *snap_id)) {
+                        gf_log (THIS->name, GF_LOG_ERROR,
+                                "Found duplicate snap %s (%s)",
+                                snap->snapname, uuid_utoa (snap->snap_id));
+                        ret = -1;
+                        break;
+                }
+        }
+        if (ret) {
+                snap = NULL;
+                goto out;
+        }
+
+        snap = glusterd_new_snap_object ();
+        if (!snap) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not create "
+                        "the snap object for snap %s", snapname);
+                goto out;
+        }
+
+        strcpy (snap->snapname, snapname);
+        uuid_copy (snap->snap_id, *snap_id);
+        snap->time_stamp = (time_t)time_stamp;
+        /* Set the status as GD_SNAP_STATUS_INIT and once the backend snapshot
+           is taken and snap is really ready to use, set the status to
+           GD_SNAP_STATUS_IN_USE. This helps in identifying the incomplete
+           snapshots and cleaning them up.
+        */
+        snap->snap_status = GD_SNAP_STATUS_INIT;
+        if (description) {
+                snap->description = gf_strdup (description);
+                if (snap->description == NULL) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Saving the Snap Description Failed");
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        ret = glusterd_store_snap (snap);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "Could not store snap"
+                        "object %s", snap->snapname);
+                goto out;
+        }
+
+        list_add_order (&snap->snap_list, &priv->snapshots,
+                        glusterd_compare_snap_time);
+
+        gf_log (this->name, GF_LOG_TRACE, "Snap %s added to the list",
+                snap->snapname);
+
+        ret = 0;
+
+out:
+        if (ret) {
+                if (snap)
+                        glusterd_snap_remove (rsp_dict, snap,
+                                              _gf_true, _gf_true);
+                snap = NULL;
+        }
+
+        return snap;
+}
+
+/* This function is called to get the device path of the snap lvm. Usually
+   if /dev/mapper/<group-name>-<lvm-name> is the device for the lvm,
+   then the snap device will be /dev/<group-name>/<snapname>.
+   This function takes care of building the path for the snap device.
+*/
+char *
+glusterd_build_snap_device_path (char *device, char *snapname)
+{
+        char        snap[PATH_MAX]      = "";
+        char        msg[1024]           = "";
+        char        volgroup[PATH_MAX]  = "";
+        char       *snap_device         = NULL;
+        xlator_t   *this                = NULL;
+        runner_t    runner              = {0,};
+        char       *ptr                 = NULL;
+        int         ret                 = -1;
+
+        this = THIS;
+        GF_ASSERT (this);
+        if (!device) {
+                gf_log (this->name, GF_LOG_ERROR, "device is NULL");
+                goto out;
+        }
+        if (!snapname) {
+                gf_log (this->name, GF_LOG_ERROR, "snapname is NULL");
+                goto out;
+        }
+
+        runinit (&runner);
+        runner_add_args (&runner, "/sbin/lvs", "--noheadings", "-o", "vg_name",
+                         device, NULL);
+        runner_redir (&runner, STDOUT_FILENO, RUN_PIPE);
+        snprintf (msg, sizeof (msg), "Get volume group for device %s", device);
+        runner_log (&runner, this->name, GF_LOG_DEBUG, msg);
+        ret = runner_start (&runner);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get volume group "
+                        "for device %s", device);
+                runner_end (&runner);
+                goto out;
+        }
+        ptr = fgets(volgroup, sizeof(volgroup),
+                    runner_chio (&runner, STDOUT_FILENO));
+        if (!ptr || !strlen(volgroup)) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get volume group "
+                        "for snap %s", snapname);
+                runner_end (&runner);
+                ret = -1;
+                goto out;
+        }
+        runner_end (&runner);
+
+        snprintf (snap, sizeof(snap), "/dev/%s/%s", gf_trim(volgroup),
+                  snapname);
+        snap_device = gf_strdup (snap);
+        if (!snap_device) {
+                gf_log (this->name, GF_LOG_WARNING, "Cannot copy the "
+                        "snapshot device name for snapname: %s)", snapname);
+        }
+
+out:
+        return snap_device;
+}
+
+/* This function actually calls the command (or the API) for taking the
+   snapshot of the backend brick filesystem. If this is successful,
+   then call the glusterd_snap_create function to create the snap object
+   for glusterd
+*/
+char *
+glusterd_take_lvm_snapshot (glusterd_volinfo_t *snap_vol,
+                            glusterd_brickinfo_t *brickinfo)
+{
+        char             msg[NAME_MAX]    = "";
+        char             buf[PATH_MAX]    = "";
+        char            *snap_device      = NULL;
+        char            *ptr              = NULL;
+        char            *device           = NULL;
+        int              ret              = -1;
+        gf_boolean_t     match            = _gf_false;
+        runner_t         runner           = {0,};
+        xlator_t        *this             = NULL;
+
+        this = THIS;
+
+        if (!brickinfo) {
+                gf_log (this->name, GF_LOG_ERROR, "brickinfo NULL");
+                goto out;
+        }
+
+        device = glusterd_get_brick_mount_details (brickinfo);
+        if (!device) {
+                gf_log (this->name, GF_LOG_ERROR, "getting device name for "
+                        "the brick %s:%s failed", brickinfo->hostname,
+                        brickinfo->path);
+                goto out;
+        }
+
+        /* Figuring out if setactivationskip flag is supported or not */
+        runinit (&runner);
+        snprintf (msg, sizeof (msg), "running lvcreate help");
+        runner_add_args (&runner, "/sbin/lvcreate", "--help", NULL);
+        runner_log (&runner, "", GF_LOG_DEBUG, msg);
+        runner_redir (&runner, STDOUT_FILENO, RUN_PIPE);
+        ret = runner_start (&runner);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to run lvcreate help");
+                runner_end (&runner);
+                goto out;
+        }
+
+        /* Looking for setactivationskip in lvcreate --help */
+        do {
+                ptr = fgets(buf, sizeof(buf),
+                            runner_chio (&runner, STDOUT_FILENO));
+                if (ptr) {
+                        if (strstr(buf, "setactivationskip")) {
+                                match = _gf_true;
+                                break;
+                        }
+                }
+        } while (ptr != NULL);
+        runner_end (&runner);
+
+        /* Takng the actual snapshot */
+        runinit (&runner);
+        snprintf (msg, sizeof (msg), "taking snapshot of the brick %s:%s",
+                  brickinfo->hostname, brickinfo->path);
+        if (match == _gf_true)
+                runner_add_args (&runner, "/sbin/lvcreate", "-s", device,
+                                 "--setactivationskip", "n", "--name",
+                                 snap_vol->volname, NULL);
+        else
+                runner_add_args (&runner, "/sbin/lvcreate", "-s", device,
+                                 "--name", snap_vol->volname, NULL);
+        runner_log (&runner, "", GF_LOG_DEBUG, msg);
+        ret = runner_start (&runner);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "taking snapshot of the "
+                        "brick (%s:%s) of device %s failed",
+                        brickinfo->hostname, brickinfo->path, device);
+                runner_end (&runner);
+                goto out;
+        }
+        runner_end (&runner);
+
+        snap_device = glusterd_build_snap_device_path (device,
+                                                       snap_vol->volname);
+        if (!snap_device) {
+                gf_log (this->name, GF_LOG_WARNING, "Cannot copy the snapshot "
+                        "device name for snap %s (volume id: %s)",
+                        snap_vol->snapshot->snapname, snap_vol->volname);
+                ret = -1;
+                goto out;
+        }
+
+out:
+        return snap_device;
+}
+
+int32_t
+glusterd_snap_brick_create (char *device, glusterd_volinfo_t *snap_volinfo,
+                            glusterd_brickinfo_t *original_brickinfo,
+                            int32_t brick_count, char *snap_brick_dir)
+{
+        int32_t          ret                             = -1;
+        xlator_t        *this                            = NULL;
+        glusterd_conf_t *priv                            = NULL;
+        char             snap_brick_mount_path[PATH_MAX] = "";
+        char             snap_brick_path[PATH_MAX]       = "";
+        char             msg[1024]                       = "";
+        struct stat      statbuf                         = {0, };
+        runner_t         runner                          = {0, };
+
+        this = THIS;
+        priv = this->private;
+
+        GF_ASSERT (device);
+        GF_ASSERT (snap_volinfo);
+        GF_ASSERT (original_brickinfo);
+        GF_ASSERT (snap_brick_dir);
+
+        snprintf (snap_brick_mount_path, sizeof (snap_brick_mount_path),
+                  "%s/%s/brick%d",  snap_mount_folder, snap_volinfo->volname,
+                  brick_count+1);
+
+        snprintf (snap_brick_path, sizeof (snap_brick_path), "%s%s",
+                  snap_brick_mount_path, snap_brick_dir);
+
+        ret = mkdir_p (snap_brick_mount_path, 0777, _gf_true);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "creating the brick directory"
+                        " %s for the snapshot %s(device: %s) failed",
+                        snap_brick_mount_path, snap_volinfo->volname, device);
+                goto out;
+        }
+        /* mount the snap logical device on the directory inside
+           /run/gluster/snaps/<snapname>/@snap_brick_mount_path
+           Way to mount the snap brick via mount api is this.
+           ret = mount (device, snap_brick_mount_path, entry->mnt_type,
+                        MS_MGC_VAL, "nouuid");
+           But for now, mounting using runner apis.
+        */
+        runinit (&runner);
+        snprintf (msg, sizeof (msg), "mounting snapshot of the brick %s:%s",
+                  original_brickinfo->hostname, original_brickinfo->path);
+        runner_add_args (&runner, "mount", "-o", "nouuid", device,
+                         snap_brick_mount_path, NULL);
+        runner_log (&runner, "", GF_LOG_DEBUG, msg);
+
+        /* let glusterd get blocked till snapshot is over */
+        synclock_unlock (&priv->big_lock);
+        ret = runner_run (&runner);
+        synclock_lock (&priv->big_lock);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "mounting the snapshot "
+                        "logical device %s failed (error: %s)", device,
+                        strerror (errno));
+                goto out;
+        } else
+                gf_log (this->name, GF_LOG_DEBUG, "mounting the snapshot "
+                        "logical device %s successful", device);
+
+        ret = stat (snap_brick_path, &statbuf);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "stat of the brick %s"
+                        "(brick mount: %s) failed (%s)", snap_brick_path,
+                        snap_brick_mount_path, strerror (errno));
+                goto out;
+        }
+        ret = sys_lsetxattr (snap_brick_path,
+                             GF_XATTR_VOL_ID_KEY,
+                             snap_volinfo->volume_id, 16,
+                             XATTR_REPLACE);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+                        "extended attribute %s on %s. Reason: "
+                        "%s, snap: %s", GF_XATTR_VOL_ID_KEY,
+                        snap_brick_path, strerror (errno),
+                        snap_volinfo->volname);
+                goto out;
+        }
+
+out:
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "unmounting the snap brick"
+                        " mount %s", snap_brick_mount_path);
+                umount (snap_brick_mount_path);
+        }
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+/* Added missed_snap_entry to rsp_dict */
+int32_t
+glusterd_add_missed_snaps_to_dict (dict_t *rsp_dict, char *snap_uuid,
+                                   glusterd_brickinfo_t *brickinfo,
+                                   int32_t brick_number, int32_t op)
+{
+        char                   *buf                             = NULL;
+        char                    missed_snap_entry[PATH_MAX]     = "";
+        char                    name_buf[PATH_MAX]              = "";
+        int32_t                 missed_snap_count               = -1;
+        int32_t                 ret                             = -1;
+        xlator_t               *this                            = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (snap_uuid);
+        GF_ASSERT (brickinfo);
+
+        snprintf (missed_snap_entry, sizeof(missed_snap_entry),
+                  "%s:%s=%d:%s:%d:%d", uuid_utoa(brickinfo->uuid),
+                  snap_uuid, brick_number, brickinfo->path, op,
+                  GD_MISSED_SNAP_PENDING);
+
+        buf = gf_strdup (missed_snap_entry);
+        if (!buf) {
+                ret = -1;
+                goto out;
+        }
+
+        /* Fetch the missed_snap_count from the dict */
+        ret = dict_get_int32 (rsp_dict, "missed_snap_count",
+                              &missed_snap_count);
+        if (ret) {
+                /* Initialize the missed_snap_count for the first time */
+                missed_snap_count = 0;
+        }
+
+        /* Setting the missed_snap_entry in the rsp_dict */
+        snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d",
+                  missed_snap_count);
+        ret = dict_set_dynstr (rsp_dict, name_buf, buf);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set missed_snap_entry (%s) "
+                        "in the rsp_dict.", buf);
+                GF_FREE (buf);
+                goto out;
+        }
+        missed_snap_count++;
+
+        /* Setting the new missed_snap_count in the dict */
+        ret = dict_set_int32 (rsp_dict, "missed_snap_count",
+                              missed_snap_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set missed_snap_count for %s "
+                        "in the rsp_dict.", missed_snap_entry);
+                goto out;
+        }
+
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+static int32_t
+glusterd_add_bricks_to_snap_volume (dict_t *dict, dict_t *rsp_dict,
+                                    glusterd_volinfo_t  *snap_vol,
+                                    glusterd_brickinfo_t *original_brickinfo,
+                                    glusterd_brickinfo_t *snap_brickinfo,
+                                    char **snap_brick_dir, int64_t volcount,
+                                    int32_t brick_count)
+{
+        char                    key[PATH_MAX]                   = "";
+        char                    snap_brick_path[PATH_MAX]       = "";
+        char                   *snap_device                     = NULL;
+        gf_boolean_t            add_missed_snap                 = _gf_false;
+        int32_t                 ret                             = -1;
+        xlator_t               *this                            = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (snap_vol);
+        GF_ASSERT (original_brickinfo);
+        GF_ASSERT (snap_brickinfo);
+        GF_ASSERT (snap_brick_dir);
+
+        snprintf (key, sizeof(key) - 1, "vol%ld.brickdir%d", volcount,
+                  brick_count);
+        ret = dict_get_ptr (dict, key, (void **)snap_brick_dir);
+        if (ret) {
+                /* Using original brickinfo here because it will be a
+                 * pending snapshot and storing the original brickinfo
+                 * will help in mapping while recreating the missed snapshot
+                 */
+                gf_log (this->name, GF_LOG_WARNING, "Unable to fetch "
+                        "snap mount path (%s). Using original brickinfo", key);
+                snap_brickinfo->snap_status = -1;
+                strcpy (snap_brick_path, original_brickinfo->path);
+
+                /* In origiator node add snaps missed
+                 * from different nodes to the dict
+                 */
+                if (is_origin_glusterd (dict) == _gf_true)
+                        add_missed_snap = _gf_true;
+        } else {
+                /* Create brick-path in the format /var/run/gluster/snaps/ *
+                 * <snap-uuid>/<original-brick#>/snap-brick-dir *
+                 */
+                snprintf (snap_brick_path, sizeof(snap_brick_path),
+                          "%s/%s/brick%d%s", snap_mount_folder,
+                          snap_vol->volname, brick_count+1,
+                          *snap_brick_dir);
+        }
+
+        if ((snap_brickinfo->snap_status != -1) &&
+            (!uuid_compare (original_brickinfo->uuid, MY_UUID)) &&
+            (!glusterd_is_brick_started (original_brickinfo))) {
+                /* In case if the brick goes down after prevalidate. */
+                gf_log (this->name, GF_LOG_WARNING, "brick %s:%s is not"
+                        " started (snap: %s)",
+                        original_brickinfo->hostname,
+                        original_brickinfo->path,
+                        snap_vol->snapshot->snapname);
+
+                snap_brickinfo->snap_status = -1;
+                strcpy (snap_brick_path, original_brickinfo->path);
+                add_missed_snap = _gf_true;
+        }
+
+        if (add_missed_snap) {
+                ret = glusterd_add_missed_snaps_to_dict (rsp_dict,
+                                            snap_vol->volname,
+                                            original_brickinfo,
+                                            brick_count + 1,
+                                            GF_SNAP_OPTION_TYPE_CREATE);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to add missed"
+                                " snapshot info for %s:%s in the rsp_dict",
+                                original_brickinfo->hostname,
+                                original_brickinfo->path);
+                        goto out;
+                }
+        }
+
+        snprintf (key, sizeof(key), "vol%ld.brick_snapdevice%d",
+                  volcount, brick_count);
+        ret = dict_get_ptr (dict, key, (void **)&snap_device);
+        if (ret) {
+                /* If the device name is empty, so will be the brick path
+                 * Hence the missed snap has already been added above
+                 */
+                gf_log (this->name, GF_LOG_ERROR, "Unable to fetch "
+                        "snap device (%s). Leaving empty", key);
+        } else
+                strcpy (snap_brickinfo->device_path, snap_device);
+
+        ret = gf_canonicalize_path (snap_brick_path);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to canonicalize path");
+                goto out;
+        }
+
+        strcpy (snap_brickinfo->hostname, original_brickinfo->hostname);
+        strcpy (snap_brickinfo->path, snap_brick_path);
+        uuid_copy (snap_brickinfo->uuid, original_brickinfo->uuid);
+        list_add_tail (&snap_brickinfo->brick_list, &snap_vol->bricks);
+
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+static int32_t
+glusterd_take_brick_snapshot (glusterd_volinfo_t *origin_vol,
+                              glusterd_volinfo_t *snap_vol, dict_t *rsp_dict,
+                              glusterd_brickinfo_t *original_brickinfo,
+                              glusterd_brickinfo_t *snap_brickinfo,
+                              char *snap_brick_dir, int32_t brick_count)
+{
+        char                   *device                          = NULL;
+        int32_t                 ret                             = -1;
+        xlator_t               *this                            = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (origin_vol);
+        GF_ASSERT (snap_vol);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (original_brickinfo);
+        GF_ASSERT (snap_brickinfo);
+        GF_ASSERT (snap_brick_dir);
+
+        device = glusterd_take_lvm_snapshot (snap_vol, original_brickinfo);
+        /* Fail the snapshot even though snapshot on one of
+           the bricks fails. At the end when we check whether
+           the snapshot volume meets quorum or not, then the
+           the snapshot can either be treated as success, or
+           in case of failure we can undo the changes and return
+           failure to cli.  */
+        if (!device) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to take snapshot of %s:%s",
+                        original_brickinfo->hostname,
+                        original_brickinfo->path);
+                goto out;
+        }
+
+        /* create the complete brick here */
+        ret = glusterd_snap_brick_create (device, snap_vol,
+                                          original_brickinfo,
+                                          brick_count, snap_brick_dir);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "not able to"
+                        " create the brickinfo for the snap %s"
+                        ", volume %s", snap_vol->snapshot->snapname,
+                        origin_vol->volname);
+                goto out;
+        }
+
+out:
+        if (device)
+                GF_FREE (device);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+/* Look for disconnected peers, for missed snap creates or deletes */
+static int32_t
+glusterd_find_missed_snap (dict_t *rsp_dict, glusterd_volinfo_t *vol,
+                           char *snap_uuid, struct list_head *peers,
+                           int32_t op)
+{
+        int32_t                   brick_count          = -1;
+        int32_t                   ret                  = -1;
+        xlator_t                 *this                 = NULL;
+        glusterd_peerinfo_t      *peerinfo             = NULL;
+        glusterd_brickinfo_t     *brickinfo            = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (peers);
+        GF_ASSERT (vol);
+        GF_ASSERT (snap_uuid);
+
+        brick_count = 0;
+        list_for_each_entry (brickinfo, &vol->bricks, brick_list) {
+                if (!uuid_compare (brickinfo->uuid, MY_UUID)) {
+                        /* If the brick belongs to the same node */
+                        brick_count++;
+                        continue;
+                }
+
+                list_for_each_entry (peerinfo, peers, uuid_list) {
+                        if (uuid_compare (peerinfo->uuid, brickinfo->uuid)) {
+                                /* If the brick doesnt belong to this peer */
+                                continue;
+                        }
+
+                        /* Found peer who owns the brick,    *
+                         * if peer is not connected or not   *
+                         * friend add it to missed snap list */
+                        if (!(peerinfo->connected) ||
+                           (peerinfo->state.state !=
+                             GD_FRIEND_STATE_BEFRIENDED)) {
+                                ret = glusterd_add_missed_snaps_to_dict
+                                                   (rsp_dict,
+                                                    snap_uuid,
+                                                    brickinfo,
+                                                    brick_count + 1,
+                                                    op);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Failed to add missed snapshot "
+                                                "info for %s:%s in the "
+                                                "rsp_dict", brickinfo->hostname,
+                                                brickinfo->path);
+                                        goto out;
+                                }
+                        }
+                }
+                brick_count++;
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+glusterd_volinfo_t *
+glusterd_do_snap_vol (glusterd_volinfo_t *origin_vol, glusterd_snap_t *snap,
+                      dict_t *dict, dict_t *rsp_dict, int64_t volcount)
+{
+        char                    key[PATH_MAX]                   = "";
+        char                   *snap_brick_dir                  = NULL;
+        char                   *username                        = NULL;
+        char                   *password                        = NULL;
+        glusterd_brickinfo_t   *brickinfo                       = NULL;
+        glusterd_conf_t        *priv                            = NULL;
+        glusterd_volinfo_t     *snap_vol                        = NULL;
+        uuid_t                 *snap_volid                      = NULL;
+        int32_t                 ret                             = -1;
+        int32_t                 brick_count                     = 0;
+        glusterd_brickinfo_t   *snap_brickinfo                  = NULL;
+        xlator_t               *this                            = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (origin_vol);
+        GF_ASSERT (dict);
+        GF_ASSERT (rsp_dict);
+
+        /* fetch username, password and vol_id from dict*/
+        snprintf (key, sizeof(key), "volume%ld_username", volcount);
+        ret = dict_get_str (dict, key, &username);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get %s for "
+                        "snap %s", key, snap->snapname);
+                goto out;
+        }
+
+        snprintf (key, sizeof(key), "volume%ld_password", volcount);
+        ret = dict_get_str (dict, key, &password);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get %s for "
+                        "snap %s", key, snap->snapname);
+                goto out;
+        }
+        snprintf (key, sizeof(key) - 1, "vol%ld_volid", volcount);
+        ret = dict_get_bin (dict, key, (void **)&snap_volid);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to fetch snap_volid");
+                goto out;
+        }
+
+        /* We are not setting the username and password here as
+         * we need to set the user name and password passed in
+         * the dictionary
+         */
+        ret = glusterd_volinfo_dup (origin_vol, &snap_vol, _gf_false);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to duplicate volinfo "
+                        "for the snapshot %s", snap->snapname);
+                goto out;
+        }
+
+        /* uuid is used as lvm snapshot name.
+           This will avoid restrictions on snapshot names provided by user */
+        GLUSTERD_GET_UUID_NOHYPHEN (snap_vol->volname, *snap_volid);
+        uuid_copy (snap_vol->volume_id, *snap_volid);
+        snap_vol->is_snap_volume = _gf_true;
+        strcpy (snap_vol->parent_volname, origin_vol->volname);
+        snap_vol->snapshot = snap;
+
+        glusterd_auth_set_username (snap_vol, username);
+        glusterd_auth_set_password (snap_vol, password);
+
+        /* Adding snap brickinfos to the snap volinfo */
+        brick_count = 0;
+        list_for_each_entry (brickinfo, &origin_vol->bricks, brick_list) {
+                snap_brickinfo = NULL;
+
+                ret = glusterd_brickinfo_new (&snap_brickinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "initializing the brick for the snap "
+                                "volume failed (snapname: %s)", snap->snapname);
+                        goto out;
+                }
+
+                ret = glusterd_add_bricks_to_snap_volume (dict, rsp_dict,
+                                                          snap_vol,
+                                                          brickinfo,
+                                                          snap_brickinfo,
+                                                          &snap_brick_dir,
+                                                          volcount,
+                                                          brick_count);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to add the snap brick for "
+                                "%s:%s to the snap volume",
+                                brickinfo->hostname, brickinfo->path);
+                        GF_FREE (snap_brickinfo);
+                        goto out;
+                }
+
+                /* Take snapshot of the brick */
+                if ((uuid_compare (brickinfo->uuid, MY_UUID)) ||
+                    (snap_brickinfo->snap_status == -1)) {
+                        brick_count++;
+                        continue;
+                }
+
+                ret = glusterd_take_brick_snapshot (origin_vol, snap_vol,
+                                                    rsp_dict, brickinfo,
+                                                    snap_brickinfo,
+                                                    snap_brick_dir,
+                                                    brick_count);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to take snapshot for %s:%s",
+                                brickinfo->hostname, brickinfo->path);
+                        goto out;
+                }
+
+                brick_count++;
+        }
+
+        /*TODO: the quorum check of the snap volume here */
+
+        ret = glusterd_store_volinfo (snap_vol,
+                                      GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to store snapshot "
+                        "volinfo (%s) for snap %s", snap_vol->volname,
+                        snap->snapname);
+                goto out;
+        }
+
+        ret = generate_brick_volfiles (snap_vol);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "generating the brick "
+                        "volfiles for the snap %s (volume: %s) failed",
+                        snap->snapname, origin_vol->volname);
+                goto out;
+        }
+
+        ret = generate_client_volfiles (snap_vol, GF_CLIENT_TRUSTED);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "generating the trusted "
+                        "client volfiles for the snap %s (volume: %s) failed",
+                        snap->snapname, origin_vol->volname);
+                goto out;
+        }
+        ret = generate_client_volfiles (snap_vol, GF_CLIENT_OTHER);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "generating the client "
+                        "volfiles for the snap %s (volume: %s) failed",
+                        snap->snapname, origin_vol->volname);
+                goto out;
+        }
+
+        ret = glusterd_list_add_snapvol (origin_vol, snap_vol);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "could not add the snap "
+                        "volume %s to the list", snap_vol->volname);
+                goto out;
+        }
+
+        list_for_each_entry (brickinfo, &snap_vol->bricks, brick_list) {
+                if (uuid_compare (brickinfo->uuid, MY_UUID))
+                        continue;
+
+                if (brickinfo->snap_status == -1) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "not starting snap brick %s:%s for "
+                                "for the snap %s (volume: %s)",
+                                brickinfo->hostname, brickinfo->path,
+                                snap->snapname, origin_vol->volname);
+                        continue;
+                }
+
+                ret = glusterd_brick_start (snap_vol, brickinfo, _gf_true);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "starting the "
+                                "brick %s:%s for the snap %s (volume: %s) "
+                                "failed", brickinfo->hostname, brickinfo->path,
+                                snap->snapname, origin_vol->volname);
+                        goto out;
+                }
+        }
+
+        snap_vol->status = GLUSTERD_STATUS_STARTED;
+        ret = glusterd_store_volinfo (snap_vol,
+                                      GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to store snap volinfo");
+                goto out;
+        }
+
+out:
+        if (ret) {
+                if (snap_vol)
+                        glusterd_snap_volume_remove (rsp_dict, snap_vol,
+                                                     _gf_true, _gf_true);
+                snap_vol = NULL;
+        }
+
+        return snap_vol;
+}
+
+/* This is a snapshot remove handler function. This function will be
+ * executed in the originator node. This function is responsible for
+ * calling mgmt v3 framework to do the actual remove on all the bricks
+ *
+ * @param req           RPC request object
+ * @param op            gluster operation
+ * @param dict          dictionary containing snapshot remove request
+ * @param err_str       In case of an err this string should be populated
+ * @param len           length of err_str buffer
+ *
+ * @return              Negative value on Failure and 0 in success
+ */
+int
+glusterd_handle_snapshot_remove (rpcsvc_request_t *req, glusterd_op_t op,
+                                 dict_t *dict, char *err_str, size_t len)
+{
+        int                      ret            = -1;
+        int64_t                  volcount       = 0;
+        char                    *snapname       = NULL;
+        char                    *volname        = NULL;
+        char                     key[PATH_MAX]  = "";
+        glusterd_snap_t         *snap           = NULL;
+        glusterd_volinfo_t      *snap_vol       = NULL;
+        glusterd_volinfo_t      *tmp            = NULL;
+        xlator_t                *this           = NULL;
+
+        this = THIS;
+
+        GF_ASSERT (req);
+        GF_ASSERT (dict);
+        GF_ASSERT (err_str);
+
+        ret = dict_get_str (dict, "snapname", &snapname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get snapname");
+                goto out;
+        }
+
+        snap = glusterd_find_snap_by_name (snapname);
+        if (!snap) {
+                snprintf (err_str, len, "Snap (%s) does not exist", snapname);
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s", err_str);
+                ret = -1;
+                goto out;
+        }
+
+        /* Set volnames in the dict to get mgmt_v3 lock */
+        list_for_each_entry_safe (snap_vol, tmp, &snap->volumes, vol_list) {
+                volcount++;
+                volname = gf_strdup (snap_vol->parent_volname);
+                if (!volname) {
+                        ret = -1;
+                        gf_log (this->name, GF_LOG_ERROR, "strdup failed");
+                        goto out;
+                }
+
+                snprintf (key, sizeof (key), "volname%ld", volcount);
+                ret = dict_set_dynstr (dict, key, volname);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+                                "volume name in dictionary");
+                        GF_FREE (volname);
+                        goto out;
+                }
+                volname = NULL;
+        }
+        ret = dict_set_int64 (dict, "volcount", volcount);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set volcount");
+                goto out;
+        }
+
+        ret = glusterd_mgmt_v3_initiate_snap_phases (req, op, dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to initiate snap "
+                        "phases");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_snapshot_remove_prevalidate (dict_t *dict, char **op_errstr,
+                                      dict_t *rsp_dict)
+{
+        int32_t             ret          = -1;
+        char               *snapname     = NULL;
+        xlator_t           *this        = NULL;
+        glusterd_snap_t    *snap        = NULL;
+
+        this = THIS;
+
+        if (!dict || !op_errstr) {
+                gf_log (this->name, GF_LOG_ERROR, "input parameters NULL");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "snapname", &snapname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Getting the snap name "
+                        "failed");
+                goto out;
+        }
+
+        snap = glusterd_find_snap_by_name (snapname);
+        if (!snap) {
+                gf_log (this->name, GF_LOG_ERROR, "Snap %s does not exist",
+                        snapname);
+                ret = -1;
+                goto out;
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_snapshot_status_prevalidate (dict_t *dict, char **op_errstr,
+                                      dict_t *rsp_dict)
+{
+        int                     ret             =       -1;
+        char                    *snapname       =       NULL;
+        glusterd_conf_t         *conf           =       NULL;
+        xlator_t                *this           =       NULL;
+        int32_t                 cmd             =       -1;
+        glusterd_volinfo_t      *volinfo        =       NULL;
+        char                    *volname        =       NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+
+        GF_ASSERT (conf);
+        GF_ASSERT (op_errstr);
+        if (!dict) {
+                gf_log (this->name, GF_LOG_ERROR, "Input dict is NULL");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "cmd", &cmd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not fetch status cmd");
+                goto out;
+        }
+
+        switch (cmd) {
+                case GF_SNAP_STATUS_TYPE_ALL:
+                {
+                        break;
+                }
+                case GF_SNAP_STATUS_TYPE_SNAP:
+                {
+                        ret = dict_get_str (dict, "snapname", &snapname);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Could not fetch snapname");
+                                goto out;
+                        }
+
+                        if (!glusterd_find_snap_by_name (snapname)) {
+                                ret = gf_asprintf (op_errstr, "Snap (%s) "
+                                                  "not found", snapname);
+                                if (ret < 0) {
+                                        goto out;
+                                }
+                                ret = -1;
+                                gf_log (this->name, GF_LOG_ERROR, "Snap (%s) "
+                                        "not found", snapname);
+                                goto out;
+                        }
+                        break;
+                }
+                case GF_SNAP_STATUS_TYPE_VOL:
+                {
+                        ret = dict_get_str (dict, "volname", &volname);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Could not fetch volname");
+                                goto out;
+                        }
+
+                        ret = glusterd_volinfo_find (volname, &volinfo);
+                        if (ret) {
+                                ret = gf_asprintf (op_errstr, "Volume (%s)"
+                                                  "not found", volname);
+                                if (ret < 0) {
+                                        goto out;
+                                }
+                                ret = -1;
+                                gf_log (this->name, GF_LOG_ERROR, "Volume "
+                                        "%s not present", volname);
+                                goto out;
+                        }
+                        break;
+
+                }
+                default:
+                {
+                        gf_log (this->name, GF_LOG_ERROR, "Invalid command");
+                        break;
+                }
+        }
+        ret = 0;
+
+out:
+        return ret;
+}
+
+int32_t
+glusterd_snapshot_remove_commit (dict_t *dict, char **op_errstr,
+                                 dict_t *rsp_dict)
+{
+        int32_t                   ret                  = -1;
+        char                     *snapname             = NULL;
+        char                     *dup_snapname         = NULL;
+        glusterd_snap_t          *snap                 = NULL;
+        glusterd_conf_t          *priv                 = NULL;
+        glusterd_volinfo_t       *snap_volinfo         = NULL;
+        xlator_t                 *this                 = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (op_errstr);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        if (!dict || !op_errstr) {
+                gf_log (this->name, GF_LOG_ERROR, "input parameters NULL");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "snapname", &snapname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Getting the snap name "
+                        "failed");
+                goto out;
+        }
+
+        snap = glusterd_find_snap_by_name (snapname);
+        if (!snap) {
+                gf_log (this->name, GF_LOG_ERROR, "Snap %s does not exist",
+                        snapname);
+                ret = -1;
+                goto out;
+        }
+
+        if (is_origin_glusterd (dict) == _gf_true) {
+                /* TODO : As of now there is only volume in snapshot.
+                 * Change this when multiple volume snapshot is introduced
+                 */
+                snap_volinfo = list_entry (snap->volumes.next,
+                                           glusterd_volinfo_t,
+                                           vol_list);
+                if (!snap_volinfo) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to fetch snap_volinfo");
+                        ret = -1;
+                        goto out;
+                }
+
+                /* From origin glusterd check if      *
+                 * any peers with snap bricks is down */
+                ret = glusterd_find_missed_snap (rsp_dict, snap_volinfo,
+                                                 snap_volinfo->volname,
+                                                 &priv->peers,
+                                                 GF_SNAP_OPTION_TYPE_DELETE);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to find missed snap deletes");
+                        goto out;
+                }
+        }
+
+        ret = glusterd_snap_remove (rsp_dict, snap, _gf_true, _gf_false);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to remove snap %s",
+                        snapname);
+                goto out;
+        }
+
+        dup_snapname = gf_strdup (snapname);
+        if (!dup_snapname) {
+                gf_log (this->name, GF_LOG_ERROR, "Strdup failed");
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_dynstr (rsp_dict, "snapname", dup_snapname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set the snapname");
+                GF_FREE (dup_snapname);
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+glusterd_do_snap_cleanup (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
+{
+        int32_t                  ret                   = -1;
+        char                     *name                 = NULL;
+        xlator_t                 *this                 = NULL;
+        glusterd_conf_t          *conf                 = NULL;
+        glusterd_volinfo_t       *volinfo              = NULL;
+        glusterd_snap_t          *snap                 = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        if (!dict || !op_errstr) {
+                gf_log (this->name, GF_LOG_ERROR, "input parameters NULL");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "snapname", &name);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "getting the snap "
+                        "name failed (volume: %s)", volinfo->volname);
+                goto out;
+        }
+
+        /*
+          If the snapname is not found that means the failure happened at
+          staging, or in commit, before the snap object is created, in which
+          case there is nothing to cleanup. So set ret to 0.
+        */
+        snap = glusterd_find_snap_by_name (name);
+        if (!snap) {
+                gf_log (this->name, GF_LOG_INFO, "snap %s is not found", name);
+                ret = 0;
+                goto out;
+        }
+
+        ret = glusterd_snap_remove (rsp_dict, snap, _gf_true, _gf_true);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "removing the snap %s failed",
+                        name);
+                goto out;
+        }
+
+        name = NULL;
+
+        ret = 0;
+
+out:
+
+        return ret;
+}
+
+/* In case of a successful, delete or create operation, during post_validate *
+ * look for missed snap operations and update the missed snap lists */
+int32_t
+glusterd_snapshot_update_snaps_post_validate (dict_t *dict, char **op_errstr,
+                                              dict_t *rsp_dict)
+{
+        int32_t                  ret                    = -1;
+        int32_t                  missed_snap_count      = -1;
+        xlator_t                *this                   = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (op_errstr);
+
+        ret = dict_get_int32 (dict, "missed_snap_count",
+                              &missed_snap_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "No missed snaps");
+                ret = 0;
+                goto out;
+        }
+
+        ret = glusterd_store_update_missed_snaps (dict, missed_snap_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to update missed_snaps_list");
+                goto out;
+        }
+
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_snapshot_create_commit (dict_t *dict, char **op_errstr,
+                                 dict_t *rsp_dict)
+{
+        int                     ret                     = -1;
+        int64_t                 i                       = 0;
+        int64_t                 volcount                = 0;
+        char                    *snapname               = NULL;
+        char                    *volname                = NULL;
+        char                    *tmp_name               = NULL;
+        char                    key[PATH_MAX]           = "";
+        xlator_t                *this                   = NULL;
+        glusterd_snap_t         *snap                   = NULL;
+        glusterd_volinfo_t      *origin_vol             = NULL;
+        glusterd_volinfo_t      *snap_vol               = NULL;
+        glusterd_conf_t *priv   = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+        GF_ASSERT(dict);
+        GF_ASSERT(op_errstr);
+        GF_ASSERT(rsp_dict);
+        priv = this->private;
+        GF_ASSERT(priv);
+
+        ret = dict_get_int64 (dict, "volcount", &volcount);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to "
+                        "get the volume count");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "snapname", &snapname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to fetch snapname");
+                goto out;
+        }
+        tmp_name = gf_strdup (snapname);
+        if (!tmp_name) {
+                gf_log (this->name, GF_LOG_ERROR, "Out of memory");
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_dynstr (rsp_dict, "snapname", tmp_name);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to set snapname in rsp_dict");
+                GF_FREE (tmp_name);
+                goto out;
+        }
+        tmp_name = NULL;
+
+        snap = glusterd_create_snap_object (dict, rsp_dict);
+        if (!snap) {
+                gf_log (this->name, GF_LOG_ERROR, "creating the"
+                        "snap object %s failed", snapname);
+                ret = -1;
+                goto out;
+        }
+
+        for (i = 1; i <= volcount; i++) {
+                snprintf (key, sizeof (key), "volname%ld", i);
+                ret = dict_get_str (dict, key, &volname);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to get volume name");
+                        goto out;
+                }
+
+                ret = glusterd_volinfo_find (volname, &origin_vol);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to get the volinfo for "
+                                "the volume %s", volname);
+                        goto out;
+                }
+
+                /* TODO: Create a stub where the bricks are
+                   added parallely by worker threads so that
+                   the snap creating happens parallely.  */
+                snap_vol = glusterd_do_snap_vol (origin_vol, snap, dict,
+                                                 rsp_dict, i);
+                if (!snap_vol) {
+                        ret = -1;
+                        gf_log (this->name, GF_LOG_WARNING, "taking the "
+                                "snapshot of the volume %s failed", volname);
+                        goto out;
+                }
+        }
+
+        snap->snap_status = GD_SNAP_STATUS_IN_USE;
+        ret = glusterd_store_snap (snap);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "Could not store snap"
+                        "object %s", snap->snapname);
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        if (ret) {
+                if (snap)
+                        glusterd_snap_remove (rsp_dict, snap,
+                                              _gf_true, _gf_true);
+                snap = NULL;
+        }
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int
+snap_max_hard_limit_set_commit (dict_t *dict, uint64_t value,
+                                char *volname, char **op_errstr)
+{
+        char                err_str[PATH_MAX]    = "";
+        glusterd_conf_t    *conf                 = NULL;
+        glusterd_volinfo_t *volinfo              = NULL;
+        int                 ret                  = -1;
+        xlator_t           *this                 = NULL;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (volname);
+        GF_ASSERT (op_errstr);
+
+        conf = this->private;
+
+        GF_ASSERT (conf);
+
+        /* TODO: Initiate auto deletion when there is a limit change */
+        if (!volname) {
+                /* For system limit */
+                conf->snap_max_hard_limit = value;
+
+                ret = glusterd_store_global_info (this);
+                if (ret) {
+                        snprintf (err_str, PATH_MAX, "Failed to store "
+                                  "snap-max-hard-limit for system");
+                        goto out;
+                }
+        } else {
+               /*  For one volume */
+                ret = glusterd_volinfo_find (volname, &volinfo);
+                if (ret) {
+                        snprintf (err_str, PATH_MAX, "Failed to get the"
+                                   " volinfo for volume %s", volname);
+                        goto out;
+                }
+
+                volinfo->snap_max_hard_limit = value;
+
+                ret = glusterd_store_volinfo (volinfo,
+                                        GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+                if (ret) {
+                        snprintf (err_str, PATH_MAX, "Failed to store "
+                                 "snap-max-hard-limit for volume %s", volname);
+                        goto out;
+                }
+        }
+
+        ret = 0;
+out:
+        if (ret) {
+                *op_errstr = gf_strdup (err_str);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+        }
+        return ret;
+}
+
+int
+snap_max_limits_display_commit (dict_t *rsp_dict, char *volname,
+                                char **op_errstr)
+{
+        char                err_str[PATH_MAX]    = "";
+        char                buf[PATH_MAX]        = "";
+        glusterd_conf_t    *conf                 = NULL;
+        glusterd_volinfo_t *volinfo              = NULL;
+        int                 ret                  = -1;
+        uint64_t            active_hard_limit    = 0;
+        uint64_t            snap_max_limit       = 0;
+        uint64_t            soft_limit_value     = -1;
+        uint64_t            count                = 0;
+        xlator_t           *this                 = NULL;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (volname);
+        GF_ASSERT (op_errstr);
+
+        conf = this->private;
+
+        GF_ASSERT (conf);
+
+        if (!volname) {
+                /* For system limit */
+                list_for_each_entry (volinfo, &conf->volumes, vol_list) {
+                        if (volinfo->is_snap_volume == _gf_true)
+                                continue;
+                        snap_max_limit = volinfo->snap_max_hard_limit;
+                        if (snap_max_limit > conf->snap_max_hard_limit)
+                                active_hard_limit = conf->snap_max_hard_limit;
+                        else
+                                active_hard_limit = snap_max_limit;
+                        soft_limit_value = (active_hard_limit *
+                                            conf->snap_max_soft_limit) / 100;
+
+                        snprintf (buf, sizeof(buf), "volume%ld-volname", count);
+                        ret = dict_set_str (rsp_dict, buf, volinfo->volname);
+                        if (ret) {
+                                snprintf (err_str, PATH_MAX,
+                                          "Failed to set %s", buf);
+                                goto out;
+                        }
+
+                        snprintf (buf, sizeof(buf),
+                                  "volume%ld-snap-max-hard-limit", count);
+                        ret = dict_set_uint64 (rsp_dict, buf, snap_max_limit);
+                        if (ret) {
+                                snprintf (err_str, PATH_MAX,
+                                          "Failed to set %s", buf);
+                                goto out;
+                        }
+
+                        snprintf (buf, sizeof(buf),
+                                  "volume%ld-active-hard-limit", count);
+                        ret = dict_set_uint64 (rsp_dict, buf,
+                                               active_hard_limit);
+                        if (ret) {
+                                snprintf (err_str, PATH_MAX,
+                                          "Failed to set %s", buf);
+                                goto out;
+                        }
+
+                        snprintf (buf, sizeof(buf),
+                                  "volume%ld-snap-max-soft-limit", count);
+                        ret = dict_set_uint64 (rsp_dict, buf, soft_limit_value);
+                        if (ret) {
+                                snprintf (err_str, PATH_MAX,
+                                          "Failed to set %s", buf);
+                                goto out;
+                        }
+                        count++;
+                }
+
+                ret = dict_set_uint64 (rsp_dict, "voldisplaycount", count);
+                if (ret) {
+                        snprintf (err_str, PATH_MAX,
+                                  "Failed to set voldisplaycount");
+                        goto out;
+                }
+        } else {
+                /*  For one volume */
+                ret = glusterd_volinfo_find (volname, &volinfo);
+                if (ret) {
+                        snprintf (err_str, PATH_MAX, "Failed to get the"
+                                   " volinfo for volume %s", volname);
+                        goto out;
+                }
+
+                snap_max_limit = volinfo->snap_max_hard_limit;
+                if (snap_max_limit > conf->snap_max_hard_limit)
+                        active_hard_limit = conf->snap_max_hard_limit;
+                else
+                        active_hard_limit = snap_max_limit;
+
+                soft_limit_value = (active_hard_limit *
+                                    conf->snap_max_soft_limit) / 100;
+
+                snprintf (buf, sizeof(buf), "volume%ld-volname", count);
+                ret = dict_set_str (rsp_dict, buf, volinfo->volname);
+                if (ret) {
+                        snprintf (err_str, PATH_MAX,
+                                  "Failed to set %s", buf);
+                        goto out;
+                }
+
+                snprintf (buf, sizeof(buf),
+                          "volume%ld-snap-max-hard-limit", count);
+                ret = dict_set_uint64 (rsp_dict, buf, snap_max_limit);
+                if (ret) {
+                        snprintf (err_str, PATH_MAX,
+                                  "Failed to set %s", buf);
+                        goto out;
+                }
+
+                snprintf (buf, sizeof(buf),
+                          "volume%ld-active-hard-limit", count);
+                ret = dict_set_uint64 (rsp_dict, buf, active_hard_limit);
+                if (ret) {
+                        snprintf (err_str, PATH_MAX,
+                                  "Failed to set %s", buf);
+                        goto out;
+                }
+
+                snprintf (buf, sizeof(buf),
+                          "volume%ld-snap-max-soft-limit", count);
+                ret = dict_set_uint64 (rsp_dict, buf, soft_limit_value);
+                if (ret) {
+                        snprintf (err_str, PATH_MAX,
+                                  "Failed to set %s", buf);
+                        goto out;
+                }
+
+                count++;
+
+                ret = dict_set_uint64 (rsp_dict, "voldisplaycount", count);
+                if (ret) {
+                        snprintf (err_str, PATH_MAX,
+                                  "Failed to set voldisplaycount");
+                        goto out;
+                }
+
+        }
+
+        ret = dict_set_uint64 (rsp_dict, "snap-max-hard-limit",
+                               conf->snap_max_hard_limit);
+        if (ret) {
+                snprintf (err_str, PATH_MAX,
+                          "Failed to set sys-snap-max-hard-limit ");
+                goto out;
+        }
+
+        ret = dict_set_uint64 (rsp_dict, "snap-max-soft-limit",
+                               conf->snap_max_soft_limit);
+        if (ret) {
+                snprintf (err_str, PATH_MAX,
+                          "Failed to set sys-snap-max-hard-limit ");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if (ret) {
+                *op_errstr = gf_strdup (err_str);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+        }
+        return ret;
+}
+
+int
+glusterd_snapshot_config_commit (dict_t *dict, char **op_errstr,
+                                 dict_t *rsp_dict)
+{
+        char               *volname              = NULL;
+        xlator_t           *this                 = NULL;
+        int                 ret                  = -1;
+        char                err_str[PATH_MAX]    = {0,};
+        glusterd_conf_t    *conf                 = NULL;
+        int                 config_command       = 0;
+        uint64_t            hard_limit           = 0;
+        uint64_t            soft_limit           = 0;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+
+        conf = this->private;
+
+        GF_ASSERT (conf);
+
+        ret = dict_get_int32 (dict, "config-command", &config_command);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get config-command type");
+                goto out;
+        }
+
+        /* Ignore the return value of the following dict_get,
+         * as they are optional
+         */
+        ret = dict_get_str (dict, "volname", &volname);
+
+        ret = dict_get_uint64 (dict, "snap-max-hard-limit", &hard_limit);
+
+        ret = dict_get_uint64 (dict, "snap-max-soft-limit", &soft_limit);
+
+        switch (config_command) {
+        case GF_SNAP_CONFIG_TYPE_SET:
+                if (hard_limit) {
+                        /* Commit ops for snap-max-hard-limit */
+                        ret = snap_max_hard_limit_set_commit (dict, hard_limit,
+                                                              volname,
+                                                              op_errstr);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "snap-max-hard-limit set "
+                                        "commit failed.");
+                                goto out;
+                        }
+                }
+
+                if (soft_limit) {
+                        /* For system limit */
+                        conf->snap_max_soft_limit = soft_limit;
+
+                        ret = glusterd_store_global_info (this);
+                        if (ret) {
+                                snprintf (err_str, PATH_MAX, "Failed to store "
+                                          "snap-max-soft-limit for system");
+                                *op_errstr = gf_strdup (err_str);
+                                gf_log (this->name, GF_LOG_ERROR, "%s",
+                                        err_str);
+                                goto out;
+                        }
+                }
+                break;
+
+        case GF_SNAP_CONFIG_DISPLAY:
+                /* Reading data from local node only */
+                if (!is_origin_glusterd (dict)) {
+                        ret = 0;
+                        break;
+                }
+
+                ret = snap_max_limits_display_commit (rsp_dict, volname,
+                                                      op_errstr);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "snap-max-limit "
+                                "display commit failed.");
+                        goto out;
+                }
+                break;
+        default:
+                break;
+        }
+
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_get_brick_lvm_details (dict_t *rsp_dict,
+                               glusterd_brickinfo_t *brickinfo, char *volname,
+                                char *device, char *key_prefix)
+{
+
+        int                     ret             =       -1;
+        glusterd_conf_t         *priv           =       NULL;
+        runner_t                runner          =       {0,};
+        xlator_t                *this           =       NULL;
+        char                    msg[PATH_MAX]   =       "";
+        char                    buf[PATH_MAX]   =       "";
+        char                    *ptr            =       NULL;
+        char                    *token          =       NULL;
+        char                    key[PATH_MAX]   =       "";
+        char                    *value          =       NULL;
+
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (brickinfo);
+        GF_ASSERT (volname);
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        device = glusterd_get_brick_mount_details (brickinfo);
+        if (!device) {
+                gf_log (this->name, GF_LOG_ERROR, "Getting device name for "
+                        "the brick %s:%s failed", brickinfo->hostname,
+                         brickinfo->path);
+                goto out;
+        }
+        runinit (&runner);
+        snprintf (msg, sizeof (msg), "running lvs command, "
+                  "for getting snap status");
+        /* Using lvs command fetch the Volume Group name,
+         * Percentage of data filled and Logical Volume size
+         *
+         * "-o" argument is used to get the desired information,
+         * example : "lvs /dev/VolGroup/thin_vol -o vgname,lv_size",
+         * will get us Volume Group name and Logical Volume size.
+         *
+         * Here separator used is ":",
+         * for the above given command with separator ":",
+         * The output will be "vgname:lvsize"
+         */
+        runner_add_args (&runner, "lvs", device, "--noheading", "-o",
+                         "vg_name,data_percent,lv_size",
+                         "--separator", ":", NULL);
+        runner_redir (&runner, STDOUT_FILENO, RUN_PIPE);
+        runner_log (&runner, "", GF_LOG_DEBUG, msg);
+        ret = runner_start (&runner);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not perform lvs action");
+                goto end;
+        }
+        do {
+                ptr = fgets (buf, sizeof (buf),
+                             runner_chio (&runner, STDOUT_FILENO));
+
+                if (ptr == NULL)
+                        break;
+                token = strtok (buf, ":");
+                if (token != NULL) {
+                        while (token && token[0] == ' ')
+                                token++;
+                        if (!token) {
+                                ret = -1;
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Invalid vg entry");
+                                goto end;
+                        }
+                        value = gf_strdup (token);
+                        if (!value) {
+                                ret = -1;
+                                goto end;
+                        }
+                        ret = snprintf (key, sizeof (key), "%s.vgname",
+                                        key_prefix);
+                        if (ret < 0) {
+                                goto end;
+                        }
+
+                        ret = dict_set_dynstr (rsp_dict, key, value);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Could not save vgname ");
+                                goto end;
+                        }
+                }
+
+                token = strtok (NULL, ":");
+                if (token != NULL) {
+                        value = gf_strdup (token);
+                        if (!value) {
+                                ret = -1;
+                                goto end;
+                        }
+                        ret = snprintf (key, sizeof (key), "%s.data",
+                                        key_prefix);
+                        if (ret < 0) {
+                                goto end;
+                        }
+
+                        ret = dict_set_dynstr (rsp_dict, key, value);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Could not save data percent ");
+                                goto end;
+                        }
+                }
+                token = strtok (NULL, ":");
+                if (token != NULL) {
+                        value = gf_strdup (token);
+                        if (!value) {
+                                ret = -1;
+                                goto end;
+                        }
+                        ret = snprintf (key, sizeof (key), "%s.lvsize",
+                                        key_prefix);
+                        if (ret < 0) {
+                                goto end;
+                        }
+
+                        ret = dict_set_dynstr (rsp_dict, key, value);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Could not save meta data percent ");
+                                goto end;
+                        }
+                }
+
+        } while (ptr != NULL);
+
+        ret = 0;
+
+end:
+        runner_end (&runner);
+
+out:
+        if (ret && value) {
+                GF_FREE (value);
+        }
+
+        return ret;
+}
+
+int
+glusterd_get_single_brick_status (char **op_errstr, dict_t *rsp_dict,
+                                 char *keyprefix, int index,
+                                 glusterd_volinfo_t *snap_volinfo,
+                                 glusterd_brickinfo_t *brickinfo)
+{
+        int             ret                     = -1;
+        xlator_t        *this                   = NULL;
+        glusterd_conf_t *priv                   = NULL;
+        char            key[PATH_MAX]           = "";
+        char            *device                 = NULL;
+        char            *value                  = NULL;
+        char            brick_path[PATH_MAX]    = "";
+        char            pidfile[PATH_MAX]       = "";
+        pid_t           pid                     = -1;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (keyprefix);
+        GF_ASSERT (snap_volinfo);
+        GF_ASSERT (brickinfo);
+
+        ret = snprintf (key, sizeof (key), "%s.brick%d.path", keyprefix,
+                        index);
+        if (ret < 0) {
+                goto out;
+        }
+
+        ret = snprintf (brick_path, sizeof (brick_path),
+                        "%s:%s", brickinfo->hostname, brickinfo->path);
+        if (ret < 0) {
+                goto out;
+        }
+
+        value = gf_strdup (brick_path);
+        if (!value) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_dynstr (rsp_dict, key, value);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to store "
+                        "brick_path %s", brickinfo->path);
+                goto out;
+        }
+
+        if (brickinfo->snap_status == -1) {
+                /* Setting vgname as "Pending Snapshot" */
+                value = gf_strdup ("Pending Snapshot");
+                if (!value) {
+                        ret = -1;
+                        goto out;
+                }
+
+                snprintf (key, sizeof (key), "%s.brick%d.vgname",
+                          keyprefix, index);
+                ret = dict_set_dynstr (rsp_dict, key, value);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Could not save vgname ");
+                        goto out;
+                }
+
+                ret = 0;
+                goto out;
+        }
+        value = NULL;
+
+        ret = snprintf (key, sizeof (key), "%s.brick%d.status",
+                        keyprefix, index);
+        if (ret < 0) {
+                goto out;
+        }
+
+        if (brickinfo->status == GF_BRICK_STOPPED) {
+                value = gf_strdup ("No");
+                if (!value) {
+                        ret = -1;
+                        goto out;
+                }
+                ret = dict_set_str (rsp_dict, key, value);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Could not save brick status");
+                        goto out;
+                }
+                value = NULL;
+        } else {
+                value = gf_strdup ("Yes");
+                if (!value) {
+                        ret = -1;
+                        goto out;
+                }
+                ret = dict_set_str (rsp_dict, key, value);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Could not save brick status");
+                        goto out;
+                }
+                value = NULL;
+
+                GLUSTERD_GET_BRICK_PIDFILE (pidfile, snap_volinfo,
+                                            brickinfo, priv);
+                ret = glusterd_is_service_running (pidfile, &pid);
+
+                ret = snprintf (key, sizeof (key), "%s.brick%d.pid",
+                                keyprefix, index);
+                if (ret < 0) {
+                        goto out;
+                }
+
+                ret = dict_set_int32 (rsp_dict, key, pid);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Could not save pid %d", pid);
+                        goto out;
+                }
+        }
+
+        ret = snprintf (key, sizeof (key), "%s.brick%d",
+                        keyprefix, index);
+        if (ret < 0) {
+                goto out;
+        }
+
+        ret = glusterd_get_brick_lvm_details (rsp_dict, brickinfo,
+                                              snap_volinfo->volname,
+                                              device, key);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                        "brick LVM details");
+                goto out;
+        }
+out:
+        if (ret && value) {
+                GF_FREE (value);
+        }
+
+        return ret;
+}
+
+int
+glusterd_get_single_snap_status (char **op_errstr, dict_t *rsp_dict,
+                                 char *keyprefix, glusterd_snap_t *snap)
+{
+        int                      ret                 =       -1;
+        xlator_t                *this                =       NULL;
+        char                     key[PATH_MAX]       =       "";
+        char                     brickkey[PATH_MAX]  =       "";
+        glusterd_volinfo_t      *snap_volinfo        =       NULL;
+        glusterd_volinfo_t      *tmp_volinfo         =       NULL;
+        glusterd_brickinfo_t    *brickinfo           =       NULL;
+        int                      volcount            =       0;
+        int                      brickcount          =       0;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (keyprefix);
+        GF_ASSERT (snap);
+
+        list_for_each_entry_safe (snap_volinfo, tmp_volinfo, &snap->volumes,
+                                  vol_list) {
+                ret = snprintf (key, sizeof (key), "%s.vol%d", keyprefix,
+                                volcount);
+                if (ret < 0) {
+                        goto out;
+                }
+                list_for_each_entry (brickinfo, &snap_volinfo->bricks,
+                                     brick_list) {
+                        if (!glusterd_is_local_brick (this, snap_volinfo,
+                            brickinfo)) {
+                                brickcount++;
+                                continue;
+                        }
+
+                        ret = glusterd_get_single_brick_status (op_errstr,
+                                                  rsp_dict, key, brickcount,
+                                                  snap_volinfo, brickinfo);
+
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Getting "
+                                        "single snap status failed");
+                                goto out;
+                        }
+                        brickcount++;
+                }
+                ret = snprintf (brickkey, sizeof (brickkey), "%s.brickcount",
+                                key);
+                if (ret < 0) {
+                        goto out;
+                }
+
+                ret = dict_set_int32 (rsp_dict, brickkey, brickcount);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Could not save brick count");
+                        goto out;
+                }
+                volcount++;
+        }
+
+        ret = snprintf (key, sizeof (key), "%s.volcount", keyprefix);
+        if (ret < 0) {
+                goto out;
+        }
+
+        ret = dict_set_int32 (rsp_dict, key, volcount);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not save volcount");
+                goto out;
+        }
+
+out:
+
+        return ret;
+}
+
+int
+glusterd_get_each_snap_object_status (char **op_errstr, dict_t *rsp_dict,
+                                      glusterd_snap_t *snap, char *keyprefix)
+{
+        int                     ret             =       -1;
+        char                    key[PATH_MAX]   =       "";
+        char                    *temp           =       NULL;
+        xlator_t                *this           =       NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (snap);
+        GF_ASSERT (keyprefix);
+
+        /* TODO : Get all the snap volume info present in snap object,
+         * as of now, There will be only one snapvolinfo per snap object
+         */
+        ret = snprintf (key, sizeof (key), "%s.snapname", keyprefix);
+        if (ret < 0) {
+                goto out;
+        }
+
+        temp = gf_strdup (snap->snapname);
+        if (temp == NULL) {
+                ret = -1;
+                goto out;
+        }
+        ret = dict_set_dynstr (rsp_dict, key, temp);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not save "
+                        "snap name");
+                goto out;
+        }
+
+        temp = NULL;
+
+        ret = snprintf (key, sizeof (key), "%s.uuid", keyprefix);
+        if (ret < 0) {
+                goto out;
+        }
+
+        temp = gf_strdup (uuid_utoa (snap->snap_id));
+        if (temp == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_dynstr (rsp_dict, key, temp);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not save "
+                        "snap UUID");
+                goto out;
+        }
+
+        temp = NULL;
+
+        ret = glusterd_get_single_snap_status (op_errstr, rsp_dict, keyprefix,
+                                               snap);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not get single snap status");
+                goto out;
+        }
+
+        ret = snprintf (key, sizeof (key), "%s.volcount", keyprefix);
+        if (ret < 0) {
+                goto out;
+        }
+
+        ret = dict_set_int32 (rsp_dict, key, 1);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not save volcount");
+                goto out;
+        }
+out:
+        if (ret && temp)
+                GF_FREE (temp);
+
+        return ret;
+}
+
+int
+glusterd_get_snap_status_of_volume (char **op_errstr, dict_t *rsp_dict,
+                                    char *volname, char *keyprefix) {
+        int                     ret              =       -1;
+        glusterd_volinfo_t      *snap_volinfo    =       NULL;
+        glusterd_volinfo_t      *temp_volinfo    =       NULL;
+        glusterd_volinfo_t      *volinfo         =       NULL;
+        char                    key[PATH_MAX]    =        "";
+        xlator_t                *this            =       NULL;
+        glusterd_conf_t         *priv            =       NULL;
+        int                     i                =        0;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (volname);
+        GF_ASSERT (keyprefix);
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get volinfo of "
+                        "volume %s", volname);
+                goto out;
+        }
+
+        list_for_each_entry_safe (snap_volinfo, temp_volinfo,
+                             &volinfo->snap_volumes, snapvol_list) {
+                ret = snprintf (key, sizeof (key), "status.snap%d", i);
+                if (ret < 0) {
+                        goto out;
+                }
+
+                ret = glusterd_get_each_snap_object_status (op_errstr,
+                                       rsp_dict, snap_volinfo->snapshot, key);
+
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Function : "
+                                "glusterd_get_single_snap_status failed");
+                        goto out;
+                }
+                i++;
+        }
+
+        ret = dict_set_int32 (rsp_dict, "status.snapcount", i);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to save snapcount");
+                ret = -1;
+                goto out;
+        }
+out:
+        return ret;
+}
+
+int
+glusterd_get_all_snapshot_status (dict_t *dict, char **op_errstr,
+                              dict_t *rsp_dict)
+{
+        int32_t                 i               =       0;
+        int                     ret             =       -1;
+        char                    key[PATH_MAX]   =       "";
+        glusterd_conf_t         *priv           =       NULL;
+        glusterd_snap_t         *snap           =       NULL;
+        glusterd_snap_t         *tmp_snap       =       NULL;
+        xlator_t                *this           =       NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+
+        list_for_each_entry_safe (snap, tmp_snap,
+                                  &priv->snapshots, snap_list) {
+                ret = snprintf (key, sizeof (key), "status.snap%d", i);
+                if (ret < 0) {
+                        goto out;
+                }
+
+                ret = glusterd_get_each_snap_object_status (op_errstr,
+                                                         rsp_dict, snap, key);
+
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Could not get "
+                                "the details of a snap object: %s",
+                                snap->snapname);
+                        goto out;
+                }
+                i++;
+        }
+
+        ret = dict_set_int32 (rsp_dict, "status.snapcount", i);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not save snapcount");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+int
+glusterd_snapshot_status_commit (dict_t *dict, char **op_errstr,
+                                 dict_t *rsp_dict)
+{
+        xlator_t        *this             =       NULL;
+        int             ret               =       -1;
+        glusterd_conf_t *conf             =       NULL;
+        char            *get_buffer       =       NULL;
+        int32_t         cmd               =       -1;
+        char            *snapname         =       NULL;
+        glusterd_snap_t *snap             =       NULL;
+        char            *volname          =       NULL;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (op_errstr);
+
+        conf = this->private;
+
+        GF_ASSERT (conf);
+        ret = dict_get_int32 (dict, "cmd", &cmd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get status cmd type");
+                goto out;
+        }
+
+        ret = dict_set_int32 (rsp_dict, "cmd", cmd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not save status cmd in rsp dictionary");
+                goto out;
+        }
+        switch (cmd) {
+                case GF_SNAP_STATUS_TYPE_ALL:
+                {
+                        ret = glusterd_get_all_snapshot_status (dict, op_errstr,
+                                                           rsp_dict);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Unable to "
+                                        "get snapshot status");
+                                goto out;
+                        }
+                        break;
+                }
+                case GF_SNAP_STATUS_TYPE_SNAP:
+                {
+
+                        ret = dict_get_str (dict, "snapname", &snapname);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Unable to "
+                                        "get snap name");
+                                goto out;
+                        }
+
+                        snap = glusterd_find_snap_by_name (snapname);
+                        if (!snap) {
+                                ret = gf_asprintf (op_errstr, "Snap (%s) "
+                                                  "not found", snapname);
+                                if (ret < 0) {
+                                        goto out;
+                                }
+                                ret = -1;
+                                gf_log (this->name, GF_LOG_ERROR, "Unable to "
+                                        "get snap volinfo");
+                                goto out;
+                        }
+                        ret = glusterd_get_each_snap_object_status (op_errstr,
+                                          rsp_dict, snap, "status.snap0");
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Unable to "
+                                        "get status of snap %s", get_buffer);
+                                goto out;
+                        }
+                        break;
+                }
+                case GF_SNAP_STATUS_TYPE_VOL:
+                {
+                        ret = dict_get_str (dict, "volname", &volname);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Unable to"
+                                        " get volume name");
+                                goto out;
+                        }
+
+                        ret = glusterd_get_snap_status_of_volume (op_errstr,
+                                           rsp_dict, volname, "status.vol0");
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Function :"
+                                        " glusterd_get_snap_status_of_volume "
+                                        "failed");
+                                goto out;
+                        }
+                }
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+glusterd_snapshot_create_postvalidate (dict_t *dict, int32_t op_ret,
+                                       char **op_errstr, dict_t *rsp_dict)
+{
+        xlator_t        *this           = NULL;
+        glusterd_conf_t *priv           = NULL;
+        int              ret            = -1;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (rsp_dict);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        if (op_ret) {
+                ret = glusterd_do_snap_cleanup (dict, op_errstr, rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "cleanup operation "
+                                "failed");
+                        goto out;
+                }
+        } else {
+                ret = glusterd_snapshot_update_snaps_post_validate (dict,
+                                                                    op_errstr,
+                                                                    rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                "create snapshot");
+                        goto out;
+                }
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+glusterd_snapshot (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
+{
+
+        xlator_t        *this           = NULL;
+        glusterd_conf_t *priv           = NULL;
+        int32_t          snap_command   = 0;
+        int              ret            = -1;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (rsp_dict);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_int32 (dict, "type", &snap_command);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "unable to get the type of "
+                        "the snapshot command");
+                goto out;
+        }
+
+        switch (snap_command) {
+        case (GF_SNAP_OPTION_TYPE_CREATE):
+                ret = glusterd_snapshot_create_commit (dict, op_errstr,
+                                                       rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                "create snapshot");
+                        goto out;
+                }
+                break;
+
+        case GF_SNAP_OPTION_TYPE_CONFIG:
+                ret = glusterd_snapshot_config_commit (dict, op_errstr,
+                                                       rsp_dict);
+                break;
+
+        case GF_SNAP_OPTION_TYPE_DELETE:
+                ret = glusterd_snapshot_remove_commit (dict, op_errstr,
+                                                       rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                "delete snapshot");
+                        goto out;
+                }
+                break;
+
+        case GF_SNAP_OPTION_TYPE_RESTORE:
+                ret = glusterd_snapshot_restore (dict, op_errstr,
+                                                 rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Failed to "
+                                "restore snapshot");
+                        goto out;
+                }
+
+                break;
+
+        case GF_SNAP_OPTION_TYPE_STATUS:
+                ret = glusterd_snapshot_status_commit (dict, op_errstr,
+                                                       rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                "show snapshot status");
+                        goto out;
+                }
+                break;
+
+
+        default:
+                gf_log (this->name, GF_LOG_WARNING, "invalid snap command");
+                goto out;
+                break;
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+int
+glusterd_snapshot_brickop (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
+{
+        int            ret       = -1;
+        int64_t        vol_count = 0;
+        int64_t        count     = 1;
+        char           key[1024] = {0,};
+        char           *volname  = NULL;
+        int32_t        snap_command = 0;
+        xlator_t       *this     = NULL;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (rsp_dict);
+
+        ret = dict_get_int32 (dict, "type", &snap_command);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "unable to get the type of "
+                        "the snapshot command");
+                goto out;
+        }
+
+        switch (snap_command) {
+        case GF_SNAP_OPTION_TYPE_CREATE:
+                ret = dict_get_int64 (dict, "volcount", &vol_count);
+                if (ret)
+                        goto out;
+                while (count <= vol_count) {
+                        snprintf (key, 1024, "volname%"PRId64, count);
+                        ret = dict_get_str (dict, key, &volname);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to get volname");
+                                goto out;
+                        }
+                        ret = dict_set_str (dict, "volname", volname);
+                        if (ret)
+                                goto out;
+
+                        ret = gd_brick_op_phase (GD_OP_SNAP, NULL, dict,
+                                        op_errstr);
+                        if (ret)
+                                goto out;
+                        volname = NULL;
+                        count++;
+                }
+
+                dict_del (dict, "volname");
+                ret = 0;
+                break;
+        case GF_SNAP_OPTION_TYPE_DELETE:
+                break;
+        default:
+                break;
+        }
+
+out:
+        return ret;
+}
+
+int
+glusterd_snapshot_prevalidate (dict_t *dict, char **op_errstr,
+                               dict_t *rsp_dict)
+{
+        int                snap_command          = 0;
+        xlator_t           *this                 = NULL;
+        int                ret                   = -1;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (rsp_dict);
+
+        ret = dict_get_int32 (dict, "type", &snap_command);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "unable to get the type of "
+                        "the snapshot command");
+                goto out;
+        }
+
+        switch (snap_command) {
+        case (GF_SNAP_OPTION_TYPE_CREATE):
+                ret = glusterd_snapshot_create_prevalidate (dict, op_errstr,
+                                                            rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot create "
+                                "pre-validation failed");
+                        goto out;
+                }
+                break;
+
+        case (GF_SNAP_OPTION_TYPE_CONFIG):
+                ret = glusterd_snapshot_config_prevalidate (dict, op_errstr);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot config "
+                                "pre-validation failed");
+                        goto out;
+                }
+                break;
+
+        case GF_SNAP_OPTION_TYPE_RESTORE:
+                ret = glusterd_snapshot_restore_prevalidate (dict, op_errstr,
+                                                             rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot restore "
+                                        "validation failed");
+                        goto out;
+                }
+                break;
+        case GF_SNAP_OPTION_TYPE_DELETE:
+                ret = glusterd_snapshot_remove_prevalidate (dict, op_errstr,
+                                                            rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot remove "
+                                "validation failed");
+                        goto out;
+                }
+                break;
+
+        case GF_SNAP_OPTION_TYPE_STATUS:
+                ret = glusterd_snapshot_status_prevalidate (dict, op_errstr,
+                                                            rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot status "
+                                "validation failed");
+                        goto out;
+                }
+                break;
+
+        default:
+                gf_log (this->name, GF_LOG_WARNING, "invalid snap command");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_snapshot_postvalidate (dict_t *dict, int32_t op_ret, char **op_errstr,
+                                dict_t *rsp_dict)
+{
+        int                snap_command          = 0;
+        xlator_t           *this                 = NULL;
+        int                ret                   = -1;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (dict);
+        GF_ASSERT (rsp_dict);
+
+        ret = dict_get_int32 (dict, "type", &snap_command);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "unable to get the type of "
+                        "the snapshot command");
+                goto out;
+        }
+
+        switch (snap_command) {
+        case GF_SNAP_OPTION_TYPE_CREATE:
+                ret = glusterd_snapshot_create_postvalidate (dict, op_ret,
+                                                             op_errstr,
+                                                             rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot create "
+                                "post-validation failed");
+                        goto out;
+                }
+                break;
+
+        case GF_SNAP_OPTION_TYPE_DELETE:
+        case GF_SNAP_OPTION_TYPE_RESTORE:
+                ret = glusterd_snapshot_update_snaps_post_validate (dict,
+                                                                    op_errstr,
+                                                                    rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                "update missed snaps list");
+                        goto out;
+                }
+                break;
+
+        default:
+                gf_log (this->name, GF_LOG_WARNING, "invalid snap command");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_handle_snapshot_fn (rpcsvc_request_t *req)
+{
+        int32_t               ret            = 0;
+        dict_t               *dict           = NULL;
+        gf_cli_req            cli_req        = {{0},};
+        glusterd_op_t         cli_op         = GD_OP_SNAP;
+        int                   type           = 0;
+        glusterd_conf_t      *conf           = NULL;
+        char                 *host_uuid      = NULL;
+        char                  err_str[2048]  = {0,};
+        xlator_t             *this           = NULL;
+        char                 *volname        = NULL;
+
+        GF_ASSERT (req);
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req,
+                              (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (cli_req.dict.dict_len > 0) {
+                dict = dict_new ();
+                if (!dict)
+                        goto out;
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
+                        goto out;
+                }
+
+                dict->extra_stdfree = cli_req.dict.dict_val;
+
+                host_uuid = gf_strdup (uuid_utoa(MY_UUID));
+                if (host_uuid == NULL) {
+                        snprintf (err_str, sizeof (err_str), "Failed to get "
+                                  "the uuid of local glusterd");
+                        ret = -1;
+                        goto out;
+                }
+                ret = dict_set_dynstr (dict, "host-uuid", host_uuid);
+                if (ret) {
+                        GF_FREE (host_uuid);
+                        goto out;
+                }
+
+
+        } else {
+                gf_log (this->name, GF_LOG_ERROR, "request dict length is %d",
+                        cli_req.dict.dict_len);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "type", &type);
+        if (ret < 0) {
+                snprintf (err_str, sizeof (err_str), "Command type not found");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        switch (type) {
+        case GF_SNAP_OPTION_TYPE_CREATE:
+                ret = glusterd_handle_snapshot_create (req, cli_op, dict,
+                                                err_str, sizeof (err_str));
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot create "
+                                        "failed: %s", err_str);
+                }
+                break;
+        case GF_SNAP_OPTION_TYPE_RESTORE:
+                ret = glusterd_handle_snapshot_restore (req, cli_op, dict,
+                                                err_str, sizeof (err_str));
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot restore "
+                                        "failed: %s", err_str);
+                }
+
+                break;
+        case GF_SNAP_OPTION_TYPE_INFO:
+                ret = glusterd_handle_snapshot_info (req, cli_op, dict,
+                                                err_str, sizeof (err_str));
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot info "
+                                        "failed");
+                }
+                break;
+        case GF_SNAP_OPTION_TYPE_LIST:
+                ret = glusterd_handle_snapshot_list (req, cli_op, dict,
+                                                err_str, sizeof (err_str));
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot list "
+                                        "failed");
+                }
+                break;
+        case GF_SNAP_OPTION_TYPE_CONFIG:
+                /* TODO : Type of lock to be taken when we are setting
+                 * limits system wide
+                 */
+                ret = dict_get_str (dict, "volname", &volname);
+                if (!volname) {
+                        ret = dict_set_int32 (dict, "hold_vol_locks",
+                                              _gf_false);
+                        if (ret) {
+                                gf_log ("cli", GF_LOG_ERROR,
+                                        "Unable to set hold_vol_locks value "
+                                        "as _gf_false");
+                                goto out;
+                        }
+
+                }
+                ret = glusterd_mgmt_v3_initiate_all_phases (req, cli_op, dict);
+                break;
+        case GF_SNAP_OPTION_TYPE_DELETE:
+                ret = glusterd_handle_snapshot_remove (req, cli_op, dict,
+                                                       err_str,
+                                                       sizeof (err_str));
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot delete "
+                                "failed: %s", err_str);
+                }
+                break;
+        case GF_SNAP_OPTION_TYPE_START:
+        case GF_SNAP_OPTION_TYPE_STOP:
+        case GF_SNAP_OPTION_TYPE_STATUS:
+                ret = glusterd_handle_snapshot_status (req, cli_op, dict,
+                                                       err_str,
+                                                       sizeof (err_str));
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING, "Snapshot status "
+                                "failed: %s", err_str);
+                }
+                break;
+        default:
+                gf_log (this->name, GF_LOG_ERROR, "Unkown snapshot request "
+                        "type (%d)", type);
+                ret = -1; /* Failure */
+        }
+
+out:
+        if (ret) {
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
+        }
+
+        return ret;
+}
+
+int
+glusterd_handle_snapshot (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req, glusterd_handle_snapshot_fn);
+}
+
+static inline void
+glusterd_free_snap_op (glusterd_snap_op_t *snap_op)
+{
+        if (snap_op) {
+                if (snap_op->brick_path)
+                        GF_FREE (snap_op->brick_path);
+
+                GF_FREE (snap_op);
+        }
+}
+
+/* Look for duplicates and accordingly update the list */
+int32_t
+glusterd_update_missed_snap_entry (glusterd_missed_snap_info *missed_snapinfo,
+                                   glusterd_snap_op_t *missed_snap_op)
+{
+        int32_t                        ret                         = -1;
+        glusterd_snap_op_t            *snap_opinfo                 = NULL;
+        gf_boolean_t                   match                       = _gf_false;
+        xlator_t                      *this                        = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+        GF_ASSERT(missed_snapinfo);
+        GF_ASSERT(missed_snap_op);
+
+        list_for_each_entry (snap_opinfo, &missed_snapinfo->snap_ops,
+                             snap_ops_list) {
+                if ((!strcmp (snap_opinfo->brick_path,
+                              missed_snap_op->brick_path)) &&
+                    (snap_opinfo->op == missed_snap_op->op)) {
+                        /* If two entries have conflicting status
+                         * GD_MISSED_SNAP_DONE takes precedence
+                         */
+                        if ((snap_opinfo->status == GD_MISSED_SNAP_PENDING) &&
+                            (missed_snap_op->status == GD_MISSED_SNAP_DONE)) {
+                                snap_opinfo->status = GD_MISSED_SNAP_DONE;
+                                gf_log (this->name, GF_LOG_INFO,
+                                        "Updating missed snap status "
+                                        "for %s:%d:%s:%d as DONE",
+                                        missed_snapinfo->node_snap_info,
+                                        snap_opinfo->brick_num,
+                                        snap_opinfo->brick_path,
+                                        snap_opinfo->op);
+                                ret = 0;
+                                glusterd_free_snap_op (missed_snap_op);
+                                goto out;
+                        }
+                        match = _gf_true;
+                        break;
+                } else if ((snap_opinfo->brick_num ==
+                            missed_snap_op->brick_num) &&
+                            (snap_opinfo->op == GF_SNAP_OPTION_TYPE_CREATE) &&
+                            (missed_snap_op->op ==
+                             GF_SNAP_OPTION_TYPE_DELETE)) {
+                        /* Optimizing create and delete entries for the same
+                         * brick and same node
+                         */
+                        gf_log (this->name, GF_LOG_INFO,
+                                "Updating missed snap status "
+                                "for %s:%d:%s:%d as DONE",
+                                missed_snapinfo->node_snap_info,
+                                snap_opinfo->brick_num,
+                                snap_opinfo->brick_path,
+                                snap_opinfo->op);
+                        snap_opinfo->status = GD_MISSED_SNAP_DONE;
+                        ret = 0;
+                        glusterd_free_snap_op (missed_snap_op);
+                        goto out;
+                }
+        }
+
+        if (match == _gf_true) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Duplicate entry. Not updating");
+                glusterd_free_snap_op (missed_snap_op);
+        } else {
+                list_add_tail (&missed_snap_op->snap_ops_list,
+                               &missed_snapinfo->snap_ops);
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+/* Add new missed snap entry to the missed_snaps list. */
+int32_t
+glusterd_store_missed_snaps_list (char *missed_info, int32_t brick_num,
+                                  char *brick_path, int32_t snap_op,
+                                  int32_t snap_status)
+{
+        int32_t                        ret                         = -1;
+        glusterd_missed_snap_info     *missed_snapinfo             = NULL;
+        glusterd_snap_op_t            *missed_snap_op              = NULL;
+        glusterd_conf_t               *priv                        = NULL;
+        gf_boolean_t                   match                       = _gf_false;
+        gf_boolean_t                   free_missed_snap_info       = _gf_false;
+        xlator_t                      *this                        = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+        GF_ASSERT(missed_info);
+        GF_ASSERT(brick_path);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        /* Create the snap_op object consisting of the *
+         * snap id and the op */
+        ret = glusterd_missed_snap_op_new (&missed_snap_op);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to create new missed snap object.");
+                ret = -1;
+                goto out;
+        }
+
+        missed_snap_op->brick_path = gf_strdup(brick_path);
+        if (!missed_snap_op->brick_path) {
+                ret = -1;
+                goto out;
+        }
+        missed_snap_op->brick_num = brick_num;
+        missed_snap_op->op = snap_op;
+        missed_snap_op->status = snap_status;
+
+        /* Look for other entries for the same node and same snap */
+        list_for_each_entry (missed_snapinfo, &priv->missed_snaps_list,
+                             missed_snaps) {
+                if (!strcmp (missed_snapinfo->node_snap_info,
+                             missed_info)) {
+                        /* Found missed snapshot info for *
+                         * the same node and same snap */
+                        match = _gf_true;
+                        break;
+                }
+        }
+
+        if (match == _gf_false) {
+                /* First snap op missed for the brick */
+                ret = glusterd_missed_snapinfo_new (&missed_snapinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to create missed snapinfo");
+                        goto out;
+                }
+                free_missed_snap_info = _gf_true;
+                missed_snapinfo->node_snap_info = gf_strdup(missed_info);
+                if (!missed_snapinfo->node_snap_info) {
+                        ret = -1;
+                        goto out;
+                }
+
+                list_add_tail (&missed_snap_op->snap_ops_list,
+                               &missed_snapinfo->snap_ops);
+                list_add_tail (&missed_snapinfo->missed_snaps,
+                               &priv->missed_snaps_list);
+
+                ret = 0;
+                goto out;
+        } else {
+                ret = glusterd_update_missed_snap_entry (missed_snapinfo,
+                                                         missed_snap_op);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to update existing missed snap entry.");
+                        goto out;
+                }
+        }
+
+out:
+        if (ret) {
+                glusterd_free_snap_op (missed_snap_op);
+
+                if (missed_snapinfo &&
+                    (free_missed_snap_info == _gf_true)) {
+                        if (missed_snapinfo->node_snap_info)
+                                GF_FREE (missed_snapinfo->node_snap_info);
+
+                        GF_FREE (missed_snapinfo);
+                }
+        }
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+/* Add  missing snap entries to the in-memory conf->missed_snap_list */
+int32_t
+glusterd_add_missed_snaps_to_list (dict_t *dict, int32_t missed_snap_count)
+{
+        char                          *buf                         = NULL;
+        char                          *tmp                         = NULL;
+        char                          *save_ptr                    = NULL;
+        char                          *nodeid                      = NULL;
+        char                          *snap_uuid                   = NULL;
+        char                          *brick_path                  = NULL;
+        char                           missed_info[PATH_MAX]       = "";
+        char                           name_buf[PATH_MAX]          = "";
+        int32_t                        i                           = -1;
+        int32_t                        ret                         = -1;
+        int32_t                        brick_num                   = -1;
+        int32_t                        snap_op                     = -1;
+        int32_t                        snap_status                 = -1;
+        glusterd_conf_t               *priv                        = NULL;
+        xlator_t                      *this                        = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+        GF_ASSERT(dict);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        /* We can update the missed_snaps_list without acquiring *
+         * any additional locks as big lock will be held.        */
+        for (i = 0; i < missed_snap_count; i++) {
+                snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d",
+                          i);
+                ret = dict_get_str (dict, name_buf, &buf);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to fetch %s", name_buf);
+                        goto out;
+                }
+
+                gf_log (this->name, GF_LOG_DEBUG, "missed_snap_entry = %s",
+                        buf);
+
+                /* Need to make a duplicate string coz the same dictionary *
+                 * is resent to the non-originator nodes */
+                tmp = gf_strdup (buf);
+                if (!tmp) {
+                        ret = -1;
+                        goto out;
+                }
+
+                /* Fetch the node-id, snap-id, brick_num,
+                 * brick_path, snap_op and snap status
+                 */
+                nodeid = strtok_r (tmp, ":", &save_ptr);
+                snap_uuid = strtok_r (NULL, "=", &save_ptr);
+                brick_num = atoi(strtok_r (NULL, ":", &save_ptr));
+                brick_path = strtok_r (NULL, ":", &save_ptr);
+                snap_op = atoi(strtok_r (NULL, ":", &save_ptr));
+                snap_status = atoi(strtok_r (NULL, ":", &save_ptr));
+
+                if (!nodeid || !snap_uuid || !brick_path ||
+                    brick_num < 1 || snap_op < 1 ||
+                    snap_status < 1) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Invalid missed_snap_entry");
+                        ret = -1;
+                        goto out;
+                }
+
+                snprintf (missed_info, sizeof(missed_info), "%s:%s",
+                          nodeid, snap_uuid);
+
+                ret = glusterd_store_missed_snaps_list (missed_info,
+                                                        brick_num,
+                                                        brick_path,
+                                                        snap_op,
+                                                        snap_status);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to store missed snaps_list");
+                        goto out;
+                }
+
+                GF_FREE (tmp);
+                tmp = NULL;
+        }
+
+out:
+        if (tmp)
+                GF_FREE (tmp);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-store.c b/xlators/mgmt/glusterd/src/glusterd-store.c
index 498a063a5..1c2ec58e8 100644
--- a/xlators/mgmt/glusterd/src/glusterd-store.c
+++ b/xlators/mgmt/glusterd/src/glusterd-store.c
@@ -1,25 +1,19 @@
 /*
-  Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+   Copyright (c) 2007-2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
+
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
 #endif
+
+#include "glusterd-op-sm.h"
 #include <inttypes.h>
 
 
@@ -40,150 +34,311 @@
 #include "glusterd-sm.h"
 #include "glusterd-op-sm.h"
 #include "glusterd-utils.h"
+#include "glusterd-hooks.h"
+#include "store.h"
 #include "glusterd-store.h"
 
-#include "glusterd1.h"
-#include "cli1.h"
 #include "rpc-clnt.h"
+#include "common-utils.h"
 
 #include <sys/resource.h>
 #include <inttypes.h>
 #include <dirent.h>
 
-static int32_t
-glusterd_store_create_volume_dir (char *volname)
+void
+glusterd_replace_slash_with_hyphen (char *str)
+{
+        char                    *ptr = NULL;
+
+        ptr = strchr (str, '/');
+
+        while (ptr) {
+                *ptr = '-';
+                ptr = strchr (str, '/');
+        }
+}
+
+int32_t
+glusterd_store_create_brick_dir (glusterd_volinfo_t *volinfo)
 {
         int32_t                 ret = -1;
-        char                    path[PATH_MAX] = {0,};
+        char                    brickdirpath[PATH_MAX] = {0,};
         glusterd_conf_t         *priv = NULL;
 
-        GF_ASSERT (volname);
+        GF_ASSERT (volinfo);
+
         priv = THIS->private;
+        GF_ASSERT (priv);
+
+        GLUSTERD_GET_BRICK_DIR (brickdirpath, volinfo, priv);
+        ret = gf_store_mkdir (brickdirpath);
+
+        return ret;
+}
+
+static void
+glusterd_store_key_vol_brick_set (glusterd_brickinfo_t *brickinfo,
+                                char *key_vol_brick, size_t len)
+{
+        GF_ASSERT (brickinfo);
+        GF_ASSERT (key_vol_brick);
+        GF_ASSERT (len >= PATH_MAX);
+
+        snprintf (key_vol_brick, len, "%s", brickinfo->path);
+        glusterd_replace_slash_with_hyphen (key_vol_brick);
+}
+
+static void
+glusterd_store_brickinfofname_set (glusterd_brickinfo_t *brickinfo,
+                                char *brickfname, size_t len)
+{
+        char                    key_vol_brick[PATH_MAX] = {0};
+
+        GF_ASSERT (brickfname);
+        GF_ASSERT (brickinfo);
+        GF_ASSERT (len >= PATH_MAX);
 
+        glusterd_store_key_vol_brick_set (brickinfo, key_vol_brick,
+                                        sizeof (key_vol_brick));
+        snprintf (brickfname, len, "%s:%s", brickinfo->hostname, key_vol_brick);
+}
+
+static void
+glusterd_store_brickinfopath_set (glusterd_volinfo_t *volinfo,
+                              glusterd_brickinfo_t *brickinfo,
+                              char *brickpath, size_t len)
+{
+        char                    brickfname[PATH_MAX] = {0};
+        char                    brickdirpath[PATH_MAX] = {0,};
+        glusterd_conf_t         *priv = NULL;
+
+        GF_ASSERT (brickpath);
+        GF_ASSERT (brickinfo);
+        GF_ASSERT (len >= PATH_MAX);
+
+        priv = THIS->private;
         GF_ASSERT (priv);
 
-        snprintf (path, 1024, "%s/vols/%s", priv->workdir,
-                  volname);
+        GLUSTERD_GET_BRICK_DIR (brickdirpath, volinfo, priv);
+        glusterd_store_brickinfofname_set (brickinfo, brickfname,
+                                        sizeof (brickfname));
+        snprintf (brickpath, len, "%s/%s", brickdirpath, brickfname);
+}
+
+gf_boolean_t
+glusterd_store_is_valid_brickpath (char *volname, char *brick)
+{
+        char                    brickpath[PATH_MAX] = {0};
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        glusterd_volinfo_t      *volinfo = NULL;
+        int32_t                 ret = 0;
+        size_t                  volname_len = strlen (volname);
+        xlator_t                *this = NULL;
 
-        ret = mkdir (path, 0777);
+        this = THIS;
+        GF_ASSERT (this);
 
-        if (-1 == ret) {
-                gf_log ("", GF_LOG_ERROR, "mkdir() failed on path %s,"
-                        "errno: %d", path, errno);
+        ret = glusterd_brickinfo_new_from_brick (brick, &brickinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "Failed to create brick "
+                        "info for brick %s", brick);
+                ret = 0;
+                goto out;
+        }
+        ret = glusterd_volinfo_new (&volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "Failed to create volinfo");
+                ret = 0;
+                goto out;
+        }
+        if (volname_len >= sizeof (volinfo->volname)) {
+                gf_log (this->name, GF_LOG_WARNING, "volume name too long");
+                ret = 0;
                 goto out;
         }
+        memcpy (volinfo->volname, volname, volname_len+1);
+        glusterd_store_brickinfopath_set (volinfo, brickinfo, brickpath,
+                                                sizeof (brickpath));
+
+        ret = (strlen (brickpath) < _POSIX_PATH_MAX);
 
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        if (brickinfo)
+                glusterd_brickinfo_delete (brickinfo);
+        if (volinfo)
+                glusterd_volinfo_delete (volinfo);
+
         return ret;
 }
 
 int32_t
-glusterd_store_create_brick (glusterd_volinfo_t *volinfo,
-                             glusterd_brickinfo_t *brickinfo)
+glusterd_store_volinfo_brick_fname_write (int vol_fd,
+                                         glusterd_brickinfo_t *brickinfo,
+                                         int32_t brick_count)
 {
+        char                    key[PATH_MAX] = {0,};
+        char                    brickfname[PATH_MAX] = {0,};
         int32_t                 ret = -1;
-        glusterd_conf_t         *priv = NULL;
-        char                    path[PATH_MAX] = {0,};
+
+        snprintf (key, sizeof (key), "%s-%d", GLUSTERD_STORE_KEY_VOL_BRICK,
+                  brick_count);
+        glusterd_store_brickinfofname_set (brickinfo, brickfname,
+                                        sizeof (brickfname));
+        ret = gf_store_save_value (vol_fd, key, brickfname);
+        if (ret)
+                goto out;
+
+out:
+        return ret;
+}
+
+int32_t
+glusterd_store_create_brick_shandle_on_absence (glusterd_volinfo_t *volinfo,
+                                                glusterd_brickinfo_t *brickinfo)
+{
         char                    brickpath[PATH_MAX] = {0,};
-        struct  stat            stbuf = {0,};
-        char                    buf[4096] = {0,};
-        char                    *tmppath = NULL;
-        char                    *ptr = NULL;
-        glusterd_store_handle_t *shandle = NULL;
+        int32_t                 ret = 0;
 
         GF_ASSERT (volinfo);
         GF_ASSERT (brickinfo);
 
-        priv = THIS->private;
+        glusterd_store_brickinfopath_set (volinfo, brickinfo, brickpath,
+                                          sizeof (brickpath));
+        ret = gf_store_handle_create_on_absence (&brickinfo->shandle,
+                                                 brickpath);
+        return ret;
+}
 
-        GF_ASSERT (priv);
+int32_t
+glusterd_store_brickinfo_write (int fd, glusterd_brickinfo_t *brickinfo)
+{
+        char                    value[256] = {0,};
+        int32_t                 ret = 0;
+
+        GF_ASSERT (brickinfo);
+        GF_ASSERT (fd > 0);
+
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_HOSTNAME,
+                                   brickinfo->hostname);
+        if (ret)
+                goto out;
+
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_PATH,
+                                   brickinfo->path);
+        if (ret)
+                goto out;
 
-        GLUSTERD_GET_BRICK_DIR (path, volinfo, priv);
+        snprintf (value, sizeof(value), "%d", brickinfo->port);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_PORT, value);
 
-        ret = stat (path, &stbuf);
+        snprintf (value, sizeof(value), "%d", brickinfo->rdma_port);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_RDMA_PORT,
+                                   value);
 
-        if (ret == -1 && ENOENT == errno) {
-                ret = mkdir (path, 0777);
+        snprintf (value, sizeof(value), "%d", brickinfo->decommissioned);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED,
+                                   value);
+        if (ret)
+                goto out;
+
+        if (strlen(brickinfo->device_path) > 0) {
+                snprintf (value, sizeof(value), "%s", brickinfo->device_path);
+                ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH,
+                                           value);
                 if (ret)
                         goto out;
         }
 
-        tmppath = gf_strdup (brickinfo->path);
+        snprintf (value, sizeof(value), "%d", brickinfo->snap_status);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS,
+                                   value);
+        if (ret)
+                goto out;
 
-        ptr = strchr (tmppath, '/');
+        if (!brickinfo->vg[0])
+                goto out;
 
-        while (ptr) {
-                *ptr = '-';
-                ptr = strchr (tmppath, '/');
-        }
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_VGNAME,
+                                         brickinfo->vg);
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
 
-        snprintf (brickpath, sizeof (brickpath), "%s/%s:%s",
-                  path, brickinfo->hostname, tmppath);
+int32_t
+glusterd_store_perform_brick_store (glusterd_brickinfo_t *brickinfo)
+{
+        int                         fd = -1;
+        int32_t                     ret = -1;
+        GF_ASSERT (brickinfo);
 
-        GF_FREE (tmppath);
-        ret = glusterd_store_handle_new (brickpath, &brickinfo->shandle);
+        fd = gf_store_mkstemp (brickinfo->shandle);
+        if (fd <= 0) {
+                ret = -1;
+                goto out;
+        }
 
+        ret = glusterd_store_brickinfo_write (fd, brickinfo);
         if (ret)
                 goto out;
 
-        shandle = brickinfo->shandle;
-        shandle->fd = open (brickpath, O_RDWR | O_CREAT | O_APPEND, 0666);
+out:
+        if (ret && (fd > 0))
+                gf_store_unlink_tmppath (brickinfo->shandle);
+        if (fd > 0)
+                close (fd);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
 
-        if (shandle->fd < 0) {
-                gf_log ("", GF_LOG_ERROR, "Open failed on %s",
-                        brickpath);
-                ret = -1;
-                goto out;
-        }
+int32_t
+glusterd_store_brickinfo (glusterd_volinfo_t *volinfo,
+                          glusterd_brickinfo_t *brickinfo, int32_t brick_count,
+                          int vol_fd)
+{
+        int32_t                 ret = -1;
 
+        GF_ASSERT (volinfo);
+        GF_ASSERT (brickinfo);
 
-        snprintf (buf, sizeof(buf), "%s=%s\n",
-                  GLUSTERD_STORE_KEY_BRICK_HOSTNAME, brickinfo->hostname);
-        ret = write (shandle->fd, buf, strlen(buf));
+        ret = glusterd_store_volinfo_brick_fname_write (vol_fd, brickinfo,
+                                                       brick_count);
         if (ret)
-                gf_log ("", GF_LOG_TRACE, "failed to write brick->hostname");
-        snprintf (buf, sizeof(buf), "%s=%s\n",
-                  GLUSTERD_STORE_KEY_BRICK_PATH, brickinfo->path);
-        ret = write (shandle->fd, buf, strlen(buf));
+                goto out;
+
+        ret = glusterd_store_create_brick_dir (volinfo);
         if (ret)
-                gf_log ("", GF_LOG_TRACE, "failed to write brick->path");
-        snprintf (buf, sizeof(buf), "%s=%d\n",
-                  GLUSTERD_STORE_KEY_BRICK_PORT, brickinfo->port);
-        ret = write (shandle->fd, buf, strlen(buf));
+                goto out;
+
+        ret = glusterd_store_create_brick_shandle_on_absence (volinfo,
+                                                              brickinfo);
         if (ret)
-                gf_log ("", GF_LOG_TRACE, "failed to write brick->port");
+                goto out;
 
-        ret = 0;
+        ret = glusterd_store_perform_brick_store (brickinfo);
 
 out:
-        if (shandle->fd > 0) {
-                close (shandle->fd);
-        }
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret);
         return ret;
 }
 
 int32_t
-glusterd_store_delete_brick (glusterd_volinfo_t *volinfo,
-                             glusterd_brickinfo_t *brickinfo)
+glusterd_store_delete_brick (glusterd_brickinfo_t *brickinfo, char *delete_path)
 {
         int32_t                 ret = -1;
         glusterd_conf_t         *priv = NULL;
-        char                    path[PATH_MAX] = {0,};
         char                    brickpath[PATH_MAX] = {0,};
         char                    *ptr = NULL;
         char                    *tmppath = NULL;
+        xlator_t                *this = NULL;
 
-        GF_ASSERT (volinfo);
+        this = THIS;
+        GF_ASSERT (this);
         GF_ASSERT (brickinfo);
 
-        priv = THIS->private;
-
+        priv = this->private;
         GF_ASSERT (priv);
 
-        GLUSTERD_GET_BRICK_DIR (path, volinfo, priv);
-
         tmppath = gf_strdup (brickinfo->path);
 
         ptr = strchr (tmppath, '/');
@@ -193,29 +348,34 @@ glusterd_store_delete_brick (glusterd_volinfo_t *volinfo,
                 ptr = strchr (tmppath, '/');
         }
 
-        snprintf (brickpath, sizeof (brickpath), "%s/%s:%s",
-                  path, brickinfo->hostname, tmppath);
+        snprintf (brickpath, sizeof (brickpath),
+                  "%s/"GLUSTERD_BRICK_INFO_DIR"/%s:%s", delete_path,
+                  brickinfo->hostname, tmppath);
 
         GF_FREE (tmppath);
 
         ret = unlink (brickpath);
 
-        if (ret < 0) {
-                gf_log ("", GF_LOG_ERROR, "Unlink failed on %s",
-                        brickpath);
+        if ((ret < 0) && (errno != ENOENT)) {
+                gf_log (this->name, GF_LOG_DEBUG, "Unlink failed on %s, "
+                        "reason: %s", brickpath, strerror(errno));
                 ret = -1;
                 goto out;
+        } else {
+                ret = 0;
         }
 
 out:
-        if (brickinfo->shandle)
-                glusterd_store_handle_destroy (brickinfo->shandle);
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        if (brickinfo->shandle) {
+                gf_store_handle_destroy (brickinfo->shandle);
+                brickinfo->shandle = NULL;
+        }
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
         return ret;
 }
 
 int32_t
-glusterd_store_remove_bricks (glusterd_volinfo_t *volinfo)
+glusterd_store_remove_bricks (glusterd_volinfo_t *volinfo, char *delete_path)
 {
         int32_t                 ret = 0;
         glusterd_brickinfo_t    *tmp = NULL;
@@ -224,19 +384,24 @@ glusterd_store_remove_bricks (glusterd_volinfo_t *volinfo)
         DIR                     *dir = NULL;
         struct dirent           *entry = NULL;
         char                    path[PATH_MAX] = {0,};
+        xlator_t                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
 
         GF_ASSERT (volinfo);
 
         list_for_each_entry (tmp, &volinfo->bricks, brick_list) {
-                ret = glusterd_store_delete_brick (volinfo, tmp);
+                ret = glusterd_store_delete_brick (tmp, delete_path);
                 if (ret)
                         goto out;
         }
 
-        priv = THIS->private;
+        priv = this->private;
         GF_ASSERT (priv);
 
-        GLUSTERD_GET_BRICK_DIR (brickdir, volinfo, priv);
+        snprintf (brickdir, sizeof (brickdir), "%s/%s", delete_path,
+                  GLUSTERD_BRICK_INFO_DIR);
 
         dir = opendir (brickdir);
 
@@ -246,9 +411,9 @@ glusterd_store_remove_bricks (glusterd_volinfo_t *volinfo)
                 snprintf (path, sizeof (path), "%s/%s",
                           brickdir, entry->d_name);
                 ret = unlink (path);
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to unlink %s",
-                                path);
+                if (ret && errno != ENOENT) {
+                        gf_log (this->name, GF_LOG_DEBUG, "Unable to unlink %s, "
+                                "reason: %s", path, strerror(errno));
                 }
                 glusterd_for_each_entry (entry, dir);
         }
@@ -258,587 +423,1482 @@ glusterd_store_remove_bricks (glusterd_volinfo_t *volinfo)
         ret = rmdir (brickdir);
 
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
         return ret;
 }
 
-int32_t
-glusterd_store_create_volume (glusterd_volinfo_t *volinfo)
+static int
+_storeslaves (dict_t *this, char *key, data_t *value, void *data)
 {
-        int32_t                 ret = -1;
-        char                    filepath[PATH_MAX] = {0,};
-        char                    buf[4096] = {0,};
-        glusterd_conf_t         *priv = NULL;
-        glusterd_brickinfo_t    *brickinfo = NULL;
+        int32_t                      ret = 0;
+        gf_store_handle_t           *shandle = NULL;
+        xlator_t                    *xl = NULL;
 
-        GF_ASSERT (volinfo);
-        priv = THIS->private;
+        xl = THIS;
+        GF_ASSERT (xl);
 
-        GF_ASSERT (priv);
+        shandle = (gf_store_handle_t*)data;
 
-        ret = glusterd_store_create_volume_dir (volinfo->volname);
+        GF_ASSERT (shandle);
+        GF_ASSERT (shandle->fd > 0);
+        GF_ASSERT (shandle->path);
+        GF_ASSERT (key);
+        GF_ASSERT (value && value->data);
 
-        if (ret)
-                goto out;
+        if ((!shandle) || (shandle->fd <= 0) || (!shandle->path))
+                return -1;
 
-        snprintf (filepath, 1024, "%s/%s/%s/%s", priv->workdir,
-                  GLUSTERD_VOLUME_DIR_PREFIX, volinfo->volname,
-                  GLUSTERD_VOLUME_INFO_FILE);
+        if (!key)
+                return -1;
+        if (!value || !value->data)
+                return -1;
 
-        ret = glusterd_store_handle_new (filepath, &volinfo->shandle);
+        gf_log (xl->name, GF_LOG_DEBUG, "Storing in volinfo:key= %s, val=%s",
+                key, value->data);
 
+        ret = gf_store_save_value (shandle->fd, key, (char*)value->data);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to create store"
-                        " handle for path: %s", filepath);
-                goto out;
+                gf_log (xl->name, GF_LOG_ERROR, "Unable to write into store"
+                                " handle for path: %s", shandle->path);
+                return -1;
+        }
+        return 0;
+}
+
+
+int _storeopts (dict_t *this, char *key, data_t *value, void *data)
+{
+        int32_t                      ret = 0;
+        int32_t                      exists = 0;
+        gf_store_handle_t           *shandle = NULL;
+        xlator_t                    *xl = NULL;
+
+        xl = THIS;
+        GF_ASSERT (xl);
+
+        shandle = (gf_store_handle_t*)data;
+
+        GF_ASSERT (shandle);
+        GF_ASSERT (shandle->fd > 0);
+        GF_ASSERT (shandle->path);
+        GF_ASSERT (key);
+        GF_ASSERT (value && value->data);
+
+        if ((!shandle) || (shandle->fd <= 0) || (!shandle->path))
+                return -1;
+
+        if (!key)
+                return -1;
+        if (!value || !value->data)
+                return -1;
+
+        if (is_key_glusterd_hooks_friendly (key)) {
+                exists = 1;
+
+        } else {
+                exists = glusterd_check_option_exists (key, NULL);
+        }
+
+        if (1 == exists) {
+                gf_log (xl->name, GF_LOG_DEBUG, "Storing in volinfo:key= %s, "
+                        "val=%s", key, value->data);
+
+        } else {
+                gf_log (xl->name, GF_LOG_DEBUG, "Discarding:key= %s, val=%s",
+                        key, value->data);
+                return 0;
         }
 
+        ret = gf_store_save_value (shandle->fd, key, (char*)value->data);
+        if (ret) {
+                gf_log (xl->name, GF_LOG_ERROR, "Unable to write into store"
+                        " handle for path: %s", shandle->path);
+                return -1;
+        }
+        return 0;
+}
+
+int32_t
+glusterd_volume_exclude_options_write (int fd, glusterd_volinfo_t *volinfo)
+{
+        char    *str            = NULL;
+        char    buf[PATH_MAX]   = {0,};
+        int32_t ret             = -1;
+
+        GF_ASSERT (fd > 0);
+        GF_ASSERT (volinfo);
+
         snprintf (buf, sizeof (buf), "%d", volinfo->type);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_TYPE, buf);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_TYPE, buf);
         if (ret)
                 goto out;
 
         snprintf (buf, sizeof (buf), "%d", volinfo->brick_count);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_COUNT, buf);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_COUNT, buf);
         if (ret)
                 goto out;
 
         snprintf (buf, sizeof (buf), "%d", volinfo->status);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_STATUS, buf);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_STATUS, buf);
         if (ret)
                 goto out;
 
-/*        snprintf (buf, sizeof (buf), "%d", volinfo->port);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_PORT, buf);
+        snprintf (buf, sizeof (buf), "%d", volinfo->sub_count);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_SUB_COUNT, buf);
         if (ret)
                 goto out;
-*/
-        snprintf (buf, sizeof (buf), "%d", volinfo->sub_count);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_SUB_COUNT, buf);
+
+        snprintf (buf, sizeof (buf), "%d", volinfo->stripe_count);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_STRIPE_CNT, buf);
+        if (ret)
+                goto out;
+
+        snprintf (buf, sizeof (buf), "%d", volinfo->replica_count);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_REPLICA_CNT,
+                                   buf);
         if (ret)
                 goto out;
 
         snprintf (buf, sizeof (buf), "%d", volinfo->version);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_VERSION, buf);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_VERSION, buf);
         if (ret)
                 goto out;
 
         snprintf (buf, sizeof (buf), "%d", volinfo->transport_type);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                         GLUSTERD_STORE_KEY_VOL_TRANSPORT, buf);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_TRANSPORT, buf);
         if (ret)
                 goto out;
 
-        uuid_unparse (volinfo->volume_id, buf);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_ID, buf);
+        snprintf (buf, sizeof (buf), "%s", volinfo->parent_volname);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_PARENT_VOLNAME, buf);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to store "
+                        GLUSTERD_STORE_KEY_PARENT_VOLNAME);
+                goto out;
+        }
+
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_ID,
+                                   uuid_utoa (volinfo->volume_id));
         if (ret)
                 goto out;
 
-        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
-                ret = glusterd_store_create_brick (volinfo, brickinfo);
+        str = glusterd_auth_get_username (volinfo);
+        if (str) {
+                ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_USERNAME,
+                                           str);
+                if (ret)
+                goto out;
+        }
+
+        str = glusterd_auth_get_password (volinfo);
+        if (str) {
+                ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_PASSWORD,
+                                           str);
                 if (ret)
                         goto out;
         }
 
-        ret = 0;
+        snprintf (buf, sizeof (buf), "%d", volinfo->op_version);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_OP_VERSION, buf);
+        if (ret)
+                goto out;
 
-out:
+        snprintf (buf, sizeof (buf), "%d", volinfo->client_op_version);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION,
+                                   buf);
+        if (ret)
+                goto out;
+        if (volinfo->caps) {
+                snprintf (buf, sizeof (buf), "%d", volinfo->caps);
+                ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_CAPS,
+                                           buf);
+                if (ret)
+                        goto out;
+        }
+
+        snprintf (buf, sizeof (buf), "%d", volinfo->is_volume_restored);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_IS_RESTORED, buf);
         if (ret) {
-                glusterd_store_delete_volume (volinfo);
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "Unable to write is_volume_restored");
+                goto out;
         }
 
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        snprintf (buf, sizeof (buf), "%"PRIu64, volinfo->snap_max_hard_limit);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT,
+                                   buf);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "Unable to write snap-max-hard-limit");
+                goto out;
+        }
 
+out:
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR, "Unable to write volume "
+                        "values for %s", volinfo->volname);
         return ret;
 }
 
-
-int32_t
-glusterd_store_delete_volume (glusterd_volinfo_t *volinfo)
+static void
+glusterd_store_voldirpath_set (glusterd_volinfo_t *volinfo, char *voldirpath,
+                               size_t len)
 {
-        char    pathname[PATH_MAX] = {0,};
-        int32_t ret = 0;
-        glusterd_conf_t *priv = NULL;
-        DIR     *dir = NULL;
-        struct dirent *entry = NULL;
-        char path[PATH_MAX] = {0,};
-        struct stat     st = {0, };
+        glusterd_conf_t         *priv = NULL;
 
         GF_ASSERT (volinfo);
         priv = THIS->private;
+        GF_ASSERT (priv);
+
+        GLUSTERD_GET_VOLUME_DIR (voldirpath, volinfo, priv);
+}
+
+static int32_t
+glusterd_store_create_volume_dir (glusterd_volinfo_t *volinfo)
+{
+        int32_t                 ret = -1;
+        char                    voldirpath[PATH_MAX] = {0,};
+
+        GF_ASSERT (volinfo);
+
+        glusterd_store_voldirpath_set (volinfo, voldirpath,
+                                       sizeof (voldirpath));
+        ret = gf_store_mkdir (voldirpath);
+
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
 
+static int32_t
+glusterd_store_create_snap_dir (glusterd_snap_t *snap)
+{
+        int32_t            ret                   = -1;
+        char               snapdirpath[PATH_MAX] = {0,};
+        glusterd_conf_t   *priv                  = NULL;
+
+        priv = THIS->private;
         GF_ASSERT (priv);
-        snprintf (pathname, 1024, "%s/vols/%s", priv->workdir,
-                  volinfo->volname);
+        GF_ASSERT (snap);
 
-        dir = opendir (pathname);
-        if (!dir)
-                goto out;
-        ret = glusterd_store_remove_bricks (volinfo);
+        GLUSTERD_GET_SNAP_DIR (snapdirpath, snap, priv);
 
+        ret = mkdir_p (snapdirpath, 0755, _gf_true);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Remove bricks failed");
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to create snaps dir "
+                        "%s", snapdirpath);
         }
+        return ret;
+}
 
-        glusterd_for_each_entry (entry, dir);
-        while (entry) {
+int32_t
+glusterd_store_volinfo_write (int fd, glusterd_volinfo_t *volinfo)
+{
+        int32_t                         ret = -1;
+        gf_store_handle_t              *shandle = NULL;
+        GF_ASSERT (fd > 0);
+        GF_ASSERT (volinfo);
+        GF_ASSERT (volinfo->shandle);
 
-                snprintf (path, PATH_MAX, "%s/%s", pathname, entry->d_name);
-                ret = stat (path, &st);
-                if (ret == -1) {
-                        gf_log ("", GF_LOG_ERROR, "Failed to stat entry: %s:%s",
-                                path, strerror (errno));
-                        goto stat_failed;
-                }
+        shandle = volinfo->shandle;
+        ret = glusterd_volume_exclude_options_write (fd, volinfo);
+        if (ret)
+                goto out;
 
-                if (S_ISDIR (st.st_mode))
-                        ret = rmdir (path);
-                else
-                        ret = unlink (path);
+        shandle->fd = fd;
+        dict_foreach (volinfo->dict, _storeopts, shandle);
 
-                gf_log ("", GF_LOG_NORMAL, "%s %s",
-                                ret?"Failed to remove":"Removed",
-                                entry->d_name);
-                if (ret)
-                        gf_log ("", GF_LOG_NORMAL, "errno:%d", errno);
-stat_failed:
-                memset (path, 0, sizeof(path));
-                glusterd_for_each_entry (entry, dir);
-        }
+        dict_foreach (volinfo->gsync_slaves, _storeslaves, shandle);
+        shandle->fd = 0;
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
 
-        ret = closedir (dir);
-        if (ret) {
-                gf_log ("", GF_LOG_NORMAL, "Failed to close dir, errno:%d",
-                        errno);
-        }
+int32_t
+glusterd_store_snapinfo_write (glusterd_snap_t *snap)
+{
+        int32_t  ret            = -1;
+        int      fd             = 0;
+        char     buf[PATH_MAX]  = "";
 
-        ret = rmdir (pathname);
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Failed to rmdir: %s, errno: %d",
-                        pathname, errno);
+        GF_ASSERT (snap);
+
+        fd = gf_store_mkstemp (snap->shandle);
+        if (fd <= 0)
+               goto out;
+
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_ID,
+                                   uuid_utoa (snap->snap_id));
+        if (ret)
+                goto out;
+
+        snprintf (buf, sizeof (buf), "%d", snap->snap_status);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_STATUS, buf);
+        if (ret)
+                goto out;
+
+        snprintf (buf, sizeof (buf), "%d", snap->snap_restored);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_RESTORED, buf);
+        if (ret)
+                goto out;
+
+        if (snap->description) {
+                ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_DESC,
+                                           snap->description);
+                if (ret)
+                        goto out;
         }
 
+        snprintf (buf, sizeof (buf), "%ld", snap->time_stamp);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_SNAP_TIMESTAMP, buf);
 
 out:
-        if (volinfo->shandle)
-                glusterd_store_handle_destroy (volinfo->shandle);
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+static void
+glusterd_store_rbstatepath_set (glusterd_volinfo_t *volinfo, char *rbstatepath,
+                                size_t len)
+{
+        char    voldirpath[PATH_MAX] = {0,};
+        GF_ASSERT (volinfo);
+        GF_ASSERT (rbstatepath);
+        GF_ASSERT (len <= PATH_MAX);
+
+        glusterd_store_voldirpath_set (volinfo, voldirpath,
+                                       sizeof (voldirpath));
+        snprintf (rbstatepath, len, "%s/%s", voldirpath,
+                  GLUSTERD_VOLUME_RBSTATE_FILE);
+}
+
+static void
+glusterd_store_volfpath_set (glusterd_volinfo_t *volinfo, char *volfpath,
+                             size_t len)
+{
+        char    voldirpath[PATH_MAX] = {0,};
+        GF_ASSERT (volinfo);
+        GF_ASSERT (volfpath);
+        GF_ASSERT (len <= PATH_MAX);
+
+        glusterd_store_voldirpath_set (volinfo, voldirpath,
+                                       sizeof (voldirpath));
+        snprintf (volfpath, len, "%s/%s", voldirpath, GLUSTERD_VOLUME_INFO_FILE);
+}
+
+static void
+glusterd_store_node_state_path_set (glusterd_volinfo_t *volinfo,
+                                    char *node_statepath, size_t len)
+{
+        char    voldirpath[PATH_MAX] = {0,};
+        GF_ASSERT (volinfo);
+        GF_ASSERT (node_statepath);
+        GF_ASSERT (len <= PATH_MAX);
+
+        glusterd_store_voldirpath_set (volinfo, voldirpath,
+                                       sizeof (voldirpath));
+        snprintf (node_statepath, len, "%s/%s", voldirpath,
+                  GLUSTERD_NODE_STATE_FILE);
+}
+
+static void
+glusterd_store_missed_snaps_list_path_set (char *missed_snaps_list,
+                                           size_t len)
+{
+        glusterd_conf_t *priv = NULL;
+
+        priv = THIS->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (missed_snaps_list);
+        GF_ASSERT (len <= PATH_MAX);
+
+        snprintf (missed_snaps_list, len, "%s/snaps/"
+                  GLUSTERD_MISSED_SNAPS_LIST_FILE, priv->workdir);
+}
+
+static void
+glusterd_store_snapfpath_set (glusterd_snap_t *snap, char *snap_fpath,
+                              size_t len)
+{
+        glusterd_conf_t *priv = NULL;
+
+        priv = THIS->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (snap);
+        GF_ASSERT (snap_fpath);
+        GF_ASSERT (len <= PATH_MAX);
 
+        snprintf (snap_fpath, len, "%s/snaps/%s/%s", priv->workdir,
+                  snap->snapname, GLUSTERD_SNAP_INFO_FILE);
+}
+
+int32_t
+glusterd_store_create_rbstate_shandle_on_absence (glusterd_volinfo_t *volinfo)
+{
+        char            rbstatepath[PATH_MAX] = {0};
+        int32_t         ret                   = 0;
+
+        GF_ASSERT (volinfo);
+
+        glusterd_store_rbstatepath_set (volinfo, rbstatepath, sizeof (rbstatepath));
+        ret = gf_store_handle_create_on_absence (&volinfo->rb_shandle,
+                                                 rbstatepath);
         return ret;
 }
 
+int32_t
+glusterd_store_create_vol_shandle_on_absence (glusterd_volinfo_t *volinfo)
+{
+        char            volfpath[PATH_MAX] = {0};
+        int32_t         ret = 0;
+
+        GF_ASSERT (volinfo);
 
+        glusterd_store_volfpath_set (volinfo, volfpath, sizeof (volfpath));
+        ret = gf_store_handle_create_on_absence (&volinfo->shandle, volfpath);
+        return ret;
+}
 
 int32_t
-glusterd_store_retrieve_value (glusterd_store_handle_t *handle,
-                               char *key, char **value)
+glusterd_store_create_nodestate_sh_on_absence (glusterd_volinfo_t *volinfo)
 {
-        int32_t         ret = -1;
-        char            scan_str[4096] = {0,};
-        char            *iter_key = NULL;
-        char            *iter_val = NULL;
-        char            *str = NULL;
-        char            *free_str = NULL;
+        char            node_state_path[PATH_MAX] = {0};
+        int32_t         ret                   = 0;
 
-        GF_ASSERT (handle);
+        GF_ASSERT (volinfo);
 
-        handle->fd = open (handle->path, O_RDWR);
+        glusterd_store_node_state_path_set (volinfo, node_state_path,
+                                            sizeof (node_state_path));
+        ret =
+          gf_store_handle_create_on_absence (&volinfo->node_state_shandle,
+                                             node_state_path);
 
-        if (!handle->read)
-                handle->read = fdopen (handle->fd, "r");
+        return ret;
+}
 
-        if (!handle->read) {
-                gf_log ("", GF_LOG_ERROR, "Unable to open file %s errno: %d",
-                        handle->path, errno);
-                goto out;
+static int32_t
+glusterd_store_create_missed_snaps_list_shandle_on_absence ()
+{
+        char               missed_snaps_list[PATH_MAX] = "";
+        int32_t            ret                         = -1;
+        glusterd_conf_t   *priv                        = NULL;
+        xlator_t          *this                        = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        glusterd_store_missed_snaps_list_path_set (missed_snaps_list,
+                                                   sizeof(missed_snaps_list));
+
+        ret = gf_store_handle_create_on_absence
+                                            (&priv->missed_snaps_list_shandle,
+                                             missed_snaps_list);
+        return ret;
+}
+
+int32_t
+glusterd_store_create_snap_shandle_on_absence (glusterd_snap_t *snap)
+{
+        char            snapfpath[PATH_MAX] = {0};
+        int32_t         ret = 0;
+
+        GF_ASSERT (snap);
+
+        glusterd_store_snapfpath_set (snap, snapfpath, sizeof (snapfpath));
+        ret = gf_store_handle_create_on_absence (&snap->shandle, snapfpath);
+        return ret;
+}
+
+int32_t
+glusterd_store_brickinfos (glusterd_volinfo_t *volinfo, int vol_fd)
+{
+        int32_t                 ret = 0;
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        int32_t                 brick_count = 0;
+
+        GF_ASSERT (volinfo);
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                ret = glusterd_store_brickinfo (volinfo, brickinfo,
+                                            brick_count, vol_fd);
+                if (ret)
+                        goto out;
+                brick_count++;
         }
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_store_rbstate_write (int fd, glusterd_volinfo_t *volinfo)
+{
+        int     ret             = -1;
+        int     port            = 0;
+        char    buf[PATH_MAX]   = {0, };
 
-        ret = fscanf (handle->read, "%s", scan_str);
+        GF_ASSERT (fd > 0);
+        GF_ASSERT (volinfo);
 
-        while (ret != EOF) {
-                if (free_str) {
-                        GF_FREE (free_str);
-                        free_str = NULL;
-                }
-                str = gf_strdup (scan_str);
-                if (!str)
+        snprintf (buf, sizeof (buf), "%d", volinfo->rep_brick.rb_status);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_RB_STATUS, buf);
+        if (ret)
+                goto out;
+
+        if (volinfo->rep_brick.rb_status > GF_RB_STATUS_NONE) {
+
+                snprintf (buf, sizeof (buf), "%s:%s",
+                          volinfo->rep_brick.src_brick->hostname,
+                          volinfo->rep_brick.src_brick->path);
+                ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_RB_SRC_BRICK,
+                                           buf);
+                if (ret)
                         goto out;
-                else
-                        free_str = str;
-                iter_key = strtok (str, "=");
-                gf_log ("", GF_LOG_DEBUG, "key %s read", iter_key);
 
-                if (!strcmp (key, iter_key)) {
-                        gf_log ("", GF_LOG_DEBUG, "key %s found", key);
-                        iter_val = strtok (NULL, "=");
-                        ret = 0;
-                        if (iter_val)
-                                *value = gf_strdup (iter_val);
+                snprintf (buf, sizeof (buf), "%s:%s",
+                          volinfo->rep_brick.dst_brick->hostname,
+                          volinfo->rep_brick.dst_brick->path);
+                ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_RB_DST_BRICK,
+                                           buf);
+                if (ret)
                         goto out;
+
+                switch (volinfo->transport_type) {
+                        case GF_TRANSPORT_RDMA:
+                                port = volinfo->rep_brick.dst_brick->rdma_port;
+                                break;
+
+                        case GF_TRANSPORT_TCP:
+                        case GF_TRANSPORT_BOTH_TCP_RDMA:
+                                port = volinfo->rep_brick.dst_brick->port;
+                                break;
                 }
 
-                ret = fscanf (handle->read, "%s", scan_str);
+                snprintf (buf, sizeof (buf), "%d", port);
+                ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_RB_DST_PORT,
+                                           buf);
+                if (ret)
+                        goto out;
+                uuid_unparse (volinfo->rep_brick.rb_id, buf);
+                ret = gf_store_save_value (fd, GF_REPLACE_BRICK_TID_KEY, buf);
         }
 
-        if (EOF == ret)
-                ret = -1;
+        ret = 0;
 out:
-        if (handle->fd > 0) {
-                close (handle->fd);
-                handle->read = NULL;
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_store_perform_rbstate_store (glusterd_volinfo_t *volinfo)
+{
+        int                         fd = -1;
+        int32_t                     ret = -1;
+        GF_ASSERT (volinfo);
+
+        fd = gf_store_mkstemp (volinfo->rb_shandle);
+        if (fd <= 0) {
+                ret = -1;
+                goto out;
         }
 
-        if (free_str)
-                GF_FREE (free_str);
+        ret = glusterd_store_rbstate_write (fd, volinfo);
+        if (ret)
+                goto out;
+
+        ret = gf_store_rename_tmppath (volinfo->rb_shandle);
+        if (ret)
+                goto out;
 
+out:
+        if (ret && (fd > 0))
+                gf_store_unlink_tmppath (volinfo->rb_shandle);
+        if (fd > 0)
+                close (fd);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
 int32_t
-glusterd_store_save_value (glusterd_store_handle_t *handle,
-                           char *key, char *value)
+glusterd_store_node_state_write (int fd, glusterd_volinfo_t *volinfo)
 {
-        int32_t         ret = -1;
-        char            buf[4096] = {0,};
+        int     ret             = -1;
+        char    buf[PATH_MAX]   = {0, };
 
-        GF_ASSERT (handle);
-        GF_ASSERT (key);
-        GF_ASSERT (value);
+        GF_ASSERT (fd > 0);
+        GF_ASSERT (volinfo);
 
-        handle->fd = open (handle->path, O_RDWR | O_APPEND);
+        if (volinfo->rebal.defrag_cmd == GF_DEFRAG_CMD_STATUS) {
+                ret = 0;
+                goto out;
+        }
 
-        if (handle->fd < 0) {
-                gf_log ("", GF_LOG_ERROR, "Unable to open %s, errno: %d",
-                        handle->path, errno);
+        snprintf (buf, sizeof (buf), "%d", volinfo->rebal.defrag_cmd);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_VOL_DEFRAG, buf);
+        if (ret)
+                goto out;
+
+        snprintf (buf, sizeof (buf), "%d", volinfo->rebal.op);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_DEFRAG_OP, buf);
+        if (ret)
                 goto out;
+
+        if (volinfo->rebal.defrag_cmd) {
+                uuid_unparse (volinfo->rebal.rebalance_id, buf);
+                ret = gf_store_save_value (fd, GF_REBALANCE_TID_KEY, buf);
         }
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
 
-        snprintf (buf, sizeof (buf), "%s=%s\n", key, value);
-        ret = write (handle->fd, buf, strlen (buf));
+int32_t
+glusterd_store_perform_node_state_store (glusterd_volinfo_t *volinfo)
+{
+        int                         fd = -1;
+        int32_t                     ret = -1;
+        GF_ASSERT (volinfo);
 
-        if (ret < 0) {
-                gf_log ("", GF_LOG_CRITICAL, "Unable to store key: %s,"
-                        "value: %s, error: %s", key, value,
-                        strerror (errno));
+        fd = gf_store_mkstemp (volinfo->node_state_shandle);
+        if (fd <= 0) {
                 ret = -1;
                 goto out;
         }
 
-        ret = 0;
+        ret = glusterd_store_node_state_write (fd, volinfo);
+        if (ret)
+                goto out;
+
+        ret = gf_store_rename_tmppath (volinfo->node_state_shandle);
+        if (ret)
+                goto out;
 
 out:
+        if (ret && (fd > 0))
+                gf_store_unlink_tmppath (volinfo->node_state_shandle);
+        if (fd > 0)
+                close (fd);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
 
-        if (handle->fd > 0) {
-                close (handle->fd);
-                handle->fd = -1;
+int32_t
+glusterd_store_perform_volume_store (glusterd_volinfo_t *volinfo)
+{
+        int                         fd = -1;
+        int32_t                     ret = -1;
+        GF_ASSERT (volinfo);
+
+        fd = gf_store_mkstemp (volinfo->shandle);
+        if (fd <= 0) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_store_volinfo_write (fd, volinfo);
+        if (ret)
+                goto out;
+
+        ret = glusterd_store_brickinfos (volinfo, fd);
+        if (ret)
+                goto out;
+
+out:
+        if (ret && (fd > 0))
+                gf_store_unlink_tmppath (volinfo->shandle);
+        if (fd > 0)
+                close (fd);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+void
+glusterd_perform_volinfo_version_action (glusterd_volinfo_t *volinfo,
+                                         glusterd_volinfo_ver_ac_t ac)
+{
+        GF_ASSERT (volinfo);
+
+        switch (ac) {
+        case GLUSTERD_VOLINFO_VER_AC_NONE:
+        break;
+        case GLUSTERD_VOLINFO_VER_AC_INCREMENT:
+                volinfo->version++;
+        break;
+        case GLUSTERD_VOLINFO_VER_AC_DECREMENT:
+                volinfo->version--;
+        break;
+        }
+}
+
+void
+glusterd_store_bricks_cleanup_tmp (glusterd_volinfo_t *volinfo)
+{
+        glusterd_brickinfo_t    *brickinfo           = NULL;
+
+        GF_ASSERT (volinfo);
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                gf_store_unlink_tmppath (brickinfo->shandle);
         }
+}
+
+void
+glusterd_store_volume_cleanup_tmp (glusterd_volinfo_t *volinfo)
+{
+        GF_ASSERT (volinfo);
+
+        glusterd_store_bricks_cleanup_tmp (volinfo);
+
+        gf_store_unlink_tmppath (volinfo->shandle);
+
+        gf_store_unlink_tmppath (volinfo->rb_shandle);
+
+        gf_store_unlink_tmppath (volinfo->node_state_shandle);
+}
 
-        gf_log ("", GF_LOG_DEBUG, "returning: %d", ret);
+void
+glusterd_store_snap_cleanup_tmp (glusterd_snap_t *snap)
+{
+        GF_ASSERT (snap);
+
+        gf_store_unlink_tmppath (snap->shandle);
+}
+
+int32_t
+glusterd_store_brickinfos_atomic_update (glusterd_volinfo_t *volinfo)
+{
+        int                      ret            = -1;
+        glusterd_brickinfo_t    *brickinfo      = NULL;
+
+        GF_ASSERT (volinfo);
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                ret = gf_store_rename_tmppath (brickinfo->shandle);
+                if (ret)
+                        goto out;
+        }
+out:
         return ret;
 }
 
 int32_t
-glusterd_store_handle_new (char *path, glusterd_store_handle_t **handle)
+glusterd_store_volinfo_atomic_update (glusterd_volinfo_t *volinfo)
 {
-        int32_t                 ret = -1;
-        glusterd_store_handle_t *shandle = NULL;
+        int ret = -1;
+        GF_ASSERT (volinfo);
 
-        shandle = GF_CALLOC (1, sizeof (*shandle), gf_gld_mt_store_handle_t);
-        if (!shandle)
+        ret = gf_store_rename_tmppath (volinfo->shandle);
+        if (ret)
                 goto out;
 
-        shandle->path = gf_strdup (path);
+out:
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR, "Couldn't rename "
+                        "temporary file(s): Reason %s", strerror (errno));
+        return ret;
+}
 
-        if (!shandle->path)
-                goto out;
+int32_t
+glusterd_store_volume_atomic_update (glusterd_volinfo_t *volinfo)
+{
+        int ret = -1;
+        GF_ASSERT (volinfo);
 
-        shandle->fd = open (path, O_RDWR | O_CREAT | O_APPEND, 0644);
-        if (!shandle->fd)
+        ret = glusterd_store_brickinfos_atomic_update (volinfo);
+        if (ret)
                 goto out;
 
-        *handle = shandle;
-
-        ret = 0;
+        ret = glusterd_store_volinfo_atomic_update (volinfo);
 
 out:
-        if (ret == -1) {
-                if (shandle) {
-                        if (shandle->path)
-                                GF_FREE (shandle->path);
-                        if (shandle->fd > 0)
-                                close (shandle->fd);
-                        GF_FREE (shandle);
-                }
-        } else {
-                close (shandle->fd);
-        }
+        return ret;
+}
 
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+int32_t
+glusterd_store_snap_atomic_update (glusterd_snap_t *snap)
+{
+        int ret = -1;
+        GF_ASSERT (snap);
+
+        ret = gf_store_rename_tmppath (snap->shandle);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR, "Couldn't rename "
+                        "temporary file(s): Reason %s", strerror (errno));
 
         return ret;
 }
 
 int32_t
-glusterd_store_handle_destroy (glusterd_store_handle_t *handle)
+glusterd_store_snap (glusterd_snap_t *snap)
 {
         int32_t                 ret = -1;
 
-        if (!handle) {
-                ret = 0;
+        GF_ASSERT (snap);
+
+        ret = glusterd_store_create_snap_dir (snap);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to create snap dir");
                 goto out;
         }
 
-        GF_FREE (handle->path);
+        ret = glusterd_store_create_snap_shandle_on_absence (snap);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to create snap info "
+                        "file");
+                goto out;
+        }
 
-        GF_FREE (handle);
+        ret = glusterd_store_snapinfo_write (snap);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to write snap info");
+                goto out;
+        }
 
-        ret = 0;
+        ret = glusterd_store_snap_atomic_update (snap);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR,"Failed to do automic update");
+                goto out;
+        }
 
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        if (ret)
+                glusterd_store_snap_cleanup_tmp (snap);
 
+        gf_log (THIS->name, GF_LOG_TRACE, "Returning %d", ret);
         return ret;
 }
 
-
 int32_t
-glusterd_store_handle_truncate (glusterd_store_handle_t *handle)
+glusterd_store_volinfo (glusterd_volinfo_t *volinfo, glusterd_volinfo_ver_ac_t ac)
 {
-        int32_t         ret = -1;
+        int32_t                 ret = -1;
 
-        GF_ASSERT (handle);
-        GF_ASSERT (handle->path);
+        GF_ASSERT (volinfo);
 
-        ret = truncate (handle->path, 0);
+        glusterd_perform_volinfo_version_action (volinfo, ac);
+        ret = glusterd_store_create_volume_dir (volinfo);
+        if (ret)
+                goto out;
 
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        ret = glusterd_store_create_vol_shandle_on_absence (volinfo);
+        if (ret)
+                goto out;
+
+        ret = glusterd_store_create_rbstate_shandle_on_absence (volinfo);
+        if (ret)
+                goto out;
+
+        ret = glusterd_store_create_nodestate_sh_on_absence (volinfo);
+        if (ret)
+                goto out;
+
+        ret = glusterd_store_perform_volume_store (volinfo);
+        if (ret)
+                goto out;
+
+        ret = glusterd_store_volume_atomic_update (volinfo);
+        if (ret) {
+                glusterd_perform_volinfo_version_action (volinfo,
+                                                         GLUSTERD_VOLINFO_VER_AC_DECREMENT);
+                goto out;
+        }
+
+        ret = glusterd_store_perform_rbstate_store (volinfo);
+        if (ret)
+                goto out;
+
+        ret = glusterd_store_perform_node_state_store (volinfo);
+        if (ret)
+                goto out;
+
+        //checksum should be computed at the end
+        ret = glusterd_volume_compute_cksum (volinfo);
+        if (ret)
+                goto out;
+
+out:
+        if (ret)
+                glusterd_store_volume_cleanup_tmp (volinfo);
+
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
 
         return ret;
 }
 
 int32_t
-glusterd_store_uuid ()
+glusterd_store_delete_volume (glusterd_volinfo_t *volinfo)
 {
-        char            str[GLUSTERD_UUID_LEN] = {0,};
-        glusterd_conf_t *priv = NULL;
-        char            path[PATH_MAX] = {0,};
-        int32_t         ret = -1;
-        glusterd_store_handle_t *handle = NULL;
+        char             pathname[PATH_MAX]    = {0,};
+        int32_t          ret                   = 0;
+        glusterd_conf_t *priv                  = NULL;
+        DIR             *dir                   = NULL;
+        struct dirent   *entry                 = NULL;
+        char             path[PATH_MAX]        = {0,};
+        char             delete_path[PATH_MAX] = {0,};
+        char             trashdir[PATH_MAX]    = {0,};
+        struct stat      st                    = {0, };
+        xlator_t        *this                  = NULL;
+        gf_boolean_t     rename_fail           = _gf_false;
 
-        priv = THIS->private;
+        this = THIS;
+        GF_ASSERT (this);
 
-        uuid_unparse (priv->uuid, str);
+        GF_ASSERT (volinfo);
+        priv = this->private;
 
-        snprintf (path, PATH_MAX, "%s/%s", priv->workdir,
-                  GLUSTERD_INFO_FILE);
+        GF_ASSERT (priv);
 
-        if (!priv->handle) {
-                ret = glusterd_store_handle_new (path, &handle);
+        GLUSTERD_GET_VOLUME_DIR (pathname, volinfo, priv);
+
+        snprintf (delete_path, sizeof (delete_path),
+        "%s/"GLUSTERD_TRASH"/%s.deleted", priv->workdir,
+        uuid_utoa (volinfo->volume_id));
+
+        snprintf (trashdir, sizeof (trashdir), "%s/"GLUSTERD_TRASH,
+                  priv->workdir);
+
+        ret = mkdir (trashdir, 0777);
+        if (ret && errno != EEXIST) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to create trash "
+                        "directory, reason : %s", strerror (errno));
+                ret = -1;
+                goto out;
+        }
+
+        ret = rename (pathname, delete_path);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to rename volume "
+                        "directory for volume %s", volinfo->volname);
+                rename_fail = _gf_true;
+                goto out;
+        }
+
+        dir = opendir (delete_path);
+        if (!dir) {
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to open directory %s."
+                        " Reason : %s", delete_path, strerror (errno));
+                ret = 0;
+                goto out;
+        }
+        ret = glusterd_store_remove_bricks (volinfo, delete_path);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "Remove bricks failed for %s",
+                        volinfo->volname);
+        }
+
+        glusterd_for_each_entry (entry, dir);
+        while (entry) {
+
+                snprintf (path, PATH_MAX, "%s/%s", delete_path, entry->d_name);
+                ret = stat (path, &st);
+                if (ret == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG, "Failed to stat "
+                                "entry %s : %s", path, strerror (errno));
+                        goto stat_failed;
+                }
+
+                if (S_ISDIR (st.st_mode))
+                        ret = rmdir (path);
+                else
+                        ret = unlink (path);
 
                 if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to get store"
-                                " handle!");
-                        goto out;
+                        gf_log (this->name, GF_LOG_DEBUG, " Failed to remove "
+                                "%s. Reason : %s", path, strerror (errno));
                 }
 
-                priv->handle = handle;
+                gf_log (this->name, GF_LOG_DEBUG, "%s %s",
+                                ret ? "Failed to remove":"Removed",
+                                entry->d_name);
+stat_failed:
+                memset (path, 0, sizeof(path));
+                glusterd_for_each_entry (entry, dir);
         }
 
-        ret = glusterd_store_save_value (priv->handle, GLUSTERD_STORE_UUID_KEY,
-                                         str);
-
+        ret = closedir (dir);
         if (ret) {
-                gf_log ("", GF_LOG_CRITICAL, "Storing uuid failed"
-                        "ret = %d", ret);
-                goto out;
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to close dir %s. "
+                        "Reason : %s",delete_path, strerror (errno));
         }
 
+        ret = rmdir (delete_path);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to rmdir: %s,err: %s",
+                        delete_path, strerror (errno));
+        }
+        ret = rmdir (trashdir);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to rmdir: %s, Reason:"
+                        " %s", trashdir, strerror (errno));
+        }
 
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        if (volinfo->shandle) {
+                gf_store_handle_destroy (volinfo->shandle);
+                volinfo->shandle = NULL;
+        }
+        ret = (rename_fail == _gf_true) ? -1: 0;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
-
+/*TODO: cleanup the duplicate code and implement a generic function for
+ * deleting snap/volume depending on the parameter flag */
 int32_t
-glusterd_retrieve_uuid ()
+glusterd_store_delete_snap (glusterd_snap_t *snap)
 {
-        char            *uuid_str = NULL;
-        int32_t         ret = -1;
-        glusterd_store_handle_t *handle = NULL;
-        glusterd_conf_t *priv = NULL;
-        char            path[PATH_MAX] = {0,};
+        char             pathname[PATH_MAX]    = {0,};
+        int32_t          ret                   = 0;
+        glusterd_conf_t *priv                  = NULL;
+        DIR             *dir                   = NULL;
+        struct dirent   *entry                 = NULL;
+        char             path[PATH_MAX]        = {0,};
+        char             delete_path[PATH_MAX] = {0,};
+        char             trashdir[PATH_MAX]    = {0,};
+        struct stat      st                    = {0, };
+        xlator_t        *this                  = NULL;
+        gf_boolean_t     rename_fail           = _gf_false;
 
-        priv = THIS->private;
+        this = THIS;
+        priv = this->private;
+        GF_ASSERT (priv);
 
+        GF_ASSERT (snap);
+        GLUSTERD_GET_SNAP_DIR (pathname, snap, priv);
 
-        if (!priv->handle) {
-                snprintf (path, PATH_MAX, "%s/%s", priv->workdir,
-                          GLUSTERD_INFO_FILE);
-                ret = glusterd_store_handle_new (path, &handle);
+        snprintf (delete_path, sizeof (delete_path),
+        "%s/"GLUSTERD_TRASH"/snap-%s.deleted", priv->workdir,
+        uuid_utoa (snap->snap_id));
+
+        snprintf (trashdir, sizeof (trashdir), "%s/"GLUSTERD_TRASH,
+                  priv->workdir);
+
+        ret = mkdir (trashdir, 0777);
+        if (ret && errno != EEXIST) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to create trash "
+                        "directory, reason : %s", strerror (errno));
+                ret = -1;
+                goto out;
+        }
+
+        ret = rename (pathname, delete_path);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to rename snap "
+                        "directory %s to %s", snap->snapname, delete_path);
+                rename_fail = _gf_true;
+                goto out;
+        }
+
+        dir = opendir (delete_path);
+        if (!dir) {
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to open directory %s."
+                        " Reason : %s", delete_path, strerror (errno));
+                ret = 0;
+                goto out;
+        }
+
+        glusterd_for_each_entry (entry, dir);
+        while (entry) {
+                snprintf (path, PATH_MAX, "%s/%s", delete_path, entry->d_name);
+                ret = stat (path, &st);
+                if (ret == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG, "Failed to stat "
+                                "entry %s : %s", path, strerror (errno));
+                        goto stat_failed;
+                }
+
+                if (S_ISDIR (st.st_mode))
+                        ret = rmdir (path);
+                else
+                        ret = unlink (path);
 
                 if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to get store "
-                                "handle!");
+                        gf_log (this->name, GF_LOG_DEBUG, " Failed to remove "
+                                "%s. Reason : %s", path, strerror (errno));
+                }
+
+                gf_log (this->name, GF_LOG_DEBUG, "%s %s",
+                                ret ? "Failed to remove":"Removed",
+                                entry->d_name);
+stat_failed:
+                memset (path, 0, sizeof(path));
+                glusterd_for_each_entry (entry, dir);
+        }
+
+        ret = closedir (dir);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to close dir %s. "
+                        "Reason : %s",delete_path, strerror (errno));
+        }
+
+        ret = rmdir (delete_path);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to rmdir: %s,err: %s",
+                        delete_path, strerror (errno));
+        }
+        ret = rmdir (trashdir);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to rmdir: %s, Reason:"
+                        " %s", trashdir, strerror (errno));
+        }
+
+out:
+        if (snap->shandle) {
+                gf_store_handle_destroy (snap->shandle);
+                snap->shandle = NULL;
+        }
+        ret = (rename_fail == _gf_true) ? -1: 0;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_store_global_info (xlator_t *this)
+{
+        int                     ret                     = -1;
+        glusterd_conf_t         *conf                   = NULL;
+        char                    op_version_str[15]      = {0,};
+        char                    path[PATH_MAX]          = {0,};
+        gf_store_handle_t       *handle                 = NULL;
+        char                    *uuid_str               = NULL;
+        char                     buf[256]               = {0, };
+
+        conf = this->private;
+
+        uuid_str = gf_strdup (uuid_utoa (MY_UUID));
+        if (!uuid_str)
+                goto out;
+
+        if (!conf->handle) {
+                snprintf (path, PATH_MAX, "%s/%s", conf->workdir,
+                          GLUSTERD_INFO_FILE);
+                ret = gf_store_handle_new (path, &handle);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to get store handle");
                         goto out;
                 }
 
-                priv->handle = handle;
+                conf->handle = handle;
+        } else
+                handle = conf->handle;
+
+        /* These options need to be available for all users */
+        ret = chmod (handle->path, 0644);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "chmod error for %s: %s",
+                        GLUSTERD_INFO_FILE, strerror (errno));
+                goto out;
         }
 
-        ret = glusterd_store_retrieve_value (priv->handle,
-                                             GLUSTERD_STORE_UUID_KEY,
-                                             &uuid_str);
+        handle->fd = gf_store_mkstemp (handle);
+        if (handle->fd <= 0) {
+                ret = -1;
+                goto out;
+        }
 
+        ret = gf_store_save_value (handle->fd, GLUSTERD_STORE_UUID_KEY,
+                                   uuid_str);
         if (ret) {
-                gf_log ("", GF_LOG_CRITICAL, "Retrieving uuid failed"
-                        " ret = %d", ret);
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "Storing uuid failed ret = %d", ret);
                 goto out;
         }
 
-        uuid_parse (uuid_str, priv->uuid);
+        snprintf (op_version_str, 15, "%d", conf->op_version);
+        ret = gf_store_save_value (handle->fd, GD_OP_VERSION_KEY,
+                                   op_version_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Storing op-version failed ret = %d", ret);
+                goto out;
+        }
+
+        snprintf (buf, sizeof (buf), "%"PRIu64, conf->snap_max_hard_limit);
+        ret = gf_store_save_value (handle->fd,
+                                   GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT, buf);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Storing snap-max-hard-limit failed ret = %d", ret);
+                goto out;
+        }
+
+        snprintf (buf, sizeof (buf), "%"PRIu64, conf->snap_max_soft_limit);
+        ret = gf_store_save_value (handle->fd,
+                                   GLUSTERD_STORE_KEY_SNAP_MAX_SOFT_LIMIT, buf);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Storing snap-max-soft-limit failed ret = %d", ret);
+                goto out;
+        }
 
+        ret = gf_store_rename_tmppath (handle);
 out:
+        if (ret && (handle->fd > 0))
+                gf_store_unlink_tmppath (handle);
+
+        if (handle->fd > 0) {
+                close (handle->fd);
+                handle->fd = 0;
+        }
+
         if (uuid_str)
                 GF_FREE (uuid_str);
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to store glusterd global-info");
+
         return ret;
 }
 
-int32_t
-glusterd_store_iter_new (glusterd_store_handle_t  *shandle,
-                         glusterd_store_iter_t  **iter)
+int
+glusterd_retrieve_op_version (xlator_t *this, int *op_version)
 {
-        int32_t                 ret = -1;
-        glusterd_store_iter_t   *tmp_iter = NULL;
-        int                     fd = -1;
+        char                    *op_version_str = NULL;
+        glusterd_conf_t         *priv           = NULL;
+        int                     ret             = -1;
+        int                     tmp_version     = 0;
+        char                    *tmp            = NULL;
+        char                    path[PATH_MAX]  = {0,};
+        gf_store_handle_t       *handle         = NULL;
 
-        GF_ASSERT (shandle);
-        GF_ASSERT (iter);
+        priv = this->private;
 
-        tmp_iter = GF_CALLOC (1, sizeof (*tmp_iter),
-                             gf_gld_mt_store_iter_t);
+        if (!priv->handle) {
+                snprintf (path, PATH_MAX, "%s/%s", priv->workdir,
+                          GLUSTERD_INFO_FILE);
+                ret = gf_store_handle_retrieve (path, &handle);
 
-        if (!tmp_iter) {
-                gf_log ("", GF_LOG_ERROR, "Out of Memory");
-                goto out;
-        }
+                if (ret) {
+                        gf_log ("", GF_LOG_DEBUG, "Unable to get store "
+                                "handle!");
+                        goto out;
+                }
 
-        fd = open (shandle->path, O_RDWR);
+                priv->handle = handle;
+        }
 
-        if (fd < 0) {
-                gf_log ("", GF_LOG_ERROR, "Unable to open %s, errno: %d",
-                        shandle->path, errno);
+        ret = gf_store_retrieve_value (priv->handle, GD_OP_VERSION_KEY,
+                                       &op_version_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "No previous op_version present");
                 goto out;
         }
 
-        tmp_iter->fd = fd;
-
-        tmp_iter->file = fdopen (shandle->fd, "r");
-
-        if (!tmp_iter->file) {
-                gf_log ("", GF_LOG_ERROR, "Unable to open file %s errno: %d",
-                        shandle->path, errno);
+        tmp_version = strtol (op_version_str, &tmp, 10);
+        if ((tmp_version <= 0) || (tmp && strlen (tmp) > 1)) {
+                gf_log (this->name, GF_LOG_WARNING, "invalid version number");
                 goto out;
         }
 
-        *iter = tmp_iter;
-        ret = 0;
+        *op_version = tmp_version;
 
+        ret = 0;
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        if (op_version_str)
+                GF_FREE (op_version_str);
+
         return ret;
 }
 
-int32_t
-glusterd_store_iter_get_next (glusterd_store_iter_t *iter,
-                              char  **key, char **value)
+int
+glusterd_retrieve_sys_snap_max_limit (xlator_t *this, uint64_t *limit,
+                                      char *key)
 {
-        int32_t         ret = -1;
-        char            scan_str[4096] = {0,};
-        char            *str = NULL;
-        char            *free_str = NULL;
-        char            *iter_key = NULL;
-        char            *iter_val = NULL;
+        char                    *limit_str      = NULL;
+        glusterd_conf_t         *priv           = NULL;
+        int                     ret             = -1;
+        uint64_t                tmp_limit       = 0;
+        char                    *tmp            = NULL;
+        char                    path[PATH_MAX]  = {0,};
+        gf_store_handle_t       *handle         = NULL;
 
-        GF_ASSERT (iter);
-        GF_ASSERT (iter->file);
+        GF_ASSERT (this);
+        priv = this->private;
 
-        ret = fscanf (iter->file, "%s", scan_str);
+        GF_ASSERT (priv);
+        GF_ASSERT (limit);
+        GF_ASSERT (key);
 
-        if (ret <= 0) {
-                ret = -1;
+        if (!priv->handle) {
+                snprintf (path, PATH_MAX, "%s/%s", priv->workdir,
+                          GLUSTERD_INFO_FILE);
+                ret = gf_store_handle_retrieve (path, &handle);
+
+                if (ret) {
+                        gf_log ("", GF_LOG_DEBUG, "Unable to get store "
+                                "handle!");
+                        goto out;
+                }
+
+                priv->handle = handle;
+        }
+
+        ret = gf_store_retrieve_value (priv->handle,
+                                       key,
+                                       &limit_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "No previous %s present", key);
                 goto out;
         }
 
-        str = gf_strdup (scan_str);
-        if (!str)
+        tmp_limit = strtoul (limit_str, &tmp, 10);
+        if ((tmp_limit <= 0) || (tmp && strlen (tmp) > 1)) {
+                gf_log (this->name, GF_LOG_WARNING, "invalid version number");
                 goto out;
-        else
-                free_str = str;
+        }
 
-        iter_key = strtok (str, "=");
-        gf_log ("", GF_LOG_DEBUG, "key %s read", iter_key);
+        *limit = tmp_limit;
 
+        ret = 0;
+out:
+        if (limit_str)
+                GF_FREE (limit_str);
 
-        iter_val = strtok (NULL, "=");
-        gf_log ("", GF_LOG_DEBUG, "value %s read", iter_val);
+        return ret;
+}
+static int
+glusterd_restore_op_version (xlator_t *this)
+{
+        glusterd_conf_t *conf           = NULL;
+        int              ret            = 0;
+        int              op_version     = 0;
 
-        if (iter_val)
-                *value = gf_strdup (iter_val);
-        *key   = gf_strdup (iter_key);
+        conf = this->private;
 
-        ret = 0;
+        ret = glusterd_retrieve_sys_snap_max_limit (this,
+                                                    &conf->snap_max_hard_limit,
+                                                    GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Unable to retrieve system snap-max-hard-limit, "
+                        "setting it to default value(%d)",
+                        GLUSTERD_SNAPS_MAX_HARD_LIMIT);
+                conf->snap_max_hard_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT;
+        }
 
-out:
-        if (free_str)
-                GF_FREE (free_str);
+        ret = glusterd_retrieve_sys_snap_max_limit (this,
+                                                    &conf->snap_max_soft_limit,
+                                                    GLUSTERD_STORE_KEY_SNAP_MAX_SOFT_LIMIT);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Unable to retrieve system snap-max-soft-limit, "
+                        "setting it to default value(%d)",
+                        GLUSTERD_SNAPS_DEF_SOFT_LIMIT_PERCENT);
+                conf->snap_max_soft_limit = GLUSTERD_SNAPS_DEF_SOFT_LIMIT_PERCENT;
+        }
 
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        ret = glusterd_retrieve_op_version (this, &op_version);
+        if (!ret) {
+                if ((op_version < GD_OP_VERSION_MIN) ||
+                    (op_version > GD_OP_VERSION_MAX)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "wrong op-version (%d) retrieved", op_version);
+                        ret = -1;
+                        goto out;
+                }
+                conf->op_version = op_version;
+                gf_log ("glusterd", GF_LOG_INFO,
+                        "retrieved op-version: %d", conf->op_version);
+                goto out;
+        }
+
+        /* op-version can be missing from the store file in 2 cases,
+         * 1. This is a new install of glusterfs
+         * 2. This is an upgrade of glusterfs from a version without op-version
+         *    to a version with op-version (eg. 3.3 -> 3.4)
+         *
+         * Detection of a new install or an upgrade from an older install can be
+         * done by checking for the presence of the its peer-id in the store
+         * file.  If peer-id is present, the installation is an upgrade else, it
+         * is a new install.
+         *
+         * For case 1, set op-version to GD_OP_VERSION_MAX.
+         * For case 2, set op-version to GD_OP_VERSION_MIN.
+         */
+        ret = glusterd_retrieve_uuid();
+        if (ret) {
+                gf_log (this->name, GF_LOG_INFO, "Detected new install. Setting"
+                        " op-version to maximum : %d", GD_OP_VERSION_MAX);
+                conf->op_version = GD_OP_VERSION_MAX;
+        } else {
+                gf_log (this->name, GF_LOG_INFO, "Upgrade detected. Setting"
+                        " op-version to minimum : %d", GD_OP_VERSION_MIN);
+                conf->op_version = GD_OP_VERSION_MIN;
+        }
+        ret = 0;
+out:
         return ret;
 }
 
 int32_t
-glusterd_store_iter_destroy (glusterd_store_iter_t *iter)
+glusterd_retrieve_uuid ()
 {
+        char            *uuid_str = NULL;
         int32_t         ret = -1;
+        gf_store_handle_t *handle = NULL;
+        glusterd_conf_t *priv = NULL;
+        char            path[PATH_MAX] = {0,};
+
+        priv = THIS->private;
+
+        if (!priv->handle) {
+                snprintf (path, PATH_MAX, "%s/%s", priv->workdir,
+                          GLUSTERD_INFO_FILE);
+                ret = gf_store_handle_retrieve (path, &handle);
 
-        GF_ASSERT (iter);
-        GF_ASSERT (iter->fd > 0);
+                if (ret) {
+                        gf_log ("", GF_LOG_DEBUG, "Unable to get store"
+                                "handle!");
+                        goto out;
+                }
 
-        ret = fclose (iter->file);
+                priv->handle = handle;
+        }
+
+        ret = gf_store_retrieve_value (priv->handle, GLUSTERD_STORE_UUID_KEY,
+                                       &uuid_str);
 
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to close fd: %d, ret: %d, "
-                        "errno: %d" ,iter->fd, ret, errno);
+                gf_log ("", GF_LOG_DEBUG, "No previous uuid is present");
+                goto out;
         }
 
-        GF_FREE (iter);
+        uuid_parse (uuid_str, priv->uuid);
 
+out:
+        GF_FREE (uuid_str);
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
+
 int32_t
 glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo)
 {
-
         int32_t                 ret = 0;
         glusterd_brickinfo_t    *brickinfo = NULL;
-        glusterd_store_iter_t   *iter = NULL;
+        gf_store_iter_t         *iter = NULL;
         char                    *key = NULL;
         char                    *value = NULL;
         char                    brickdir[PATH_MAX] = {0,};
         char                    path[PATH_MAX] = {0,};
         glusterd_conf_t         *priv = NULL;
-        DIR                     *dir = NULL;
-        struct dirent           *entry = NULL;
+        int32_t                 brick_count = 0;
+        char                    tmpkey[4096] = {0,};
+        gf_store_iter_t         *tmpiter = NULL;
+        char                    *tmpvalue = NULL;
+        struct pmap_registry    *pmap = NULL;
+        gf_store_op_errno_t     op_errno = GD_STORE_SUCCESS;
 
         GF_ASSERT (volinfo);
         GF_ASSERT (volinfo->volname);
@@ -847,31 +1907,41 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo)
 
         GLUSTERD_GET_BRICK_DIR (brickdir, volinfo, priv);
 
-        dir = opendir (brickdir);
-
-        glusterd_for_each_entry (entry, dir);
+        ret = gf_store_iter_new (volinfo->shandle, &tmpiter);
+        if (ret)
+                goto out;
 
-        while (entry) {
+        while (brick_count < volinfo->brick_count) {
                 ret = glusterd_brickinfo_new (&brickinfo);
 
                 if (ret)
                         goto out;
+                snprintf (tmpkey, sizeof (tmpkey), "%s-%d",
+                          GLUSTERD_STORE_KEY_VOL_BRICK,brick_count);
+                ret = gf_store_iter_get_matching (tmpiter, tmpkey, &tmpvalue);
+                snprintf (path, sizeof (path), "%s/%s", brickdir, tmpvalue);
+
+                GF_FREE (tmpvalue);
 
-                snprintf (path, sizeof (path), "%s/%s", brickdir,
-                          entry->d_name);
+                tmpvalue = NULL;
 
-                ret = glusterd_store_handle_new (path, &brickinfo->shandle);
+                ret = gf_store_handle_retrieve (path, &brickinfo->shandle);
 
                 if (ret)
                         goto out;
 
-                ret = glusterd_store_iter_new (brickinfo->shandle, &iter);
+                ret = gf_store_iter_new (brickinfo->shandle, &iter);
 
                 if (ret)
                         goto out;
 
-                ret = glusterd_store_iter_get_next (iter, &key, &value);
-
+                ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR, "Unable to iterate "
+                                "the store for brick: %s, reason: %s", path,
+                                gf_store_strerror (op_errno));
+                        goto out;
+                }
                 while (!ret) {
                         if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_HOSTNAME,
                                       strlen (GLUSTERD_STORE_KEY_BRICK_HOSTNAME))) {
@@ -883,6 +1953,52 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo)
                         } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_PORT,
                                     strlen (GLUSTERD_STORE_KEY_BRICK_PORT))) {
                                 gf_string2int (value, &brickinfo->port);
+
+                                if (brickinfo->port < priv->base_port) {
+                                        /* This is required to adhere to the
+                                           IANA standards */
+                                        brickinfo->port = 0;
+                                } else {
+                                        /* This is required to have proper ports
+                                           assigned to bricks after restart */
+                                        pmap = pmap_registry_get (THIS);
+                                        if (pmap->last_alloc <= brickinfo->port)
+                                                pmap->last_alloc =
+                                                        brickinfo->port + 1;
+                                }
+                        } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_RDMA_PORT,
+                                    strlen (GLUSTERD_STORE_KEY_BRICK_RDMA_PORT))) {
+                                gf_string2int (value, &brickinfo->rdma_port);
+
+                                if (brickinfo->rdma_port < priv->base_port) {
+                                        /* This is required to adhere to the
+                                           IANA standards */
+                                        brickinfo->rdma_port = 0;
+                                } else {
+                                        /* This is required to have proper ports
+                                           assigned to bricks after restart */
+                                        pmap = pmap_registry_get (THIS);
+                                        if (pmap->last_alloc <=
+                                            brickinfo->rdma_port)
+                                                pmap->last_alloc =
+                                                        brickinfo->rdma_port +1;
+                                }
+
+                        } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED,
+                                             strlen (GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED))) {
+                                gf_string2int (value, &brickinfo->decommissioned);
+                        } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH,
+                                             strlen (GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH))) {
+                                strncpy (brickinfo->device_path, value,
+                                         sizeof (brickinfo->device_path));
+                        } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS,
+                                             strlen (GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS))) {
+                                gf_string2int (value, &brickinfo->snap_status);
+                        } else if (!strncmp (key,
+                                    GLUSTERD_STORE_KEY_BRICK_VGNAME,
+                                    strlen (GLUSTERD_STORE_KEY_BRICK_VGNAME))) {
+                                strncpy (brickinfo->vg, value,
+                                         sizeof (brickinfo->vg));
                         } else {
                                 gf_log ("", GF_LOG_ERROR, "Unknown key: %s",
                                         key);
@@ -893,21 +2009,25 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo)
                         key = NULL;
                         value = NULL;
 
-                        ret = glusterd_store_iter_get_next (iter, &key, &value);
+                        ret = gf_store_iter_get_next (iter, &key, &value,
+                                                      &op_errno);
                 }
 
-                ret = glusterd_store_iter_destroy (iter);
+                if (op_errno != GD_STORE_EOF)
+                        goto out;
+                ret = gf_store_iter_destroy (iter);
 
                 if (ret)
                         goto out;
 
                 list_add_tail (&brickinfo->brick_list, &volinfo->bricks);
-                glusterd_for_each_entry (entry, dir);
+                brick_count++;
         }
 
+        ret = gf_store_iter_destroy (tmpiter);
+        if (ret)
+                goto out;
 out:
-        if (dir)
-                closedir (dir);
         gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
 
         return ret;
@@ -915,73 +2035,334 @@ out:
 
 
 int32_t
-glusterd_store_retrieve_volume (char    *volname)
+glusterd_store_retrieve_rbstate (glusterd_volinfo_t *volinfo)
 {
-        int32_t                 ret = -1;
-        glusterd_volinfo_t      *volinfo = NULL;
-        glusterd_store_iter_t   *iter = NULL;
-        char                    *key = NULL;
-        char                    *value = NULL;
-        char                    volpath[PATH_MAX] = {0,};
-        glusterd_conf_t         *priv = NULL;
-        char                    path[PATH_MAX] = {0,};
+        int32_t                   ret                   = -1;
+        gf_store_iter_t           *iter                 = NULL;
+        char                      *key                  = NULL;
+        char                      *value                = NULL;
+        char                      volpath[PATH_MAX]     = {0,};
+        glusterd_conf_t           *priv                 = NULL;
+        char                      path[PATH_MAX]        = {0,};
+        gf_store_op_errno_t       op_errno              = GD_STORE_SUCCESS;
+        xlator_t                 *this                  = NULL;
 
-        ret = glusterd_volinfo_new (&volinfo);
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (volinfo);
+
+        GLUSTERD_GET_VOLUME_DIR(volpath, volinfo, priv);
+        snprintf (path, sizeof (path), "%s/%s", volpath,
+                  GLUSTERD_VOLUME_RBSTATE_FILE);
+
+        ret = gf_store_handle_retrieve (path, &volinfo->rb_shandle);
 
         if (ret)
                 goto out;
 
-        strncpy (volinfo->volname, volname, GLUSTERD_MAX_VOLUME_NAME);
+        ret = gf_store_iter_new (volinfo->rb_shandle, &iter);
 
-        priv = THIS->private;
+        if (ret)
+                goto out;
+
+        ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+        if (ret)
+                goto out;
+
+        while (!ret) {
+                if (!strncmp (key, GLUSTERD_STORE_KEY_RB_STATUS,
+                              strlen (GLUSTERD_STORE_KEY_RB_STATUS))) {
+                        volinfo->rep_brick.rb_status = atoi (value);
+                }
+
+                if (volinfo->rep_brick.rb_status > GF_RB_STATUS_NONE) {
+                        if (!strncmp (key, GLUSTERD_STORE_KEY_RB_SRC_BRICK,
+                                      strlen (GLUSTERD_STORE_KEY_RB_SRC_BRICK))) {
+                                ret = glusterd_brickinfo_new_from_brick (value,
+                                                &volinfo->rep_brick.src_brick);
+                                if (ret)
+                                        goto out;
+                        } else if (!strncmp (key, GLUSTERD_STORE_KEY_RB_DST_BRICK,
+                                             strlen (GLUSTERD_STORE_KEY_RB_DST_BRICK))) {
+                                ret = glusterd_brickinfo_new_from_brick (value,
+                                                &volinfo->rep_brick.dst_brick);
+                                if (ret)
+                                        goto out;
+                        } else if (!strncmp (key, GLUSTERD_STORE_KEY_RB_DST_PORT,
+                                             strlen (GLUSTERD_STORE_KEY_RB_DST_PORT))) {
+                                switch (volinfo->transport_type) {
+                                case GF_TRANSPORT_RDMA:
+                                volinfo->rep_brick.dst_brick->rdma_port =
+                                                 atoi (value);
+                                        break;
+
+                                case GF_TRANSPORT_TCP:
+                                case GF_TRANSPORT_BOTH_TCP_RDMA:
+                                volinfo->rep_brick.dst_brick->port =
+                                                atoi (value);
+                                        break;
+                                }
+                        } else if (!strncmp (key, GF_REPLACE_BRICK_TID_KEY,
+                                             strlen (GF_REPLACE_BRICK_TID_KEY))) {
+                                        uuid_parse (value,
+                                                    volinfo->rep_brick.rb_id);
+                        }
+                }
+
+                GF_FREE (key);
+                GF_FREE (value);
+                key = NULL;
+                value = NULL;
+
+                ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+        }
+
+        if (op_errno != GD_STORE_EOF)
+                goto out;
+
+        ret = gf_store_iter_destroy (iter);
+
+        if (ret)
+                goto out;
+
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret);
+
+        return ret;
+}
+
+int32_t
+glusterd_store_retrieve_node_state (glusterd_volinfo_t *volinfo)
+{
+        int32_t                   ret                   = -1;
+        gf_store_iter_t           *iter                 = NULL;
+        char                      *key                  = NULL;
+        char                      *value                = NULL;
+        char                      volpath[PATH_MAX]     = {0,};
+        glusterd_conf_t           *priv                 = NULL;
+        char                      path[PATH_MAX]        = {0,};
+        gf_store_op_errno_t       op_errno              = GD_STORE_SUCCESS;
+        xlator_t                 *this                  = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (volinfo);
 
         GLUSTERD_GET_VOLUME_DIR(volpath, volinfo, priv);
         snprintf (path, sizeof (path), "%s/%s", volpath,
-                  GLUSTERD_VOLUME_INFO_FILE);
+                  GLUSTERD_NODE_STATE_FILE);
+
+        ret = gf_store_handle_retrieve (path, &volinfo->node_state_shandle);
+        if (ret)
+                goto out;
 
-        ret = glusterd_store_handle_new (path, &volinfo->shandle);
+        ret = gf_store_iter_new (volinfo->node_state_shandle, &iter);
 
         if (ret)
                 goto out;
 
-        ret = glusterd_store_iter_new (volinfo->shandle, &iter);
+        ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+        if (ret)
+                goto out;
+
+        while (ret == 0) {
+                if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_DEFRAG,
+                              strlen (GLUSTERD_STORE_KEY_VOL_DEFRAG))) {
+                        volinfo->rebal.defrag_cmd = atoi (value);
+                }
+
+                if (volinfo->rebal.defrag_cmd) {
+                        if (!strncmp (key, GF_REBALANCE_TID_KEY,
+                                       strlen (GF_REBALANCE_TID_KEY)))
+                                uuid_parse (value, volinfo->rebal.rebalance_id);
+
+                        if (!strncmp (key, GLUSTERD_STORE_KEY_DEFRAG_OP,
+                                      strlen (GLUSTERD_STORE_KEY_DEFRAG_OP)))
+                                volinfo->rebal.op = atoi (value);
+                }
+
+                GF_FREE (key);
+                GF_FREE (value);
+                key = NULL;
+                value = NULL;
+
+                ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+        }
+
+        if (op_errno != GD_STORE_EOF)
+                goto out;
+
+        ret = gf_store_iter_destroy (iter);
 
         if (ret)
                 goto out;
 
-        ret = glusterd_store_iter_get_next (iter, &key, &value);
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret);
+
+        return ret;
+}
+
+
+int
+glusterd_store_update_volinfo (glusterd_volinfo_t *volinfo)
+{
+        int                     ret = -1;
+        int                     exists                  = 0;
+        char                    *key                    = NULL;
+        char                    *value                  = NULL;
+        char                    volpath[PATH_MAX]       = {0,};
+        char                    path[PATH_MAX]          = {0,};
+        xlator_t                *this                   = NULL;
+        glusterd_conf_t         *conf                   = NULL;
+        gf_store_iter_t         *iter                   = NULL;
+        gf_store_op_errno_t     op_errno                = GD_STORE_SUCCESS;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = THIS->private;
+        GF_ASSERT (volinfo);
+
+        GLUSTERD_GET_VOLUME_DIR(volpath, volinfo, conf);
+
+        snprintf (path, sizeof (path), "%s/%s", volpath,
+                  GLUSTERD_VOLUME_INFO_FILE);
+
+        ret = gf_store_handle_retrieve (path, &volinfo->shandle);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "volinfo handle is NULL");
+                goto out;
+        }
+
+        ret = gf_store_iter_new (volinfo->shandle, &iter);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get new store "
+                        "iter");
+                goto out;
+        }
+
+        ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get next store "
+                        "iter");
+                goto out;
+        }
 
         while (!ret) {
+                gf_log ("", GF_LOG_DEBUG, "key = %s value = %s", key, value);
                 if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_TYPE,
                               strlen (GLUSTERD_STORE_KEY_VOL_TYPE))) {
                         volinfo->type = atoi (value);
                 } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_COUNT,
-                            strlen (GLUSTERD_STORE_KEY_VOL_COUNT))) {
+                                     strlen (GLUSTERD_STORE_KEY_VOL_COUNT))) {
                         volinfo->brick_count = atoi (value);
                 } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_STATUS,
-                            strlen (GLUSTERD_STORE_KEY_VOL_STATUS))) {
+                                     strlen (GLUSTERD_STORE_KEY_VOL_STATUS))) {
                         volinfo->status = atoi (value);
                 } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_VERSION,
-                            strlen (GLUSTERD_STORE_KEY_VOL_VERSION))) {
+                                     strlen (GLUSTERD_STORE_KEY_VOL_VERSION))) {
                         volinfo->version = atoi (value);
                 } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_PORT,
-                            strlen (GLUSTERD_STORE_KEY_VOL_PORT))) {
+                                     strlen (GLUSTERD_STORE_KEY_VOL_PORT))) {
                         volinfo->port = atoi (value);
                 } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_SUB_COUNT,
-                            strlen (GLUSTERD_STORE_KEY_VOL_SUB_COUNT))) {
+                                     strlen (GLUSTERD_STORE_KEY_VOL_SUB_COUNT))) {
                         volinfo->sub_count = atoi (value);
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_STRIPE_CNT,
+                                     strlen (GLUSTERD_STORE_KEY_VOL_STRIPE_CNT))) {
+                        volinfo->stripe_count = atoi (value);
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_REPLICA_CNT,
+                                     strlen (GLUSTERD_STORE_KEY_VOL_REPLICA_CNT))) {
+                        volinfo->replica_count = atoi (value);
                 } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_TRANSPORT,
-                            strlen (GLUSTERD_STORE_KEY_VOL_TRANSPORT))) {
+                                     strlen (GLUSTERD_STORE_KEY_VOL_TRANSPORT))) {
                         volinfo->transport_type = atoi (value);
+                        volinfo->nfs_transport_type = volinfo->transport_type;
+                        if (volinfo->transport_type == GF_TRANSPORT_BOTH_TCP_RDMA) {
+                                volinfo->nfs_transport_type = GF_DEFAULT_NFS_TRANSPORT;
+                        }
                 } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_ID,
-                            strlen (GLUSTERD_STORE_KEY_VOL_ID))) {
+                                     strlen (GLUSTERD_STORE_KEY_VOL_ID))) {
                         ret = uuid_parse (value, volinfo->volume_id);
                         if (ret)
                                 gf_log ("", GF_LOG_WARNING,
                                         "failed to parse uuid");
+
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_USERNAME,
+                                     strlen (GLUSTERD_STORE_KEY_USERNAME))) {
+
+                        glusterd_auth_set_username (volinfo, value);
+
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_PASSWORD,
+                                     strlen (GLUSTERD_STORE_KEY_PASSWORD))) {
+
+                        glusterd_auth_set_password (volinfo, value);
+
+                } else if (strstr (key, "slave")) {
+                        ret = dict_set_dynstr (volinfo->gsync_slaves, key,
+                                                gf_strdup (value));
+                        if (ret) {
+                                gf_log ("",GF_LOG_ERROR, "Error in "
+                                                "dict_set_str");
+                                goto out;
+                        }
+                        gf_log ("", GF_LOG_DEBUG, "Parsed as "GEOREP" "
+                                " slave:key=%s,value:%s", key, value);
+
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_OP_VERSION,
+                                strlen (GLUSTERD_STORE_KEY_VOL_OP_VERSION))) {
+                        volinfo->op_version = atoi (value);
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION,
+                                strlen (GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION))) {
+                        volinfo->client_op_version = atoi (value);
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_CAPS,
+                                     strlen (GLUSTERD_STORE_KEY_VOL_CAPS))) {
+                        volinfo->caps = atoi (value);
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT,
+                                strlen (GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT))) {
+                        volinfo->snap_max_hard_limit = (uint64_t) atoll (value);
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_VOL_IS_RESTORED,
+                                     strlen (GLUSTERD_STORE_KEY_VOL_IS_RESTORED))) {
+                        volinfo->is_volume_restored = atoi (value);
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_PARENT_VOLNAME,
+                                strlen (GLUSTERD_STORE_KEY_PARENT_VOLNAME))) {
+                        strncpy (volinfo->parent_volname, value, sizeof(volinfo->parent_volname) - 1);
                 } else {
-                        gf_log ("", GF_LOG_ERROR, "Unknown key: %s",
+
+                        if (is_key_glusterd_hooks_friendly (key)) {
+                                exists = 1;
+
+                        } else  {
+                                exists = glusterd_check_option_exists (key,
+                                                                       NULL);
+                        }
+
+                        switch (exists) {
+                        case -1:
+                                ret = -1;
+                                goto out;
+
+                        case 0:
+                                gf_log ("", GF_LOG_ERROR, "Unknown key: %s",
                                         key);
+                                break;
+
+                        case 1:
+                                ret = dict_set_str(volinfo->dict, key,
+                                                   gf_strdup (value));
+                                if (ret) {
+                                        gf_log ("",GF_LOG_ERROR, "Error in "
+                                                "dict_set_str");
+                                        goto out;
+                                }
+                                gf_log ("", GF_LOG_DEBUG, "Parsed as Volume-"
+                                        "set:key=%s,value:%s", key, value);
+                                break;
+                        }
                 }
 
                 GF_FREE (key);
@@ -989,62 +2370,293 @@ glusterd_store_retrieve_volume (char    *volname)
                 key = NULL;
                 value = NULL;
 
-                ret = glusterd_store_iter_get_next (iter, &key, &value);
+                ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+        }
+
+        /* backward compatibility */
+        {
+
+                switch (volinfo->type) {
+
+                        case GF_CLUSTER_TYPE_NONE:
+                                volinfo->stripe_count  = 1;
+                                volinfo->replica_count = 1;
+                        break;
+
+                        case GF_CLUSTER_TYPE_STRIPE:
+                                volinfo->stripe_count  = volinfo->sub_count;
+                                volinfo->replica_count = 1;
+                        break;
+
+                        case GF_CLUSTER_TYPE_REPLICATE:
+                                volinfo->stripe_count  = 1;
+                                volinfo->replica_count = volinfo->sub_count;
+                        break;
+
+                        case GF_CLUSTER_TYPE_STRIPE_REPLICATE:
+                                /* Introduced in 3.3 */
+                                GF_ASSERT (volinfo->stripe_count > 0);
+                                GF_ASSERT (volinfo->replica_count > 0);
+                        break;
+
+                        default:
+                                GF_ASSERT (0);
+                        break;
+                }
+
+                volinfo->dist_leaf_count = glusterd_get_dist_leaf_count (volinfo);
+
+                volinfo->subvol_count = (volinfo->brick_count /
+                                         volinfo->dist_leaf_count);
+
+                /* Only calculate volume op-versions if they are not found */
+                if (!volinfo->op_version && !volinfo->client_op_version)
+                        gd_update_volume_op_versions (volinfo);
         }
 
-        ret = glusterd_store_iter_destroy (iter);
+        if (op_errno != GD_STORE_EOF)
+                goto out;
+
+        ret = gf_store_iter_destroy (iter);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to destroy store "
+                        "iter");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+glusterd_volinfo_t*
+glusterd_store_retrieve_volume (char *volname, glusterd_snap_t *snap)
+{
+        int32_t                    ret                  = -1;
+        glusterd_volinfo_t        *volinfo              = NULL;
+        glusterd_volinfo_t        *origin_volinfo       = NULL;
+        glusterd_conf_t           *priv                 = NULL;
+        xlator_t                  *this                 = NULL;
 
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (volname);
+
+        ret = glusterd_volinfo_new (&volinfo);
         if (ret)
                 goto out;
 
+        priv = THIS->private;
+
+        strncpy (volinfo->volname, volname, GLUSTERD_MAX_VOLUME_NAME);
+        volinfo->snapshot = snap;
+        if (snap)
+                volinfo->is_snap_volume = _gf_true;
+
+        ret = glusterd_store_update_volinfo (volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to update volinfo "
+                        "for %s volume", volname);
+                goto out;
+        }
+
         ret = glusterd_store_retrieve_bricks (volinfo);
+        if (ret)
+                goto out;
 
-        list_add_tail (&volinfo->vol_list, &priv->volumes);
+        ret = glusterd_volume_compute_cksum (volinfo);
+        if (ret)
+                goto out;
 
+        if (!snap) {
+                list_add_tail (&volinfo->vol_list, &priv->volumes);
+        } else {
+                ret = glusterd_volinfo_find (volinfo->parent_volname,
+                                             &origin_volinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Parent volinfo "
+                                "not found for %s volume", volname);
+                        goto out;
+                }
+                glusterd_list_add_snapvol (origin_volinfo, volinfo);
+        }
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        if (ret) {
+                if (volinfo)
+                        glusterd_volinfo_delete (volinfo);
+                volinfo = NULL;
+        }
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret);
+
+        return volinfo;
+}
+
+inline void
+glusterd_store_set_options_path (glusterd_conf_t *conf, char *path, size_t len)
+{
+        snprintf (path, len, "%s/options", conf->workdir);
+}
+
+int
+_store_global_opts (dict_t *this, char *key, data_t *value, void *data)
+{
+        gf_store_handle_t *shandle = data;
+
+        gf_store_save_value (shandle->fd, key, (char*)value->data);
+        return 0;
+}
+
+int32_t
+glusterd_store_options (xlator_t *this, dict_t *opts)
+{
+        gf_store_handle_t               *shandle = NULL;
+        glusterd_conf_t                 *conf = NULL;
+        char                            path[PATH_MAX] = {0};
+        int                             fd = -1;
+        int32_t                         ret = -1;
+
+        conf = this->private;
+        glusterd_store_set_options_path (conf, path, sizeof (path));
 
+        ret = gf_store_handle_new (path, &shandle);
+        if (ret)
+                goto out;
+
+        fd = gf_store_mkstemp (shandle);
+        if (fd <= 0) {
+                ret = -1;
+                goto out;
+        }
+
+        shandle->fd = fd;
+        dict_foreach (opts, _store_global_opts, shandle);
+        shandle->fd = 0;
+        ret = gf_store_rename_tmppath (shandle);
+        if (ret)
+                goto out;
+out:
+        gf_store_handle_destroy (shandle);
+        if (fd >=0 )
+                close (fd);
         return ret;
 }
 
+int32_t
+glusterd_store_retrieve_options (xlator_t *this)
+{
+        char                            path[PATH_MAX] = {0};
+        glusterd_conf_t                 *conf = NULL;
+        gf_store_handle_t               *shandle = NULL;
+        gf_store_iter_t                 *iter = NULL;
+        char                            *key = NULL;
+        char                            *value = NULL;
+        gf_store_op_errno_t             op_errno = 0;
+        int                             ret = -1;
+
+        conf = this->private;
+        glusterd_store_set_options_path (conf, path, sizeof (path));
+
+        ret = gf_store_handle_retrieve (path, &shandle);
+        if (ret)
+                goto out;
+
+        ret = gf_store_iter_new (shandle, &iter);
+        if (ret)
+                goto out;
+
+        ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+        while (!ret) {
+                ret = dict_set_dynstr (conf->opts, key, value);
+                if (ret) {
+                        GF_FREE (key);
+                        GF_FREE (value);
+                        goto out;
+                }
+                GF_FREE (key);
+                key = NULL;
+                value = NULL;
+
+                ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+        }
+        if (op_errno != GD_STORE_EOF)
+                goto out;
+        ret = 0;
+out:
+        gf_store_iter_destroy (iter);
+        gf_store_handle_destroy (shandle);
+        return ret;
+}
 
 int32_t
-glusterd_store_retrieve_volumes (xlator_t  *this)
+glusterd_store_retrieve_volumes (xlator_t  *this, glusterd_snap_t *snap)
 {
-        int32_t         ret = 0;
-        char            path[PATH_MAX] = {0,};
-        glusterd_conf_t *priv = NULL;
-        DIR             *dir = NULL;
-        struct dirent   *entry = NULL;
+        int32_t                ret              = -1;
+        char                   path[PATH_MAX]   = {0,};
+        glusterd_conf_t       *priv             = NULL;
+        DIR                   *dir              = NULL;
+        struct dirent         *entry            = NULL;
+        glusterd_volinfo_t    *volinfo          = NULL;
 
         GF_ASSERT (this);
         priv = this->private;
 
         GF_ASSERT (priv);
 
-        snprintf (path, PATH_MAX, "%s/%s", priv->workdir,
-                        GLUSTERD_VOLUME_DIR_PREFIX);
+        if (snap)
+                snprintf (path, PATH_MAX, "%s/snaps/%s", priv->workdir,
+                          snap->snapname);
+        else
+                snprintf (path, PATH_MAX, "%s/%s", priv->workdir,
+                          GLUSTERD_VOLUME_DIR_PREFIX);
 
         dir = opendir (path);
 
         if (!dir) {
                 gf_log ("", GF_LOG_ERROR, "Unable to open dir %s", path);
-                ret = -1;
                 goto out;
         }
 
         glusterd_for_each_entry (entry, dir);
 
         while (entry) {
-                ret = glusterd_store_retrieve_volume (entry->d_name);
-                if (ret) {
+                if ( entry->d_type != DT_DIR )
+                        goto next;
+
+                volinfo = glusterd_store_retrieve_volume (entry->d_name, snap);
+                if (!volinfo) {
                         gf_log ("", GF_LOG_ERROR, "Unable to restore "
                                 "volume: %s", entry->d_name);
+                        ret = -1;
                         goto out;
                 }
+
+                ret = glusterd_store_retrieve_rbstate (volinfo);
+                if (ret) {
+                        /* Backward compatibility */
+                        gf_log ("", GF_LOG_INFO, "Creating a new rbstate "
+                                "for volume: %s.", entry->d_name);
+                        ret = glusterd_store_create_rbstate_shandle_on_absence (volinfo);
+                        ret = glusterd_store_perform_rbstate_store (volinfo);
+                }
+
+                ret = glusterd_store_retrieve_node_state (volinfo);
+                if (ret) {
+                        /* Backward compatibility */
+                        gf_log ("", GF_LOG_INFO, "Creating a new node_state "
+                                "for volume: %s.", entry->d_name);
+                        glusterd_store_create_nodestate_sh_on_absence (volinfo);
+                        ret = glusterd_store_perform_node_state_store (volinfo);
+
+                }
+
+next:
                 glusterd_for_each_entry (entry, dir);
         }
 
+        ret = 0;
 out:
         if (dir)
                 closedir (dir);
@@ -1054,79 +2666,481 @@ out:
 }
 
 int32_t
-glusterd_store_update_volume (glusterd_volinfo_t *volinfo)
+glusterd_resolve_snap_bricks (xlator_t  *this, glusterd_snap_t *snap)
 {
-        int32_t         ret = -1;
-        char            buf[1024] = {0,};
+        int32_t                 ret = -1;
+        glusterd_volinfo_t      *volinfo = NULL;
         glusterd_brickinfo_t    *brickinfo = NULL;
-        glusterd_brickinfo_t    *tmp = NULL;
 
+        GF_ASSERT (this);
+        GF_VALIDATE_OR_GOTO (this->name, snap, out);
 
-        list_for_each_entry (tmp, &volinfo->bricks, brick_list) {
-                ret = glusterd_store_delete_brick (volinfo, tmp);
-                //if (ret)
-                  //      goto out;
+        list_for_each_entry (volinfo, &snap->volumes, vol_list) {
+                list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                        ret = glusterd_resolve_brick (brickinfo);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "resolve brick failed in restore");
+                                goto out;
+                        }
+                }
         }
 
-        ret = glusterd_store_handle_truncate (volinfo->shandle);
+        ret = 0;
+
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret);
 
-        snprintf (buf, sizeof (buf), "%d", volinfo->type);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_TYPE, buf);
-        if (ret)
+        return ret;
+}
+
+int
+glusterd_store_update_snap (glusterd_snap_t *snap)
+{
+        int                     ret = -1;
+        char                    *key                    = NULL;
+        char                    *value                  = NULL;
+        char                    snappath[PATH_MAX]      = {0,};
+        char                    path[PATH_MAX]          = {0,};
+        xlator_t                *this                   = NULL;
+        glusterd_conf_t         *conf                   = NULL;
+        gf_store_iter_t         *iter                   = NULL;
+        gf_store_op_errno_t     op_errno                = GD_STORE_SUCCESS;
+
+        this = THIS;
+        conf = this->private;
+        GF_ASSERT (snap);
+
+        GLUSTERD_GET_SNAP_DIR (snappath, snap, conf);
+
+        snprintf (path, sizeof (path), "%s/%s", snappath,
+                  GLUSTERD_SNAP_INFO_FILE);
+
+        ret = gf_store_handle_retrieve (path, &snap->shandle);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "snap handle is NULL");
                 goto out;
+        }
 
-        snprintf (buf, sizeof (buf), "%d", volinfo->brick_count);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_COUNT, buf);
-        if (ret)
+        ret = gf_store_iter_new (snap->shandle, &iter);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get new store "
+                        "iter");
                 goto out;
+        }
 
-        snprintf (buf, sizeof (buf), "%d", volinfo->status);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_STATUS, buf);
-        if (ret)
+        ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get next store "
+                        "iter");
                 goto out;
+        }
 
-        snprintf (buf, sizeof (buf), "%d", volinfo->sub_count);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_SUB_COUNT, buf);
-        if (ret)
+        while (!ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "key = %s value = %s",
+                        key, value);
+
+                if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_ID,
+                                     strlen (GLUSTERD_STORE_KEY_SNAP_ID))) {
+                        ret = uuid_parse (value, snap->snap_id);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "Failed to parse uuid");
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_RESTORED,
+                                   strlen (GLUSTERD_STORE_KEY_SNAP_RESTORED))) {
+                        snap->snap_restored = atoi (value);
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_STATUS,
+                                     strlen (GLUSTERD_STORE_KEY_SNAP_STATUS))) {
+                        snap->snap_status = atoi (value);
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_DESC,
+                                     strlen (GLUSTERD_STORE_KEY_SNAP_DESC))) {
+                        snap->description = gf_strdup (value);
+                } else if (!strncmp (key, GLUSTERD_STORE_KEY_SNAP_TIMESTAMP,
+                                  strlen (GLUSTERD_STORE_KEY_SNAP_TIMESTAMP))) {
+                        snap->time_stamp = atoi (value);
+                }
+
+                GF_FREE (key);
+                GF_FREE (value);
+                key = NULL;
+                value = NULL;
+
+                ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+        }
+
+        if (op_errno != GD_STORE_EOF)
                 goto out;
 
-        snprintf (buf, sizeof (buf), "%d", volinfo->version);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_VERSION, buf);
-        if (ret)
+        ret = gf_store_iter_destroy (iter);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to destroy store "
+                        "iter");
+        }
+
+out:
+        return ret;
+}
+
+int32_t
+glusterd_store_retrieve_snap (char *snapname)
+{
+        int32_t                ret    = -1;
+        dict_t                *dict   = NULL;
+        glusterd_snap_t       *snap   = NULL;
+        glusterd_conf_t       *priv   = NULL;
+        xlator_t              *this   = NULL;
+
+        this = THIS;
+        priv = this->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (snapname);
+
+        dict = dict_new();
+        if (!dict) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to create dict");
+                ret = -1;
+                goto out;
+        }
+
+        snap = glusterd_new_snap_object ();
+        if (!snap) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to create "
+                                                  " snap object");
                 goto out;
+        }
 
-        snprintf (buf, sizeof (buf), "%d", volinfo->transport_type);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                         GLUSTERD_STORE_KEY_VOL_TRANSPORT, buf);
-        if (ret)
+        strncpy (snap->snapname, snapname, strlen(snapname));
+        ret = glusterd_store_update_snap (snap);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to update snapshot "
+                        "for %s snap", snapname);
                 goto out;
+        }
 
-        uuid_unparse (volinfo->volume_id, buf);
-        ret = glusterd_store_save_value (volinfo->shandle,
-                                        GLUSTERD_STORE_KEY_VOL_ID, buf);
-        if (ret)
+        ret = glusterd_store_retrieve_volumes (this, snap);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to retrieve "
+                        "snap volumes for snap %s", snapname);
                 goto out;
+        }
 
-        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
-                ret = glusterd_store_create_brick (volinfo, brickinfo);
+        /* Unlike bricks of normal volumes which are resolved at the end of
+           the glusterd restore, the bricks belonging to the snap volumes of
+           each snap should be resolved as part of snapshot restore itself.
+           Because if the snapshot has to be removed, then resolving bricks
+           helps glusterd in understanding what all bricks have its own uuid
+           and killing those bricks.
+        */
+        ret = glusterd_resolve_snap_bricks (this, snap);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING, "resolving the snap bricks"
+                        " failed (snap: %s)", snap?snap->snapname:"");
+
+        /* When the snapshot command from cli is received, the on disk and
+           in memory structures for the snapshot are created (with the status)
+           being marked as GD_SNAP_STATUS_INIT. Once the backend snapshot is
+           taken, the status is changed to GD_SNAP_STATUS_IN_USE. If glusterd
+           dies after taking the backend snapshot, but before updating the
+           status, then when glusterd comes up, it should treat that snapshot
+           as a failed snapshot and clean it up.
+        */
+        if (snap->snap_status != GD_SNAP_STATUS_IN_USE) {
+                ret = glusterd_snap_remove (dict, snap, _gf_true, _gf_true);
                 if (ret)
-                        goto out;
+                        gf_log (this->name, GF_LOG_WARNING, "failed to remove"
+                                " the snapshot %s", snap->snapname);
+                goto out;
+        }
+
+        /* TODO: list_add_order can do 'N-square' comparisions and
+           is not efficient. Find a better solution to store the snap
+           in order */
+        list_add_order (&snap->snap_list, &priv->snapshots,
+                        glusterd_compare_snap_time);
+
+out:
+        if (dict)
+                dict_unref (dict);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret);
+        return ret;
+}
+
+/* Read the missed_snap_list and update the in-memory structs */
+int32_t
+glusterd_store_retrieve_missed_snaps_list (xlator_t  *this)
+{
+        char                   buf[PATH_MAX]    = "";
+        char                   path[PATH_MAX]   = "";
+        char                  *missed_node_info = NULL;
+        char                  *brick_path       = NULL;
+        char                  *value            = NULL;
+        char                  *save_ptr         = NULL;
+        FILE                  *fp               = NULL;
+        int32_t                brick_num        = -1;
+        int32_t                snap_op          = -1;
+        int32_t                snap_status      = -1;
+        int32_t                ret              = -1;
+        glusterd_conf_t       *priv             = NULL;
+        gf_store_op_errno_t    store_errno      = GD_STORE_SUCCESS;
+
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        /* Get the path of the missed_snap_list */
+        glusterd_store_missed_snaps_list_path_set (path, sizeof(path));
+
+        fp = fopen (path, "r");
+        if (!fp) {
+                /* If errno is ENOENT then there are no missed snaps yet */
+                if (errno != ENOENT) {
+                        ret = -1;
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to open %s. "
+                                "Error: %s", path, strerror(errno));
+                } else {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "No missed snaps list.");
+                        ret = 0;
+                }
+                goto out;
         }
 
+        do {
+                ret = gf_store_read_and_tokenize (fp, buf,
+                                                  &missed_node_info, &value,
+                                                  &store_errno);
+                if (ret) {
+                        if (store_errno == GD_STORE_EOF) {
+                                gf_log (this->name,
+                                        GF_LOG_DEBUG,
+                                        "EOF for missed_snap_list");
+                                ret = 0;
+                                break;
+                        }
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to fetch data from "
+                                "missed_snaps_list. Error: %s",
+                                gf_store_strerror (store_errno));
+                        goto out;
+                }
+
+                /* Fetch the brick_num, brick_path, snap_op and snap status */
+                brick_num = atoi(strtok_r (value, ":", &save_ptr));
+                brick_path = strtok_r (NULL, ":", &save_ptr);
+                snap_op = atoi(strtok_r (NULL, ":", &save_ptr));
+                snap_status = atoi(strtok_r (NULL, ":", &save_ptr));
+
+                if (!missed_node_info || !brick_path ||
+                    brick_num < 1 || snap_op < 1 ||
+                    snap_status < 1) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Invalid missed_snap_entry");
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = glusterd_store_missed_snaps_list (missed_node_info,
+                                                        brick_num,
+                                                        brick_path,
+                                                        snap_op,
+                                                        snap_status);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to store missed snaps_list");
+                        goto out;
+                }
+
+        } while (store_errno == GD_STORE_SUCCESS);
+
         ret = 0;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning with %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_store_retrieve_snaps (xlator_t  *this)
+{
+        int32_t                ret              = 0;
+        char                   path[PATH_MAX]   = {0,};
+        glusterd_conf_t       *priv             = NULL;
+        DIR                   *dir              = NULL;
+        struct dirent         *entry            = NULL;
+
+        GF_ASSERT (this);
+        priv = this->private;
+
+        GF_ASSERT (priv);
+
+        snprintf (path, PATH_MAX, "%s/snaps", priv->workdir);
+
+        dir = opendir (path);
+
+        if (!dir) {
+                /* If snaps dir doesn't exists ignore the error for
+                   backward compatibility */
+                if (errno != ENOENT) {
+                        ret = -1;
+                        gf_log ("", GF_LOG_ERROR, "Unable to open dir %s", path);
+                }
+                goto out;
+        }
+
+        glusterd_for_each_entry (entry, dir);
 
+        while (entry) {
+                if (entry->d_type == DT_DIR) {
+                        ret = glusterd_store_retrieve_snap (entry->d_name);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to restore snapshot: %s",
+                                        entry->d_name);
+                                goto out;
+                        }
+                }
+
+                glusterd_for_each_entry (entry, dir);
+        }
+
+        /* Retrieve missed_snaps_list */
+        ret = glusterd_store_retrieve_missed_snaps_list (this);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Failed to retrieve missed_snaps_list");
+                goto out;
+        }
 
 out:
+        if (dir)
+                closedir (dir);
         gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
 
         return ret;
 }
 
+/* Writes all the contents of conf->missed_snap_list */
+int32_t
+glusterd_store_write_missed_snapinfo (int32_t fd)
+{
+        char                           value[PATH_MAX]             = "";
+        int32_t                        ret                         = -1;
+        glusterd_conf_t               *priv                        = NULL;
+        glusterd_missed_snap_info     *missed_snapinfo             = NULL;
+        glusterd_snap_op_t            *snap_opinfo                 = NULL;
+        xlator_t                      *this                        = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        /* Write the missed_snap_entry */
+        list_for_each_entry (missed_snapinfo, &priv->missed_snaps_list,
+                             missed_snaps) {
+                list_for_each_entry (snap_opinfo,
+                                     &missed_snapinfo->snap_ops,
+                                     snap_ops_list) {
+                        snprintf (value, sizeof(value), "%d:%s:%d:%d",
+                                  snap_opinfo->brick_num,
+                                  snap_opinfo->brick_path,
+                                  snap_opinfo->op, snap_opinfo->status);
+                        ret = gf_store_save_value
+                                           (fd,
+                                            missed_snapinfo->node_snap_info,
+                                            value);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to write missed snapinfo");
+                                goto out;
+                        }
+                }
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+/* Adds the missed snap entries to the in-memory conf->missed_snap_list *
+ * and writes them to disk */
+int32_t
+glusterd_store_update_missed_snaps (dict_t *dict, int32_t missed_snap_count)
+{
+        int32_t                        fd                          = -1;
+        int32_t                        ret                         = -1;
+        glusterd_conf_t               *priv                        = NULL;
+        xlator_t                      *this                        = NULL;
+
+        this = THIS;
+        GF_ASSERT(this);
+        GF_ASSERT(dict);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        if (missed_snap_count < 1) {
+                gf_log (this->name, GF_LOG_DEBUG, "No missed snaps");
+                ret = 0;
+                goto out;
+        }
+
+        ret = glusterd_store_create_missed_snaps_list_shandle_on_absence ();
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to obtain "
+                        "missed_snaps_list store handle.");
+                goto out;
+        }
+
+        fd = gf_store_mkstemp (priv->missed_snaps_list_shandle);
+        if (fd <= 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to create tmp file");
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_add_missed_snaps_to_list (dict, missed_snap_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to add missed snaps to list");
+                goto out;
+        }
+
+        ret = glusterd_store_write_missed_snapinfo (fd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to write missed snaps to disk");
+                goto out;
+        }
+
+        ret = gf_store_rename_tmppath (priv->missed_snaps_list_shandle);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to rename the tmp file");
+                goto out;
+        }
+out:
+        if (ret && (fd > 0)) {
+                ret = gf_store_unlink_tmppath (priv->missed_snaps_list_shandle);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to unlink the tmp file");
+                }
+                ret = -1;
+        }
+
+        if (fd > 0)
+                close (fd);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
 int32_t
 glusterd_store_delete_peerinfo (glusterd_peerinfo_t *peerinfo)
 {
@@ -1134,7 +3148,6 @@ glusterd_store_delete_peerinfo (glusterd_peerinfo_t *peerinfo)
         glusterd_conf_t                 *priv = NULL;
         char                            peerdir[PATH_MAX] = {0,};
         char                            filepath[PATH_MAX] = {0,};
-        char                            str[512] = {0,};
         char                            hostname_path[PATH_MAX] = {0,};
 
 
@@ -1158,9 +3171,9 @@ glusterd_store_delete_peerinfo (glusterd_peerinfo_t *peerinfo)
                        goto out;
                 }
         } else {
-                uuid_unparse (peerinfo->uuid, str);
 
-                snprintf (filepath, PATH_MAX, "%s/%s", peerdir, str);
+                snprintf (filepath, PATH_MAX, "%s/%s", peerdir,
+                          uuid_utoa (peerinfo->uuid));
                 snprintf (hostname_path, PATH_MAX, "%s/%s",
                           peerdir, peerinfo->hostname);
 
@@ -1171,107 +3184,204 @@ glusterd_store_delete_peerinfo (glusterd_peerinfo_t *peerinfo)
         }
 
         ret = unlink (filepath);
+        if (ret && (errno == ENOENT))
+                ret = 0;
 
 out:
-        if (peerinfo->shandle)
-                glusterd_store_handle_destroy(peerinfo->shandle);
+        if (peerinfo->shandle) {
+                gf_store_handle_destroy (peerinfo->shandle);
+                peerinfo->shandle = NULL;
+        }
         gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
 
         return ret;
 }
 
+void
+glusterd_store_peerinfo_dirpath_set (char *path, size_t len)
+{
+        glusterd_conf_t                 *priv = NULL;
+        GF_ASSERT (path);
+        GF_ASSERT (len >= PATH_MAX);
+
+        priv = THIS->private;
+        snprintf (path, len, "%s/peers", priv->workdir);
+}
 
 int32_t
-glusterd_store_update_peerinfo (glusterd_peerinfo_t *peerinfo)
+glusterd_store_create_peer_dir ()
 {
-        int32_t                         ret = -1;
-        struct  stat                    stbuf = {0,};
-        glusterd_conf_t                 *priv = NULL;
-        char                            peerdir[PATH_MAX] = {0,};
-        char                            filepath[PATH_MAX] = {0,};
-        char                            str[512] = {0,};
-        char                            buf[4096] = {0,};
-        glusterd_peer_hostname_t        *hname = NULL;
-        int                             i = 0;
-        char                            hostname_path[PATH_MAX] = {0,};
+        int32_t                             ret = 0;
+        char                            path[PATH_MAX];
 
+        glusterd_store_peerinfo_dirpath_set (path, sizeof (path));
+        ret = gf_store_mkdir (path);
 
-        if (!peerinfo) {
-                ret = 0;
-                goto out;
-        }
+        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
 
-        priv = THIS->private;
+static void
+glusterd_store_uuid_peerpath_set (glusterd_peerinfo_t *peerinfo, char *peerfpath,
+                             size_t len)
+{
+        char                    peerdir[PATH_MAX];
+        char                    str[50] = {0};
 
-        snprintf (peerdir, PATH_MAX, "%s/peers", priv->workdir);
+        GF_ASSERT (peerinfo);
+        GF_ASSERT (peerfpath);
+        GF_ASSERT (len >= PATH_MAX);
 
-        ret = stat (peerdir, &stbuf);
+        glusterd_store_peerinfo_dirpath_set (peerdir, sizeof (peerdir));
+        uuid_unparse (peerinfo->uuid, str);
+        snprintf (peerfpath, len, "%s/%s", peerdir, str);
+}
 
-        if (-1 == ret) {
-                ret = mkdir (peerdir, 0777);
-                if (ret)
-                        goto out;
-        }
+static void
+glusterd_store_hostname_peerpath_set (glusterd_peerinfo_t *peerinfo,
+                                       char *peerfpath, size_t len)
+{
+        char                    peerdir[PATH_MAX];
 
-        if (uuid_is_null (peerinfo->uuid)) {
+        GF_ASSERT (peerinfo);
+        GF_ASSERT (peerfpath);
+        GF_ASSERT (len >= PATH_MAX);
 
-                if (peerinfo->hostname) {
-                        snprintf (filepath, PATH_MAX, "%s/%s", peerdir,
-                                  peerinfo->hostname);
-                } else {
-                       ret = 0;
-                       goto out;
-                }
-        } else {
-                uuid_unparse (peerinfo->uuid, str);
+        glusterd_store_peerinfo_dirpath_set (peerdir, sizeof (peerdir));
+        snprintf (peerfpath, len, "%s/%s", peerdir, peerinfo->hostname);
+}
 
-                snprintf (filepath, PATH_MAX, "%s/%s", peerdir, str);
-                snprintf (hostname_path, PATH_MAX, "%s/%s",
-                          peerdir, peerinfo->hostname);
+int32_t
+glusterd_store_peerinfo_hostname_shandle_create (glusterd_peerinfo_t *peerinfo)
+{
+        char                    peerfpath[PATH_MAX];
+        int32_t                 ret = -1;
 
-                ret = stat (hostname_path, &stbuf);
+        glusterd_store_hostname_peerpath_set (peerinfo, peerfpath,
+                                              sizeof (peerfpath));
+        ret = gf_store_handle_create_on_absence (&peerinfo->shandle,
+                                                 peerfpath);
+        return ret;
+}
 
-                if (!ret) {
-                        gf_log ("", GF_LOG_DEBUG, "Destroying store handle");
-                        glusterd_store_handle_destroy (peerinfo->shandle);
-                        peerinfo->shandle = NULL;
-                        ret = remove (hostname_path);
-                }
+int32_t
+glusterd_store_peerinfo_uuid_shandle_create (glusterd_peerinfo_t *peerinfo)
+{
+        char                    peerfpath[PATH_MAX];
+        int32_t                 ret = -1;
+
+        glusterd_store_uuid_peerpath_set (peerinfo, peerfpath,
+                                          sizeof (peerfpath));
+        ret = gf_store_handle_create_on_absence (&peerinfo->shandle,
+                                                 peerfpath);
+        return ret;
+}
+
+int32_t
+glusterd_peerinfo_hostname_shandle_check_destroy (glusterd_peerinfo_t *peerinfo)
+{
+        char                    peerfpath[PATH_MAX];
+        int32_t                 ret = -1;
+        struct  stat            stbuf = {0,};
+
+        glusterd_store_hostname_peerpath_set (peerinfo, peerfpath,
+                                              sizeof (peerfpath));
+        ret = stat (peerfpath, &stbuf);
+        if (!ret) {
+                if (peerinfo->shandle)
+                        gf_store_handle_destroy (peerinfo->shandle);
+                peerinfo->shandle = NULL;
+                ret = unlink (peerfpath);
         }
+        return ret;
+}
 
+int32_t
+glusterd_store_create_peer_shandle (glusterd_peerinfo_t *peerinfo)
+{
+        int32_t                 ret = 0;
 
-        if (!peerinfo->shandle) {
-                ret = glusterd_store_handle_new (filepath, &peerinfo->shandle);
-                if (ret)
-                        goto out;
-                ret = glusterd_store_handle_truncate (peerinfo->shandle);
+        GF_ASSERT (peerinfo);
+
+        if (glusterd_peerinfo_is_uuid_unknown (peerinfo)) {
+                ret = glusterd_store_peerinfo_hostname_shandle_create (peerinfo);
         } else {
-                ret = glusterd_store_handle_truncate (peerinfo->shandle);
-                if (ret)
-                        goto out;
+                ret = glusterd_peerinfo_hostname_shandle_check_destroy (peerinfo);
+                ret = glusterd_store_peerinfo_uuid_shandle_create (peerinfo);
         }
+        return ret;
+}
+
+int32_t
+glusterd_store_peer_write (int fd, glusterd_peerinfo_t *peerinfo)
+{
+        char                    buf[50] = {0};
+        int32_t                 ret = 0;
 
-        ret = glusterd_store_save_value (peerinfo->shandle,
-                                         GLUSTERD_STORE_KEY_PEER_UUID, str);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_PEER_UUID,
+                                   uuid_utoa (peerinfo->uuid));
         if (ret)
                 goto out;
 
         snprintf (buf, sizeof (buf), "%d", peerinfo->state.state);
-        ret = glusterd_store_save_value (peerinfo->shandle,
-                                         GLUSTERD_STORE_KEY_PEER_STATE, buf);
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_PEER_STATE, buf);
         if (ret)
                 goto out;
 
-        list_for_each_entry (hname, &peerinfo->hostnames, hostname_list) {
-                i++;
-                snprintf (buf, sizeof (buf), "%s%d",
-                          GLUSTERD_STORE_KEY_PEER_HOSTNAME, i);
-                ret = glusterd_store_save_value (peerinfo->shandle,
-                                                 buf, hname->hostname);
-                if (ret)
-                        goto out;
+        ret = gf_store_save_value (fd, GLUSTERD_STORE_KEY_PEER_HOSTNAME "1",
+                                   peerinfo->hostname);
+        if (ret)
+                goto out;
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_store_perform_peer_store (glusterd_peerinfo_t *peerinfo)
+{
+        int                                 fd = -1;
+        int32_t                             ret = -1;
+
+        GF_ASSERT (peerinfo);
+
+        fd = gf_store_mkstemp (peerinfo->shandle);
+        if (fd <= 0) {
+                ret = -1;
+                goto out;
         }
 
+        ret = glusterd_store_peer_write (fd, peerinfo);
+        if (ret)
+                goto out;
+
+        ret = gf_store_rename_tmppath (peerinfo->shandle);
+out:
+        if (ret && (fd > 0))
+                gf_store_unlink_tmppath (peerinfo->shandle);
+        if (fd > 0)
+                close (fd);
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_store_peerinfo (glusterd_peerinfo_t *peerinfo)
+{
+        int32_t                         ret = -1;
+
+        GF_ASSERT (peerinfo);
+
+        ret = glusterd_store_create_peer_dir ();
+        if (ret)
+                goto out;
+
+        ret = glusterd_store_create_peer_shandle (peerinfo);
+        if (ret)
+                goto out;
+
+        ret = glusterd_store_perform_peer_store (peerinfo);
 out:
         gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
         return ret;
@@ -1289,11 +3399,13 @@ glusterd_store_retrieve_peers (xlator_t *this)
         uuid_t                  uuid = {0,};
         char                    *hostname = NULL;
         int32_t                 state = 0;
-        glusterd_store_handle_t *shandle = NULL;
+        gf_store_handle_t       *shandle = NULL;
         char                    filepath[PATH_MAX] = {0,};
-        glusterd_store_iter_t   *iter = NULL;
+        gf_store_iter_t         *iter = NULL;
         char                    *key = NULL;
         char                    *value = NULL;
+        glusterd_peerctx_args_t args = {0};
+        gf_store_op_errno_t     op_errno = GD_STORE_SUCCESS;
 
         GF_ASSERT (this);
         priv = this->private;
@@ -1315,15 +3427,17 @@ glusterd_store_retrieve_peers (xlator_t *this)
 
         while (entry) {
                 snprintf (filepath, PATH_MAX, "%s/%s", path, entry->d_name);
-                ret = glusterd_store_handle_new (filepath, &shandle);
+                ret = gf_store_handle_retrieve (filepath, &shandle);
                 if (ret)
                         goto out;
 
-                ret = glusterd_store_iter_new (shandle, &iter);
+                ret = gf_store_iter_new (shandle, &iter);
                 if (ret)
                         goto out;
 
-                ret = glusterd_store_iter_get_next (iter, &key, &value);
+                ret = gf_store_iter_get_next (iter, &key, &value, &op_errno);
+                if (ret)
+                        goto out;
 
                 while (!ret) {
 
@@ -1349,14 +3463,18 @@ glusterd_store_retrieve_peers (xlator_t *this)
                         key = NULL;
                         value = NULL;
 
-                        ret = glusterd_store_iter_get_next (iter, &key, &value);
+                        ret = gf_store_iter_get_next (iter, &key, &value,
+                                                      &op_errno);
                 }
+                if (op_errno != GD_STORE_EOF)
+                        goto out;
 
-                (void) glusterd_store_iter_destroy (iter);
+                (void) gf_store_iter_destroy (iter);
 
                 ret = glusterd_friend_add (hostname, 0, state, &uuid,
-                                           NULL, &peerinfo, 1);
+                                           &peerinfo, 1, NULL);
 
+                GF_FREE (hostname);
                 if (ret)
                         goto out;
 
@@ -1364,6 +3482,13 @@ glusterd_store_retrieve_peers (xlator_t *this)
                 glusterd_for_each_entry (entry, dir);
         }
 
+        args.mode = GD_MODE_ON;
+        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
+                ret = glusterd_friend_rpc_create (this, peerinfo, &args);
+                if (ret)
+                        goto out;
+        }
+
 out:
         if (dir)
                 closedir (dir);
@@ -1405,13 +3530,23 @@ out:
 int32_t
 glusterd_restore ()
 {
-        int             ret = -1;
+        int32_t         ret = -1;
         xlator_t        *this = NULL;
 
         this = THIS;
 
-        ret = glusterd_store_retrieve_volumes (this);
+        ret = glusterd_restore_op_version (this);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to restore op_version");
+                goto out;
+        }
+
+        ret = glusterd_store_retrieve_volumes (this, NULL);
+        if (ret)
+                goto out;
 
+        ret = glusterd_store_retrieve_snaps (this);
         if (ret)
                 goto out;
 
diff --git a/xlators/mgmt/glusterd/src/glusterd-store.h b/xlators/mgmt/glusterd/src/glusterd-store.h
index af05815d3..1b5cebc0c 100644
--- a/xlators/mgmt/glusterd/src/glusterd-store.h
+++ b/xlators/mgmt/glusterd/src/glusterd-store.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _GLUSTERD_HA_H_
 #define _GLUSTERD_HA_H_
 
@@ -30,6 +20,7 @@
 
 #include "glusterfs.h"
 #include "xlator.h"
+#include "run.h"
 #include "logging.h"
 #include "call-stub.h"
 #include "fd.h"
@@ -37,26 +28,62 @@
 #include "glusterd.h"
 #include "rpcsvc.h"
 
-
-#define GLUSTERD_STORE_UUID_KEY           "UUID"
-
-#define GLUSTERD_STORE_KEY_VOL_TYPE       "type"
-#define GLUSTERD_STORE_KEY_VOL_COUNT      "count"
-#define GLUSTERD_STORE_KEY_VOL_STATUS     "status"
-#define GLUSTERD_STORE_KEY_VOL_PORT       "port"
-#define GLUSTERD_STORE_KEY_VOL_SUB_COUNT  "sub_count"
-#define GLUSTERD_STORE_KEY_VOL_BRICK      "brick"
-#define GLUSTERD_STORE_KEY_VOL_VERSION    "version"
-#define GLUSTERD_STORE_KEY_VOL_TRANSPORT  "transport-type"
-#define GLUSTERD_STORE_KEY_VOL_ID         "volume-id"
-
-#define GLUSTERD_STORE_KEY_BRICK_HOSTNAME "hostname"
-#define GLUSTERD_STORE_KEY_BRICK_PATH     "path"
-#define GLUSTERD_STORE_KEY_BRICK_PORT     "listen-port"
-
-#define GLUSTERD_STORE_KEY_PEER_UUID      "uuid"
-#define GLUSTERD_STORE_KEY_PEER_HOSTNAME  "hostname"
-#define GLUSTERD_STORE_KEY_PEER_STATE     "state"
+typedef enum glusterd_store_ver_ac_{
+        GLUSTERD_VOLINFO_VER_AC_NONE = 0,
+        GLUSTERD_VOLINFO_VER_AC_INCREMENT = 1,
+        GLUSTERD_VOLINFO_VER_AC_DECREMENT = 2,
+} glusterd_volinfo_ver_ac_t;
+
+
+#define GLUSTERD_STORE_UUID_KEY                 "UUID"
+
+#define GLUSTERD_STORE_KEY_VOL_TYPE             "type"
+#define GLUSTERD_STORE_KEY_VOL_COUNT            "count"
+#define GLUSTERD_STORE_KEY_VOL_STATUS           "status"
+#define GLUSTERD_STORE_KEY_VOL_PORT             "port"
+#define GLUSTERD_STORE_KEY_VOL_SUB_COUNT        "sub_count"
+#define GLUSTERD_STORE_KEY_VOL_STRIPE_CNT       "stripe_count"
+#define GLUSTERD_STORE_KEY_VOL_REPLICA_CNT      "replica_count"
+#define GLUSTERD_STORE_KEY_VOL_BRICK            "brick"
+#define GLUSTERD_STORE_KEY_VOL_VERSION          "version"
+#define GLUSTERD_STORE_KEY_VOL_TRANSPORT        "transport-type"
+#define GLUSTERD_STORE_KEY_VOL_ID               "volume-id"
+#define GLUSTERD_STORE_KEY_VOL_IS_RESTORED      "is-volume-restored"
+#define GLUSTERD_STORE_KEY_RB_STATUS            "rb_status"
+#define GLUSTERD_STORE_KEY_RB_SRC_BRICK         "rb_src"
+#define GLUSTERD_STORE_KEY_RB_DST_BRICK         "rb_dst"
+#define GLUSTERD_STORE_KEY_RB_DST_PORT          "rb_port"
+#define GLUSTERD_STORE_KEY_VOL_DEFRAG           "rebalance_status"
+#define GLUSTERD_STORE_KEY_DEFRAG_OP            "rebalance_op"
+#define GLUSTERD_STORE_KEY_USERNAME             "username"
+#define GLUSTERD_STORE_KEY_PASSWORD             "password"
+#define GLUSTERD_STORE_KEY_PARENT_VOLNAME       "parent_volname"
+#define GLUSTERD_STORE_KEY_VOL_OP_VERSION       "op-version"
+#define GLUSTERD_STORE_KEY_VOL_CLIENT_OP_VERSION "client-op-version"
+
+#define GLUSTERD_STORE_KEY_SNAP_NAME            "name"
+#define GLUSTERD_STORE_KEY_SNAP_ID              "snap-id"
+#define GLUSTERD_STORE_KEY_SNAP_DESC            "desc"
+#define GLUSTERD_STORE_KEY_SNAP_TIMESTAMP       "time-stamp"
+#define GLUSTERD_STORE_KEY_SNAP_STATUS          "status"
+#define GLUSTERD_STORE_KEY_SNAP_RESTORED        "snap-restored"
+#define GLUSTERD_STORE_KEY_SNAP_MAX_HARD_LIMIT  "snap-max-hard-limit"
+#define GLUSTERD_STORE_KEY_SNAP_MAX_SOFT_LIMIT  "snap-max-soft-limit"
+
+#define GLUSTERD_STORE_KEY_BRICK_HOSTNAME       "hostname"
+#define GLUSTERD_STORE_KEY_BRICK_PATH           "path"
+#define GLUSTERD_STORE_KEY_BRICK_PORT           "listen-port"
+#define GLUSTERD_STORE_KEY_BRICK_RDMA_PORT      "rdma.listen-port"
+#define GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED "decommissioned"
+#define GLUSTERD_STORE_KEY_BRICK_VGNAME         "vg"
+#define GLUSTERD_STORE_KEY_BRICK_DEVICE_PATH    "device_path"
+#define GLUSTERD_STORE_KEY_BRICK_SNAP_STATUS    "snap-status"
+
+#define GLUSTERD_STORE_KEY_PEER_UUID            "uuid"
+#define GLUSTERD_STORE_KEY_PEER_HOSTNAME        "hostname"
+#define GLUSTERD_STORE_KEY_PEER_STATE           "state"
+
+#define GLUSTERD_STORE_KEY_VOL_CAPS             "caps"
 
 #define glusterd_for_each_entry(entry, dir) \
         do {\
@@ -64,51 +91,74 @@
                 if (dir) {\
                         entry = readdir (dir);\
                         while (entry && (!strcmp (entry->d_name, ".") ||\
+                            !fnmatch ("*.tmp", entry->d_name, 0) ||\
                             !strcmp (entry->d_name, ".."))) {\
                                 entry = readdir (dir);\
                         }\
                 }\
         } while (0); \
 
+
 int32_t
-glusterd_store_create_volume (glusterd_volinfo_t *volinfo);
+glusterd_store_volinfo (glusterd_volinfo_t *volinfo, glusterd_volinfo_ver_ac_t ac);
 
 int32_t
 glusterd_store_delete_volume (glusterd_volinfo_t *volinfo);
 
 int32_t
-glusterd_store_uuid ();
+glusterd_store_delete_snap (glusterd_snap_t *snap);
 
 int32_t
-glusterd_store_handle_new (char *path, glusterd_store_handle_t **handle);
+glusterd_retrieve_uuid ();
 
 int32_t
-glusterd_store_save_value (glusterd_store_handle_t *handle,
-                           char *key, char *value);
+glusterd_store_peerinfo (glusterd_peerinfo_t *peerinfo);
 
 int32_t
-glusterd_store_retrieve_value (glusterd_store_handle_t *handle,
-                               char *key, char **value);
+glusterd_store_delete_peerinfo (glusterd_peerinfo_t *peerinfo);
 
 int32_t
-glusterd_store_update_volume (glusterd_volinfo_t *volinfo);
+glusterd_store_delete_brick (glusterd_brickinfo_t *brickinfo,
+                             char *delete_path);
 
 int32_t
-glusterd_retrieve_uuid ();
+glusterd_restore ();
+
+void
+glusterd_perform_volinfo_version_action (glusterd_volinfo_t *volinfo,
+                                         glusterd_volinfo_ver_ac_t ac);
+gf_boolean_t
+glusterd_store_is_valid_brickpath (char *volname, char *brick);
 
 int32_t
-glusterd_store_update_peerinfo (glusterd_peerinfo_t *peerinfo);
+glusterd_store_perform_node_state_store (glusterd_volinfo_t *volinfo);
+
+int
+glusterd_retrieve_op_version (xlator_t *this, int *op_version);
+
+int
+glusterd_store_global_info (xlator_t *this);
 
 int32_t
-glusterd_store_delete_peerinfo (glusterd_peerinfo_t *peerinfo);
+glusterd_store_retrieve_options (xlator_t *this);
 
 int32_t
-glusterd_store_delete_brick (glusterd_volinfo_t *volinfo,
-                             glusterd_brickinfo_t *brickinfo);
+glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo);
 
 int32_t
-glusterd_store_handle_destroy (glusterd_store_handle_t *handle);
+glusterd_store_options (xlator_t *this, dict_t *opts);
+
+void
+glusterd_replace_slash_with_hyphen (char *str);
 
 int32_t
-glusterd_restore ();
+glusterd_store_perform_volume_store (glusterd_volinfo_t *volinfo);
+
+int32_t
+glusterd_store_snap (glusterd_snap_t *snap);
+
+int32_t
+glusterd_store_update_missed_snaps (dict_t *dict,
+                                    int32_t missed_snap_count);
+
 #endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-syncop.c b/xlators/mgmt/glusterd/src/glusterd-syncop.c
new file mode 100644
index 000000000..438df8266
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-syncop.c
@@ -0,0 +1,1639 @@
+/*
+   Copyright (c) 2012-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+/* rpc related syncops */
+#include "rpc-clnt.h"
+#include "protocol-common.h"
+#include "xdr-generic.h"
+#include "glusterd1-xdr.h"
+#include "glusterd-syncop.h"
+
+#include "glusterd.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-utils.h"
+#include "glusterd-locks.h"
+
+extern glusterd_op_info_t opinfo;
+
+void
+gd_synctask_barrier_wait (struct syncargs *args, int count)
+{
+        glusterd_conf_t *conf = THIS->private;
+
+        synclock_unlock (&conf->big_lock);
+        synctask_barrier_wait (args, count);
+        synclock_lock (&conf->big_lock);
+
+	syncbarrier_destroy (&args->barrier);
+}
+
+static void
+gd_mgmt_v3_collate_errors (struct syncargs *args, int op_ret, int op_errno,
+                           char *op_errstr, int op_code,
+                           glusterd_peerinfo_t *peerinfo, u_char *uuid)
+{
+        char     err_str[PATH_MAX] = "Please check log file for details.";
+        char     op_err[PATH_MAX] = "";
+        int      len               = -1;
+        char    *peer_str      = NULL;
+
+        if (op_ret) {
+                args->op_ret = op_ret;
+                args->op_errno = op_errno;
+
+                if (peerinfo)
+                        peer_str = peerinfo->hostname;
+                else
+                        peer_str = uuid_utoa (uuid);
+
+                if (op_errstr && strcmp (op_errstr, "")) {
+                        len = snprintf (err_str, sizeof(err_str) - 1,
+                                        "Error: %s", op_errstr);
+                        err_str[len] = '\0';
+                }
+
+                switch (op_code){
+                        case GLUSTERD_MGMT_V3_LOCK:
+                        {
+                                len = snprintf (op_err, sizeof(op_err) - 1,
+                                                "Locking failed "
+                                                "on %s. %s", peer_str, err_str);
+                                break;
+                        }
+                        case GLUSTERD_MGMT_V3_UNLOCK:
+                        {
+                                len = snprintf (op_err, sizeof(op_err) - 1,
+                                                "Unlocking failed "
+                                                "on %s. %s", peer_str, err_str);
+                                break;
+                        }
+                }
+                op_err[len] = '\0';
+
+                if (args->errstr) {
+                        len = snprintf (err_str, sizeof(err_str) - 1,
+                                        "%s\n%s", args->errstr,
+                                        op_err);
+                        GF_FREE (args->errstr);
+                        args->errstr = NULL;
+                } else
+                        len = snprintf (err_str, sizeof(err_str) - 1,
+                                        "%s", op_err);
+                err_str[len] = '\0';
+
+                gf_log ("", GF_LOG_ERROR, "%s", op_err);
+                args->errstr = gf_strdup (err_str);
+        }
+
+        return;
+}
+
+static void
+gd_collate_errors (struct syncargs *args, int op_ret, int op_errno,
+                   char *op_errstr, int op_code,
+                   glusterd_peerinfo_t *peerinfo, u_char *uuid)
+{
+        char     err_str[PATH_MAX] = "Please check log file for details.";
+        char     op_err[PATH_MAX] = "";
+        int      len               = -1;
+        char    *peer_str      = NULL;
+
+        if (op_ret) {
+                args->op_ret = op_ret;
+                args->op_errno = op_errno;
+
+                if (peerinfo)
+                        peer_str = peerinfo->hostname;
+                else
+                        peer_str = uuid_utoa (uuid);
+
+                if (op_errstr && strcmp (op_errstr, "")) {
+                        len = snprintf (err_str, sizeof(err_str) - 1,
+                                        "Error: %s", op_errstr);
+                        err_str[len] = '\0';
+                }
+
+                switch (op_code){
+                        case GLUSTERD_MGMT_CLUSTER_UNLOCK :
+                        {
+                                len = snprintf (op_err, sizeof(op_err) - 1,
+                                                "Unlocking failed on %s. %s",
+                                                peer_str, err_str);
+                                break;
+                        }
+                        case GLUSTERD_MGMT_STAGE_OP :
+                        {
+                                len = snprintf (op_err, sizeof(op_err) - 1,
+                                                "Staging failed on %s. %s",
+                                                peer_str, err_str);
+                                break;
+                        }
+                        case GLUSTERD_MGMT_COMMIT_OP :
+                        {
+                                len = snprintf (op_err, sizeof(op_err) - 1,
+                                                "Commit failed on %s. %s",
+                                                peer_str, err_str);
+                                break;
+                        }
+                }
+                op_err[len] = '\0';
+
+                if (args->errstr) {
+                        len = snprintf (err_str, sizeof(err_str) - 1,
+                                        "%s\n%s", args->errstr,
+                                        op_err);
+                        GF_FREE (args->errstr);
+                        args->errstr = NULL;
+                } else
+                        len = snprintf (err_str, sizeof(err_str) - 1,
+                                        "%s", op_err);
+                err_str[len] = '\0';
+
+                gf_log ("", GF_LOG_ERROR, "%s", op_err);
+                args->errstr = gf_strdup (err_str);
+        }
+
+        return;
+}
+
+void
+gd_syncargs_init (struct syncargs *args, dict_t *op_ctx)
+{
+        args->dict = op_ctx;
+        pthread_mutex_init (&args->lock_dict, NULL);
+}
+
+static void
+gd_stage_op_req_free (gd1_mgmt_stage_op_req *req)
+{
+        if (!req)
+                return;
+
+        GF_FREE (req->buf.buf_val);
+        GF_FREE (req);
+}
+
+static void
+gd_commit_op_req_free (gd1_mgmt_commit_op_req *req)
+{
+        if (!req)
+                return;
+
+        GF_FREE (req->buf.buf_val);
+        GF_FREE (req);
+}
+
+static void
+gd_brick_op_req_free (gd1_mgmt_brick_op_req *req)
+{
+        if (!req)
+                return;
+
+        if (strcmp (req->name, "") != 0)
+                GF_FREE (req->name);
+        GF_FREE (req->input.input_val);
+        GF_FREE (req);
+}
+
+int
+gd_syncop_submit_request (struct rpc_clnt *rpc, void *req, void *local,
+                          void *cookie, rpc_clnt_prog_t *prog, int procnum,
+                          fop_cbk_fn_t cbkfn, xdrproc_t xdrproc)
+{
+        int            ret      = -1;
+        struct iobuf  *iobuf    = NULL;
+        struct iobref *iobref   = NULL;
+        int            count    = 0;
+        struct iovec   iov      = {0, };
+        ssize_t        req_size = 0;
+        call_frame_t  *frame    = NULL;
+
+        GF_ASSERT (rpc);
+        if (!req)
+                goto out;
+
+        req_size = xdr_sizeof (xdrproc, req);
+        iobuf = iobuf_get2 (rpc->ctx->iobuf_pool, req_size);
+        if (!iobuf)
+                goto out;
+
+        iobref = iobref_new ();
+        if (!iobref)
+                goto out;
+
+        frame = create_frame (THIS, THIS->ctx->pool);
+        if (!frame)
+                goto out;
+
+        iobref_add (iobref, iobuf);
+
+        iov.iov_base = iobuf->ptr;
+        iov.iov_len  = iobuf_pagesize (iobuf);
+
+        /* Create the xdr payload */
+        ret = xdr_serialize_generic (iov, req, xdrproc);
+        if (ret == -1)
+                goto out;
+
+        iov.iov_len = ret;
+        count = 1;
+
+        frame->local = local;
+        frame->cookie = cookie;
+
+        /* Send the msg */
+        ret = rpc_clnt_submit (rpc, prog, procnum, cbkfn,
+                               &iov, count, NULL, 0, iobref,
+                               frame, NULL, 0, NULL, 0, NULL);
+
+        /* TODO: do we need to start ping also? */
+
+out:
+        iobref_unref (iobref);
+        iobuf_unref (iobuf);
+
+        return ret;
+}
+
+/* Defined in glusterd-rpc-ops.c */
+extern struct rpc_clnt_program gd_mgmt_prog;
+extern struct rpc_clnt_program gd_brick_prog;
+extern struct rpc_clnt_program gd_mgmt_v3_prog;
+
+int
+glusterd_syncop_aggr_rsp_dict (glusterd_op_t op, dict_t *aggr, dict_t *rsp)
+{
+        int ret = 0;
+
+        switch (op) {
+        case GD_OP_REPLACE_BRICK:
+                ret = glusterd_rb_use_rsp_dict (aggr, rsp);
+                if (ret)
+                        goto out;
+        break;
+
+        case GD_OP_SYNC_VOLUME:
+                ret = glusterd_sync_use_rsp_dict (aggr, rsp);
+                if (ret)
+                        goto out;
+        break;
+
+        case GD_OP_PROFILE_VOLUME:
+                ret = glusterd_profile_volume_use_rsp_dict (aggr, rsp);
+                if (ret)
+                        goto out;
+        break;
+
+        case GD_OP_GSYNC_CREATE:
+        break;
+
+        case GD_OP_GSYNC_SET:
+                ret = glusterd_gsync_use_rsp_dict (aggr, rsp, NULL);
+                if (ret)
+                        goto out;
+        break;
+
+        case GD_OP_STATUS_VOLUME:
+                ret = glusterd_volume_status_copy_to_op_ctx_dict (aggr, rsp);
+                if (ret)
+                        goto out;
+        break;
+
+        case GD_OP_REBALANCE:
+        case GD_OP_DEFRAG_BRICK_VOLUME:
+                ret = glusterd_volume_rebalance_use_rsp_dict (aggr, rsp);
+                if (ret)
+                        goto out;
+        break;
+
+        case GD_OP_HEAL_VOLUME:
+                ret = glusterd_volume_heal_use_rsp_dict (aggr, rsp);
+                if (ret)
+                        goto out;
+
+        break;
+
+        case GD_OP_QUOTA:
+        case GD_OP_CLEARLOCKS_VOLUME:
+                ret = glusterd_use_rsp_dict (aggr, rsp);
+                if (ret)
+                        goto out;
+
+        break;
+
+        case GD_OP_SYS_EXEC:
+                ret = glusterd_sys_exec_output_rsp_dict (aggr, rsp);
+                if (ret)
+                        goto out;
+        break;
+
+        case GD_OP_SNAP:
+                ret = glusterd_snap_use_rsp_dict (aggr, rsp);
+                if (ret)
+                        goto out;
+        break;
+
+        default:
+        break;
+        }
+out:
+        return ret;
+}
+
+int32_t
+_gd_syncop_mgmt_lock_cbk (struct rpc_req *req, struct iovec *iov,
+                          int count, void *myframe)
+{
+        int                         ret         = -1;
+        struct syncargs             *args       = NULL;
+        glusterd_peerinfo_t         *peerinfo   = NULL;
+        gd1_mgmt_cluster_lock_rsp   rsp         = {{0},};
+        call_frame_t                *frame      = NULL;
+        int                         op_ret      = -1;
+        int                         op_errno    = -1;
+
+        frame  = myframe;
+        args   = frame->local;
+        peerinfo = frame->cookie;
+        frame->local = NULL;
+        frame->cookie = NULL;
+
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_cluster_lock_rsp);
+        if (ret < 0)
+                goto out;
+
+        uuid_copy (args->uuid, rsp.uuid);
+
+        /* Set peer as locked, so we unlock only the locked peers */
+        if (rsp.op_ret == 0)
+                peerinfo->locked = _gf_true;
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+out:
+        gd_collate_errors (args, op_ret, op_errno, NULL,
+                           GLUSTERD_MGMT_CLUSTER_LOCK, peerinfo, rsp.uuid);
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+        return 0;
+}
+
+int32_t
+gd_syncop_mgmt_lock_cbk (struct rpc_req *req, struct iovec *iov, int count,
+                         void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        _gd_syncop_mgmt_lock_cbk);
+}
+
+int
+gd_syncop_mgmt_lock (glusterd_peerinfo_t *peerinfo, struct syncargs *args,
+                     uuid_t my_uuid, uuid_t recv_uuid)
+{
+        int                       ret = -1;
+        gd1_mgmt_cluster_lock_req req  = {{0},};
+        glusterd_conf_t           *conf = THIS->private;
+
+        uuid_copy (req.uuid, my_uuid);
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo,
+                                        &gd_mgmt_prog,
+                                        GLUSTERD_MGMT_CLUSTER_LOCK,
+                                        gd_syncop_mgmt_lock_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_cluster_lock_req);
+        synclock_lock (&conf->big_lock);
+        return ret;
+}
+
+int32_t
+gd_syncop_mgmt_v3_lock_cbk_fn (struct rpc_req *req, struct iovec *iov,
+                               int count, void *myframe)
+{
+        int                         ret         = -1;
+        struct syncargs             *args       = NULL;
+        glusterd_peerinfo_t         *peerinfo   = NULL;
+        gd1_mgmt_v3_lock_rsp         rsp        = {{0},};
+        call_frame_t                *frame      = NULL;
+        int                         op_ret      = -1;
+        int                         op_errno    = -1;
+
+        GF_ASSERT(req);
+        GF_ASSERT(iov);
+        GF_ASSERT(myframe);
+
+        frame  = myframe;
+        args   = frame->local;
+        peerinfo = frame->cookie;
+        frame->local = NULL;
+        frame->cookie = NULL;
+
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_lock_rsp);
+        if (ret < 0)
+                goto out;
+
+        uuid_copy (args->uuid, rsp.uuid);
+
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+out:
+        gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL,
+                                   GLUSTERD_MGMT_V3_LOCK,
+                                   peerinfo, rsp.uuid);
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+        return 0;
+}
+
+int32_t
+gd_syncop_mgmt_v3_lock_cbk (struct rpc_req *req, struct iovec *iov,
+                            int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        gd_syncop_mgmt_v3_lock_cbk_fn);
+}
+
+int
+gd_syncop_mgmt_v3_lock (glusterd_op_t op, dict_t *op_ctx,
+                        glusterd_peerinfo_t *peerinfo,
+                        struct syncargs *args, uuid_t my_uuid,
+                        uuid_t recv_uuid, uuid_t txn_id)
+{
+        int                      ret  = -1;
+        gd1_mgmt_v3_lock_req     req  = {{0},};
+        glusterd_conf_t         *conf = THIS->private;
+
+        GF_ASSERT(op_ctx);
+        GF_ASSERT(peerinfo);
+        GF_ASSERT(args);
+
+        ret = dict_allocate_and_serialize (op_ctx,
+                                           &req.dict.dict_val,
+                                           &req.dict.dict_len);
+        if (ret)
+                goto out;
+
+        uuid_copy (req.uuid, my_uuid);
+        uuid_copy (req.txn_id, txn_id);
+        req.op = op;
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo,
+                                        &gd_mgmt_v3_prog,
+                                        GLUSTERD_MGMT_V3_LOCK,
+                                        gd_syncop_mgmt_v3_lock_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_v3_lock_req);
+        synclock_lock (&conf->big_lock);
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+gd_syncop_mgmt_v3_unlock_cbk_fn (struct rpc_req *req, struct iovec *iov,
+                                 int count, void *myframe)
+{
+        int                         ret         = -1;
+        struct syncargs             *args       = NULL;
+        glusterd_peerinfo_t         *peerinfo   = NULL;
+        gd1_mgmt_v3_unlock_rsp       rsp        = {{0},};
+        call_frame_t                *frame      = NULL;
+        int                         op_ret      = -1;
+        int                         op_errno    = -1;
+
+        GF_ASSERT(req);
+        GF_ASSERT(iov);
+        GF_ASSERT(myframe);
+
+        frame  = myframe;
+        args   = frame->local;
+        peerinfo = frame->cookie;
+        frame->local = NULL;
+        frame->cookie = NULL;
+
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_v3_unlock_rsp);
+        if (ret < 0)
+                goto out;
+
+        uuid_copy (args->uuid, rsp.uuid);
+
+        /* Set peer as locked, so we unlock only the locked peers */
+        if (rsp.op_ret == 0)
+                peerinfo->locked = _gf_true;
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+out:
+        gd_mgmt_v3_collate_errors (args, op_ret, op_errno, NULL,
+                                   GLUSTERD_MGMT_V3_UNLOCK,
+                                   peerinfo, rsp.uuid);
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+        return 0;
+}
+
+int32_t
+gd_syncop_mgmt_v3_unlock_cbk (struct rpc_req *req, struct iovec *iov,
+                              int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        gd_syncop_mgmt_v3_unlock_cbk_fn);
+}
+
+int
+gd_syncop_mgmt_v3_unlock (dict_t *op_ctx, glusterd_peerinfo_t *peerinfo,
+                          struct syncargs *args, uuid_t my_uuid,
+                          uuid_t recv_uuid, uuid_t txn_id)
+{
+        int                          ret  = -1;
+        gd1_mgmt_v3_unlock_req       req  = {{0},};
+        glusterd_conf_t             *conf = THIS->private;
+
+        GF_ASSERT(op_ctx);
+        GF_ASSERT(peerinfo);
+        GF_ASSERT(args);
+
+        ret = dict_allocate_and_serialize (op_ctx,
+                                           &req.dict.dict_val,
+                                           &req.dict.dict_len);
+        if (ret)
+                goto out;
+
+        uuid_copy (req.uuid, my_uuid);
+        uuid_copy (req.txn_id, txn_id);
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo,
+                                        &gd_mgmt_v3_prog,
+                                        GLUSTERD_MGMT_V3_UNLOCK,
+                                        gd_syncop_mgmt_v3_unlock_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_v3_unlock_req);
+        synclock_lock (&conf->big_lock);
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+_gd_syncop_mgmt_unlock_cbk (struct rpc_req *req, struct iovec *iov,
+                            int count, void *myframe)
+{
+        int                         ret         = -1;
+        struct syncargs             *args       = NULL;
+        glusterd_peerinfo_t         *peerinfo   = NULL;
+        gd1_mgmt_cluster_unlock_rsp rsp         = {{0},};
+        call_frame_t                *frame      = NULL;
+        int                         op_ret      = -1;
+        int                         op_errno    = -1;
+
+        frame = myframe;
+        args  = frame->local;
+        peerinfo = frame->cookie;
+        frame->local = NULL;
+
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_cluster_unlock_rsp);
+        if (ret < 0)
+                goto out;
+
+        uuid_copy (args->uuid, rsp.uuid);
+
+        peerinfo->locked = _gf_false;
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+out:
+        gd_collate_errors (args, op_ret, op_errno, NULL,
+                           GLUSTERD_MGMT_CLUSTER_UNLOCK, peerinfo, rsp.uuid);
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+        return 0;
+}
+
+int32_t
+gd_syncop_mgmt_unlock_cbk (struct rpc_req *req, struct iovec *iov,
+                           int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        _gd_syncop_mgmt_unlock_cbk);
+}
+
+
+int
+gd_syncop_mgmt_unlock (glusterd_peerinfo_t *peerinfo, struct syncargs *args,
+                       uuid_t my_uuid, uuid_t recv_uuid)
+{
+        int                         ret     = -1;
+        gd1_mgmt_cluster_unlock_req req     = {{0},};
+        glusterd_conf_t             *conf   = THIS->private;
+
+        uuid_copy (req.uuid, my_uuid);
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (peerinfo->rpc, &req, args, peerinfo,
+                                        &gd_mgmt_prog,
+                                        GLUSTERD_MGMT_CLUSTER_UNLOCK,
+                                        gd_syncop_mgmt_unlock_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_cluster_lock_req);
+        synclock_lock (&conf->big_lock);
+        return ret;
+}
+
+int32_t
+_gd_syncop_stage_op_cbk (struct rpc_req *req, struct iovec *iov,
+                         int count, void *myframe)
+{
+        int                         ret         = -1;
+        gd1_mgmt_stage_op_rsp       rsp         = {{0},};
+        struct syncargs             *args       = NULL;
+        xlator_t                    *this       = NULL;
+        dict_t                      *rsp_dict   = NULL;
+        call_frame_t                *frame      = NULL;
+        glusterd_peerinfo_t         *peerinfo   = NULL;
+        int                         op_ret      = -1;
+        int                         op_errno    = -1;
+
+        this  = THIS;
+        frame = myframe;
+        args  = frame->local;
+        frame->local = NULL;
+
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_stage_op_rsp);
+        if (ret < 0)
+                goto out;
+
+        if (rsp.dict.dict_len) {
+                /* Unserialize the dictionary */
+                rsp_dict  = dict_new ();
+
+                ret = dict_unserialize (rsp.dict.dict_val,
+                                        rsp.dict.dict_len,
+                                        &rsp_dict);
+                if (ret < 0) {
+                        GF_FREE (rsp.dict.dict_val);
+                        goto out;
+                } else {
+                        rsp_dict->extra_stdfree = rsp.dict.dict_val;
+                }
+        }
+
+        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL, "Staging response "
+                        "for 'Volume %s' received from unknown "
+                        "peer: %s", gd_op_list[rsp.op],
+                        uuid_utoa (rsp.uuid));
+                goto out;
+        }
+
+        uuid_copy (args->uuid, rsp.uuid);
+        if (rsp.op == GD_OP_REPLACE_BRICK) {
+                pthread_mutex_lock (&args->lock_dict);
+                {
+                        ret = glusterd_syncop_aggr_rsp_dict (rsp.op, args->dict,
+                                                             rsp_dict);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR, "%s",
+                                        "Failed to aggregate response from "
+                                        " node/brick");
+                }
+                pthread_mutex_unlock (&args->lock_dict);
+        }
+
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+
+out:
+        gd_collate_errors (args, op_ret, op_errno, rsp.op_errstr,
+                           GLUSTERD_MGMT_STAGE_OP, peerinfo, rsp.uuid);
+
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+        return 0;
+}
+
+int32_t
+gd_syncop_stage_op_cbk (struct rpc_req *req, struct iovec *iov,
+                        int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        _gd_syncop_stage_op_cbk);
+}
+
+
+int
+gd_syncop_mgmt_stage_op (struct rpc_clnt *rpc, struct syncargs *args,
+                         uuid_t my_uuid, uuid_t recv_uuid, int op,
+                         dict_t *dict_out, dict_t *op_ctx)
+{
+        gd1_mgmt_stage_op_req *req  = NULL;
+        glusterd_conf_t       *conf = THIS->private;
+        int                   ret   = -1;
+
+        req = GF_CALLOC (1, sizeof (*req), gf_gld_mt_mop_stage_req_t);
+        if (!req)
+                goto out;
+
+        uuid_copy (req->uuid, my_uuid);
+        req->op = op;
+
+        ret = dict_allocate_and_serialize (dict_out,
+                                           &req->buf.buf_val, &req->buf.buf_len);
+        if (ret)
+                goto out;
+
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (rpc, req, args, NULL, &gd_mgmt_prog,
+                                        GLUSTERD_MGMT_STAGE_OP,
+                                        gd_syncop_stage_op_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_stage_op_req);
+        synclock_lock (&conf->big_lock);
+out:
+        gd_stage_op_req_free (req);
+        return ret;
+
+}
+
+int32_t
+_gd_syncop_brick_op_cbk (struct rpc_req *req, struct iovec *iov,
+                        int count, void *myframe)
+{
+        struct syncargs        *args  = NULL;
+        gd1_mgmt_brick_op_rsp  rsp   = {0,};
+        int                    ret   = -1;
+        call_frame_t           *frame = NULL;
+
+        frame = myframe;
+        args = frame->local;
+        frame->local = NULL;
+
+        /* initialize */
+        args->op_ret   = -1;
+        args->op_errno = EINVAL;
+
+        if (-1 == req->rpc_status) {
+                args->op_errno = ENOTCONN;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_brick_op_rsp);
+        if (ret < 0)
+                goto out;
+
+        if (rsp.output.output_len) {
+                args->dict  = dict_new ();
+                if (!args->dict) {
+                        ret = -1;
+                        args->op_errno = ENOMEM;
+                        goto out;
+                }
+
+                ret = dict_unserialize (rsp.output.output_val,
+                                        rsp.output.output_len,
+                                        &args->dict);
+                if (ret < 0)
+                        goto out;
+        }
+
+        args->op_ret = rsp.op_ret;
+        args->op_errno = rsp.op_errno;
+        args->errstr = gf_strdup (rsp.op_errstr);
+
+out:
+        if ((rsp.op_errstr) && (strcmp (rsp.op_errstr, "") != 0))
+                free (rsp.op_errstr);
+        free (rsp.output.output_val);
+
+        STACK_DESTROY (frame->root);
+        __wake (args);
+
+        return 0;
+}
+
+int32_t
+gd_syncop_brick_op_cbk (struct rpc_req *req, struct iovec *iov,
+                        int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        _gd_syncop_brick_op_cbk);
+}
+
+int
+gd_syncop_mgmt_brick_op (struct rpc_clnt *rpc, glusterd_pending_node_t *pnode,
+                         int op, dict_t *dict_out, dict_t *op_ctx,
+                         char **errstr)
+{
+        struct syncargs        args = {0, };
+        gd1_mgmt_brick_op_req  *req  = NULL;
+        int                    ret  = 0;
+        xlator_t               *this = NULL;
+
+        this = THIS;
+        args.op_ret = -1;
+        args.op_errno = ENOTCONN;
+
+        if ((pnode->type == GD_NODE_NFS) ||
+            ((pnode->type == GD_NODE_SHD) &&
+            (op == GD_OP_STATUS_VOLUME))) {
+                ret = glusterd_node_op_build_payload
+                        (op, &req, dict_out);
+
+        } else {
+                ret = glusterd_brick_op_build_payload
+                        (op, pnode->node, &req, dict_out);
+
+        }
+
+        if (ret)
+                goto out;
+
+        GD_SYNCOP (rpc, (&args), NULL, gd_syncop_brick_op_cbk, req,
+                   &gd_brick_prog, req->op, xdr_gd1_mgmt_brick_op_req);
+
+        if (args.errstr) {
+                if ((strlen(args.errstr) > 0) && errstr)
+                        *errstr = args.errstr;
+                else
+                        GF_FREE (args.errstr);
+        }
+
+        if (GD_OP_STATUS_VOLUME == op) {
+                ret = dict_set_int32 (args.dict, "index", pnode->index);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Error setting index on brick status"
+                                " rsp dict");
+                        args.op_ret = -1;
+                        goto out;
+                }
+        }
+        if (args.op_ret == 0)
+                glusterd_handle_node_rsp (dict_out, pnode->node, op,
+                                          args.dict, op_ctx, errstr,
+                                          pnode->type);
+
+out:
+        errno = args.op_errno;
+        if (args.dict)
+                dict_unref (args.dict);
+        gd_brick_op_req_free (req);
+        return args.op_ret;
+
+}
+
+int32_t
+_gd_syncop_commit_op_cbk (struct rpc_req *req, struct iovec *iov,
+                          int count, void *myframe)
+{
+        int                          ret           = -1;
+        gd1_mgmt_commit_op_rsp       rsp           = {{0},};
+        struct syncargs             *args          = NULL;
+        xlator_t                    *this          = NULL;
+        dict_t                      *rsp_dict      = NULL;
+        call_frame_t                *frame         = NULL;
+        glusterd_peerinfo_t         *peerinfo      = NULL;
+        int                          op_ret        = -1;
+        int                          op_errno      = -1;
+
+        this  = THIS;
+        frame = myframe;
+        args  = frame->local;
+        frame->local = NULL;
+
+        if (-1 == req->rpc_status) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        ret = xdr_to_generic (*iov, &rsp,
+                              (xdrproc_t)xdr_gd1_mgmt_commit_op_rsp);
+        if (ret < 0) {
+                goto out;
+        }
+
+        if (rsp.dict.dict_len) {
+                /* Unserialize the dictionary */
+                rsp_dict  = dict_new ();
+
+                ret = dict_unserialize (rsp.dict.dict_val,
+                                        rsp.dict.dict_len,
+                                        &rsp_dict);
+                if (ret < 0) {
+                        GF_FREE (rsp.dict.dict_val);
+                        goto out;
+                } else {
+                        rsp_dict->extra_stdfree = rsp.dict.dict_val;
+                }
+        }
+
+        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL, "Commit response "
+                        "for 'Volume %s' received from unknown "
+                        "peer: %s", gd_op_list[rsp.op],
+                        uuid_utoa (rsp.uuid));
+                goto out;
+        }
+
+        uuid_copy (args->uuid, rsp.uuid);
+        pthread_mutex_lock (&args->lock_dict);
+        {
+                ret = glusterd_syncop_aggr_rsp_dict (rsp.op, args->dict,
+                                                     rsp_dict);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR, "%s",
+                                "Failed to aggregate response from "
+                                " node/brick");
+        }
+        pthread_mutex_unlock (&args->lock_dict);
+
+        op_ret = rsp.op_ret;
+        op_errno = rsp.op_errno;
+
+out:
+        gd_collate_errors (args, op_ret, op_errno, rsp.op_errstr,
+                           GLUSTERD_MGMT_COMMIT_OP, peerinfo, rsp.uuid);
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        STACK_DESTROY (frame->root);
+        synctask_barrier_wake(args);
+
+        return 0;
+}
+
+int32_t
+gd_syncop_commit_op_cbk (struct rpc_req *req, struct iovec *iov,
+                         int count, void *myframe)
+{
+        return glusterd_big_locked_cbk (req, iov, count, myframe,
+                                        _gd_syncop_commit_op_cbk);
+}
+
+
+int
+gd_syncop_mgmt_commit_op (struct rpc_clnt *rpc, struct syncargs *args,
+                          uuid_t my_uuid, uuid_t recv_uuid,
+                          int op, dict_t *dict_out, dict_t *op_ctx)
+{
+        glusterd_conf_t        *conf = THIS->private;
+        gd1_mgmt_commit_op_req *req  = NULL;
+        int                    ret  = -1;
+
+        req = GF_CALLOC (1, sizeof (*req), gf_gld_mt_mop_commit_req_t);
+        if (!req)
+                goto out;
+
+        uuid_copy (req->uuid, my_uuid);
+        req->op = op;
+
+        ret = dict_allocate_and_serialize (dict_out,
+                                           &req->buf.buf_val, &req->buf.buf_len);
+        if (ret)
+                goto out;
+
+        synclock_unlock (&conf->big_lock);
+        ret = gd_syncop_submit_request (rpc, req, args, NULL, &gd_mgmt_prog,
+                                        GLUSTERD_MGMT_COMMIT_OP ,
+                                        gd_syncop_commit_op_cbk,
+                                        (xdrproc_t) xdr_gd1_mgmt_commit_op_req);
+        synclock_lock (&conf->big_lock);
+out:
+        gd_commit_op_req_free (req);
+        return ret;
+}
+
+
+int
+gd_build_peers_list (struct list_head *peers, struct list_head *xact_peers,
+                     glusterd_op_t op)
+{
+        glusterd_peerinfo_t *peerinfo = NULL;
+        int                 npeers      = 0;
+
+        list_for_each_entry (peerinfo, peers, uuid_list) {
+                if (!peerinfo->connected)
+                        continue;
+                if (op != GD_OP_SYNC_VOLUME &&
+                    peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED)
+                        continue;
+
+                list_add_tail (&peerinfo->op_peers_list, xact_peers);
+                npeers++;
+        }
+        return npeers;
+}
+
+int
+gd_lock_op_phase (glusterd_conf_t  *conf, glusterd_op_t op, dict_t *op_ctx,
+                  char **op_errstr, int npeers, uuid_t txn_id)
+{
+        int                 ret         = -1;
+        int                 peer_cnt    = 0;
+        uuid_t              peer_uuid   = {0};
+        xlator_t            *this       = NULL;
+        glusterd_peerinfo_t *peerinfo   = NULL;
+        struct syncargs     args        = {0};
+        struct list_head   *peers       = NULL;
+
+        peers = &conf->xaction_peers;
+
+        if (!npeers) {
+                ret = 0;
+                goto out;
+        }
+
+        this = THIS;
+        synctask_barrier_init((&args));
+        peer_cnt = 0;
+        list_for_each_entry (peerinfo, peers, op_peers_list) {
+                if (conf->op_version < 3) {
+                        /* Reset lock status */
+                        peerinfo->locked = _gf_false;
+                        gd_syncop_mgmt_lock (peerinfo, &args,
+                                             MY_UUID, peer_uuid);
+                } else
+                        gd_syncop_mgmt_v3_lock (op, op_ctx, peerinfo, &args,
+                                                MY_UUID, peer_uuid, txn_id);
+                peer_cnt++;
+        }
+        gd_synctask_barrier_wait((&args), peer_cnt);
+
+        if (args.op_ret) {
+                if (args.errstr)
+                         *op_errstr = gf_strdup (args.errstr);
+                else {
+                        ret = gf_asprintf (op_errstr, "Another transaction could be "
+                                           "in progress. Please try again after "
+                                           "sometime.");
+                        if (ret == -1)
+                                *op_errstr = NULL;
+
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to acquire lock");
+
+                }
+        }
+
+        ret = args.op_ret;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Sent lock op req for 'Volume %s' "
+                "to %d peers. Returning %d", gd_op_list[op], peer_cnt, ret);
+out:
+        return ret;
+}
+
+int
+gd_stage_op_phase (struct list_head *peers, glusterd_op_t op, dict_t *op_ctx,
+                   dict_t *req_dict, char **op_errstr, int npeers)
+{
+        int                 ret             = -1;
+        int                 peer_cnt           = 0;
+        dict_t              *rsp_dict       = NULL;
+        char                *hostname       = NULL;
+        xlator_t            *this           = NULL;
+        glusterd_peerinfo_t *peerinfo       = NULL;
+        uuid_t              tmp_uuid        = {0};
+        char                *errstr         = NULL;
+        struct syncargs     args            = {0};
+
+        this = THIS;
+        rsp_dict = dict_new ();
+        if (!rsp_dict)
+                goto out;
+
+        ret = glusterd_op_stage_validate (op, req_dict, op_errstr, rsp_dict);
+        if (ret) {
+                hostname = "localhost";
+                goto stage_done;
+        }
+
+        if ((op == GD_OP_REPLACE_BRICK)) {
+                ret = glusterd_syncop_aggr_rsp_dict (op, op_ctx, rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s",
+                                "Failed to aggregate response from node/brick");
+                        goto out;
+                }
+        }
+        dict_unref (rsp_dict);
+        rsp_dict = NULL;
+
+stage_done:
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, LOGSTR_STAGE_FAIL,
+                        gd_op_list[op], hostname, (*op_errstr) ? ":" : " ",
+                        (*op_errstr) ? *op_errstr : " ");
+                if (*op_errstr == NULL)
+                        gf_asprintf (op_errstr, OPERRSTR_STAGE_FAIL, hostname);
+                goto out;
+        }
+
+        if (!npeers) {
+                ret = 0;
+                goto out;
+        }
+
+        gd_syncargs_init (&args, op_ctx);
+        synctask_barrier_init((&args));
+        peer_cnt = 0;
+        list_for_each_entry (peerinfo, peers, op_peers_list) {
+                ret = gd_syncop_mgmt_stage_op (peerinfo->rpc, &args,
+                                               MY_UUID, tmp_uuid,
+                                               op, req_dict, op_ctx);
+                peer_cnt++;
+        }
+        gd_synctask_barrier_wait((&args), peer_cnt);
+
+        if (args.errstr)
+                 *op_errstr = gf_strdup (args.errstr);
+        else if (dict_get_str (op_ctx, "errstr", &errstr) == 0)
+                *op_errstr = gf_strdup (errstr);
+
+        ret = args.op_ret;
+
+        gf_log (this->name, GF_LOG_DEBUG, "Sent stage op req for 'Volume %s' "
+                "to %d peers", gd_op_list[op], peer_cnt);
+out:
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+        return ret;
+}
+
+int
+gd_commit_op_phase (struct list_head *peers, glusterd_op_t op, dict_t *op_ctx,
+                    dict_t *req_dict, char **op_errstr, int npeers)
+{
+        dict_t              *rsp_dict       = NULL;
+        int                 peer_cnt           = -1;
+        int                 ret             = -1;
+        char                *hostname       = NULL;
+        glusterd_peerinfo_t *peerinfo       = NULL;
+        xlator_t            *this           = NULL;
+        uuid_t              tmp_uuid        = {0};
+        char                *errstr         = NULL;
+        struct syncargs     args            = {0};
+
+        this = THIS;
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = glusterd_op_commit_perform (op, req_dict, op_errstr, rsp_dict);
+        if (ret) {
+                hostname = "localhost";
+                goto commit_done;
+        }
+        if (op != GD_OP_SYNC_VOLUME) {
+                ret =  glusterd_syncop_aggr_rsp_dict (op, op_ctx, rsp_dict);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s",
+                                "Failed to aggregate response "
+                                "from node/brick");
+                        goto out;
+                }
+        }
+        dict_unref (rsp_dict);
+        rsp_dict = NULL;
+
+commit_done:
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, LOGSTR_COMMIT_FAIL,
+                        gd_op_list[op], hostname, (*op_errstr) ? ":" : " ",
+                        (*op_errstr) ? *op_errstr : " ");
+                if (*op_errstr == NULL)
+                        gf_asprintf (op_errstr, OPERRSTR_COMMIT_FAIL,
+                                     hostname);
+                goto out;
+        }
+
+        if (!npeers) {
+                ret = 0;
+                goto out;
+        }
+
+        gd_syncargs_init (&args, op_ctx);
+        synctask_barrier_init((&args));
+        peer_cnt = 0;
+        list_for_each_entry (peerinfo, peers, op_peers_list) {
+                ret = gd_syncop_mgmt_commit_op (peerinfo->rpc, &args,
+                                                MY_UUID, tmp_uuid,
+                                                op, req_dict, op_ctx);
+                peer_cnt++;
+        }
+        gd_synctask_barrier_wait((&args), peer_cnt);
+        ret = args.op_ret;
+        if (args.errstr)
+                 *op_errstr = gf_strdup (args.errstr);
+        else if (dict_get_str (op_ctx, "errstr", &errstr) == 0)
+                *op_errstr = gf_strdup (errstr);
+
+        gf_log (this->name, GF_LOG_DEBUG, "Sent commit op req for 'Volume %s' "
+                "to %d peers", gd_op_list[op], peer_cnt);
+out:
+        if (!ret)
+                glusterd_op_modify_op_ctx (op, op_ctx);
+
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        GF_FREE (args.errstr);
+        args.errstr = NULL;
+
+        return ret;
+}
+
+int
+gd_unlock_op_phase (glusterd_conf_t  *conf, glusterd_op_t op, int op_ret,
+                    rpcsvc_request_t *req, dict_t *op_ctx, char *op_errstr,
+                    int npeers, char *volname, gf_boolean_t is_acquired,
+                    uuid_t txn_id)
+{
+        glusterd_peerinfo_t *peerinfo   = NULL;
+        glusterd_peerinfo_t *tmp        = NULL;
+        uuid_t              tmp_uuid    = {0};
+        int                 peer_cnt       = 0;
+        int                 ret         = -1;
+        xlator_t            *this       = NULL;
+        struct syncargs     args        = {0};
+        struct list_head   *peers       = NULL;
+
+        peers = &conf->xaction_peers;
+
+        if (!npeers) {
+                ret = 0;
+                goto out;
+        }
+
+        /* If the lock has not been held during this
+         * transaction, do not send unlock requests */
+        if (!is_acquired)
+                goto out;
+
+        this = THIS;
+        synctask_barrier_init((&args));
+        peer_cnt = 0;
+        list_for_each_entry_safe (peerinfo, tmp, peers, op_peers_list) {
+                if (conf->op_version < 3) {
+                        /* Only unlock peers that were locked */
+                        if (peerinfo->locked)
+                                gd_syncop_mgmt_unlock (peerinfo, &args,
+                                                       MY_UUID, tmp_uuid);
+                } else
+                        gd_syncop_mgmt_v3_unlock (op_ctx, peerinfo,
+                                                  &args, MY_UUID,
+                                                  tmp_uuid, txn_id);
+                peer_cnt++;
+                list_del_init (&peerinfo->op_peers_list);
+        }
+        gd_synctask_barrier_wait((&args), peer_cnt);
+        ret = args.op_ret;
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to unlock "
+                        "on some peer(s)");
+        }
+
+out:
+        glusterd_op_send_cli_response (op, op_ret, 0, req, op_ctx, op_errstr);
+        glusterd_op_clear_op (op);
+        if (is_acquired) {
+                /* Based on the op-version, we release *
+                 * the cluster or mgmt_v3 lock */
+                if (conf->op_version < 3)
+                        glusterd_unlock (MY_UUID);
+                else {
+                        ret = glusterd_mgmt_v3_unlock (volname, MY_UUID,
+                                                       "vol");
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to release lock for %s",
+                                        volname);
+                }
+        }
+
+        return 0;
+}
+
+int
+gd_get_brick_count (struct list_head *bricks)
+{
+        glusterd_pending_node_t *pending_node = NULL;
+        int                     npeers        = 0;
+        list_for_each_entry (pending_node, bricks, list) {
+                npeers++;
+        }
+        return npeers;
+}
+
+int
+gd_brick_op_phase (glusterd_op_t op, dict_t *op_ctx, dict_t *req_dict,
+                   char **op_errstr)
+{
+        glusterd_pending_node_t *pending_node = NULL;
+        struct list_head        selected = {0,};
+        xlator_t                *this = NULL;
+        int                     brick_count = 0;
+        int                     ret = -1;
+        rpc_clnt_t              *rpc = NULL;
+        dict_t                  *rsp_dict = NULL;
+        glusterd_conf_t         *conf = NULL;
+
+        this = THIS;
+        conf = this->private;
+        rsp_dict = dict_new ();
+        if (!rsp_dict) {
+                ret = -1;
+                goto out;
+        }
+
+        INIT_LIST_HEAD (&selected);
+        ret = glusterd_op_bricks_select (op, req_dict, op_errstr, &selected, rsp_dict);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s",
+                       (*op_errstr)? *op_errstr: "Brick op failed. Check "
+                       "glusterd log file for more details.");
+                goto out;
+        }
+
+        if (op == GD_OP_HEAL_VOLUME) {
+                ret = glusterd_syncop_aggr_rsp_dict (op, op_ctx, rsp_dict);
+                if (ret)
+                        goto out;
+        }
+        dict_unref (rsp_dict);
+        rsp_dict = NULL;
+
+        brick_count = 0;
+        list_for_each_entry (pending_node, &selected, list) {
+                rpc = glusterd_pending_node_get_rpc (pending_node);
+                if (!rpc) {
+                        if (pending_node->type == GD_NODE_REBALANCE) {
+                                ret = 0;
+                                glusterd_defrag_volume_node_rsp (req_dict,
+                                                                 NULL, op_ctx);
+                                goto out;
+                        }
+
+                        ret = -1;
+                        gf_log (this->name, GF_LOG_ERROR, "Brick Op failed "
+                                "due to rpc failure.");
+                        goto out;
+                }
+                ret = gd_syncop_mgmt_brick_op (rpc, pending_node, op, req_dict,
+                                               op_ctx, op_errstr);
+                if (ret)
+                        goto out;
+
+                brick_count++;
+        }
+
+        ret = 0;
+out:
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+        gf_log (this->name, GF_LOG_DEBUG, "Sent op req to %d bricks",
+                brick_count);
+        return ret;
+}
+
+void
+gd_sync_task_begin (dict_t *op_ctx, rpcsvc_request_t * req)
+{
+        int                         ret              = -1;
+        int                         npeers           = 0;
+        dict_t                      *req_dict        = NULL;
+        glusterd_conf_t             *conf            = NULL;
+        glusterd_op_t               op               = 0;
+        int32_t                     tmp_op           = 0;
+        char                        *op_errstr       = NULL;
+        char                        *tmp             = NULL;
+        char                        *volname         = NULL;
+        xlator_t                    *this            = NULL;
+        gf_boolean_t                is_acquired      = _gf_false;
+        uuid_t                      *txn_id          = NULL;
+        uuid_t                      *originator_uuid = NULL;
+        glusterd_op_info_t          txn_opinfo;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        /* Generate a transaction-id for this operation and
+         * save it in the dict */
+        txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t);
+        if (!txn_id) {
+                ret = -1;
+                goto out;
+        }
+
+        uuid_generate (*txn_id);
+
+        ret = dict_set_bin (op_ctx, "transaction_id",
+                            txn_id, sizeof(*txn_id));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set transaction id.");
+                goto out;
+        } else
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Transaction_id = %s", uuid_utoa (*txn_id));
+
+        /* Save the MY_UUID as the originator_uuid */
+        originator_uuid = GF_CALLOC (1, sizeof(uuid_t),
+                                     gf_common_mt_uuid_t);
+        if (!originator_uuid) {
+                ret = -1;
+                goto out;
+        }
+
+        uuid_copy (*originator_uuid, MY_UUID);
+        ret = dict_set_bin (op_ctx, "originator_uuid",
+                            originator_uuid, sizeof (uuid_t));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to set originator_uuid.");
+                goto out;
+        }
+
+        ret = dict_get_int32 (op_ctx, GD_SYNC_OPCODE_KEY, &tmp_op);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get volume "
+                        "operation");
+                goto out;
+        }
+
+        op = tmp_op;
+
+        /* Based on the op_version, acquire a cluster or mgmt_v3 lock */
+        if (conf->op_version < 3) {
+                ret = glusterd_lock (MY_UUID);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to acquire lock");
+                        gf_asprintf (&op_errstr,
+                                     "Another transaction is in progress. "
+                                     "Please try again after sometime.");
+                        goto out;
+                }
+        } else {
+
+                /* If no volname is given as a part of the command, locks will
+                 * not be held */
+                ret = dict_get_str (op_ctx, "volname", &tmp);
+                if (ret) {
+                        gf_log ("", GF_LOG_DEBUG, "Failed to get volume "
+                                "name");
+                        goto local_locking_done;
+                } else {
+                        /* Use a copy of volname, as cli response will be
+                         * sent before the unlock, and the volname in the
+                         * dict, might be removed */
+                        volname = gf_strdup (tmp);
+                        if (!volname)
+                                goto out;
+                }
+
+                ret = glusterd_mgmt_v3_lock (volname, MY_UUID, "vol");
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Unable to acquire lock for %s", volname);
+                        gf_asprintf (&op_errstr,
+                                     "Another transaction is in progress "
+                                     "for %s. Please try again after sometime.",
+                                     volname);
+                        goto out;
+                }
+        }
+
+        is_acquired = _gf_true;
+
+local_locking_done:
+
+        /* Save opinfo for this transaction with the transaction id */
+        glusterd_txn_opinfo_init (&txn_opinfo, NULL, &op, NULL, NULL);
+        ret = glusterd_set_txn_opinfo (txn_id, &txn_opinfo);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to set transaction's opinfo");
+
+        opinfo = txn_opinfo;
+
+        INIT_LIST_HEAD (&conf->xaction_peers);
+
+        /* Make 'volume status tasks' command a local operation.
+         * This is accomplished by setting npeers to 0.
+         */
+        if (!glusterd_is_status_tasks_op (op, op_ctx))
+                npeers = gd_build_peers_list  (&conf->peers,
+                                               &conf->xaction_peers, op);
+
+        /* If no volname is given as a part of the command, locks will
+         * not be held */
+        if (volname) {
+                ret = gd_lock_op_phase (conf, op, op_ctx, &op_errstr, npeers, *txn_id);
+                if (ret)
+                        goto out;
+        }
+
+        ret = glusterd_op_build_payload (&req_dict, &op_errstr, op_ctx);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, LOGSTR_BUILD_PAYLOAD,
+                        gd_op_list[op]);
+                if (op_errstr == NULL)
+                        gf_asprintf (&op_errstr, OPERRSTR_BUILD_PAYLOAD);
+                goto out;
+        }
+
+        ret = gd_stage_op_phase (&conf->xaction_peers, op, op_ctx, req_dict,
+                                 &op_errstr, npeers);
+        if (ret)
+                goto out;
+
+        ret = gd_brick_op_phase (op, op_ctx, req_dict, &op_errstr);
+        if (ret)
+                goto out;
+
+        ret = gd_commit_op_phase (&conf->xaction_peers, op, op_ctx, req_dict,
+                                  &op_errstr, npeers);
+        if (ret)
+                goto out;
+
+        ret = 0;
+out:
+        (void) gd_unlock_op_phase (conf, op, ret, req, op_ctx, op_errstr,
+                                   npeers, volname, is_acquired, *txn_id);
+
+        /* Clearing the transaction opinfo */
+        ret = glusterd_clear_txn_opinfo (txn_id);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to clear transaction's opinfo");
+
+        if (volname)
+                GF_FREE (volname);
+
+        if (req_dict)
+                dict_unref (req_dict);
+
+        if (op_errstr) {
+                GF_FREE (op_errstr);
+                op_errstr = NULL;
+        }
+
+        return;
+}
+
+int32_t
+glusterd_op_begin_synctask (rpcsvc_request_t *req, glusterd_op_t op,
+                            void *dict)
+{
+        int              ret = 0;
+
+        ret = dict_set_int32 (dict, GD_SYNC_OPCODE_KEY, op);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "dict set failed for setting operations");
+                goto out;
+        }
+
+        gd_sync_task_begin (dict, req);
+        ret = 0;
+out:
+
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-syncop.h b/xlators/mgmt/glusterd/src/glusterd-syncop.h
new file mode 100644
index 000000000..e83ea2f4c
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-syncop.h
@@ -0,0 +1,71 @@
+/*
+   Copyright (c) 2012-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef __RPC_SYNCOP_H
+#define __RPC_SYNCOP_H
+
+#include "syncop.h"
+#include "glusterd-sm.h"
+#include "glusterd.h"
+
+#define GD_SYNC_OPCODE_KEY "sync-mgmt-operation"
+
+/* gd_syncop_* */
+#define GD_SYNCOP(rpc, stb, cookie, cbk, req, prog, procnum, xdrproc) do {     \
+                int ret = 0;                                                   \
+                struct  synctask        *task = NULL;                          \
+                glusterd_conf_t         *conf= THIS->private;                  \
+                                                                               \
+                task = synctask_get ();                                        \
+                stb->task = task;                                              \
+                                                                               \
+                /*This is to ensure that the brick_op_cbk is able to           \
+                 * take the big lock*/                                         \
+                synclock_unlock (&conf->big_lock);                             \
+                ret = gd_syncop_submit_request (rpc, req, stb, cookie,         \
+                                                prog, procnum, cbk,            \
+                                                (xdrproc_t)xdrproc);           \
+                if (!ret)                                                      \
+                        synctask_yield (stb->task);                            \
+                synclock_lock (&conf->big_lock);                               \
+        } while (0)
+
+
+int gd_syncop_submit_request (struct rpc_clnt *rpc, void *req, void *local,
+                              void *cookie, rpc_clnt_prog_t *prog, int procnum,
+                              fop_cbk_fn_t cbkfn, xdrproc_t xdrproc);
+
+
+int gd_syncop_mgmt_lock (glusterd_peerinfo_t *peerinfo, struct syncargs *arg,
+                         uuid_t my_uuid, uuid_t recv_uuid);
+int gd_syncop_mgmt_unlock (glusterd_peerinfo_t *peerinfo, struct syncargs *arg,
+                           uuid_t my_uuid, uuid_t recv_uuid);
+int gd_syncop_mgmt_stage_op (struct rpc_clnt *rpc, struct syncargs *arg,
+                             uuid_t my_uuid, uuid_t recv_uuid, int op,
+                             dict_t *dict_out, dict_t *op_ctx);
+int gd_syncop_mgmt_commit_op (struct rpc_clnt *rpc, struct syncargs *arg,
+                              uuid_t my_uuid, uuid_t recv_uuid, int op,
+                              dict_t *dict_out, dict_t *op_ctx);
+
+void
+gd_synctask_barrier_wait (struct syncargs *args, int count);
+
+int
+gd_build_peers_list (struct list_head *peers, struct list_head *xact_peers,
+                     glusterd_op_t op);
+int
+gd_brick_op_phase (glusterd_op_t op, dict_t *op_ctx, dict_t *req_dict,
+                   char **op_errstr);
+
+int
+glusterd_syncop_aggr_rsp_dict (glusterd_op_t op, dict_t *aggr, dict_t *rsp);
+
+void
+gd_syncargs_init (struct syncargs *args, dict_t *op_ctx);
+#endif /* __RPC_SYNCOP_H */
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index e3592891b..e8ae05851 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -1,29 +1,18 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
 #endif
 #include <inttypes.h>
 
-
 #include "globals.h"
 #include "glusterfs.h"
 #include "compat.h"
@@ -33,15 +22,23 @@
 #include "timer.h"
 #include "defaults.h"
 #include "compat.h"
+#include "syncop.h"
+#include "run.h"
 #include "compat-errno.h"
 #include "statedump.h"
+#include "syscall.h"
 #include "glusterd-mem-types.h"
 #include "glusterd.h"
+#include "glusterd-op-sm.h"
 #include "glusterd-sm.h"
 #include "glusterd-utils.h"
 #include "glusterd-store.h"
 #include "glusterd-volgen.h"
+#include "glusterd-pmap.h"
+#include "glusterfs-acl.h"
+#include "glusterd-locks.h"
 
+#include "xdr-generic.h"
 #include <sys/resource.h>
 #include <inttypes.h>
 #include <signal.h>
@@ -49,14 +46,53 @@
 #include <net/if.h>
 #include <sys/ioctl.h>
 #include <sys/socket.h>
+#include <rpc/pmap_clnt.h>
+#include <unistd.h>
+#include <fnmatch.h>
+#include <sys/statvfs.h>
+#include <ifaddrs.h>
+#ifdef HAVE_BD_XLATOR
+#include <lvm2app.h>
+#endif
+
+
+#ifdef GF_LINUX_HOST_OS
+#include <mntent.h>
+#endif
 
 #ifdef GF_SOLARIS_HOST_OS
 #include <sys/sockio.h>
 #endif
 
+#define NFS_PROGRAM         100003
+#define NFSV3_VERSION       3
+
+#define MOUNT_PROGRAM       100005
+#define MOUNTV3_VERSION     3
+#define MOUNTV1_VERSION     1
+
+#define NLM_PROGRAM         100021
+#define NLMV4_VERSION       4
+#define NLMV1_VERSION       1
+
+#define CEILING_POS(X) (((X)-(int)(X)) > 0 ? (int)((X)+1) : (int)(X))
+
 static glusterd_lock_t lock;
 
-static int32_t
+char*
+gd_peer_uuid_str (glusterd_peerinfo_t *peerinfo)
+{
+        if ((peerinfo == NULL) || uuid_is_null (peerinfo->uuid))
+                return NULL;
+
+        if (peerinfo->uuid_str[0] == '\0')
+                uuid_utoa_r (peerinfo->uuid, peerinfo->uuid_str);
+
+        return peerinfo->uuid_str;
+}
+
+
+int32_t
 glusterd_get_lock_owner (uuid_t *uuid)
 {
         uuid_copy (*uuid, lock.owner) ;
@@ -79,117 +115,18 @@ glusterd_unset_lock_owner (uuid_t owner)
         return 0;
 }
 
-int32_t
-glusterd_is_local_addr (char *hostname)
+gf_boolean_t
+glusterd_is_fuse_available ()
 {
-        int32_t         ret = -1;
-        struct          addrinfo *result = NULL;
-        struct          addrinfo *res = NULL;
-        int32_t         found = 0;
-        struct          ifconf buf = {0,};
-        char            nodename[256] = {0,};
-
-        if ((!strcmp (hostname, "localhost")) ||
-             (!strcmp (hostname, "127.0.0.1"))) {
-                found = 1;
-                goto out;
-        }
-
-        ret = gethostname (nodename, 256);
-        if (ret)
-                goto out;
-
-        if ((!strcmp (nodename, hostname))) {
-                found = 1;
-                goto out;
-        }
-
-        ret = getaddrinfo (hostname, NULL, NULL, &result);
-
-        if (ret != 0) {
-                gf_log ("", GF_LOG_ERROR, "error in getaddrinfo: %s\n",
-                        gai_strerror(ret));
-                goto out;
-        }
 
-        for (res = result; res != NULL; res = res->ai_next) {
-                char hname[1024] = "";
-
-                ret = getnameinfo (res->ai_addr, res->ai_addrlen, hname,
-                                   NI_MAXHOST, NULL, 0, NI_NUMERICHOST);
-                if (ret)
-                        goto out;
-
-                if (!strncasecmp (hname, "127", 3)) {
-                        ret = 0;
-                        gf_log ("", GF_LOG_NORMAL, "local addr found");
-                        found = 1;
-                        break;
-                }
-        }
-
-        if (!found) {
-                int sd = -1;
-                struct ifreq  *ifr = NULL;
-                int32_t       size = 0;
-                int32_t       num_req = 0;
-                struct sockaddr_in sa = {0,};
-
-                sd = socket (PF_UNIX, SOCK_DGRAM, 0);
-                if (sd == -1)
-                        goto out;
-
-                buf.ifc_len = sizeof (struct ifreq);
-                buf.ifc_req = GF_CALLOC (1, sizeof (struct ifreq),
-                                         gf_gld_mt_ifreq);
-                size = buf.ifc_len;
-
-                ret = ioctl (sd, SIOCGIFCONF, &buf);
-                if (ret) {
-                        close (sd);
-                        goto out;
-                }
-
-                while (size <= buf.ifc_len) {
-                        size += sizeof (struct ifreq);
-                        buf.ifc_len = size;
-                        buf.ifc_req = GF_REALLOC (buf.ifc_req, size);
-                        ret = ioctl (sd, SIOCGIFCONF, &buf);
-                        if (ret) {
-                                close (sd);
-                                goto out;
-                        }
-                }
+        int     fd = 0;
 
-                ifr = buf.ifc_req;
-                num_req = size / sizeof (struct ifreq) - 1;
+        fd = open ("/dev/fuse", O_RDWR);
 
-                while (num_req--) {
-                        char *addr = inet_ntoa ( *(struct in_addr *)
-                                &ifr->ifr_addr.sa_data[sizeof(sa.sin_port)]);
-                        if (!strcmp (addr, hostname)) {
-                                gf_log ("", GF_LOG_DEBUG, "%s found as local",
-                                        addr);
-                                found = 1;
-                        }
-                        ifr++;
-                }
-
-                if (sd > 0)
-                        close (sd);
-        }
-
-
-
-
-out:
-        if (result)
-                freeaddrinfo (result);
-
-        if (buf.ifc_req)
-                GF_FREE (buf.ifc_req);
-
-        return !found;
+        if (fd > -1 && !close (fd))
+                return _gf_true;
+        else
+                return _gf_false;
 }
 
 int32_t
@@ -200,25 +137,28 @@ glusterd_lock (uuid_t   uuid)
         char    new_owner_str[50];
         char    owner_str[50];
         int     ret = -1;
+        xlator_t *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
 
         GF_ASSERT (uuid);
-        uuid_unparse (uuid, new_owner_str);
 
         glusterd_get_lock_owner (&owner);
 
         if (!uuid_is_null (owner)) {
-                uuid_unparse (owner, owner_str);
-                gf_log ("glusterd", GF_LOG_ERROR, "Unable to get lock"
-                        " for uuid: %s, lock held by: %s", new_owner_str,
-                        owner_str);
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get lock"
+                        " for uuid: %s, lock held by: %s",
+                        uuid_utoa_r (uuid, new_owner_str),
+                        uuid_utoa_r (owner, owner_str));
                 goto out;
         }
 
         ret = glusterd_set_lock_owner (uuid);
 
         if (!ret) {
-                gf_log ("glusterd", GF_LOG_NORMAL, "Cluster lock held by"
-                         " %s", new_owner_str);
+                gf_log (this->name, GF_LOG_DEBUG, "Cluster lock held by"
+                         " %s", uuid_utoa (uuid));
         }
 
 out:
@@ -233,30 +173,33 @@ glusterd_unlock (uuid_t uuid)
         char    new_owner_str[50];
         char    owner_str[50];
         int32_t ret = -1;
+        xlator_t *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
 
         GF_ASSERT (uuid);
-        uuid_unparse (uuid, new_owner_str);
 
         glusterd_get_lock_owner (&owner);
 
-        if (NULL == owner) {
-                gf_log ("glusterd", GF_LOG_ERROR, "Cluster lock not held!");
+        if (uuid_is_null (owner)) {
+                gf_log (this->name, GF_LOG_ERROR, "Cluster lock not held!");
                 goto out;
         }
 
         ret = uuid_compare (uuid, owner);
 
         if (ret) {
-               uuid_unparse (owner, owner_str);
-               gf_log ("glusterd", GF_LOG_ERROR, "Cluster lock held by %s"
-                        " ,unlock req from %s!", owner_str, new_owner_str);
+               gf_log (this->name, GF_LOG_ERROR, "Cluster lock held by %s ,"
+                       "unlock req from %s!", uuid_utoa_r (owner ,owner_str)
+                        , uuid_utoa_r (uuid, new_owner_str));
                goto out;
         }
 
         ret = glusterd_unset_lock_owner (uuid);
 
         if (ret) {
-                gf_log ("glusterd", GF_LOG_ERROR, "Unable to clear cluster "
+                gf_log (this->name, GF_LOG_ERROR, "Unable to clear cluster "
                         "lock");
                 goto out;
         }
@@ -277,68 +220,70 @@ glusterd_get_uuid (uuid_t *uuid)
 
         GF_ASSERT (priv);
 
-        uuid_copy (*uuid, priv->uuid);
+        uuid_copy (*uuid, MY_UUID);
 
         return 0;
 }
 
 int
-glusterd_submit_request (glusterd_peerinfo_t *peerinfo, void *req,
+glusterd_submit_request (struct rpc_clnt *rpc, void *req,
                          call_frame_t *frame, rpc_clnt_prog_t *prog,
                          int procnum, struct iobref *iobref,
-                         gd_serialize_t sfunc, xlator_t *this,
-                         fop_cbk_fn_t cbkfn)
+                         xlator_t *this, fop_cbk_fn_t cbkfn, xdrproc_t xdrproc)
 {
         int                     ret         = -1;
         struct iobuf            *iobuf      = NULL;
         int                     count      = 0;
         char                    new_iobref = 0, start_ping = 0;
         struct iovec            iov         = {0, };
+        ssize_t                 req_size    = 0;
 
-        GF_ASSERT (peerinfo);
+        GF_ASSERT (rpc);
         GF_ASSERT (this);
 
-        iobuf = iobuf_get (this->ctx->iobuf_pool);
-        if (!iobuf) {
-                goto out;
-        };
+        if (req) {
+                req_size = xdr_sizeof (xdrproc, req);
+                iobuf = iobuf_get2 (this->ctx->iobuf_pool, req_size);
+                if (!iobuf) {
+                        goto out;
+                };
 
-        if (!iobref) {
-                iobref = iobref_new ();
                 if (!iobref) {
-                        goto out;
-                }
+                        iobref = iobref_new ();
+                        if (!iobref) {
+                                goto out;
+                        }
 
-                new_iobref = 1;
-        }
+                        new_iobref = 1;
+                }
 
-        iobref_add (iobref, iobuf);
+                iobref_add (iobref, iobuf);
 
-        iov.iov_base = iobuf->ptr;
-        iov.iov_len  = 128 * GF_UNIT_KB;
+                iov.iov_base = iobuf->ptr;
+                iov.iov_len  = iobuf_pagesize (iobuf);
 
-        /* Create the xdr payload */
-        if (req && sfunc) {
-                ret = sfunc (iov, req);
+                /* Create the xdr payload */
+                ret = xdr_serialize_generic (iov, req, xdrproc);
                 if (ret == -1) {
                         goto out;
                 }
                 iov.iov_len = ret;
                 count = 1;
         }
+
         /* Send the msg */
-        ret = rpc_clnt_submit (peerinfo->rpc, prog, procnum, cbkfn,
+        ret = rpc_clnt_submit (rpc, prog, procnum, cbkfn,
                                &iov, count,
                                NULL, 0, iobref, frame, NULL, 0, NULL, 0, NULL);
 
         if (ret == 0) {
-                pthread_mutex_lock (&peerinfo->rpc->conn.lock);
+                pthread_mutex_lock (&rpc->conn.lock);
                 {
-                        if (!peerinfo->rpc->conn.ping_started) {
+                        if (!rpc->conn.ping_started) {
                                 start_ping = 1;
                         }
                 }
-                pthread_mutex_unlock (&peerinfo->rpc->conn.lock);
+                pthread_mutex_unlock (&rpc->conn.lock);
         }
 
         if (start_ping)
@@ -355,18 +300,19 @@ out:
         return ret;
 }
 
-
 struct iobuf *
 glusterd_serialize_reply (rpcsvc_request_t *req, void *arg,
-                          gd_serialize_t sfunc, struct iovec *outmsg)
+                          struct iovec *outmsg, xdrproc_t xdrproc)
 {
         struct iobuf            *iob = NULL;
         ssize_t                  retlen = -1;
+        ssize_t                  rsp_size = 0;
 
         /* First, get the io buffer into which the reply in arg will
          * be serialized.
          */
-        iob = iobuf_get (req->svc->ctx->iobuf_pool);
+        rsp_size = xdr_sizeof (xdrproc, arg);
+        iob = iobuf_get2 (req->svc->ctx->iobuf_pool, rsp_size);
         if (!iob) {
                 gf_log ("", GF_LOG_ERROR, "Failed to get iobuf");
                 goto ret;
@@ -379,7 +325,7 @@ glusterd_serialize_reply (rpcsvc_request_t *req, void *arg,
         /* retlen is used to received the error since size_t is unsigned and we
          * need -1 for error notification during encoding.
          */
-        retlen = sfunc (*outmsg, arg);
+        retlen = xdr_serialize_generic (*outmsg, arg, xdrproc);
         if (retlen == -1) {
                 gf_log ("", GF_LOG_ERROR, "Failed to encode message");
                 goto ret;
@@ -398,7 +344,7 @@ ret:
 int
 glusterd_submit_reply (rpcsvc_request_t *req, void *arg,
                        struct iovec *payload, int payloadcount,
-                       struct iobref *iobref, gd_serialize_t sfunc)
+                       struct iobref *iobref, xdrproc_t xdrproc)
 {
         struct iobuf           *iob        = NULL;
         int                     ret        = -1;
@@ -410,7 +356,6 @@ glusterd_submit_reply (rpcsvc_request_t *req, void *arg,
                 goto out;
         }
 
-
         if (!iobref) {
                 iobref = iobref_new ();
                 if (!iobref) {
@@ -421,14 +366,13 @@ glusterd_submit_reply (rpcsvc_request_t *req, void *arg,
                 new_iobref = 1;
         }
 
-        iob = glusterd_serialize_reply (req, arg, sfunc, &rsp);
+        iob = glusterd_serialize_reply (req, arg, &rsp, xdrproc);
         if (!iob) {
                 gf_log ("", GF_LOG_ERROR, "Failed to serialize reply");
-                goto out;
+        } else {
+                iobref_add (iobref, iob);
         }
 
-        iobref_add (iobref, iob);
-
         ret = rpcsvc_submit_generic (req, &rsp, 1, payload, payloadcount,
                                      iobref);
 
@@ -436,7 +380,6 @@ glusterd_submit_reply (rpcsvc_request_t *req, void *arg,
          * we can safely unref the iob in the hope that RPC layer must have
          * ref'ed the iob on receiving into the txlist.
          */
-        iobuf_unref (iob);
         if (ret == -1) {
                 gf_log ("", GF_LOG_ERROR, "Reply submission failed");
                 goto out;
@@ -449,6 +392,8 @@ out:
                 iobref_unref (iobref);
         }
 
+        if (iob)
+                iobuf_unref (iob);
         return ret;
 }
 
@@ -468,7 +413,7 @@ glusterd_check_volume_exists (char *volname)
         ret = stat (pathname, &stbuf);
 
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG, "Volume %s does not exist."
+                gf_log (THIS->name, GF_LOG_DEBUG, "Volume %s does not exist."
                         "stat failed with errno : %d on path: %s",
                         volname, errno, pathname);
                 return _gf_false;
@@ -491,26 +436,300 @@ glusterd_volinfo_new (glusterd_volinfo_t **volinfo)
         if (!new_volinfo)
                 goto out;
 
+        LOCK_INIT (&new_volinfo->lock);
         INIT_LIST_HEAD (&new_volinfo->vol_list);
+        INIT_LIST_HEAD (&new_volinfo->snapvol_list);
         INIT_LIST_HEAD (&new_volinfo->bricks);
+        INIT_LIST_HEAD (&new_volinfo->snap_volumes);
 
-        ret = glusterd_default_xlator_options (new_volinfo);
-        if (ret) {
-                if (new_volinfo)
-                        GF_FREE (new_volinfo);
+        new_volinfo->dict = dict_new ();
+        if (!new_volinfo->dict) {
+                GF_FREE (new_volinfo);
+
+                goto out;
+        }
+
+        new_volinfo->gsync_slaves = dict_new ();
+        if (!new_volinfo->gsync_slaves) {
+                GF_FREE (new_volinfo);
 
                 goto out;
         }
 
+        snprintf (new_volinfo->parent_volname, GLUSTERD_MAX_VOLUME_NAME, "N/A");
+
+        new_volinfo->snap_max_hard_limit = GLUSTERD_SNAPS_MAX_HARD_LIMIT;
+
+        new_volinfo->xl = THIS;
+
         *volinfo = new_volinfo;
 
         ret = 0;
 
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
+/* This function will create a new volinfo and then
+ * dup the entries from volinfo to the new_volinfo.
+ *
+ * @param volinfo       volinfo which will be duplicated
+ * @param dup_volinfo   new volinfo which will be created
+ * @param set_userauth  if this true then auth info is also set
+ *
+ * @return 0 on success else -1
+ */
+int32_t
+glusterd_volinfo_dup (glusterd_volinfo_t *volinfo,
+                      glusterd_volinfo_t **dup_volinfo,
+                      gf_boolean_t set_userauth)
+{
+        int32_t                 ret             = -1;
+        xlator_t                *this           = NULL;
+        glusterd_volinfo_t      *new_volinfo    = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_VALIDATE_OR_GOTO (this->name, volinfo, out);
+        GF_VALIDATE_OR_GOTO (this->name, dup_volinfo, out);
+
+        ret = glusterd_volinfo_new (&new_volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "not able to create the "
+                        "duplicate volinfo for the volume %s",
+                        volinfo->volname);
+                goto out;
+        }
+
+        new_volinfo->type = volinfo->type;
+        new_volinfo->replica_count = volinfo->replica_count;
+        new_volinfo->stripe_count = volinfo->stripe_count;
+        new_volinfo->dist_leaf_count = volinfo->dist_leaf_count;
+        new_volinfo->sub_count = volinfo->sub_count;
+        new_volinfo->transport_type = volinfo->transport_type;
+        new_volinfo->nfs_transport_type = volinfo->nfs_transport_type;
+        new_volinfo->brick_count = volinfo->brick_count;
+
+        dict_copy (volinfo->dict, new_volinfo->dict);
+        gd_update_volume_op_versions (new_volinfo);
+
+        if (set_userauth) {
+                glusterd_auth_set_username (new_volinfo,
+                                            volinfo->auth.username);
+                glusterd_auth_set_password (new_volinfo,
+                                            volinfo->auth.password);
+        }
+
+        *dup_volinfo = new_volinfo;
+        ret = 0;
+out:
+        if (ret && (NULL != new_volinfo)) {
+                (void) glusterd_volinfo_delete (new_volinfo);
+        }
+        return ret;
+}
+
+/* This function will duplicate brickinfo
+ *
+ * @param brickinfo     Source brickinfo
+ * @param dup_brickinfo Destination brickinfo
+ *
+ * @return 0 on success else -1
+ */
+int32_t
+glusterd_brickinfo_dup (glusterd_brickinfo_t *brickinfo,
+                        glusterd_brickinfo_t *dup_brickinfo)
+{
+        int32_t         ret     = -1;
+        xlator_t        *this   = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        GF_VALIDATE_OR_GOTO (this->name, brickinfo, out);
+        GF_VALIDATE_OR_GOTO (this->name, dup_brickinfo, out);
+
+        strcpy (dup_brickinfo->hostname, brickinfo->hostname);
+        strcpy (dup_brickinfo->path, brickinfo->path);
+        strcpy (dup_brickinfo->device_path, brickinfo->device_path);
+        ret = gf_canonicalize_path (dup_brickinfo->path);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to canonicalize "
+                        "brick path");
+                goto out;
+        }
+        uuid_copy (dup_brickinfo->uuid, brickinfo->uuid);
+
+        dup_brickinfo->port = brickinfo->port;
+        dup_brickinfo->rdma_port = brickinfo->rdma_port;
+        if (NULL != brickinfo->logfile) {
+                dup_brickinfo->logfile = gf_strdup (brickinfo->logfile);
+                if (NULL == dup_brickinfo->logfile) {
+                        ret = -1;
+                        goto out;
+                }
+        }
+        dup_brickinfo->status = brickinfo->status;
+        dup_brickinfo->snap_status = brickinfo->snap_status;
+out:
+        return ret;
+}
+
+/* This function will copy snap volinfo to the new
+ * passed volinfo and regenerate backend store files
+ * for the restored snap.
+ *
+ * @param new_volinfo   new volinfo
+ * @param snap_volinfo  volinfo of snap volume
+ *
+ * @return 0 on success and -1 on failure
+ *
+ * TODO: Duplicate all members of volinfo, e.g. geo-rep sync slaves
+ */
+int32_t
+glusterd_snap_volinfo_restore (dict_t *rsp_dict,
+                               glusterd_volinfo_t *new_volinfo,
+                               glusterd_volinfo_t *snap_volinfo)
+{
+        int32_t                 brick_count     = -1;
+        int32_t                 ret             = -1;
+        xlator_t                *this           = NULL;
+        glusterd_brickinfo_t    *brickinfo      = NULL;
+        glusterd_brickinfo_t    *new_brickinfo  = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (rsp_dict);
+
+        GF_VALIDATE_OR_GOTO (this->name, new_volinfo, out);
+        GF_VALIDATE_OR_GOTO (this->name, snap_volinfo, out);
+
+        brick_count = 0;
+        list_for_each_entry (brickinfo, &snap_volinfo->bricks, brick_list) {
+                ret = glusterd_brickinfo_new (&new_brickinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to create "
+                                "new brickinfo");
+                        goto out;
+                }
+
+                /* Duplicate brickinfo */
+                ret = glusterd_brickinfo_dup (brickinfo, new_brickinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to dup "
+                                "brickinfo");
+                        goto out;
+                }
+
+                /* If the brick is not of this peer, or snapshot is missed *
+                 * for the brick do not replace the xattr for it */
+                if ((!uuid_compare (brickinfo->uuid, MY_UUID)) &&
+                    (brickinfo->snap_status != -1)) {
+                        /* We need to replace the volume id of all the bricks
+                         * to the volume id of the origin volume. new_volinfo
+                         * has the origin volume's volume id*/
+                        ret = sys_lsetxattr (new_brickinfo->path,
+                                             GF_XATTR_VOL_ID_KEY,
+                                             new_volinfo->volume_id,
+                                             sizeof (new_volinfo->volume_id),
+                                             XATTR_REPLACE);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                        "set extended attribute %s on %s. "
+                                        "Reason: %s, snap: %s",
+                                        GF_XATTR_VOL_ID_KEY,
+                                        new_brickinfo->path, strerror (errno),
+                                        new_volinfo->volname);
+                                goto out;
+                        }
+                }
+
+                /* If a snapshot is pending for this brick then
+                 * restore should also be pending
+                 */
+                if (brickinfo->snap_status == -1) {
+                        /* Adding missed delete to the dict */
+                        ret = glusterd_add_missed_snaps_to_dict
+                                                (rsp_dict,
+                                                 snap_volinfo->volname,
+                                                 brickinfo,
+                                                 brick_count + 1,
+                                                 GF_SNAP_OPTION_TYPE_RESTORE);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to add missed snapshot info "
+                                        "for %s:%s in the rsp_dict",
+                                        brickinfo->hostname,
+                                        brickinfo->path);
+                                goto out;
+                        }
+                }
+
+                list_add_tail (&new_brickinfo->brick_list,
+                                &new_volinfo->bricks);
+                /* ownership of new_brickinfo is passed to new_volinfo */
+                new_brickinfo = NULL;
+                brick_count++;
+        }
+
+        /* Regenerate all volfiles */
+        ret = glusterd_create_volfiles_and_notify_services (new_volinfo);
+
+out:
+        if (ret && (NULL != new_brickinfo)) {
+                (void) glusterd_brickinfo_delete (new_brickinfo);
+        }
+
+        return ret;
+}
+
+void
+glusterd_auth_cleanup (glusterd_volinfo_t *volinfo) {
+
+        GF_ASSERT (volinfo);
+
+        GF_FREE (volinfo->auth.username);
+
+        GF_FREE (volinfo->auth.password);
+}
+
+char *
+glusterd_auth_get_username (glusterd_volinfo_t *volinfo) {
+
+        GF_ASSERT (volinfo);
+
+        return volinfo->auth.username;
+}
+
+char *
+glusterd_auth_get_password (glusterd_volinfo_t *volinfo) {
+
+        GF_ASSERT (volinfo);
+
+        return volinfo->auth.password;
+}
+
+int32_t
+glusterd_auth_set_username (glusterd_volinfo_t *volinfo, char *username) {
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (username);
+
+        volinfo->auth.username = gf_strdup (username);
+        return 0;
+}
+
+int32_t
+glusterd_auth_set_password (glusterd_volinfo_t *volinfo, char *password) {
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (password);
+
+        volinfo->auth.password = gf_strdup (password);
+        return 0;
+}
+
 int32_t
 glusterd_brickinfo_delete (glusterd_brickinfo_t *brickinfo)
 {
@@ -520,8 +739,7 @@ glusterd_brickinfo_delete (glusterd_brickinfo_t *brickinfo)
 
         list_del_init (&brickinfo->brick_list);
 
-        if (brickinfo->logfile)
-                GF_FREE (brickinfo->logfile);
+        GF_FREE (brickinfo->logfile);
         GF_FREE (brickinfo);
 
         ret = 0;
@@ -530,16 +748,14 @@ glusterd_brickinfo_delete (glusterd_brickinfo_t *brickinfo)
 }
 
 int32_t
-glusterd_volinfo_delete (glusterd_volinfo_t *volinfo)
+glusterd_volume_brickinfos_delete (glusterd_volinfo_t *volinfo)
 {
-        int32_t                 ret = -1;
         glusterd_brickinfo_t    *brickinfo = NULL;
         glusterd_brickinfo_t    *tmp = NULL;
+        int32_t                 ret = 0;
 
         GF_ASSERT (volinfo);
 
-        list_del_init (&volinfo->vol_list);
-
         list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks,
                                    brick_list) {
                 ret = glusterd_brickinfo_delete (brickinfo);
@@ -547,17 +763,40 @@ glusterd_volinfo_delete (glusterd_volinfo_t *volinfo)
                         goto out;
         }
 
-        dict_unref (volinfo->dict);
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_volinfo_delete (glusterd_volinfo_t *volinfo)
+{
+        int32_t                 ret = -1;
+
+        GF_ASSERT (volinfo);
+
+        list_del_init (&volinfo->vol_list);
+        list_del_init (&volinfo->snapvol_list);
+
+        ret = glusterd_volume_brickinfos_delete (volinfo);
+        if (ret)
+                goto out;
+        if (volinfo->dict)
+                dict_unref (volinfo->dict);
+        if (volinfo->gsync_slaves)
+                dict_unref (volinfo->gsync_slaves);
+        GF_FREE (volinfo->logdir);
+
+        glusterd_auth_cleanup (volinfo);
 
         GF_FREE (volinfo);
         ret = 0;
 
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
-
 int32_t
 glusterd_brickinfo_new (glusterd_brickinfo_t **brickinfo)
 {
@@ -579,7 +818,7 @@ glusterd_brickinfo_new (glusterd_brickinfo_t **brickinfo)
         ret = 0;
 
 out:
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
@@ -587,69 +826,59 @@ int32_t
 glusterd_resolve_brick (glusterd_brickinfo_t *brickinfo)
 {
         int32_t                 ret = -1;
-        glusterd_peerinfo_t     *peerinfo = NULL;
-        glusterd_conf_t         *priv = NULL;
-        glusterd_peer_hostname_t        *host = NULL;
+        xlator_t                *this = NULL;
 
-        priv = THIS->private;
-        GF_ASSERT (priv);
+        this = THIS;
+        GF_ASSERT (this);
 
         GF_ASSERT (brickinfo);
 
-        ret = glusterd_friend_find (NULL, brickinfo->hostname, &peerinfo);
-
-        if (!ret) {
-                uuid_copy (brickinfo->uuid, peerinfo->uuid);
-        } else {
-                list_for_each_entry (host, &priv->hostnames, hostname_list) {
-                        if (!strcmp (host->hostname, brickinfo->hostname)) {
-                                uuid_copy (brickinfo->uuid, priv->uuid);
-                                ret = 0;
-                                break;
-                        }
-
-                }
-        }
-
-        if (ret) {
-                ret = glusterd_is_local_addr (brickinfo->hostname);
-                if (!ret)
-                        uuid_copy (brickinfo->uuid, priv->uuid);
-        }
-
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        ret = glusterd_hostname_to_uuid (brickinfo->hostname, brickinfo->uuid);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
 int32_t
-glusterd_brickinfo_from_brick (char *brick,
-                               glusterd_brickinfo_t **brickinfo)
+glusterd_brickinfo_new_from_brick (char *brick,
+                                   glusterd_brickinfo_t **brickinfo)
 {
         int32_t                 ret = -1;
         glusterd_brickinfo_t    *new_brickinfo = NULL;
         char                    *hostname = NULL;
         char                    *path = NULL;
-        char                    *tmp = NULL;
-        char                    *tmpstr = NULL;
+        char                    *tmp_host = NULL;
+        char                    *tmp_path = NULL;
+        char                    *vg       = NULL;
 
         GF_ASSERT (brick);
         GF_ASSERT (brickinfo);
 
-        tmp = gf_strdup (brick);
-        if (!tmp) {
-                gf_log ("glusterd", GF_LOG_ERROR,
-                        "Out of memory");
+        tmp_host = gf_strdup (brick);
+        if (tmp_host && !get_host_name (tmp_host, &hostname))
+                goto out;
+        tmp_path = gf_strdup (brick);
+        if (tmp_path && !get_path_name (tmp_path, &path))
                 goto out;
-        }
-
-        hostname = strtok_r (tmp, ":", &tmpstr);
-        path = strtok_r (NULL, ":", &tmpstr);
 
         GF_ASSERT (hostname);
         GF_ASSERT (path);
 
         ret = glusterd_brickinfo_new (&new_brickinfo);
+        if (ret)
+                goto out;
 
+#ifdef HAVE_BD_XLATOR
+        vg = strchr (path, '?');
+        /* ? is used as a delimiter for vg */
+        if (vg) {
+                strncpy (new_brickinfo->vg, vg + 1, PATH_MAX - 1);
+                *vg = '\0';
+        }
+        new_brickinfo->caps = CAPS_BD;
+#else
+        vg = NULL; /* Avoid compiler warnings when BD not enabled */
+#endif
+        ret = gf_canonicalize_path (path);
         if (ret)
                 goto out;
 
@@ -660,83 +889,440 @@ glusterd_brickinfo_from_brick (char *brick,
 
         ret = 0;
 out:
-        if (tmp)
-                GF_FREE (tmp);
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        GF_FREE (tmp_host);
+        if (tmp_host)
+                GF_FREE (tmp_path);
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
-int32_t
-glusterd_brickinfo_get (char *brick, glusterd_volinfo_t *volinfo,
-                        glusterd_brickinfo_t **brickinfo)
+static gf_boolean_t
+_is_prefix (char *str1, char *str2)
 {
-        int32_t                 ret = -1;
-        char                    *hostname = NULL;
-        char                    *path = NULL;
-        char                    *dup_brick = NULL;
-        char                    *free_ptr = NULL;
-        glusterd_brickinfo_t    *tmp = NULL;
+        GF_ASSERT (str1);
+        GF_ASSERT (str2);
+
+        int             i = 0;
+        int             len1 = 0;
+        int             len2 = 0;
+        int             small_len = 0;
+        char            *bigger = NULL;
+        gf_boolean_t    prefix = _gf_true;
+
+        len1 = strlen (str1);
+        len2 = strlen (str2);
+        small_len = min (len1, len2);
+        for (i = 0; i < small_len; i++) {
+                if (str1[i] != str2[i]) {
+                        prefix = _gf_false;
+                        break;
+                }
+        }
 
-        GF_ASSERT (brick);
-        GF_ASSERT (brickinfo);
-        GF_ASSERT (volinfo);
+        if (len1 < len2)
+            bigger = str2;
 
-        gf_log ("", GF_LOG_NORMAL, "brick: %s", brick);
+        else if (len1 > len2)
+            bigger = str1;
 
-        dup_brick = gf_strdup (brick);
-        if (!dup_brick) {
-                gf_log ("", GF_LOG_ERROR,
-                        "Out of memory");
-                ret = -1;
+        else
+            return prefix;
+
+        if (bigger[small_len] != '/')
+            prefix = _gf_false;
+
+        return prefix;
+}
+
+/* Checks if @path is available in the peer identified by @uuid
+ * 'availability' is determined by querying current state of volumes
+ * in the cluster. */
+gf_boolean_t
+glusterd_is_brickpath_available (uuid_t uuid, char *path)
+{
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        glusterd_volinfo_t      *volinfo   = NULL;
+        glusterd_conf_t         *priv      = NULL;
+        gf_boolean_t            available  = _gf_false;
+        char                    tmp_path[PATH_MAX+1] = {0};
+        char                    tmp_brickpath[PATH_MAX+1] = {0};
+
+        priv = THIS->private;
+
+        strncpy (tmp_path, path, PATH_MAX);
+        /* path may not yet exist */
+        if (!realpath (path, tmp_path)) {
+                if (errno != ENOENT) {
+                        goto out;
+                }
+                /* When realpath(3) fails, tmp_path is undefined. */
+                strncpy(tmp_path,path,PATH_MAX);
+        }
+
+        list_for_each_entry (volinfo, &priv->volumes, vol_list) {
+                list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                        if (uuid_compare (uuid, brickinfo->uuid))
+                                continue;
+
+                        if (!realpath (brickinfo->path, tmp_brickpath)) {
+                            if (errno == ENOENT)
+                                strncpy (tmp_brickpath, brickinfo->path,
+                                         PATH_MAX);
+                            else
+                                goto out;
+                        }
+
+                        if (_is_prefix (tmp_brickpath, tmp_path))
+                                goto out;
+                }
+        }
+        available = _gf_true;
+out:
+        return available;
+}
+
+#ifdef HAVE_BD_XLATOR
+/*
+ * Sets the tag of the format "trusted.glusterfs.volume-id:<uuid>" in
+ * the brick VG. It is used to avoid using same VG for another brick.
+ * @volume-id - gfid, @brick - brick info, @msg - Error message returned
+ * to the caller
+ */
+int
+glusterd_bd_set_vg_tag (unsigned char *volume_id, glusterd_brickinfo_t *brick,
+                        char *msg, int msg_size)
+{
+        lvm_t        handle    = NULL;
+        vg_t         vg        = NULL;
+        char        *uuid      = NULL;
+        int          ret       = -1;
+
+        gf_asprintf (&uuid, "%s:%s", GF_XATTR_VOL_ID_KEY,
+                     uuid_utoa (volume_id));
+        if (!uuid) {
+                snprintf (msg, sizeof(*msg), "Could not allocate memory "
+                          "for tag");
+                return -1;
+        }
+
+        handle = lvm_init (NULL);
+        if (!handle) {
+                snprintf (msg, sizeof(*msg), "lvm_init failed");
+                goto out;
+        }
+
+        vg = lvm_vg_open (handle, brick->vg, "w", 0);
+        if (!vg) {
+                snprintf (msg, sizeof(*msg), "Could not open VG %s",
+                          brick->vg);
                 goto out;
+        }
+
+        if (lvm_vg_add_tag (vg, uuid) < 0) {
+                snprintf (msg, sizeof(*msg), "Could not set tag %s for "
+                          "VG %s", uuid, brick->vg);
+                goto out;
+        }
+        lvm_vg_write (vg);
+        ret = 0;
+out:
+        GF_FREE (uuid);
+
+        if (vg)
+                lvm_vg_close (vg);
+        if (handle)
+                lvm_quit (handle);
+
+        return ret;
+}
+#endif
+
+int
+glusterd_validate_and_create_brickpath (glusterd_brickinfo_t *brickinfo,
+                                        uuid_t volume_id, char **op_errstr,
+                                        gf_boolean_t is_force)
+{
+        int          ret                 = -1;
+        char         parentdir[PATH_MAX] = {0,};
+        struct stat  parent_st           = {0,};
+        struct stat  brick_st            = {0,};
+        struct stat  root_st             = {0,};
+        char         msg[2048]           = {0,};
+        gf_boolean_t is_created          = _gf_false;
+
+        ret = mkdir (brickinfo->path, 0777);
+        if (ret) {
+                if (errno != EEXIST) {
+                        snprintf (msg, sizeof (msg), "Failed to create brick "
+                                  "directory for brick %s:%s. Reason : %s ",
+                                  brickinfo->hostname, brickinfo->path,
+                                  strerror (errno));
+                        goto out;
+                }
         } else {
-                free_ptr = dup_brick;
+                is_created = _gf_true;
         }
 
-        hostname = strtok (dup_brick, ":");
-        path = strtok (NULL, ":");
+        ret = lstat (brickinfo->path, &brick_st);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "lstat failed on %s. Reason : %s",
+                          brickinfo->path, strerror (errno));
+                goto out;
+        }
 
-        if (!hostname || !path) {
-                gf_log ("", GF_LOG_ERROR,
-                        "brick %s is not of form <HOSTNAME>:<export-dir>",
-                        brick);
+        if ((!is_created) && (!S_ISDIR (brick_st.st_mode))) {
+                snprintf (msg, sizeof (msg), "The provided path %s which is "
+                          "already present, is not a directory",
+                          brickinfo->path);
                 ret = -1;
                 goto out;
         }
 
-        list_for_each_entry (tmp, &volinfo->bricks, brick_list) {
+        snprintf (parentdir, sizeof (parentdir), "%s/..", brickinfo->path);
+
+        ret = lstat ("/", &root_st);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "lstat failed on /. Reason : %s",
+                          strerror (errno));
+                goto out;
+        }
+
+        ret = lstat (parentdir, &parent_st);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "lstat failed on %s. Reason : %s",
+                          parentdir, strerror (errno));
+                goto out;
+        }
 
-                if ((!strcmp (tmp->hostname, hostname)) &&
-                        !strcmp (tmp->path, path)) {
-                        gf_log ("", GF_LOG_NORMAL, "Found brick");
+        if (!is_force) {
+                if (brick_st.st_dev != parent_st.st_dev) {
+                        snprintf (msg, sizeof (msg), "The brick %s:%s is a "
+                                  "mount point. Please create a sub-directory "
+                                  "under the mount point and use that as the "
+                                  "brick directory. Or use 'force' at the end "
+                                  "of the command if you want to override this "
+                                  "behavior.", brickinfo->hostname,
+                                  brickinfo->path);
+                        ret = -1;
+                        goto out;
+                }
+                else if (parent_st.st_dev == root_st.st_dev) {
+                        snprintf (msg, sizeof (msg), "The brick %s:%s is "
+                                  "is being created in the root partition. It "
+                                  "is recommended that you don't use the "
+                                  "system's root partition for storage backend."
+                                  " Or use 'force' at the end of the command if"
+                                  " you want to override this behavior.",
+                                  brickinfo->hostname, brickinfo->path);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+#ifdef HAVE_BD_XLATOR
+        if (brickinfo->vg[0]) {
+                ret = glusterd_bd_set_vg_tag (volume_id, brickinfo, msg,
+                                              sizeof(msg));
+                if (ret)
+                        goto out;
+        }
+#endif
+        ret = glusterd_check_and_set_brick_xattr (brickinfo->hostname,
+                                                  brickinfo->path, volume_id,
+                                                  op_errstr, is_force);
+        if (ret)
+                goto out;
+
+        ret = 0;
+
+out:
+        if (ret && is_created)
+                rmdir (brickinfo->path);
+        if (ret && !*op_errstr && msg[0] != '\0')
+                *op_errstr = gf_strdup (msg);
+
+        return ret;
+}
+
+int32_t
+glusterd_volume_brickinfo_get (uuid_t uuid, char *hostname, char *path,
+                               glusterd_volinfo_t *volinfo,
+                               glusterd_brickinfo_t **brickinfo)
+{
+        glusterd_brickinfo_t    *brickiter = NULL;
+        uuid_t                  peer_uuid = {0};
+        int32_t                 ret = -1;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+
+        if (uuid) {
+                uuid_copy (peer_uuid, uuid);
+        } else {
+                ret = glusterd_hostname_to_uuid (hostname, peer_uuid);
+                if (ret)
+                        goto out;
+        }
+        ret = -1;
+        list_for_each_entry (brickiter, &volinfo->bricks, brick_list) {
+
+                if ((uuid_is_null (brickiter->uuid)) &&
+                    (glusterd_resolve_brick (brickiter) != 0))
+                        goto out;
+                if (uuid_compare (peer_uuid, brickiter->uuid))
+                        continue;
+
+                if (strcmp (brickiter->path, path) == 0) {
+                        gf_log (this->name, GF_LOG_DEBUG, LOGSTR_FOUND_BRICK,
+                                brickiter->hostname, brickiter->path,
+                                volinfo->volname);
                         ret = 0;
+                        if (brickinfo)
+                                *brickinfo = brickiter;
                         break;
                 }
         }
 
-        *brickinfo = tmp;
-
 out:
-        if (free_ptr)
-                GF_FREE (free_ptr);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_volume_brickinfo_get_by_brick (char *brick,
+                                        glusterd_volinfo_t *volinfo,
+                                        glusterd_brickinfo_t **brickinfo)
+{
+        int32_t                 ret = -1;
+        glusterd_brickinfo_t    *tmp_brickinfo = NULL;
+
+        GF_ASSERT (brick);
+        GF_ASSERT (volinfo);
+
+        ret = glusterd_brickinfo_new_from_brick (brick, &tmp_brickinfo);
+        if (ret)
+                goto out;
 
+        ret = glusterd_volume_brickinfo_get (NULL, tmp_brickinfo->hostname,
+                                             tmp_brickinfo->path, volinfo,
+                                             brickinfo);
+        (void) glusterd_brickinfo_delete (tmp_brickinfo);
+out:
         gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
+gf_boolean_t
+glusterd_is_brick_decommissioned (glusterd_volinfo_t *volinfo, char *hostname,
+                                  char *path)
+{
+        gf_boolean_t            decommissioned = _gf_false;
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        int                     ret = -1;
+
+        ret = glusterd_volume_brickinfo_get (NULL, hostname, path, volinfo,
+                                             &brickinfo);
+        if (ret)
+                goto out;
+        decommissioned = brickinfo->decommissioned;
+out:
+        return decommissioned;
+}
+
 int32_t
 glusterd_friend_cleanup (glusterd_peerinfo_t *peerinfo)
 {
         GF_ASSERT (peerinfo);
+        glusterd_peerctx_t      *peerctx = NULL;
+        gf_boolean_t            quorum_action = _gf_false;
+        glusterd_conf_t         *priv = THIS->private;
+
+        if (peerinfo->quorum_contrib != QUORUM_NONE)
+                quorum_action = _gf_true;
         if (peerinfo->rpc) {
+                /* cleanup the saved-frames before last unref */
+                synclock_unlock (&priv->big_lock);
+                rpc_clnt_connection_cleanup (&peerinfo->rpc->conn);
+                synclock_lock (&priv->big_lock);
+
+                peerctx = peerinfo->rpc->mydata;
+                peerinfo->rpc->mydata = NULL;
                 peerinfo->rpc = rpc_clnt_unref (peerinfo->rpc);
                 peerinfo->rpc = NULL;
+                if (peerctx) {
+                        GF_FREE (peerctx->errstr);
+                        GF_FREE (peerctx);
+                }
         }
         glusterd_peer_destroy (peerinfo);
 
+        if (quorum_action)
+                glusterd_do_quorum_action ();
         return 0;
 }
 
+int
+glusterd_volinfo_find_by_volume_id (uuid_t volume_id, glusterd_volinfo_t **volinfo)
+{
+        int32_t                 ret = -1;
+        xlator_t                *this = NULL;
+        glusterd_volinfo_t      *voliter = NULL;
+        glusterd_conf_t         *priv = NULL;
+
+        if (!volume_id)
+                return -1;
+
+        this = THIS;
+        priv = this->private;
+
+        list_for_each_entry (voliter, &priv->volumes, vol_list) {
+                if (uuid_compare (volume_id, voliter->volume_id))
+                        continue;
+                *volinfo = voliter;
+                ret = 0;
+                gf_log (this->name, GF_LOG_DEBUG, "Volume %s found",
+                        voliter->volname);
+                break;
+        }
+        return ret;
+}
+
+int
+glusterd_snap_volinfo_find_by_volume_id (uuid_t volume_id,
+                                         glusterd_volinfo_t **volinfo)
+{
+        int32_t                  ret     = -1;
+        xlator_t                *this    = NULL;
+        glusterd_volinfo_t      *voliter = NULL;
+        glusterd_snap_t         *snap    = NULL;
+        glusterd_conf_t         *priv    = NULL;
+
+        this = THIS;
+        priv = this->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (volinfo);
+
+        if (uuid_is_null(volume_id)) {
+                gf_log (this->name, GF_LOG_WARNING, "Volume UUID is NULL");
+                goto out;
+        }
+
+        list_for_each_entry (snap, &priv->snapshots, snap_list) {
+                list_for_each_entry (voliter, &snap->volumes, vol_list) {
+                        if (uuid_compare (volume_id, voliter->volume_id))
+                                continue;
+                        *volinfo = voliter;
+                        ret = 0;
+                        goto out;
+                }
+        }
+
+        gf_log (this->name, GF_LOG_WARNING, "Snap volume not found");
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
 int32_t
 glusterd_volinfo_find (char *volname, glusterd_volinfo_t **volinfo)
 {
@@ -746,104 +1332,259 @@ glusterd_volinfo_find (char *volname, glusterd_volinfo_t **volinfo)
         glusterd_conf_t         *priv = NULL;
 
         GF_ASSERT (volname);
-
         this = THIS;
         GF_ASSERT (this);
 
         priv = this->private;
+        GF_ASSERT (priv);
 
         list_for_each_entry (tmp_volinfo, &priv->volumes, vol_list) {
                 if (!strcmp (tmp_volinfo->volname, volname)) {
-                        gf_log ("", GF_LOG_DEBUG, "Volume %s found", volname);
+                        gf_log (this->name, GF_LOG_DEBUG, "Volume %s found",
+                                volname);
                         ret = 0;
                         *volinfo = tmp_volinfo;
                         break;
                 }
         }
 
-
-        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
-
 int32_t
-glusterd_service_stop(const char *service, char *pidfile, int sig,
-                      gf_boolean_t keep_pidfile)
+glusterd_snap_volinfo_find (char *snap_volname, glusterd_snap_t *snap,
+                            glusterd_volinfo_t **volinfo)
 {
-        int32_t  ret = -1;
-        pid_t    pid = -1;
-        FILE    *file = NULL;
+        int32_t                  ret         = -1;
+        xlator_t                *this        = NULL;
+        glusterd_volinfo_t      *snap_vol    = NULL;
+        glusterd_conf_t         *priv        = NULL;
 
-        file = fopen (pidfile, "r+");
+        this = THIS;
+        priv = this->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (snap);
+        GF_ASSERT (snap_volname);
 
-        if (!file) {
-                gf_log ("", GF_LOG_ERROR, "Unable to open pidfile: %s",
-                                pidfile);
-                if (errno == ENOENT) {
-                        gf_log ("",GF_LOG_TRACE, "%s may not be running",
-                                service);
+        list_for_each_entry (snap_vol, &snap->volumes, vol_list) {
+                if (!strcmp (snap_vol->volname, snap_volname)) {
                         ret = 0;
+                        *volinfo = snap_vol;
                         goto out;
                 }
-                ret = -1;
-                goto out;
         }
 
-        ret = fscanf (file, "%d", &pid);
-        if (ret <= 0) {
-                gf_log ("", GF_LOG_ERROR, "Unable to read pidfile: %s",
-                                pidfile);
-                ret = -1;
-                goto out;
+        gf_log (this->name, GF_LOG_WARNING, "Snap volume %s not found",
+                snap_volname);
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_snap_volinfo_find_from_parent_volname (char *origin_volname,
+                                      glusterd_snap_t *snap,
+                                      glusterd_volinfo_t **volinfo)
+{
+        int32_t                  ret         = -1;
+        xlator_t                *this        = NULL;
+        glusterd_volinfo_t      *snap_vol    = NULL;
+        glusterd_conf_t         *priv        = NULL;
+
+        this = THIS;
+        priv = this->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (snap);
+        GF_ASSERT (origin_volname);
+
+        list_for_each_entry (snap_vol, &snap->volumes, vol_list) {
+                if (!strcmp (snap_vol->parent_volname, origin_volname)) {
+                        ret = 0;
+                        *volinfo = snap_vol;
+                        goto out;
+                }
         }
-        fclose (file);
-        file = NULL;
 
-        gf_log ("", GF_LOG_NORMAL, "Stopping gluster %s running in pid: %d",
-                service, pid);
+        gf_log (this->name, GF_LOG_DEBUG, "Snap volume not found(snap: %s, "
+                "origin-volume: %s", snap->snapname, origin_volname);
 
-        ret = kill (pid, sig);
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to kill pid %d", pid);
+int32_t
+glusterd_service_stop (const char *service, char *pidfile, int sig,
+                       gf_boolean_t force_kill)
+{
+        int32_t  ret = -1;
+        pid_t    pid = -1;
+        xlator_t *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        if (!glusterd_is_service_running (pidfile, &pid)) {
+                ret = 0;
+                gf_log (this->name, GF_LOG_INFO, "%s already stopped", service);
                 goto out;
         }
+        gf_log (this->name, GF_LOG_DEBUG, "Stopping gluster %s running in pid: "
+                "%d", service, pid);
 
-        if (keep_pidfile)
+        ret = kill (pid, sig);
+        if (!force_kill)
                 goto out;
 
-        ret = unlink (pidfile);
-
-        if (ret && (ENOENT != errno)) {
-                gf_log ("", GF_LOG_ERROR, "Unable to unlink pidfile: %s",
-                                pidfile);
-                goto out;
+        sleep (1);
+        if (glusterd_is_service_running (pidfile, NULL)) {
+                ret = kill (pid, SIGKILL);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Unable to "
+                                "kill pid %d reason: %s", pid,
+                                strerror(errno));
+                        goto out;
+                }
         }
 
         ret = 0;
 out:
-        if (file)
-                fclose (file);
+        return ret;
+}
+
+void
+glusterd_set_socket_filepath (char *sock_filepath, char *sockpath, size_t len)
+{
+        char md5_sum[MD5_DIGEST_LENGTH*2+1] = {0,};
+
+        md5_wrapper ((unsigned char *) sock_filepath, strlen(sock_filepath), md5_sum);
+        snprintf (sockpath, len, "%s/%s.socket", GLUSTERD_SOCK_DIR, md5_sum);
+}
+
+void
+glusterd_set_brick_socket_filepath (glusterd_volinfo_t *volinfo,
+                                    glusterd_brickinfo_t *brickinfo,
+                                    char *sockpath, size_t len)
+{
+        char                    export_path[PATH_MAX] = {0,};
+        char                    sock_filepath[PATH_MAX] = {0,};
+        char                    volume_dir[PATH_MAX] = {0,};
+        xlator_t                *this = NULL;
+        glusterd_conf_t         *priv = NULL;
+        int                     expected_file_len = 0;
+
+        expected_file_len = strlen (GLUSTERD_SOCK_DIR) + strlen ("/") +
+                            MD5_DIGEST_LENGTH*2 + strlen (".socket") + 1;
+        GF_ASSERT (len >= expected_file_len);
+        this = THIS;
+        GF_ASSERT (this);
+
+        priv = this->private;
+
+        GLUSTERD_GET_VOLUME_DIR (volume_dir, volinfo, priv);
+        GLUSTERD_REMOVE_SLASH_FROM_PATH (brickinfo->path, export_path);
+        snprintf (sock_filepath, PATH_MAX, "%s/run/%s-%s",
+                  volume_dir, brickinfo->hostname, export_path);
+
+        glusterd_set_socket_filepath (sock_filepath, sockpath, len);
+}
+
+/* connection happens only if it is not aleady connected,
+ * reconnections are taken care by rpc-layer
+ */
+int32_t
+glusterd_brick_connect (glusterd_volinfo_t  *volinfo,
+                        glusterd_brickinfo_t  *brickinfo, char *socketpath)
+{
+        int                     ret = 0;
+        char                    volume_id_str[64];
+        char                    *brickid = NULL;
+        dict_t                  *options = NULL;
+        struct rpc_clnt         *rpc = NULL;
+        glusterd_conf_t         *priv = THIS->private;
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (brickinfo);
+        GF_ASSERT (socketpath);
+
+        if (brickinfo->rpc == NULL) {
+                /* Setting frame-timeout to 10mins (600seconds).
+                 * Unix domain sockets ensures that the connection is reliable.
+                 * The default timeout of 30mins used for unreliable network
+                 * connections is too long for unix domain socket connections.
+                 */
+                ret = rpc_transport_unix_options_build (&options, socketpath,
+                                                        600);
+                if (ret)
+                        goto out;
+
+                uuid_utoa_r (volinfo->volume_id, volume_id_str);
+                ret = gf_asprintf (&brickid, "%s:%s:%s", volume_id_str,
+                                   brickinfo->hostname, brickinfo->path);
+                if (ret < 0)
+                        goto out;
+
+                synclock_unlock (&priv->big_lock);
+                ret = glusterd_rpc_create (&rpc, options,
+                                           glusterd_brick_rpc_notify,
+                                           brickid);
+                synclock_lock (&priv->big_lock);
+                if (ret)
+                        goto out;
+                brickinfo->rpc = rpc;
+        }
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+/* Caller should ensure that brick process is not running*/
+static void
+_reap_brick_process (char *pidfile, char *brickpath)
+{
+        unlink (pidfile);
+        /* Brick process is not running and pmap may have an entry for it.*/
+        pmap_registry_remove (THIS, 0, brickpath,
+                              GF_PMAP_PORT_BRICKSERVER, NULL);
+}
+
+static int
+_mk_rundir_p (glusterd_volinfo_t *volinfo)
+{
+        char voldir[PATH_MAX]   = {0,};
+        char rundir[PATH_MAX]   = {0,};
+        glusterd_conf_t *priv   = NULL;
+        xlator_t        *this   = NULL;
+        int             ret     = -1;
+
+        this = THIS;
+        priv = this->private;
+        GLUSTERD_GET_VOLUME_DIR (voldir, volinfo, priv);
+        snprintf (rundir, sizeof (rundir)-1, "%s/run", voldir);
+        ret =  mkdir_p (rundir, 0777, _gf_true);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "Failed to create rundir");
         return ret;
 }
 
 int32_t
 glusterd_volume_start_glusterfs (glusterd_volinfo_t  *volinfo,
                                  glusterd_brickinfo_t  *brickinfo,
-                                 int32_t count)
+                                 gf_boolean_t wait)
 {
         int32_t                 ret = -1;
         xlator_t                *this = NULL;
         glusterd_conf_t         *priv = NULL;
-        char                    pidfile[PATH_MAX] = {0,};
+        char                    pidfile[PATH_MAX+1] = {0,};
         char                    volfile[PATH_MAX] = {0,};
-        char                    path[PATH_MAX] = {0,};
-        char                    cmd_str[8192] = {0,};
-        char                    rundir[PATH_MAX] = {0,};
+        runner_t                runner = {0,};
         char                    exp_path[PATH_MAX] = {0,};
         char                    logfile[PATH_MAX] = {0,};
         int                     port = 0;
+        int                     rdma_port = 0;
+        char                    socketpath[PATH_MAX] = {0};
+        char                    glusterd_uuid[1024] = {0,};
+        char                    valgrind_logfile[PATH_MAX] = {0};
 
         GF_ASSERT (volinfo);
         GF_ASSERT (brickinfo);
@@ -852,61 +1593,149 @@ glusterd_volume_start_glusterfs (glusterd_volinfo_t  *volinfo,
         GF_ASSERT (this);
 
         priv = this->private;
+        GF_ASSERT (priv);
 
-        GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv);
-        snprintf (rundir, PATH_MAX, "%s/run", path);
-        ret = mkdir (rundir, 0777);
-
-        if ((ret == -1) && (EEXIST != errno)) {
-                gf_log ("", GF_LOG_ERROR, "Unable to create rundir %s",
-                        rundir);
+        if (brickinfo->snap_status == -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Snapshot is pending on %s:%s. "
+                        "Hence not starting the brick",
+                        brickinfo->hostname,
+                        brickinfo->path);
+                ret = 0;
                 goto out;
         }
 
+        ret = _mk_rundir_p (volinfo);
+        if (ret)
+                goto out;
+
+        glusterd_set_brick_socket_filepath (volinfo, brickinfo, socketpath,
+                                            sizeof (socketpath));
+
+        GLUSTERD_GET_BRICK_PIDFILE (pidfile, volinfo, brickinfo, priv);
+        if (glusterd_is_service_running (pidfile, NULL))
+                goto connect;
+
+        _reap_brick_process (pidfile, brickinfo->path);
+
         port = brickinfo->port;
         if (!port)
                 port = pmap_registry_alloc (THIS);
 
-        GLUSTERD_GET_BRICK_PIDFILE (pidfile, path, brickinfo->hostname,
-                                    brickinfo->path);
-
+        /* Build the exp_path, before starting the glusterfsd even in
+           valgrind mode. Otherwise all the glusterfsd processes start
+           writing the valgrind log to the same file.
+        */
         GLUSTERD_REMOVE_SLASH_FROM_PATH (brickinfo->path, exp_path);
-        snprintf (volfile, PATH_MAX, "%s.%s.%s", volinfo->volname,
-                  brickinfo->hostname, exp_path);
+        runinit (&runner);
+
+        if (priv->valgrind) {
+                /* Run bricks with valgrind */
+                if (volinfo->logdir) {
+                        snprintf (valgrind_logfile, PATH_MAX,
+                                  "%s/valgrind-%s-%s.log",
+                                  volinfo->logdir,
+                                  volinfo->volname, exp_path);
+                } else {
+                        snprintf (valgrind_logfile, PATH_MAX,
+                                  "%s/bricks/valgrind-%s-%s.log",
+                                  DEFAULT_LOG_FILE_DIRECTORY,
+                                  volinfo->volname, exp_path);
+                }
+
+                runner_add_args (&runner, "valgrind", "--leak-check=full",
+                                 "--trace-children=yes", "--track-origins=yes",
+                                 NULL);
+                runner_argprintf (&runner, "--log-file=%s", valgrind_logfile);
+        }
+
+        if (volinfo->is_snap_volume) {
+                snprintf (volfile, PATH_MAX,"/%s/%s/%s.%s.%s",
+                          GLUSTERD_VOL_SNAP_DIR_PREFIX,
+                          volinfo->snapshot->snapname, volinfo->volname,
+                          brickinfo->hostname, exp_path);
+        } else {
+                snprintf (volfile, PATH_MAX, "%s.%s.%s", volinfo->volname,
+                          brickinfo->hostname, exp_path);
+        }
 
-        if (!brickinfo->logfile) {
-                snprintf (logfile, PATH_MAX, "%s/logs/bricks/%s.log",
-                          priv->workdir, exp_path);
+        if (volinfo->logdir) {
+                snprintf (logfile, PATH_MAX, "%s/%s.log",
+                          volinfo->logdir, exp_path);
+        } else {
+                snprintf (logfile, PATH_MAX, "%s/bricks/%s.log",
+                          DEFAULT_LOG_FILE_DIRECTORY, exp_path);
+        }
+        if (!brickinfo->logfile)
                 brickinfo->logfile = gf_strdup (logfile);
+
+        (void) snprintf (glusterd_uuid, 1024, "*-posix.glusterd-uuid=%s",
+                         uuid_utoa (MY_UUID));
+        runner_add_args (&runner, SBIN_DIR"/glusterfsd",
+                         "-s", brickinfo->hostname, "--volfile-id", volfile,
+                         "-p", pidfile, "-S", socketpath,
+                         "--brick-name", brickinfo->path,
+                         "-l", brickinfo->logfile,
+                         "--xlator-option", glusterd_uuid,
+                         NULL);
+
+        runner_add_arg (&runner, "--brick-port");
+        if (volinfo->transport_type != GF_TRANSPORT_BOTH_TCP_RDMA) {
+                runner_argprintf (&runner, "%d", port);
+        } else {
+                rdma_port = brickinfo->rdma_port;
+                if (!rdma_port)
+                        rdma_port = pmap_registry_alloc (THIS);
+                runner_argprintf (&runner, "%d,%d", port, rdma_port);
+                runner_add_arg (&runner, "--xlator-option");
+                runner_argprintf (&runner, "%s-server.transport.rdma.listen-port=%d",
+                                  volinfo->volname, rdma_port);
         }
 
-        snprintf (cmd_str, 8192,
-                  "%s/sbin/glusterfs --xlator-option %s-server.listen-port=%d "
-                  "-s localhost --volfile-id %s -p %s --brick-name %s "
-                  "--brick-port %d -l %s", GFS_PREFIX, volinfo->volname,
-                  port, volfile, pidfile, brickinfo->path, port,
-                  brickinfo->logfile);
+        runner_add_arg (&runner, "--xlator-option");
+        runner_argprintf (&runner, "%s-server.listen-port=%d",
+                          volinfo->volname, port);
 
-	gf_log ("",GF_LOG_DEBUG,"Starting GlusterFS Command Executed: \n %s \n", cmd_str);
-        ret = gf_system (cmd_str);
+        if (volinfo->memory_accounting)
+                runner_add_arg (&runner, "--mem-accounting");
 
-        if (ret == 0) {
-                //pmap_registry_bind (THIS, port, brickinfo->path);
-                brickinfo->port = port;
+        runner_log (&runner, "", GF_LOG_DEBUG, "Starting GlusterFS");
+        if (wait) {
+                synclock_unlock (&priv->big_lock);
+                ret = runner_run (&runner);
+                synclock_lock (&priv->big_lock);
+
+        } else {
+                ret = runner_run_nowait (&runner);
+        }
+
+        if (ret)
+                goto out;
+
+        brickinfo->port = port;
+        brickinfo->rdma_port = rdma_port;
+
+connect:
+        ret = glusterd_brick_connect (volinfo, brickinfo, socketpath);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to connect to brick %s:%s on %s",
+                        brickinfo->hostname, brickinfo->path, socketpath);
+                goto out;
         }
 out:
         return ret;
 }
 
 int32_t
-glusterd_volume_stop_glusterfs (glusterd_volinfo_t  *volinfo,
-                                glusterd_brickinfo_t   *brickinfo,
-                                int32_t count)
+glusterd_brick_unlink_socket_file (glusterd_volinfo_t *volinfo,
+                                   glusterd_brickinfo_t *brickinfo)
 {
+        char                    path[PATH_MAX] = {0,};
+        char                    socketpath[PATH_MAX] = {0};
         xlator_t                *this = NULL;
         glusterd_conf_t         *priv = NULL;
-        char                    pidfile[PATH_MAX] = {0,};
-        char                    path[PATH_MAX] = {0,};
+        int                     ret = 0;
 
         GF_ASSERT (volinfo);
         GF_ASSERT (brickinfo);
@@ -915,12 +1744,79 @@ glusterd_volume_stop_glusterfs (glusterd_volinfo_t  *volinfo,
         GF_ASSERT (this);
 
         priv = this->private;
-
         GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv);
-        GLUSTERD_GET_BRICK_PIDFILE (pidfile, path, brickinfo->hostname,
-                                    brickinfo->path);
+        glusterd_set_brick_socket_filepath (volinfo, brickinfo, socketpath,
+                                            sizeof (socketpath));
+        ret = unlink (socketpath);
+        if (ret && (ENOENT == errno)) {
+                ret = 0;
+        } else {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to remove %s"
+                        " error: %s", socketpath, strerror (errno));
+        }
 
-        return glusterd_service_stop ("brick", pidfile, SIGTERM, _gf_false);
+        return ret;
+}
+
+int32_t
+glusterd_brick_disconnect (glusterd_brickinfo_t *brickinfo)
+{
+        rpc_clnt_t              *rpc = NULL;
+        glusterd_conf_t         *priv = THIS->private;
+
+        GF_ASSERT (brickinfo);
+
+        if (!brickinfo) {
+                gf_log_callingfn ("glusterd", GF_LOG_WARNING, "!brickinfo");
+                return -1;
+        }
+
+        rpc            = brickinfo->rpc;
+        brickinfo->rpc = NULL;
+
+        if (rpc) {
+                synclock_unlock (&priv->big_lock);
+                rpc_clnt_unref (rpc);
+                synclock_lock (&priv->big_lock);
+        }
+
+        return 0;
+}
+
+int32_t
+glusterd_volume_stop_glusterfs (glusterd_volinfo_t  *volinfo,
+                                glusterd_brickinfo_t   *brickinfo,
+                                gf_boolean_t del_brick)
+{
+        xlator_t        *this                   = NULL;
+        glusterd_conf_t *priv                   = NULL;
+        char            pidfile[PATH_MAX]       = {0,};
+        int             ret                     = 0;
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (brickinfo);
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        priv = this->private;
+        if (del_brick)
+                list_del_init (&brickinfo->brick_list);
+
+        if (GLUSTERD_STATUS_STARTED == volinfo->status) {
+                (void) glusterd_brick_disconnect (brickinfo);
+                GLUSTERD_GET_BRICK_PIDFILE (pidfile, volinfo, brickinfo, priv);
+                ret = glusterd_service_stop ("brick", pidfile, SIGTERM, _gf_false);
+                if (ret == 0) {
+                        glusterd_set_brick_status (brickinfo, GF_BRICK_STOPPED);
+                        (void) glusterd_brick_unlink_socket_file (volinfo, brickinfo);
+                }
+        }
+
+        if (del_brick)
+                glusterd_delete_brick (volinfo, brickinfo);
+
+        return ret;
 }
 
 int32_t
@@ -949,65 +1845,113 @@ out:
         return ret;
 }
 
-int32_t
-glusterd_peer_destroy (glusterd_peerinfo_t *peerinfo)
+/* Free LINE[0..N-1] and then the LINE buffer.  */
+static void
+free_lines (char **line, size_t n)
 {
-        int32_t                         ret = -1;
-        glusterd_peer_hostname_t        *name = NULL;
-        glusterd_peer_hostname_t        *tmp = NULL;
+  size_t i;
+  for (i = 0; i < n; i++)
+    GF_FREE (line[i]);
+  GF_FREE (line);
+}
 
-        if (!peerinfo)
+char **
+glusterd_readin_file (const char *filepath, int *line_count)
+{
+        int         ret                    = -1;
+        int         n                      = 8;
+        int         counter                = 0;
+        char        buffer[PATH_MAX + 256] = {0};
+        char      **lines                  = NULL;
+        FILE       *fp                     = NULL;
+        void       *p;
+
+        fp = fopen (filepath, "r");
+        if (!fp)
                 goto out;
 
-        ret = glusterd_store_delete_peerinfo (peerinfo);
+        lines = GF_CALLOC (1, n * sizeof (*lines), gf_gld_mt_charptr);
+        if (!lines)
+                goto out;
 
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Deleting peer info failed");
-        }
+        for (counter = 0; fgets (buffer, sizeof (buffer), fp); counter++) {
 
-        list_del_init (&peerinfo->uuid_list);
-        list_for_each_entry_safe (name, tmp, &peerinfo->hostnames,
-                                  hostname_list) {
-                list_del_init (&name->hostname_list);
-                GF_FREE (name->hostname);
+                if (counter == n-1) {
+                        n *= 2;
+                        p = GF_REALLOC (lines, n * sizeof (char *));
+                        if (!p) {
+                                free_lines (lines, n/2);
+                                lines = NULL;
+                                goto out;
+                        }
+                        lines = p;
+                }
+
+                lines[counter] = gf_strdup (buffer);
         }
 
-        list_del_init (&peerinfo->hostnames);
-        if (peerinfo->hostname)
-                GF_FREE (peerinfo->hostname);
-        GF_FREE (peerinfo);
-        peerinfo = NULL;
+        lines[counter] = NULL;
+        /* Reduce allocation to minimal size.  */
+        p = GF_REALLOC (lines, (counter + 1) * sizeof (char *));
+        if (!p) {
+                free_lines (lines, counter);
+                lines = NULL;
+                goto out;
+        }
+        lines = p;
 
+        *line_count = counter;
         ret = 0;
 
-out:
-        return ret;
+ out:
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", strerror (errno));
+        if (fp)
+                fclose (fp);
+
+        return lines;
 }
 
+int
+glusterd_compare_lines (const void *a, const void *b) {
 
-gf_boolean_t
-glusterd_is_cli_op_req (int32_t op)
+        return strcmp(* (char * const *) a, * (char * const *) b);
+}
+
+int
+glusterd_sort_and_redirect (const char *src_filepath, int dest_fd)
 {
-        switch (op) {
-                case GD_MGMT_CLI_CREATE_VOLUME:
-                case GD_MGMT_CLI_START_VOLUME:
-                case GD_MGMT_CLI_STOP_VOLUME:
-                case GD_MGMT_CLI_DELETE_VOLUME:
-                case GD_MGMT_CLI_DEFRAG_VOLUME:
-                case GD_MGMT_CLI_ADD_BRICK:
-                case GD_MGMT_CLI_REMOVE_BRICK:
-                case GD_MGMT_CLI_REPLACE_BRICK:
-                case GD_MGMT_CLI_LOG_FILENAME:
-                case GD_MGMT_CLI_LOG_LOCATE:
-                case GD_MGMT_CLI_LOG_ROTATE:
-                case GD_MGMT_CLI_SYNC_VOLUME:
-                        return _gf_true;
-                        break;
+        int            ret          = -1;
+        int            line_count   = 0;
+        int            counter      = 0;
+        char         **lines        = NULL;
+
+
+        if (!src_filepath || dest_fd < 0)
+                goto out;
+
+        lines = glusterd_readin_file (src_filepath, &line_count);
+        if (!lines)
+                goto out;
+
+        qsort (lines, line_count, sizeof (*lines), glusterd_compare_lines);
+
+        for (counter = 0; lines[counter]; counter++) {
+
+                ret = write (dest_fd, lines[counter],
+                             strlen (lines[counter]));
+                if (ret < 0)
+                        goto out;
+
+                GF_FREE (lines[counter]);
         }
 
-        return _gf_false;
-}
+        ret = 0;
+ out:
+        GF_FREE (lines);
 
+        return ret;
+}
 
 int
 glusterd_volume_compute_cksum (glusterd_volinfo_t  *volinfo)
@@ -1017,14 +1961,16 @@ glusterd_volume_compute_cksum (glusterd_volinfo_t  *volinfo)
         char                    path[PATH_MAX] = {0,};
         char                    cksum_path[PATH_MAX] = {0,};
         char                    filepath[PATH_MAX] = {0,};
-        DIR                     *dir = NULL;
-        struct dirent           *entry = NULL;
         int                     fd = -1;
         uint32_t                cksum = 0;
         char                    buf[4096] = {0,};
+        char                    sort_filepath[PATH_MAX] = {0};
+        gf_boolean_t            unlink_sortfile = _gf_false;
+        int                     sort_fd = 0;
+        xlator_t               *this = NULL;
 
         GF_ASSERT (volinfo);
-
+        this = THIS;
         priv = THIS->private;
         GF_ASSERT (priv);
 
@@ -1033,59 +1979,48 @@ glusterd_volume_compute_cksum (glusterd_volinfo_t  *volinfo)
         snprintf (cksum_path, sizeof (cksum_path), "%s/%s",
                   path, GLUSTERD_CKSUM_FILE);
 
-        fd = open (cksum_path, O_RDWR | O_APPEND | O_CREAT| O_TRUNC, 0644);
+        fd = open (cksum_path, O_RDWR | O_APPEND | O_CREAT| O_TRUNC, 0600);
 
         if (-1 == fd) {
-                gf_log ("", GF_LOG_ERROR, "Unable to open %s, errno: %d",
+                gf_log (this->name, GF_LOG_ERROR, "Unable to open %s, errno: %d",
                         cksum_path, errno);
                 ret = -1;
                 goto out;
         }
 
+        snprintf (filepath, sizeof (filepath), "%s/%s", path,
+                  GLUSTERD_VOLUME_INFO_FILE);
+        snprintf (sort_filepath, sizeof (sort_filepath),
+                  "/tmp/%s.XXXXXX", volinfo->volname);
+
+        sort_fd = mkstemp (sort_filepath);
+        if (sort_fd < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not generate temp "
+                        "file, reason: %s for %s: %s", strerror (errno),
+                        (volinfo->is_snap_volume)?"snap":"volume",
+                        volinfo->volname);
+                goto out;
+        } else {
+                unlink_sortfile = _gf_true;
+        }
 
-        dir = opendir (path);
-
-        glusterd_for_each_entry (entry, dir);
-
-/*        while (entry) {
-
-                snprintf (filepath, sizeof (filepath), "%s/%s", path,
-                          entry->d_name);
-
-                if (!strcmp (entry->d_name, "bricks") ||
-                     !strcmp (entry->d_name, "run")) {
-                        glusterd_for_each_entry (entry, dir);
-                        continue;
-                }
-
-                ret = get_checksum_for_path (filepath, &cksum);
-
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to get checksum"
-                                " for path: %s", filepath);
-                        goto out;
-                }
-
-                snprintf (buf, sizeof (buf), "%s=%u\n", entry->d_name, cksum);
-                ret = write (fd, buf, strlen (buf));
-
-                if (ret <= 0) {
-                        ret = -1;
-                        goto out;
-                }
+        /* sort the info file, result in sort_filepath */
 
-                glusterd_for_each_entry (entry, dir);
+        ret = glusterd_sort_and_redirect (filepath, sort_fd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "sorting info file failed");
+                goto out;
         }
-*/
 
-        snprintf (filepath, sizeof (filepath), "%s/%s", path,
-                  GLUSTERD_VOLUME_INFO_FILE);
+        ret = close (sort_fd);
+        if (ret)
+                goto out;
 
-        ret = get_checksum_for_path (filepath, &cksum);
+        ret = get_checksum_for_path (sort_filepath, &cksum);
 
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Unable to get checksum"
-                        " for path: %s", filepath);
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get checksum"
+                        " for path: %s", sort_filepath);
                 goto out;
         }
 
@@ -1102,31 +2037,89 @@ glusterd_volume_compute_cksum (glusterd_volinfo_t  *volinfo)
         if (ret)
                 goto out;
 
-       volinfo->cksum = cksum;
-
+        volinfo->cksum = cksum;
 out:
-       if (fd > 0)
+        if (fd > 0)
                close (fd);
+        if (unlink_sortfile)
+               unlink (sort_filepath);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
 
-       if (dir)
-               closedir (dir);
+        return ret;
+}
+
+int
+_add_dict_to_prdict (dict_t *this, char *key, data_t *value, void *data)
+{
+        glusterd_dict_ctx_t     *ctx = NULL;
+        char                    optkey[512] = {0,};
+        int                     ret = -1;
+
+        ctx = data;
+        snprintf (optkey, sizeof (optkey), "%s.%s%d", ctx->prefix,
+                  ctx->key_name, ctx->opt_count);
+        ret = dict_set_str (ctx->dict, optkey, key);
+        if (ret)
+                gf_log ("", GF_LOG_ERROR, "option add for %s%d %s",
+                        ctx->key_name, ctx->opt_count, key);
+        snprintf (optkey, sizeof (optkey), "%s.%s%d", ctx->prefix,
+                  ctx->val_name, ctx->opt_count);
+        ret = dict_set_str (ctx->dict, optkey, value->data);
+        if (ret)
+                gf_log ("", GF_LOG_ERROR, "option add for %s%d %s",
+                        ctx->val_name, ctx->opt_count, value->data);
+        ctx->opt_count++;
+
+        return ret;
+}
+
+int32_t
+glusterd_add_bricks_hname_path_to_dict (dict_t *dict,
+                                        glusterd_volinfo_t *volinfo)
+{
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        int                     ret = 0;
+        char                    key[256] = {0};
+        int                     index = 0;
+
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                snprintf (key, sizeof (key), "%d-hostname", index);
+                ret = dict_set_str (dict, key, brickinfo->hostname);
+                if (ret)
+                        goto out;
 
-       gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+                snprintf (key, sizeof (key), "%d-path", index);
+                ret = dict_set_str (dict, key, brickinfo->path);
+                if (ret)
+                        goto out;
 
-       return ret;
+                index++;
+        }
+out:
+        return ret;
 }
 
 int32_t
 glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo,
                              dict_t  *dict, int32_t count)
 {
-        int32_t                 ret = -1;
-        char                    key[512] = {0,};
-        glusterd_brickinfo_t    *brickinfo = NULL;
-        int32_t                 i = 1;
-        char                    uuid_str[50] = {0,};
-        char                    *volume_id_str = NULL;
+        int32_t                 ret               = -1;
+        char                    prefix[512]       = {0,};
+        char                    key[512]          = {0,};
+        glusterd_brickinfo_t    *brickinfo        = NULL;
+        int32_t                 i                 = 1;
+        char                    *volume_id_str    = NULL;
+        char                    *src_brick        = NULL;
+        char                    *dst_brick        = NULL;
+        char                    *str              = NULL;
+        glusterd_dict_ctx_t     ctx               = {0};
+        char                    *rebalance_id_str = NULL;
+        char                    *rb_id_str        = NULL;
+        xlator_t                *this             = NULL;
 
+        this = THIS;
+        GF_ASSERT (this);
         GF_ASSERT (dict);
         GF_ASSERT (volinfo);
 
@@ -1135,73 +2128,277 @@ glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo,
         if (ret)
                 goto out;
 
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.type", count);
         ret = dict_set_int32 (dict, key, volinfo->type);
         if (ret)
                 goto out;
 
-        memset (&key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.is_volume_restored", count);
+        ret = dict_set_int32 (dict, key, volinfo->is_volume_restored);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to set "
+                        "is_volume_restored option for %s volume",
+                        volinfo->volname);
+                goto out;
+        }
+
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.brick_count", count);
         ret = dict_set_int32 (dict, key, volinfo->brick_count);
         if (ret)
                 goto out;
 
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.version", count);
         ret = dict_set_int32 (dict, key, volinfo->version);
         if (ret)
                 goto out;
 
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.status", count);
         ret = dict_set_int32 (dict, key, volinfo->status);
         if (ret)
                 goto out;
 
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.sub_count", count);
         ret = dict_set_int32 (dict, key, volinfo->sub_count);
         if (ret)
                 goto out;
 
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.stripe_count", count);
+        ret = dict_set_int32 (dict, key, volinfo->stripe_count);
+        if (ret)
+                goto out;
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.replica_count", count);
+        ret = dict_set_int32 (dict, key, volinfo->replica_count);
+        if (ret)
+                goto out;
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.dist_count", count);
+        ret = dict_set_int32 (dict, key, volinfo->dist_leaf_count);
+        if (ret)
+                goto out;
+
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.ckusm", count);
         ret = dict_set_int64 (dict, key, volinfo->cksum);
         if (ret)
                 goto out;
 
-        uuid_unparse (volinfo->volume_id, uuid_str);
-        volume_id_str = gf_strdup (uuid_str);
-        if (!volume_id_str)
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.transport_type", count);
+        ret = dict_set_uint32 (dict, key, volinfo->transport_type);
+        if (ret)
                 goto out;
 
-        memset (&key, 0, sizeof (key));
-        snprintf (key, 256, "volume%d.volume_id", count);
+        snprintf (key, sizeof (key), "volume%d.is_snap_volume", count);
+        ret = dict_set_uint32 (dict, key, volinfo->is_snap_volume);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Unable to set %s", key);
+                goto out;
+        }
+
+        snprintf (key, sizeof (key), "volume%d.snap-max-hard-limit", count);
+        ret = dict_set_uint64 (dict, key, volinfo->snap_max_hard_limit);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Unable to set %s", key);
+                goto out;
+        }
+
+        volume_id_str = gf_strdup (uuid_utoa (volinfo->volume_id));
+        if (!volume_id_str) {
+                ret = -1;
+                goto out;
+        }
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.volume_id", count);
         ret = dict_set_dynstr (dict, key, volume_id_str);
         if (ret)
                 goto out;
+        volume_id_str = NULL;
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.username", count);
+        str = glusterd_auth_get_username (volinfo);
+        if (str) {
+                ret = dict_set_dynstr (dict, key, gf_strdup (str));
+                if (ret)
+                        goto out;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.password", count);
+        str = glusterd_auth_get_password (volinfo);
+        if (str) {
+                ret = dict_set_dynstr (dict, key, gf_strdup (str));
+                if (ret)
+                        goto out;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, 256, "volume%d.rebalance", count);
+        ret = dict_set_int32 (dict, key, volinfo->rebal.defrag_cmd);
+        if (ret)
+                goto out;
+
+        if (volinfo->rebal.defrag_cmd) {
+                rebalance_id_str = gf_strdup (uuid_utoa
+                                                (volinfo->rebal.rebalance_id));
+                if (!rebalance_id_str) {
+                        ret = -1;
+                        goto out;
+                }
+                memset (key, 0, sizeof (key));
+                snprintf (key, 256, "volume%d.rebalance-id", count);
+                ret = dict_set_dynstr (dict, key, rebalance_id_str);
+                if (ret)
+                        goto out;
+                rebalance_id_str = NULL;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.rebalance-op", count);
+        ret = dict_set_uint32 (dict, key, volinfo->rebal.op);
+        if (ret)
+                goto out;
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, 256, "volume%d."GLUSTERD_STORE_KEY_RB_STATUS, count);
+        ret = dict_set_int32 (dict, key, volinfo->rep_brick.rb_status);
+        if (ret)
+                goto out;
+
+        if (volinfo->rep_brick.rb_status > GF_RB_STATUS_NONE) {
+
+                memset (key, 0, sizeof (key));
+                snprintf (key, 256, "volume%d."GLUSTERD_STORE_KEY_RB_SRC_BRICK,
+                          count);
+                gf_asprintf (&src_brick, "%s:%s",
+                             volinfo->rep_brick.src_brick->hostname,
+                             volinfo->rep_brick.src_brick->path);
+                ret = dict_set_dynstr (dict, key, src_brick);
+                if (ret)
+                        goto out;
+
+                memset (key, 0, sizeof (key));
+                snprintf (key, 256, "volume%d."GLUSTERD_STORE_KEY_RB_DST_BRICK,
+                          count);
+                gf_asprintf (&dst_brick, "%s:%s",
+                             volinfo->rep_brick.dst_brick->hostname,
+                             volinfo->rep_brick.dst_brick->path);
+                ret = dict_set_dynstr (dict, key, dst_brick);
+                if (ret)
+                        goto out;
+
+                rb_id_str = gf_strdup (uuid_utoa (volinfo->rep_brick.rb_id));
+                if (!rb_id_str) {
+                        ret = -1;
+                        goto out;
+                }
+
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), "volume%d.rb_id", count);
+                ret = dict_set_dynstr (dict, key, rb_id_str);
+                if (ret)
+                        goto out;
+                rb_id_str = NULL;
+        }
+
+        snprintf (prefix, sizeof (prefix), "volume%d", count);
+        ctx.dict = dict;
+        ctx.prefix = prefix;
+        ctx.opt_count = 1;
+        ctx.key_name = "key";
+        ctx.val_name = "value";
+        GF_ASSERT (volinfo->dict);
+
+        dict_foreach (volinfo->dict, _add_dict_to_prdict, &ctx);
+        ctx.opt_count--;
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.opt-count", count);
+        ret = dict_set_int32 (dict, key, ctx.opt_count);
+        if (ret)
+                goto out;
+
+        ctx.dict = dict;
+        ctx.prefix = prefix;
+        ctx.opt_count = 1;
+        ctx.key_name = "slave-num";
+        ctx.val_name = "slave-val";
+        GF_ASSERT (volinfo->gsync_slaves);
+
+        dict_foreach (volinfo->gsync_slaves, _add_dict_to_prdict, &ctx);
+        ctx.opt_count--;
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.gsync-count", count);
+        ret = dict_set_int32 (dict, key, ctx.opt_count);
+        if (ret)
+                goto out;
 
         list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
-                memset (&key, 0, sizeof (key));
+                memset (key, 0, sizeof (key));
                 snprintf (key, sizeof (key), "volume%d.brick%d.hostname",
                           count, i);
                 ret = dict_set_str (dict, key, brickinfo->hostname);
                 if (ret)
                         goto out;
 
-                memset (&key, 0, sizeof (key));
+                memset (key, 0, sizeof (key));
                 snprintf (key, sizeof (key), "volume%d.brick%d.path",
                           count, i);
                 ret = dict_set_str (dict, key, brickinfo->path);
                 if (ret)
                         goto out;
 
+                snprintf (key, sizeof (key), "volume%d.brick%d.snap_status",
+                          count, i);
+                ret = dict_set_int32 (dict, key, brickinfo->snap_status);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to set snap_status for %s:%s",
+                                brickinfo->hostname,
+                                brickinfo->path);
+                        goto out;
+                }
+
+                snprintf (key, sizeof (key), "volume%d.brick%d.device_path",
+                          count, i);
+                ret = dict_set_str (dict, key, brickinfo->device_path);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to set snap_device for %s:%s",
+                                brickinfo->hostname,
+                                brickinfo->path);
+                        goto out;
+                }
+
                 i++;
         }
 
+        /* Add volume op-versions to dict. This prevents volume inconsistencies
+         * in the cluster
+         */
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.op-version", count);
+        ret = dict_set_int32 (dict, key, volinfo->op_version);
+        if (ret)
+                goto out;
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.client-op-version", count);
+        ret = dict_set_int32 (dict, key, volinfo->client_op_version);
 
 out:
+        GF_FREE (volume_id_str);
+        GF_FREE (rebalance_id_str);
+        GF_FREE (rb_id_str);
+
         gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
 
         return ret;
@@ -1215,6 +2412,7 @@ glusterd_build_volume_dict (dict_t **vols)
         glusterd_conf_t         *priv = NULL;
         glusterd_volinfo_t      *volinfo = NULL;
         int32_t                 count = 0;
+        glusterd_dict_ctx_t     ctx            = {0};
 
         priv = THIS->private;
 
@@ -1235,6 +2433,17 @@ glusterd_build_volume_dict (dict_t **vols)
         if (ret)
                 goto out;
 
+        ctx.dict = dict;
+        ctx.prefix = "global";
+        ctx.opt_count = 1;
+        ctx.key_name = "key";
+        ctx.val_name = "val";
+        dict_foreach (priv->opts, _add_dict_to_prdict, &ctx);
+        ctx.opt_count--;
+        ret = dict_set_int32 (dict, "global-opt-count", ctx.opt_count);
+        if (ret)
+                goto out;
+
         *vols = dict;
 out:
         gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
@@ -1245,7 +2454,8 @@ out:
 }
 
 int32_t
-glusterd_compare_friend_volume (dict_t *vols, int32_t count, int32_t *status)
+glusterd_compare_friend_volume (dict_t *vols, int32_t count, int32_t *status,
+                                char *hostname)
 {
 
         int32_t                 ret = -1;
@@ -1271,7 +2481,7 @@ glusterd_compare_friend_volume (dict_t *vols, int32_t count, int32_t *status)
                 goto out;
         }
 
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.version", count);
         ret = dict_get_int32 (vols, key, &version);
         if (ret)
@@ -1281,8 +2491,8 @@ glusterd_compare_friend_volume (dict_t *vols, int32_t count, int32_t *status)
                 //Mismatch detected
                 ret = 0;
                 gf_log ("", GF_LOG_ERROR, "Version of volume %s differ."
-                        "local version = %d, remote version = %d",
-                        volinfo->volname, volinfo->version, version);
+                        "local version = %d, remote version = %d on peer %s",
+                        volinfo->volname, volinfo->version, version, hostname);
                 *status = GLUSTERD_VOL_COMP_UPDATE_REQ;
                 goto out;
         } else if (version < volinfo->version) {
@@ -1292,7 +2502,7 @@ glusterd_compare_friend_volume (dict_t *vols, int32_t count, int32_t *status)
 
         //Now, versions are same, compare cksums.
         //
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.ckusm", count);
         ret = dict_get_uint32 (vols, key, &cksum);
         if (ret)
@@ -1301,8 +2511,8 @@ glusterd_compare_friend_volume (dict_t *vols, int32_t count, int32_t *status)
         if (cksum != volinfo->cksum) {
                 ret = 0;
                 gf_log ("", GF_LOG_ERROR, "Cksums of volume %s differ."
-                        " local cksum = %d, remote cksum = %d",
-                        volinfo->volname, volinfo->cksum, cksum);
+                        " local cksum = %u, remote cksum = %u on peer %s",
+                        volinfo->volname, volinfo->cksum, cksum, hostname);
                 *status = GLUSTERD_VOL_COMP_RJT;
                 goto out;
         }
@@ -1315,139 +2525,962 @@ out:
         return ret;
 }
 
-int32_t
-glusterd_import_friend_volume (dict_t *vols, int count)
+static int32_t
+import_prdict_dict (dict_t *vols, dict_t  *dst_dict, char *key_prefix,
+                    char *value_prefix, int opt_count, char *prefix)
+{
+        char                    key[512] = {0,};
+        int32_t                 ret = 0;
+        int                     i = 1;
+        char                    *opt_key = NULL;
+        char                    *opt_val = NULL;
+        char                    *dup_opt_val = NULL;
+        char                    msg[2048] = {0};
+
+        while (i <= opt_count) {
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), "%s.%s%d",
+                          prefix, key_prefix, i);
+                ret = dict_get_str (vols, key, &opt_key);
+                if (ret) {
+                        snprintf (msg, sizeof (msg), "Volume dict key not "
+                                  "specified");
+                        goto out;
+                }
+
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), "%s.%s%d",
+                          prefix, value_prefix, i);
+                ret = dict_get_str (vols, key, &opt_val);
+                if (ret) {
+                        snprintf (msg, sizeof (msg), "Volume dict value not "
+                                  "specified");
+                        goto out;
+                }
+                dup_opt_val = gf_strdup (opt_val);
+                if (!dup_opt_val) {
+                        ret = -1;
+                        goto out;
+                }
+                ret = dict_set_dynstr (dst_dict, opt_key, dup_opt_val);
+                if (ret) {
+                        snprintf (msg, sizeof (msg), "Volume set %s %s "
+                                  "unsuccessful", opt_key, dup_opt_val);
+                        goto out;
+                }
+                i++;
+        }
+
+out:
+        if (msg[0])
+                gf_log ("glusterd", GF_LOG_ERROR, "%s", msg);
+        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+
+}
+
+gf_boolean_t
+glusterd_is_quorum_option (char *option)
+{
+        gf_boolean_t    res = _gf_false;
+        int             i = 0;
+        char            *keys[] = {GLUSTERD_QUORUM_TYPE_KEY,
+                                   GLUSTERD_QUORUM_RATIO_KEY, NULL};
+
+        for (i = 0; keys[i]; i++) {
+                if (strcmp (option, keys[i]) == 0) {
+                        res = _gf_true;
+                        break;
+                }
+        }
+        return res;
+}
+
+gf_boolean_t
+glusterd_is_quorum_changed (dict_t *options, char *option, char *value)
+{
+        int             ret = 0;
+        gf_boolean_t    reconfigured = _gf_false;
+        gf_boolean_t    all = _gf_false;
+        char            *oldquorum = NULL;
+        char            *newquorum = NULL;
+        char            *oldratio = NULL;
+        char            *newratio = NULL;
+
+        if ((strcmp ("all", option) != 0) &&
+            !glusterd_is_quorum_option (option))
+                goto out;
+
+        if (strcmp ("all", option) == 0)
+                all = _gf_true;
+
+        if (all || (strcmp (GLUSTERD_QUORUM_TYPE_KEY, option) == 0)) {
+                newquorum = value;
+                ret = dict_get_str (options, GLUSTERD_QUORUM_TYPE_KEY,
+                                    &oldquorum);
+        }
+
+        if (all || (strcmp (GLUSTERD_QUORUM_RATIO_KEY, option) == 0)) {
+                newratio = value;
+                ret = dict_get_str (options, GLUSTERD_QUORUM_RATIO_KEY,
+                                    &oldratio);
+        }
+
+        reconfigured = _gf_true;
+
+        if (oldquorum && newquorum && (strcmp (oldquorum, newquorum) == 0))
+                reconfigured = _gf_false;
+        if (oldratio && newratio && (strcmp (oldratio, newratio) == 0))
+                reconfigured = _gf_false;
+
+        if ((oldratio == NULL) && (newratio == NULL) && (oldquorum == NULL) &&
+            (newquorum == NULL))
+                reconfigured = _gf_false;
+out:
+        return reconfigured;
+}
+
+static inline gf_boolean_t
+_is_contributing_to_quorum (gd_quorum_contrib_t contrib)
+{
+        if ((contrib == QUORUM_UP) || (contrib == QUORUM_DOWN))
+                return _gf_true;
+        return _gf_false;
+}
+
+static inline gf_boolean_t
+_does_quorum_meet (int active_count, int quorum_count)
+{
+        return (active_count >= quorum_count);
+}
+
+int
+glusterd_get_quorum_cluster_counts (xlator_t *this, int *active_count,
+                                    int *quorum_count)
+{
+        glusterd_peerinfo_t *peerinfo      = NULL;
+        glusterd_conf_t     *conf          = NULL;
+        int                 ret            = -1;
+        int                 inquorum_count = 0;
+        char                *val           = NULL;
+        double              quorum_percentage = 0.0;
+        gf_boolean_t        ratio          = _gf_false;
+        int                 count          = 0;
+
+        conf = this->private;
+        //Start with counting self
+        inquorum_count = 1;
+        if (active_count)
+                *active_count = 1;
+        list_for_each_entry (peerinfo, &conf->peers, uuid_list) {
+                if (peerinfo->quorum_contrib == QUORUM_WAITING)
+                        goto out;
+
+                if (_is_contributing_to_quorum (peerinfo->quorum_contrib))
+                        inquorum_count = inquorum_count + 1;
+
+                if (active_count && (peerinfo->quorum_contrib == QUORUM_UP))
+                        *active_count = *active_count + 1;
+        }
+
+        ret = dict_get_str (conf->opts, GLUSTERD_QUORUM_RATIO_KEY, &val);
+        if (ret == 0) {
+                ratio = _gf_true;
+                ret = gf_string2percent (val, &quorum_percentage);
+                if (!ret)
+                        ratio = _gf_true;
+        }
+        if (ratio)
+                count = CEILING_POS (inquorum_count *
+                                     quorum_percentage / 100.0);
+        else
+                count = (inquorum_count * 50 / 100) + 1;
+
+        *quorum_count = count;
+        ret = 0;
+out:
+        return ret;
+}
+
+gf_boolean_t
+glusterd_is_volume_in_server_quorum (glusterd_volinfo_t *volinfo)
+{
+        gf_boolean_t    res = _gf_false;
+        char            *quorum_type = NULL;
+        int             ret = 0;
+
+        ret = dict_get_str (volinfo->dict, GLUSTERD_QUORUM_TYPE_KEY,
+                            &quorum_type);
+        if (ret)
+                goto out;
+
+        if (strcmp (quorum_type, GLUSTERD_SERVER_QUORUM) == 0)
+                res = _gf_true;
+out:
+        return res;
+}
+
+gf_boolean_t
+glusterd_is_any_volume_in_server_quorum (xlator_t *this)
+{
+        glusterd_conf_t         *conf = NULL;
+        glusterd_volinfo_t      *volinfo = NULL;
+
+        conf = this->private;
+        list_for_each_entry (volinfo, &conf->volumes, vol_list) {
+                if (glusterd_is_volume_in_server_quorum (volinfo)) {
+                        return _gf_true;
+                }
+        }
+        return _gf_false;
+}
+
+gf_boolean_t
+does_gd_meet_server_quorum (xlator_t *this)
+{
+        int                     quorum_count = 0;
+        int                     active_count   = 0;
+        gf_boolean_t            in = _gf_false;
+        glusterd_conf_t         *conf = NULL;
+        int                     ret = -1;
+
+        conf = this->private;
+        ret = glusterd_get_quorum_cluster_counts (this, &active_count,
+                                                  &quorum_count);
+        if (ret)
+                goto out;
+
+        if (!_does_quorum_meet (active_count, quorum_count)) {
+                goto out;
+        }
+
+        in = _gf_true;
+out:
+        return in;
+}
+
+int
+glusterd_spawn_daemons (void *opaque)
 {
+        glusterd_conf_t *conf = THIS->private;
+        gf_boolean_t    start_bricks = !conf->restart_done;
+
+        if (start_bricks) {
+                glusterd_restart_bricks (conf);
+                conf->restart_done = _gf_true;
+        }
+        glusterd_restart_gsyncds (conf);
+        glusterd_restart_rebalance (conf);
+        return 0;
+}
+
+void
+glusterd_do_volume_quorum_action (xlator_t *this, glusterd_volinfo_t *volinfo,
+                                  gf_boolean_t meets_quorum)
+{
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        glusterd_conf_t         *conf      = NULL;
+
+        conf = this->private;
+        if (volinfo->status != GLUSTERD_STATUS_STARTED)
+                goto out;
+
+        if (!glusterd_is_volume_in_server_quorum (volinfo))
+                meets_quorum = _gf_true;
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                if (!glusterd_is_local_brick (this, volinfo, brickinfo))
+                        continue;
+                if (meets_quorum)
+                        glusterd_brick_start (volinfo, brickinfo, _gf_false);
+                else
+                        glusterd_brick_stop (volinfo, brickinfo, _gf_false);
+        }
+out:
+        return;
+}
+
+int
+glusterd_do_quorum_action ()
+{
+        xlator_t            *this          = NULL;
+        glusterd_conf_t     *conf          = NULL;
+        glusterd_volinfo_t  *volinfo       = NULL;
+        int                 ret            = 0;
+        int                 active_count   = 0;
+        int                 quorum_count   = 0;
+        gf_boolean_t        meets          = _gf_false;
+
+        this = THIS;
+        conf = this->private;
+
+        conf->pending_quorum_action = _gf_true;
+        ret = glusterd_lock (conf->uuid);
+        if (ret)
+                goto out;
+
+        {
+                ret = glusterd_get_quorum_cluster_counts (this, &active_count,
+                                                          &quorum_count);
+                if (ret)
+                        goto unlock;
 
+                if (_does_quorum_meet (active_count, quorum_count))
+                        meets = _gf_true;
+                list_for_each_entry (volinfo, &conf->volumes, vol_list) {
+                        glusterd_do_volume_quorum_action (this, volinfo, meets);
+                }
+        }
+unlock:
+        (void)glusterd_unlock (conf->uuid);
+        conf->pending_quorum_action = _gf_false;
+out:
+        return ret;
+}
+
+int32_t
+glusterd_import_friend_volume_opts (dict_t *vols, int count,
+                                    glusterd_volinfo_t *volinfo)
+{
+        char                    key[512] = {0,};
         int32_t                 ret = -1;
-        glusterd_conf_t         *priv = NULL;
+        int                     opt_count = 0;
+        char                    msg[2048] = {0};
+        char                    volume_prefix[1024] = {0};
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.opt-count", count);
+        ret = dict_get_int32 (vols, key, &opt_count);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Volume option count not "
+                          "specified for %s", volinfo->volname);
+                goto out;
+        }
+
+        snprintf (volume_prefix, sizeof (volume_prefix), "volume%d", count);
+        ret = import_prdict_dict (vols, volinfo->dict, "key", "value",
+                                  opt_count, volume_prefix);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Unable to import options dict "
+                          "specified for %s", volinfo->volname);
+                goto out;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.gsync-count", count);
+        ret = dict_get_int32 (vols, key, &opt_count);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Gsync count not "
+                          "specified for %s", volinfo->volname);
+                goto out;
+        }
+
+        ret = import_prdict_dict (vols, volinfo->gsync_slaves, "slave-num",
+                                  "slave-val", opt_count, volume_prefix);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Unable to import gsync sessions "
+                          "specified for %s", volinfo->volname);
+                goto out;
+        }
+
+out:
+        if (msg[0])
+                gf_log ("glusterd", GF_LOG_ERROR, "%s", msg);
+        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_import_new_brick (dict_t *vols, int32_t vol_count,
+                           int32_t brick_count,
+                           glusterd_brickinfo_t **brickinfo)
+{
         char                    key[512] = {0,};
-        glusterd_volinfo_t      *volinfo = NULL;
-        char                    *volname = NULL;
+        int                     ret = -1;
+        int32_t                 snap_status = 0;
+        char                    *snap_device = NULL;
         char                    *hostname = NULL;
         char                    *path = NULL;
-        glusterd_brickinfo_t    *brickinfo = NULL;
-        glusterd_brickinfo_t    *tmp = NULL;
-        int                     new_volinfo = 0;
-        int                     i = 1;
-        char                    *volume_id_str = NULL;
+        glusterd_brickinfo_t    *new_brickinfo = NULL;
+        char                    msg[2048] = {0};
 
         GF_ASSERT (vols);
+        GF_ASSERT (vol_count >= 0);
+        GF_ASSERT (brickinfo);
 
-        snprintf (key, sizeof (key), "volume%d.name", count);
-        ret = dict_get_str (vols, key, &volname);
-        if (ret)
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.brick%d.hostname",
+                  vol_count, brick_count);
+        ret = dict_get_str (vols, key, &hostname);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload", key);
                 goto out;
+        }
 
-        priv = THIS->private;
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.brick%d.path",
+                  vol_count, brick_count);
+        ret = dict_get_str (vols, key, &path);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload", key);
+                goto out;
+        }
 
-        ret = glusterd_volinfo_find (volname, &volinfo);
+        snprintf (key, sizeof (key), "volume%d.brick%d.snap_status",
+                  vol_count, brick_count);
+        ret = dict_get_int32 (vols, key, &snap_status);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload", key);
+                goto out;
+        }
 
+        snprintf (key, sizeof (key), "volume%d.brick%d.device_path",
+                  vol_count, brick_count);
+        ret = dict_get_str (vols, key, &snap_device);
         if (ret) {
-                ret = glusterd_volinfo_new (&volinfo);
+                snprintf (msg, sizeof (msg), "%s missing in payload", key);
+                goto out;
+        }
+
+        ret = glusterd_brickinfo_new (&new_brickinfo);
+        if (ret)
+                goto out;
+
+        strcpy (new_brickinfo->path, path);
+        strcpy (new_brickinfo->hostname, hostname);
+        strcpy (new_brickinfo->device_path, snap_device);
+        new_brickinfo->snap_status = snap_status;
+
+        //peerinfo might not be added yet
+        (void) glusterd_resolve_brick (new_brickinfo);
+        ret = 0;
+        *brickinfo = new_brickinfo;
+out:
+        if (msg[0])
+                gf_log ("glusterd", GF_LOG_ERROR, "%s", msg);
+        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_import_bricks (dict_t *vols, int32_t vol_count,
+                        glusterd_volinfo_t *new_volinfo)
+{
+        int                     ret = -1;
+        int                     brick_count = 1;
+        glusterd_brickinfo_t     *new_brickinfo = NULL;
+
+        GF_ASSERT (vols);
+        GF_ASSERT (vol_count >= 0);
+        GF_ASSERT (new_volinfo);
+        while (brick_count <= new_volinfo->brick_count) {
+
+                ret = glusterd_import_new_brick (vols, vol_count, brick_count,
+                                                 &new_brickinfo);
                 if (ret)
                         goto out;
-                strncpy (volinfo->volname, volname, sizeof (volinfo->volname));
-                new_volinfo = 1;
+                list_add_tail (&new_brickinfo->brick_list, &new_volinfo->bricks);
+                brick_count++;
         }
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
 
+int32_t
+glusterd_import_volinfo (dict_t *vols, int count,
+                         glusterd_volinfo_t **volinfo)
+{
+        int                ret               = -1;
+        char               key[256]          = {0};
+        char               *volname          = NULL;
+        glusterd_volinfo_t *new_volinfo      = NULL;
+        char               *volume_id_str    = NULL;
+        char               msg[2048]         = {0};
+        char               *src_brick        = NULL;
+        char               *dst_brick        = NULL;
+        char               *str              = NULL;
+        int                rb_status         = 0;
+        char               *rebalance_id_str = NULL;
+        char               *rb_id_str        = NULL;
+        int                op_version        = 0;
+        int                client_op_version = 0;
+        uint32_t           is_snap_volume    = 0;
 
-        memset (&key, 0, sizeof (key));
-        snprintf (key, sizeof (key), "volume%d.type", count);
-        ret = dict_get_int32 (vols, key, &volinfo->type);
+        GF_ASSERT (vols);
+        GF_ASSERT (volinfo);
+
+        snprintf (key, sizeof (key), "volume%d.name", count);
+        ret = dict_get_str (vols, key, &volname);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload", key);
+                goto out;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.is_snap_volume", count);
+        ret = dict_get_uint32 (vols, key, &is_snap_volume);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload for %s",
+                          key, volname);
+                goto out;
+        }
+
+        if (is_snap_volume == _gf_true) {
+                gf_log (THIS->name, GF_LOG_DEBUG,
+                        "Not syncing snap volume %s", volname);
+                ret = 0;
+                goto out;
+        }
+
+        ret = glusterd_volinfo_new (&new_volinfo);
         if (ret)
                 goto out;
+        strncpy (new_volinfo->volname, volname, sizeof (new_volinfo->volname));
+
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.type", count);
+        ret = dict_get_int32 (vols, key, &new_volinfo->type);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload for %s",
+                          key, volname);
+                goto out;
+        }
 
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.brick_count", count);
-        ret = dict_get_int32 (vols, key, &volinfo->brick_count);
-        if (ret)
+        ret = dict_get_int32 (vols, key, &new_volinfo->brick_count);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload for %s",
+                          key, volname);
                 goto out;
+        }
 
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.version", count);
-        ret = dict_get_int32 (vols, key, &volinfo->version);
-        if (ret)
+        ret = dict_get_int32 (vols, key, &new_volinfo->version);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload for %s",
+                          key, volname);
                 goto out;
+        }
 
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.status", count);
-        ret = dict_get_int32 (vols, key, (int32_t *)&volinfo->status);
-        if (ret)
+        ret = dict_get_int32 (vols, key, (int32_t *)&new_volinfo->status);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload for %s",
+                          key, volname);
                 goto out;
+        }
 
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.sub_count", count);
-        ret = dict_get_int32 (vols, key, &volinfo->sub_count);
-        if (ret)
+        ret = dict_get_int32 (vols, key, &new_volinfo->sub_count);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload for %s",
+                          key, volname);
                 goto out;
+        }
 
-        memset (&key, 0, sizeof (key));
-        snprintf (key, sizeof (key), "volume%d.ckusm", count);
-        ret = dict_get_uint32 (vols, key, &volinfo->cksum);
+        /* not having a 'stripe_count' key is not a error
+           (as peer may be of old version) */
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.stripe_count", count);
+        ret = dict_get_int32 (vols, key, &new_volinfo->stripe_count);
         if (ret)
+                gf_log (THIS->name, GF_LOG_INFO,
+                        "peer is possibly old version");
+
+        /* not having a 'replica_count' key is not a error
+           (as peer may be of old version) */
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.replica_count", count);
+        ret = dict_get_int32 (vols, key, &new_volinfo->replica_count);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_INFO,
+                        "peer is possibly old version");
+
+        /* not having a 'dist_count' key is not a error
+           (as peer may be of old version) */
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.dist_count", count);
+        ret = dict_get_int32 (vols, key, &new_volinfo->dist_leaf_count);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_INFO,
+                        "peer is possibly old version");
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.ckusm", count);
+        ret = dict_get_uint32 (vols, key, &new_volinfo->cksum);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload for %s",
+                          key, volname);
                 goto out;
+        }
 
-        memset (&key, 0, sizeof (key));
+        memset (key, 0, sizeof (key));
         snprintf (key, sizeof (key), "volume%d.volume_id", count);
         ret = dict_get_str (vols, key, &volume_id_str);
-        if (ret)
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload for %s",
+                          key, volname);
                 goto out;
-        uuid_parse (volume_id_str, volinfo->volume_id);
+        }
 
-        list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks,
-                                   brick_list) {
-                ret = glusterd_brickinfo_delete (brickinfo);
+        uuid_parse (volume_id_str, new_volinfo->volume_id);
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.username", count);
+        ret = dict_get_str (vols, key, &str);
+        if (!ret) {
+                ret = glusterd_auth_set_username (new_volinfo, str);
                 if (ret)
                         goto out;
         }
 
-        while (i <= volinfo->brick_count) {
-
-                memset (&key, 0, sizeof (key));
-                snprintf (key, sizeof (key), "volume%d.brick%d.hostname",
-                          count, i);
-                ret = dict_get_str (vols, key, &hostname);
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.password", count);
+        ret = dict_get_str (vols, key, &str);
+        if (!ret) {
+                ret = glusterd_auth_set_password (new_volinfo, str);
                 if (ret)
                         goto out;
+        }
 
-                memset (&key, 0, sizeof (key));
-                snprintf (key, sizeof (key), "volume%d.brick%d.path",
-                          count, i);
-                ret = dict_get_str (vols, key, &path);
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.transport_type", count);
+        ret = dict_get_uint32 (vols, key, &new_volinfo->transport_type);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload for %s",
+                          key, volname);
+                goto out;
+        }
+
+        new_volinfo->is_snap_volume = is_snap_volume;
+
+        snprintf (key, sizeof (key), "volume%d.is_volume_restored", count);
+        ret = dict_get_uint32 (vols, key, &new_volinfo->is_volume_restored);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to get "
+                        "is_volume_restored option for %s",
+                        volname);
+                goto out;
+        }
+
+        snprintf (key, sizeof (key), "volume%d.snap-max-hard-limit", count);
+        ret = dict_get_uint64 (vols, key, &new_volinfo->snap_max_hard_limit);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload for %s",
+                          key, volname);
+                goto out;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.rebalance", count);
+        ret = dict_get_uint32 (vols, key, &new_volinfo->rebal.defrag_cmd);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "%s missing in payload for %s",
+                          key, volname);
+                goto out;
+        }
+
+        if (new_volinfo->rebal.defrag_cmd) {
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), "volume%d.rebalance-id", count);
+                ret = dict_get_str (vols, key, &rebalance_id_str);
+                if (ret) {
+                        /* This is not present in older glusterfs versions,
+                         * so don't error out
+                         */
+                        ret = 0;
+                } else {
+                        uuid_parse (rebalance_id_str,
+                                    new_volinfo->rebal.rebalance_id);
+                }
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.rebalance-op", count);
+        ret = dict_get_uint32 (vols, key,(uint32_t *) &new_volinfo->rebal.op);
+        if (ret) {
+                /* This is not present in older glusterfs versions,
+                 * so don't error out
+                 */
+                ret = 0;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, 256, "volume%d."GLUSTERD_STORE_KEY_RB_STATUS, count);
+        ret = dict_get_int32 (vols, key, &rb_status);
+        if (ret)
+                goto out;
+        new_volinfo->rep_brick.rb_status = rb_status;
+
+        if (new_volinfo->rep_brick.rb_status > GF_RB_STATUS_NONE) {
+
+                memset (key, 0, sizeof (key));
+                snprintf (key, 256, "volume%d."GLUSTERD_STORE_KEY_RB_SRC_BRICK,
+                          count);
+                ret = dict_get_str (vols, key, &src_brick);
                 if (ret)
                         goto out;
 
-                ret = glusterd_brickinfo_new (&brickinfo);
-                if (ret)
+                ret = glusterd_brickinfo_new_from_brick (src_brick,
+                                        &new_volinfo->rep_brick.src_brick);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to create"
+                                " src brickinfo");
                         goto out;
+                }
 
-                strcpy (brickinfo->path, path);
-                strcpy (brickinfo->hostname, hostname);
+                memset (key, 0, sizeof (key));
+                snprintf (key, 256, "volume%d."GLUSTERD_STORE_KEY_RB_DST_BRICK,
+                          count);
+                ret = dict_get_str (vols, key, &dst_brick);
+                if (ret)
+                        goto out;
 
-                list_add_tail (&brickinfo->brick_list, &volinfo->bricks);
+                ret = glusterd_brickinfo_new_from_brick (dst_brick,
+                                     &new_volinfo->rep_brick.dst_brick);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to create"
+                                " dst brickinfo");
+                        goto out;
+                }
 
-                i++;
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), "volume%d.rb_id", count);
+                ret = dict_get_str (vols, key, &rb_id_str);
+                if (ret) {
+                        /* This is not present in older glusterfs versions,
+                         * so don't error out
+                         */
+                        ret = 0;
+                } else {
+                        uuid_parse (rb_id_str, new_volinfo->rep_brick.rb_id);
+                }
         }
 
-        if (new_volinfo) {
-                list_add_tail (&volinfo->vol_list, &priv->volumes);
-                ret = glusterd_store_create_volume (volinfo);
+
+        ret = glusterd_import_friend_volume_opts (vols, count, new_volinfo);
+        if (ret)
+                goto out;
+
+        /* Import the volume's op-versions if available else set it to 1.
+         * Not having op-versions implies this informtation was obtained from a
+         * op-version 1 friend (gluster-3.3), ergo the cluster is at op-version
+         * 1 and all volumes are at op-versions 1.
+         *
+         * Either both the volume op-versions should be absent or both should be
+         * present. Only one being present is a failure
+         */
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.op-version", count);
+        ret = dict_get_int32 (vols, key, &op_version);
+        if (ret)
+                ret = 0;
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "volume%d.client-op-version", count);
+        ret = dict_get_int32 (vols, key, &client_op_version);
+        if (ret)
+                ret = 0;
+
+        if (op_version && client_op_version) {
+                new_volinfo->op_version = op_version;
+                new_volinfo->client_op_version = client_op_version;
+        } else if (((op_version == 0) && (client_op_version != 0)) ||
+                   ((op_version != 0) && (client_op_version == 0))) {
+                ret = -1;
+                gf_log ("glusterd", GF_LOG_ERROR,
+                        "Only one volume op-version found");
+                goto out;
         } else {
-                ret = glusterd_store_update_volume (volinfo);
+                new_volinfo->op_version = 1;
+                new_volinfo->client_op_version = 1;
         }
 
-        ret = glusterd_create_volfiles (volinfo);
+        ret = glusterd_import_bricks (vols, count, new_volinfo);
         if (ret)
                 goto out;
+        *volinfo = new_volinfo;
+out:
+        if (msg[0])
+                gf_log ("glusterd", GF_LOG_ERROR, "%s", msg);
+        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_volume_disconnect_all_bricks (glusterd_volinfo_t *volinfo)
+{
+        int                  ret = 0;
+        glusterd_brickinfo_t *brickinfo = NULL;
+        GF_ASSERT (volinfo);
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                if (glusterd_is_brick_started (brickinfo)) {
+                        ret = glusterd_brick_disconnect (brickinfo);
+                        if (ret) {
+                                gf_log ("glusterd", GF_LOG_ERROR, "Failed to "
+                                        "disconnect %s:%s", brickinfo->hostname,
+                                        brickinfo->path);
+                                break;
+                        }
+                }
+        }
+
+        return ret;
+}
+
+int32_t
+glusterd_volinfo_copy_brick_portinfo (glusterd_volinfo_t *new_volinfo,
+                                      glusterd_volinfo_t *old_volinfo)
+{
+        glusterd_brickinfo_t *new_brickinfo = NULL;
+        glusterd_brickinfo_t *old_brickinfo = NULL;
+
+        int             ret = 0;
+        GF_ASSERT (new_volinfo);
+        GF_ASSERT (old_volinfo);
+        if (_gf_false == glusterd_is_volume_started (new_volinfo))
+                goto out;
+        list_for_each_entry (new_brickinfo, &new_volinfo->bricks, brick_list) {
+                ret = glusterd_volume_brickinfo_get (new_brickinfo->uuid,
+                                                     new_brickinfo->hostname,
+                                                     new_brickinfo->path,
+                                                     old_volinfo, &old_brickinfo);
+                if ((0 == ret) && glusterd_is_brick_started (old_brickinfo)) {
+                        new_brickinfo->port = old_brickinfo->port;
+                }
+        }
+out:
+        ret = 0;
+        return ret;
+}
+
+int32_t
+glusterd_volinfo_stop_stale_bricks (glusterd_volinfo_t *new_volinfo,
+                                    glusterd_volinfo_t *old_volinfo)
+{
+        glusterd_brickinfo_t *new_brickinfo = NULL;
+        glusterd_brickinfo_t *old_brickinfo = NULL;
+
+        int             ret = 0;
+        GF_ASSERT (new_volinfo);
+        GF_ASSERT (old_volinfo);
+        if (_gf_false == glusterd_is_volume_started (old_volinfo))
+                goto out;
+        list_for_each_entry (old_brickinfo, &old_volinfo->bricks, brick_list) {
+                ret = glusterd_volume_brickinfo_get (old_brickinfo->uuid,
+                                                     old_brickinfo->hostname,
+                                                     old_brickinfo->path,
+                                                     new_volinfo, &new_brickinfo);
+                if (ret) {
+                        /*TODO: may need to switch to 'atomic' flavour of
+                         * brick_stop, once we make peer rpc program also
+                         * synctask enabled*/
+                        ret = glusterd_brick_stop (old_volinfo, old_brickinfo,
+                                                   _gf_false);
+                        if (ret)
+                                gf_log ("glusterd", GF_LOG_ERROR, "Failed to "
+                                        "stop brick %s:%s", old_brickinfo->hostname,
+                                        old_brickinfo->path);
+                }
+        }
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
 
-        //volinfo->version++;
+int32_t
+glusterd_delete_stale_volume (glusterd_volinfo_t *stale_volinfo,
+                              glusterd_volinfo_t *valid_volinfo)
+{
+        GF_ASSERT (stale_volinfo);
+        GF_ASSERT (valid_volinfo);
+
+        /* If stale volume is in started state, copy the port numbers of the
+         * local bricks if they exist in the valid volume information.
+         * stop stale bricks. Stale volume information is going to be deleted.
+         * Which deletes the valid brick information inside stale volinfo.
+         * We dont want brick_rpc_notify to access already deleted brickinfo.
+         * Disconnect all bricks from stale_volinfo (unconditionally), since
+         * they are being deleted subsequently.
+         */
+        if (glusterd_is_volume_started (stale_volinfo)) {
+                if (glusterd_is_volume_started (valid_volinfo)) {
+                        (void) glusterd_volinfo_stop_stale_bricks (valid_volinfo,
+                                                                   stale_volinfo);
+                        //Only valid bricks will be running now.
+                        (void) glusterd_volinfo_copy_brick_portinfo (valid_volinfo,
+                                                                     stale_volinfo);
+
+                } else {
+                        (void) glusterd_stop_bricks (stale_volinfo);
+                }
+
+                (void) glusterd_volume_disconnect_all_bricks (stale_volinfo);
+        }
+        /* Delete all the bricks and stores and vol files. They will be created
+         * again by the valid_volinfo. Volume store delete should not be
+         * performed because some of the bricks could still be running,
+         * keeping pid files under run directory
+         */
+        (void) glusterd_delete_all_bricks (stale_volinfo);
+        if (stale_volinfo->shandle) {
+                unlink (stale_volinfo->shandle->path);
+                (void) gf_store_handle_destroy (stale_volinfo->shandle);
+                stale_volinfo->shandle = NULL;
+        }
+        (void) glusterd_volinfo_delete (stale_volinfo);
+        return 0;
+}
 
-        ret = glusterd_volume_compute_cksum (volinfo);
+int32_t
+glusterd_import_friend_volume (dict_t *vols, size_t count)
+{
+
+        int32_t                 ret = -1;
+        glusterd_conf_t         *priv = NULL;
+        xlator_t                *this = NULL;
+        glusterd_volinfo_t      *old_volinfo = NULL;
+        glusterd_volinfo_t      *new_volinfo = NULL;
+
+        GF_ASSERT (vols);
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+        ret = glusterd_import_volinfo (vols, count, &new_volinfo);
         if (ret)
                 goto out;
 
+        if (!new_volinfo) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Not importing snap volume");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (new_volinfo->volname, &old_volinfo);
+        if (0 == ret) {
+                (void) glusterd_delete_stale_volume (old_volinfo, new_volinfo);
+        }
 
+        if (glusterd_is_volume_started (new_volinfo)) {
+                (void) glusterd_start_bricks (new_volinfo);
+        }
+
+        ret = glusterd_store_volinfo (new_volinfo, GLUSTERD_VOLINFO_VER_AC_NONE);
+        ret = glusterd_create_volfiles_and_notify_services (new_volinfo);
+        if (ret)
+                goto out;
+
+        list_add_tail (&new_volinfo->vol_list, &priv->volumes);
 out:
         gf_log ("", GF_LOG_DEBUG, "Returning with ret: %d", ret);
         return ret;
@@ -1478,12 +3511,104 @@ out:
         return ret;
 }
 
+int
+glusterd_get_global_opt_version (dict_t *opts, uint32_t *version)
+{
+        int     ret = -1;
+        char    *version_str = NULL;
+
+        ret = dict_get_str (opts, GLUSTERD_GLOBAL_OPT_VERSION, &version_str);
+        if (ret)
+                goto out;
+
+        ret = gf_string2uint (version_str, version);
+        if (ret)
+                goto out;
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_get_next_global_opt_version_str (dict_t *opts, char **version_str)
+{
+        int             ret = -1;
+        char            version_string[64] = {0};
+        uint32_t        version = 0;
+
+        ret = glusterd_get_global_opt_version (opts, &version);
+        if (ret)
+                goto out;
+        version++;
+        snprintf (version_string, sizeof (version_string), "%"PRIu32, version);
+        *version_str = gf_strdup (version_string);
+        if (*version_str)
+                ret = 0;
+out:
+        return ret;
+}
+
 int32_t
-glusterd_compare_friend_data (dict_t  *vols, int32_t *status)
+glusterd_import_global_opts (dict_t *friend_data)
+{
+        xlator_t        *this = NULL;
+        glusterd_conf_t *conf = NULL;
+        int             ret = -1;
+        dict_t          *import_options = NULL;
+        int             count = 0;
+        uint32_t        local_version = 0;
+        uint32_t        remote_version = 0;
+
+        this = THIS;
+        conf = this->private;
+
+        ret = dict_get_int32 (friend_data, "global-opt-count", &count);
+        if (ret) {
+                //old version peer
+                ret = 0;
+                goto out;
+        }
+
+        import_options = dict_new ();
+        if (!import_options)
+                goto out;
+        ret = import_prdict_dict (friend_data, import_options, "key", "val",
+                                  count, "global");
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to import"
+                        " global options");
+                goto out;
+        }
+
+        ret = glusterd_get_global_opt_version (conf->opts, &local_version);
+        if (ret)
+                goto out;
+        ret = glusterd_get_global_opt_version (import_options, &remote_version);
+        if (ret)
+                goto out;
+        if (remote_version > local_version) {
+                ret = glusterd_store_options (this, import_options);
+                if (ret)
+                        goto out;
+                dict_unref (conf->opts);
+                conf->opts = dict_ref (import_options);
+        }
+        ret = 0;
+out:
+        if (import_options)
+                dict_unref (import_options);
+        return ret;
+}
+
+int32_t
+glusterd_compare_friend_data (dict_t  *vols, int32_t *status, char *hostname)
 {
         int32_t                 ret = -1;
         int32_t                 count = 0;
         int                     i = 1;
+        gf_boolean_t            update = _gf_false;
+        gf_boolean_t            stale_nfs = _gf_false;
+        gf_boolean_t            stale_shd = _gf_false;
 
         GF_ASSERT (vols);
         GF_ASSERT (status);
@@ -1493,7 +3618,8 @@ glusterd_compare_friend_data (dict_t  *vols, int32_t *status)
                 goto out;
 
         while (i <= count) {
-                ret = glusterd_compare_friend_volume (vols, i, status);
+                ret = glusterd_compare_friend_volume (vols, i, status,
+                                                      hostname);
                 if (ret)
                         goto out;
 
@@ -1501,14 +3627,31 @@ glusterd_compare_friend_data (dict_t  *vols, int32_t *status)
                         ret = 0;
                         goto out;
                 }
+                if (GLUSTERD_VOL_COMP_UPDATE_REQ == *status)
+                        update = _gf_true;
 
                 i++;
         }
 
-        if (GLUSTERD_VOL_COMP_UPDATE_REQ == *status) {
+        if (update) {
+                if (glusterd_is_nodesvc_running ("nfs"))
+                        stale_nfs = _gf_true;
+                if (glusterd_is_nodesvc_running ("glustershd"))
+                        stale_shd = _gf_true;
+                ret = glusterd_import_global_opts (vols);
+                if (ret)
+                        goto out;
                 ret = glusterd_import_friend_volumes (vols);
                 if (ret)
                         goto out;
+                if (_gf_false == glusterd_are_all_volumes_stopped ()) {
+                        ret = glusterd_nodesvcs_handle_graph_change (NULL);
+                } else {
+                        if (stale_nfs)
+                                glusterd_nfs_server_stop ();
+                        if (stale_shd)
+                                glusterd_shd_stop ();
+                }
         }
 
 out:
@@ -1518,78 +3661,279 @@ out:
         return ret;
 }
 
-int
-glusterd_file_copy (int out, int in)
+/* Valid only in if service is 'local' to glusterd.
+ * pid can be -1, if reading pidfile failed */
+gf_boolean_t
+glusterd_is_service_running (char *pidfile, int *pid)
 {
-        int     read_size = 0;
-        char    buffer[16 * 1024];
-        int     ret = -1;
+        FILE            *file = NULL;
+        gf_boolean_t    running = _gf_false;
+        int             ret = 0;
+        int             fno = 0;
+
+        file = fopen (pidfile, "r+");
+        if (!file)
+                goto out;
 
-        if (out <= 0 || in < 0) {
-                gf_log ("", GF_LOG_ERROR, "Invalid File descriptors");
+        fno = fileno (file);
+        ret = lockf (fno, F_TEST, 0);
+        if (ret == -1)
+                running = _gf_true;
+        if (!pid)
                 goto out;
+
+        ret = fscanf (file, "%d", pid);
+        if (ret <= 0) {
+                gf_log ("", GF_LOG_ERROR, "Unable to read pidfile: %s, %s",
+                        pidfile, strerror (errno));
+                *pid = -1;
         }
 
-        while (1) {
-                read_size = read(in, buffer, sizeof(buffer));
+out:
+        if (file)
+                fclose (file);
+        return running;
+}
 
-                if (read_size == 0) {
-                        ret = 0;
-                        break;              /* end of file */
-                }
+void
+glusterd_get_nodesvc_dir (char *server, char *workdir,
+                                char *path, size_t len)
+{
+        GF_ASSERT (len == PATH_MAX);
+        snprintf (path, len, "%s/%s", workdir, server);
+}
 
-                if (read_size < 0) {
-                        ret = -1;
-                        break; /*error reading file); */
-                }
-                write (out, buffer, (unsigned int) read_size);
+void
+glusterd_get_nodesvc_rundir (char *server, char *workdir,
+                                   char *path, size_t len)
+{
+        char    dir[PATH_MAX] = {0};
+        GF_ASSERT (len == PATH_MAX);
+
+        glusterd_get_nodesvc_dir (server, workdir, dir, sizeof (dir));
+        snprintf (path, len, "%s/run", dir);
+}
+
+void
+glusterd_get_nodesvc_pidfile (char *server, char *workdir,
+                                    char *path, size_t len)
+{
+        char    dir[PATH_MAX] = {0};
+        GF_ASSERT (len == PATH_MAX);
+
+        glusterd_get_nodesvc_rundir (server, workdir, dir, sizeof (dir));
+        snprintf (path, len, "%s/%s.pid", dir, server);
+}
+
+void
+glusterd_get_nodesvc_volfile (char *server, char *workdir,
+                                    char *volfile, size_t len)
+{
+        char  dir[PATH_MAX] = {0,};
+        GF_ASSERT (len == PATH_MAX);
+
+        glusterd_get_nodesvc_dir (server, workdir, dir, sizeof (dir));
+        snprintf (volfile, len, "%s/%s-server.vol", dir, server);
+}
+
+void
+glusterd_nodesvc_set_online_status (char *server, gf_boolean_t status)
+{
+        glusterd_conf_t *priv = NULL;
+
+        GF_ASSERT (server);
+        priv = THIS->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (priv->shd);
+        GF_ASSERT (priv->nfs);
+
+        if (!strcmp("glustershd", server))
+                priv->shd->online = status;
+        else if (!strcmp ("nfs", server))
+                priv->nfs->online = status;
+}
+
+gf_boolean_t
+glusterd_is_nodesvc_online (char *server)
+{
+        glusterd_conf_t *conf = NULL;
+        gf_boolean_t    online = _gf_false;
+
+        GF_ASSERT (server);
+        conf = THIS->private;
+        GF_ASSERT (conf);
+        GF_ASSERT (conf->shd);
+        GF_ASSERT (conf->nfs);
+
+        if (!strcmp (server, "glustershd"))
+                online = conf->shd->online;
+        else if (!strcmp (server, "nfs"))
+                online = conf->nfs->online;
+
+        return online;
+}
+
+int32_t
+glusterd_nodesvc_set_socket_filepath (char *rundir, uuid_t uuid,
+                                      char *socketpath, int len)
+{
+        char                    sockfilepath[PATH_MAX] = {0,};
+
+        snprintf (sockfilepath, sizeof (sockfilepath), "%s/run-%s",
+                  rundir, uuid_utoa (uuid));
+
+        glusterd_set_socket_filepath (sockfilepath, socketpath, len);
+        return 0;
+}
+
+struct rpc_clnt*
+glusterd_pending_node_get_rpc (glusterd_pending_node_t *pending_node)
+{
+        struct rpc_clnt *rpc = NULL;
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        nodesrv_t               *shd       = NULL;
+        glusterd_volinfo_t      *volinfo   = NULL;
+        nodesrv_t               *nfs       = NULL;
+
+        GF_VALIDATE_OR_GOTO (THIS->name, pending_node, out);
+        GF_VALIDATE_OR_GOTO (THIS->name, pending_node->node, out);
+
+        if (pending_node->type == GD_NODE_BRICK) {
+                brickinfo = pending_node->node;
+                rpc       = brickinfo->rpc;
+
+        } else if (pending_node->type == GD_NODE_SHD) {
+                shd       = pending_node->node;
+                rpc       = shd->rpc;
+
+        } else if (pending_node->type == GD_NODE_REBALANCE) {
+                volinfo = pending_node->node;
+                if (volinfo->rebal.defrag)
+                        rpc = volinfo->rebal.defrag->rpc;
+
+        } else if (pending_node->type == GD_NODE_NFS) {
+                nfs = pending_node->node;
+                rpc = nfs->rpc;
+
+        } else {
+                GF_ASSERT (0);
         }
+
 out:
-        return ret;
+        return rpc;
 }
 
-gf_boolean_t
-glusterd_is_nfs_started ()
+struct rpc_clnt*
+glusterd_nodesvc_get_rpc (char *server)
 {
-        int32_t                 ret = -1;
-        xlator_t                *this = NULL;
-        glusterd_conf_t         *priv = NULL;
-        char                    pidfile[PATH_MAX] = {0,};
+        glusterd_conf_t *priv   = NULL;
+        struct rpc_clnt *rpc    = NULL;
 
-        this = THIS;
-        GF_ASSERT(this);
+        GF_ASSERT (server);
+        priv = THIS->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (priv->shd);
+        GF_ASSERT (priv->nfs);
 
+        if (!strcmp (server, "glustershd"))
+                rpc = priv->shd->rpc;
+        else if (!strcmp (server, "nfs"))
+                rpc = priv->nfs->rpc;
+
+        return rpc;
+}
+
+int32_t
+glusterd_nodesvc_set_rpc (char *server, struct rpc_clnt *rpc)
+{
+        int             ret   = 0;
+        xlator_t        *this = NULL;
+        glusterd_conf_t *priv = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
         priv = this->private;
+        GF_ASSERT (priv);
+        GF_ASSERT (priv->shd);
+        GF_ASSERT (priv->nfs);
 
-        GLUSTERD_GET_NFS_PIDFILE(pidfile);
-        ret = access (pidfile, F_OK);
+        if (!strcmp ("glustershd", server))
+                priv->shd->rpc = rpc;
+        else if (!strcmp ("nfs", server))
+                priv->nfs->rpc = rpc;
 
-        if (ret == 0)
-                return _gf_true;
-        else
-                return _gf_false;
+        return ret;
 }
 
 int32_t
-glusterd_nfs_server_start ()
+glusterd_nodesvc_connect (char *server, char *socketpath) {
+        int                     ret = 0;
+        dict_t                  *options = NULL;
+        struct rpc_clnt         *rpc = NULL;
+        glusterd_conf_t         *priv = THIS->private;
+
+        rpc = glusterd_nodesvc_get_rpc (server);
+
+        if (rpc == NULL) {
+                /* Setting frame-timeout to 10mins (600seconds).
+                 * Unix domain sockets ensures that the connection is reliable.
+                 * The default timeout of 30mins used for unreliable network
+                 * connections is too long for unix domain socket connections.
+                 */
+                ret = rpc_transport_unix_options_build (&options, socketpath,
+                                                        600);
+                if (ret)
+                        goto out;
+                synclock_unlock (&priv->big_lock);
+                ret = glusterd_rpc_create (&rpc, options,
+                                           glusterd_nodesvc_rpc_notify,
+                                           server);
+                synclock_lock (&priv->big_lock);
+                if (ret)
+                        goto out;
+                (void) glusterd_nodesvc_set_rpc (server, rpc);
+        }
+out:
+        return ret;
+}
+
+int32_t
+glusterd_nodesvc_disconnect (char *server)
 {
-        int32_t                 ret = -1;
-        xlator_t                *this = NULL;
-        glusterd_conf_t         *priv = NULL;
-        char                    pidfile[PATH_MAX] = {0,};
-        char                    logfile[PATH_MAX] = {0,};
-        char                    *volfile = NULL;
-        char                    path[PATH_MAX] = {0,};
-        char                    cmd_str[8192] = {0,};
-        char                    rundir[PATH_MAX] = {0,};
+        struct rpc_clnt         *rpc = NULL;
+
+        rpc = glusterd_nodesvc_get_rpc (server);
+        (void)glusterd_nodesvc_set_rpc (server, NULL);
+
+        if (rpc)
+                rpc_clnt_unref (rpc);
+
+        return 0;
+}
+
+int32_t
+glusterd_nodesvc_start (char *server)
+{
+        int32_t                 ret                        = -1;
+        xlator_t               *this                       = NULL;
+        glusterd_conf_t        *priv                       = NULL;
+        runner_t                runner                     = {0,};
+        char                    pidfile[PATH_MAX]          = {0,};
+        char                    logfile[PATH_MAX]          = {0,};
+        char                    volfile[PATH_MAX]          = {0,};
+        char                    rundir[PATH_MAX]           = {0,};
+        char                    sockfpath[PATH_MAX] = {0,};
+        char                    volfileid[256]             = {0};
+        char                    glusterd_uuid_option[1024] = {0};
+        char                    valgrind_logfile[PATH_MAX] = {0};
 
         this = THIS;
         GF_ASSERT(this);
 
         priv = this->private;
 
-        GLUSTERD_GET_NFS_DIR(path, priv);
-        snprintf (rundir, PATH_MAX, "%s/run", path);
+        glusterd_get_nodesvc_rundir (server, priv->workdir,
+                                     rundir, sizeof (rundir));
         ret = mkdir (rundir, 0777);
 
         if ((ret == -1) && (EEXIST != errno)) {
@@ -1598,50 +3942,266 @@ glusterd_nfs_server_start ()
                 goto out;
         }
 
-        GLUSTERD_GET_NFS_PIDFILE(pidfile);
-        volfile = glusterd_get_nfs_filepath ();
-        if (!volfile) {
-                ret = -1;
-                goto out;
-        }
-
+        glusterd_get_nodesvc_pidfile (server, priv->workdir,
+                                      pidfile, sizeof (pidfile));
+        glusterd_get_nodesvc_volfile (server, priv->workdir,
+                                      volfile, sizeof (volfile));
         ret = access (volfile, F_OK);
         if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Nfs Volfile %s is not present",
-                        volfile);
+                gf_log ("", GF_LOG_ERROR, "%s Volfile %s is not present",
+                        server, volfile);
                 goto out;
         }
 
-        snprintf (logfile, PATH_MAX, "%s/logs/nfs.log", priv->workdir);
+        snprintf (logfile, PATH_MAX, "%s/%s.log", DEFAULT_LOG_FILE_DIRECTORY,
+                  server);
+        snprintf (volfileid, sizeof (volfileid), "gluster/%s", server);
 
-        snprintf (cmd_str, 8192,
-                  "%s/sbin/glusterfs -f %s -p %s -l %s",
-                  GFS_PREFIX, volfile, pidfile, logfile);
-        ret = gf_system (cmd_str);
+        glusterd_nodesvc_set_socket_filepath (rundir, MY_UUID,
+                                              sockfpath, sizeof (sockfpath));
 
+        runinit (&runner);
+
+        if (priv->valgrind) {
+                snprintf (valgrind_logfile, PATH_MAX,
+                          "%s/valgrind-%s.log",
+                          DEFAULT_LOG_FILE_DIRECTORY,
+                          server);
+
+                runner_add_args (&runner, "valgrind", "--leak-check=full",
+                                 "--trace-children=yes", "--track-origins=yes",
+                                 NULL);
+                runner_argprintf (&runner, "--log-file=%s", valgrind_logfile);
+        }
+
+        runner_add_args (&runner, SBIN_DIR"/glusterfs",
+                         "-s", "localhost",
+                         "--volfile-id", volfileid,
+                         "-p", pidfile,
+                         "-l", logfile,
+                         "-S", sockfpath, NULL);
+
+        if (!strcmp (server, "glustershd")) {
+                snprintf (glusterd_uuid_option, sizeof (glusterd_uuid_option),
+                          "*replicate*.node-uuid=%s", uuid_utoa (MY_UUID));
+                runner_add_args (&runner, "--xlator-option",
+                                 glusterd_uuid_option, NULL);
+        }
+        runner_log (&runner, "", GF_LOG_DEBUG,
+                    "Starting the nfs/glustershd services");
+
+        ret = runner_run_nowait (&runner);
+        if (ret == 0) {
+                glusterd_nodesvc_connect (server, sockfpath);
+        }
 out:
-        if (volfile)
-                GF_FREE(volfile);
+        return ret;
+}
+
+int
+glusterd_nfs_server_start ()
+{
+        return glusterd_nodesvc_start ("nfs");
+}
+
+int
+glusterd_shd_start ()
+{
+        return glusterd_nodesvc_start ("glustershd");
+}
+
+gf_boolean_t
+glusterd_is_nodesvc_running (char *server)
+{
+        char                    pidfile[PATH_MAX] = {0,};
+        glusterd_conf_t         *priv = THIS->private;
+
+        glusterd_get_nodesvc_pidfile (server, priv->workdir,
+                                            pidfile, sizeof (pidfile));
+        return glusterd_is_service_running (pidfile, NULL);
+}
+
+int32_t
+glusterd_nodesvc_unlink_socket_file (char *server)
+{
+        int             ret = 0;
+        char            sockfpath[PATH_MAX] = {0,};
+        char            rundir[PATH_MAX] = {0,};
+        glusterd_conf_t *priv = THIS->private;
+
+        glusterd_get_nodesvc_rundir (server, priv->workdir,
+                                     rundir, sizeof (rundir));
+
+        glusterd_nodesvc_set_socket_filepath (rundir, MY_UUID,
+                                              sockfpath, sizeof (sockfpath));
+
+        ret = unlink (sockfpath);
+        if (ret && (ENOENT == errno)) {
+                ret = 0;
+        } else {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to remove %s"
+                        " error: %s", sockfpath, strerror (errno));
+        }
+
         return ret;
 }
 
 int32_t
+glusterd_nodesvc_stop (char *server, int sig)
+{
+        char                    pidfile[PATH_MAX] = {0,};
+        glusterd_conf_t         *priv = THIS->private;
+        int                     ret = 0;
+
+        if (!glusterd_is_nodesvc_running (server))
+                goto out;
+
+        (void)glusterd_nodesvc_disconnect (server);
+
+        glusterd_get_nodesvc_pidfile (server, priv->workdir,
+                                            pidfile, sizeof (pidfile));
+        ret = glusterd_service_stop (server, pidfile, sig, _gf_true);
+
+        if (ret == 0) {
+                glusterd_nodesvc_set_online_status (server, _gf_false);
+                (void)glusterd_nodesvc_unlink_socket_file (server);
+        }
+out:
+        return ret;
+}
+
+void
+glusterd_nfs_pmap_deregister ()
+{
+        if (pmap_unset (MOUNT_PROGRAM, MOUNTV3_VERSION))
+                gf_log ("", GF_LOG_INFO, "De-registered MOUNTV3 successfully");
+        else
+                gf_log ("", GF_LOG_ERROR, "De-register MOUNTV3 is unsuccessful");
+
+        if (pmap_unset (MOUNT_PROGRAM, MOUNTV1_VERSION))
+                gf_log ("", GF_LOG_INFO, "De-registered MOUNTV1 successfully");
+        else
+                gf_log ("", GF_LOG_ERROR, "De-register MOUNTV1 is unsuccessful");
+
+        if (pmap_unset (NFS_PROGRAM, NFSV3_VERSION))
+                gf_log ("", GF_LOG_INFO, "De-registered NFSV3 successfully");
+        else
+                gf_log ("", GF_LOG_ERROR, "De-register NFSV3 is unsuccessful");
+
+        if (pmap_unset (NLM_PROGRAM, NLMV4_VERSION))
+                gf_log ("", GF_LOG_INFO, "De-registered NLM v4 successfully");
+        else
+                gf_log ("", GF_LOG_ERROR, "De-registration of NLM v4 failed");
+
+        if (pmap_unset (NLM_PROGRAM, NLMV1_VERSION))
+                gf_log ("", GF_LOG_INFO, "De-registered NLM v1 successfully");
+        else
+                gf_log ("", GF_LOG_ERROR, "De-registration of NLM v1 failed");
+
+        if (pmap_unset (ACL_PROGRAM, ACLV3_VERSION))
+                gf_log ("", GF_LOG_INFO, "De-registered ACL v3 successfully");
+        else
+                gf_log ("", GF_LOG_ERROR, "De-registration of ACL v3 failed");
+}
+
+int
 glusterd_nfs_server_stop ()
 {
-        xlator_t                *this = NULL;
-        glusterd_conf_t         *priv = NULL;
+        int                     ret = 0;
+        gf_boolean_t            deregister = _gf_false;
+
+        if (glusterd_is_nodesvc_running ("nfs"))
+                deregister = _gf_true;
+        ret = glusterd_nodesvc_stop ("nfs", SIGKILL);
+        if (ret)
+                goto out;
+        if (deregister)
+                glusterd_nfs_pmap_deregister ();
+out:
+        return ret;
+}
+
+int
+glusterd_shd_stop ()
+{
+        return glusterd_nodesvc_stop ("glustershd", SIGTERM);
+}
+
+int
+glusterd_add_node_to_dict (char *server, dict_t *dict, int count,
+                           dict_t *vol_opts)
+{
+        int                     ret = -1;
+        glusterd_conf_t         *priv = THIS->private;
         char                    pidfile[PATH_MAX] = {0,};
-        char                    path[PATH_MAX] = {0,};
+        gf_boolean_t            running = _gf_false;
+        int                     pid = -1;
+        int                     port = 0;
+        char                    key[1024] = {0,};
+
+        glusterd_get_nodesvc_pidfile (server, priv->workdir, pidfile,
+                                      sizeof (pidfile));
+        //Consider service to be running only when glusterd sees it Online
+        if (glusterd_is_nodesvc_online (server))
+                running = glusterd_is_service_running (pidfile, &pid);
+
+        /* For nfs-servers/self-heal-daemon setting
+         * brick<n>.hostname = "NFS Server" / "Self-heal Daemon"
+         * brick<n>.path = uuid
+         * brick<n>.port = 0
+         *
+         * This might be confusing, but cli displays the name of
+         * the brick as hostname+path, so this will make more sense
+         * when output.
+         */
+        snprintf (key, sizeof (key), "brick%d.hostname", count);
+        if (!strcmp (server, "nfs"))
+                ret = dict_set_str (dict, key, "NFS Server");
+        else if (!strcmp (server, "glustershd"))
+                ret = dict_set_str (dict, key, "Self-heal Daemon");
+        if (ret)
+                goto out;
 
-        this = THIS;
-        GF_ASSERT(this);
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "brick%d.path", count);
+        ret = dict_set_dynstr (dict, key, gf_strdup (uuid_utoa (MY_UUID)));
+        if (ret)
+                goto out;
 
-        priv = this->private;
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "brick%d.port", count);
+        /* Port is available only for the NFS server.
+         * Self-heal daemon doesn't provide any port for access
+         * by entities other than gluster.
+         */
+        if (!strcmp (server, "nfs")) {
+                if (dict_get (vol_opts, "nfs.port")) {
+                        ret = dict_get_int32 (vol_opts, "nfs.port", &port);
+                        if (ret)
+                                goto out;
+                } else
+                        port = GF_NFS3_PORT;
+        }
+        ret = dict_set_int32 (dict, key, port);
+        if (ret)
+                goto out;
 
-        GLUSTERD_GET_NFS_DIR(path, priv);
-        GLUSTERD_GET_NFS_PIDFILE(pidfile);
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "brick%d.pid", count);
+        ret = dict_set_int32 (dict, key, pid);
+        if (ret)
+                goto out;
 
-        return glusterd_service_stop ("nfsd", pidfile, SIGTERM, _gf_false);
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "brick%d.status", count);
+        ret = dict_set_int32 (dict, key, running);
+        if (ret)
+                goto out;
+
+
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
 }
 
 int
@@ -1652,49 +4212,251 @@ glusterd_remote_hostname_get (rpcsvc_request_t *req, char *remote_host, int len)
         GF_ASSERT (req->trans);
 
         char *name = NULL;
-        char *delimiter = NULL;
+        char *hostname = NULL;
+        char *tmp_host = NULL;
+        int  ret = 0;
 
         name = req->trans->peerinfo.identifier;
-        strncpy (remote_host, name, len);
-        delimiter = strchr (remote_host, ':');
+        tmp_host = gf_strdup (name);
+        if (tmp_host)
+                get_host_name (tmp_host, &hostname);
 
-        GF_ASSERT (delimiter);
-        if (!delimiter) {
+        GF_ASSERT (hostname);
+        if (!hostname) {
                 memset (remote_host, 0, len);
-                return -1;
+                ret = -1;
+                goto out;
         }
 
-        *delimiter = '\0';
+        strncpy (remote_host, hostname, strlen (hostname));
 
-        return 0;
+
+out:
+        GF_FREE (tmp_host);
+        return ret;
 }
 
 int
-glusterd_check_generate_start_nfs (glusterd_volinfo_t *volinfo)
+glusterd_check_generate_start_service (int (*create_volfile) (),
+                                       int (*stop) (), int (*start) ())
 {
         int ret = -1;
 
-        if (!volinfo) {
-                gf_log ("", GF_LOG_ERROR, "Invalid Arguments");
+        ret = create_volfile ();
+        if (ret)
+                goto out;
+
+        ret = stop ();
+        if (ret)
+                goto out;
+
+        ret = start ();
+out:
+        return ret;
+}
+
+int
+glusterd_reconfigure_nodesvc (int (*create_volfile) ())
+{
+        int ret = -1;
+
+        ret = create_volfile ();
+        if (ret)
+                goto out;
+
+        ret = glusterd_fetchspec_notify (THIS);
+out:
+        return ret;
+}
+
+int
+glusterd_reconfigure_shd ()
+{
+        int (*create_volfile) () = glusterd_create_shd_volfile;
+        return glusterd_reconfigure_nodesvc (create_volfile);
+}
+
+int
+glusterd_reconfigure_nfs ()
+{
+        int             ret             = -1;
+        gf_boolean_t    identical       = _gf_false;
+
+        /*
+         * Check both OLD and NEW volfiles, if they are SAME by size
+         * and cksum i.e. "character-by-character". If YES, then
+         * NOTHING has been changed, just return.
+         */
+        ret = glusterd_check_nfs_volfile_identical (&identical);
+        if (ret)
+                goto out;
+
+        if (identical) {
+                ret = 0;
                 goto out;
         }
 
-        ret = volgen_generate_nfs_volfile (volinfo);
+        /*
+         * They are not identical. Find out if the topology is changed
+         * OR just the volume options. If just the options which got
+         * changed, then inform the xlator to reconfigure the options.
+         */
+        identical = _gf_false; /* RESET the FLAG */
+        ret = glusterd_check_nfs_topology_identical (&identical);
         if (ret)
                 goto out;
 
-        if (glusterd_is_nfs_started ()) {
-                ret = glusterd_nfs_server_stop ();
-                if (ret)
-                        goto out;
+        /* Topology is not changed, but just the options. But write the
+         * options to NFS volfile, so that NFS will be reconfigured.
+         */
+        if (identical) {
+                ret = glusterd_create_nfs_volfile();
+                if (ret == 0) {/* Only if above PASSES */
+                        ret = glusterd_fetchspec_notify (THIS);
+                }
+                goto out;
         }
 
-        ret = glusterd_nfs_server_start ();
+        /*
+         * NFS volfile's topology has been changed. NFS server needs
+         * to be RESTARTED to ACT on the changed volfile.
+         */
+        ret = glusterd_check_generate_start_nfs ();
+
+out:
+        return ret;
+}
+
+
+int
+glusterd_check_generate_start_nfs ()
+{
+        int ret = 0;
+
+        ret = glusterd_check_generate_start_service (glusterd_create_nfs_volfile,
+                                                     glusterd_nfs_server_stop,
+                                                     glusterd_nfs_server_start);
+        return ret;
+}
+
+int
+glusterd_check_generate_start_shd ()
+{
+        int ret = 0;
+
+        ret = glusterd_check_generate_start_service (glusterd_create_shd_volfile,
+                                                     glusterd_shd_stop,
+                                                     glusterd_shd_start);
+        if (ret == -EINVAL)
+                ret = 0;
+        return ret;
+}
+
+int
+glusterd_nodesvcs_batch_op (glusterd_volinfo_t *volinfo,
+                             int (*nfs_op) (), int (*shd_op) ())
+{
+        int     ret = 0;
+
+        ret = nfs_op ();
+        if (ret)
+                goto out;
+
+        if (volinfo && !glusterd_is_volume_replicate (volinfo))
+                goto out;
+
+        ret = shd_op ();
+        if (ret)
+                goto out;
 out:
         return ret;
 }
 
 int
+glusterd_nodesvcs_start (glusterd_volinfo_t *volinfo)
+{
+        return glusterd_nodesvcs_batch_op (volinfo,
+                                           glusterd_nfs_server_start,
+                                           glusterd_shd_start);
+}
+
+int
+glusterd_nodesvcs_stop (glusterd_volinfo_t *volinfo)
+{
+        return glusterd_nodesvcs_batch_op (volinfo,
+                                            glusterd_nfs_server_stop,
+                                            glusterd_shd_stop);
+}
+
+gf_boolean_t
+glusterd_are_all_volumes_stopped ()
+{
+        glusterd_conf_t                         *priv = NULL;
+        xlator_t                                *this = NULL;
+        glusterd_volinfo_t                      *voliter = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        list_for_each_entry (voliter, &priv->volumes, vol_list) {
+                if (voliter->status == GLUSTERD_STATUS_STARTED)
+                        return _gf_false;
+        }
+
+        return _gf_true;
+
+}
+
+gf_boolean_t
+glusterd_all_replicate_volumes_stopped ()
+{
+        glusterd_conf_t                         *priv = NULL;
+        xlator_t                                *this = NULL;
+        glusterd_volinfo_t                      *voliter = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        list_for_each_entry (voliter, &priv->volumes, vol_list) {
+                if (!glusterd_is_volume_replicate (voliter))
+                        continue;
+                if (voliter->status == GLUSTERD_STATUS_STARTED)
+                        return _gf_false;
+        }
+
+        return _gf_true;
+}
+
+int
+glusterd_nodesvcs_handle_graph_change (glusterd_volinfo_t *volinfo)
+{
+        int (*shd_op) () = NULL;
+        int (*nfs_op) () = NULL;
+
+        shd_op = glusterd_check_generate_start_shd;
+        nfs_op = glusterd_check_generate_start_nfs;
+        if (glusterd_are_all_volumes_stopped ()) {
+                shd_op = glusterd_shd_stop;
+                nfs_op = glusterd_nfs_server_stop;
+        } else if (glusterd_all_replicate_volumes_stopped()) {
+                shd_op = glusterd_shd_stop;
+        }
+        return glusterd_nodesvcs_batch_op (volinfo, nfs_op, shd_op);
+}
+
+int
+glusterd_nodesvcs_handle_reconfigure (glusterd_volinfo_t *volinfo)
+{
+        return glusterd_nodesvcs_batch_op (volinfo,
+                                           glusterd_reconfigure_nfs,
+                                           glusterd_reconfigure_shd);
+}
+
+int
 glusterd_volume_count_get (void)
 {
         glusterd_volinfo_t      *tmp_volinfo = NULL;
@@ -1718,35 +4480,4415 @@ glusterd_volume_count_get (void)
 }
 
 int
-glusterd_is_exisiting_brick (char *hostname, char *path)
+glusterd_brickinfo_get (uuid_t uuid, char *hostname, char *path,
+                        glusterd_brickinfo_t **brickinfo)
 {
-        glusterd_brickinfo_t            *tmpbrkinfo = NULL;
         glusterd_volinfo_t              *volinfo     = NULL;
         glusterd_conf_t                 *priv = NULL;
         xlator_t                        *this = NULL;
-        int                             ret = 0;
+        int                             ret = -1;
 
-        GF_ASSERT (hostname);
         GF_ASSERT (path);
 
         this = THIS;
         GF_ASSERT (this);
 
         priv = this->private;
+
         list_for_each_entry (volinfo, &priv->volumes, vol_list) {
 
+                ret = glusterd_volume_brickinfo_get (uuid, hostname, path,
+                                                     volinfo, brickinfo);
+                if (ret == 0)
+                        /*Found*/
+                        goto out;
+        }
+out:
+        return ret;
+}
+
+int
+glusterd_brick_start (glusterd_volinfo_t *volinfo,
+                      glusterd_brickinfo_t *brickinfo,
+                      gf_boolean_t wait)
+{
+        int                                     ret   = -1;
+        xlator_t                                *this = NULL;
+        glusterd_conf_t                         *conf = NULL;
+
+        if ((!brickinfo) || (!volinfo))
+                goto out;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        if (uuid_is_null (brickinfo->uuid)) {
+                ret = glusterd_resolve_brick (brickinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK,
+                                brickinfo->hostname, brickinfo->path);
+                        goto out;
+                }
+        }
+
+        if (uuid_compare (brickinfo->uuid, MY_UUID)) {
+                ret = 0;
+                goto out;
+        }
+        ret = glusterd_volume_start_glusterfs (volinfo, brickinfo, wait);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to start brick %s:%s",
+                        brickinfo->hostname, brickinfo->path);
+                goto out;
+        }
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d ", ret);
+        return ret;
+}
+
+int
+glusterd_restart_bricks (glusterd_conf_t *conf)
+{
+        int                   ret            = 0;
+        glusterd_volinfo_t   *volinfo        = NULL;
+        glusterd_brickinfo_t *brickinfo      = NULL;
+        glusterd_snap_t      *snap           = NULL;
+        gf_boolean_t          start_nodesvcs = _gf_false;
+        xlator_t             *this           = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        list_for_each_entry (volinfo, &conf->volumes, vol_list) {
+                if (volinfo->status != GLUSTERD_STATUS_STARTED)
+                        continue;
+                start_nodesvcs = _gf_true;
+                gf_log (this->name, GF_LOG_DEBUG, "starting the volume %s",
+                        volinfo->volname);
+                list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                        glusterd_brick_start (volinfo, brickinfo, _gf_false);
+                }
+        }
+
+        list_for_each_entry (snap, &conf->snapshots, snap_list) {
+                list_for_each_entry (volinfo, &snap->volumes, vol_list) {
+                        if (volinfo->status != GLUSTERD_STATUS_STARTED)
+                                continue;
+                        start_nodesvcs = _gf_true;
+                        gf_log (this->name, GF_LOG_DEBUG, "starting the snap "
+                                "volume %s", volinfo->volname);
+                        list_for_each_entry (brickinfo, &volinfo->bricks,
+                                             brick_list) {
+                                glusterd_brick_start (volinfo, brickinfo,
+                                                      _gf_false);
+                        }
+                }
+        }
+
+        if (start_nodesvcs)
+                glusterd_nodesvcs_handle_graph_change (NULL);
+
+        return ret;
+}
+
+int
+_local_gsyncd_start (dict_t *this, char *key, data_t *value, void *data)
+{
+        char                        *path_list = NULL;
+        char                        *slave = NULL;
+        int                          uuid_len = 0;
+        int                          ret = 0;
+        char                         uuid_str[64] = {0};
+        glusterd_volinfo_t           *volinfo = NULL;
+        char                         *conf_path = NULL;
+
+        volinfo = data;
+        GF_ASSERT (volinfo);
+        slave = strchr(value->data, ':');
+        if (slave)
+                slave ++;
+        else
+                return 0;
+        uuid_len = (slave - value->data - 1);
+
+        strncpy (uuid_str, (char*)value->data, uuid_len);
+
+        ret = glusterd_get_local_brickpaths (volinfo, &path_list);
+
+        ret = dict_get_str (this, "conf_path", &conf_path);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Unable to fetch conf file path.");
+                goto out;
+        }
+
+        glusterd_start_gsync (volinfo, slave, path_list, conf_path,
+                              uuid_str, NULL);
+
+        GF_FREE (path_list);
+        path_list = NULL;
+
+out:
+        return ret;
+}
+
+int
+glusterd_volume_restart_gsyncds (glusterd_volinfo_t *volinfo)
+{
+        GF_ASSERT (volinfo);
+
+        dict_foreach (volinfo->gsync_slaves, _local_gsyncd_start, volinfo);
+        return 0;
+}
+
+int
+glusterd_restart_gsyncds (glusterd_conf_t *conf)
+{
+        glusterd_volinfo_t       *volinfo = NULL;
+        int                      ret = 0;
+
+        list_for_each_entry (volinfo, &conf->volumes, vol_list) {
+                glusterd_volume_restart_gsyncds (volinfo);
+        }
+        return ret;
+}
+
+inline int
+glusterd_get_dist_leaf_count (glusterd_volinfo_t *volinfo)
+{
+    int rcount = volinfo->replica_count;
+    int scount = volinfo->stripe_count;
+
+    return (rcount ? rcount : 1) * (scount ? scount : 1);
+}
+
+int
+glusterd_get_brickinfo (xlator_t *this, const char *brickname, int port,
+                        gf_boolean_t localhost, glusterd_brickinfo_t **brickinfo)
+{
+        glusterd_conf_t         *priv = NULL;
+        glusterd_volinfo_t      *volinfo = NULL;
+        glusterd_brickinfo_t    *tmpbrkinfo = NULL;
+        int                     ret = -1;
+
+        GF_ASSERT (brickname);
+        GF_ASSERT (this);
+
+        priv = this->private;
+        list_for_each_entry (volinfo, &priv->volumes, vol_list) {
                 list_for_each_entry (tmpbrkinfo, &volinfo->bricks,
                                      brick_list) {
+                        if (localhost && !gf_is_local_addr (tmpbrkinfo->hostname))
+                                continue;
+                        if (!strcmp(tmpbrkinfo->path, brickname) &&
+                            (tmpbrkinfo->port == port)) {
+                                *brickinfo = tmpbrkinfo;
+                                return 0;
+                        }
+                }
+        }
+        return ret;
+}
 
-                        if ((!strcmp(hostname, tmpbrkinfo-> hostname)
-                            && !strcmp(path, tmpbrkinfo->path))) {
-                                gf_log ("glusterd", GF_LOG_ERROR, "Brick %s:%s"
-                                        " already in use", hostname, path);
-                                ret = 1;
+glusterd_brickinfo_t*
+glusterd_get_brickinfo_by_position (glusterd_volinfo_t *volinfo, uint32_t pos)
+{
+        glusterd_brickinfo_t    *tmpbrkinfo = NULL;
+
+        list_for_each_entry (tmpbrkinfo, &volinfo->bricks,
+                             brick_list) {
+                if (pos == 0)
+                        return tmpbrkinfo;
+                pos--;
+        }
+        return NULL;
+}
+
+void
+glusterd_set_brick_status (glusterd_brickinfo_t  *brickinfo,
+                           gf_brick_status_t status)
+{
+        GF_ASSERT (brickinfo);
+        brickinfo->status = status;
+        if (GF_BRICK_STARTED == status) {
+                gf_log ("glusterd", GF_LOG_DEBUG, "Setting brick %s:%s status "
+                        "to started", brickinfo->hostname, brickinfo->path);
+        } else {
+                gf_log ("glusterd", GF_LOG_DEBUG, "Setting brick %s:%s status "
+                        "to stopped", brickinfo->hostname, brickinfo->path);
+        }
+}
+
+gf_boolean_t
+glusterd_is_brick_started (glusterd_brickinfo_t  *brickinfo)
+{
+        GF_ASSERT (brickinfo);
+        return (brickinfo->status == GF_BRICK_STARTED);
+}
+
+int
+glusterd_friend_brick_belongs (glusterd_volinfo_t *volinfo,
+                               glusterd_brickinfo_t *brickinfo, void* uuid)
+{
+        int             ret = -1;
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (brickinfo);
+        GF_ASSERT (uuid);
+
+        if (uuid_is_null (brickinfo->uuid)) {
+                ret = glusterd_resolve_brick (brickinfo);
+                if (ret) {
+                        GF_ASSERT (0);
+                        goto out;
+                }
+        }
+        if (!uuid_compare (brickinfo->uuid, *((uuid_t *)uuid)))
+                return 0;
+out:
+        return -1;
+}
+
+#ifdef GF_LINUX_HOST_OS
+int
+glusterd_get_brick_root (char *path, char **mount_point)
+{
+        char           *ptr            = NULL;
+        char           *mnt_pt         = NULL;
+        struct stat     brickstat      = {0};
+        struct stat     buf            = {0};
+
+        if (!path)
+                goto err;
+        mnt_pt = gf_strdup (path);
+        if (!mnt_pt)
+                goto err;
+        if (stat (mnt_pt, &brickstat))
+                goto err;
+
+        while ((ptr = strrchr (mnt_pt, '/')) &&
+               ptr != mnt_pt) {
+
+                *ptr = '\0';
+                if (stat (mnt_pt, &buf)) {
+                        gf_log (THIS->name, GF_LOG_ERROR, "error in "
+                                "stat: %s", strerror (errno));
+                        goto err;
+                }
+
+                if (brickstat.st_dev != buf.st_dev) {
+                        *ptr = '/';
+                        break;
+                }
+        }
+
+        if (ptr == mnt_pt) {
+                if (stat ("/", &buf)) {
+                        gf_log (THIS->name, GF_LOG_ERROR, "error in "
+                                "stat: %s", strerror (errno));
+                        goto err;
+                }
+                if (brickstat.st_dev == buf.st_dev)
+                        strcpy (mnt_pt, "/");
+        }
+
+        *mount_point = mnt_pt;
+        return 0;
+
+ err:
+        GF_FREE (mnt_pt);
+        return -1;
+}
+
+static char*
+glusterd_parse_inode_size (char *stream, char *pattern)
+{
+        char *needle = NULL;
+        char *trail  = NULL;
+
+        needle = strstr (stream, pattern);
+        if (!needle)
+                goto out;
+
+        needle = nwstrtail (needle, pattern);
+
+        trail = needle;
+        while (trail && isdigit (*trail)) trail++;
+        if (trail)
+                *trail = '\0';
+
+out:
+        return needle;
+}
+
+static int
+glusterd_add_inode_size_to_dict (dict_t *dict, int count)
+{
+        int             ret               = -1;
+        char            key[1024]         = {0};
+        char            buffer[4096]      = {0};
+        char           *inode_size        = NULL;
+        char           *device            = NULL;
+        char           *fs_name           = NULL;
+        char           *cur_word          = NULL;
+        char           *pattern           = NULL;
+        char           *trail             = NULL;
+        runner_t        runner            = {0, };
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "brick%d.device", count);
+        ret = dict_get_str (dict, key, &device);
+        if (ret)
+                goto out;
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "brick%d.fs_name", count);
+        ret = dict_get_str (dict, key, &fs_name);
+        if (ret)
+                goto out;
+
+        runinit (&runner);
+        runner_redir (&runner, STDOUT_FILENO, RUN_PIPE);
+        /* get inode size for xfs or ext2/3/4 */
+        if (!strcmp (fs_name, "xfs")) {
+
+                runner_add_args (&runner, "xfs_info", device, NULL);
+                pattern = "isize=";
+
+        } else if (IS_EXT_FS(fs_name)) {
+
+                runner_add_args (&runner, "tune2fs", "-l", device, NULL);
+                pattern = "Inode size:";
+
+        } else {
+                ret = 0;
+                gf_log (THIS->name, GF_LOG_INFO, "Skipped fetching "
+                        "inode size for %s: FS type not recommended",
+                        fs_name);
+                goto out;
+        }
+
+        ret = runner_start (&runner);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "could not get inode "
+                        "size for %s : %s package missing", fs_name,
+                        ((strcmp (fs_name, "xfs")) ?
+                         "e2fsprogs" : "xfsprogs"));
+                goto out;
+        }
+
+        for (;;) {
+                if (fgets (buffer, sizeof (buffer),
+                    runner_chio (&runner, STDOUT_FILENO)) == NULL)
+                        break;
+                trail = strrchr (buffer, '\n');
+                if (trail)
+                        *trail = '\0';
+
+                cur_word = glusterd_parse_inode_size (buffer, pattern);
+                if (cur_word)
+                        break;
+        }
+
+        ret = runner_end (&runner);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "%s exited with non-zero "
+                        "exit status", ((!strcmp (fs_name, "xfs")) ?
+                        "xfs_info" : "tune2fs"));
+                goto out;
+        }
+        if (!cur_word) {
+                ret = -1;
+                gf_log (THIS->name, GF_LOG_ERROR, "Unable to retrieve inode "
+                        "size using %s",
+                        (!strcmp (fs_name, "xfs")? "xfs_info": "tune2fs"));
+                goto out;
+        }
+
+        inode_size = gf_strdup (cur_word);
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "brick%d.inode_size", count);
+
+        ret = dict_set_dynstr (dict, key, inode_size);
+
+ out:
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR, "failed to get inode size");
+        return ret;
+}
+
+struct mntent *
+glusterd_get_mnt_entry_info (char *mnt_pt, FILE *mtab)
+{
+        struct mntent  *entry                = NULL;
+
+        mtab = setmntent (_PATH_MOUNTED, "r");
+        if (!mtab)
+                goto out;
+
+        entry = getmntent (mtab);
+
+        while (1) {
+                if (!entry)
+                        goto out;
+
+                if (!strcmp (entry->mnt_dir, mnt_pt) &&
+                    strcmp (entry->mnt_type, "rootfs"))
+                        break;
+                entry = getmntent (mtab);
+        }
+
+out:
+        return entry;
+}
+
+static int
+glusterd_add_brick_mount_details (glusterd_brickinfo_t *brickinfo,
+                                  dict_t *dict, int count)
+{
+        int             ret                  = -1;
+        char            key[1024]            = {0};
+        char            base_key[1024]       = {0};
+        char           *mnt_pt               = NULL;
+        char           *fs_name              = NULL;
+        char           *mnt_options          = NULL;
+        char           *device               = NULL;
+        struct mntent  *entry                = NULL;
+        FILE           *mtab                 = NULL;
+
+        snprintf (base_key, sizeof (base_key), "brick%d", count);
+
+        ret = glusterd_get_brick_root (brickinfo->path, &mnt_pt);
+        if (ret)
+                goto out;
+
+        entry = glusterd_get_mnt_entry_info (mnt_pt, mtab);
+        if (!entry) {
+                ret = -1;
+                goto out;
+        }
+
+        /* get device file */
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%s.device", base_key);
+
+        device = gf_strdup (entry->mnt_fsname);
+        ret = dict_set_dynstr (dict, key, device);
+        if (ret)
+                goto out;
+
+        /* fs type */
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%s.fs_name", base_key);
+
+        fs_name = gf_strdup (entry->mnt_type);
+        ret = dict_set_dynstr (dict, key, fs_name);
+        if (ret)
+                goto out;
+
+        /* mount options */
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%s.mnt_options", base_key);
+
+        mnt_options = gf_strdup (entry->mnt_opts);
+        ret = dict_set_dynstr (dict, key, mnt_options);
+
+ out:
+        GF_FREE (mnt_pt);
+        if (mtab)
+                endmntent (mtab);
+
+        return ret;
+}
+
+char*
+glusterd_get_brick_mount_details (glusterd_brickinfo_t *brickinfo)
+{
+        int             ret                  = -1;
+        char           *mnt_pt               = NULL;
+        char           *device               = NULL;
+        FILE           *mtab                 = NULL;
+        struct mntent  *entry                = NULL;
+        xlator_t       *this                 = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (brickinfo);
+
+        ret = glusterd_get_brick_root (brickinfo->path, &mnt_pt);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get mount point "
+                        "for %s brick", brickinfo->path);
+                goto out;
+        }
+
+        entry = glusterd_get_mnt_entry_info (mnt_pt, mtab);
+        if (NULL == entry) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get mnt entry "
+                        "for %s mount path", mnt_pt);
+                goto out;
+        }
+
+        /* get the fs_name/device */
+        device = gf_strdup (entry->mnt_fsname);
+
+out:
+        if (NULL != mtab) {
+                endmntent (mtab);
+        }
+
+        return device;
+}
+#endif
+
+int
+glusterd_add_brick_detail_to_dict (glusterd_volinfo_t *volinfo,
+                                   glusterd_brickinfo_t *brickinfo,
+                                   dict_t *dict, int count)
+{
+        int             ret               = -1;
+        uint64_t        memtotal          = 0;
+        uint64_t        memfree           = 0;
+        uint64_t        inodes_total      = 0;
+        uint64_t        inodes_free       = 0;
+        uint64_t        block_size        = 0;
+        char            key[1024]         = {0};
+        char            base_key[1024]    = {0};
+        struct statvfs  brickstat         = {0};
+        xlator_t       *this              = NULL;
+
+        this = THIS;
+        GF_ASSERT (volinfo);
+        GF_ASSERT (brickinfo);
+        GF_ASSERT (dict);
+
+        snprintf (base_key, sizeof (base_key), "brick%d", count);
+
+        ret = statvfs (brickinfo->path, &brickstat);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "statfs error: %s ",
+                        strerror (errno));
+                goto out;
+        }
+
+        /* file system block size */
+        block_size = brickstat.f_bsize;
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%s.block_size", base_key);
+        ret = dict_set_uint64 (dict, key, block_size);
+        if (ret)
+                goto out;
+
+        /* free space in brick */
+        memfree = brickstat.f_bfree * brickstat.f_bsize;
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%s.free", base_key);
+        ret = dict_set_uint64 (dict, key, memfree);
+        if (ret)
+                goto out;
+
+        /* total space of brick */
+        memtotal = brickstat.f_blocks * brickstat.f_bsize;
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%s.total", base_key);
+        ret = dict_set_uint64 (dict, key, memtotal);
+        if (ret)
+                goto out;
+
+        /* inodes: total and free counts only for ext2/3/4 and xfs */
+        inodes_total = brickstat.f_files;
+        if (inodes_total) {
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), "%s.total_inodes", base_key);
+                ret = dict_set_uint64 (dict, key, inodes_total);
+                if (ret)
+                        goto out;
+        }
+
+        inodes_free = brickstat.f_ffree;
+        if (inodes_free) {
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), "%s.free_inodes", base_key);
+                ret = dict_set_uint64 (dict, key, inodes_free);
+                if (ret)
+                        goto out;
+        }
+#ifdef GF_LINUX_HOST_OS
+        ret = glusterd_add_brick_mount_details (brickinfo, dict, count);
+        if (ret)
+                goto out;
+
+        ret = glusterd_add_inode_size_to_dict (dict, count);
+#endif
+ out:
+        if (ret)
+                gf_log (this->name, GF_LOG_DEBUG, "Error adding brick"
+                        " detail to dict: %s", strerror (errno));
+        return ret;
+}
+
+int32_t
+glusterd_add_brick_to_dict (glusterd_volinfo_t *volinfo,
+                            glusterd_brickinfo_t *brickinfo,
+                            dict_t  *dict, int32_t count)
+{
+
+        int             ret                   = -1;
+        int32_t         pid                   = -1;
+        int32_t         brick_online          = -1;
+        char            key[1024]             = {0};
+        char            base_key[1024]        = {0};
+        char            pidfile[PATH_MAX]     = {0};
+        xlator_t        *this                 = NULL;
+        glusterd_conf_t *priv                 = NULL;
+
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (brickinfo);
+        GF_ASSERT (dict);
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        priv = this->private;
+
+        snprintf (base_key, sizeof (base_key), "brick%d", count);
+        snprintf (key, sizeof (key), "%s.hostname", base_key);
+
+        ret = dict_set_str (dict, key, brickinfo->hostname);
+        if (ret)
+                goto out;
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%s.path", base_key);
+        ret = dict_set_str (dict, key, brickinfo->path);
+        if (ret)
+                goto out;
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%s.port", base_key);
+        ret = dict_set_int32 (dict, key, brickinfo->port);
+        if (ret)
+                goto out;
+
+        GLUSTERD_GET_BRICK_PIDFILE (pidfile, volinfo, brickinfo, priv);
+
+        brick_online = glusterd_is_service_running (pidfile, &pid);
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%s.pid", base_key);
+        ret = dict_set_int32 (dict, key, pid);
+        if (ret)
+                goto out;
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%s.status", base_key);
+        ret = dict_set_int32 (dict, key, brick_online);
+
+out:
+        if (ret)
+                gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+int32_t
+glusterd_get_all_volnames (dict_t *dict)
+{
+        int                    ret        = -1;
+        int32_t                vol_count  = 0;
+        char                   key[256]   = {0};
+        glusterd_volinfo_t    *entry      = NULL;
+        glusterd_conf_t       *priv       = NULL;
+
+        priv = THIS->private;
+        GF_ASSERT (priv);
+
+        list_for_each_entry (entry, &priv->volumes, vol_list) {
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), "vol%d", vol_count);
+                ret = dict_set_str (dict, key, entry->volname);
+                if (ret)
+                        goto out;
+
+                vol_count++;
+        }
+
+        ret = dict_set_int32 (dict, "vol_count", vol_count);
+
+ out:
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR, "failed to get all "
+                        "volume names for status");
+        return ret;
+}
+
+int
+glusterd_all_volume_cond_check (glusterd_condition_func func, int status,
+                                void *ctx)
+{
+        glusterd_conf_t         *priv = NULL;
+        glusterd_volinfo_t      *volinfo = NULL;
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        int                     ret = -1;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+        priv = this->private;
+
+        list_for_each_entry (volinfo, &priv->volumes, vol_list) {
+                list_for_each_entry (brickinfo, &volinfo->bricks,
+                                     brick_list) {
+                        ret = func (volinfo, brickinfo, ctx);
+                        if (ret != status) {
+                                ret = -1;
+                                goto out;
+                        }
+                }
+        }
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_friend_find_by_uuid (uuid_t uuid,
+                              glusterd_peerinfo_t  **peerinfo)
+{
+        int                     ret = -1;
+        glusterd_conf_t         *priv = NULL;
+        glusterd_peerinfo_t     *entry = NULL;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (peerinfo);
+
+        *peerinfo = NULL;
+        priv    = this->private;
+
+        GF_ASSERT (priv);
+
+        if (uuid_is_null (uuid))
+                return -1;
+
+        list_for_each_entry (entry, &priv->peers, uuid_list) {
+                if (!uuid_compare (entry->uuid, uuid)) {
+
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                 "Friend found... state: %s",
+                        glusterd_friend_sm_state_name_get (entry->state.state));
+                        *peerinfo = entry;
+                        return 0;
+                }
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "Friend with uuid: %s, not found",
+                uuid_utoa (uuid));
+        return ret;
+}
+
+
+int
+glusterd_friend_find_by_hostname (const char *hoststr,
+                                  glusterd_peerinfo_t  **peerinfo)
+{
+        int                     ret = -1;
+        glusterd_conf_t         *priv = NULL;
+        glusterd_peerinfo_t     *entry = NULL;
+        struct addrinfo         *addr = NULL;
+        struct addrinfo         *p = NULL;
+        char                    *host = NULL;
+        struct sockaddr_in6     *s6 = NULL;
+        struct sockaddr_in      *s4 = NULL;
+        struct in_addr          *in_addr = NULL;
+        char                    hname[1024] = {0,};
+        xlator_t                *this  = NULL;
+
+
+        this = THIS;
+        GF_ASSERT (hoststr);
+        GF_ASSERT (peerinfo);
+
+        *peerinfo = NULL;
+        priv    = this->private;
+
+        GF_ASSERT (priv);
+
+        list_for_each_entry (entry, &priv->peers, uuid_list) {
+                if (!strncasecmp (entry->hostname, hoststr,
+                                  1024)) {
+
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                 "Friend %s found.. state: %d", hoststr,
+                                  entry->state.state);
+                        *peerinfo = entry;
+                        return 0;
+                }
+        }
+
+        ret = getaddrinfo (hoststr, NULL, NULL, &addr);
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "error in getaddrinfo: %s\n",
+                        gai_strerror(ret));
+                goto out;
+        }
+
+        for (p = addr; p != NULL; p = p->ai_next) {
+                switch (p->ai_family) {
+                        case AF_INET:
+                                s4 = (struct sockaddr_in *) p->ai_addr;
+                                in_addr = &s4->sin_addr;
+                                break;
+                        case AF_INET6:
+                                s6 = (struct sockaddr_in6 *) p->ai_addr;
+                                in_addr =(struct in_addr *) &s6->sin6_addr;
+                                break;
+                       default: ret = -1;
+                                goto out;
+                }
+                host = inet_ntoa(*in_addr);
+
+                ret = getnameinfo (p->ai_addr, p->ai_addrlen, hname,
+                                   1024, NULL, 0, 0);
+                if (ret)
+                        goto out;
+
+                list_for_each_entry (entry, &priv->peers, uuid_list) {
+                        if (!strncasecmp (entry->hostname, host,
+                            1024) || !strncasecmp (entry->hostname,hname,
+                            1024)) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "Friend %s found.. state: %d",
+                                        hoststr, entry->state.state);
+                                *peerinfo = entry;
+                                freeaddrinfo (addr);
+                                return 0;
+                        }
+                }
+        }
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Unable to find friend: %s", hoststr);
+        if (addr)
+                freeaddrinfo (addr);
+        return -1;
+}
+
+int
+glusterd_hostname_to_uuid (char *hostname, uuid_t uuid)
+{
+        GF_ASSERT (hostname);
+        GF_ASSERT (uuid);
+
+        glusterd_peerinfo_t     *peerinfo = NULL;
+        glusterd_conf_t         *priv = NULL;
+        int                     ret = -1;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = glusterd_friend_find_by_hostname (hostname, &peerinfo);
+        if (ret) {
+                if (gf_is_local_addr (hostname)) {
+                        uuid_copy (uuid, MY_UUID);
+                        ret = 0;
+                } else {
+                        ret = 0;
+                        goto out;
+                }
+        } else {
+                uuid_copy (uuid, peerinfo->uuid);
+        }
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_brick_stop (glusterd_volinfo_t *volinfo,
+                     glusterd_brickinfo_t *brickinfo,
+                     gf_boolean_t del_brick)
+{
+        int                                     ret   = -1;
+        xlator_t                                *this = NULL;
+        glusterd_conf_t                         *conf = NULL;
+
+        if ((!brickinfo) || (!volinfo))
+                goto out;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        if (uuid_is_null (brickinfo->uuid)) {
+                ret = glusterd_resolve_brick (brickinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK,
+                                brickinfo->hostname, brickinfo->path);
+                        goto out;
+                }
+        }
+
+        if (uuid_compare (brickinfo->uuid, MY_UUID)) {
+                ret = 0;
+                if (del_brick)
+                        glusterd_delete_brick (volinfo, brickinfo);
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "About to stop glusterfs"
+                " for brick %s:%s", brickinfo->hostname,
+                brickinfo->path);
+        ret = glusterd_volume_stop_glusterfs (volinfo, brickinfo, del_brick);
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL, "Unable to stop"
+                        " brick: %s:%s", brickinfo->hostname,
+                        brickinfo->path);
+                goto out;
+        }
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d ", ret);
+        return ret;
+}
+
+int
+glusterd_is_defrag_on (glusterd_volinfo_t *volinfo)
+{
+        return (volinfo->rebal.defrag != NULL);
+}
+
+gf_boolean_t
+glusterd_is_rb_ongoing (glusterd_volinfo_t *volinfo)
+{
+        gf_boolean_t     ret = _gf_false;
+
+        GF_ASSERT (volinfo);
+
+        if (glusterd_is_rb_started (volinfo) ||
+            glusterd_is_rb_paused (volinfo))
+                ret = _gf_true;
+
+        return ret;
+}
+
+int
+glusterd_new_brick_validate (char *brick, glusterd_brickinfo_t *brickinfo,
+                             char *op_errstr, size_t len)
+{
+        glusterd_brickinfo_t    *newbrickinfo = NULL;
+        int                     ret = -1;
+        gf_boolean_t            is_allocated = _gf_false;
+        glusterd_peerinfo_t     *peerinfo = NULL;
+        glusterd_conf_t         *priv = NULL;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+
+        GF_ASSERT (brick);
+        GF_ASSERT (op_errstr);
+
+        if (!brickinfo) {
+                ret = glusterd_brickinfo_new_from_brick (brick, &newbrickinfo);
+                if (ret)
+                        goto out;
+                is_allocated = _gf_true;
+        } else {
+                newbrickinfo = brickinfo;
+        }
+
+        ret = glusterd_resolve_brick (newbrickinfo);
+        if (ret) {
+                snprintf(op_errstr, len, "Host %s is not in \'Peer "
+                         "in Cluster\' state", newbrickinfo->hostname);
+                goto out;
+        }
+
+        if (!uuid_compare (MY_UUID, newbrickinfo->uuid)) {
+                /* brick is local */
+                if (!glusterd_is_brickpath_available (newbrickinfo->uuid,
+                                                      newbrickinfo->path)) {
+                        snprintf(op_errstr, len, "Brick: %s not available."
+                                 " Brick may be containing or be contained "
+                                 "by an existing brick", brick);
+                        ret = -1;
+                        goto out;
+                }
+
+        } else {
+                ret = glusterd_friend_find_by_uuid (newbrickinfo->uuid,
+                                                    &peerinfo);
+                if (ret) {
+                        snprintf (op_errstr, len, "Failed to find host %s",
+                                  newbrickinfo->hostname);
+                        goto out;
+                }
+
+                if ((!peerinfo->connected)) {
+                        snprintf(op_errstr, len, "Host %s not connected",
+                                 newbrickinfo->hostname);
+                        ret = -1;
+                        goto out;
+                }
+
+                if (peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED) {
+                        snprintf(op_errstr, len, "Host %s is not in \'Peer "
+                                 "in Cluster\' state",
+                                 newbrickinfo->hostname);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        ret = 0;
+out:
+        if (is_allocated)
+                glusterd_brickinfo_delete (newbrickinfo);
+        if (op_errstr[0] != '\0')
+                gf_log (this->name, GF_LOG_ERROR, "%s", op_errstr);
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d ", ret);
+        return ret;
+}
+
+int
+glusterd_is_rb_started(glusterd_volinfo_t *volinfo)
+{
+        gf_log ("", GF_LOG_DEBUG,
+                "is_rb_started:status=%d", volinfo->rep_brick.rb_status);
+        return (volinfo->rep_brick.rb_status == GF_RB_STATUS_STARTED);
+
+}
+
+int
+glusterd_is_rb_paused ( glusterd_volinfo_t *volinfo)
+{
+        gf_log ("", GF_LOG_DEBUG,
+                "is_rb_paused:status=%d", volinfo->rep_brick.rb_status);
+
+        return (volinfo->rep_brick.rb_status == GF_RB_STATUS_PAUSED);
+}
+
+inline int
+glusterd_set_rb_status (glusterd_volinfo_t *volinfo, gf_rb_status_t status)
+{
+        gf_log ("", GF_LOG_DEBUG,
+                "setting status from %d to %d",
+                volinfo->rep_brick.rb_status,
+                status);
+
+        volinfo->rep_brick.rb_status = status;
+        return 0;
+}
+
+inline int
+glusterd_rb_check_bricks (glusterd_volinfo_t *volinfo,
+                          glusterd_brickinfo_t *src, glusterd_brickinfo_t *dst)
+{
+        glusterd_replace_brick_t        *rb = NULL;
+
+        GF_ASSERT (volinfo);
+
+        rb = &volinfo->rep_brick;
+
+        if (!rb->src_brick || !rb->dst_brick)
+                return -1;
+
+        if (strcmp (rb->src_brick->hostname, src->hostname) ||
+            strcmp (rb->src_brick->path, src->path)) {
+                gf_log("", GF_LOG_ERROR, "Replace brick src bricks differ");
+                return -1;
+        }
+
+        if (strcmp (rb->dst_brick->hostname, dst->hostname) ||
+            strcmp (rb->dst_brick->path, dst->path)) {
+                gf_log ("", GF_LOG_ERROR, "Replace brick dst bricks differ");
+                return -1;
+        }
+
+        return 0;
+}
+
+/*path needs to be absolute; works only on gfid, volume-id*/
+static int
+glusterd_is_uuid_present (char *path, char *xattr, gf_boolean_t *present)
+{
+        GF_ASSERT (path);
+        GF_ASSERT (xattr);
+        GF_ASSERT (present);
+
+        int     ret      = -1;
+        uuid_t  uid     = {0,};
+
+        if (!path || !xattr || !present)
+                goto out;
+
+        ret = sys_lgetxattr (path, xattr, &uid, 16);
+
+        if (ret >= 0) {
+                *present = _gf_true;
+                ret = 0;
+                goto out;
+        }
+
+        switch (errno) {
+#if defined(ENODATA)
+                case ENODATA: /* FALLTHROUGH */
+#endif
+#if defined(ENOATTR) && (ENOATTR != ENODATA)
+                case ENOATTR: /* FALLTHROUGH */
+#endif
+                case ENOTSUP:
+                        *present = _gf_false;
+                        ret = 0;
+                        break;
+                default:
+                        break;
+        }
+out:
+        return ret;
+}
+
+/*path needs to be absolute*/
+static int
+glusterd_is_path_in_use (char *path, gf_boolean_t *in_use, char **op_errstr)
+{
+        int             i               = 0;
+        int             ret             = -1;
+        gf_boolean_t    used            = _gf_false;
+        char            dir[PATH_MAX]   = {0,};
+        char            *curdir         = NULL;
+        char            msg[2048]       = {0};
+        char            *keys[3]         = {GFID_XATTR_KEY,
+                                            GF_XATTR_VOL_ID_KEY,
+                                            NULL};
+
+        GF_ASSERT (path);
+        if (!path)
+                goto out;
+
+        strcpy (dir, path);
+        curdir = dir;
+        do {
+                for (i = 0; !used && keys[i]; i++) {
+                        ret = glusterd_is_uuid_present (curdir, keys[i], &used);
+                        if (ret)
+                                goto out;
+                }
+
+                if (used)
+                        break;
+
+                curdir = dirname (curdir);
+                if (!strcmp (curdir, "."))
+                        goto out;
+
+
+        } while (strcmp (curdir, "/"));
+
+        if (!strcmp (curdir, "/")) {
+                for (i = 0; !used && keys[i]; i++) {
+                        ret = glusterd_is_uuid_present (curdir, keys[i], &used);
+                        if (ret)
+                                goto out;
+                }
+        }
+
+        ret = 0;
+        *in_use = used;
+out:
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Failed to get extended "
+                          "attribute %s, reason: %s", keys[i],
+                          strerror (errno));
+        }
+
+        if (*in_use) {
+                if (!strcmp (path, curdir)) {
+                        snprintf (msg, sizeof (msg), "%s is already part of a "
+                          "volume", path);
+                } else {
+                        snprintf (msg, sizeof (msg), "parent directory %s is "
+                          "already part of a volume", curdir);
+                }
+        }
+
+        if (strlen (msg)) {
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+        }
+
+        return ret;
+}
+
+int
+glusterd_check_and_set_brick_xattr (char *host, char *path, uuid_t uuid,
+                                    char **op_errstr, gf_boolean_t is_force)
+{
+        int             ret             = -1;
+        char            msg[2048]       = {0,};
+        gf_boolean_t    in_use          = _gf_false;
+        int             flags           = 0;
+
+        /* Check for xattr support in backend fs */
+        ret = sys_lsetxattr (path, "trusted.glusterfs.test",
+                             "working", 8, 0);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Glusterfs is not"
+                          " supported on brick: %s:%s.\nSetting"
+                          " extended attributes failed, reason:"
+                          " %s.", host, path, strerror(errno));
+                goto out;
+
+        } else {
+                sys_lremovexattr (path, "trusted.glusterfs.test");
+        }
+
+        ret = glusterd_is_path_in_use (path, &in_use, op_errstr);
+        if (ret)
+                goto out;
+
+        if (in_use && !is_force) {
+                ret = -1;
+                goto out;
+        }
+
+
+        if (!is_force)
+                flags = XATTR_CREATE;
+
+        ret = sys_lsetxattr (path, GF_XATTR_VOL_ID_KEY, uuid, 16,
+                             flags);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Failed to set extended "
+                          "attributes %s, reason: %s",
+                          GF_XATTR_VOL_ID_KEY, strerror (errno));
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if (strlen (msg))
+                *op_errstr = gf_strdup (msg);
+
+        return ret;
+}
+
+int
+glusterd_sm_tr_log_transition_add_to_dict (dict_t *dict,
+                                           glusterd_sm_tr_log_t *log, int i,
+                                           int count)
+{
+        int     ret = -1;
+        char    key[512] = {0};
+	char    timestr[64] = {0,};
+        char    *str = NULL;
+
+        GF_ASSERT (dict);
+        GF_ASSERT (log);
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "log%d-old-state", count);
+        str = log->state_name_get (log->transitions[i].old_state);
+        ret = dict_set_str (dict, key, str);
+        if (ret)
+                goto out;
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "log%d-event", count);
+        str = log->event_name_get (log->transitions[i].event);
+        ret = dict_set_str (dict, key, str);
+        if (ret)
+                goto out;
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "log%d-new-state", count);
+        str = log->state_name_get (log->transitions[i].new_state);
+        ret = dict_set_str (dict, key, str);
+        if (ret)
+                goto out;
+
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "log%d-time", count);
+        gf_time_fmt (timestr, sizeof timestr, log->transitions[i].time,
+                     gf_timefmt_FT);
+        str = gf_strdup (timestr);
+        ret = dict_set_dynstr (dict, key, str);
+        if (ret)
+                goto out;
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_sm_tr_log_add_to_dict (dict_t *dict,
+                                glusterd_sm_tr_log_t *circular_log)
+{
+        int     ret = -1;
+        int     i = 0;
+        int     start = 0;
+        int     end     = 0;
+        int     index = 0;
+        char    key[256] = {0};
+        glusterd_sm_tr_log_t *log = NULL;
+        int     count = 0;
+
+        GF_ASSERT (dict);
+        GF_ASSERT (circular_log);
+
+        log = circular_log;
+        if (!log->count)
+                return 0;
+
+        if (log->count == log->size)
+                start = log->current + 1;
+
+        end = start + log->count;
+        for (i = start; i < end; i++, count++) {
+                index = i % log->count;
+                ret = glusterd_sm_tr_log_transition_add_to_dict (dict, log, index,
+                                                                 count);
+                if (ret)
+                        goto out;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "count");
+        ret = dict_set_int32 (dict, key, log->count);
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_sm_tr_log_init (glusterd_sm_tr_log_t *log,
+                         char * (*state_name_get) (int),
+                         char * (*event_name_get) (int),
+                         size_t  size)
+{
+        glusterd_sm_transition_t *transitions = NULL;
+        int                      ret = -1;
+
+        GF_ASSERT (size > 0);
+        GF_ASSERT (log && state_name_get && event_name_get);
+
+        if (!log || !state_name_get || !event_name_get || (size <= 0))
+                goto out;
+
+        transitions = GF_CALLOC (size, sizeof (*transitions),
+                                 gf_gld_mt_sm_tr_log_t);
+        if (!transitions)
+                goto out;
+
+        log->transitions = transitions;
+        log->size        = size;
+        log->state_name_get = state_name_get;
+        log->event_name_get = event_name_get;
+        ret = 0;
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+void
+glusterd_sm_tr_log_delete (glusterd_sm_tr_log_t *log)
+{
+        if (!log)
+                return;
+        GF_FREE (log->transitions);
+        return;
+}
+
+int
+glusterd_sm_tr_log_transition_add (glusterd_sm_tr_log_t *log,
+                                   int old_state, int new_state,
+                                   int event)
+{
+        glusterd_sm_transition_t *transitions = NULL;
+        int                      ret = -1;
+        int                      next = 0;
+        xlator_t                 *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        GF_ASSERT (log);
+        if (!log)
+                goto out;
+
+        transitions = log->transitions;
+        if (!transitions)
+                goto out;
+
+        if (log->count)
+                next = (log->current + 1) % log->size;
+        else
+                next = 0;
+
+        transitions[next].old_state = old_state;
+        transitions[next].new_state = new_state;
+        transitions[next].event     = event;
+        time (&transitions[next].time);
+        log->current = next;
+        if (log->count < log->size)
+                log->count++;
+        ret = 0;
+        gf_log (this->name, GF_LOG_DEBUG, "Transitioning from '%s' to '%s' "
+                "due to event '%s'", log->state_name_get (old_state),
+                log->state_name_get (new_state), log->event_name_get (event));
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_peerinfo_new (glusterd_peerinfo_t **peerinfo,
+                       glusterd_friend_sm_state_t state, uuid_t *uuid,
+                       const char *hostname, int port)
+{
+        glusterd_peerinfo_t      *new_peer = NULL;
+        int                      ret = -1;
+
+        GF_ASSERT (peerinfo);
+        if (!peerinfo)
+                goto out;
+
+        new_peer = GF_CALLOC (1, sizeof (*new_peer), gf_gld_mt_peerinfo_t);
+        if (!new_peer)
+                goto out;
+
+        new_peer->state.state = state;
+        if (hostname)
+                new_peer->hostname = gf_strdup (hostname);
+
+        INIT_LIST_HEAD (&new_peer->uuid_list);
+
+        if (uuid) {
+                uuid_copy (new_peer->uuid, *uuid);
+        }
+
+        ret = glusterd_sm_tr_log_init (&new_peer->sm_log,
+                                       glusterd_friend_sm_state_name_get,
+                                       glusterd_friend_sm_event_name_get,
+                                       GLUSTERD_TR_LOG_SIZE);
+        if (ret)
+                goto out;
+
+        if (new_peer->state.state == GD_FRIEND_STATE_BEFRIENDED)
+                new_peer->quorum_contrib = QUORUM_WAITING;
+        new_peer->port = port;
+        *peerinfo = new_peer;
+out:
+        if (ret && new_peer)
+                glusterd_friend_cleanup (new_peer);
+        gf_log ("", GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_peer_destroy (glusterd_peerinfo_t *peerinfo)
+{
+        int32_t                         ret = -1;
+
+        if (!peerinfo)
+                goto out;
+
+        ret = glusterd_store_delete_peerinfo (peerinfo);
+
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Deleting peer info failed");
+        }
+
+        list_del_init (&peerinfo->uuid_list);
+        GF_FREE (peerinfo->hostname);
+        glusterd_sm_tr_log_delete (&peerinfo->sm_log);
+        GF_FREE (peerinfo);
+        peerinfo = NULL;
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+int
+glusterd_remove_pending_entry (struct list_head *list, void *elem)
+{
+        glusterd_pending_node_t *pending_node = NULL;
+        glusterd_pending_node_t *tmp = NULL;
+        int                     ret = 0;
+
+        list_for_each_entry_safe (pending_node, tmp, list, list) {
+                if (elem == pending_node->node) {
+                        list_del_init (&pending_node->list);
+                        GF_FREE (pending_node);
+                        ret = 0;
+                        goto out;
+                }
+        }
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+
+}
+
+int
+glusterd_clear_pending_nodes (struct list_head *list)
+{
+        glusterd_pending_node_t *pending_node = NULL;
+        glusterd_pending_node_t *tmp = NULL;
+
+        list_for_each_entry_safe (pending_node, tmp, list, list) {
+                list_del_init (&pending_node->list);
+                GF_FREE (pending_node);
+        }
+
+        return 0;
+}
+
+gf_boolean_t
+glusterd_peerinfo_is_uuid_unknown (glusterd_peerinfo_t *peerinfo)
+{
+        GF_ASSERT (peerinfo);
+
+        if (uuid_is_null (peerinfo->uuid))
+                return _gf_true;
+        return _gf_false;
+}
+
+int32_t
+glusterd_delete_volume (glusterd_volinfo_t *volinfo)
+{
+        int             ret = -1;
+        GF_ASSERT (volinfo);
+
+        ret = glusterd_store_delete_volume (volinfo);
+
+        if (ret)
+                goto out;
+
+        ret = glusterd_volinfo_delete (volinfo);
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_delete_brick (glusterd_volinfo_t* volinfo,
+                       glusterd_brickinfo_t *brickinfo)
+{
+        int             ret = 0;
+        char      voldir[PATH_MAX] = {0,};
+        glusterd_conf_t *priv = THIS->private;
+        GF_ASSERT (volinfo);
+        GF_ASSERT (brickinfo);
+
+        GLUSTERD_GET_VOLUME_DIR(voldir, volinfo, priv);
+
+        glusterd_delete_volfile (volinfo, brickinfo);
+        glusterd_store_delete_brick (brickinfo, voldir);
+        glusterd_brickinfo_delete (brickinfo);
+        volinfo->brick_count--;
+        return ret;
+}
+
+int32_t
+glusterd_delete_all_bricks (glusterd_volinfo_t* volinfo)
+{
+        int             ret = 0;
+        glusterd_brickinfo_t *brickinfo = NULL;
+        glusterd_brickinfo_t *tmp = NULL;
+
+        GF_ASSERT (volinfo);
+
+        list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks, brick_list) {
+                ret = glusterd_delete_brick (volinfo, brickinfo);
+        }
+        return ret;
+}
+
+int
+glusterd_get_local_brickpaths (glusterd_volinfo_t *volinfo, char **pathlist)
+{
+        char                 **path_tokens  = NULL;
+        char                  *tmp_path_list = NULL;
+        char                   path[PATH_MAX] = "";
+        int32_t                count          = 0;
+        int32_t                pathlen        = 0;
+        int32_t                total_len      = 0;
+        int32_t                ret            = 0;
+        int                    i              = 0;
+        glusterd_brickinfo_t  *brickinfo      = NULL;
+
+        if ((!volinfo) || (!pathlist))
+            goto out;
+
+        path_tokens = GF_CALLOC (sizeof(char*), volinfo->brick_count,
+                                 gf_gld_mt_charptr);
+        if (!path_tokens) {
+                gf_log ("", GF_LOG_DEBUG, "Could not allocate memory.");
+                ret = -1;
+                goto out;
+        }
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                if (uuid_compare (brickinfo->uuid, MY_UUID))
+                        continue;
+
+                 pathlen = snprintf (path, sizeof(path),
+                                     "--path=%s ", brickinfo->path);
+                 if (pathlen < sizeof(path))
+                        path[pathlen] = '\0';
+                 else
+                        path[sizeof(path)-1] = '\0';
+                 path_tokens[count] = gf_strdup (path);
+                 if (!path_tokens[count]) {
+                        gf_log ("", GF_LOG_DEBUG,
+                                "Could not allocate memory.");
+                        ret = -1;
+                        goto out;
+                 }
+                 count++;
+                 total_len += pathlen;
+        }
+
+        tmp_path_list = GF_CALLOC (sizeof(char), total_len + 1,
+                                   gf_gld_mt_char);
+        if (!tmp_path_list) {
+                gf_log ("", GF_LOG_DEBUG, "Could not allocate memory.");
+                ret = -1;
+                goto out;
+        }
+
+        for (i = 0; i < count; i++)
+                strcat (tmp_path_list, path_tokens[i]);
+
+        if (count)
+                *pathlist = tmp_path_list;
+
+        ret = count;
+out:
+        for (i = 0; i < count; i++) {
+                GF_FREE (path_tokens[i]);
+                path_tokens[i] = NULL;
+        }
+
+        GF_FREE (path_tokens);
+        path_tokens = NULL;
+
+        if (ret == 0) {
+                gf_log ("", GF_LOG_DEBUG, "No Local Bricks Present.");
+                GF_FREE (tmp_path_list);
+                tmp_path_list = NULL;
+        }
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_start_gsync (glusterd_volinfo_t *master_vol, char *slave,
+                      char *path_list, char *conf_path,
+                      char *glusterd_uuid_str,
+                      char **op_errstr)
+{
+        int32_t         ret     = 0;
+        int32_t         status  = 0;
+        char            uuid_str [64] = {0};
+        runner_t        runner = {0,};
+        xlator_t        *this = NULL;
+        glusterd_conf_t *priv = NULL;
+        int             errcode = 0;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        uuid_utoa_r (MY_UUID, uuid_str);
+
+        if (!path_list) {
+                ret = 0;
+                gf_log ("", GF_LOG_DEBUG, "No Bricks in this node."
+                        " Not starting gsyncd.");
+                goto out;
+        }
+
+        ret = gsync_status (master_vol->volname, slave, conf_path, &status);
+        if (status == 0)
+                goto out;
+
+        uuid_utoa_r (master_vol->volume_id, uuid_str);
+        runinit (&runner);
+        runner_add_args  (&runner, GSYNCD_PREFIX"/gsyncd",
+                          path_list, "-c", NULL);
+        runner_argprintf (&runner, "%s", conf_path);
+        runner_argprintf (&runner, ":%s", master_vol->volname);
+        runner_add_args  (&runner, slave, "--config-set", "session-owner",
+                          uuid_str, NULL);
+        synclock_unlock (&priv->big_lock);
+        ret = runner_run (&runner);
+        synclock_lock (&priv->big_lock);
+        if (ret == -1) {
+                errcode = -1;
+                goto out;
+        }
+
+        runinit (&runner);
+        runner_add_args  (&runner, GSYNCD_PREFIX"/gsyncd",
+                          path_list, "--monitor", "-c", NULL);
+        runner_argprintf (&runner, "%s", conf_path);
+        runner_argprintf (&runner, ":%s", master_vol->volname);
+        runner_argprintf (&runner, "--glusterd-uuid=%s",
+                          uuid_utoa (priv->uuid));
+        runner_add_arg   (&runner, slave);
+        synclock_unlock (&priv->big_lock);
+        ret = runner_run (&runner);
+        synclock_lock (&priv->big_lock);
+        if (ret == -1) {
+                gf_asprintf (op_errstr, GEOREP" start failed for %s %s",
+                             master_vol->volname, slave);
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        if ((ret != 0) && errcode == -1) {
+                if (op_errstr)
+                        *op_errstr = gf_strdup ("internal error, cannot start "
+                                                "the " GEOREP " session");
+        }
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_recreate_volfiles (glusterd_conf_t *conf)
+{
+
+        glusterd_volinfo_t      *volinfo = NULL;
+        int                      ret = 0;
+        int                      op_ret = 0;
+
+        GF_ASSERT (conf);
+        list_for_each_entry (volinfo, &conf->volumes, vol_list) {
+                ret = generate_brick_volfiles (volinfo);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR, "Failed to "
+                                "regenerate brick volfiles for %s",
+                                volinfo->volname);
+                        op_ret = ret;
+                }
+                ret = generate_client_volfiles (volinfo, GF_CLIENT_TRUSTED);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR, "Failed to "
+                                "regenerate trusted client volfiles for %s",
+                                volinfo->volname);
+                        op_ret = ret;
+                }
+                ret = generate_client_volfiles (volinfo, GF_CLIENT_OTHER);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR, "Failed to "
+                                "regenerate client volfiles for %s",
+                                volinfo->volname);
+                        op_ret = ret;
+                }
+        }
+        return op_ret;
+}
+
+int32_t
+glusterd_handle_upgrade_downgrade (dict_t *options, glusterd_conf_t *conf)
+{
+        int              ret                            = 0;
+        char            *type                           = NULL;
+        gf_boolean_t     upgrade                        = _gf_false;
+        gf_boolean_t     downgrade                      = _gf_false;
+        gf_boolean_t     regenerate_volfiles            = _gf_false;
+        gf_boolean_t     terminate                      = _gf_false;
+
+        ret = dict_get_str (options, "upgrade", &type);
+        if (!ret) {
+                ret = gf_string2boolean (type, &upgrade);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR, "upgrade option "
+                                "%s is not a valid boolean type", type);
+                        ret = -1;
+                        goto out;
+                }
+                if (_gf_true == upgrade)
+                        regenerate_volfiles = _gf_true;
+        }
+
+        ret = dict_get_str (options, "downgrade", &type);
+        if (!ret) {
+                ret = gf_string2boolean (type, &downgrade);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR, "downgrade option "
+                                "%s is not a valid boolean type", type);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        if (upgrade && downgrade) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Both upgrade and downgrade"
+                        " options are set. Only one should be on");
+                ret = -1;
+                goto out;
+        }
+
+        if (!upgrade && !downgrade)
+                ret = 0;
+        else
+                terminate = _gf_true;
+        if (regenerate_volfiles) {
+                ret = glusterd_recreate_volfiles (conf);
+        }
+out:
+        if (terminate && (ret == 0))
+                kill (getpid(), SIGTERM);
+        return ret;
+}
+
+gf_boolean_t
+glusterd_is_volume_replicate (glusterd_volinfo_t *volinfo)
+{
+        gf_boolean_t    replicates = _gf_false;
+        if (volinfo && ((volinfo->type == GF_CLUSTER_TYPE_REPLICATE) ||
+            (volinfo->type == GF_CLUSTER_TYPE_STRIPE_REPLICATE)))
+                replicates = _gf_true;
+        return replicates;
+}
+
+int
+glusterd_set_dump_options (char *dumpoptions_path, char *options,
+                           int option_cnt)
+{
+        int     ret = 0;
+        char    *dup_options = NULL;
+        char    *option = NULL;
+        char    *tmpptr = NULL;
+        FILE    *fp = NULL;
+        int     nfs_cnt = 0;
+
+        if (0 == option_cnt ||
+            (option_cnt == 1 && (!strcmp (options, "nfs ")))) {
+                ret = 0;
+                goto out;
+        }
+
+        fp = fopen (dumpoptions_path, "w");
+        if (!fp) {
+                ret = -1;
+                goto out;
+        }
+        dup_options = gf_strdup (options);
+        gf_log ("", GF_LOG_INFO, "Received following statedump options: %s",
+                dup_options);
+        option = strtok_r (dup_options, " ", &tmpptr);
+        while (option) {
+                if (!strcmp (option, "nfs")) {
+                        if (nfs_cnt > 0) {
+                                unlink (dumpoptions_path);
+                                ret = 0;
                                 goto out;
                         }
+                        nfs_cnt++;
+                        option = strtok_r (NULL, " ", &tmpptr);
+                        continue;
+                }
+                fprintf (fp, "%s=yes\n", option);
+                option = strtok_r (NULL, " ", &tmpptr);
+        }
+
+out:
+        if (fp)
+                fclose (fp);
+        GF_FREE (dup_options);
+        return ret;
+}
+
+int
+glusterd_brick_statedump (glusterd_volinfo_t *volinfo,
+                          glusterd_brickinfo_t *brickinfo,
+                          char *options, int option_cnt, char **op_errstr)
+{
+        int                     ret = -1;
+        xlator_t                *this = NULL;
+        glusterd_conf_t         *conf = NULL;
+        char                    pidfile_path[PATH_MAX] = {0,};
+        char                    dumpoptions_path[PATH_MAX] = {0,};
+        FILE                    *pidfile = NULL;
+        pid_t                   pid = -1;
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        if (uuid_is_null (brickinfo->uuid)) {
+                ret = glusterd_resolve_brick (brickinfo);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR,
+                                "Cannot resolve brick %s:%s",
+                                brickinfo->hostname, brickinfo->path);
+                        goto out;
                 }
         }
+
+        if (uuid_compare (brickinfo->uuid, MY_UUID)) {
+                ret = 0;
+                goto out;
+        }
+
+        GLUSTERD_GET_BRICK_PIDFILE (pidfile_path, volinfo, brickinfo, conf);
+
+        pidfile = fopen (pidfile_path, "r");
+        if (!pidfile) {
+                gf_log ("", GF_LOG_ERROR, "Unable to open pidfile: %s",
+                        pidfile_path);
+                ret = -1;
+                goto out;
+        }
+
+        ret = fscanf (pidfile, "%d", &pid);
+        if (ret <= 0) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get pid of brick process");
+                ret = -1;
+                goto out;
+        }
+
+        snprintf (dumpoptions_path, sizeof (dumpoptions_path),
+                  DEFAULT_VAR_RUN_DIRECTORY"/glusterdump.%d.options", pid);
+        ret = glusterd_set_dump_options (dumpoptions_path, options, option_cnt);
+        if (ret < 0) {
+                gf_log ("", GF_LOG_ERROR, "error while parsing the statedump "
+                        "options");
+                ret = -1;
+                goto out;
+        }
+
+        gf_log ("", GF_LOG_INFO, "Performing statedump on brick with pid %d",
+                pid);
+
+        kill (pid, SIGUSR1);
+
+        sleep (1);
+        ret = 0;
+out:
+        unlink (dumpoptions_path);
+        if (pidfile)
+                fclose (pidfile);
+        return ret;
+}
+
+int
+glusterd_nfs_statedump (char *options, int option_cnt, char **op_errstr)
+{
+        int                     ret = -1;
+        xlator_t                *this = NULL;
+        glusterd_conf_t         *conf = NULL;
+        char                    pidfile_path[PATH_MAX] = {0,};
+        char                    path[PATH_MAX] = {0,};
+        FILE                    *pidfile = NULL;
+        pid_t                   pid = -1;
+        char                    dumpoptions_path[PATH_MAX] = {0,};
+        char                    *option = NULL;
+        char                    *tmpptr = NULL;
+        char                    *dup_options = NULL;
+        char                    msg[256] = {0,};
+
+        this = THIS;
+        GF_ASSERT (this);
+        conf = this->private;
+        GF_ASSERT (conf);
+
+        dup_options = gf_strdup (options);
+        option = strtok_r (dup_options, " ", &tmpptr);
+        if (strcmp (option, "nfs")) {
+                snprintf (msg, sizeof (msg), "for nfs statedump, options should"
+                          " be after the key nfs");
+                *op_errstr = gf_strdup (msg);
+                ret = -1;
+                goto out;
+        }
+
+        GLUSTERD_GET_NFS_DIR (path, conf);
+        GLUSTERD_GET_NFS_PIDFILE (pidfile_path, path);
+
+        pidfile = fopen (pidfile_path, "r");
+        if (!pidfile) {
+                gf_log ("", GF_LOG_ERROR, "Unable to open pidfile: %s",
+                        pidfile_path);
+                ret = -1;
+                goto out;
+        }
+
+        ret = fscanf (pidfile, "%d", &pid);
+        if (ret <= 0) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get pid of brick process");
+                ret = -1;
+                goto out;
+        }
+
+        snprintf (dumpoptions_path, sizeof (dumpoptions_path),
+                  DEFAULT_VAR_RUN_DIRECTORY"/glusterdump.%d.options", pid);
+        ret = glusterd_set_dump_options (dumpoptions_path, options, option_cnt);
+        if (ret < 0) {
+                gf_log ("", GF_LOG_ERROR, "error while parsing the statedump "
+                        "options");
+                ret = -1;
+                goto out;
+        }
+
+        gf_log ("", GF_LOG_INFO, "Performing statedump on nfs server with "
+                "pid %d", pid);
+
+        kill (pid, SIGUSR1);
+
+        sleep (1);
+
+        ret = 0;
+out:
+        if (pidfile)
+                fclose (pidfile);
+        unlink (dumpoptions_path);
+        GF_FREE (dup_options);
+        return ret;
+}
+
+/* Checks if the given peer contains all the bricks belonging to the
+ * given volume. Returns true if it does else returns false
+ */
+gf_boolean_t
+glusterd_friend_contains_vol_bricks (glusterd_volinfo_t *volinfo,
+                                     uuid_t friend_uuid)
+{
+        gf_boolean_t            ret = _gf_true;
+        glusterd_brickinfo_t    *brickinfo = NULL;
+
+        GF_ASSERT (volinfo);
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                if (uuid_compare (friend_uuid, brickinfo->uuid)) {
+                        ret = _gf_false;
+                        break;
+                }
+        }
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+/* Remove all volumes which completely belong to given friend
+ */
+int
+glusterd_friend_remove_cleanup_vols (uuid_t uuid)
+{
+        int                     ret = -1;
+        glusterd_conf_t         *priv = NULL;
+        glusterd_volinfo_t      *volinfo = NULL;
+        glusterd_volinfo_t      *tmp_volinfo = NULL;
+
+        priv = THIS->private;
+        GF_ASSERT (priv);
+
+        list_for_each_entry_safe (volinfo, tmp_volinfo,
+                                  &priv->volumes, vol_list) {
+                if (glusterd_friend_contains_vol_bricks (volinfo, uuid)) {
+                        gf_log (THIS->name, GF_LOG_INFO,
+                                "Deleting stale volume %s", volinfo->volname);
+                        ret = glusterd_delete_volume (volinfo);
+                        if (ret) {
+                                gf_log (THIS->name, GF_LOG_ERROR,
+                                        "Error deleting stale volume");
+                                goto out;
+                        }
+                }
+        }
+        ret = 0;
+out:
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+/* Check if the all peers are connected and befriended, except the peer
+ * specified (the peer being detached)
+ */
+gf_boolean_t
+glusterd_chk_peers_connected_befriended (uuid_t skip_uuid)
+{
+        gf_boolean_t            ret = _gf_true;
+        glusterd_peerinfo_t     *peerinfo = NULL;
+        glusterd_conf_t         *priv = NULL;
+
+        priv= THIS->private;
+        GF_ASSERT (priv);
+
+        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
+
+                if (!uuid_is_null (skip_uuid) && !uuid_compare (skip_uuid,
+                                                           peerinfo->uuid))
+                        continue;
+
+                if ((GD_FRIEND_STATE_BEFRIENDED != peerinfo->state.state)
+                    || !(peerinfo->connected)) {
+                        ret = _gf_false;
+                        break;
+                }
+        }
+        gf_log (THIS->name, GF_LOG_DEBUG, "Returning %s",
+                (ret?"TRUE":"FALSE"));
+        return ret;
+}
+
+void
+glusterd_get_client_filepath (char *filepath, glusterd_volinfo_t *volinfo,
+                              gf_transport_type type)
+{
+        char  path[PATH_MAX] = {0,};
+        glusterd_conf_t *priv = NULL;
+
+        priv = THIS->private;
+
+        GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv);
+
+        if ((volinfo->transport_type == GF_TRANSPORT_BOTH_TCP_RDMA) &&
+            (type == GF_TRANSPORT_RDMA))
+                snprintf (filepath, PATH_MAX, "%s/%s.rdma-fuse.vol",
+                          path, volinfo->volname);
+        else
+                snprintf (filepath, PATH_MAX, "%s/%s-fuse.vol",
+                          path, volinfo->volname);
+}
+
+void
+glusterd_get_trusted_client_filepath (char *filepath,
+                                      glusterd_volinfo_t *volinfo,
+                                      gf_transport_type type)
+{
+        char  path[PATH_MAX] = {0,};
+        glusterd_conf_t *priv = NULL;
+
+        priv = THIS->private;
+
+        GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv);
+
+        if ((volinfo->transport_type == GF_TRANSPORT_BOTH_TCP_RDMA) &&
+            (type == GF_TRANSPORT_RDMA))
+                snprintf (filepath, PATH_MAX,
+                          "%s/trusted-%s.rdma-fuse.vol",
+                          path, volinfo->volname);
+        else
+                snprintf (filepath, PATH_MAX,
+                          "%s/trusted-%s-fuse.vol",
+                          path, volinfo->volname);
+}
+
+int
+glusterd_volume_defrag_restart (glusterd_volinfo_t *volinfo, char *op_errstr,
+                              size_t len, int cmd, defrag_cbk_fn_t cbk)
+{
+        glusterd_conf_t         *priv                   = NULL;
+        char                     pidfile[PATH_MAX];
+        int                      ret                    = -1;
+        pid_t                    pid;
+
+        priv = THIS->private;
+        if (!priv)
+                return ret;
+
+        GLUSTERD_GET_DEFRAG_PID_FILE(pidfile, volinfo, priv);
+
+        if (!glusterd_is_service_running (pidfile, &pid)) {
+                glusterd_handle_defrag_start (volinfo, op_errstr, len, cmd,
+                                              cbk, volinfo->rebal.op);
+        } else {
+                glusterd_rebalance_rpc_create (volinfo, priv, cmd);
+        }
+
+        return ret;
+}
+
+int
+glusterd_restart_rebalance (glusterd_conf_t *conf)
+{
+        glusterd_volinfo_t       *volinfo = NULL;
+        int                      ret = 0;
+        char                     op_errstr[256];
+
+        list_for_each_entry (volinfo, &conf->volumes, vol_list) {
+                if (!volinfo->rebal.defrag_cmd)
+                        continue;
+                glusterd_volume_defrag_restart (volinfo, op_errstr, 256,
+                                        volinfo->rebal.defrag_cmd, NULL);
+        }
+        return ret;
+}
+
+void
+glusterd_volinfo_reset_defrag_stats (glusterd_volinfo_t *volinfo)
+{
+        glusterd_rebalance_t *rebal = NULL;
+        GF_ASSERT (volinfo);
+
+        rebal = &volinfo->rebal;
+        rebal->rebalance_files = 0;
+        rebal->rebalance_data = 0;
+        rebal->lookedup_files = 0;
+        rebal->rebalance_failures = 0;
+        rebal->rebalance_time = 0;
+        rebal->skipped_files = 0;
+
+}
+
+/* Return hostname for given uuid if it exists
+ * else return NULL
+ */
+char *
+glusterd_uuid_to_hostname (uuid_t uuid)
+{
+        char                    *hostname = NULL;
+        glusterd_conf_t         *priv = NULL;
+        glusterd_peerinfo_t     *entry = NULL;
+
+        priv = THIS->private;
+        GF_ASSERT (priv);
+
+        if (!uuid_compare (MY_UUID, uuid)) {
+                hostname = gf_strdup ("localhost");
+        }
+        if (!list_empty (&priv->peers)) {
+                list_for_each_entry (entry, &priv->peers, uuid_list) {
+                        if (!uuid_compare (entry->uuid, uuid)) {
+                                hostname = gf_strdup (entry->hostname);
+                                break;
+                        }
+                }
+        }
+
+        return hostname;
+}
+
+gf_boolean_t
+glusterd_is_local_brick (xlator_t *this, glusterd_volinfo_t *volinfo,
+                         glusterd_brickinfo_t *brickinfo)
+{
+        gf_boolean_t    local = _gf_false;
+        int             ret = 0;
+        glusterd_conf_t *conf = NULL;
+
+        if (uuid_is_null (brickinfo->uuid)) {
+                ret = glusterd_resolve_brick (brickinfo);
+                if (ret)
+                        goto out;
+        }
+        conf = this->private;
+        local = !uuid_compare (brickinfo->uuid, MY_UUID);
+out:
+        return local;
+}
+int
+glusterd_validate_volume_id (dict_t *op_dict, glusterd_volinfo_t *volinfo)
+{
+        int     ret             = -1;
+        char    *volid_str      = NULL;
+        uuid_t  vol_uid         = {0, };
+        xlator_t *this          = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = dict_get_str (op_dict, "vol-id", &volid_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get volume id for "
+                        "volume %s", volinfo->volname);
+                goto out;
+        }
+        ret = uuid_parse (volid_str, vol_uid);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to parse volume id "
+                        "for volume %s", volinfo->volname);
+                goto out;
+        }
+
+        if (uuid_compare (vol_uid, volinfo->volume_id)) {
+                gf_log (this->name, GF_LOG_ERROR, "Volume ids of volume %s - %s"
+                        " and %s - are different. Possibly a split brain among "
+                        "peers.", volinfo->volname, volid_str,
+                        uuid_utoa (volinfo->volume_id));
+                ret = -1;
+                goto out;
+        }
+
+out:
+        return ret;
+}
+
+int
+glusterd_defrag_volume_status_update (glusterd_volinfo_t *volinfo,
+                                      dict_t *rsp_dict)
+{
+        int                             ret = 0;
+        uint64_t                        files = 0;
+        uint64_t                        size = 0;
+        uint64_t                        lookup = 0;
+        gf_defrag_status_t              status = GF_DEFRAG_STATUS_NOT_STARTED;
+        uint64_t                        failures = 0;
+        uint64_t                        skipped = 0;
+        xlator_t                       *this = NULL;
+        double                          run_time = 0;
+
+        this = THIS;
+
+        ret = dict_get_uint64 (rsp_dict, "files", &files);
+        if (ret)
+                gf_log (this->name, GF_LOG_TRACE,
+                        "failed to get file count");
+
+        ret = dict_get_uint64 (rsp_dict, "size", &size);
+        if (ret)
+                gf_log (this->name, GF_LOG_TRACE,
+                        "failed to get size of xfer");
+
+        ret = dict_get_uint64 (rsp_dict, "lookups", &lookup);
+        if (ret)
+                gf_log (this->name, GF_LOG_TRACE,
+                        "failed to get lookedup file count");
+
+        ret = dict_get_int32 (rsp_dict, "status", (int32_t *)&status);
+        if (ret)
+                gf_log (this->name, GF_LOG_TRACE,
+                        "failed to get status");
+
+        ret = dict_get_uint64 (rsp_dict, "failures", &failures);
+        if (ret)
+                gf_log (this->name, GF_LOG_TRACE,
+                        "failed to get failure count");
+
+        ret = dict_get_uint64 (rsp_dict, "skipped", &skipped);
+        if (ret)
+                gf_log (this->name, GF_LOG_TRACE,
+                        "failed to get skipped count");
+
+        ret = dict_get_double (rsp_dict, "run-time", &run_time);
+        if (ret)
+                gf_log (this->name, GF_LOG_TRACE,
+                        "failed to get run-time");
+
+        if (files)
+                volinfo->rebal.rebalance_files = files;
+        if (size)
+                volinfo->rebal.rebalance_data = size;
+        if (lookup)
+                volinfo->rebal.lookedup_files = lookup;
+        if (status)
+                volinfo->rebal.defrag_status = status;
+        if (failures)
+                volinfo->rebal.rebalance_failures = failures;
+        if (skipped)
+                volinfo->rebal.skipped_files = skipped;
+        if (run_time)
+                volinfo->rebal.rebalance_time = run_time;
+
+        return ret;
+}
+
+int
+glusterd_check_topology_identical (const char *filename1,
+                                   const char *filename2,
+                                   gf_boolean_t *identical)
+{
+        int                     ret = -1; /* FAILURE */
+        xlator_t                *this = NULL;
+        FILE                    *fp1 = NULL;
+        FILE                    *fp2 = NULL;
+        glusterfs_graph_t       *grph1 = NULL;
+        glusterfs_graph_t       *grph2 = NULL;
+
+        if ((!filename1) || (!filename2) || (!identical))
+                goto out;
+
+        this = THIS;
+
+        errno = 0; /* RESET the errno */
+
+        /* fopen() the volfile1 to create the graph */
+        fp1 = fopen (filename1, "r");
+        if (fp1 == NULL) {
+                gf_log (this->name, GF_LOG_ERROR, "fopen() on file: %s failed "
+                        "(%s)", filename1, strerror (errno));
+                goto out;
+        }
+
+        /* fopen() the volfile2 to create the graph */
+        fp2 = fopen (filename2, "r");
+        if (fp2 == NULL) {
+                gf_log (this->name, GF_LOG_ERROR, "fopen() on file: %s failed "
+                        "(%s)", filename2, strerror (errno));
+                goto out;
+        }
+
+        /* create the graph for filename1 */
+        grph1 = glusterfs_graph_construct(fp1);
+        if (grph1 == NULL)
+                goto out;
+
+        /* create the graph for filename2 */
+        grph2 = glusterfs_graph_construct(fp2);
+        if (grph2 == NULL)
+                goto out;
+
+        /* compare the graph topology */
+        *identical = is_graph_topology_equal(grph1, grph2);
+        ret = 0; /* SUCCESS */
+out:
+        if (fp1)
+                fclose(fp1);
+        if (fp2)
+                fclose(fp2);
+        if (grph1)
+                glusterfs_graph_destroy(grph1);
+        if (grph2)
+                glusterfs_graph_destroy(grph2);
+
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
+
+int
+glusterd_check_files_identical (char *filename1, char *filename2,
+                                gf_boolean_t *identical)
+{
+        int                     ret = -1;
+        struct stat             buf1 = {0,};
+        struct stat             buf2 = {0,};
+        uint32_t                cksum1 = 0;
+        uint32_t                cksum2 = 0;
+        xlator_t                *this = NULL;
+
+        GF_ASSERT (filename1);
+        GF_ASSERT (filename2);
+        GF_ASSERT (identical);
+
+        this = THIS;
+
+        ret = stat (filename1, &buf1);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "stat on file: %s failed "
+                        "(%s)", filename1, strerror (errno));
+                goto out;
+        }
+
+        ret = stat (filename2, &buf2);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "stat on file: %s failed "
+                        "(%s)", filename2, strerror (errno));
+                goto out;
+        }
+
+        if (buf1.st_size != buf2.st_size) {
+                *identical = _gf_false;
+                goto out;
+        }
+
+        ret = get_checksum_for_path (filename1, &cksum1);
+        if (ret)
+                goto out;
+
+
+        ret = get_checksum_for_path (filename2, &cksum2);
+        if (ret)
+                goto out;
+
+        if (cksum1 != cksum2)
+                *identical = _gf_false;
+        else
+                *identical = _gf_true;
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret);
+        return ret;
+}
+
+int
+glusterd_volset_help (dict_t *dict, char **op_errstr)
+{
+        int                     ret = -1;
+        gf_boolean_t            xml_out = _gf_false;
+        xlator_t                *this = NULL;
+
+        this = THIS;
+
+        if (!dict) {
+                if (!(dict = glusterd_op_get_ctx ())) {
+                        ret = 0;
+                        goto out;
+                }
+        }
+
+        if (dict_get (dict, "help" )) {
+                xml_out = _gf_false;
+
+        } else if (dict_get (dict, "help-xml" )) {
+                xml_out = _gf_true;
+#if (HAVE_LIB_XML)
+                ret = 0;
+#else
+                gf_log (this->name, GF_LOG_ERROR,
+                        "libxml not present in the system");
+                if (op_errstr)
+                        *op_errstr = gf_strdup ("Error: xml libraries not "
+                                                "present to produce "
+                                                "xml-output");
+                goto out;
+#endif
+
+        } else {
+                goto out;
+        }
+
+        ret = glusterd_get_volopt_content (dict, xml_out);
+        if (ret && op_errstr)
+                *op_errstr = gf_strdup ("Failed to get volume options help");
+ out:
+
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_to_cli (rpcsvc_request_t *req, gf_cli_rsp *arg, struct iovec *payload,
+                 int payloadcount, struct iobref *iobref, xdrproc_t xdrproc,
+                 dict_t *dict)
+{
+        int                ret = -1;
+        char               *cmd = NULL;
+        int                op_ret = 0;
+        char               *op_errstr = NULL;
+        int                op_errno = 0;
+        xlator_t           *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        op_ret = arg->op_ret;
+        op_errstr = arg->op_errstr;
+        op_errno = arg->op_errno;
+
+        ret = dict_get_str (dict, "cmd-str", &cmd);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get command "
+                        "string");
+
+        if (cmd) {
+                if (op_ret)
+                        gf_cmd_log ("", "%s : FAILED %s %s", cmd,
+                                       (op_errstr)? ":" : " ",
+                                       (op_errstr)? op_errstr : " ");
+                else
+                        gf_cmd_log ("", "%s : SUCCESS", cmd);
+        }
+
+        glusterd_submit_reply (req, arg, payload, payloadcount, iobref,
+                               (xdrproc_t) xdrproc);
+        if (dict)
+                dict_unref (dict);
+
+        return ret;
+}
+
+static int32_t
+glusterd_append_gsync_status (dict_t *dst, dict_t *src)
+{
+        int                ret = 0;
+        char               *stop_msg = NULL;
+
+        ret = dict_get_str (src, "gsync-status", &stop_msg);
+        if (ret) {
+                ret = 0;
+                goto out;
+        }
+
+        ret = dict_set_dynstr (dst, "gsync-status", gf_strdup (stop_msg));
+        if (ret) {
+                gf_log ("glusterd", GF_LOG_WARNING, "Unable to set the stop"
+                        "message in the ctx dictionary");
+                goto out;
+        }
+
+        ret = 0;
+ out:
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+
+}
+
+static int32_t
+glusterd_append_status_dicts (dict_t *dst, dict_t *src)
+{
+        int              dst_count = 0;
+        int              src_count = 0;
+        int              i = 0;
+        int              ret = 0;
+        char             mst[PATH_MAX] = {0,};
+        char             slv[PATH_MAX] = {0, };
+        char             sts[PATH_MAX] = {0, };
+        char             nds[PATH_MAX] = {0, };
+        char             *mst_val = NULL;
+        char             *slv_val = NULL;
+        char             *sts_val = NULL;
+        char             *nds_val = NULL;
+
+        GF_ASSERT (dst);
+
+        if (src == NULL)
+                goto out;
+
+        ret = dict_get_int32 (dst, "gsync-count", &dst_count);
+        if (ret)
+                dst_count = 0;
+
+        ret = dict_get_int32 (src, "gsync-count", &src_count);
+        if (ret || !src_count) {
+                gf_log ("", GF_LOG_DEBUG, "Source brick empty");
+                ret = 0;
+                goto out;
+        }
+
+        for (i = 1; i <= src_count; i++) {
+                snprintf (nds, sizeof(nds), "node%d", i);
+                snprintf (mst, sizeof(mst), "master%d", i);
+                snprintf (slv, sizeof(slv), "slave%d", i);
+                snprintf (sts, sizeof(sts), "status%d", i);
+
+                ret = dict_get_str (src, nds, &nds_val);
+                if (ret)
+                        goto out;
+
+                ret = dict_get_str (src, mst, &mst_val);
+                if (ret)
+                        goto out;
+
+                ret = dict_get_str (src, slv, &slv_val);
+                if (ret)
+                        goto out;
+
+                ret = dict_get_str (src, sts, &sts_val);
+                if (ret)
+                        goto out;
+
+                snprintf (nds, sizeof(nds), "node%d", i+dst_count);
+                snprintf (mst, sizeof(mst), "master%d", i+dst_count);
+                snprintf (slv, sizeof(slv), "slave%d", i+dst_count);
+                snprintf (sts, sizeof(sts), "status%d", i+dst_count);
+
+                ret = dict_set_dynstr (dst, nds, gf_strdup (nds_val));
+                if (ret)
+                        goto out;
+
+                ret = dict_set_dynstr (dst, mst, gf_strdup (mst_val));
+                if (ret)
+                        goto out;
+
+                ret = dict_set_dynstr (dst, slv, gf_strdup (slv_val));
+                if (ret)
+                        goto out;
+
+                ret = dict_set_dynstr (dst, sts, gf_strdup (sts_val));
+                if (ret)
+                        goto out;
+
+        }
+
+        ret = dict_set_int32 (dst, "gsync-count", dst_count+src_count);
+
+ out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+
+}
+
+int32_t
+glusterd_gsync_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict, char *op_errstr)
+{
+        dict_t             *ctx = NULL;
+        int                ret = 0;
+        char               *conf_path = NULL;
+
+        if (aggr) {
+                ctx = aggr;
+
+        } else {
+                ctx = glusterd_op_get_ctx ();
+                if (!ctx) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Operation Context is not present");
+                        GF_ASSERT (0);
+                }
+        }
+
+        if (rsp_dict) {
+                ret = glusterd_append_status_dicts (ctx, rsp_dict);
+                if (ret)
+                        goto out;
+
+                ret = glusterd_append_gsync_status (ctx, rsp_dict);
+                if (ret)
+                        goto out;
+
+                ret = dict_get_str (rsp_dict, "conf_path", &conf_path);
+                if (!ret && conf_path) {
+                        ret = dict_set_dynstr (ctx, "conf_path",
+                                            gf_strdup(conf_path));
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to store conf path.");
+                                goto out;
+                        }
+                }
+        }
+        if ((op_errstr) && (strcmp ("", op_errstr))) {
+                ret = dict_set_dynstr (ctx, "errstr", gf_strdup(op_errstr));
+                if (ret)
+                        goto out;
+        }
+
+        ret = 0;
+ out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d ", ret);
+        return ret;
+}
+
+int32_t
+glusterd_rb_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict)
+{
+        int32_t  src_port = 0;
+        int32_t  dst_port = 0;
+        int      ret      = 0;
+        dict_t  *ctx      = NULL;
+
+
+        if (aggr) {
+                ctx = aggr;
+
+        } else {
+                ctx = glusterd_op_get_ctx ();
+                if (!ctx) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Operation Context is not present");
+                        GF_ASSERT (0);
+                }
+        }
+
+        if (rsp_dict) {
+                ret = dict_get_int32 (rsp_dict, "src-brick-port", &src_port);
+                if (ret == 0) {
+                        gf_log ("", GF_LOG_DEBUG,
+                                "src-brick-port=%d found", src_port);
+                }
+
+                ret = dict_get_int32 (rsp_dict, "dst-brick-port", &dst_port);
+                if (ret == 0) {
+                        gf_log ("", GF_LOG_DEBUG,
+                                "dst-brick-port=%d found", dst_port);
+                }
+
+        }
+
+        if (src_port) {
+                ret = dict_set_int32 (ctx, "src-brick-port",
+                                      src_port);
+                if (ret) {
+                        gf_log ("", GF_LOG_DEBUG,
+                                "Could not set src-brick");
+                        goto out;
+                }
+        }
+
+        if (dst_port) {
+                ret = dict_set_int32 (ctx, "dst-brick-port",
+                                      dst_port);
+                if (ret) {
+                        gf_log ("", GF_LOG_DEBUG,
+                                "Could not set dst-brick");
+                        goto out;
+                }
+
+        }
+
+out:
+        return ret;
+
+}
+
+int32_t
+glusterd_sync_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict)
+{
+        int      ret      = 0;
+
+        GF_ASSERT (rsp_dict);
+
+        if (!rsp_dict) {
+                goto out;
+        }
+
+        ret = glusterd_import_friend_volumes (rsp_dict);
+out:
+        return ret;
+
+}
+
+static int
+_profile_volume_add_friend_rsp (dict_t *this, char *key, data_t *value,
+                               void *data)
+{
+        char    new_key[256] = {0};
+        glusterd_pr_brick_rsp_conv_t *rsp_ctx = NULL;
+        data_t  *new_value = NULL;
+        int     brick_count = 0;
+        char    brick_key[256];
+
+        if (strcmp (key, "count") == 0)
+                return 0;
+        sscanf (key, "%d%s", &brick_count, brick_key);
+        rsp_ctx = data;
+        new_value = data_copy (value);
+        GF_ASSERT (new_value);
+        snprintf (new_key, sizeof (new_key), "%d%s",
+                  rsp_ctx->count + brick_count, brick_key);
+        dict_set (rsp_ctx->dict, new_key, new_value);
+        return 0;
+}
+
+int
+glusterd_profile_volume_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict)
+{
+        int     ret = 0;
+        glusterd_pr_brick_rsp_conv_t rsp_ctx = {0};
+        int32_t brick_count = 0;
+        int32_t count = 0;
+        dict_t  *ctx_dict = NULL;
+        glusterd_op_t   op = GD_OP_NONE;
+
+        GF_ASSERT (rsp_dict);
+
+        ret = dict_get_int32 (rsp_dict, "count", &brick_count);
+        if (ret) {
+                ret = 0; //no bricks in the rsp
+                goto out;
+        }
+
+        op = glusterd_op_get_op ();
+        GF_ASSERT (GD_OP_PROFILE_VOLUME == op);
+        if (aggr) {
+                ctx_dict = aggr;
+
+        } else {
+                ctx_dict = glusterd_op_get_ctx ();
+        }
+
+        ret = dict_get_int32 (ctx_dict, "count", &count);
+        rsp_ctx.count = count;
+        rsp_ctx.dict = ctx_dict;
+        dict_foreach (rsp_dict, _profile_volume_add_friend_rsp, &rsp_ctx);
+        dict_del (ctx_dict, "count");
+        ret = dict_set_int32 (ctx_dict, "count", count + brick_count);
+out:
+        return ret;
+}
+
+static int
+glusterd_volume_status_add_peer_rsp (dict_t *this, char *key, data_t *value,
+                                     void *data)
+{
+        glusterd_status_rsp_conv_t      *rsp_ctx = NULL;
+        data_t                          *new_value = NULL;
+        char                            brick_key[1024] = {0,};
+        char                            new_key[1024] = {0,};
+        int32_t                         index = 0;
+        int32_t                         ret = 0;
+
+        /* Skip the following keys, they are already present in the ctx_dict */
+        if (!strcmp (key, "count") || !strcmp (key, "cmd") ||
+            !strcmp (key, "brick-index-max") || !strcmp (key, "other-count"))
+                return 0;
+
+        rsp_ctx = data;
+        new_value = data_copy (value);
+        GF_ASSERT (new_value);
+
+        sscanf (key, "brick%d.%s", &index, brick_key);
+
+        if (index > rsp_ctx->brick_index_max) {
+                snprintf (new_key, sizeof (new_key), "brick%d.%s",
+                          index + rsp_ctx->other_count, brick_key);
+        } else {
+                strncpy (new_key, key, sizeof (new_key));
+                new_key[sizeof (new_key) - 1] = 0;
+        }
+
+        ret = dict_set (rsp_ctx->dict, new_key, new_value);
+        if (ret)
+                gf_log ("", GF_LOG_ERROR, "Unable to set key: %s in dict",
+                        key);
+
+        return 0;
+}
+
+int
+glusterd_volume_status_copy_to_op_ctx_dict (dict_t *aggr, dict_t *rsp_dict)
+{
+        int                             ret = 0;
+        glusterd_status_rsp_conv_t      rsp_ctx = {0};
+        int32_t                         cmd = GF_CLI_STATUS_NONE;
+        int32_t                         node_count = 0;
+        int32_t                         other_count = 0;
+        int32_t                         brick_index_max = -1;
+        int32_t                         rsp_node_count = 0;
+        int32_t                         rsp_other_count = 0;
+        int                             vol_count = -1;
+        int                             i = 0;
+        dict_t                          *ctx_dict = NULL;
+        char                            key[PATH_MAX] = {0,};
+        char                            *volname = NULL;
+
+        GF_ASSERT (rsp_dict);
+
+        if (aggr) {
+                ctx_dict = aggr;
+
+        } else {
+                ctx_dict = glusterd_op_get_ctx (GD_OP_STATUS_VOLUME);
+
+        }
+
+        ret = dict_get_int32 (ctx_dict, "cmd", &cmd);
+        if (ret)
+                goto out;
+
+        if (cmd & GF_CLI_STATUS_ALL && is_origin_glusterd (ctx_dict)) {
+                ret = dict_get_int32 (rsp_dict, "vol_count", &vol_count);
+                if (ret == 0) {
+                        ret = dict_set_int32 (ctx_dict, "vol_count",
+                                              vol_count);
+                        if (ret)
+                                goto out;
+
+                        for (i = 0; i < vol_count; i++) {
+                                memset (key, 0, sizeof (key));
+                                snprintf (key, sizeof (key), "vol%d", i);
+                                ret = dict_get_str (rsp_dict, key, &volname);
+                                if (ret)
+                                        goto out;
+
+                                ret = dict_set_str (ctx_dict, key, volname);
+                                if (ret)
+                                        goto out;
+                        }
+                }
+        }
+
+        if ((cmd & GF_CLI_STATUS_TASKS) != 0) {
+                dict_copy (rsp_dict, aggr);
+                ret = 0;
+                goto out;
+        }
+
+        ret = dict_get_int32 (rsp_dict, "count", &rsp_node_count);
+        if (ret) {
+                ret = 0; //no bricks in the rsp
+                goto out;
+        }
+
+        ret = dict_get_int32 (rsp_dict, "other-count", &rsp_other_count);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "Failed to get other count from rsp_dict");
+                goto out;
+        }
+
+        ret = dict_get_int32 (ctx_dict, "count", &node_count);
+        ret = dict_get_int32 (ctx_dict, "other-count", &other_count);
+        if (!dict_get (ctx_dict, "brick-index-max")) {
+                ret = dict_get_int32 (rsp_dict, "brick-index-max", &brick_index_max);
+                if (ret)
+                        goto out;
+                ret = dict_set_int32 (ctx_dict, "brick-index-max", brick_index_max);
+                if (ret)
+                        goto out;
+
+        } else {
+                ret = dict_get_int32 (ctx_dict, "brick-index-max", &brick_index_max);
+        }
+
+        rsp_ctx.count = node_count;
+        rsp_ctx.brick_index_max = brick_index_max;
+        rsp_ctx.other_count = other_count;
+        rsp_ctx.dict = ctx_dict;
+
+        dict_foreach (rsp_dict, glusterd_volume_status_add_peer_rsp, &rsp_ctx);
+
+        ret = dict_set_int32 (ctx_dict, "count", node_count + rsp_node_count);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "Failed to update node count");
+                goto out;
+        }
+
+        ret = dict_set_int32 (ctx_dict, "other-count",
+                              (other_count + rsp_other_count));
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "Failed to update other-count");
+out:
+        return ret;
+}
+
+int
+glusterd_volume_rebalance_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict)
+{
+        char                 key[256]      = {0,};
+        char                *node_uuid     = NULL;
+        char                *node_uuid_str = NULL;
+        char                *volname       = NULL;
+        dict_t              *ctx_dict      = NULL;
+        double               elapsed_time  = 0;
+        glusterd_conf_t     *conf          = NULL;
+        glusterd_op_t        op            = GD_OP_NONE;
+        glusterd_peerinfo_t *peerinfo      = NULL;
+        glusterd_volinfo_t  *volinfo       = NULL;
+        int                  ret           = 0;
+        int32_t              index         = 0;
+        int32_t              count         = 0;
+        int32_t              current_index = 2;
+        int32_t              value32       = 0;
+        uint64_t             value         = 0;
+        char                *peer_uuid_str = NULL;
+
+        GF_ASSERT (rsp_dict);
+        conf = THIS->private;
+
+        op = glusterd_op_get_op ();
+        GF_ASSERT ((GD_OP_REBALANCE == op) ||
+                   (GD_OP_DEFRAG_BRICK_VOLUME == op));
+
+        if (aggr) {
+                ctx_dict = aggr;
+
+        } else {
+                ctx_dict = glusterd_op_get_ctx (op);
+
+        }
+
+        if (!ctx_dict)
+                goto out;
+
+        ret = dict_get_str (ctx_dict, "volname", &volname);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret  = glusterd_volinfo_find (volname, &volinfo);
+
+        if (ret)
+                goto out;
+
+        ret = dict_get_int32 (rsp_dict, "count", &index);
+        if (ret)
+                gf_log ("", GF_LOG_ERROR, "failed to get index");
+
+        memset (key, 0, 256);
+        snprintf (key, 256, "node-uuid-%d", index);
+        ret = dict_get_str (rsp_dict, key, &node_uuid);
+        if (!ret) {
+                node_uuid_str = gf_strdup (node_uuid);
+
+                /* Finding the index of the node-uuid in the peer-list */
+                list_for_each_entry (peerinfo, &conf->peers, uuid_list) {
+                        peer_uuid_str = gd_peer_uuid_str (peerinfo);
+                        if (strcmp (peer_uuid_str, node_uuid_str) == 0)
+                                break;
+
+                        current_index++;
+                }
+
+                /* Setting the largest index value as the total count. */
+                ret = dict_get_int32 (ctx_dict, "count", &count);
+                if (count < current_index) {
+                    ret = dict_set_int32 (ctx_dict, "count", current_index);
+                    if (ret)
+                            gf_log ("", GF_LOG_ERROR, "Failed to set count");
+                }
+
+                /* Setting the same index for the node, as is in the peerlist.*/
+                memset (key, 0, 256);
+                snprintf (key, 256, "node-uuid-%d", current_index);
+                ret = dict_set_dynstr (ctx_dict, key, node_uuid_str);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_DEBUG,
+                                "failed to set node-uuid");
+                }
+        }
+
+        snprintf (key, 256, "files-%d", index);
+        ret = dict_get_uint64 (rsp_dict, key, &value);
+        if (!ret) {
+                memset (key, 0, 256);
+                snprintf (key, 256, "files-%d", current_index);
+                ret = dict_set_uint64 (ctx_dict, key, value);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_DEBUG,
+                                "failed to set the file count");
+                }
+        }
+
+        memset (key, 0, 256);
+        snprintf (key, 256, "size-%d", index);
+        ret = dict_get_uint64 (rsp_dict, key, &value);
+        if (!ret) {
+                memset (key, 0, 256);
+                snprintf (key, 256, "size-%d", current_index);
+                ret = dict_set_uint64 (ctx_dict, key, value);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_DEBUG,
+                                "failed to set the size of migration");
+                }
+        }
+
+        memset (key, 0, 256);
+        snprintf (key, 256, "lookups-%d", index);
+        ret = dict_get_uint64 (rsp_dict, key, &value);
+        if (!ret) {
+                memset (key, 0, 256);
+                snprintf (key, 256, "lookups-%d", current_index);
+                ret = dict_set_uint64 (ctx_dict, key, value);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_DEBUG,
+                                "failed to set lookuped file count");
+                }
+        }
+
+        memset (key, 0, 256);
+        snprintf (key, 256, "status-%d", index);
+        ret = dict_get_int32 (rsp_dict, key, &value32);
+        if (!ret) {
+                memset (key, 0, 256);
+                snprintf (key, 256, "status-%d", current_index);
+                ret = dict_set_int32 (ctx_dict, key, value32);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_DEBUG,
+                                "failed to set status");
+                }
+        }
+
+        memset (key, 0, 256);
+        snprintf (key, 256, "failures-%d", index);
+        ret = dict_get_uint64 (rsp_dict, key, &value);
+        if (!ret) {
+                memset (key, 0, 256);
+                snprintf (key, 256, "failures-%d", current_index);
+                ret = dict_set_uint64 (ctx_dict, key, value);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_DEBUG,
+                                "failed to set failure count");
+                }
+        }
+
+        memset (key, 0, 256);
+        snprintf (key, 256, "skipped-%d", index);
+        ret = dict_get_uint64 (rsp_dict, key, &value);
+        if (!ret) {
+                memset (key, 0, 256);
+                snprintf (key, 256, "skipped-%d", current_index);
+                ret = dict_set_uint64 (ctx_dict, key, value);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_DEBUG,
+                                "failed to set skipped count");
+                }
+        }
+        memset (key, 0, 256);
+        snprintf (key, 256, "run-time-%d", index);
+        ret = dict_get_double (rsp_dict, key, &elapsed_time);
+        if (!ret) {
+                memset (key, 0, 256);
+                snprintf (key, 256, "run-time-%d", current_index);
+                ret = dict_set_double (ctx_dict, key, elapsed_time);
+                if (ret) {
+                        gf_log (THIS->name, GF_LOG_DEBUG,
+                                "failed to set run-time");
+                }
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+int
+glusterd_snap_config_use_rsp_dict (dict_t *dst, dict_t *src)
+{
+        char           buf[PATH_MAX]        = "";
+        char          *volname              = NULL;
+        int            ret                  = -1;
+        int            config_command       = 0;
+        uint64_t       i                    = 0;
+        uint64_t       value                = 0;
+        uint64_t       voldisplaycount      = 0;
+
+        if (!dst || !src) {
+                gf_log ("", GF_LOG_ERROR, "Source or Destination "
+                        "dict is empty.");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dst, "config-command", &config_command);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR,
+                        "failed to get config-command type");
+                goto out;
+        }
+
+        switch (config_command) {
+        case GF_SNAP_CONFIG_DISPLAY:
+                ret = dict_get_uint64 (src, "snap-max-hard-limit", &value);
+                if (!ret) {
+                        ret = dict_set_uint64 (dst, "snap-max-hard-limit", value);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to set snap_max_hard_limit");
+                                goto out;
+                        }
+                } else {
+                        /* Received dummy response from other nodes */
+                        ret = 0;
+                        goto out;
+                }
+
+                ret = dict_get_uint64 (src, "snap-max-soft-limit", &value);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Unable to get snap_max_soft_limit");
+                        goto out;
+                }
+
+                ret = dict_set_uint64 (dst, "snap-max-soft-limit", value);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Unable to set snap_max_soft_limit");
+                        goto out;
+                }
+
+                ret = dict_get_uint64 (src, "voldisplaycount",
+                                       &voldisplaycount);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Unable to get voldisplaycount");
+                        goto out;
+                }
+
+                ret = dict_set_uint64 (dst, "voldisplaycount",
+                                       voldisplaycount);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "Unable to set voldisplaycount");
+                        goto out;
+                }
+
+                for (i = 0; i < voldisplaycount; i++) {
+                        snprintf (buf, sizeof(buf), "volume%ld-volname", i);
+                        ret = dict_get_str (src, buf, &volname);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to get %s", buf);
+                                goto out;
+                        }
+                        ret = dict_set_str (dst, buf, volname);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to set %s", buf);
+                                goto out;
+                        }
+
+                        snprintf (buf, sizeof(buf),
+                                  "volume%ld-snap-max-hard-limit", i);
+                        ret = dict_get_uint64 (src, buf, &value);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to get %s", buf);
+                                goto out;
+                        }
+                        ret = dict_set_uint64 (dst, buf, value);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to set %s", buf);
+                                goto out;
+                        }
+
+                        snprintf (buf, sizeof(buf),
+                                  "volume%ld-active-hard-limit", i);
+                        ret = dict_get_uint64 (src, buf, &value);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to get %s", buf);
+                                goto out;
+                        }
+                        ret = dict_set_uint64 (dst, buf, value);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to set %s", buf);
+                                goto out;
+                        }
+
+                        snprintf (buf, sizeof(buf),
+                                  "volume%ld-snap-max-soft-limit", i);
+                        ret = dict_get_uint64 (src, buf, &value);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to get %s", buf);
+                                goto out;
+                        }
+                        ret = dict_set_uint64 (dst, buf, value);
+                        if (ret) {
+                                gf_log ("", GF_LOG_ERROR,
+                                        "Unable to set %s", buf);
+                                goto out;
+                        }
+                }
+
+                break;
+        default:
+                break;
+        }
+
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+/* Aggregate missed_snap_counts from different nodes and save it *
+ * in the req_dict of the originator node */
+int
+glusterd_snap_create_use_rsp_dict (dict_t *dst, dict_t *src)
+{
+        char          *buf                      = NULL;
+        char          *tmp_str                  = NULL;
+        char           name_buf[PATH_MAX]       = "";
+        int32_t        i                        = -1;
+        int32_t        ret                      = -1;
+        int32_t        src_missed_snap_count    = -1;
+        int32_t        dst_missed_snap_count    = -1;
+        xlator_t      *this                     = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        if (!dst || !src) {
+                gf_log (this->name, GF_LOG_ERROR, "Source or Destination "
+                        "dict is empty.");
+                goto out;
+        }
+
+        ret = dict_get_int32 (src, "missed_snap_count",
+                              &src_missed_snap_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "No missed snaps");
+                ret = 0;
+                goto out;
+        }
+
+        ret = dict_get_int32 (dst, "missed_snap_count",
+                              &dst_missed_snap_count);
+        if (ret) {
+                /* Initialize dst_missed_count for the first time */
+                dst_missed_snap_count = 0;
+        }
+
+        for (i = 0; i < src_missed_snap_count; i++) {
+                 snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d",
+                           i);
+                 ret = dict_get_str (src, name_buf, &buf);
+                 if (ret) {
+                         gf_log (this->name, GF_LOG_ERROR,
+                                 "Unable to fetch %s", name_buf);
+                         goto out;
+                 }
+
+                 snprintf (name_buf, sizeof(name_buf), "missed_snaps_%d",
+                           dst_missed_snap_count);
+
+                 tmp_str = gf_strdup (buf);
+                 if (!tmp_str) {
+                         ret = -1;
+                         goto out;
+                 }
+
+                 ret = dict_set_dynstr (dst, name_buf, tmp_str);
+                 if (ret) {
+                         gf_log (this->name, GF_LOG_ERROR,
+                                 "Unable to set %s", name_buf);
+                         goto out;
+                 }
+
+                 tmp_str = NULL;
+                 dst_missed_snap_count++;
+        }
+
+        ret = dict_set_int32 (dst, "missed_snap_count", dst_missed_snap_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to set dst_missed_snap_count");
+                goto out;
+        }
+
+out:
+        if (ret && tmp_str)
+                GF_FREE(tmp_str);
+
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_snap_use_rsp_dict (dict_t *dst, dict_t *src)
+{
+        int            ret            = -1;
+        int32_t        snap_command   = 0;
+
+        if (!dst || !src) {
+                gf_log ("", GF_LOG_ERROR, "Source or Destination "
+                        "dict is empty.");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dst, "type", &snap_command);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "unable to get the type of "
+                        "the snapshot command");
+                goto out;
+        }
+
+        switch (snap_command) {
+        case GF_SNAP_OPTION_TYPE_CREATE:
+        case GF_SNAP_OPTION_TYPE_DELETE:
+                ret = glusterd_snap_create_use_rsp_dict (dst, src);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to use rsp dict");
+                        goto out;
+                }
+                break;
+        case GF_SNAP_OPTION_TYPE_CONFIG:
+                ret = glusterd_snap_config_use_rsp_dict (dst, src);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to use rsp dict");
+                        goto out;
+                }
+                break;
+        default:
+                // copy the response dictinary's contents to the dict to be
+                // sent back to the cli
+                dict_copy (src, dst);
+                break;
+        }
+
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_sys_exec_output_rsp_dict (dict_t *dst, dict_t *src)
+{
+        char           output_name[PATH_MAX] = "";
+        char          *output = NULL;
+        int            ret      = 0;
+        int            i      = 0;
+        int            len    = 0;
+        int            src_output_count      = 0;
+        int            dst_output_count      = 0;
+
+        if (!dst || !src) {
+                gf_log ("", GF_LOG_ERROR, "Source or Destination "
+                        "dict is empty.");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dst, "output_count", &dst_output_count);
+
+        ret = dict_get_int32 (src, "output_count", &src_output_count);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "No output from source");
+                ret = 0;
+                goto out;
+        }
+
+        for (i = 1; i <= src_output_count; i++) {
+                len = snprintf (output_name, sizeof(output_name) - 1,
+                                "output_%d", i);
+                output_name[len] = '\0';
+                ret = dict_get_str (src, output_name, &output);
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to fetch %s",
+                                output_name);
+                        goto out;
+                }
+
+                len = snprintf (output_name, sizeof(output_name) - 1,
+                                "output_%d", i+dst_output_count);
+                output_name[len] = '\0';
+                ret = dict_set_dynstr (dst, output_name, gf_strdup (output));
+                if (ret) {
+                        gf_log ("", GF_LOG_ERROR, "Unable to set %s",
+                                output_name);
+                        goto out;
+                }
+        }
+
+        ret = dict_set_int32 (dst, "output_count",
+                              dst_output_count+src_output_count);
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict)
+{
+        int            ret      = 0;
+        glusterd_op_t  op       = GD_OP_NONE;
+
+        op = glusterd_op_get_op ();
+        GF_ASSERT (aggr);
+        GF_ASSERT (rsp_dict);
+
+        if (!aggr)
+                goto out;
+        dict_copy (rsp_dict, aggr);
+out:
+        return ret;
+}
+
+int
+glusterd_volume_heal_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict)
+{
+        int            ret      = 0;
+        dict_t        *ctx_dict = NULL;
+        glusterd_op_t  op       = GD_OP_NONE;
+
+        GF_ASSERT (rsp_dict);
+
+        op = glusterd_op_get_op ();
+        GF_ASSERT (GD_OP_HEAL_VOLUME == op);
+
+        if (aggr) {
+                ctx_dict = aggr;
+
+        } else {
+                ctx_dict = glusterd_op_get_ctx (op);
+        }
+
+        if (!ctx_dict)
+                goto out;
+        dict_copy (rsp_dict, ctx_dict);
+out:
+        return ret;
+}
+
+int
+_profile_volume_add_brick_rsp (dict_t *this, char *key, data_t *value,
+                             void *data)
+{
+        char    new_key[256] = {0};
+        glusterd_pr_brick_rsp_conv_t *rsp_ctx = NULL;
+        data_t  *new_value = NULL;
+
+        rsp_ctx = data;
+        new_value = data_copy (value);
+        GF_ASSERT (new_value);
+        snprintf (new_key, sizeof (new_key), "%d-%s", rsp_ctx->count, key);
+        dict_set (rsp_ctx->dict, new_key, new_value);
+        return 0;
+}
+
+int
+glusterd_profile_volume_brick_rsp (void *pending_entry,
+                                   dict_t *rsp_dict, dict_t *op_ctx,
+                                   char **op_errstr, gd_node_type type)
+{
+        int                             ret = 0;
+        glusterd_pr_brick_rsp_conv_t    rsp_ctx = {0};
+        int32_t                         count = 0;
+        char                            brick[PATH_MAX+1024] = {0};
+        char                            key[256] = {0};
+        char                            *full_brick = NULL;
+        glusterd_brickinfo_t            *brickinfo = NULL;
+        xlator_t                        *this = NULL;
+        glusterd_conf_t                 *priv = NULL;
+
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (op_ctx);
+        GF_ASSERT (op_errstr);
+        GF_ASSERT (pending_entry);
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_int32 (op_ctx, "count", &count);
+        if (ret) {
+                count = 1;
+        } else {
+                count++;
+        }
+        snprintf (key, sizeof (key), "%d-brick", count);
+        if (type == GD_NODE_BRICK) {
+                brickinfo = pending_entry;
+                snprintf (brick, sizeof (brick), "%s:%s", brickinfo->hostname,
+                          brickinfo->path);
+        } else if (type == GD_NODE_NFS) {
+                snprintf (brick, sizeof (brick), "%s", uuid_utoa (MY_UUID));
+        }
+        full_brick = gf_strdup (brick);
+        GF_ASSERT (full_brick);
+        ret = dict_set_dynstr (op_ctx, key, full_brick);
+
+        rsp_ctx.count = count;
+        rsp_ctx.dict = op_ctx;
+        dict_foreach (rsp_dict, _profile_volume_add_brick_rsp, &rsp_ctx);
+        dict_del (op_ctx, "count");
+        ret = dict_set_int32 (op_ctx, "count", count);
+        return ret;
+}
+
+//input-key: <replica-id>:<child-id>-*
+//output-key: <brick-id>-*
+int
+_heal_volume_add_shd_rsp (dict_t *this, char *key, data_t *value, void *data)
+{
+        char                            new_key[256] = {0,};
+        char                            int_str[16] = {0};
+        data_t                          *new_value = NULL;
+        char                            *rxl_end = NULL;
+        char                            *rxl_child_end = NULL;
+        glusterd_volinfo_t              *volinfo = NULL;
+        int                             rxl_id = 0;
+        int                             rxl_child_id = 0;
+        int                             brick_id = 0;
+        int                             int_len = 0;
+        int                             ret = 0;
+        glusterd_heal_rsp_conv_t        *rsp_ctx = NULL;
+        glusterd_brickinfo_t            *brickinfo = NULL;
+
+        rsp_ctx = data;
+        rxl_end = strchr (key, '-');
+        if (!rxl_end)
+                goto out;
+
+        int_len = strlen (key) - strlen (rxl_end);
+        strncpy (int_str, key, int_len);
+        int_str[int_len] = '\0';
+        ret = gf_string2int (int_str, &rxl_id);
+        if (ret)
+                goto out;
+
+        rxl_child_end = strchr (rxl_end + 1, '-');
+        if (!rxl_child_end)
+                goto out;
+
+        int_len = strlen (rxl_end) - strlen (rxl_child_end) - 1;
+        strncpy (int_str, rxl_end + 1, int_len);
+        int_str[int_len] = '\0';
+        ret = gf_string2int (int_str, &rxl_child_id);
+        if (ret)
+                goto out;
+
+        volinfo = rsp_ctx->volinfo;
+        brick_id = rxl_id * volinfo->replica_count + rxl_child_id;
+
+        if (!strcmp (rxl_child_end, "-status")) {
+                brickinfo = glusterd_get_brickinfo_by_position (volinfo,
+                                                                brick_id);
+                if (!brickinfo)
+                        goto out;
+                if (!glusterd_is_local_brick (rsp_ctx->this, volinfo,
+                                              brickinfo))
+                        goto out;
+        }
+        new_value = data_copy (value);
+        snprintf (new_key, sizeof (new_key), "%d%s", brick_id, rxl_child_end);
+        dict_set (rsp_ctx->dict, new_key, new_value);
+
+out:
+        return 0;
+}
+
+int
+_heal_volume_add_shd_rsp_of_statistics (dict_t *this, char *key, data_t
+                                             *value, void *data)
+{
+        char                            new_key[256] = {0,};
+        char                            int_str[16] = {0,};
+        char                            key_begin_string[128] = {0,};
+        data_t                          *new_value = NULL;
+        char                            *rxl_end = NULL;
+        char                            *rxl_child_end = NULL;
+        glusterd_volinfo_t              *volinfo = NULL;
+        char                            *key_begin_str = NULL;
+        int                             rxl_id = 0;
+        int                             rxl_child_id = 0;
+        int                             brick_id = 0;
+        int                             int_len = 0;
+        int                             ret = 0;
+        glusterd_heal_rsp_conv_t        *rsp_ctx = NULL;
+        glusterd_brickinfo_t            *brickinfo = NULL;
+
+        rsp_ctx = data;
+        key_begin_str = strchr (key, '-');
+        if (!key_begin_str)
+                goto out;
+
+        int_len = strlen (key) - strlen (key_begin_str);
+        strncpy (key_begin_string, key, int_len);
+        key_begin_string[int_len] = '\0';
+
+        rxl_end = strchr (key_begin_str + 1, '-');
+        if (!rxl_end)
+                goto out;
+
+        int_len = strlen (key_begin_str) - strlen (rxl_end) - 1;
+        strncpy (int_str, key_begin_str + 1, int_len);
+        int_str[int_len] = '\0';
+        ret = gf_string2int (int_str, &rxl_id);
+        if (ret)
+                goto out;
+
+
+        rxl_child_end = strchr (rxl_end + 1, '-');
+        if (!rxl_child_end)
+                goto out;
+
+        int_len = strlen (rxl_end) - strlen (rxl_child_end) - 1;
+        strncpy (int_str, rxl_end + 1, int_len);
+        int_str[int_len] = '\0';
+        ret = gf_string2int (int_str, &rxl_child_id);
+        if (ret)
+                goto out;
+
+        volinfo = rsp_ctx->volinfo;
+        brick_id = rxl_id * volinfo->replica_count + rxl_child_id;
+
+        brickinfo = glusterd_get_brickinfo_by_position (volinfo, brick_id);
+        if (!brickinfo)
+                goto out;
+        if (!glusterd_is_local_brick (rsp_ctx->this, volinfo, brickinfo))
+                goto out;
+
+        new_value = data_copy (value);
+        snprintf (new_key, sizeof (new_key), "%s-%d%s", key_begin_string,
+                  brick_id, rxl_child_end);
+        dict_set (rsp_ctx->dict, new_key, new_value);
+
+out:
+        return 0;
+
+}
+
+int
+glusterd_heal_volume_brick_rsp (dict_t *req_dict, dict_t *rsp_dict,
+                                dict_t *op_ctx, char **op_errstr)
+{
+        int                             ret = 0;
+        glusterd_heal_rsp_conv_t        rsp_ctx = {0};
+        char                            *volname = NULL;
+        glusterd_volinfo_t              *volinfo = NULL;
+        int                             heal_op = -1;
+
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (op_ctx);
+        GF_ASSERT (op_errstr);
+
+        ret = dict_get_str (req_dict, "volname", &volname);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret = dict_get_int32 (req_dict, "heal-op", &heal_op);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get heal_op");
+                goto out;
+        }
+
+
+        ret  = glusterd_volinfo_find (volname, &volinfo);
+
+        if (ret)
+                goto out;
+
+        rsp_ctx.dict = op_ctx;
+        rsp_ctx.volinfo = volinfo;
+        rsp_ctx.this = THIS;
+        if (heal_op == GF_AFR_OP_STATISTICS)
+                dict_foreach (rsp_dict, _heal_volume_add_shd_rsp_of_statistics,
+                              &rsp_ctx);
+        else
+                dict_foreach (rsp_dict, _heal_volume_add_shd_rsp, &rsp_ctx);
+
+
+out:
+        return ret;
+}
+
+int
+_status_volume_add_brick_rsp (dict_t *this, char *key, data_t *value,
+                              void *data)
+{
+        char                            new_key[256] = {0,};
+        data_t                          *new_value = 0;
+        glusterd_pr_brick_rsp_conv_t    *rsp_ctx = NULL;
+
+        rsp_ctx = data;
+        new_value = data_copy (value);
+        snprintf (new_key, sizeof (new_key), "brick%d.%s", rsp_ctx->count, key);
+        dict_set (rsp_ctx->dict, new_key, new_value);
+
+        return 0;
+}
+
+int
+glusterd_status_volume_brick_rsp (dict_t *rsp_dict, dict_t *op_ctx,
+                                  char **op_errstr)
+{
+        int                             ret = 0;
+        glusterd_pr_brick_rsp_conv_t    rsp_ctx = {0};
+        int32_t                         count = 0;
+        int                             index = 0;
+
+        GF_ASSERT (rsp_dict);
+        GF_ASSERT (op_ctx);
+        GF_ASSERT (op_errstr);
+
+        ret = dict_get_int32 (op_ctx, "count", &count);
+        if (ret) {
+                count = 0;
+        } else {
+                count++;
+        }
+        ret = dict_get_int32 (rsp_dict, "index", &index);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Couldn't get node index");
+                goto out;
+        }
+        dict_del (rsp_dict, "index");
+
+        rsp_ctx.count = index;
+        rsp_ctx.dict = op_ctx;
+        dict_foreach (rsp_dict, _status_volume_add_brick_rsp, &rsp_ctx);
+        ret = dict_set_int32 (op_ctx, "count", count);
+
+out:
+        return ret;
+}
+
+int
+glusterd_defrag_volume_node_rsp (dict_t *req_dict, dict_t *rsp_dict,
+                                 dict_t *op_ctx)
+{
+        int                             ret = 0;
+        char                            *volname = NULL;
+        glusterd_volinfo_t              *volinfo = NULL;
+        char                            key[256] = {0,};
+        int32_t                         i = 0;
+        char                            buf[1024] = {0,};
+        char                            *node_str = NULL;
+        glusterd_conf_t                 *priv = NULL;
+
+        priv = THIS->private;
+        GF_ASSERT (req_dict);
+
+        ret = dict_get_str (req_dict, "volname", &volname);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret  = glusterd_volinfo_find (volname, &volinfo);
+
+        if (ret)
+                goto out;
+
+        if (rsp_dict) {
+                ret = glusterd_defrag_volume_status_update (volinfo,
+                                                            rsp_dict);
+        }
+
+        if (!op_ctx) {
+                dict_copy (rsp_dict, op_ctx);
+                goto out;
+        }
+
+        ret = dict_get_int32 (op_ctx, "count", &i);
+        i++;
+
+        ret = dict_set_int32 (op_ctx, "count", i);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to set count");
+
+        snprintf (buf, 1024, "%s", uuid_utoa (MY_UUID));
+        node_str = gf_strdup (buf);
+
+        snprintf (key, 256, "node-uuid-%d",i);
+        ret = dict_set_dynstr (op_ctx, key, node_str);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "failed to set node-uuid");
+
+        memset (key, 0 , 256);
+        snprintf (key, 256, "files-%d", i);
+        ret = dict_set_uint64 (op_ctx, key, volinfo->rebal.rebalance_files);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "failed to set file count");
+
+        memset (key, 0 , 256);
+        snprintf (key, 256, "size-%d", i);
+        ret = dict_set_uint64 (op_ctx, key, volinfo->rebal.rebalance_data);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "failed to set size of xfer");
+
+        memset (key, 0 , 256);
+        snprintf (key, 256, "lookups-%d", i);
+        ret = dict_set_uint64 (op_ctx, key, volinfo->rebal.lookedup_files);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "failed to set lookedup file count");
+
+        memset (key, 0 , 256);
+        snprintf (key, 256, "status-%d", i);
+        ret = dict_set_int32 (op_ctx, key, volinfo->rebal.defrag_status);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "failed to set status");
+
+        memset (key, 0 , 256);
+        snprintf (key, 256, "failures-%d", i);
+        ret = dict_set_uint64 (op_ctx, key, volinfo->rebal.rebalance_failures);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "failed to set failure count");
+
+        memset (key, 0 , 256);
+        snprintf (key, 256, "skipped-%d", i);
+        ret = dict_set_uint64 (op_ctx, key, volinfo->rebal.skipped_files);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "failed to set skipped count");
+
+        memset (key, 0, 256);
+        snprintf (key, 256, "run-time-%d", i);
+        ret = dict_set_double (op_ctx, key, volinfo->rebal.rebalance_time);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "failed to set run-time");
+
+out:
+        return ret;
+}
+int32_t
+glusterd_handle_node_rsp (dict_t *req_dict, void *pending_entry,
+                          glusterd_op_t op, dict_t *rsp_dict, dict_t *op_ctx,
+                          char **op_errstr, gd_node_type type)
+{
+        int                     ret = 0;
+
+        GF_ASSERT (op_errstr);
+
+        switch (op) {
+        case GD_OP_PROFILE_VOLUME:
+                ret = glusterd_profile_volume_brick_rsp (pending_entry,
+                                                         rsp_dict, op_ctx,
+                                                         op_errstr, type);
+                break;
+        case GD_OP_STATUS_VOLUME:
+                ret = glusterd_status_volume_brick_rsp (rsp_dict, op_ctx,
+                                                        op_errstr);
+                break;
+
+        case GD_OP_DEFRAG_BRICK_VOLUME:
+                glusterd_defrag_volume_node_rsp (req_dict,
+                                                 rsp_dict, op_ctx);
+                break;
+
+        case GD_OP_HEAL_VOLUME:
+                ret = glusterd_heal_volume_brick_rsp (req_dict, rsp_dict,
+                                                      op_ctx, op_errstr);
+                break;
+        default:
+                break;
+        }
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+/* Should be used only when an operation is in progress, as that is the only
+ * time a lock_owner is set
+ */
+gf_boolean_t
+is_origin_glusterd (dict_t *dict)
+{
+        gf_boolean_t  ret              = _gf_false;
+        uuid_t        lock_owner       = {0,};
+        uuid_t        *originator_uuid = NULL;
+
+        GF_ASSERT (dict);
+
+        ret = dict_get_bin (dict, "originator_uuid",
+                            (void **) &originator_uuid);
+        if (ret) {
+                /* If not originator_uuid has been set, then the command
+                 * has been originated from a glusterd running on older version
+                 * Hence fetching the lock owner */
+                ret = glusterd_get_lock_owner (&lock_owner);
+                if (ret) {
+                        ret = _gf_false;
+                        goto out;
+                }
+                ret = !uuid_compare (MY_UUID, lock_owner);
+        } else
+                ret = !uuid_compare (MY_UUID, *originator_uuid);
+
+out:
+        return ret;
+}
+
+int
+glusterd_generate_and_set_task_id (dict_t *dict, char *key)
+{
+        int             ret = -1;
+        uuid_t          task_id = {0,};
+        char            *uuid_str = NULL;
+        xlator_t        *this = NULL;
+
+        GF_ASSERT (dict);
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        uuid_generate (task_id);
+        uuid_str = gf_strdup (uuid_utoa (task_id));
+        if (!uuid_str) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_dynstr (dict, key, uuid_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set %s in dict",
+                        key);
+                goto out;
+        }
+        gf_log (this->name, GF_LOG_INFO, "Generated task-id %s for key %s",
+                uuid_str, key);
+
+out:
+        if (ret)
+                GF_FREE (uuid_str);
+        return ret;
+}
+
+int
+glusterd_copy_uuid_to_dict (uuid_t uuid, dict_t *dict, char *key)
+{
+        int             ret = -1;
+        char            tmp_str[40] = {0,};
+        char            *task_id_str = NULL;
+
+        GF_ASSERT (dict);
+        GF_ASSERT (key);
+
+        uuid_unparse (uuid, tmp_str);
+        task_id_str = gf_strdup (tmp_str);
+        if (!task_id_str)
+                return -1;
+
+        ret = dict_set_dynstr (dict, key, task_id_str);
+        if (ret) {
+                GF_FREE (task_id_str);
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "Error setting uuid in dict with key %s", key);
+        }
+
+        return 0;
+}
+
+int
+_update_volume_op_versions (dict_t *this, char *key, data_t *value, void *data)
+{
+        int                op_version = 0;
+        glusterd_volinfo_t *ctx       = NULL;
+        gf_boolean_t       enabled    = _gf_true;
+        int                ret        = -1;
+
+        GF_ASSERT (data);
+        ctx = data;
+
+        op_version = glusterd_get_op_version_for_key (key);
+
+        if (gd_is_xlator_option (key) || gd_is_boolean_option (key)) {
+                ret = gf_string2boolean (value->data, &enabled);
+                if (ret)
+                        return 0;
+
+                if (!enabled)
+                        return 0;
+        }
+
+        if (op_version > ctx->op_version)
+                ctx->op_version = op_version;
+
+        if (gd_is_client_option (key) &&
+            (op_version > ctx->client_op_version))
+                ctx->client_op_version = op_version;
+
+        return 0;
+}
+
+void
+gd_update_volume_op_versions (glusterd_volinfo_t *volinfo)
+{
+        glusterd_conf_t *conf = NULL;
+        gf_boolean_t    ob_enabled = _gf_false;
+
+        GF_ASSERT (volinfo);
+
+        conf = THIS->private;
+        GF_ASSERT (conf);
+
+        /* Reset op-versions to minimum */
+        volinfo->op_version = 1;
+        volinfo->client_op_version = 1;
+
+        dict_foreach (volinfo->dict, _update_volume_op_versions, volinfo);
+
+        /* Special case for open-behind
+         * If cluster op-version >= 2 and open-behind hasn't been explicitly
+         * disabled, volume op-versions must be updated to account for it
+         */
+
+        /* TODO: Remove once we have a general way to update automatically
+         * enabled features
+         */
+        if (conf->op_version >= 2) {
+                ob_enabled = dict_get_str_boolean (volinfo->dict,
+                                                   "performance.open-behind",
+                                                   _gf_true);
+                if (ob_enabled) {
+
+                        if (volinfo->op_version < 2)
+                                volinfo->op_version = 2;
+                        if (volinfo->client_op_version < 2)
+                                volinfo->client_op_version = 2;
+                }
+        }
+
+        return;
+}
+
+/* A task is committed/completed once the task-id for it is cleared */
+gf_boolean_t
+gd_is_remove_brick_committed (glusterd_volinfo_t *volinfo)
+{
+        GF_ASSERT (volinfo);
+
+        if ((GD_OP_REMOVE_BRICK == volinfo->rebal.op) &&
+            !uuid_is_null (volinfo->rebal.rebalance_id))
+                        return _gf_false;
+
+        return _gf_true;
+}
+
+gf_boolean_t
+glusterd_are_vol_all_peers_up (glusterd_volinfo_t *volinfo,
+                               struct list_head *peers,
+                               char **down_peerstr)
+{
+        glusterd_peerinfo_t   *peerinfo  = NULL;
+        glusterd_brickinfo_t  *brickinfo = NULL;
+        gf_boolean_t           ret       = _gf_false;
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                if (!uuid_compare (brickinfo->uuid, MY_UUID))
+                        continue;
+
+                list_for_each_entry (peerinfo, peers, uuid_list) {
+                        if (uuid_compare (peerinfo->uuid, brickinfo->uuid))
+                                continue;
+
+                        /*Found peer who owns the brick, return false
+                         * if peer is not connected or not friend */
+                        if (!(peerinfo->connected) ||
+                           (peerinfo->state.state !=
+                             GD_FRIEND_STATE_BEFRIENDED)) {
+                                *down_peerstr = gf_strdup (peerinfo->hostname);
+                                gf_log ("", GF_LOG_DEBUG, "Peer %s is down. ",
+                                        peerinfo->hostname);
+                                goto out;
+                        }
+                }
+       }
+
+        ret = _gf_true;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+gf_boolean_t
+glusterd_is_status_tasks_op (glusterd_op_t op, dict_t *dict)
+{
+        int           ret             = -1;
+        uint32_t      cmd             = GF_CLI_STATUS_NONE;
+        gf_boolean_t  is_status_tasks = _gf_false;
+
+        if (op != GD_OP_STATUS_VOLUME)
+                goto out;
+
+        ret = dict_get_uint32 (dict, "cmd", &cmd);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to get opcode");
+                goto out;
+        }
+
+        if (cmd & GF_CLI_STATUS_TASKS)
+                is_status_tasks = _gf_true;
+
+out:
+        return is_status_tasks;
+}
+
+int
+glusterd_compare_snap_time(struct list_head *list1, struct list_head *list2)
+{
+        glusterd_snap_t *snap1 = NULL;
+        glusterd_snap_t *snap2 = NULL;
+        double diff_time       = 0;
+
+        GF_ASSERT (list1);
+        GF_ASSERT (list2);
+
+        snap1 = list_entry(list1, glusterd_snap_t, snap_list);
+        snap2 = list_entry(list2, glusterd_snap_t, snap_list);
+        diff_time = difftime(snap1->time_stamp, snap2->time_stamp);
+
+        return ((int)diff_time);
+}
+
+int
+glusterd_compare_snap_vol_time(struct list_head *list1, struct list_head *list2)
+{
+        glusterd_volinfo_t *snapvol1 = NULL;
+        glusterd_volinfo_t *snapvol2 = NULL;
+        double diff_time             = 0;
+
+        GF_ASSERT (list1);
+        GF_ASSERT (list2);
+
+        snapvol1 = list_entry(list1, glusterd_volinfo_t, snapvol_list);
+        snapvol2 = list_entry(list2, glusterd_volinfo_t, snapvol_list);
+        diff_time = difftime(snapvol1->snapshot->time_stamp,
+                             snapvol2->snapshot->time_stamp);
+
+        return ((int)diff_time);
+}
+
+int32_t
+glusterd_missed_snapinfo_new (glusterd_missed_snap_info **missed_snapinfo)
+{
+        glusterd_missed_snap_info      *new_missed_snapinfo = NULL;
+        int32_t                         ret                 = -1;
+        xlator_t                       *this                = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (missed_snapinfo);
+
+        new_missed_snapinfo = GF_CALLOC (1, sizeof(*new_missed_snapinfo),
+                                         gf_gld_mt_missed_snapinfo_t);
+
+        if (!new_missed_snapinfo)
+                goto out;
+
+        new_missed_snapinfo->node_snap_info = NULL;
+        INIT_LIST_HEAD (&new_missed_snapinfo->missed_snaps);
+        INIT_LIST_HEAD (&new_missed_snapinfo->snap_ops);
+
+        *missed_snapinfo = new_missed_snapinfo;
+
+        ret = 0;
+
+out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
+        return ret;
+}
+
+int32_t
+glusterd_missed_snap_op_new (glusterd_snap_op_t **snap_op)
+{
+        glusterd_snap_op_t      *new_snap_op = NULL;
+        int32_t                  ret         = -1;
+        xlator_t                *this        = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (snap_op);
+
+        new_snap_op = GF_CALLOC (1, sizeof(*new_snap_op),
+                                 gf_gld_mt_missed_snapinfo_t);
+
+        if (!new_snap_op)
+                goto out;
+
+        new_snap_op->brick_path = NULL;
+        new_snap_op->brick_num = -1;
+        new_snap_op->op = -1;
+        new_snap_op->status = -1;
+        INIT_LIST_HEAD (&new_snap_op->snap_ops_list);
+
+        *snap_op = new_snap_op;
+
+        ret = 0;
 out:
+        gf_log (this->name, GF_LOG_TRACE, "Returning %d", ret);
         return ret;
 }
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h
index 72715a0f5..56bb799bf 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.h
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _GLUSTERD_UTILS_H
 #define _GLUSTERD_UTILS_H_
 
@@ -38,14 +28,33 @@
 #include "rpc-clnt.h"
 #include "protocol-common.h"
 
+#define GLUSTERD_SOCK_DIR "/var/run"
+
 struct glusterd_lock_ {
         uuid_t  owner;
         time_t  timestamp;
 };
 
+typedef struct glusterd_dict_ctx_ {
+        dict_t  *dict;
+        int     opt_count;
+        char    *key_name;
+        char    *val_name;
+        char    *prefix;
+} glusterd_dict_ctx_t;
+
+int
+glusterd_compare_lines (const void *a, const void *b);
+
+typedef int (*glusterd_condition_func) (glusterd_volinfo_t *volinfo,
+                                        glusterd_brickinfo_t *brickinfo,
+                                        void *ctx);
 typedef struct glusterd_lock_ glusterd_lock_t;
 
 int32_t
+glusterd_get_lock_owner (uuid_t *cur_owner);
+
+int32_t
 glusterd_lock (uuid_t new_owner);
 
 int32_t
@@ -57,18 +66,41 @@ glusterd_get_uuid (uuid_t *uuid);
 int
 glusterd_submit_reply (rpcsvc_request_t *req, void *arg,
                        struct iovec *payload, int payloadcount,
-                       struct iobref *iobref, gd_serialize_t sfunc);
+                       struct iobref *iobref, xdrproc_t xdrproc);
 
 int
-glusterd_submit_request (glusterd_peerinfo_t *peerinfo, void *req,
-                         call_frame_t *frame, struct rpc_clnt_program *prog,
-                         int procnum, struct iobref *iobref,
-                         gd_serialize_t sfunc, xlator_t *this,
-                         fop_cbk_fn_t cbkfn);
+glusterd_to_cli (rpcsvc_request_t *req, gf_cli_rsp *arg, struct iovec *payload,
+                 int payloadcount, struct iobref *iobref, xdrproc_t xdrproc,
+                 dict_t *dict);
 
+int
+glusterd_submit_request (struct rpc_clnt *rpc, void *req,
+                         call_frame_t *frame, rpc_clnt_prog_t *prog,
+                         int procnum, struct iobref *iobref,
+                         xlator_t *this, fop_cbk_fn_t cbkfn, xdrproc_t xdrproc);
 int32_t
 glusterd_volinfo_new (glusterd_volinfo_t **volinfo);
 
+int32_t
+glusterd_volinfo_dup (glusterd_volinfo_t *volinfo,
+                      glusterd_volinfo_t **dup_volinfo,
+                      gf_boolean_t set_userauth);
+
+char *
+glusterd_auth_get_username (glusterd_volinfo_t *volinfo);
+
+char *
+glusterd_auth_get_password (glusterd_volinfo_t *volinfo);
+
+int32_t
+glusterd_auth_set_username (glusterd_volinfo_t *volinfo, char *username);
+
+int32_t
+glusterd_auth_set_password (glusterd_volinfo_t *volinfo, char *password);
+
+void
+glusterd_auth_cleanup (glusterd_volinfo_t *volinfo);
+
 gf_boolean_t
 glusterd_check_volume_exists (char *volname);
 
@@ -76,7 +108,7 @@ int32_t
 glusterd_brickinfo_new (glusterd_brickinfo_t **brickinfo);
 
 int32_t
-glusterd_brickinfo_from_brick (char *brick, glusterd_brickinfo_t **brickinfo);
+glusterd_brickinfo_new_from_brick (char *brick, glusterd_brickinfo_t **brickinfo);
 
 int32_t
 glusterd_friend_cleanup (glusterd_peerinfo_t *peerinfo);
@@ -88,11 +120,26 @@ int32_t
 glusterd_peer_hostname_new (char *hostname, glusterd_peer_hostname_t **name);
 
 int32_t
+glusterd_snap_volinfo_find (char *volname, glusterd_snap_t *snap,
+                            glusterd_volinfo_t **volinfo);
+int32_t
+glusterd_snap_volinfo_find_from_parent_volname (char *origin_volname,
+                                      glusterd_snap_t *snap,
+                                      glusterd_volinfo_t **volinfo);
+
+int32_t
 glusterd_volinfo_find (char *volname, glusterd_volinfo_t **volinfo);
 
+int
+glusterd_volinfo_find_by_volume_id (uuid_t volume_id, glusterd_volinfo_t **volinfo);
+
+int
+glusterd_snap_volinfo_find_by_volume_id (uuid_t volume_id,
+                                         glusterd_volinfo_t **volinfo);
+
 int32_t
 glusterd_service_stop(const char *service, char *pidfile, int sig,
-                      gf_boolean_t keep_pidfile);
+                      gf_boolean_t force_kill);
 
 int32_t
 glusterd_resolve_brick (glusterd_brickinfo_t *brickinfo);
@@ -100,12 +147,12 @@ glusterd_resolve_brick (glusterd_brickinfo_t *brickinfo);
 int32_t
 glusterd_volume_start_glusterfs (glusterd_volinfo_t  *volinfo,
                                  glusterd_brickinfo_t   *brickinfo,
-                                 int32_t count);
+                                 gf_boolean_t wait);
 
 int32_t
 glusterd_volume_stop_glusterfs (glusterd_volinfo_t  *volinfo,
                                 glusterd_brickinfo_t   *brickinfo,
-                                int32_t count);
+                                gf_boolean_t del_brick);
 
 int32_t
 glusterd_volinfo_delete (glusterd_volinfo_t *volinfo);
@@ -117,31 +164,71 @@ gf_boolean_t
 glusterd_is_cli_op_req (int32_t op);
 
 int32_t
-glusterd_brickinfo_get (char *brick, glusterd_volinfo_t *volinfo,
-                        glusterd_brickinfo_t **brickinfo);
-int32_t
-glusterd_is_local_addr (char *hostname);
+glusterd_volume_brickinfo_get_by_brick (char *brick,
+                                        glusterd_volinfo_t *volinfo,
+                                        glusterd_brickinfo_t **brickinfo);
 
 int32_t
 glusterd_build_volume_dict (dict_t **vols);
 
 int32_t
-glusterd_compare_friend_data (dict_t  *vols, int32_t *status);
+glusterd_compare_friend_data (dict_t  *vols, int32_t *status, char *hostname);
 
 int
 glusterd_volume_compute_cksum (glusterd_volinfo_t  *volinfo);
 
+void
+glusterd_get_nodesvc_volfile (char *server, char *workdir,
+                                    char *volfile, size_t len);
+
+gf_boolean_t
+glusterd_is_service_running (char *pidfile, int *pid);
+
 gf_boolean_t
-glusterd_is_nfs_started ();
+glusterd_is_nodesvc_running ();
 
+gf_boolean_t
+glusterd_is_nodesvc_running ();
+
+void
+glusterd_get_nodesvc_dir (char *server, char *workdir,
+                                char *path, size_t len);
 int32_t
 glusterd_nfs_server_start ();
 
 int32_t
 glusterd_nfs_server_stop ();
 
-int
-glusterd_file_copy (int out, int in);
+int32_t
+glusterd_shd_start ();
+
+int32_t
+glusterd_shd_stop ();
+
+void
+glusterd_set_socket_filepath (char *sock_filepath, char *sockpath, size_t len);
+
+int32_t
+glusterd_nodesvc_set_socket_filepath (char *rundir, uuid_t uuid,
+                                      char *socketpath, int len);
+
+struct rpc_clnt*
+glusterd_pending_node_get_rpc (glusterd_pending_node_t *pending_node);
+
+struct rpc_clnt*
+glusterd_nodesvc_get_rpc (char *server);
+
+int32_t
+glusterd_nodesvc_set_rpc (char *server, struct rpc_clnt *rpc);
+
+int32_t
+glusterd_nodesvc_connect (char *server, char *socketpath);
+
+void
+glusterd_nodesvc_set_online_status (char *server, gf_boolean_t status);
+
+gf_boolean_t
+glusterd_is_nodesvc_online (char *server);
 
 int
 glusterd_remote_hostname_get (rpcsvc_request_t *req,
@@ -152,12 +239,396 @@ void
 glusterd_set_volume_status (glusterd_volinfo_t  *volinfo,
                             glusterd_volume_status status);
 int
-glusterd_check_generate_start_nfs (glusterd_volinfo_t *volinfo);
+glusterd_check_generate_start_nfs (void);
+
+int
+glusterd_check_generate_start_shd (void);
+
+int
+glusterd_nodesvcs_handle_graph_change (glusterd_volinfo_t *volinfo);
+
+int
+glusterd_nodesvcs_handle_reconfigure (glusterd_volinfo_t *volinfo);
+
+int
+glusterd_nodesvcs_start (glusterd_volinfo_t *volinfo);
+
+int
+glusterd_nodesvcs_stop (glusterd_volinfo_t *volinfo);
+
 int32_t
 glusterd_volume_count_get (void);
 int32_t
 glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo,
                              dict_t  *dict, int32_t count);
 int
-glusterd_is_exisiting_brick (char *hostname, char *path);
+glusterd_get_brickinfo (xlator_t *this, const char *brickname,
+                        int port, gf_boolean_t localhost,
+                        glusterd_brickinfo_t **brickinfo);
+
+void
+glusterd_set_brick_status (glusterd_brickinfo_t  *brickinfo,
+                            gf_brick_status_t status);
+
+gf_boolean_t
+glusterd_is_brick_started (glusterd_brickinfo_t  *brickinfo);
+
+int
+glusterd_friend_find_by_hostname (const char *hoststr,
+                                  glusterd_peerinfo_t  **peerinfo);
+int
+glusterd_hostname_to_uuid (char *hostname, uuid_t uuid);
+
+int
+glusterd_friend_brick_belongs (glusterd_volinfo_t *volinfo,
+                               glusterd_brickinfo_t *brickinfo, void *uuid);
+int
+glusterd_all_volume_cond_check (glusterd_condition_func func, int status,
+                                void *ctx);
+int
+glusterd_brick_start (glusterd_volinfo_t *volinfo,
+                      glusterd_brickinfo_t *brickinfo,
+                      gf_boolean_t wait);
+int
+glusterd_brick_stop (glusterd_volinfo_t *volinfo,
+                     glusterd_brickinfo_t *brickinfo,
+                     gf_boolean_t del_brick);
+
+int
+glusterd_is_defrag_on (glusterd_volinfo_t *volinfo);
+
+int32_t
+glusterd_volinfo_bricks_delete (glusterd_volinfo_t *volinfo);
+
+int
+glusterd_friend_find_by_uuid (uuid_t uuid,
+                              glusterd_peerinfo_t  **peerinfo);
+int
+glusterd_new_brick_validate (char *brick, glusterd_brickinfo_t *brickinfo,
+                             char *op_errstr, size_t len);
+int32_t
+glusterd_volume_brickinfos_delete (glusterd_volinfo_t *volinfo);
+
+int32_t
+glusterd_volume_brickinfo_get (uuid_t uuid, char *hostname, char *path,
+                               glusterd_volinfo_t *volinfo,
+                               glusterd_brickinfo_t **brickinfo);
+
+int
+glusterd_brickinfo_get (uuid_t uuid, char *hostname, char *path,
+                       glusterd_brickinfo_t **brickinfo);
+int
+glusterd_is_rb_started (glusterd_volinfo_t *volinfo);
+
+int
+glusterd_is_rb_paused (glusterd_volinfo_t *volinfo);
+
+int
+glusterd_set_rb_status (glusterd_volinfo_t *volinfo, gf_rb_status_t status);
+
+gf_boolean_t
+glusterd_is_rb_ongoing (glusterd_volinfo_t *volinfo);
+
+int
+glusterd_rb_check_bricks (glusterd_volinfo_t *volinfo,
+                          glusterd_brickinfo_t *src_brick,
+                          glusterd_brickinfo_t *dst_brick);
+
+int
+glusterd_check_and_set_brick_xattr (char *host, char *path, uuid_t uuid,
+                                    char **op_errstr, gf_boolean_t is_force);
+
+int
+glusterd_validate_and_create_brickpath (glusterd_brickinfo_t *brickinfo,
+                                        uuid_t volume_id, char **op_errstr,
+                                        gf_boolean_t is_force);
+int
+glusterd_sm_tr_log_transition_add (glusterd_sm_tr_log_t *log,
+                                           int old_state, int new_state,
+                                           int event);
+int
+glusterd_peerinfo_new (glusterd_peerinfo_t **peerinfo,
+                       glusterd_friend_sm_state_t state, uuid_t *uuid,
+                       const char *hostname, int port);
+int
+glusterd_sm_tr_log_init (glusterd_sm_tr_log_t *log,
+                         char * (*state_name_get) (int),
+                         char * (*event_name_get) (int),
+                         size_t  size);
+void
+glusterd_sm_tr_log_delete (glusterd_sm_tr_log_t *log);
+
+int
+glusterd_sm_tr_log_add_to_dict (dict_t *dict,
+                                glusterd_sm_tr_log_t *circular_log);
+int
+glusterd_remove_pending_entry (struct list_head *list, void *elem);
+int
+glusterd_clear_pending_nodes (struct list_head *list);
+gf_boolean_t
+glusterd_peerinfo_is_uuid_unknown (glusterd_peerinfo_t *peerinfo);
+int32_t
+glusterd_brick_connect (glusterd_volinfo_t  *volinfo,
+                        glusterd_brickinfo_t  *brickinfo, char *socketpath);
+int32_t
+glusterd_brick_disconnect (glusterd_brickinfo_t *brickinfo);
+int32_t
+glusterd_delete_volume (glusterd_volinfo_t *volinfo);
+int32_t
+glusterd_delete_brick (glusterd_volinfo_t* volinfo,
+                       glusterd_brickinfo_t *brickinfo);
+
+int32_t
+glusterd_delete_all_bricks (glusterd_volinfo_t* volinfo);
+
+int
+glusterd_spawn_daemons (void *opaque);
+
+int
+glusterd_restart_gsyncds (glusterd_conf_t *conf);
+
+int
+glusterd_start_gsync (glusterd_volinfo_t *master_vol, char *slave,
+                      char *path_list, char *conf_path,
+                      char *glusterd_uuid_str,
+                      char **op_errstr);
+int
+glusterd_get_local_brickpaths (glusterd_volinfo_t *volinfo,
+                               char **pathlist);
+
+int32_t
+glusterd_recreate_bricks (glusterd_conf_t *conf);
+int32_t
+glusterd_handle_upgrade_downgrade (dict_t *options, glusterd_conf_t *conf);
+
+int
+glusterd_add_brick_detail_to_dict (glusterd_volinfo_t *volinfo,
+                                   glusterd_brickinfo_t *brickinfo,
+                                   dict_t  *dict, int32_t count);
+
+int32_t
+glusterd_add_brick_to_dict (glusterd_volinfo_t *volinfo,
+                            glusterd_brickinfo_t *brickinfo,
+                            dict_t  *dict, int32_t count);
+
+int32_t
+glusterd_get_all_volnames (dict_t *dict);
+
+gf_boolean_t
+glusterd_is_fuse_available ();
+
+int
+glusterd_brick_statedump (glusterd_volinfo_t *volinfo,
+                          glusterd_brickinfo_t *brickinfo,
+                          char *options, int option_cnt, char **op_errstr);
+int
+glusterd_nfs_statedump (char *options, int option_cnt, char **op_errstr);
+gf_boolean_t
+glusterd_is_volume_replicate (glusterd_volinfo_t *volinfo);
+gf_boolean_t
+glusterd_is_brick_decommissioned (glusterd_volinfo_t *volinfo, char *hostname,
+                                  char *path);
+gf_boolean_t
+glusterd_friend_contains_vol_bricks (glusterd_volinfo_t *volinfo,
+                                     uuid_t friend_uuid);
+int
+glusterd_friend_remove_cleanup_vols (uuid_t uuid);
+
+gf_boolean_t
+glusterd_chk_peers_connected_befriended (uuid_t skip_uuid);
+
+void
+glusterd_get_client_filepath (char *filepath,
+                              glusterd_volinfo_t *volinfo,
+                              gf_transport_type type);
+void
+glusterd_get_trusted_client_filepath (char *filepath,
+                                      glusterd_volinfo_t *volinfo,
+                                      gf_transport_type type);
+int
+glusterd_restart_rebalance (glusterd_conf_t *conf);
+
+int32_t
+glusterd_add_bricks_hname_path_to_dict (dict_t *dict,
+                                        glusterd_volinfo_t *volinfo);
+
+int
+glusterd_add_node_to_dict (char *server, dict_t *dict, int count,
+                           dict_t *vol_opts);
+
+char *
+glusterd_uuid_to_hostname (uuid_t uuid);
+
+int
+glusterd_get_dist_leaf_count (glusterd_volinfo_t *volinfo);
+
+glusterd_brickinfo_t*
+glusterd_get_brickinfo_by_position (glusterd_volinfo_t *volinfo, uint32_t pos);
+
+gf_boolean_t
+glusterd_is_local_brick (xlator_t *this, glusterd_volinfo_t *volinfo,
+                         glusterd_brickinfo_t *brickinfo);
+int
+glusterd_validate_volume_id (dict_t *op_dict, glusterd_volinfo_t *volinfo);
+
+int
+glusterd_defrag_volume_status_update (glusterd_volinfo_t *volinfo,
+                                      dict_t *rsp_dict);
+
+int
+glusterd_check_files_identical (char *filename1, char *filename2,
+                                gf_boolean_t *identical);
+
+int
+glusterd_check_topology_identical (const char *filename1,
+                                   const char *filename2,
+                                   gf_boolean_t *identical);
+
+void
+glusterd_volinfo_reset_defrag_stats (glusterd_volinfo_t *volinfo);
+int
+glusterd_volset_help (dict_t *dict, char **op_errstr);
+
+int32_t
+glusterd_sync_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict);
+int32_t
+glusterd_gsync_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict, char *op_errstr);
+int32_t
+glusterd_rb_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict);
+int
+glusterd_profile_volume_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict);
+int
+glusterd_volume_status_copy_to_op_ctx_dict (dict_t *aggr, dict_t *rsp_dict);
+int
+glusterd_volume_rebalance_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict);
+int
+glusterd_volume_heal_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict);
+int
+glusterd_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict);
+int
+glusterd_sys_exec_output_rsp_dict (dict_t *aggr, dict_t *rsp_dict);
+int
+glusterd_snap_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict);
+int32_t
+glusterd_handle_node_rsp (dict_t *req_ctx, void *pending_entry,
+                          glusterd_op_t op, dict_t *rsp_dict, dict_t *op_ctx,
+                          char **op_errstr, gd_node_type type);
+int
+glusterd_volume_rebalance_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict);
+int
+glusterd_volume_heal_use_rsp_dict (dict_t *aggr, dict_t *rsp_dict);
+int
+_profile_volume_add_brick_rsp (dict_t *this, char *key, data_t *value,
+                             void *data);
+int
+glusterd_profile_volume_brick_rsp (void *pending_entry,
+                                   dict_t *rsp_dict, dict_t *op_ctx,
+                                   char **op_errstr, gd_node_type type);
+
+gf_boolean_t
+glusterd_are_vol_all_peers_up (glusterd_volinfo_t *volinfo,
+                               struct list_head *peers,
+                               char **down_peerstr);
+
+/* Should be used only when an operation is in progress, as that is the only
+ * time a lock_owner is set
+ */
+gf_boolean_t
+is_origin_glusterd (dict_t *dict);
+
+gf_boolean_t
+glusterd_is_quorum_changed (dict_t *options, char *option, char *value);
+
+int
+glusterd_do_quorum_action ();
+
+int
+glusterd_get_quorum_cluster_counts (xlator_t *this, int *active_count,
+                                    int *quorum_count);
+
+int
+glusterd_get_next_global_opt_version_str (dict_t *opts, char **version_str);
+gf_boolean_t
+glusterd_is_quorum_option (char *option);
+gf_boolean_t
+glusterd_is_volume_in_server_quorum (glusterd_volinfo_t *volinfo);
+gf_boolean_t
+glusterd_is_any_volume_in_server_quorum (xlator_t *this);
+gf_boolean_t
+does_gd_meet_server_quorum (xlator_t *this);
+
+int
+glusterd_generate_and_set_task_id (dict_t *dict, char *key);
+
+int
+glusterd_copy_uuid_to_dict (uuid_t uuid, dict_t *dict, char *key);
+
+gf_boolean_t
+glusterd_is_same_address (char *name1, char *name2);
+
+void
+gd_update_volume_op_versions (glusterd_volinfo_t *volinfo);
+
+char*
+gd_peer_uuid_str (glusterd_peerinfo_t *peerinfo);
+
+gf_boolean_t
+gd_is_remove_brick_committed (glusterd_volinfo_t *volinfo);
+
+gf_boolean_t
+glusterd_are_vol_all_peers_up (glusterd_volinfo_t *volinfo,
+                               struct list_head *peers,
+                               char **down_peerstr);
+
+int
+glusterd_get_slave_details_confpath (glusterd_volinfo_t *volinfo, dict_t *dict,
+                                     char **slave_ip, char **slave_vol,
+                                     char **conf_path, char **op_errstr);
+
+int
+glusterd_check_restart_gsync_session (glusterd_volinfo_t *volinfo, char *slave,
+                                      dict_t *resp_dict, char *path_list,
+                                      char *conf_path, gf_boolean_t is_force);
+
+int
+glusterd_check_gsync_running_local (char *master, char *slave,
+                                    char *conf_path,
+                                    gf_boolean_t *is_run);
+
+gf_boolean_t
+glusterd_is_status_tasks_op (glusterd_op_t op, dict_t *dict);
+
+#ifdef GF_LINUX_HOST_OS
+char*
+glusterd_get_brick_mount_details (glusterd_brickinfo_t *brickinfo);
+struct mntent *
+glusterd_get_mnt_entry_info (char *mnt_pt, FILE *mtab);
+int
+glusterd_get_brick_root (char *path, char **mount_point);
+#endif //LINUX_HOST
+
+int
+glusterd_compare_snap_time(struct list_head *, struct list_head *);
+
+int
+glusterd_compare_snap_vol_time(struct list_head *, struct list_head *);
+
+int32_t
+glusterd_snap_volinfo_restore (dict_t *rsp_dict,
+                               glusterd_volinfo_t *new_volinfo,
+                               glusterd_volinfo_t *snap_volinfo);
+int32_t
+glusterd_lvm_snapshot_remove (dict_t *rsp_dict, glusterd_volinfo_t *snap_vol);
+
+int32_t
+glusterd_missed_snapinfo_new (glusterd_missed_snap_info **missed_snapinfo);
+
+int32_t
+glusterd_missed_snap_op_new (glusterd_snap_op_t **snap_op);
+
+int32_t
+glusterd_add_missed_snaps_to_dict (dict_t *rsp_dict, char *snap_uuid,
+                                   glusterd_brickinfo_t *brickinfo,
+                                   int32_t brick_number, int32_t op);
+
 #endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-volgen.c b/xlators/mgmt/glusterd/src/glusterd-volgen.c
index 46a05d965..6f3c69e7d 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volgen.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volgen.c
@@ -1,580 +1,677 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is GF_FREE software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
 #endif
 
+#include <fnmatch.h>
+#include <sys/wait.h>
+#include <dlfcn.h>
+
+#if (HAVE_LIB_XML)
+#include <libxml/encoding.h>
+#include <libxml/xmlwriter.h>
+#endif
+
 #include "xlator.h"
-#include "protocol-common.h"
 #include "glusterd.h"
 #include "defaults.h"
-#include "list.h"
+#include "logging.h"
 #include "dict.h"
-#include "compat.h"
-#include "compat-errno.h"
-#include "glusterd-sm.h"
-#include "glusterd-op-sm.h"
-#include "cli1.h"
+#include "graph-utils.h"
+#include "glusterd-store.h"
+#include "glusterd-hooks.h"
+#include "trie.h"
 #include "glusterd-mem-types.h"
+#include "cli1-xdr.h"
 #include "glusterd-volgen.h"
+#include "glusterd-op-sm.h"
 #include "glusterd-utils.h"
+#include "run.h"
+#include "options.h"
 
-int
-set_xlator_option (dict_t *dict, char *key,
-                   char *value)
-{
-        int  ret      = 0;
-        char *str     = NULL;
+extern struct volopt_map_entry glusterd_volopt_map[];
 
-        str = GF_CALLOC (1, strlen (value) + 1,
-                         gf_gld_mt_char);
+/*********************************************
+ *
+ * xlator generation / graph manipulation API
+ *
+ *********************************************/
 
-        if (!str)
-                return -1;
 
-        strncpy (str, value, strlen (value));
+struct volgen_graph {
+        char **errstr;
+        glusterfs_graph_t graph;
+};
+typedef struct volgen_graph volgen_graph_t;
 
-        ret = dict_set_dynstr (dict, key, str);
+static void
+set_graph_errstr (volgen_graph_t *graph, const char *str)
+{
+        if (!graph->errstr)
+                return;
 
-        return ret;
+        *graph->errstr = gf_strdup (str);
 }
 
-static int32_t
-set_default_options (dict_t *dict, char *volname)
+static xlator_t *
+xlator_instantiate_va (const char *type, const char *format, va_list arg)
 {
-        int     ret       = -1;
+        xlator_t *xl = NULL;
+        char *volname = NULL;
+        int ret = 0;
 
-        ret = dict_set_str (dict, "volname",
-                            volname);
-        if (ret)
-                goto out;
+        ret = gf_vasprintf (&volname, format, arg);
+        if (ret < 0) {
+                volname = NULL;
 
-        ret = set_xlator_option (dict, VOLGEN_POSIX_OPTION_ODIRECT,
-                                 "on");
-        if (ret)
-                goto out;
+                goto error;
+        }
 
-        ret = set_xlator_option (dict, VOLGEN_POSIX_OPTION_STATFSSIZE,
-                                 "on");
+        xl = GF_CALLOC (1, sizeof (*xl), gf_common_mt_xlator_t);
+        if (!xl)
+                goto error;
+        ret = xlator_set_type_virtual (xl, type);
         if (ret)
-                goto out;
+                goto error;
+        xl->options = get_new_dict();
+        if (!xl->options)
+                goto error;
+        xl->name = volname;
+        INIT_LIST_HEAD (&xl->volume_options);
 
-        ret = set_xlator_option (dict, VOLGEN_POSIX_OPTION_MANDATTR,
-                                 "on");
-        if (ret)
-                goto out;
+        xl->ctx = THIS->ctx;
 
-        ret = set_xlator_option (dict, VOLGEN_POSIX_OPTION_SPANDEVICES,
-                                 "1");
-        if (ret)
-                goto out;
+        return xl;
 
-        ret = set_xlator_option (dict, VOLGEN_POSIX_OPTION_BCKUNLINK,
-                                 "on");
-        if (ret)
-                goto out;
+ error:
+        gf_log ("", GF_LOG_ERROR, "creating xlator of type %s failed",
+                type);
+        GF_FREE (volname);
+        if (xl)
+                xlator_destroy (xl);
 
-        ret = set_xlator_option (dict, VOLGEN_LOCKS_OPTION_TRACE,
-                                 "on");
-        if (ret)
-                goto out;
+        return NULL;
+}
 
-        ret = set_xlator_option (dict, VOLGEN_LOCKS_OPTION_MAND,
-                                 "on");
-        if (ret)
-                goto out;
+#ifdef __not_used_as_of_now_
+static xlator_t *
+xlator_instantiate (const char *type, const char *format, ...)
+{
+        va_list arg;
+        xlator_t *xl;
 
-        ret = set_xlator_option (dict, VOLGEN_CLIENT_OPTION_TRANSTYPE,
-                                 "tcp");
-        if (ret)
-                goto out;
+        va_start (arg, format);
+        xl = xlator_instantiate_va (type, format, arg);
+        va_end (arg);
 
-        ret = set_xlator_option (dict, VOLGEN_CLIENT_OPTION_NODELAY,
-                                 "on");
-        if (ret)
-                goto out;
+        return xl;
+}
+#endif
 
-        ret = set_xlator_option (dict, VOLGEN_IOT_OPTION_THREADCOUNT,
-                                 "16");
-        if (ret)
-                goto out;
+static int
+volgen_xlator_link (xlator_t *pxl, xlator_t *cxl)
+{
+        int ret = 0;
 
-        ret = set_xlator_option (dict, VOLGEN_IOT_OPTION_AUTOSCALING,
-                                 "on");
-        if (ret)
-                goto out;
+        ret = glusterfs_xlator_link (pxl, cxl);
+        if (ret == -1) {
+                gf_log ("", GF_LOG_ERROR,
+                        "Out of memory, cannot link xlators %s <- %s",
+                        pxl->name, cxl->name);
+        }
 
-        ret = set_xlator_option (dict, VOLGEN_IOT_OPTION_MINTHREADS,
-                                 "on");
-        if (ret)
-                goto out;
+        return ret;
+}
 
-        ret = set_xlator_option (dict, VOLGEN_IOT_OPTION_MAXTHREADS,
-                                 "on");
-        if (ret)
-                goto out;
+static int
+volgen_graph_link (volgen_graph_t *graph, xlator_t *xl)
+{
+        int ret = 0;
 
-        ret = set_xlator_option (dict, VOLGEN_SERVER_OPTION_TRANSTYPE,
-                                 "tcp");
-        if (ret)
-                goto out;
+        /* no need to care about graph->top here */
+        if (graph->graph.first)
+                ret = volgen_xlator_link (xl, graph->graph.first);
+        if (ret == -1) {
+                gf_log ("", GF_LOG_ERROR, "failed to add graph entry %s",
+                        xl->name);
 
-        ret = set_xlator_option (dict, VOLGEN_SERVER_OPTION_NODELAY,
-                                 "on");
-        if (ret)
-                goto out;
+                return -1;
+        }
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_READSUBVOL,
-                                 "on");
-        if (ret)
-                goto out;
+        return 0;
+}
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_FAVCHILD,
-                                 "on");
-        if (ret)
-                goto out;
+static xlator_t *
+volgen_graph_add_as (volgen_graph_t *graph, const char *type,
+                     const char *format, ...)
+{
+        va_list arg;
+        xlator_t *xl = NULL;
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_BCKSHCOUNT,
-                                 "on");
-        if (ret)
-                goto out;
+        va_start (arg, format);
+        xl = xlator_instantiate_va (type, format, arg);
+        va_end (arg);
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_DATASH,
-                                 "on");
-        if (ret)
-                goto out;
+        if (!xl)
+                return NULL;
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_DATASHALGO,
-                                 "on");
-        if (ret)
-                goto out;
+        if (volgen_graph_link (graph, xl)) {
+                xlator_destroy (xl);
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_SHWINDOWSIZE,
-                                 "on");
-        if (ret)
-                goto out;
+                return NULL;
+        } else
+                glusterfs_graph_set_first (&graph->graph, xl);
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_METASH,
-                                 "on");
-        if (ret)
-                goto out;
+        return xl;
+}
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_ENTRYSH,
-                                 "on");
-        if (ret)
-                goto out;
+static xlator_t *
+volgen_graph_add_nolink (volgen_graph_t *graph, const char *type,
+                         const char *format, ...)
+{
+        va_list arg;
+        xlator_t *xl = NULL;
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_DATACHANGELOG,
-                                 "on");
-        if (ret)
-                goto out;
+        va_start (arg, format);
+        xl = xlator_instantiate_va (type, format, arg);
+        va_end (arg);
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_METADATACHANGELOG,
-                                 "on");
-        if (ret)
-                goto out;
+        if (!xl)
+                return NULL;
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_ENTRYCHANGELOG,
-                                 "on");
-        if (ret)
-                goto out;
+        glusterfs_graph_set_first (&graph->graph, xl);
 
-        ret = set_xlator_option (dict, VOLGEN_REPLICATE_OPTION_STRICTREADDIR,
-                                 "on");
-        if (ret)
-                goto out;
+        return xl;
+}
 
-        ret = set_xlator_option (dict, VOLGEN_STRIPE_OPTION_BLOCKSIZE,
-                                 "on");
-        if (ret)
-                goto out;
+static xlator_t *
+volgen_graph_add (volgen_graph_t *graph, char *type, char *volname)
+{
+        char *shorttype = NULL;
 
-        ret = set_xlator_option (dict, VOLGEN_STRIPE_OPTION_USEXATTR,
-                                 "on");
-        if (ret)
-                goto out;
+        shorttype = strrchr (type, '/');
+        GF_ASSERT (shorttype);
+        shorttype++;
+        GF_ASSERT (*shorttype);
 
-        ret = set_xlator_option (dict, VOLGEN_DHT_OPTION_LOOKUPUNHASH,
-                                 "on");
-        if (ret)
-                goto out;
+        return volgen_graph_add_as (graph, type, "%s-%s", volname, shorttype);
+}
 
-        ret = set_xlator_option (dict, VOLGEN_DHT_OPTION_MINFREEDISK,
-                                 "on");
-        if (ret)
-                goto out;
+/* XXX Seems there is no such generic routine?
+ * Maybe should put to xlator.c ??
+ */
+static int
+xlator_set_option (xlator_t *xl, char *key, char *value)
+{
+        char *dval     = NULL;
 
-        ret = set_xlator_option (dict, VOLGEN_DHT_OPTION_UNHASHSTICKY,
-                                 "on");
-        if (ret)
-                goto out;
+        dval = gf_strdup (value);
+        if (!dval) {
+                gf_log ("", GF_LOG_ERROR,
+                        "failed to set xlator opt: %s[%s] = %s",
+                        xl->name, key, value);
 
-        ret = set_xlator_option (dict, VOLGEN_WB_OPTION_FLUSHBEHIND,
-                                 "on");
-        if (ret)
-                goto out;
+                return -1;
+        }
 
-        ret = set_xlator_option (dict, VOLGEN_WB_OPTION_CACHESIZE,
-                                 "on");
-        if (ret)
-                goto out;
+        return dict_set_dynstr (xl->options, key, dval);
+}
 
-        ret = set_xlator_option (dict, VOLGEN_WB_OPTION_DISABLENBYTES,
-                                 "on");
-        if (ret)
-                goto out;
+static int
+xlator_get_option (xlator_t *xl, char *key, char **value)
+{
+        GF_ASSERT (xl);
+        return dict_get_str (xl->options, key, value);
+}
 
-        ret = set_xlator_option (dict, VOLGEN_WB_OPTION_OSYNC,
-                                 "on");
-        if (ret)
-                goto out;
+static inline xlator_t *
+first_of (volgen_graph_t *graph)
+{
+        return (xlator_t *)graph->graph.first;
+}
 
-        ret = set_xlator_option (dict, VOLGEN_WB_OPTION_TRICKLINGWRITES,
-                                 "on");
-        if (ret)
-                goto out;
 
-        ret = set_xlator_option (dict, VOLGEN_RA_OPTION_ATIME,
-                                 "on");
-        if (ret)
-                goto out;
 
-        ret = set_xlator_option (dict, VOLGEN_RA_OPTION_PAGECOUNT,
-                                 "on");
-        if (ret)
-                goto out;
 
-        ret = set_xlator_option (dict, VOLGEN_IOCACHE_OPTION_PRIORITY,
-                                 "on");
-        if (ret)
-                goto out;
+/**************************
+ *
+ * Trie glue
+ *
+ *************************/
 
-        ret = set_xlator_option (dict, VOLGEN_IOCACHE_OPTION_TIMEOUT,
-                                 "on");
-        if (ret)
-                goto out;
 
-        ret = set_xlator_option (dict, VOLGEN_IOCACHE_OPTION_CACHESIZE,
-                                 "on");
-        if (ret)
-                goto out;
+static int
+volopt_selector (int lvl, char **patt, void *param,
+                     int (*optcbk)(char *word, void *param))
+{
+        struct volopt_map_entry *vme = NULL;
+        char *w = NULL;
+        int i = 0;
+        int len = 0;
+        int ret = 0;
+        char *dot = NULL;
+
+        for (vme = glusterd_volopt_map; vme->key; vme++) {
+                w = vme->key;
+
+                for (i = 0; i < lvl; i++) {
+                        if (patt[i]) {
+                                w = strtail (w, patt[i]);
+                                GF_ASSERT (!w || *w);
+                                if (!w || *w != '.')
+                                        goto next;
+                        } else {
+                                w = strchr (w, '.');
+                                GF_ASSERT (w);
+                        }
+                        w++;
+                }
 
-        ret = set_xlator_option (dict, VOLGEN_IOCACHE_OPTION_MINFILESIZE,
-                                 "on");
-        if (ret)
-                goto out;
+                dot = strchr (w, '.');
+                if (dot) {
+                        len = dot - w;
+                        w = gf_strdup (w);
+                        if (!w)
+                                return -1;
+                        w[len] = '\0';
+                }
+                ret = optcbk (w, param);
+                if (dot)
+                        GF_FREE (w);
+                if (ret)
+                        return -1;
+ next:
+                continue;
+        }
 
-        ret = set_xlator_option (dict, VOLGEN_IOCACHE_OPTION_MAXFILESIZE,
-                                 "on");
-        if (ret)
-                goto out;
+        return 0;
+}
 
-        ret = set_xlator_option (dict, VOLGEN_QR_OPTION_PRIORITY,
-                                 "on");
-        if (ret)
-                goto out;
+static int
+volopt_trie_cbk (char *word, void *param)
+{
+        return trie_add ((trie_t *)param, word);
+}
 
-        ret = set_xlator_option (dict, VOLGEN_QR_OPTION_TIMEOUT,
-                                 "on");
-        if (ret)
-                goto out;
+static int
+process_nodevec (struct trienodevec *nodevec, char **hint)
+{
+        int ret = 0;
+        char *hint1 = NULL;
+        char *hint2 = NULL;
+        char *hintinfx = "";
+        trienode_t **nodes = nodevec->nodes;
 
-        ret = set_xlator_option (dict, VOLGEN_QR_OPTION_CACHESIZE,
-                                 "on");
-        if (ret)
-                goto out;
+        if (!nodes[0]) {
+                *hint = NULL;
+                return 0;
+        }
 
-        ret = set_xlator_option (dict, VOLGEN_QR_OPTION_MAXFILESIZE,
-                                 "on");
-        if (ret)
-                goto out;
+#if 0
+        /* Limit as in git */
+        if (trienode_get_dist (nodes[0]) >= 6) {
+                *hint = NULL;
+                return 0;
+        }
+#endif
 
-        ret = 0;
+        if (trienode_get_word (nodes[0], &hint1))
+                return -1;
 
-out:
+        if (nodevec->cnt < 2 || !nodes[1]) {
+                *hint = hint1;
+                return 0;
+        }
+
+        if (trienode_get_word (nodes[1], &hint2))
+                return -1;
+
+        if (*hint)
+                hintinfx = *hint;
+        ret = gf_asprintf (hint, "%s or %s%s", hint1, hintinfx, hint2);
+        if (ret > 0)
+                ret = 0;
         return ret;
 }
 
-int32_t
-glusterd_default_xlator_options (glusterd_volinfo_t *volinfo)
+static int
+volopt_trie_section (int lvl, char **patt, char *word, char **hint, int hints)
 {
-        int ret = -1;
+        trienode_t *nodes[] = { NULL, NULL };
+        struct trienodevec nodevec = { nodes, 2};
+        trie_t *trie = NULL;
+        int ret = 0;
 
-        volinfo->dict = dict_new ();
-        if (!volinfo->dict) {
-                ret = -1;
-                goto out;
-        }
+        trie = trie_new ();
+        if (!trie)
+                return -1;
 
-        ret = set_default_options (volinfo->dict,
-                                   volinfo->volname);
-        if (ret) {
-                dict_unref (volinfo->dict);
-                goto out;
+        if (volopt_selector (lvl, patt, trie, &volopt_trie_cbk)) {
+                trie_destroy (trie);
+
+                return -1;
         }
 
-        ret = 0;
+        GF_ASSERT (hints <= 2);
+        nodevec.cnt = hints;
+        ret = trie_measure_vec (trie, word, &nodevec);
+        if (ret || !nodevec.nodes[0])
+                trie_destroy (trie);
 
-out:
-        return ret;
+        ret = process_nodevec (&nodevec, hint);
+        trie_destroy (trie);
 
+        return ret;
 }
 
 static int
-__write_posix_xlator (FILE *file, dict_t *dict,
-                      char *posix_directory)
-{
-         char      *volname                 = NULL;
-        char       *opt_odirect             = NULL;
-        char       *opt_statfssize          = NULL;
-        char       *opt_mandattr            = NULL;
-        char       *opt_spandevices         = NULL;
-        char       *opt_bckunlink           = NULL;
-        int         ret                     = -1;
-
-        const char *posix_str = "volume %s-%s\n"
-                "    type storage/posix\n"
-                "    option directory %s\n"
-                "#   option o-direct %s\n"
-                "#   option export-statfs-size %s\n"
-                "#   option mandate-attribute %s\n"
-                "#   option span-devices %s\n"
-                "#   option background-unlink %s\n"
-                "end-volume\n\n";
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                goto out;
-        }
+volopt_trie (char *key, char **hint)
+{
+        char *patt[] = { NULL };
+        char *fullhint = NULL;
+        char *dot = NULL;
+        char *dom = NULL;
+        int   len = 0;
+        int   ret = 0;
+
+        *hint = NULL;
+
+        dot = strchr (key, '.');
+        if (!dot)
+                return volopt_trie_section (1, patt, key, hint, 2);
+
+        len = dot - key;
+        dom = gf_strdup (key);
+        if (!dom)
+                return -1;
+        dom[len] = '\0';
 
-        ret = dict_get_str (dict, VOLGEN_POSIX_OPTION_ODIRECT,
-                            &opt_odirect);
+        ret = volopt_trie_section (0, NULL, dom, patt, 1);
+        GF_FREE (dom);
         if (ret) {
+                patt[0] = NULL;
                 goto out;
         }
-
-        ret = dict_get_str (dict, VOLGEN_POSIX_OPTION_STATFSSIZE,
-                            &opt_statfssize);
-        if (ret) {
+        if (!patt[0])
                 goto out;
-        }
 
-        ret = dict_get_str (dict, VOLGEN_POSIX_OPTION_MANDATTR,
-                            &opt_mandattr);
-        if (ret) {
+        *hint = "...";
+        ret = volopt_trie_section (1, patt, dot + 1, hint, 2);
+        if (ret)
                 goto out;
+        if (*hint) {
+                ret = gf_asprintf (&fullhint, "%s.%s", patt[0], *hint);
+                GF_FREE (*hint);
+                if (ret >= 0) {
+                        ret = 0;
+                        *hint = fullhint;
+                }
         }
 
-        ret = dict_get_str (dict, VOLGEN_POSIX_OPTION_SPANDEVICES,
-                            &opt_spandevices);
-        if (ret) {
-                goto out;
-        }
+ out:
+        GF_FREE (patt[0]);
+        if (ret)
+                *hint = NULL;
 
-        ret = dict_get_str (dict, VOLGEN_POSIX_OPTION_BCKUNLINK,
-                            &opt_bckunlink);
-        if (ret) {
-                goto out;
-        }
+        return ret;
+}
 
-        fprintf (file, posix_str,
-                 volname,
-                 "posix",
-                 posix_directory,
-                 opt_odirect,
-                 opt_statfssize,
-                 opt_mandattr,
-                 opt_spandevices,
-                 opt_bckunlink);
 
-        ret = 0;
 
-out:
-        return ret;
+
+/**************************
+ *
+ * Volume generation engine
+ *
+ **************************/
+
+
+typedef int (*volgen_opthandler_t) (volgen_graph_t *graph,
+                                    struct volopt_map_entry *vme,
+                                    void *param);
+
+struct opthandler_data {
+        volgen_graph_t *graph;
+        volgen_opthandler_t handler;
+        struct volopt_map_entry *vme;
+        gf_boolean_t found;
+        gf_boolean_t data_t_fake;
+        int rv;
+        char *volname;
+        void *param;
+};
+
+static int
+process_option (char *key, data_t *value, void *param)
+{
+        struct opthandler_data *odt = param;
+        struct volopt_map_entry vme = {0,};
+
+        if (odt->rv)
+                return 0;
+        odt->found = _gf_true;
+
+        vme.key = key;
+        vme.voltype = odt->vme->voltype;
+        vme.option = odt->vme->option;
+        vme.op_version = odt->vme->op_version;
+
+        if (!vme.option) {
+                vme.option = strrchr (key, '.');
+                if (vme.option)
+                        vme.option++;
+                else
+                        vme.option = key;
+        }
+        if (odt->data_t_fake)
+                vme.value = (char *)value;
+        else
+                vme.value = value->data;
+
+        odt->rv = odt->handler (odt->graph, &vme, odt->param);
+        return 0;
 }
 
 static int
-__write_locks_xlator (FILE *file, dict_t *dict,
-                      char *subvolume)
-{
-        char       *volname           = NULL;
-        char       *opt_trace         = NULL;
-        char       *opt_mand          = NULL;
-        int         ret               = -1;
-
-        const char *locks_str = "volume %s-%s\n"
-                "    type features/locks\n"
-                "#   option trace %s\n"
-                "#   option mandatory %s\n"
-                "    subvolumes %s\n"
-                "end-volume\n\n";
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                goto out;
-        }
+volgen_graph_set_options_generic (volgen_graph_t *graph, dict_t *dict,
+                                  void *param, volgen_opthandler_t handler)
+{
+        struct volopt_map_entry *vme = NULL;
+        struct opthandler_data odt = {0,};
+        data_t *data = NULL;
 
-        ret = dict_get_str (dict, VOLGEN_LOCKS_OPTION_TRACE,
-                            &opt_trace);
-        if (ret) {
-                goto out;
+        odt.graph = graph;
+        odt.handler = handler;
+        odt.param = param;
+        (void)data;
+
+        for (vme = glusterd_volopt_map; vme->key; vme++) {
+                odt.vme = vme;
+                odt.found = _gf_false;
+                odt.data_t_fake = _gf_false;
+
+                data = dict_get (dict, vme->key);
+
+                if (data)
+                        process_option (vme->key, data, &odt);
+                if (odt.rv)
+                        return odt.rv;
+
+                if (odt.found)
+                        continue;
+
+                /* check for default value */
+
+                if (vme->value) {
+                        /* stupid hack to be able to reuse dict iterator
+                         * in this context
+                         */
+                        odt.data_t_fake = _gf_true;
+                        process_option (vme->key, (data_t *)vme->value, &odt);
+                        if (odt.rv)
+                                return odt.rv;
+                }
         }
 
-        ret = dict_get_str (dict, VOLGEN_LOCKS_OPTION_MAND,
-                            &opt_mand);
-        if (ret) {
-                goto out;
+        return 0;
+}
+
+static int
+no_filter_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme,
+                          void *param)
+{
+        xlator_t *trav;
+        int ret = 0;
+
+        for (trav = first_of (graph); trav; trav = trav->next) {
+                if (strcmp (trav->type, vme->voltype) != 0)
+                        continue;
+
+                ret = xlator_set_option (trav, vme->option, vme->value);
+                if (ret)
+                        break;
         }
+        return ret;
+}
 
-        fprintf (file, locks_str,
-                 volname,
-                 "locks",
-                 opt_trace,
-                 opt_mand,
-                 subvolume);
+static int
+basic_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme,
+                      void *param)
+{
+        int     ret = 0;
 
-        ret = 0;
+        if (vme->option[0] == '!')
+                goto out;
 
+        ret = no_filter_option_handler (graph, vme, param);
 out:
         return ret;
 }
 
 static int
-__write_client_xlator (FILE *file, dict_t *dict,
-                       char *remote_subvol,
-                       char *remote_host,
-                       int count)
-{
-        char       *volname               = NULL;
-        char       *opt_transtype         = NULL;
-        char       *opt_nodelay           = NULL;
-        int         ret                   = 0;
-	char	   *ping                  = NULL;
-	char	   *frame                 = NULL;
-	char	    ping_timeout[100]	  = {0, };
-	char 	    frame_timeout[100]	  = {0, };
-
-        const char *client_str = "volume %s-%s-%d\n"
-                "    type protocol/client\n"
-                "    option transport-type %s\n"
-                "    option remote-host %s\n"
-                "    option transport.socket.nodelay %s\n"
-		"%s" //for frame-timeout
-		"%s" //for ping-timeout
-                "    option remote-subvolume %s\n"
-                "end-volume\n\n";
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                goto out;
-        }
+volgen_graph_set_options (volgen_graph_t *graph, dict_t *dict)
+{
+        return volgen_graph_set_options_generic (graph, dict, NULL,
+                                                 &basic_option_handler);
+}
 
-        ret = dict_get_str (dict, VOLGEN_CLIENT_OPTION_TRANSTYPE,
-                            &opt_transtype);
-        if (ret) {
-                goto out;
-        }
+static int
+optget_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme,
+                       void *param)
+{
+        struct volopt_map_entry *vme2 = param;
 
-        ret = dict_get_str (dict, VOLGEN_CLIENT_OPTION_NODELAY,
-                            &opt_nodelay);
-        if (ret) {
-                goto out;
-        }
+        if (strcmp (vme->key, vme2->key) == 0)
+                vme2->value = vme->value;
 
-	if (dict_get (dict, "frame-timeout")) {
-		ret = dict_get_str (dict, "frame-timeout", &frame);
-		if (ret) {
-                	goto out;
-        	}
-		gf_log ("", GF_LOG_DEBUG, "Reconfiguring frame-timeout %s",
-			frame);
-		sprintf(frame_timeout, "    option frame-timeout %s\n",frame);
-	} 
-
-	if (dict_get (dict, "ping-timeout")) {
-		ret = dict_get_str (dict, "ping-timeout", &ping);
-		if (ret) {
-                	goto out;
-        	}
-		gf_log ("", GF_LOG_DEBUG, "Reconfiguring ping-timeout %s",
-			ping);
-		sprintf(ping_timeout, "    option ping-timeout %s\n",ping);
-	} 
-
-        fprintf (file, client_str,
-                 volname,
-                 "client",
-                 count,
-                 opt_transtype,
-                 remote_host,
-                 opt_nodelay,
-		 frame_timeout,
-		 ping_timeout,
-                 remote_subvol);
+        return 0;
+}
 
-        ret = 0;
+static glusterd_server_xlator_t
+get_server_xlator (char *xlator)
+{
+        glusterd_server_xlator_t subvol = GF_XLATOR_NONE;
+
+        if (strcmp (xlator, "posix") == 0)
+                subvol = GF_XLATOR_POSIX;
+        if (strcmp (xlator, "acl") == 0)
+                subvol = GF_XLATOR_ACL;
+        if (strcmp (xlator, "locks") == 0)
+                subvol = GF_XLATOR_LOCKS;
+        if (strcmp (xlator, "io-threads") == 0)
+                subvol = GF_XLATOR_IOT;
+        if (strcmp (xlator, "index") == 0)
+                subvol = GF_XLATOR_INDEX;
+        if (strcmp (xlator, "marker") == 0)
+                subvol = GF_XLATOR_MARKER;
+        if (strcmp (xlator, "io-stats") == 0)
+                subvol = GF_XLATOR_IO_STATS;
+        if (strcmp (xlator, "bd") == 0)
+                subvol = GF_XLATOR_BD;
+
+        return subvol;
+}
 
-out:
-        return ret;
+static glusterd_client_xlator_t
+get_client_xlator (char *xlator)
+{
+        glusterd_client_xlator_t subvol = GF_CLNT_XLATOR_NONE;
+
+        if (strcmp (xlator, "client") == 0)
+                subvol = GF_CLNT_XLATOR_FUSE;
+
+        return subvol;
 }
 
 static int
-__write_replace_brick_xlator (FILE *file, dict_t *dict)
+debugxl_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme,
+                       void *param)
 {
-        char       *volname               = NULL;
-        char       *opt_transtype         = NULL;
-        char       *opt_nodelay           = NULL;
-        int         ret                   = 0;
+        char *volname = NULL;
+        gf_boolean_t enabled = _gf_false;
 
+        volname = param;
 
-        const char *client_str = "volume %s-%s\n"
-                "    type protocol/client\n"
-                "    option transport-type %s\n"
-                "    option remote-port 34034\n"
-                "    option transport.socket.nodelay %s\n"
-                "end-volume\n\n";
+        if (strcmp (vme->option, "!debug") != 0)
+                return 0;
 
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                goto out;
+        if (!strcmp (vme->key , "debug.trace") ||
+            !strcmp (vme->key, "debug.error-gen")) {
+                if (get_server_xlator (vme->value) == GF_XLATOR_NONE &&
+                    get_client_xlator (vme->value) == GF_CLNT_XLATOR_NONE)
+                        return 0;
+                else
+                        goto add_graph;
         }
 
-        ret = dict_get_str (dict, VOLGEN_CLIENT_OPTION_TRANSTYPE,
-                            &opt_transtype);
-        if (ret) {
-                goto out;
-        }
+        if (gf_string2boolean (vme->value, &enabled) == -1)
+                return -1;
+        if (!enabled)
+                return 0;
 
-        ret = dict_get_str (dict, VOLGEN_CLIENT_OPTION_NODELAY,
-                            &opt_nodelay);
-        if (ret) {
-                goto out;
+add_graph:
+        if (volgen_graph_add (graph, vme->voltype, volname))
+                return 0;
+        else
+                return -1;
+}
+
+int
+check_and_add_debug_xl (volgen_graph_t *graph, dict_t *set_dict, char *volname,
+                        char *xlname)
+{
+        int       ret       = 0;
+        char     *value_str = NULL;
+
+        ret = dict_get_str (set_dict, "debug.trace", &value_str);
+        if (!ret) {
+                if (strcmp (xlname, value_str) == 0) {
+                        ret = volgen_graph_set_options_generic (graph, set_dict, volname,
+                                                                &debugxl_option_handler);
+                        if (ret)
+                                goto out;
+                }
         }
 
-        fprintf (file, client_str,
-                 volname,
-                 "replace-brick",
-                 opt_transtype,
-                 opt_nodelay);
+        ret = dict_get_str (set_dict, "debug.error-gen", &value_str);
+        if (!ret) {
+                if (strcmp (xlname, value_str) == 0) {
+                        ret = volgen_graph_set_options_generic (graph, set_dict, volname,
+                                                                &debugxl_option_handler);
+                        if (ret)
+                                goto out;
+                }
+        }
 
         ret = 0;
 
@@ -582,1931 +679,3387 @@ out:
         return ret;
 }
 
+/* This getter considers defaults also. */
 static int
-__write_pump_xlator (FILE *file, dict_t *dict,
-                     char *subvolume)
+volgen_dict_get (dict_t *dict, char *key, char **value)
 {
-        char *volname   = NULL;
-        int   ret       = -1;
+        struct volopt_map_entry vme = {0,};
+        int ret = 0;
 
-        const char *pump_str = "volume %s-%s\n"
-                "    type cluster/pump\n"
-                "    subvolumes %s %s-replace-brick\n"
-                "end-volume\n\n";
+        vme.key = key;
 
-        ret = dict_get_str (dict, "volname", &volname);
+        ret = volgen_graph_set_options_generic (NULL, dict, &vme,
+                                                &optget_option_handler);
         if (ret) {
-                goto out;
-        }
+                gf_log ("", GF_LOG_ERROR, "Out of memory");
 
-        fprintf (file, pump_str,
-                 volname, "pump",
-                 subvolume,
-                 volname);
+                return -1;
+        }
 
-        ret = 0;
+        *value = vme.value;
 
-out:
-        return ret;
+        return 0;
 }
 
 static int
-__write_iothreads_xlator (FILE *file, dict_t *dict,
-                          char *subvolume)
-{
-        char       *volname           = NULL;
-        char       *opt_threadcount   = NULL;
-        char       *opt_autoscaling   = NULL;
-        char       *opt_minthreads    = NULL;
-        char       *opt_maxthreads    = NULL;
-        int         ret               = -1;
-
-        const char *iot_str = "volume %s\n"
-                "    type performance/io-threads\n"
-                "    option thread-count %s\n"
-                "#   option autoscaling %s\n"
-                "#   option min-threads %s\n"
-                "#   option max-threads %s\n"
-                "    subvolumes %s\n"
-                "end-volume\n\n";
-
-        ret = dict_get_str (dict, "export-path", &volname);
-        if (ret) {
-                goto out;
-        }
+option_complete (char *key, char **completion)
+{
+        struct volopt_map_entry *vme = NULL;
 
-	if (dict_get (dict, "thread-count")) {
-		gf_log("", GF_LOG_DEBUG, "Resetting the thread-count value");
-        	ret = dict_get_str (dict, "thread-count", &opt_threadcount);
-	}
-	else 
-		ret = dict_get_str (dict, VOLGEN_IOT_OPTION_THREADCOUNT,
-                	            &opt_threadcount);
-	
-        if (ret) {
-                goto out;
-        }
+        *completion = NULL;
+        for (vme = glusterd_volopt_map; vme->key; vme++) {
+                if (strcmp (strchr (vme->key, '.') + 1, key) != 0)
+                        continue;
 
-        ret = dict_get_str (dict, VOLGEN_IOT_OPTION_AUTOSCALING,
-                            &opt_autoscaling);
-        if (ret) {
-                goto out;
+                if (*completion && strcmp (*completion, vme->key) != 0) {
+                        /* cancel on non-unique match */
+                        *completion = NULL;
+
+                        return 0;
+                } else
+                        *completion = vme->key;
         }
 
-        ret = dict_get_str (dict, VOLGEN_IOT_OPTION_MINTHREADS,
-                            &opt_minthreads);
-        if (ret) {
-                goto out;
+        if (*completion) {
+                /* For sake of unified API we want
+                 * have the completion to be a to-be-freed
+                 * string.
+                 */
+                *completion = gf_strdup (*completion);
+                return -!*completion;
         }
 
-        ret = dict_get_str (dict, VOLGEN_IOT_OPTION_MAXTHREADS,
-                            &opt_maxthreads);
+        return 0;
+}
+
+int
+glusterd_volinfo_get (glusterd_volinfo_t *volinfo, char *key, char **value)
+{
+        return volgen_dict_get (volinfo->dict, key, value);
+}
+
+int
+glusterd_volinfo_get_boolean (glusterd_volinfo_t *volinfo, char *key)
+{
+        char *val = NULL;
+        gf_boolean_t  boo = _gf_false;
+        int ret = 0;
+
+        ret = glusterd_volinfo_get (volinfo, key, &val);
+        if (ret)
+                return -1;
+
+        if (val)
+                ret = gf_string2boolean (val, &boo);
         if (ret) {
-                goto out;
+                gf_log ("", GF_LOG_ERROR, "value for %s option is not valid", key);
+
+                return -1;
         }
 
-        fprintf (file, iot_str,
-                 volname,
-                 opt_threadcount,
-                 opt_autoscaling,
-                 opt_minthreads,
-                 opt_maxthreads,
-                 subvolume);
+        return boo;
+}
 
-        ret = 0;
+gf_boolean_t
+glusterd_check_voloption_flags (char *key, int32_t flags)
+{
+        char *completion = NULL;
+        struct volopt_map_entry *vmep = NULL;
+        int   ret = 0;
+
+        COMPLETE_OPTION(key, completion, ret);
+        for (vmep = glusterd_volopt_map; vmep->key; vmep++) {
+                if (strcmp (vmep->key, key) == 0) {
+                        if (vmep->flags & flags)
+                                return _gf_true;
+                        else
+                                return _gf_false;
+                }
+        }
 
-out:
-        return ret;
+        return _gf_false;
 }
 
-static int
-__write_access_control_xlator (FILE *file, dict_t *dict,
-                               char *subvolume)
+gf_boolean_t
+glusterd_check_globaloption (char *key)
 {
-        char  *volname       = NULL;
-        int    ret           = -1;
+        char *completion = NULL;
+        struct volopt_map_entry *vmep = NULL;
+        int   ret = 0;
+
+        COMPLETE_OPTION(key, completion, ret);
+        for (vmep = glusterd_volopt_map; vmep->key; vmep++) {
+                if (strcmp (vmep->key, key) == 0) {
+                        if ((vmep->type == GLOBAL_DOC) ||
+                            (vmep->type == GLOBAL_NO_DOC))
+                                return _gf_true;
+                        else
+                                return _gf_false;
+                }
+        }
 
-        const char *ac_str = "volume %s-access-control\n"
-                             "    type features/access-control\n"
-                             "    subvolumes %s\n"
-                             "end-volume\n\n";
+        return _gf_false;
+}
 
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                goto out;
+gf_boolean_t
+glusterd_check_localoption (char *key)
+{
+        char *completion = NULL;
+        struct volopt_map_entry *vmep = NULL;
+        int   ret = 0;
+
+        COMPLETE_OPTION(key, completion, ret);
+        for (vmep = glusterd_volopt_map; vmep->key; vmep++) {
+                if (strcmp (vmep->key, key) == 0) {
+                        if ((vmep->type == DOC) ||
+                            (vmep->type == NO_DOC))
+                                return _gf_true;
+                        else
+                                return _gf_false;
+                }
         }
 
+        return _gf_false;
+}
 
-        fprintf (file, ac_str, volname, subvolume);
-        ret = 0;
+int
+glusterd_check_voloption (char *key)
+{
+        char *completion = NULL;
+        struct volopt_map_entry *vmep = NULL;
+        int ret = 0;
+
+        COMPLETE_OPTION(key, completion, ret);
+        for (vmep = glusterd_volopt_map; vmep->key; vmep++) {
+                if (strcmp (vmep->key, key) == 0) {
+                        if ((vmep->type == DOC) ||
+                            (vmep->type == DOC))
+                                return _gf_true;
+                        else
+                                return _gf_false;
+                }
+        }
+
+        return _gf_false;
 
-out:
-        return ret;
 }
 
-static int
-__write_server_xlator (FILE *file, dict_t *dict,
-                       char *subvolume)
-{
-        char  *volname              = NULL;
-        char  *opt_transtype        = NULL;
-        char  *opt_nodelay          = NULL;
-        int    ret                  = -1;
-	char  *lru_count            = NULL;
-	char   inode_lru_count[100] = {0,}; 
-
-        const char *server_str = "volume %s-%s\n"
-                "    type protocol/server\n"
-                "    option transport-type %s\n"
-                "    option auth.addr.%s.allow *\n"
-                "    option transport.socket.nodelay %s\n"
-		"%s"//for inode-lru-limit
-                "    subvolumes %s\n"
-                "end-volume\n\n";
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                goto out;
+int
+glusterd_check_option_exists (char *key, char **completion)
+{
+        struct volopt_map_entry vme = {0,};
+        struct volopt_map_entry *vmep = NULL;
+        int ret = 0;
+        xlator_t *this = THIS;
+
+        (void)vme;
+        (void)vmep;
+
+        if (!strchr (key, '.')) {
+                if (completion) {
+                        ret = option_complete (key, completion);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Out of memory");
+                                return -1;
+                        }
+
+                        ret = !!*completion;
+                        if (ret)
+                                return ret;
+                        else
+                                goto trie;
+                } else
+                        return 0;
         }
 
-        ret = dict_get_str (dict, VOLGEN_SERVER_OPTION_TRANSTYPE,
-                            &opt_transtype);
-        if (ret) {
-                goto out;
+        for (vmep = glusterd_volopt_map; vmep->key; vmep++) {
+                if (strcmp (vmep->key, key) == 0) {
+                        ret = 1;
+                        break;
+                }
         }
 
-        ret = dict_get_str (dict, VOLGEN_SERVER_OPTION_NODELAY,
-                            &opt_nodelay);
+        if (ret || !completion)
+                return ret;
+
+ trie:
+        ret = volopt_trie (key, completion);
         if (ret) {
-                goto out;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Some error occurred during keyword hinting");
         }
 
-	if (dict_get (dict, "inode-lru-limit")) {
-		ret = dict_get_str (dict, "inode-lru-limit", &lru_count);
-		if (ret) {
-                	goto out;
-        	}
-		gf_log ("", GF_LOG_DEBUG, "Reconfiguring inode-lru-limit %s",
-			lru_count);
-		sprintf (inode_lru_count, "    option inode-lru-limit %s\n", 
-			 lru_count);
-	} 
+        return ret;
+}
 
-        fprintf (file, server_str,
-                 volname, "server",
-                 opt_transtype,
-                 subvolume,
-                 opt_nodelay,
-		 inode_lru_count,
-                 subvolume
-		 );
+char*
+glusterd_get_trans_type_rb (gf_transport_type ttype)
+{
+        char *trans_type = NULL;
+
+        switch (ttype) {
+        case GF_TRANSPORT_RDMA:
+                gf_asprintf (&trans_type, "rdma");
+                break;
+        case GF_TRANSPORT_TCP:
+        case GF_TRANSPORT_BOTH_TCP_RDMA:
+                gf_asprintf (&trans_type, "tcp");
+                break;
+        default:
+                gf_log (THIS->name, GF_LOG_ERROR, "Unknown "
+                        "transport type");
+        }
+
+        return trans_type;
+}
 
-        ret = 0;
+static int
+_xl_link_children (xlator_t *parent, xlator_t *children, size_t child_count)
+{
+        xlator_t        *trav = NULL;
+        size_t          seek = 0;
+        int             ret = -1;
 
+        if (child_count == 0)
+                goto out;
+        seek = child_count;
+        for (trav = children; --seek; trav = trav->next);
+        for (; child_count--; trav = trav->prev) {
+                ret = volgen_xlator_link (parent, trav);
+                if (ret)
+                        goto out;
+        }
+        ret = 0;
 out:
         return ret;
 }
 
 static int
-__write_replicate_xlator (FILE *file, dict_t *dict,
-                          char *subvolume,
-                          int replicate_count,
-                          int subvol_count,
-                          int count)
-{
-        char *volname               = NULL;
-        char       *opt_readsubvol        = NULL;
-        char       *opt_favchild          = NULL;
-        char       *opt_bckshcount        = NULL;
-        char       *opt_datash            = NULL;
-        char       *opt_datashalgo        = NULL;
-        char       *opt_shwindowsize      = NULL;
-        char       *opt_metash            = NULL;
-        char       *opt_entrysh           = NULL;
-        char       *opt_datachangelog     = NULL;
-        char       *opt_metadatachangelog = NULL;
-        char       *opt_entrychangelog    = NULL;
-        char       *opt_strictreaddir     = NULL;
-        char        *subvol_str           = NULL;
-        char        tmp[4096]             = {0,};
-        int         ret                   = -1;
-        int         subvolume_count       = 0;
-        int         i                     = 0;
-        int         len                   = 0;
-        int         subvol_len            = 0;
-
-
-        char replicate_str[] = "volume %s-%s-%d\n"
-                "    type cluster/replicate\n"
-                "#   option read-subvolume %s\n"
-                "#   option favorite-child %s\n"
-                "#   option background-self-heal-count %s\n"
-                "#   option data-self-heal %s\n"
-                "#   option data-self-heal-algorithm %s\n"
-                "#   option data-self-heal-window-size %s\n"
-                "#   option metadata-self-heal %s\n"
-                "#   option entry-self-heal %s\n"
-                "#   option data-change-log %s\n"
-                "#   option metadata-change-log %s\n"
-                "#   option entry-change-log %s\n"
-                "#   option strict-readdir %s\n"
-                "    subvolumes %s\n"
-                "end-volume\n\n";
-
-        subvolume_count = subvol_count;
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                goto out;
-        }
+volgen_graph_merge_sub (volgen_graph_t *dgraph, volgen_graph_t *sgraph,
+                        size_t child_count)
+{
+        xlator_t *trav = NULL;
+        int      ret   = 0;
 
-	if (dict_get (dict, "read-subvolume")) {
-		ret = dict_get_str (dict, "read-subvolume", &opt_readsubvol);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring read-subvolume: %s", 
-		       &opt_readsubvol);
-		uncomment_option (replicate_str,
-				  (char *) "#   option read-subvolume %s\n");
-	}
-	else
-        	ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_READSUBVOL,
-                            &opt_readsubvol);
-        if (ret) {
-                goto out;
-        }
+        GF_ASSERT (dgraph->graph.first);
 
-        ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_FAVCHILD,
-                            &opt_favchild);
-        if (ret) {
+        ret = _xl_link_children (first_of (dgraph), first_of (sgraph),
+                                 child_count);
+        if (ret)
                 goto out;
-        }
 
-	if (dict_get (dict, "background-self-heal-count")) {
-		ret = dict_get_str (dict,"background-self-heal-count", 
-				    &opt_bckshcount);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring background-self-heal-"
-					 "count: %s", &opt_bckshcount);
-		uncomment_option (replicate_str,
-				  (char *) "#   option background-self-heal-count %s\n");
+        for (trav = first_of (dgraph); trav->next; trav = trav->next);
+
+        trav->next = first_of (sgraph);
+        trav->next->prev = trav;
+        dgraph->graph.xl_count += sgraph->graph.xl_count;
+
+out:
+        return ret;
+}
+
+static void
+volgen_apply_filters (char *orig_volfile)
+{
+	DIR           *filterdir = NULL;
+	struct dirent  entry = {0,};
+	struct dirent *next = NULL;
+	char          *filterpath = NULL;
+	struct stat    statbuf = {0,};
+
+	filterdir = opendir(FILTERDIR);
+	if (!filterdir) {
+		return;
 	}
-	else
-	        ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_BCKSHCOUNT,
-                            &opt_bckshcount);
-        if (ret) {
-                goto out;
-        }
 
-	if (dict_get (dict, "data-self-heal")) {
-		ret = dict_get_str (dict,"data-self-heal", 
-				    &opt_datash);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring data-self-heal"
-					 "count: %s", &opt_datash);
-		uncomment_option (replicate_str,
-				  (char *) "#   option data-self-heal %s\n");
+	while ((readdir_r(filterdir,&entry,&next) == 0) && next) {
+		if (!strncmp(entry.d_name,".",sizeof(entry.d_name))) {
+			continue;
+		}
+		if (!strncmp(entry.d_name,"..",sizeof(entry.d_name))) {
+			continue;
+		}
+		/*
+		 * d_type isn't guaranteed to be present/valid on all systems,
+		 * so do an explicit stat instead.
+		 */
+		if (gf_asprintf(&filterpath,"%s/%.*s",FILTERDIR,
+				sizeof(entry.d_name), entry.d_name) == (-1)) {
+			continue;
+		}
+		/* Deliberately use stat instead of lstat to allow symlinks. */
+		if (stat(filterpath,&statbuf) == (-1)) {
+			goto free_fp;
+		}
+		if (!S_ISREG(statbuf.st_mode)) {
+			goto free_fp;
+		}
+		/*
+		 * We could check the mode in statbuf directly, or just skip
+		 * this entirely and check for EPERM after exec fails, but this
+		 * is cleaner.
+		 */
+		if (access(filterpath,X_OK) != 0) {
+			goto free_fp;
+		}
+		if (runcmd(filterpath,orig_volfile,NULL)) {
+			gf_log("",GF_LOG_ERROR,"failed to run filter %.*s",
+			       (int)sizeof(entry.d_name), entry.d_name);
+		}
+free_fp:
+		GF_FREE(filterpath);
 	}
-	else
-	        ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_DATASH,
-                                    &opt_datash);
-        if (ret) {
-                goto out;
-        }
+}
 
-        ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_DATASHALGO,
-                            &opt_datashalgo);
-        if (ret) {
-                goto out;
-        }
+static int
+volgen_write_volfile (volgen_graph_t *graph, char *filename)
+{
+        char        *ftmp = NULL;
+        FILE        *f = NULL;
+        int          fd   = 0;
+        xlator_t    *this = NULL;
 
-	if (dict_get (dict, "data-self-heal-window-size")) {
-		ret = dict_get_str (dict,"data-self-heal-window-size", 
-				    &opt_shwindowsize);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring data-self-heal"
-					 "window-size: %s", &opt_shwindowsize);
-		uncomment_option (replicate_str,
-				  (char *) "#   option data-self-heal-window-size %s\n");
-	}
-	else
-	        ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_SHWINDOWSIZE,
-                            &opt_shwindowsize);
-        if (ret) {
-                goto out;
+        this = THIS;
+
+        if (gf_asprintf (&ftmp, "%s.tmp", filename) == -1) {
+                ftmp = NULL;
+
+                goto error;
         }
 
-	if (dict_get (dict, "metadata-self-heal")) {
-		ret = dict_get_str (dict,"metadata-self-heal", 
-				    &opt_metash);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring metadata-self-heal"
-					 "count: %s", &opt_metash);
-		uncomment_option (replicate_str,
-				  (char *) "#   option metadata-self-heal %s\n");
-	}
-	else
-        	ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_METASH,
-                            &opt_metash);
-        if (ret) {
-                goto out;
+        fd = creat (ftmp, S_IRUSR | S_IWUSR);
+        if (fd < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "%s",
+                        strerror (errno));
+                goto error;
         }
-	
-	if (dict_get (dict, "entry-self-heal")) {
-		ret = dict_get_str (dict,"entry-self-heal", 
-				    &opt_entrysh);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring entry-self-heal"
-					 "count: %s", &opt_entrysh);
-		uncomment_option (replicate_str,
-				  (char *) "#   option entry-self-heal %s\n");
-	}
-	else
-	        ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_ENTRYSH,
-                            &opt_entrysh);
-        if (ret) {
-                goto out;
+
+        close (fd);
+
+        f = fopen (ftmp, "w");
+        if (!f)
+                goto error;
+
+        if (glusterfs_graph_print_file (f, &graph->graph) == -1)
+                goto error;
+
+        if (fclose (f) != 0) {
+                gf_log (THIS->name, GF_LOG_ERROR, "fclose on the file %s "
+                        "failed (%s)", ftmp, strerror (errno));
+                /*
+                 * Even though fclose has failed here, we have to set f to NULL.
+                 * Otherwise when the code path goes to error, there again we
+                 * try to close it which might cause undefined behavior such as
+                 * process crash.
+                 */
+                f = NULL;
+                goto error;
         }
 
-	if (dict_get (dict, "data-change-log")) {
-		ret = dict_get_str (dict,"data-change-log", 
-				    &opt_datachangelog);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring data-change-log"
-					 "count: %s", &opt_datachangelog);
-		uncomment_option (replicate_str,
-				  (char *) "#   option data-change-log %s\n");
-	}
-	else
-        	ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_DATACHANGELOG,
-                            &opt_datachangelog);
-        if (ret) {
-                goto out;
+        f = NULL;
+
+        if (rename (ftmp, filename) == -1)
+                goto error;
+
+        GF_FREE (ftmp);
+
+	volgen_apply_filters(filename);
+
+        return 0;
+
+ error:
+
+        GF_FREE (ftmp);
+        if (f)
+                fclose (f);
+
+        gf_log (this->name, GF_LOG_ERROR,
+                "failed to create volfile %s", filename);
+
+        return -1;
+}
+
+static void
+volgen_graph_free (volgen_graph_t *graph)
+{
+        xlator_t *trav = NULL;
+        xlator_t *trav_old = NULL;
+
+        for (trav = first_of (graph) ;; trav = trav->next) {
+                if (trav_old)
+                        xlator_destroy (trav_old);
+
+                trav_old = trav;
+
+                if (!trav)
+                        break;
         }
+}
 
-	if (dict_get (dict, "metadata-change-log")) {
-		ret = dict_get_str (dict,"metadata-change-log", 
-				    &opt_metadatachangelog);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring metadata-change-log"
-					 "count: %s", &opt_metadatachangelog);
-		uncomment_option (replicate_str,
-				  (char *) "#   option metadata-change-log %s\n");
-	}
-	else
-	        ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_METADATACHANGELOG,
-                            &opt_metadatachangelog);
-        if (ret) {
-                goto out;
+static int
+build_graph_generic (volgen_graph_t *graph, glusterd_volinfo_t *volinfo,
+                     dict_t *mod_dict, void *param,
+                     int (*builder) (volgen_graph_t *graph,
+                                     glusterd_volinfo_t *volinfo,
+                                     dict_t *set_dict, void *param))
+{
+        dict_t *set_dict = NULL;
+        int ret = 0;
+
+        if (mod_dict) {
+                set_dict = dict_copy (volinfo->dict, NULL);
+                if (!set_dict)
+                        return -1;
+                 dict_copy (mod_dict, set_dict);
+                 /* XXX dict_copy swallows errors */
+        } else {
+                set_dict = volinfo->dict;
         }
 
-	if (dict_get (dict, "entry-change-log")) {
-		ret = dict_get_str (dict, "entry-change-log", &opt_entrychangelog);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring entry-change-log: %s", 
-		       &opt_entrychangelog);
-		uncomment_option (replicate_str,
-				  (char *) "#   option entry-change-log %s\n");
-	}
-	else
+        ret = builder (graph, volinfo, set_dict, param);
+        if (!ret)
+                ret = volgen_graph_set_options (graph, set_dict);
 
-        ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_ENTRYCHANGELOG,
-                            &opt_entrychangelog);
-        if (ret) {
-                goto out;
+        if (mod_dict)
+                dict_destroy (set_dict);
+
+        return ret;
+}
+
+static gf_transport_type
+transport_str_to_type (char *tt)
+{
+        gf_transport_type type = GF_TRANSPORT_TCP;
+
+        if (!strcmp ("tcp", tt))
+                type = GF_TRANSPORT_TCP;
+        else if (!strcmp ("rdma", tt))
+                type = GF_TRANSPORT_RDMA;
+        else if (!strcmp ("tcp,rdma", tt))
+                type = GF_TRANSPORT_BOTH_TCP_RDMA;
+        return type;
+}
+
+static void
+transport_type_to_str (gf_transport_type type, char *tt)
+{
+        switch (type) {
+        case GF_TRANSPORT_RDMA:
+                strcpy (tt, "rdma");
+                break;
+        case GF_TRANSPORT_TCP:
+                strcpy (tt, "tcp");
+                break;
+        case GF_TRANSPORT_BOTH_TCP_RDMA:
+                strcpy (tt, "tcp,rdma");
+                break;
         }
+}
 
-	if (dict_get (dict, "strict-readdir")) {
-		ret = dict_get_str (dict,"strict-readdir", 
-				    &opt_strictreaddir);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring sstrict-readdir"
-					 "count: %s", &opt_strictreaddir);
-		uncomment_option (replicate_str,
-				  (char *) "#   option strict-readdir %s\n");
-	}
-	else
-        	ret = dict_get_str (dict, VOLGEN_REPLICATE_OPTION_STRICTREADDIR,
-                            &opt_strictreaddir);
-        if (ret) {
-                goto out;
+static void
+get_vol_transport_type (glusterd_volinfo_t *volinfo, char *tt)
+{
+        transport_type_to_str (volinfo->transport_type, tt);
+}
+
+static void
+get_vol_nfs_transport_type (glusterd_volinfo_t *volinfo, char *tt)
+{
+        if (volinfo->nfs_transport_type == GF_TRANSPORT_BOTH_TCP_RDMA) {
+                gf_log ("", GF_LOG_ERROR, "%s:nfs transport cannot be both"
+                        " tcp and rdma", volinfo->volname);
+                GF_ASSERT (0);
         }
+        transport_type_to_str (volinfo->nfs_transport_type, tt);
+}
 
-        for (i = 0; i < replicate_count; i++) {
-                snprintf (tmp, 4096, "%s-%d ", subvolume,
-                          subvolume_count);
-                len = strlen (tmp);
-                subvol_len += len;
-                subvolume_count++;
+/*  gets the volinfo, dict, a character array for filling in
+ *  the transport type and a boolean option which says whether
+ *  the transport type is required for nfs or not. If its not
+ *  for nfs, then it is considered as the client transport
+ *  and client transport type is filled in the character array
+ */
+static void
+get_transport_type (glusterd_volinfo_t *volinfo, dict_t *set_dict,
+                    char *transt, gf_boolean_t is_nfs)
+{
+        int    ret   = -1;
+        char  *tt    = NULL;
+        char  *key   = NULL;
+        typedef void (*transport_type) (glusterd_volinfo_t *volinfo, char *tt);
+        transport_type get_transport;
+
+        if (is_nfs == _gf_false) {
+                key = "client-transport-type";
+                get_transport = get_vol_transport_type;
+        } else {
+                key = "nfs.transport-type";
+                get_transport = get_vol_nfs_transport_type;
         }
 
-        subvolume_count = subvol_count;
-        subvol_len++;
+        ret = dict_get_str (set_dict, key, &tt);
+        if (ret)
+                get_transport (volinfo, transt);
+        if (!ret)
+                strcpy (transt, tt);
+}
 
-        subvol_str = GF_CALLOC (1, subvol_len, gf_gld_mt_char);
-        if (!subvol_str) {
-                gf_log ("glusterd", GF_LOG_ERROR,
-                        "Out of memory");
-                ret = -1;
-                goto out;
+static int
+server_auth_option_handler (volgen_graph_t *graph,
+                            struct volopt_map_entry *vme, void *param)
+{
+        xlator_t *xl = NULL;
+        xlator_list_t *trav = NULL;
+        char *aa = NULL;
+        int   ret   = 0;
+        char *key = NULL;
+
+        if (strcmp (vme->option, "!server-auth") != 0)
+                return 0;
+
+        xl = first_of (graph);
+
+        /* from 'auth.allow' -> 'allow', and 'auth.reject' -> 'reject' */
+        key = strchr (vme->key, '.') + 1;
+
+        for (trav = xl->children; trav; trav = trav->next) {
+                ret = gf_asprintf (&aa, "auth.addr.%s.%s", trav->xlator->name,
+                                   key);
+                if (ret != -1) {
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
+                }
+                if (ret)
+                        return -1;
         }
 
-        for (i = 0; i < replicate_count ; i++) {
-                snprintf (tmp, 4096, "%s-%d ", subvolume,
-                          subvolume_count);
-                strncat (subvol_str, tmp, strlen (tmp));
-                subvolume_count++;
-        }
+        return 0;
+}
 
-        fprintf (file, replicate_str,
-                 volname,
-                 "replicate",
-                 count,
-                 opt_readsubvol,
-                 opt_favchild,
-                 opt_bckshcount,
-                 opt_datash,
-                 opt_datashalgo,
-                 opt_shwindowsize,
-                 opt_metash,
-                 opt_entrysh,
-                 opt_datachangelog,
-                 opt_metadatachangelog,
-                 opt_entrychangelog,
-                 opt_strictreaddir,
-                 subvol_str);
+static int
+loglevel_option_handler (volgen_graph_t *graph,
+                         struct volopt_map_entry *vme, void *param)
+{
+        char *role = param;
+        struct volopt_map_entry vme2 = {0,};
 
+        if ( (strcmp (vme->option, "!client-log-level") != 0 &&
+               strcmp (vme->option, "!brick-log-level") != 0)
+            || !strstr (vme->key, role))
+                return 0;
 
-        ret = 0;
+        memcpy (&vme2, vme, sizeof (vme2));
+        vme2.option = "log-level";
 
-out:
-        if (subvol_str)
-                GF_FREE (subvol_str);
-        return ret;
+        return basic_option_handler (graph, &vme2, NULL);
 }
 
 static int
-__write_stripe_xlator (FILE *file, dict_t *dict,
-                       char *subvolume,
-                       int stripe_count,
-                       int subvol_count,
-                       int count)
+server_check_marker_off (volgen_graph_t *graph, struct volopt_map_entry *vme,
+                         glusterd_volinfo_t *volinfo)
 {
-        char *volname = NULL;
-        char       *opt_blocksize    = NULL;
-        char       *opt_usexattr     = NULL;
-        char       *subvol_str       = NULL;
-        char        tmp[4096]        = {0,};
-        int         subvolume_count  = 0;
-        int         ret              = -1;
-        int         i                = 0;
-        int         subvol_len        = 0;
-        int         len              = 0;
-
-        char stripe_str[] = "volume %s-%s-%d\n"
-                "    type cluster/stripe\n"
-                "#   option block-size %s\n"
-                "#   option use-xattr %s\n"
-                "    subvolumes %s\n"
-                "end-volume\n\n";
-
-        subvolume_count = subvol_count;
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
+        gf_boolean_t           bool = _gf_false;
+        int                    ret = 0;
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (vme);
+
+        if (strcmp (vme->option, "!xtime") != 0)
+                return 0;
+
+        ret = gf_string2boolean (vme->value, &bool);
+        if (ret || bool)
                 goto out;
-        }
 
-	if (dict_get (dict, "block-size")) {
-		ret = dict_get_str (dict, "block-size", &opt_blocksize);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring Stripe Count %s", 
-		       opt_blocksize);
-		uncomment_option (stripe_str,
-				  (char *) "#   option block-size %s\n");
-	}
-	else
-        	ret = dict_get_str (dict, VOLGEN_STRIPE_OPTION_BLOCKSIZE,
-                            &opt_blocksize);
-        if (ret) {
+        ret = glusterd_volinfo_get_boolean (volinfo, VKEY_MARKER_XTIME);
+        if (ret < 0) {
+                gf_log ("", GF_LOG_WARNING, "failed to get the marker status");
+                ret = -1;
                 goto out;
         }
 
-        ret = dict_get_str (dict, VOLGEN_STRIPE_OPTION_USEXATTR,
-                            &opt_usexattr);
         if (ret) {
-                goto out;
-        }
+                bool = _gf_false;
+                ret = glusterd_check_gsync_running (volinfo, &bool);
 
+                if (bool) {
+                        gf_log ("", GF_LOG_WARNING, GEOREP" sessions active"
+                                "for the volume %s, cannot disable marker "
+                                ,volinfo->volname);
+                        set_graph_errstr (graph,
+                                          VKEY_MARKER_XTIME" cannot be disabled "
+                                          "while "GEOREP" sessions exist");
+                        ret = -1;
+                        goto out;
+                }
 
-        for (i = 0; i < stripe_count; i++) {
-                snprintf (tmp, 4096, "%s-%d ", subvolume, subvolume_count);
-                len = strlen (tmp);
-                subvol_len += len;
-                subvolume_count++;
+                if (ret) {
+                        gf_log ("", GF_LOG_WARNING, "Unable to get the status"
+                                 " of active gsync session");
+                        goto out;
+                }
         }
 
-        subvolume_count = subvol_count;
-        subvol_len++;
+        ret = 0;
+ out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
 
-        subvol_str = GF_CALLOC (1, subvol_len, gf_gld_mt_char);
-        if (!subvol_str) {
-                gf_log ("glusterd", GF_LOG_ERROR,
-                        "Out of memory");
-                ret = -1;
-                goto out;
-        }
+}
 
-        for (i = 0; i < stripe_count; i++) {
-                snprintf (tmp, 4096, "%s-%d ", subvolume, subvolume_count);
-                strncat (subvol_str, tmp, strlen (tmp));
-                subvolume_count++;
-        }
+static int
+sys_loglevel_option_handler (volgen_graph_t *graph,
+                             struct volopt_map_entry *vme,
+                             void *param)
+{
+        char  *role = NULL;
+        struct volopt_map_entry vme2 = {0,};
 
-        fprintf (file, stripe_str,
-                 volname,
-                 "stripe",
-                 count,
-                 opt_blocksize,
-                 opt_usexattr,
-                 subvol_str);
+        role = (char *) param;
 
+        if (strcmp (vme->option, "!sys-log-level") != 0 ||
+            !strstr (vme->key, role))
+                return 0;
 
-        ret = 0;
+        memcpy (&vme2, vme, sizeof (vme2));
+        vme2.option = "sys-log-level";
 
-out:
-        if (subvol_str)
-                GF_FREE (subvol_str);
-        return ret;
-        
+        return basic_option_handler (graph, &vme2, NULL);
 }
 
 static int
-__write_distribute_xlator (FILE *file, dict_t *dict,
-                           char *subvolume,
-                           int dist_count)
-{
-        char       *volname          = NULL;
-        char       *subvol_str       = NULL;
-        char       tmp[4096]         = {0,};
-        char       *opt_lookupunhash = NULL;
-        char       *opt_minfreedisk  = NULL;
-        char       *opt_unhashsticky = NULL;
-        int        ret               = -1;
-        int        i                 = 0;
-        int        subvol_len        = 0;
-        int        len               = 0;
-        
-        char       dht_str[] = "volume %s-%s\n"
-                "type cluster/distribute\n"
-                "#   option lookup-unhashed %s\n"
-                "#   option min-free-disk %s\n"
-                "#   option unhashed-sticky-bit %s\n"
-                "    subvolumes %s\n"
-                "end-volume\n\n";
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                goto out;
-        }
+volgen_graph_set_xl_options (volgen_graph_t *graph, dict_t *dict)
+{
+        int32_t   ret               = -1;
+        char     *xlator            = NULL;
+        char     xlator_match[1024] = {0,}; /* for posix* -> *posix* */
+        char     *loglevel          = NULL;
+        xlator_t *trav              = NULL;
 
-	if (dict_get (dict, "lookup-unhashed")) {
-		ret = dict_get_str (dict, "lookup-unhashed", &opt_lookupunhash);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring lookup-unhashed %s",
-		       opt_lookupunhash);
-		uncomment_option (dht_str, 
-				  (char *) "#   option lookup-unhashed %s\n");
-	}
-	else
-        	ret = dict_get_str (dict, VOLGEN_DHT_OPTION_LOOKUPUNHASH,
-                            &opt_lookupunhash);
-        if (ret) {
+        ret = dict_get_str (dict, "xlator", &xlator);
+        if (ret)
                 goto out;
-        }
 
-	if (dict_get (dict, "min-free-disk")) {
-		ret = dict_get_str (dict, "min-free-disk", &opt_minfreedisk);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring min-free-disk",
-		       opt_minfreedisk);
-		uncomment_option (dht_str, 
-				  (char *) "#   option min-free-disk %s\n");
-	}
-	else
-        ret = dict_get_str (dict, VOLGEN_DHT_OPTION_MINFREEDISK,
-                            &opt_minfreedisk);
-        if (ret) {
+        ret = dict_get_str (dict, "loglevel", &loglevel);
+        if (ret)
                 goto out;
-        }
 
-        ret = dict_get_str (dict, VOLGEN_DHT_OPTION_UNHASHSTICKY,
-                            &opt_unhashsticky);
-        if (ret) {
-                goto out;
-        }
+        snprintf (xlator_match, 1024, "*%s", xlator);
 
-        for (i = 0; i < dist_count; i++) {
-                snprintf (tmp, 4096, "%s-%d ", subvolume, i);
-                len = strlen (tmp);
-                subvol_len += len;
+        for (trav = first_of (graph); trav; trav = trav->next) {
+                if (fnmatch(xlator_match, trav->type, FNM_NOESCAPE) == 0) {
+                        gf_log ("glusterd", GF_LOG_DEBUG, "Setting log level for xlator: %s",
+                                trav->type);
+                        ret = xlator_set_option (trav, "log-level", loglevel);
+                        if (ret)
+                                break;
+                }
         }
 
-        subvol_len++;
-        subvol_str = GF_CALLOC (1, subvol_len, gf_gld_mt_char);
-        if (!subvol_str) {
-                gf_log ("glusterd", GF_LOG_ERROR,
-                        "Out of memory");
-                ret = -1;
-                goto out;
-        }
+ out:
+        return ret;
+}
 
-        for (i = 0; i < dist_count ; i++) {
-                snprintf (tmp, 4096, "%s-%d ", subvolume, i);
-                strncat (subvol_str, tmp, strlen (tmp));
-        }
+static int
+server_spec_option_handler (volgen_graph_t *graph,
+                            struct volopt_map_entry *vme, void *param)
+{
+        int                     ret = 0;
+        glusterd_volinfo_t      *volinfo = NULL;
 
-        fprintf (file, dht_str,
-                 volname,
-                 "dht",
-                 opt_lookupunhash,
-                 opt_minfreedisk,
-                 opt_unhashsticky,
-                 subvol_str);
+        volinfo = param;
 
+        ret = server_auth_option_handler (graph, vme, NULL);
+        if (!ret)
+                ret = server_check_marker_off (graph, vme, volinfo);
 
-        ret = 0;
+        if (!ret)
+                ret = loglevel_option_handler (graph, vme, "brick");
+
+        if (!ret)
+                ret = sys_loglevel_option_handler (graph, vme, "brick");
 
-out:
-        if (subvol_str)
-                GF_FREE (subvol_str);
         return ret;
 }
 
-
-int
-uncomment_option( char *opt_str,char *comment_str)
+static int
+server_spec_extended_option_handler (volgen_graph_t *graph,
+                                     struct volopt_map_entry *vme, void *param)
 {
-	char *ptr;
+        int     ret           = 0;
+        dict_t *dict          = NULL;
 
-	ptr = strstr (opt_str,comment_str);
-	if (!ptr)
-		return -1;
-	
-	if (*ptr != '#')
-		return -1;
+        GF_ASSERT (param);
+        dict = (dict_t *)param;
 
-	*ptr = ' ';
+        ret = server_auth_option_handler (graph, vme, NULL);
+        if (!ret)
+                ret = volgen_graph_set_xl_options (graph, dict);
 
-	return 0;
-} 
+        return ret;
+}
+
+static void get_vol_tstamp_file (char *filename, glusterd_volinfo_t *volinfo);
 
 static int
-__write_wb_xlator (FILE *file, dict_t *dict,
-                   char *subvolume)
-{
-        char        *volname              = NULL;
-        char        *opt_flushbehind     = NULL;
-        char        *opt_cachesize       = NULL;
-        char        *opt_disablenbytes   = NULL;
-        char        *opt_osync           = NULL;
-        char        *opt_tricklingwrites = NULL;
-        int          ret                 = -1;
-
-        char dht_str[] = "volume %s-%s\n"
-                "    type performance/write-behind\n"
-                "#   option flush-behind %s\n"
-                "#   option cache-size %s\n"
-                "#   option disable-for-first-nbytes %s\n"
-                "#   option enable-O_SYNC %s\n"
-                "#   option enable-trickling-writes %s\n"
-                "    subvolumes %s\n"
-                "end-volume\n\n";
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                goto out;
+server_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo,
+                      dict_t *set_dict, void *param)
+{
+        char     *volname                 = NULL;
+        char     *path                    = NULL;
+        int       pump                    = 0;
+        xlator_t *xl                      = NULL;
+        xlator_t *txl                     = NULL;
+        xlator_t *rbxl                    = NULL;
+        char      transt[16]              = {0,};
+        char     *ptranst                 = NULL;
+        char      volume_id[64]           = {0,};
+        char      tstamp_file[PATH_MAX]   = {0,};
+        int       ret                     = 0;
+        char     *xlator                  = NULL;
+        char     *loglevel                = NULL;
+        char     *username                = NULL;
+        char     *password                = NULL;
+        char     index_basepath[PATH_MAX] = {0};
+        char     key[1024]                = {0};
+        glusterd_brickinfo_t *brickinfo   = NULL;
+        char changelog_basepath[PATH_MAX] = {0,};
+
+        brickinfo = param;
+        path      = brickinfo->path;
+        volname = volinfo->volname;
+        get_vol_transport_type (volinfo, transt);
+
+        ret = dict_get_str (set_dict, "xlator", &xlator);
+
+        /* got a cli log level request */
+        if (!ret) {
+                ret = dict_get_str (set_dict, "loglevel", &loglevel);
+                if (ret) {
+                        gf_log ("glusterd", GF_LOG_ERROR, "could not get both"
+                                " translator name and loglevel for log level request");
+                        goto out;
+                }
         }
 
-        ret = dict_get_str (dict, VOLGEN_WB_OPTION_FLUSHBEHIND,
-                            &opt_flushbehind);
-        if (ret) {
-                goto out;
+        xl = volgen_graph_add (graph, "storage/posix", volname);
+        if (!xl)
+                return -1;
+
+        ret = xlator_set_option (xl, "directory", path);
+        if (ret)
+                return -1;
+
+        ret = xlator_set_option (xl, "volume-id",
+                                 uuid_utoa (volinfo->volume_id));
+        if (ret)
+                return -1;
+
+        ret = check_and_add_debug_xl (graph, set_dict, volname,
+                                      "posix");
+        if (ret)
+                return -1;
+#ifdef HAVE_BD_XLATOR
+        if (*brickinfo->vg != '\0') {
+                /* Now add BD v2 xlator if volume is BD type */
+                xl = volgen_graph_add (graph, "storage/bd", volname);
+                if (!xl)
+                        return -1;
+
+                ret = xlator_set_option (xl, "device", "vg");
+                if (ret)
+                        return -1;
+                ret = xlator_set_option (xl, "export", brickinfo->vg);
+                if (ret)
+                        return -1;
+
+                ret = check_and_add_debug_xl (graph, set_dict, volname, "bd");
+                if (ret)
+                        return -1;
+
         }
+#endif
 
-	if (dict_get (dict, "cache-size")) {
-		gf_log("",GF_LOG_DEBUG, "Uncommenting option cache-size");
-		uncomment_option (dht_str,(char *) "#   option cache-size %s\n");
-		ret = dict_get_str (dict, "cache-size", &opt_cachesize);
-	}
-	else
-        	ret = dict_get_str (dict, VOLGEN_WB_OPTION_CACHESIZE,
-                            &opt_cachesize);
-        if (ret) {
-                goto out;
+        xl = volgen_graph_add (graph, "features/changelog", volname);
+        if (!xl)
+                return -1;
+
+        ret = xlator_set_option (xl, "changelog-brick", path);
+        if (ret)
+                return -1;
+
+        snprintf (changelog_basepath, sizeof (changelog_basepath),
+                  "%s/%s", path, ".glusterfs/changelogs");
+        ret = xlator_set_option (xl, "changelog-dir", changelog_basepath);
+        if (ret)
+                return -1;
+
+        ret = check_and_add_debug_xl (graph, set_dict, volname, "changelog");
+        if (ret)
+                return -1;
+
+        xl = volgen_graph_add (graph, "features/access-control", volname);
+        if (!xl)
+                return -1;
+
+        ret = check_and_add_debug_xl (graph, set_dict, volname, "acl");
+        if (ret)
+                return -1;
+
+        xl = volgen_graph_add (graph, "features/locks", volname);
+        if (!xl)
+                return -1;
+
+        ret = check_and_add_debug_xl (graph, set_dict, volname, "locks");
+        if (ret)
+                return -1;
+
+        xl = volgen_graph_add (graph, "performance/io-threads", volname);
+        if (!xl)
+                return -1;
+
+        ret = check_and_add_debug_xl (graph, set_dict, volname, "io-threads");
+        if (ret)
+                return -1;
+
+        ret = dict_get_int32 (volinfo->dict, "enable-pump", &pump);
+        if (ret == -ENOENT)
+                ret = pump = 0;
+        if (ret)
+                return -1;
+
+        username = glusterd_auth_get_username (volinfo);
+        password = glusterd_auth_get_password (volinfo);
+
+        if (pump) {
+                txl = first_of (graph);
+
+                rbxl = volgen_graph_add_nolink (graph, "protocol/client",
+                                                "%s-replace-brick", volname);
+                if (!rbxl)
+                        return -1;
+
+		ptranst = glusterd_get_trans_type_rb (volinfo->transport_type);
+		if (NULL == ptranst)
+			return -1;
+
+                if (username) {
+                        ret = xlator_set_option (rbxl, "username", username);
+                        if (ret)
+                                return -1;
+                }
+
+                if (password) {
+                        ret = xlator_set_option (rbxl, "password", password);
+                        if (ret)
+                                return -1;
+                }
+
+                ret = xlator_set_option (rbxl, "transport-type", ptranst);
+                GF_FREE (ptranst);
+                if (ret)
+                        return -1;
+
+                xl = volgen_graph_add_nolink (graph, "cluster/pump", "%s-pump",
+                                              volname);
+                if (!xl)
+                        return -1;
+                ret = volgen_xlator_link (xl, txl);
+                if (ret)
+                        return -1;
+                ret = volgen_xlator_link (xl, rbxl);
+                if (ret)
+                        return -1;
         }
 
-        ret = dict_get_str (dict, VOLGEN_WB_OPTION_DISABLENBYTES,
-                            &opt_disablenbytes);
-        if (ret) {
+        xl = volgen_graph_add (graph, "features/index", volname);
+        if (!xl)
+                return -1;
+
+        snprintf (index_basepath, sizeof (index_basepath), "%s/%s",
+                  path, ".glusterfs/indices");
+        ret = xlator_set_option (xl, "index-base", index_basepath);
+        if (ret)
+                return -1;
+
+        ret = check_and_add_debug_xl (graph, set_dict, volname,
+                                      "index");
+        if (ret)
+                return -1;
+
+        xl = volgen_graph_add (graph, "features/marker", volname);
+        if (!xl)
+                return -1;
+
+        uuid_unparse (volinfo->volume_id, volume_id);
+        ret = xlator_set_option (xl, "volume-uuid", volume_id);
+        if (ret)
+                return -1;
+        get_vol_tstamp_file (tstamp_file, volinfo);
+        ret = xlator_set_option (xl, "timestamp-file", tstamp_file);
+        if (ret)
+                return -1;
+
+        ret = check_and_add_debug_xl (graph, set_dict, volname, "marker");
+        if (ret)
+                return -1;
+
+        if (dict_get_str_boolean (set_dict, "features.read-only", 0) &&
+            dict_get_str_boolean (set_dict, "features.worm",0)) {
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "read-only and worm cannot be set together");
+                ret = -1;
                 goto out;
         }
 
-        ret = dict_get_str (dict, VOLGEN_WB_OPTION_OSYNC,
-                            &opt_osync);
-        if (ret) {
-                goto out;
+        /* Check for read-only volume option, and add it to the graph */
+        if (dict_get_str_boolean (set_dict, "features.read-only", 0)
+                || volinfo -> is_snap_volume) {
+                xl = volgen_graph_add (graph, "features/read-only", volname);
+                if (!xl) {
+                        ret = -1;
+                        goto out;
+                }
         }
 
-        ret = dict_get_str (dict, VOLGEN_WB_OPTION_TRICKLINGWRITES,
-                            &opt_tricklingwrites);
-        if (ret) {
-                goto out;
+        /* Check for worm volume option, and add it to the graph */
+        if (dict_get_str_boolean (set_dict, "features.worm", 0)) {
+                xl = volgen_graph_add (graph, "features/worm", volname);
+                if (!xl) {
+                        ret = -1;
+                        goto out;
+                }
         }
 
-        fprintf (file, dht_str,
-                 volname,
-                 "write-behind",
-                 opt_flushbehind,
-                 opt_cachesize,
-                 opt_disablenbytes,
-                 opt_osync,
-                 opt_tricklingwrites,
-                 subvolume);
+        /* Check for compress volume option, and add it to the graph on server side */
+        if (dict_get_str_boolean (set_dict, "features.compress", 0)) {
+                xl = volgen_graph_add (graph, "features/cdc", volname);
+                if (!xl) {
+                        ret = -1;
+                        goto out;
+                }
+                ret = dict_set_str (set_dict, "compress.mode", "server");
+                if (ret)
+                        goto out;
+        }
 
+        xl = volgen_graph_add_as (graph, "debug/io-stats", path);
+        if (!xl)
+                return -1;
 
-        ret = 0;
+        xl = volgen_graph_add (graph, "protocol/server", volname);
+        if (!xl)
+                return -1;
+        ret = xlator_set_option (xl, "transport-type", transt);
+        if (ret)
+                return -1;
 
-out:
+        /*In the case of running multiple glusterds on a single machine,
+         * we should ensure that bricks don't listen on all IPs on that
+         * machine and break the IP based separation being brought about.*/
+        if (dict_get (THIS->options, "transport.socket.bind-address")) {
+                ret = xlator_set_option (xl, "transport.socket.bind-address",
+                                         brickinfo->hostname);
+                if (ret)
+                        return -1;
+        }
+
+        if (username) {
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), "auth.login.%s.allow", path);
+
+                ret = xlator_set_option (xl, key, username);
+                if (ret)
+                        return -1;
+        }
+
+        if (password) {
+                memset (key, 0, sizeof (key));
+                snprintf (key, sizeof (key), "auth.login.%s.password",
+                          username);
+
+                ret = xlator_set_option (xl, key, password);
+                if (ret)
+                        return -1;
+        }
+
+        ret = volgen_graph_set_options_generic (graph, set_dict,
+                                                (xlator && loglevel) ? (void *)set_dict : volinfo,
+                                                (xlator && loglevel) ?  &server_spec_extended_option_handler :
+                                                 &server_spec_option_handler);
+
+ out:
         return ret;
 }
 
+
+/* builds a graph for server role , with option overrides in mod_dict */
 static int
-__write_ra_xlator (FILE *file, dict_t *dict,
-                   char *subvolume)
-{
-        char       *volname       = NULL;
-        char       *opt_atime     = NULL;
-        char       *opt_pagecount = NULL;
-        int         ret           = -1;
-
-        const char *ra_str = "volume %s-%s\n"
-                "    type performance/read-ahead\n"
-                "#   option force-atime-update %s\n"
-                "#   option page-count %s\n"
-                "    subvolumes %s\n"
-                "end-volume\n\n";
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                goto out;
-        }
+build_server_graph (volgen_graph_t *graph, glusterd_volinfo_t *volinfo,
+                    dict_t *mod_dict, glusterd_brickinfo_t *brickinfo)
+{
+        return build_graph_generic (graph, volinfo, mod_dict, brickinfo,
+                                    &server_graph_builder);
+}
 
-        ret = dict_get_str (dict, VOLGEN_RA_OPTION_ATIME,
-                            &opt_atime);
-        if (ret) {
-                goto out;
-        }
+static int
+perfxl_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme,
+                       void *param)
+{
+        gf_boolean_t enabled = _gf_false;
+        glusterd_volinfo_t *volinfo = NULL;
 
-        ret = dict_get_str (dict, VOLGEN_RA_OPTION_PAGECOUNT,
-                            &opt_pagecount);
-        if (ret) {
-                goto out;
-        }
+        GF_ASSERT (param);
+        volinfo = param;
 
-        fprintf (file, ra_str,
-                 volname,
-                 "read-ahead",
-                 opt_atime,
-                 opt_pagecount,
-                 subvolume);
+        if (strcmp (vme->option, "!perf") != 0)
+                return 0;
 
+        if (gf_string2boolean (vme->value, &enabled) == -1)
+                return -1;
+        if (!enabled)
+                return 0;
 
-        ret = 0;
+        /* Check op-version before adding the 'open-behind' xlator in the graph
+         */
+        if (!strcmp (vme->key, "performance.open-behind") &&
+            (vme->op_version > volinfo->client_op_version))
+                return 0;
 
-out:
-        return ret;
+        if (volgen_graph_add (graph, vme->voltype, volinfo->volname))
+                return 0;
+        else
+                return -1;
 }
 
 static int
-__write_iocache_xlator (FILE *file, dict_t *dict,
-                        char *subvolume)
-{
-        char       *volname         = NULL;
-        char       *opt_priority    = NULL;
-        char       *opt_timeout     = NULL;
-        char       *opt_cachesize   = NULL;
-        char       *opt_minfilesize = NULL;
-        char       *opt_maxfilesize = NULL;
-        int         ret             = -1;
-
-        char        iocache_str[] = "volume %s-%s\n"
-                "    type performance/io-cache\n"
-                "#   option priority %s\n"
-                "#   option cache-timeout %s\n"
-                "#   option cache-size %s\n"
-                "#   option min-file-size %s\n"
-                "#   option max-file-size %s\n"
-                "    subvolumes %s\n"
-                "end-volume\n\n";
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
+nfsperfxl_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme,
+                       void *param)
+{
+        char *volname = NULL;
+        gf_boolean_t enabled = _gf_false;
+
+        volname = param;
+
+        if (strcmp (vme->option, "!nfsperf") != 0)
+                return 0;
+
+        if (gf_string2boolean (vme->value, &enabled) == -1)
+                return -1;
+        if (!enabled)
+                return 0;
+
+        if (volgen_graph_add (graph, vme->voltype, volname))
+                return 0;
+        else
+                return -1;
+}
+
+#if (HAVE_LIB_XML)
+static int
+end_sethelp_xml_doc (xmlTextWriterPtr writer)
+{
+        int             ret = -1;
+
+        ret = xmlTextWriterEndElement(writer);
+        if (ret < 0) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Could not end an "
+                        "xmlElemetnt");
+                ret = -1;
                 goto out;
         }
-
-	if (dict_get (dict, "priority")) {
-		ret = dict_get_str (dict, "priority", &opt_priority);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring priority",
-		       opt_priority);
-		uncomment_option (iocache_str, 
-				  (char *) "#   option priority %s\n");
-	}
-	else
-        	ret = dict_get_str (dict, VOLGEN_IOCACHE_OPTION_PRIORITY,
-                            &opt_priority);
-        if (ret) {
+        ret = xmlTextWriterEndDocument (writer);
+        if (ret < 0) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Could not end an "
+                        "xmlDocument");
+                ret = -1;
                 goto out;
         }
+        ret = 0;
+ out:
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
 
-	if (dict_get (dict, "cache-timeout")) {
-		ret = dict_get_str (dict, "cache-timeout", &opt_timeout);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring cache-timeout",
-		       opt_timeout);
-		uncomment_option (iocache_str, 
-				  (char *) "#   option cache-timeout %s\n");
-	}
-	else
-        	ret = dict_get_str (dict, VOLGEN_IOCACHE_OPTION_TIMEOUT,
-                            &opt_timeout);
-        if (ret) {
+}
+
+static int
+init_sethelp_xml_doc (xmlTextWriterPtr *writer, xmlBufferPtr  *buf)
+{
+        int ret;
+
+        *buf = xmlBufferCreateSize (8192);
+        if (buf == NULL) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Error creating the xml "
+                          "buffer");
+                ret = -1;
                 goto out;
         }
 
-	if (dict_get (dict, "cache-size")) {
-		ret = dict_get_str (dict, "cache-size", &opt_cachesize);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring cache-size :%s",
-		       opt_cachesize);
-		uncomment_option (iocache_str, 
-				  (char *) "#   option cache-size %s\n");
-	}
-	else
-	        ret = dict_get_str (dict, VOLGEN_IOCACHE_OPTION_CACHESIZE,
-                            &opt_cachesize);
-        if (ret) {
+        xmlBufferSetAllocationScheme (*buf,XML_BUFFER_ALLOC_DOUBLEIT);
+
+        *writer = xmlNewTextWriterMemory(*buf, 0);
+        if (writer == NULL) {
+                gf_log ("glusterd", GF_LOG_ERROR, " Error creating the xml "
+                         "writer");
+                ret = -1;
                 goto out;
         }
 
-	if (dict_get (dict, "min-file-size")) {
-		ret = dict_get_str (dict, "min-file-size", &opt_minfilesize);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring min-file-size: %s",
-		       opt_minfilesize);
-		uncomment_option (iocache_str, 
-				  (char *) "#   option min-file-size %s\n");
-	}
-	else
-        	ret = dict_get_str (dict, VOLGEN_IOCACHE_OPTION_MINFILESIZE,
-                            &opt_minfilesize);
-        if (ret) {
+        ret = xmlTextWriterStartDocument(*writer, "1.0", "UTF-8", "yes");
+        if (ret < 0) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Error While starting the "
+                         "xmlDoc");
                 goto out;
         }
 
-	if (dict_get (dict, "max-file-size")) {
-		ret = dict_get_str (dict, "max-file-size", &opt_maxfilesize);
-		gf_log("", GF_LOG_DEBUG, "Reconfiguring max-file-size: %s",
-		       opt_maxfilesize);
-		uncomment_option (iocache_str, 
-				  (char *) "#   option max-file-size %s\n");
-	}
-	else
-        	ret = dict_get_str (dict, VOLGEN_IOCACHE_OPTION_MAXFILESIZE,
-                            &opt_maxfilesize);
-        if (ret) {
+        ret = xmlTextWriterStartElement(*writer, (xmlChar *)"options");
+        if (ret < 0) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Could not create an "
+                        "xmlElemetnt");
+                ret = -1;
                 goto out;
         }
 
-        fprintf (file, iocache_str,
-                 volname,
-                 "io-cache",
-                 opt_priority,
-                 opt_timeout,
-                 opt_cachesize,
-                 opt_minfilesize,
-                 opt_maxfilesize,
-                 subvolume);
-
 
         ret = 0;
 
-out:
+ out:
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
+
 }
 
 static int
-__write_qr_xlator (FILE *file, dict_t *dict,
-                        char *subvolume)
-{
-        char       *volname         = NULL;
-        char       *opt_priority    = NULL;
-        char       *opt_timeout     = NULL;
-        char       *opt_cachesize   = NULL;
-        char       *opt_maxfilesize = NULL;
-        int         ret             = -1;
-
-        const char *qr_str = "volume %s-%s\n"
-                "    type performance/quick-read\n"
-                "#   option priority %s\n"
-                "#   option cache-timeout %s\n"
-                "#   option cache-size %s\n"
-                "#   option max-file-size %s\n"
-                "    subvolumes %s\n"
-                "end-volume\n\n";
-
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
+xml_add_volset_element (xmlTextWriterPtr writer, const char *name,
+                                 const char *def_val, const char *dscrpt)
+{
+
+        int                     ret = -1;
+
+        GF_ASSERT (name);
+
+        ret = xmlTextWriterStartElement(writer, (xmlChar *) "option");
+        if (ret < 0) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Could not create an "
+                        "xmlElemetnt");
+                ret = -1;
                 goto out;
         }
 
-        ret = dict_get_str (dict, VOLGEN_QR_OPTION_PRIORITY,
-                            &opt_priority);
-        if (ret) {
+        ret = xmlTextWriterWriteFormatElement(writer, (xmlChar*)"defaultValue",
+                                              "%s", def_val);
+        if (ret < 0) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Could not create an "
+                        "xmlElemetnt");
+                ret = -1;
                 goto out;
         }
 
-        ret = dict_get_str (dict, VOLGEN_QR_OPTION_TIMEOUT,
-                            &opt_timeout);
-        if (ret) {
+        ret = xmlTextWriterWriteFormatElement(writer, (xmlChar *)"description",
+                                              "%s",  dscrpt );
+        if (ret < 0) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Could not create an "
+                        "xmlElemetnt");
+                ret = -1;
                 goto out;
         }
 
-        ret = dict_get_str (dict, VOLGEN_QR_OPTION_CACHESIZE,
-                            &opt_cachesize);
-        if (ret) {
+        ret = xmlTextWriterWriteFormatElement(writer, (xmlChar *) "name", "%s",
+                                               name);
+        if (ret < 0) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Could not create an "
+                        "xmlElemetnt");
+                ret = -1;
                 goto out;
         }
 
-        ret = dict_get_str (dict, VOLGEN_QR_OPTION_MAXFILESIZE,
-                            &opt_maxfilesize);
-        if (ret) {
+        ret = xmlTextWriterEndElement(writer);
+        if (ret < 0) {
+                gf_log ("glusterd", GF_LOG_ERROR, "Could not end an "
+                        "xmlElemetnt");
+                ret = -1;
                 goto out;
         }
 
-        fprintf (file, qr_str,
-                 volname,
-                 "quick-read",
-                 opt_priority,
-                 opt_timeout,
-                 opt_cachesize,
-                 opt_maxfilesize,
-                 subvolume);
-
         ret = 0;
-
-out:
+ out:
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
+
 }
 
+#endif
+
 static int
-__write_statprefetch_xlator (FILE *file, dict_t *dict,
-                             char *subvolume)
+_get_xlator_opt_key_from_vme ( struct volopt_map_entry *vme, char **key)
 {
-        char *volname = NULL;
-        int   ret     = -1;
+        int ret = 0;
 
-        const char *statprefetch_str = "volume %s-%s\n"
-                "    type performance/stat-prefetch\n"
-                "    subvolumes %s\n"
-                "end-volume\n\n";
+        GF_ASSERT (vme);
+        GF_ASSERT (key);
 
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
-                goto out;
+
+        if (!strcmp (vme->key, AUTH_ALLOW_MAP_KEY))
+                *key = gf_strdup (AUTH_ALLOW_OPT_KEY);
+        else if (!strcmp (vme->key, AUTH_REJECT_MAP_KEY))
+                *key = gf_strdup (AUTH_REJECT_OPT_KEY);
+        else if (!strcmp (vme->key, NFS_DISABLE_MAP_KEY))
+                *key = gf_strdup (NFS_DISABLE_OPT_KEY);
+        else {
+                if (vme->option) {
+                        if  (vme->option[0] == '!') {
+                                *key = vme->option + 1;
+                                if (!*key[0])
+                                        ret = -1;
+                        } else {
+                                *key = vme->option;
+                        }
+                } else {
+                        *key = strchr (vme->key, '.');
+                        if (*key) {
+                                (*key) ++;
+                                if (!*key[0])
+                                        ret = -1;
+                        } else {
+                                ret = -1;
+                        }
+                }
         }
+        if (ret)
+                gf_log ("glusterd", GF_LOG_ERROR, "Wrong entry found in  "
+                        "glusterd_volopt_map entry %s", vme->key);
+        else
+                gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
 
-        fprintf (file, statprefetch_str,
-                 volname,
-                 "stat-prefetch",
-                 subvolume);
+        return ret;
+}
 
-        ret = 0;
+static void
+_free_xlator_opt_key (char *key)
+{
+        GF_ASSERT (key);
 
-out:
-        return ret;
+        if (!strcmp (key, AUTH_ALLOW_OPT_KEY) ||
+            !strcmp (key, AUTH_REJECT_OPT_KEY) ||
+            !strcmp (key, NFS_DISABLE_OPT_KEY))
+                GF_FREE (key);
+
+        return;
 }
 
-static int
-generate_server_volfile (glusterd_brickinfo_t *brickinfo,
-                         dict_t *dict,
-                         const char *filename)
+int
+glusterd_get_volopt_content (dict_t * ctx, gf_boolean_t xml_out)
 {
-        FILE *file          = NULL;
-        char  subvol[2048]  = {0,};
-        char *volname       = NULL;
-        int   ret           = -1;
-        int   activate_pump = 0;
+        void                    *dl_handle = NULL;
+        volume_opt_list_t        vol_opt_handle = {{0},};
+        char                    *key = NULL;
+        struct volopt_map_entry *vme = NULL;
+        int                      ret = -1;
+        char                    *def_val = NULL;
+        char                    *descr = NULL;
+        char                     output_string[25600] = {0, };
+        char                    *output = NULL;
+        char                     tmp_str[2048] = {0, };
+#if (HAVE_LIB_XML)
+        xmlTextWriterPtr         writer = NULL;
+        xmlBufferPtr             buf = NULL;
+
+        if (xml_out) {
+                ret = init_sethelp_xml_doc (&writer, &buf);
+                if (ret) /*logging done in init_xml_lib*/
+                        goto out;
+        }
+#endif
 
-        GF_ASSERT (filename);
+        INIT_LIST_HEAD (&vol_opt_handle.list);
 
-        file = fopen (filename, "w+");
-        if (!file) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not open file %s", filename);
-                ret = -1;
-                goto out;
+        for (vme = &glusterd_volopt_map[0]; vme->key; vme++) {
+
+                if ((vme->type == NO_DOC) || (vme->type == GLOBAL_NO_DOC))
+                        continue;
+
+                if (vme->description) {
+                        descr = vme->description;
+                        def_val = vme->value;
+                } else {
+                        if (_get_xlator_opt_key_from_vme (vme, &key)) {
+                                gf_log ("glusterd", GF_LOG_DEBUG, "Failed to "
+                                        "get %s key from volume option entry",
+                                        vme->key);
+                                goto out; /*Some error while geting key*/
+                        }
+
+                        ret = xlator_volopt_dynload (vme->voltype,
+                                                     &dl_handle,
+                                                     &vol_opt_handle);
+
+                        if (ret) {
+                                gf_log ("glusterd", GF_LOG_DEBUG,
+                                        "xlator_volopt_dynload error(%d)", ret);
+                                ret = 0;
+                                goto cont;
+                        }
+
+                        ret = xlator_option_info_list (&vol_opt_handle, key,
+                                                       &def_val, &descr);
+                        if (ret) { /*Swallow Error i.e if option not found*/
+                                gf_log ("glusterd", GF_LOG_DEBUG,
+                                        "Failed to get option for %s key", key);
+                                ret = 0;
+                                goto cont;
+                        }
+                }
+
+                if (xml_out) {
+#if (HAVE_LIB_XML)
+                        if (xml_add_volset_element (writer,vme->key,
+                                                    def_val, descr)) {
+                                ret = -1;
+                                goto cont;
+                        }
+#else
+                        gf_log ("glusterd", GF_LOG_ERROR, "Libxml not present");
+#endif
+                } else {
+                        snprintf (tmp_str, sizeof (tmp_str), "Option: %s\nDefault "
+                                        "Value: %s\nDescription: %s\n\n",
+                                        vme->key, def_val, descr);
+                        strcat (output_string, tmp_str);
+                }
+cont:
+                if (dl_handle) {
+                        dlclose (dl_handle);
+                        dl_handle = NULL;
+                        vol_opt_handle.given_opt = NULL;
+                }
+                if (key) {
+                        _free_xlator_opt_key (key);
+                        key = NULL;
+                }
+                if (ret)
+                        goto out;
         }
 
-        ret = dict_get_str (dict, "volname", &volname);
-        if (ret) {
+#if (HAVE_LIB_XML)
+        if ((xml_out) &&
+            (ret = end_sethelp_xml_doc (writer)))
                 goto out;
-        }
+#else
+        if (xml_out)
+                gf_log ("glusterd", GF_LOG_ERROR, "Libxml not present");
+#endif
 
-        /* Call functions in the same order
-           as you'd call if you were manually
-           writing a volfile top-down
-        */
+        if (!xml_out)
+                output = gf_strdup (output_string);
+        else
+#if (HAVE_LIB_XML)
+                output = gf_strdup ((char *)buf->content);
+#else
+                gf_log ("glusterd", GF_LOG_ERROR, "Libxml not present");
+#endif
 
-        ret = __write_posix_xlator (file, dict, brickinfo->path);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not write xlator");
+        if (NULL == output) {
+                ret = -1;
                 goto out;
         }
 
-        VOLGEN_GENERATE_VOLNAME (subvol, volname, "posix");
+        ret = dict_set_dynstr (ctx, "help-str", output);
+out:
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
 
-        ret = __write_access_control_xlator (file, dict, subvol);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not write xlator");
-                goto out;
-        }
+}
 
-        VOLGEN_GENERATE_VOLNAME (subvol, volname, "access-control");
+static int
+volgen_graph_build_clients (volgen_graph_t *graph, glusterd_volinfo_t *volinfo,
+                            dict_t *set_dict, void *param)
+{
+        int                      i                  = 0;
+        int                      ret                = -1;
+        uint32_t                 client_type        = GF_CLIENT_OTHER;
+        char                     transt[16]         = {0,};
+        char                    *volname            = NULL;
+        char                    *str                = NULL;
+        glusterd_brickinfo_t    *brick              = NULL;
+        xlator_t                *xl                 = NULL;
+        char                    *ssl_str            = NULL;
+        gf_boolean_t             ssl_bool;
 
-        ret = __write_locks_xlator (file, dict, subvol);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not write xlator");
+        volname = volinfo->volname;
+
+        if (volinfo->brick_count == 0) {
+                gf_log ("", GF_LOG_ERROR,
+                        "volume inconsistency: brick count is 0");
                 goto out;
         }
 
-        ret = dict_get_int32 (dict, "enable-pump", &activate_pump);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Pump is disabled");
+        if ((volinfo->dist_leaf_count < volinfo->brick_count) &&
+            ((volinfo->brick_count % volinfo->dist_leaf_count) != 0)) {
+                gf_log ("", GF_LOG_ERROR,
+                        "volume inconsistency: "
+                        "total number of bricks (%d) is not divisible with "
+                        "number of bricks per cluster (%d) in a multi-cluster "
+                        "setup",
+                        volinfo->brick_count, volinfo->dist_leaf_count);
+                goto out;
         }
 
-        if (activate_pump) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Pump is enabled");
+        get_transport_type (volinfo, set_dict, transt, _gf_false);
 
-                VOLGEN_GENERATE_VOLNAME (subvol, volname, "locks");
+        if (!strcmp (transt, "tcp,rdma"))
+                strcpy (transt, "tcp");
 
-                ret = __write_replace_brick_xlator (file, dict);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Could not write xlator");
+        i = 0;
+        ret = -1;
+        list_for_each_entry (brick, &volinfo->bricks, brick_list) {
+                ret = -1;
+                xl = volgen_graph_add_nolink (graph, "protocol/client",
+                                              "%s-client-%d", volname, i);
+                if (!xl)
+                        goto out;
+                ret = xlator_set_option (xl, "remote-host", brick->hostname);
+                if (ret)
+                        goto out;
+                ret = xlator_set_option (xl, "remote-subvolume", brick->path);
+                if (ret)
                         goto out;
+                ret = xlator_set_option (xl, "transport-type", transt);
+                if (ret)
+                        goto out;
+
+                ret = dict_get_uint32 (set_dict, "trusted-client",
+                                       &client_type);
+
+                if (!ret && client_type == GF_CLIENT_TRUSTED) {
+                        str = NULL;
+                        str = glusterd_auth_get_username (volinfo);
+                        if (str) {
+                                ret = xlator_set_option (xl, "username",
+                                                         str);
+                                if (ret)
+                                        goto out;
+                        }
+
+                        str = glusterd_auth_get_password (volinfo);
+                        if (str) {
+                                ret = xlator_set_option (xl, "password",
+                                                         str);
+                                if (ret)
+                                        goto out;
+                        }
                 }
 
-                ret = __write_pump_xlator (file, dict, subvol);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Could not write xlator");
-                        goto out;
+                if (dict_get_str(set_dict,"client.ssl",&ssl_str) == 0) {
+                        if (gf_string2boolean(ssl_str,&ssl_bool) == 0) {
+                                if (ssl_bool) {
+                                        ret = xlator_set_option(xl,
+                                                "transport.socket.ssl-enabled",
+                                                "true");
+                                        if (ret) {
+                                                goto out;
+                                        }
+                                }
+                        }
                 }
 
-                VOLGEN_GENERATE_VOLNAME (subvol, volname, "pump");
-        } else
-                VOLGEN_GENERATE_VOLNAME (subvol, volname, "locks");
+                i++;
+        }
 
+        if (i != volinfo->brick_count) {
+                gf_log ("", GF_LOG_ERROR,
+                        "volume inconsistency: actual number of bricks (%d) "
+                        "differs from brick count (%d)", i,
+                        volinfo->brick_count);
 
-        ret = dict_set_str (dict, "export-path", brickinfo->path);
-        if (ret) {
+                ret = -1;
                 goto out;
         }
+        ret = 0;
+out:
+        return ret;
+}
 
-        ret = __write_iothreads_xlator (file, dict, subvol);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not write xlator");
+static int
+volgen_graph_build_clusters (volgen_graph_t *graph,
+                             glusterd_volinfo_t *volinfo, char *xl_type,
+                             char *xl_namefmt, size_t child_count,
+                             size_t sub_count)
+{
+        int             i = 0;
+        int             j = 0;
+        xlator_t        *txl = NULL;
+        xlator_t        *xl  = NULL;
+        xlator_t        *trav = NULL;
+        char            *volname = NULL;
+        int             ret     = -1;
+
+        if (child_count == 0)
                 goto out;
+        volname = volinfo->volname;
+        txl = first_of (graph);
+        for (trav = txl; --child_count; trav = trav->next);
+        for (;; trav = trav->prev) {
+                if ((i % sub_count) == 0) {
+                        xl = volgen_graph_add_nolink (graph, xl_type,
+                                                      xl_namefmt, volname, j);
+                        if (!xl) {
+                                ret = -1;
+                                goto out;
+                        }
+                        j++;
+                }
+
+                ret = volgen_xlator_link (xl, trav);
+                if (ret)
+                        goto out;
+
+                if (trav == txl)
+                        break;
+
+                i++;
         }
 
-        ret = __write_server_xlator (file, dict, brickinfo->path);
+        ret = j;
+out:
+        return ret;
+}
+
+gf_boolean_t
+_xl_is_client_decommissioned (xlator_t *xl, glusterd_volinfo_t *volinfo)
+{
+        int             ret = 0;
+        gf_boolean_t    decommissioned = _gf_false;
+        char            *hostname = NULL;
+        char            *path = NULL;
+
+        GF_ASSERT (!strcmp (xl->type, "protocol/client"));
+        ret = xlator_get_option (xl, "remote-host", &hostname);
         if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not write xlator");
+                GF_ASSERT (0);
+                gf_log ("glusterd", GF_LOG_ERROR, "Failed to get remote-host "
+                        "from client %s", xl->name);
+                goto out;
+        }
+        ret = xlator_get_option (xl, "remote-subvolume", &path);
+        if (ret) {
+                GF_ASSERT (0);
+                gf_log ("glusterd", GF_LOG_ERROR, "Failed to get remote-host "
+                        "from client %s", xl->name);
                 goto out;
         }
 
-        fclose (file);
-        file = NULL;
-
+        decommissioned = glusterd_is_brick_decommissioned (volinfo, hostname,
+                                                           path);
 out:
-        return ret;
+        return decommissioned;
 }
 
-static int
-__write_perf_xlator (char *xlator_name, FILE *file,
-                     dict_t *dict, char *subvol)
+gf_boolean_t
+_xl_has_decommissioned_clients (xlator_t *xl, glusterd_volinfo_t *volinfo)
 {
-        int ret = 0;
+        xlator_list_t   *xl_child = NULL;
+        gf_boolean_t    decommissioned = _gf_false;
+        xlator_t        *cxl = NULL;
 
-        if (strcmp (xlator_name, "write-behind") == 0) {
+        if (!xl)
+                goto out;
 
-                ret = __write_wb_xlator (file, dict, subvol);
+        if (!strcmp (xl->type, "protocol/client")) {
+                decommissioned = _xl_is_client_decommissioned (xl, volinfo);
+                goto out;
+        }
 
-        } else if (strcmp (xlator_name, "read-ahead") == 0) {
+        xl_child = xl->children;
+        while (xl_child) {
+                cxl = xl_child->xlator;
+                /* this can go into 2 depths if the volume type
+                   is stripe-replicate */
+                decommissioned = _xl_has_decommissioned_clients (cxl, volinfo);
+                if (decommissioned)
+                        break;
 
-                ret = __write_ra_xlator (file, dict, subvol);
+                xl_child = xl_child->next;
+        }
+out:
+        return decommissioned;
+}
 
-        } else if (strcmp (xlator_name, "io-cache") == 0) {
+static int
+_graph_get_decommissioned_children (xlator_t *dht, glusterd_volinfo_t *volinfo,
+                                    char **children)
+{
+        int             ret = -1;
+        xlator_list_t   *xl_child = NULL;
+        xlator_t        *cxl = NULL;
+        gf_boolean_t    comma = _gf_false;
+
+        *children = NULL;
+        xl_child = dht->children;
+        while (xl_child) {
+                cxl = xl_child->xlator;
+                if (_xl_has_decommissioned_clients (cxl, volinfo)) {
+                        if (!*children) {
+                                *children = GF_CALLOC (16 * GF_UNIT_KB, 1,
+                                                       gf_common_mt_char);
+                                if (!*children)
+                                        goto out;
+                        }
 
-                ret = __write_iocache_xlator (file, dict, subvol);
+                        if (comma)
+                                strcat (*children, ",");
+                        strcat (*children, cxl->name);
+                        comma = _gf_true;
+                }
 
-        } else if (strcmp (xlator_name, "quick-read") == 0) {
+                xl_child = xl_child->next;
+        }
+        ret = 0;
+out:
+        return ret;
+}
 
-                ret = __write_qr_xlator (file, dict, subvol);
+static int
+volgen_graph_build_dht_cluster (volgen_graph_t *graph,
+                                glusterd_volinfo_t *volinfo, size_t child_count)
+{
+        int32_t                 clusters                 = 0;
+        int                     ret                      = -1;
+        char                    *decommissioned_children = NULL;
+        xlator_t                *dht                     = NULL;
+        char                    *voltype                 = "cluster/distribute";
+
+        /* NUFA and Switch section */
+        if (dict_get_str_boolean (volinfo->dict, "cluster.nufa", 0) &&
+            dict_get_str_boolean (volinfo->dict, "cluster.switch", 0)) {
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "nufa and switch cannot be set together");
+                ret = -1;
+                goto out;
+        }
 
-        } else if (strcmp (xlator_name, "stat-prefetch") == 0) {
+        /* Check for NUFA volume option, and change the voltype */
+        if (dict_get_str_boolean (volinfo->dict, "cluster.nufa", 0))
+                voltype = "cluster/nufa";
 
-                ret = __write_statprefetch_xlator (file, dict, subvol);
+        /* Check for switch volume option, and change the voltype */
+        if (dict_get_str_boolean (volinfo->dict, "cluster.switch", 0))
+                voltype = "cluster/switch";
 
-        }
+        clusters = volgen_graph_build_clusters (graph,  volinfo,
+                                                voltype,
+                                                "%s-dht",
+                                                child_count,
+                                                child_count);
+        if (clusters < 0)
+                goto out;
 
+        dht = first_of (graph);
+        ret = _graph_get_decommissioned_children (dht, volinfo,
+                                                  &decommissioned_children);
+        if (ret)
+                goto out;
+        if (decommissioned_children) {
+                ret = xlator_set_option (dht, "decommissioned-bricks",
+                                         decommissioned_children);
+                if (ret)
+                        goto out;
+        }
+        ret = 0;
+out:
+        GF_FREE (decommissioned_children);
         return ret;
-
 }
 
-static gf_boolean_t
-is_perf_xlator_enabled (char *status)
+static int
+volume_volgen_graph_build_clusters (volgen_graph_t *graph,
+                                    glusterd_volinfo_t *volinfo)
 {
-        gf_boolean_t ret = _gf_false;
+        char                    *replicate_args[]   = {"cluster/replicate",
+                                                       "%s-replicate-%d"};
+        char                    *stripe_args[]      = {"cluster/stripe",
+                                                       "%s-stripe-%d"};
+        int                     rclusters           = 0;
+        int                     clusters            = 0;
+        int                     dist_count          = 0;
+        int                     ret                 = -1;
+
+        if (!volinfo->dist_leaf_count)
+                goto out;
+
+        if (volinfo->dist_leaf_count == 1)
+                goto build_distribute;
+
+        /* All other cases, it will have one or the other cluster type */
+        switch (volinfo->type) {
+        case GF_CLUSTER_TYPE_REPLICATE:
+                clusters = volgen_graph_build_clusters (graph, volinfo,
+                                                        replicate_args[0],
+                                                        replicate_args[1],
+                                                        volinfo->brick_count,
+                                                        volinfo->replica_count);
+                if (clusters < 0)
+                        goto out;
+                break;
+        case GF_CLUSTER_TYPE_STRIPE:
+                clusters = volgen_graph_build_clusters (graph, volinfo,
+                                                        stripe_args[0],
+                                                        stripe_args[1],
+                                                        volinfo->brick_count,
+                                                        volinfo->stripe_count);
+                if (clusters < 0)
+                        goto out;
+                break;
+        case GF_CLUSTER_TYPE_STRIPE_REPLICATE:
+                /* Replicate after the clients, then stripe */
+                if (volinfo->replica_count == 0)
+                        goto out;
+                clusters = volgen_graph_build_clusters (graph, volinfo,
+                                                        replicate_args[0],
+                                                        replicate_args[1],
+                                                        volinfo->brick_count,
+                                                        volinfo->replica_count);
+                if (clusters < 0)
+                        goto out;
 
-        if (strcmp (status, "enable") == 0)
-                ret = _gf_true;
+                rclusters = volinfo->brick_count / volinfo->replica_count;
+                GF_ASSERT (rclusters == clusters);
+                clusters = volgen_graph_build_clusters (graph, volinfo,
+                                                        stripe_args[0],
+                                                        stripe_args[1],
+                                                        rclusters,
+                                                        volinfo->stripe_count);
+                if (clusters < 0)
+                        goto out;
+                break;
+        default:
+                gf_log ("", GF_LOG_ERROR, "volume inconsistency: "
+                        "unrecognized clustering type");
+                goto out;
+        }
+
+build_distribute:
+        dist_count = volinfo->brick_count / volinfo->dist_leaf_count;
+        if (!dist_count) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = volgen_graph_build_dht_cluster (graph, volinfo,
+                                              dist_count);
+        if (ret == -1)
+                goto out;
 
+        ret = 0;
+out:
         return ret;
 }
 
+static int client_graph_set_perf_options(volgen_graph_t *graph,
+					 glusterd_volinfo_t *volinfo,
+					 dict_t *set_dict)
+{
+	data_t *tmp_data = NULL;
+	char *volname = NULL;
+
+	/*
+	 * Logic to make sure NFS doesn't have performance translators by
+	 * default for a volume
+	 */
+	volname = volinfo->volname;
+	tmp_data = dict_get (set_dict, "nfs-volume-file");
+	if (!tmp_data)
+		return volgen_graph_set_options_generic(graph, set_dict,
+							volname,
+							&perfxl_option_handler);
+	else
+		return volgen_graph_set_options_generic(graph, set_dict,
+							volname,
+							&nfsperfxl_option_handler);
+}
+
 static int
-generate_perf_xlator_list (dict_t *dict, char *perf_xlator_list[])
+client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo,
+                      dict_t *set_dict, void *param)
 {
-        char *status = NULL;
-        int i        = 0;
-        int dict_ret = 0;
-        int ret      = 0;
+        int       ret      = 0;
+        xlator_t *xl       = NULL;
+        char     *volname  = NULL;
 
-        GF_ASSERT (dict);
+        volname = volinfo->volname;
+        ret = volgen_graph_build_clients (graph, volinfo, set_dict, param);
+        if (ret)
+                goto out;
 
-        dict_ret = dict_get_str (dict, "write-behind", &status);
-        if (dict_ret || is_perf_xlator_enabled (status)) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Write behind is enabled");
-                perf_xlator_list[i] = gf_strdup ("write-behind");
-                if (!perf_xlator_list[i]) {
-                        gf_log ("", GF_LOG_ERROR,
-                                "Out of memory");
+        ret = volume_volgen_graph_build_clusters (graph, volinfo);
+        if (ret == -1)
+                goto out;
+
+        /* Check for compress volume option, and add it to the graph on client side */
+        if (dict_get_str_boolean (set_dict, "features.compress", 0)) {
+                xl = volgen_graph_add (graph, "features/cdc", volname);
+                if (!xl) {
                         ret = -1;
                         goto out;
                 }
+                ret = dict_set_str (set_dict, "compress.mode", "client");
+                if (ret)
+                        goto out;
 
-                i++;
-
-        } else {
-                gf_log ("", GF_LOG_DEBUG,
-                        "write-behind option is disabled");
         }
 
+        ret = glusterd_volinfo_get_boolean (volinfo, "features.encryption");
+        if (ret == -1)
+                goto out;
+        if (ret) {
+                xl = volgen_graph_add (graph, "encryption/crypt", volname);
 
-        dict_ret = dict_get_str (dict, "read-ahead", &status);
-        if (dict_ret || is_perf_xlator_enabled (status)) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "read-ahead is enabled");
-                perf_xlator_list[i] = gf_strdup ("read-ahead");
-                if (!perf_xlator_list[i]) {
-                        gf_log ("", GF_LOG_ERROR,
-                                "Out of memory");
+                if (!xl) {
                         ret = -1;
                         goto out;
                 }
-
-                i++;
-
-        } else {
-                gf_log ("", GF_LOG_DEBUG,
-                        "read-ahead option is disabled");
         }
 
-        dict_ret = dict_get_str (dict, "io-cache", &status);
-        if (dict_ret || is_perf_xlator_enabled (status)) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "io-cache is enabled");
-                perf_xlator_list[i] = gf_strdup ("io-cache");
-                if (!perf_xlator_list[i]) {
-                        gf_log ("", GF_LOG_ERROR,
-                                "Out of memory");
+        ret = glusterd_volinfo_get_boolean (volinfo, VKEY_FEATURES_QUOTA);
+        if (ret == -1)
+                goto out;
+        if (ret) {
+                xl = volgen_graph_add (graph, "features/quota", volname);
+
+                if (!xl) {
                         ret = -1;
                         goto out;
                 }
+        }
 
-                i++;
 
-        } else {
-                gf_log ("", GF_LOG_DEBUG,
-                        "io-cache option is disabled");
-        }
+        ret = glusterd_volinfo_get_boolean (volinfo, "features.file-snapshot");
+        if (ret == -1)
+                goto out;
+        if (ret) {
+                xl = volgen_graph_add (graph, "features/qemu-block", volname);
 
-        dict_ret = dict_get_str (dict, "quick-read", &status);
-        if (dict_ret || is_perf_xlator_enabled (status)) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "quick-read is enabled");
-                perf_xlator_list[i] = gf_strdup ("quick-read");
-                if (!perf_xlator_list[i]) {
-                        gf_log ("", GF_LOG_ERROR,
-                                "Out of memory");
+                if (!xl) {
                         ret = -1;
                         goto out;
                 }
+        }
 
-                i++;
+	ret = client_graph_set_perf_options(graph, volinfo, set_dict);
+        if (ret)
+                goto out;
 
-        } else {
-                gf_log ("", GF_LOG_DEBUG,
-                        "quick-read option is disabled");
-        }
+        /* add debug translators depending on the options */
+        ret = check_and_add_debug_xl (graph, set_dict, volname,
+                                      "client");
+        if (ret)
+                return -1;
 
-        dict_ret = dict_get_str (dict, "stat-prefetch", &status);
-        if (dict_ret || is_perf_xlator_enabled (status)) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "stat-prefetch is enabled");
-                perf_xlator_list[i] = gf_strdup ("stat-prefetch");
-                if (!perf_xlator_list[i]) {
-                        gf_log ("", GF_LOG_ERROR,
-                                "Out of memory");
-                        ret = -1;
-                        goto out;
-                }
+        ret = -1;
+        xl = volgen_graph_add_as (graph, "debug/io-stats", volname);
+        if (!xl)
+                goto out;
 
-                i++;
+        ret = volgen_graph_set_options_generic (graph, set_dict, "client",
+                                                &loglevel_option_handler);
 
-        } else {
-                gf_log ("", GF_LOG_DEBUG,
-                        "stat-prefetch option is disabled");
-        }
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING, "changing client log level"
+                        " failed");
 
+        ret = volgen_graph_set_options_generic (graph, set_dict, "client",
+                                                &sys_loglevel_option_handler);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING, "changing client syslog "
+                        "level failed");
 out:
         return ret;
+}
+
 
+/* builds a graph for client role , with option overrides in mod_dict */
+static int
+build_client_graph (volgen_graph_t *graph, glusterd_volinfo_t *volinfo,
+                    dict_t *mod_dict)
+{
+        return build_graph_generic (graph, volinfo, mod_dict, NULL,
+                                    &client_graph_builder);
+}
+
+char *gd_shd_options[] = {
+        "!self-heal-daemon",
+        "!heal-timeout",
+        NULL
+};
+
+char*
+gd_get_matching_option (char **options, char *option)
+{
+        while (*options && strcmp (*options, option))
+                options++;
+        return *options;
 }
 
 static int
-destroy_perf_xlator_list (char *perf_xlator_list[])
+shd_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme,
+                    void *param)
 {
-        int i = 0;
+        int                     ret = 0;
+        struct volopt_map_entry new_vme = {0};
+        char                    *shd_option = NULL;
 
-        while (perf_xlator_list[i]) {
-                GF_FREE (perf_xlator_list[i]);
-                i++;
+        shd_option = gd_get_matching_option (gd_shd_options, vme->option);
+        if ((vme->option[0] == '!') && !shd_option)
+                goto out;
+        new_vme = *vme;
+        if (shd_option) {
+                new_vme.option = shd_option + 1;//option with out '!'
         }
 
-        return 0;
+        ret = no_filter_option_handler (graph, &new_vme, param);
+out:
+        return ret;
 }
 
 static int
-write_perf_xlators (glusterd_volinfo_t *volinfo, FILE *file,
-                    int32_t dist_count, int32_t replicate_count,
-                    int32_t stripe_count)
-{
-        char *perf_xlator_list[256]     = {0,};
-        char  subvol[2048]              = {0,};
-        int    i                        = 0;
-        int  ret                        = 0;
-        char  *volname                  = NULL;
-        dict_t *dict                    = NULL;
-
-        dict    = volinfo->dict;
-        volname = volinfo->volname;
+nfs_option_handler (volgen_graph_t *graph,
+                    struct volopt_map_entry *vme, void *param)
+{
+        xlator_t *xl = NULL;
+        char *aa = NULL;
+        int   ret   = 0;
+        glusterd_volinfo_t *volinfo = NULL;
 
-        ret = generate_perf_xlator_list (
-                dict, perf_xlator_list);
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not generate perf xlator list");
-                goto out;
+        volinfo = param;
+
+        xl = first_of (graph);
+
+/*        if (vme->type  == GLOBAL_DOC || vme->type == GLOBAL_NO_DOC) {
+
+                ret = xlator_set_option (xl, vme->key, vme->value);
+        }*/
+        if (!volinfo || (volinfo->volname[0] == '\0'))
+                return 0;
+
+        if (! strcmp (vme->option, "!rpc-auth.addr.*.allow")) {
+                ret = gf_asprintf (&aa, "rpc-auth.addr.%s.allow",
+                                        volinfo->volname);
+
+                if (ret != -1) {
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
+                }
+
+                if (ret)
+                        return -1;
         }
 
-        if (dist_count > 1) {
-                VOLGEN_GENERATE_VOLNAME (subvol, volname, "dht");
-                ret = __write_perf_xlator (perf_xlator_list[i], file,
-                                           dict, subvol);
-                i++;
+        if (! strcmp (vme->option, "!rpc-auth.addr.*.reject")) {
+                ret = gf_asprintf (&aa, "rpc-auth.addr.%s.reject",
+                                        volinfo->volname);
+
+                if (ret != -1) {
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
+                }
+
+                if (ret)
+                        return -1;
         }
-        else if (replicate_count > 1) {
-                VOLGEN_GENERATE_VOLNAME (subvol, volname, "replicate-0");
-                ret = __write_perf_xlator (perf_xlator_list[i], file,
-                                           dict, subvol);
-                i++;
 
+        if (! strcmp (vme->option, "!rpc-auth.auth-unix.*")) {
+                ret = gf_asprintf (&aa, "rpc-auth.auth-unix.%s",
+                                        volinfo->volname);
+
+                if (ret != -1) {
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
+                }
+
+                if (ret)
+                        return -1;
         }
-        else if (stripe_count > 1) {
-                VOLGEN_GENERATE_VOLNAME (subvol, volname, "stripe-0");
-                ret = __write_perf_xlator (perf_xlator_list[i], file,
-                                           dict, subvol);
-                i++;
+        if (! strcmp (vme->option, "!rpc-auth.auth-null.*")) {
+                ret = gf_asprintf (&aa, "rpc-auth.auth-null.%s",
+                                        volinfo->volname);
 
+                if (ret != -1) {
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
+                }
+
+                if (ret)
+                        return -1;
         }
-        else {
-                VOLGEN_GENERATE_VOLNAME (subvol, volname, "client-0");
-                ret = __write_perf_xlator (perf_xlator_list[i], file,
-                                           dict, subvol);
-                i++;
 
+        if (! strcmp (vme->option, "!nfs3.*.trusted-sync")) {
+                ret = gf_asprintf (&aa, "nfs3.%s.trusted-sync",
+                                        volinfo->volname);
+
+                if (ret != -1) {
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
+                }
+
+                if (ret)
+                        return -1;
         }
-        if (ret) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not write xlator");
-                goto out;
+
+        if (! strcmp (vme->option, "!nfs3.*.trusted-write")) {
+                ret = gf_asprintf (&aa, "nfs3.%s.trusted-write",
+                                        volinfo->volname);
+
+                if (ret != -1) {
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
+                }
+
+                if (ret)
+                        return -1;
         }
 
-        while (perf_xlator_list[i]) {
-                VOLGEN_GENERATE_VOLNAME (subvol, volname, perf_xlator_list[i-1]);
-                ret = __write_perf_xlator (perf_xlator_list[i], file,
-                                           dict, subvol);
-                i++;
+        if (! strcmp (vme->option, "!nfs3.*.volume-access")) {
+                ret = gf_asprintf (&aa, "nfs3.%s.volume-access",
+                                        volinfo->volname);
 
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Could not write xlator");
-                        goto out;
+                if (ret != -1) {
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
                 }
+
+                if (ret)
+                        return -1;
         }
 
+        if (! strcmp (vme->option, "!nfs3.*.export-dir")) {
+                ret = gf_asprintf (&aa, "nfs3.%s.export-dir",
+                                        volinfo->volname);
 
-out:
-        destroy_perf_xlator_list (perf_xlator_list);
-        return ret;
+                if (ret != -1) {
+                        ret = gf_canonicalize_path (vme->value);
+                        if (ret)
+                                return -1;
 
-}
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
+                }
 
-static int
-generate_client_volfile (glusterd_volinfo_t *volinfo, char *filename)
-{
-        FILE    *file               = NULL;
-        dict_t  *dict               = NULL;
-        char    *volname            = NULL;
-        glusterd_brickinfo_t *brick = NULL;
-        char     subvol[2048]       = {0,};
-        int32_t  replicate_count    = 0;
-        int32_t  stripe_count       = 0;
-        int32_t  dist_count         = 0;
-        int32_t  num_bricks         = 0;
-        int      subvol_count       = 0;
-        int      count              = 0;
-        int      i                  = 0;
-        int      ret                = -1;
-
-        GF_ASSERT (filename);
+                if (ret)
+                        return -1;
+        }
 
-        volname = volinfo->volname;
-        dict    = volinfo->dict;
-
-        list_for_each_entry (brick, &volinfo->bricks, brick_list)
-                num_bricks++;
-
-        if (GF_CLUSTER_TYPE_REPLICATE == volinfo->type) {
-                if (volinfo->brick_count <= volinfo->sub_count) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Volfile is plain replicated");
-                        replicate_count = volinfo->sub_count;
-                        dist_count = num_bricks / replicate_count;
-                        if (!dist_count) {
-                                replicate_count = num_bricks;
-                                dist_count = num_bricks / replicate_count;
-                        }
-                } else {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Volfile is distributed-replicated");
-                        replicate_count = volinfo->sub_count;
-                        dist_count = num_bricks / replicate_count;
+
+
+        if (! strcmp (vme->option, "!rpc-auth.ports.*.insecure")) {
+                ret = gf_asprintf (&aa, "rpc-auth.ports.%s.insecure",
+                                        volinfo->volname);
+
+                if (ret != -1) {
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
                 }
 
-        } else if (GF_CLUSTER_TYPE_STRIPE == volinfo->type) {
-                if (volinfo->brick_count == volinfo->sub_count) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Volfile is plain striped");
-                        stripe_count = volinfo->sub_count;
-                        dist_count = num_bricks / stripe_count;
-                } else {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Volfile is distributed-striped");
-                        stripe_count = volinfo->sub_count;
-                        dist_count = num_bricks / stripe_count;
+                if (ret)
+                        return -1;
+        }
+
+
+        if (! strcmp (vme->option, "!nfs-disable")) {
+                ret = gf_asprintf (&aa, "nfs.%s.disable",
+                                        volinfo->volname);
+
+                if (ret != -1) {
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
                 }
-        } else {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Volfile is plain distributed");
-                dist_count = num_bricks;
+
+                if (ret)
+                        return -1;
         }
 
+        if ( (strcmp (vme->voltype, "nfs/server") == 0) &&
+             (vme->option && vme->option[0]!='!') ) {
+               ret = xlator_set_option (xl, vme->option, vme->value);
+                if (ret)
+                        return -1;
+        }
 
-        file = fopen (filename, "w+");
-        if (!file) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Could not open file %s", filename);
+
+        /*key = strchr (vme->key, '.') + 1;
+
+        for (trav = xl->children; trav; trav = trav->next) {
+                ret = gf_asprintf (&aa, "auth.addr.%s.%s", trav->xlator->name,
+                                   key);
+                if (ret != -1) {
+                        ret = xlator_set_option (xl, aa, vme->value);
+                        GF_FREE (aa);
+                }
+                if (ret)
+                        return -1;
+        }*/
+
+        return 0;
+}
+
+static int
+volgen_graph_set_iam_shd (volgen_graph_t *graph)
+{
+        xlator_t        *trav;
+        int             ret = 0;
+
+        for (trav = first_of (graph); trav; trav = trav->next) {
+                if (strcmp (trav->type, "cluster/replicate") != 0)
+                        continue;
+
+                ret = xlator_set_option (trav, "iam-self-heal-daemon", "yes");
+                if (ret)
+                        break;
+        }
+        return ret;
+}
+
+static int
+build_shd_graph (volgen_graph_t *graph, dict_t *mod_dict)
+{
+        volgen_graph_t     cgraph         = {0};
+        glusterd_volinfo_t *voliter       = NULL;
+        xlator_t           *this          = NULL;
+        glusterd_conf_t    *priv          = NULL;
+        dict_t             *set_dict      = NULL;
+        int                ret            = 0;
+        gf_boolean_t       valid_config   = _gf_false;
+        xlator_t           *iostxl        = NULL;
+        int                rclusters      = 0;
+        int                replica_count  = 0;
+        gf_boolean_t       graph_check    = _gf_false;
+
+        this = THIS;
+        priv = this->private;
+
+        set_dict = dict_new ();
+        if (!set_dict) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
+        graph_check = dict_get_str_boolean (mod_dict, "graph-check", 0);
+        iostxl = volgen_graph_add_as (graph, "debug/io-stats", "glustershd");
+        if (!iostxl) {
                 ret = -1;
                 goto out;
         }
 
-        /* Call functions in the same order
-           as you'd call if you were manually
-           writing a volfile top-down
-        */
+        list_for_each_entry (voliter, &priv->volumes, vol_list) {
+                if (!graph_check &&
+                   (voliter->status != GLUSTERD_STATUS_STARTED))
+                        continue;
 
-        count = 0;
+                if (!glusterd_is_volume_replicate (voliter))
+                        continue;
 
-        list_for_each_entry (brick, &volinfo->bricks, brick_list) {
+                replica_count = voliter->replica_count;
 
-                ret = __write_client_xlator (file, dict, brick->path,
-                                             brick->hostname, count);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Could not write xlator");
+                valid_config = _gf_true;
+
+                ret = dict_set_str (set_dict, "cluster.self-heal-daemon", "on");
+                if (ret)
+                        goto out;
+
+                ret = dict_set_uint32 (set_dict, "trusted-client",
+                                       GF_CLIENT_TRUSTED);
+                if (ret)
+                        goto out;
+
+                dict_copy (voliter->dict, set_dict);
+                if (mod_dict)
+                        dict_copy (mod_dict, set_dict);
+
+                memset (&cgraph, 0, sizeof (cgraph));
+                ret = volgen_graph_build_clients (&cgraph, voliter, set_dict,
+                                                  NULL);
+                if (ret)
+                        goto out;
+
+                rclusters = volgen_graph_build_clusters (&cgraph, voliter,
+                                                        "cluster/replicate",
+                                                        "%s-replicate-%d",
+                                                        voliter->brick_count,
+                                                        replica_count);
+                if (rclusters < 0) {
+                        ret = -1;
                         goto out;
                 }
 
-                count++;
+                ret = volgen_graph_set_options_generic (&cgraph, set_dict, voliter,
+                                                        shd_option_handler);
+                if (ret)
+                        goto out;
+
+                ret = volgen_graph_set_iam_shd (&cgraph);
+                if (ret)
+                        goto out;
+
+                ret = volgen_graph_merge_sub (graph, &cgraph, rclusters);
+                if (ret)
+                        goto out;
+
+                ret = volgen_graph_set_options_generic (graph, set_dict,
+                                                        "client",
+                                                 &loglevel_option_handler);
+
+                if (ret)
+                        gf_log (THIS->name, GF_LOG_WARNING, "changing loglevel "
+                                "of self-heal daemon failed");
+
+                ret = volgen_graph_set_options_generic (graph, set_dict,
+                                                        "client",
+                                                 &sys_loglevel_option_handler);
+                if (ret)
+                        gf_log (THIS->name, GF_LOG_WARNING, "changing syslog "
+                                "level of self-heal daemon failed");
+
+                ret = dict_reset (set_dict);
+                if (ret)
+                        goto out;
+        }
+out:
+        if (set_dict)
+                dict_unref (set_dict);
+        if (!valid_config)
+                ret = -EINVAL;
+        return ret;
+}
+
+/* builds a graph for nfs server role, with option overrides in mod_dict */
+static int
+build_nfs_graph (volgen_graph_t *graph, dict_t *mod_dict)
+{
+        volgen_graph_t      cgraph        = {0,};
+        glusterd_volinfo_t *voliter       = NULL;
+        xlator_t           *this          = NULL;
+        glusterd_conf_t    *priv          = NULL;
+        dict_t             *set_dict      = NULL;
+        xlator_t           *nfsxl         = NULL;
+        char               *skey          = NULL;
+        int                 ret           = 0;
+        char               nfs_xprt[16]   = {0,};
+        char               *volname       = NULL;
+        data_t             *data          = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        set_dict = dict_new ();
+        if (!set_dict) {
+                gf_log ("", GF_LOG_ERROR, "Out of memory");
+                return -1;
         }
 
-        if (stripe_count && replicate_count) {
-                gf_log ("", GF_LOG_DEBUG,
-                        "Striped Replicate config not allowed");
+        nfsxl = volgen_graph_add_as (graph, "nfs/server", "nfs-server");
+        if (!nfsxl) {
                 ret = -1;
                 goto out;
         }
+        ret = xlator_set_option (nfsxl, "nfs.dynamic-volumes", "on");
+        if (ret)
+                goto out;
 
-        if (replicate_count > 1) {
-                subvol_count = 0;
-                for (i = 0; i < dist_count; i++) {
+        ret = xlator_set_option (nfsxl, "nfs.nlm", "on");
+        if (ret)
+                goto out;
 
-                        VOLGEN_GENERATE_VOLNAME (subvol, volname, "client");
+        ret = xlator_set_option (nfsxl, "nfs.drc", "on");
+        if (ret)
+                goto out;
 
-                        ret = __write_replicate_xlator (file, dict, subvol,
-                                                        replicate_count,
-                                                        subvol_count,
-                                                        i);
-                        if (ret) {
-                                gf_log ("", GF_LOG_DEBUG,
-                                        "Count not write xlator");
-                                goto out;
-                        }
+        list_for_each_entry (voliter, &priv->volumes, vol_list) {
+                if (voliter->status != GLUSTERD_STATUS_STARTED)
+                        continue;
 
-                        subvol_count += replicate_count;
+                if (dict_get_str_boolean (voliter->dict, "nfs.disable", 0))
+                        continue;
 
+                ret = gf_asprintf (&skey, "rpc-auth.addr.%s.allow",
+                                   voliter->volname);
+                if (ret == -1) {
+                        gf_log ("", GF_LOG_ERROR, "Out of memory");
+                        goto out;
                 }
-        }
+                ret = xlator_set_option (nfsxl, skey, "*");
+                GF_FREE (skey);
+                if (ret)
+                        goto out;
 
-        if (stripe_count > 1) {
-                subvol_count = 0;
-                for (i = 0; i < dist_count; i++) {
+                ret = gf_asprintf (&skey, "nfs3.%s.volume-id",
+                                   voliter->volname);
+                if (ret == -1) {
+                        gf_log ("", GF_LOG_ERROR, "Out of memory");
+                        goto out;
+                }
+                ret = xlator_set_option (nfsxl, skey, uuid_utoa (voliter->volume_id));
+                GF_FREE (skey);
+                if (ret)
+                        goto out;
 
-                        VOLGEN_GENERATE_VOLNAME (subvol, volname, "client");
+                /* If both RDMA and TCP are the transport_type, use RDMA
+                   for NFS client protocols */
+                memset (&cgraph, 0, sizeof (cgraph));
+                if (mod_dict)
+                        get_transport_type (voliter, mod_dict, nfs_xprt, _gf_true);
+                else
+                        get_transport_type (voliter, voliter->dict, nfs_xprt, _gf_true);
 
-                        ret = __write_stripe_xlator (file, dict, subvol,
-                                                     stripe_count,
-                                                     subvol_count,
-                                                     i);
-                        if (ret) {
-                                gf_log ("", GF_LOG_DEBUG,
-                                        "Count not write xlator");
-                                goto out;
-                        }
+                ret = dict_set_str (set_dict, "performance.stat-prefetch", "off");
+                if (ret)
+                        goto out;
 
-                        subvol_count += stripe_count;
+                ret = dict_set_str (set_dict, "performance.client-io-threads",
+                                    "off");
+                if (ret)
+                        goto out;
+
+                ret = dict_set_str (set_dict, "client-transport-type",
+                                    nfs_xprt);
+                if (ret)
+                        goto out;
+
+                ret = dict_set_uint32 (set_dict, "trusted-client",
+                                       GF_CLIENT_TRUSTED);
+                if (ret)
+                        goto out;
+
+                ret = dict_set_str (set_dict, "nfs-volume-file", "yes");
+                if (ret)
+                        goto out;
+
+                if (mod_dict && (data = dict_get (mod_dict, "volume-name"))) {
+                        volname = data->data;
+                        if (strcmp (volname, voliter->volname) == 0)
+                                dict_copy (mod_dict, set_dict);
                 }
 
-        }
+                ret = build_client_graph (&cgraph, voliter, set_dict);
+                if (ret)
+                        goto out;
 
-        if (dist_count > 1) {
-                if (replicate_count) {
-                        VOLGEN_GENERATE_VOLNAME (subvol, volname,
-                                                 "replicate");
-                } else if (stripe_count) {
-                        VOLGEN_GENERATE_VOLNAME (subvol, volname,
-                                                 "stripe");
+                if (mod_dict) {
+                        dict_copy (mod_dict, set_dict);
+                        ret = volgen_graph_set_options_generic (&cgraph, set_dict, voliter,
+                                                                basic_option_handler);
                 } else {
-                        VOLGEN_GENERATE_VOLNAME (subvol, volname,
-                                                 "client");
+                        ret = volgen_graph_set_options_generic (&cgraph, voliter->dict, voliter,
+                                                                basic_option_handler);
                 }
 
+                if (ret)
+                        goto out;
 
-                ret = __write_distribute_xlator (file,
-                                                 dict,
-                                                 subvol,
-                                                 dist_count);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Count not write xlator");
+                ret = volgen_graph_merge_sub (graph, &cgraph, 1);
+                if (ret)
+                        goto out;
+                ret = dict_reset (set_dict);
+                if (ret)
                         goto out;
-                }
         }
 
+        list_for_each_entry (voliter, &priv->volumes, vol_list) {
+
+                if (mod_dict) {
+                        ret = volgen_graph_set_options_generic (graph, mod_dict, voliter,
+                                                                nfs_option_handler);
+                } else {
+                        ret = volgen_graph_set_options_generic (graph, voliter->dict, voliter,
+                                                                nfs_option_handler);
+                }
 
-        write_perf_xlators (volinfo, file, dist_count,
-                            replicate_count, stripe_count);
+                if (ret)
+                        gf_log ("glusterd", GF_LOG_WARNING, "Could not set "
+                                 "vol-options for the volume %s", voliter->volname);
+        }
 
-        fclose (file);
-        file = NULL;
+ out:
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        dict_destroy (set_dict);
 
-out:
         return ret;
 }
 
-static char *
-get_brick_filename (glusterd_volinfo_t *volinfo,
+
+
+
+/****************************
+ *
+ * Volume generation interface
+ *
+ ****************************/
+
+
+static void
+get_brick_filepath (char *filename, glusterd_volinfo_t *volinfo,
                     glusterd_brickinfo_t *brickinfo)
 {
         char  path[PATH_MAX]   = {0,};
-        char *ret              = NULL;
         char  brick[PATH_MAX]  = {0,};
-        char *filename         = NULL;
+        glusterd_conf_t *priv  = NULL;
 
-        filename = GF_CALLOC (1, PATH_MAX, gf_gld_mt_char);
-        if (!filename)
-                goto out;
+        priv = THIS->private;
 
         GLUSTERD_REMOVE_SLASH_FROM_PATH (brickinfo->path, brick);
-        VOLGEN_GET_VOLUME_DIR (path, volinfo);
+        GLUSTERD_GET_VOLUME_DIR (path, volinfo, priv);
 
         snprintf (filename, PATH_MAX, "%s/%s.%s.%s.vol",
                   path, volinfo->volname,
                   brickinfo->hostname,
                   brick);
-
-        ret = filename;
-out:
-        return ret;
 }
 
-char *
-glusterd_get_nfs_filepath ()
+gf_boolean_t
+glusterd_is_valid_volfpath (char *volname, char *brick)
 {
-        char  path[PATH_MAX] = {0,};
-        char *ret            = NULL;
-        char *filepath       = NULL;
+        char                    volfpath[PATH_MAX] = {0,};
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        glusterd_volinfo_t      *volinfo = NULL;
+        int32_t                 ret = 0;
+        xlator_t                *this = NULL;
 
-        filepath = GF_CALLOC (1, PATH_MAX, gf_common_mt_char);
-        if (!filepath) {
-                gf_log ("", GF_LOG_ERROR, "Unable to allocate nfs file path");
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = glusterd_brickinfo_new_from_brick (brick, &brickinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "Failed to create brickinfo"
+                        " for brick %s", brick );
+                ret = 0;
                 goto out;
         }
+        ret = glusterd_volinfo_new (&volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "Failed to create volinfo");
+                ret = 0;
+                goto out;
+        }
+        strncpy (volinfo->volname, volname, sizeof (volinfo->volname));
+        get_brick_filepath (volfpath, volinfo, brickinfo);
 
-        VOLGEN_GET_NFS_DIR (path);
-
-        snprintf (filepath, PATH_MAX, "%s/nfs-server.vol", path);
+        ret = (strlen (volfpath) < _POSIX_PATH_MAX);
 
-        ret = filepath;
 out:
+        if (brickinfo)
+                glusterd_brickinfo_delete (brickinfo);
+        if (volinfo)
+                glusterd_volinfo_delete (volinfo);
         return ret;
 }
 
-
-static char *
-get_client_filepath (glusterd_volinfo_t *volinfo)
+static int
+glusterd_generate_brick_volfile (glusterd_volinfo_t *volinfo,
+                                 glusterd_brickinfo_t *brickinfo)
 {
-        char  path[PATH_MAX] = {0,};
-        char *ret            = NULL;
-        char *filename       = NULL;
+        volgen_graph_t graph = {0,};
+        char    filename[PATH_MAX] = {0,};
+        int     ret = -1;
 
-        filename = GF_CALLOC (1, PATH_MAX, gf_gld_mt_char);
-        if (!filename)
-                goto out;
+        GF_ASSERT (volinfo);
+        GF_ASSERT (brickinfo);
 
-        VOLGEN_GET_VOLUME_DIR (path, volinfo);
+        get_brick_filepath (filename, volinfo, brickinfo);
 
-        snprintf (filename, PATH_MAX, "%s/%s-fuse.vol",
-                  path, volinfo->volname);
+        ret = build_server_graph (&graph, volinfo, NULL, brickinfo);
+        if (!ret)
+                ret = volgen_write_volfile (&graph, filename);
+
+        volgen_graph_free (&graph);
 
-        ret = filename;
-out:
         return ret;
 }
 
-static int
+static void
+get_vol_tstamp_file (char *filename, glusterd_volinfo_t *volinfo)
+{
+        glusterd_conf_t *priv  = NULL;
+
+        priv = THIS->private;
+
+        GLUSTERD_GET_VOLUME_DIR (filename, volinfo, priv);
+        strncat (filename, "/marker.tstamp",
+                 PATH_MAX - strlen(filename) - 1);
+}
+
+int
 generate_brick_volfiles (glusterd_volinfo_t *volinfo)
 {
-        glusterd_brickinfo_t *brickinfo = NULL;
-        char                 *filename  = NULL;
-        int ret = -1;
+        glusterd_brickinfo_t    *brickinfo = NULL;
+        char                     tstamp_file[PATH_MAX] = {0,};
+        int                      ret = -1;
+
+        ret = glusterd_volinfo_get_boolean (volinfo, VKEY_MARKER_XTIME);
+        if (ret == -1)
+                return -1;
+
+        get_vol_tstamp_file (tstamp_file, volinfo);
+
+        if (ret) {
+                ret = open (tstamp_file, O_WRONLY|O_CREAT|O_EXCL, 0600);
+                if (ret == -1 && errno == EEXIST) {
+                        gf_log ("", GF_LOG_DEBUG, "timestamp file exist");
+                        ret = -2;
+                }
+                if (ret == -1) {
+                        gf_log ("", GF_LOG_ERROR, "failed to create %s (%s)",
+                                tstamp_file, strerror (errno));
+                        return -1;
+                }
+                if (ret >= 0)
+                        close (ret);
+        } else {
+                ret = unlink (tstamp_file);
+                if (ret == -1 && errno == ENOENT)
+                        ret = 0;
+                if (ret == -1) {
+                        gf_log ("", GF_LOG_ERROR, "failed to unlink %s (%s)",
+                                tstamp_file, strerror (errno));
+                        return -1;
+                }
+        }
 
         list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
                 gf_log ("", GF_LOG_DEBUG,
                         "Found a brick - %s:%s", brickinfo->hostname,
                         brickinfo->path);
 
-                filename = get_brick_filename (volinfo, brickinfo);
-                if (!filename) {
-                        gf_log ("", GF_LOG_ERROR,
-                                "Out of memory");
-                        ret = -1;
-                        goto out;
-                }
-
-                ret = generate_server_volfile (brickinfo, volinfo->dict,
-                                               filename);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "Could not generate volfile for brick %s:%s",
-                                brickinfo->hostname, brickinfo->path);
+                ret = glusterd_generate_brick_volfile (volinfo, brickinfo);
+                if (ret)
                         goto out;
-                }
 
-                if (filename)
-                        GF_FREE (filename);
         }
 
         ret = 0;
+
 out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
 static int
-glusterfsd_write_nfs_xlator (int fd, char *subvols, char *volume_ids)
+generate_single_transport_client_volfile (glusterd_volinfo_t *volinfo,
+                                          char *filepath, dict_t *dict)
 {
-        char    *dup_subvols = NULL;
-        char    *subvols_remain = NULL;
-        char    *subvol      = NULL;
-        char    *str         = NULL;
-        char    *free_ptr    = NULL;
-        const char *nfs_str = "volume nfs-server\n"
-                              "type nfs/server\n";
+        volgen_graph_t graph = {0,};
+        int     ret = -1;
 
-        if (fd <= 0)
-                return -1;
+        ret = build_client_graph (&graph, volinfo, dict);
+        if (!ret)
+                ret = volgen_write_volfile (&graph, filepath);
 
-        dup_subvols = gf_strdup (subvols);
-        if (!dup_subvols)
-                return -1;
-        else
-                free_ptr = dup_subvols;
+        volgen_graph_free (&graph);
 
-        write (fd, nfs_str, strlen(nfs_str));
+        return ret;
+}
 
-        subvol = strtok_r (dup_subvols, " \n", &subvols_remain);
-        while (subvol) {
-                str = "option rpc-auth.addr.";
-                write (fd, str, strlen (str));
-                write (fd, subvol, strlen (subvol));
-                str = ".allow *\n";
-                write (fd, str, strlen (str));
-                subvol = strtok_r (NULL, " \n", &subvols_remain);
+static void
+enumerate_transport_reqs (gf_transport_type type, char **types)
+{
+        switch (type) {
+        case GF_TRANSPORT_TCP:
+                types[0] = "tcp";
+                break;
+        case GF_TRANSPORT_RDMA:
+                types[0] = "rdma";
+                break;
+        case GF_TRANSPORT_BOTH_TCP_RDMA:
+                types[0] = "tcp";
+                types[1] = "rdma";
+                break;
         }
-        str = "option nfs.dynamic-volumes on\n";
-        write (fd, str, strlen (str));
+}
 
-        /* Write fsids */
-        write (fd, volume_ids, strlen (volume_ids));
+int
+generate_client_volfiles (glusterd_volinfo_t *volinfo,
+                          glusterd_client_type_t client_type)
+{
+        char               filepath[PATH_MAX] = {0,};
+        int                ret = -1;
+        char               *types[] = {NULL, NULL, NULL};
+        int                i = 0;
+        dict_t             *dict = NULL;
+        gf_transport_type  type = GF_TRANSPORT_TCP;
+
+        enumerate_transport_reqs (volinfo->transport_type, types);
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+        for (i = 0; types[i]; i++) {
+                memset (filepath, 0, sizeof (filepath));
+                ret = dict_set_str (dict, "client-transport-type", types[i]);
+                if (ret)
+                        goto out;
+                type = transport_str_to_type (types[i]);
 
-        str = "subvolumes ";
-        write (fd, str, strlen (str));
-        write (fd, subvols, strlen (subvols));
-        str = "\nend-volume\n";
-        write (fd, str, strlen (str));
-        GF_FREE (free_ptr);
+                ret = dict_set_uint32 (dict, "trusted-client", client_type);
+                if (ret)
+                        goto out;
 
-        return 0;
+                if (client_type == GF_CLIENT_TRUSTED) {
+                        glusterd_get_trusted_client_filepath (filepath,
+                                                              volinfo,
+                                                              type);
+                } else {
+                        glusterd_get_client_filepath (filepath,
+                                                      volinfo,
+                                                      type);
+                }
+
+                ret = generate_single_transport_client_volfile (volinfo,
+                                                                filepath,
+                                                                dict);
+                if (ret)
+                        goto out;
+        }
+out:
+        if (dict)
+                dict_unref (dict);
+        return ret;
 }
 
 int
-volgen_generate_nfs_volfile (glusterd_volinfo_t *volinfo)
-{
-        char               *nfs_filepath             = NULL;
-        char               *fuse_filepath            = NULL;
-        int                 nfs_fd                   = -1;
-        int                 fuse_fd                  = -1;
-        int                 ret                      = -1;
-        char                nfs_orig_path[PATH_MAX]  = {0,};
-        char               *pad                      = NULL;
-        char               *nfs_subvols              = NULL;
-        char                fuse_subvols[2048]       = {0,};
-        int                 subvol_len               = 0;
-        char               *nfs_vol_id               = NULL;
-        char                nfs_vol_id_opt[512]      = {0,};
-        char                volume_id[64]            = {0,};
-        int                 nfs_volid_len            = 0;
-        glusterd_volinfo_t *voliter                  = NULL;
-        glusterd_conf_t    *priv                     = NULL;
-        xlator_t           *this                     = NULL;
+glusterd_create_rb_volfiles (glusterd_volinfo_t *volinfo,
+                             glusterd_brickinfo_t *brickinfo)
+{
+        int ret = -1;
+
+        ret = glusterd_generate_brick_volfile (volinfo, brickinfo);
+        if (!ret)
+                ret = generate_client_volfiles (volinfo, GF_CLIENT_TRUSTED);
+        if (!ret)
+                ret = glusterd_fetchspec_notify (THIS);
+
+        return ret;
+}
+
+int
+glusterd_create_volfiles_and_notify_services (glusterd_volinfo_t *volinfo)
+{
+        int        ret  = -1;
+        xlator_t  *this = NULL;
 
         this = THIS;
-        GF_ASSERT (this);
-        priv = this->private;
-        GF_ASSERT (priv);
-        if (!volinfo) {
-                gf_log ("", GF_LOG_ERROR, "Invalid Volume info");
+
+        ret = generate_brick_volfiles (volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not generate volfiles for bricks");
                 goto out;
         }
 
-        nfs_filepath = glusterd_get_nfs_filepath (volinfo);
-        if (!nfs_filepath)
+        ret = generate_client_volfiles (volinfo, GF_CLIENT_TRUSTED);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not generate trusted client volfiles");
                 goto out;
+        }
 
-        strncat (nfs_filepath, ".tmp", PATH_MAX);
-        nfs_fd = open (nfs_filepath, O_WRONLY|O_TRUNC|O_CREAT, 0666);
-        if (nfs_fd < 0) {
-                gf_log ("", GF_LOG_ERROR, "Could not open file: %s",
-                        nfs_filepath);
+        ret = generate_client_volfiles (volinfo, GF_CLIENT_OTHER);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not generate client volfiles");
                 goto out;
         }
 
+        ret = glusterd_fetchspec_notify (this);
 
-        list_for_each_entry (voliter, &priv->volumes, vol_list) {
-                if (voliter->status != GLUSTERD_STATUS_STARTED)
-                        continue;
-                else {
-                        subvol_len += (strlen (voliter->volname) + 1); // ' '
-                        // "option nfs3.<volume>.volume-id <uuid>\n"
-                        nfs_volid_len += (7 + 4 + 11 + 40 +
-                                          strlen (voliter->volname));
-                }
-        }
+out:
+        return ret;
+}
 
-        if (subvol_len == 0) {
-                gf_log ("", GF_LOG_ERROR, "No volumes started");
-                ret = -1;
+int
+glusterd_create_global_volfile (int (*builder) (volgen_graph_t *graph,
+                                                dict_t *set_dict),
+                                char *filepath, dict_t  *mod_dict)
+{
+        volgen_graph_t graph = {0,};
+        int     ret = -1;
+
+        ret = builder (&graph, mod_dict);
+        if (!ret)
+                ret = volgen_write_volfile (&graph, filepath);
+
+        volgen_graph_free (&graph);
+
+        return ret;
+}
+
+int
+glusterd_create_nfs_volfile ()
+{
+        char            filepath[PATH_MAX] = {0,};
+        glusterd_conf_t *conf = THIS->private;
+
+        glusterd_get_nodesvc_volfile ("nfs", conf->workdir,
+                                            filepath, sizeof (filepath));
+        return glusterd_create_global_volfile (build_nfs_graph,
+                                               filepath, NULL);
+}
+
+int
+glusterd_create_shd_volfile ()
+{
+        char            filepath[PATH_MAX] = {0,};
+        int             ret = -1;
+        glusterd_conf_t *conf = THIS->private;
+        dict_t          *mod_dict = NULL;
+
+        mod_dict = dict_new ();
+        if (!mod_dict)
+                goto out;
+
+        ret = dict_set_uint32 (mod_dict, "cluster.background-self-heal-count", 0);
+        if (ret)
+                goto out;
+
+        ret = dict_set_str (mod_dict, "cluster.data-self-heal", "on");
+        if (ret)
+                goto out;
+
+        ret = dict_set_str (mod_dict, "cluster.metadata-self-heal", "on");
+        if (ret)
+                goto out;
+
+        ret = dict_set_str (mod_dict, "cluster.entry-self-heal", "on");
+        if (ret)
+                goto out;
+
+        glusterd_get_nodesvc_volfile ("glustershd", conf->workdir,
+                                      filepath, sizeof (filepath));
+        ret = glusterd_create_global_volfile (build_shd_graph, filepath,
+                                              mod_dict);
+out:
+        if (mod_dict)
+                dict_unref (mod_dict);
+        return ret;
+}
+
+int
+glusterd_check_nfs_topology_identical (gf_boolean_t *identical)
+{
+        char            nfsvol[PATH_MAX]        = {0,};
+        char            tmpnfsvol[PATH_MAX]     = {0,};
+        glusterd_conf_t *conf                   = NULL;
+        xlator_t        *this                   = THIS;
+        int             ret                     = -1;
+        int             tmpclean                = 0;
+        int             tmpfd                   = -1;
+
+        if ((!identical) || (!this) || (!this->private))
+                goto out;
+
+        conf = (glusterd_conf_t *) this->private;
+
+        /* Fetch the original NFS volfile */
+        glusterd_get_nodesvc_volfile ("nfs", conf->workdir,
+                                      nfsvol, sizeof (nfsvol));
+
+        /* Create the temporary NFS volfile */
+        snprintf (tmpnfsvol, sizeof (tmpnfsvol), "/tmp/gnfs-XXXXXX");
+        tmpfd = mkstemp (tmpnfsvol);
+        if (tmpfd < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                                    "Unable to create temp file %s: (%s)",
+                                    tmpnfsvol, strerror (errno));
                 goto out;
         }
-        subvol_len++; //null character
-        nfs_subvols = GF_CALLOC (subvol_len, sizeof(*nfs_subvols),
-                                 gf_common_mt_char);
-        if (!nfs_subvols) {
-                gf_log ("", GF_LOG_ERROR, "Memory not available");
-                ret = -1;
+
+        tmpclean = 1; /* SET the flag to unlink() tmpfile */
+
+        ret = glusterd_create_global_volfile (build_nfs_graph,
+                                              tmpnfsvol, NULL);
+        if (ret)
+                goto out;
+
+        /* Compare the topology of volfiles */
+        ret = glusterd_check_topology_identical (nfsvol, tmpnfsvol,
+                                                 identical);
+out:
+        if (tmpfd >= 0)
+                close (tmpfd);
+        if (tmpclean)
+                unlink (tmpnfsvol);
+        return ret;
+}
+
+int
+glusterd_check_nfs_volfile_identical (gf_boolean_t *identical)
+{
+        char            nfsvol[PATH_MAX]        = {0,};
+        char            tmpnfsvol[PATH_MAX]     = {0,};
+        glusterd_conf_t *conf                   = NULL;
+        xlator_t        *this                   = NULL;
+        int             ret                     = -1;
+        int             need_unlink             = 0;
+        int             tmp_fd                  = -1;
+
+        this = THIS;
+
+        GF_ASSERT (this);
+        GF_ASSERT (identical);
+
+        conf = this->private;
+
+        glusterd_get_nodesvc_volfile ("nfs", conf->workdir,
+                                      nfsvol, sizeof (nfsvol));
+
+        snprintf (tmpnfsvol, sizeof (tmpnfsvol), "/tmp/gnfs-XXXXXX");
+
+        tmp_fd = mkstemp (tmpnfsvol);
+        if (tmp_fd < 0) {
+                gf_log ("", GF_LOG_WARNING, "Unable to create temp file %s: "
+                                "(%s)", tmpnfsvol, strerror (errno));
                 goto out;
         }
 
-        nfs_vol_id = GF_CALLOC (nfs_volid_len, sizeof (char),
-                                 gf_common_mt_char);
-        if (!nfs_vol_id) {
-                gf_log ("", GF_LOG_ERROR, "Memory not available");
-                ret = -1;
+        need_unlink = 1;
+
+        ret = glusterd_create_global_volfile (build_nfs_graph,
+                                              tmpnfsvol, NULL);
+        if (ret)
+                goto out;
+
+        ret = glusterd_check_files_identical (nfsvol, tmpnfsvol,
+                                              identical);
+        if (ret)
+                goto out;
+
+out:
+        if (need_unlink)
+                unlink (tmpnfsvol);
+
+        if (tmp_fd >= 0)
+                close (tmp_fd);
+
+        return ret;
+}
+
+int
+glusterd_delete_volfile (glusterd_volinfo_t *volinfo,
+                         glusterd_brickinfo_t *brickinfo)
+{
+        int  ret = 0;
+        char filename[PATH_MAX] = {0,};
+
+        GF_ASSERT (volinfo);
+        GF_ASSERT (brickinfo);
+
+        get_brick_filepath (filename, volinfo, brickinfo);
+        ret = unlink (filename);
+        if (ret)
+                gf_log ("glusterd", GF_LOG_ERROR, "failed to delete file: %s, "
+                        "reason: %s", filename, strerror (errno));
+        return ret;
+}
+
+int
+validate_shdopts (glusterd_volinfo_t *volinfo,
+                  dict_t *val_dict,
+                  char **op_errstr)
+{
+        volgen_graph_t graph = {0,};
+        int     ret = -1;
+
+        graph.errstr = op_errstr;
+
+        if (!glusterd_is_volume_replicate (volinfo)) {
+                ret = 0;
                 goto out;
         }
+        ret = dict_set_str (val_dict, "graph-check", "on");
+        if (ret)
+                goto out;
+        ret = build_shd_graph (&graph, val_dict);
+        if (!ret)
+                ret = graph_reconf_validateopt (&graph.graph, op_errstr);
 
-        voliter = NULL;
-        list_for_each_entry (voliter, &priv->volumes, vol_list) {
-                if (voliter->status != GLUSTERD_STATUS_STARTED)
-                        continue;
+        volgen_graph_free (&graph);
 
-                gf_log ("", GF_LOG_DEBUG,
-                        "adding fuse info of - %s", voliter->volname);
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+out:
+        dict_del (val_dict, "graph-check");
+        return ret;
+}
 
-                snprintf (fuse_subvols, sizeof(fuse_subvols), " %s", voliter->volname);
-                fuse_filepath = get_client_filepath (voliter);
-                if (!fuse_filepath) {
-                        ret = -1;
-                        goto out;
-                }
+int
+validate_nfsopts (glusterd_volinfo_t *volinfo,
+                    dict_t *val_dict,
+                    char **op_errstr)
+{
+        volgen_graph_t graph = {0,};
+        int     ret = -1;
+        char    transport_type[16] = {0,};
+        char    *tt                = NULL;
+        char    err_str[4096]      = {0,};
+        xlator_t *this             = THIS;
+
+        GF_ASSERT (this);
 
-                fuse_fd = open (fuse_filepath, O_RDONLY);
-                if (fuse_fd < 0) {
-                        gf_log ("", GF_LOG_ERROR, "Could not open file: %s",
-                                fuse_filepath);
+        graph.errstr = op_errstr;
+
+        get_vol_transport_type (volinfo, transport_type);
+        ret = dict_get_str (val_dict, "nfs.transport-type", &tt);
+        if (!ret) {
+                if (volinfo->transport_type != GF_TRANSPORT_BOTH_TCP_RDMA) {
+                        snprintf (err_str, sizeof (err_str), "Changing nfs "
+                                  "transport type is allowed only for volumes "
+                                  "of transport type tcp,rdma");
+                        gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                        *op_errstr = gf_strdup (err_str);
                         ret = -1;
                         goto out;
                 }
-
-                ret = glusterd_file_copy (nfs_fd, fuse_fd);
-                if (ret)
-                        goto out;
-                GF_FREE (fuse_filepath);
-                fuse_filepath = NULL;
-                close (fuse_fd);
-                fuse_fd = -1;
-                if (subvol_len > strlen (fuse_subvols)) {
-                        strncat (nfs_subvols, fuse_subvols, subvol_len - 1);
-                        subvol_len -= strlen (fuse_subvols);
-                } else {
+                if (strcmp (tt,"tcp") && strcmp (tt,"rdma")) {
+                        snprintf (err_str, sizeof (err_str), "wrong transport "
+                                  "type %s", tt);
+                        *op_errstr = gf_strdup (err_str);
                         ret = -1;
-                        gf_log ("", GF_LOG_ERROR, "Too many subvolumes");
                         goto out;
                 }
-                uuid_unparse (voliter->volume_id, volume_id);
-                snprintf (nfs_vol_id_opt, 512, "option nfs3.%s.volume-id %s\n",
-                          voliter->volname, volume_id);
-                strcat (nfs_vol_id, nfs_vol_id_opt);
         }
 
-        ret = glusterfsd_write_nfs_xlator (nfs_fd, nfs_subvols, nfs_vol_id);
-        if (ret)
+        ret = dict_set_str (val_dict, "volume-name", volinfo->volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set volume name");
                 goto out;
+        }
 
-        strncpy (nfs_orig_path, nfs_filepath, PATH_MAX);
-        pad = strrchr (nfs_orig_path, '.');
-        if (!pad) {
-                gf_log ("", GF_LOG_ERROR, "Failed to find the pad in nfs pat");
-                ret = -1;
-                goto out;
+        ret = build_nfs_graph (&graph, val_dict);
+        if (!ret)
+                ret = graph_reconf_validateopt (&graph.graph, op_errstr);
+
+        volgen_graph_free (&graph);
+
+out:
+        if (dict_get (val_dict, "volume-name"))
+                dict_del (val_dict, "volume-name");
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+
+int
+validate_clientopts (glusterd_volinfo_t *volinfo,
+                     dict_t *val_dict,
+                     char **op_errstr)
+{
+        volgen_graph_t graph = {0,};
+        int            ret   = -1;
+
+        GF_ASSERT (volinfo);
+
+        graph.errstr = op_errstr;
+
+        ret = build_client_graph (&graph, volinfo, val_dict);
+        if (!ret)
+                ret = graph_reconf_validateopt (&graph.graph, op_errstr);
+
+        volgen_graph_free (&graph);
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+validate_brickopts (glusterd_volinfo_t *volinfo,
+                    glusterd_brickinfo_t *brickinfo,
+                    dict_t *val_dict,
+                    char **op_errstr)
+{
+        volgen_graph_t graph = {0,};
+        int            ret   = -1;
+
+        GF_ASSERT (volinfo);
+
+        graph.errstr = op_errstr;
+
+        ret = build_server_graph (&graph, volinfo, val_dict, brickinfo);
+        if (!ret)
+                ret = graph_reconf_validateopt (&graph.graph, op_errstr);
+
+        volgen_graph_free (&graph);
+
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_validate_brickreconf (glusterd_volinfo_t *volinfo,
+                               dict_t *val_dict,
+                               char **op_errstr)
+{
+        glusterd_brickinfo_t *brickinfo = NULL;
+        int                   ret = -1;
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                gf_log ("", GF_LOG_DEBUG,
+                        "Validating %s", brickinfo->hostname);
+
+                ret = validate_brickopts (volinfo, brickinfo, val_dict,
+                                          op_errstr);
+                if (ret)
+                        goto out;
         }
-        *pad = '\0';
-        ret = rename (nfs_filepath, nfs_orig_path);
+
+        ret = 0;
+
 out:
-        if (ret && nfs_filepath)
-                unlink (nfs_filepath);
-        if (fuse_filepath)
-                GF_FREE (fuse_filepath);
-        if (nfs_filepath)
-                GF_FREE (nfs_filepath);
-        if (nfs_vol_id)
-                GF_FREE (nfs_vol_id);
-        if (nfs_subvols)
-                GF_FREE (nfs_subvols);
-        if (fuse_fd > 0)
-                close (fuse_fd);
-        if (nfs_fd > 0)
-                close (nfs_fd);
         return ret;
 }
 
 static int
-generate_client_volfiles (glusterd_volinfo_t *volinfo)
+_check_globalopt (dict_t *this, char *key, data_t *value, void *ret_val)
 {
-        char                 *filename    = NULL;
-        int                   ret         = -1;
+        int *ret = NULL;
 
-        filename = get_client_filepath (volinfo);
-        if (!filename) {
-                gf_log ("", GF_LOG_ERROR,
-                        "Out of memory");
-                ret = -1;
+        ret = ret_val;
+        if (*ret)
+                return 0;
+        if (!glusterd_check_globaloption (key))
+                *ret = 1;
+
+        return 0;
+}
+
+int
+glusterd_validate_globalopts (glusterd_volinfo_t *volinfo,
+                              dict_t *val_dict, char **op_errstr)
+{
+        int ret = 0;
+
+        dict_foreach (val_dict, _check_globalopt, &ret);
+        if (ret) {
+                *op_errstr = gf_strdup ( "option specified is not a global option");
+                return -1;
+        }
+        ret = glusterd_validate_brickreconf (volinfo, val_dict, op_errstr);
+
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG,
+                        "Could not Validate  bricks");
                 goto out;
         }
 
-        ret = generate_client_volfile (volinfo, filename);
+        ret = validate_clientopts (volinfo, val_dict, op_errstr);
         if (ret) {
                 gf_log ("", GF_LOG_DEBUG,
-                        "Could not generate volfile for client");
+                        "Could not Validate client");
                 goto out;
         }
 
-out:
-        if (filename)
-                GF_FREE (filename);
+        ret = validate_nfsopts (volinfo, val_dict, op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Could not Validate nfs");
+                goto out;
+        }
 
+        ret = validate_shdopts (volinfo, val_dict, op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Could not Validate self-heald");
+                goto out;
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
 static int
-glusterd_volgen_set_transport (glusterd_volinfo_t *volinfo)
+_check_localopt (dict_t *this, char *key, data_t *value, void *ret_val)
 {
-        int ret = 0;
+        int *ret = NULL;
 
-        GF_ASSERT (volinfo->dict);
-
-        if(volinfo->transport_type == GF_TRANSPORT_RDMA) {
-                ret = set_xlator_option (volinfo->dict, VOLGEN_CLIENT_OPTION_TRANSTYPE,
-                                        "rdma");
-                ret = set_xlator_option (volinfo->dict, VOLGEN_SERVER_OPTION_TRANSTYPE,
-                                        "rdma");
-        } else {
-                ret = set_xlator_option (volinfo->dict, VOLGEN_CLIENT_OPTION_TRANSTYPE,
-                                 "tcp");
-                ret = set_xlator_option (volinfo->dict, VOLGEN_SERVER_OPTION_TRANSTYPE,
-                                 "tcp");
-        }
+        ret = ret_val;
+        if (*ret)
+                return 0;
+        if (!glusterd_check_localoption (key))
+                *ret = 1;
 
         return 0;
 }
 
 int
-glusterd_create_volfiles (glusterd_volinfo_t *volinfo)
+glusterd_validate_reconfopts (glusterd_volinfo_t *volinfo, dict_t *val_dict,
+                              char **op_errstr)
 {
-        int ret = -1;
+        int ret = 0;
 
-        glusterd_volgen_set_transport (volinfo);
+        dict_foreach (val_dict, _check_localopt, &ret);
+        if (ret) {
+                *op_errstr = gf_strdup ( "option specified is not a local option");
+                return -1;
+        }
+        ret = glusterd_validate_brickreconf (volinfo, val_dict, op_errstr);
 
-        ret = generate_brick_volfiles (volinfo);
         if (ret) {
                 gf_log ("", GF_LOG_DEBUG,
-                        "Could not generate volfiles for bricks");
+                        "Could not Validate  bricks");
                 goto out;
         }
 
-        ret = generate_client_volfiles (volinfo);
+        ret = validate_clientopts (volinfo, val_dict, op_errstr);
         if (ret) {
                 gf_log ("", GF_LOG_DEBUG,
-                        "Could not generate volfile for client");
+                        "Could not Validate client");
+                goto out;
+        }
+
+        ret = validate_nfsopts (volinfo, val_dict, op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Could not Validate nfs");
+                goto out;
+        }
+
+
+        ret = validate_shdopts (volinfo, val_dict, op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Could not Validate self-heald");
                 goto out;
         }
 
-        ret = glusterd_fetchspec_notify (THIS);
 
 out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
         return ret;
 }
 
+static struct volopt_map_entry *
+_gd_get_vmep (char *key) {
+        char *completion = NULL;
+        struct volopt_map_entry *vmep = NULL;
+        int ret = 0;
+
+        COMPLETE_OPTION ((char *)key, completion, ret);
+        for (vmep = glusterd_volopt_map; vmep->key; vmep++) {
+                if (strcmp (vmep->key, key) == 0)
+                        return vmep;
+        }
+
+        return NULL;
+}
+
+uint32_t
+glusterd_get_op_version_for_key (char *key)
+{
+        struct volopt_map_entry *vmep = NULL;
+
+        GF_ASSERT (key);
+
+        vmep = _gd_get_vmep (key);
+        if (vmep)
+                return vmep->op_version;
+
+        return 0;
+}
+
+gf_boolean_t
+gd_is_client_option (char *key)
+{
+        struct volopt_map_entry *vmep = NULL;
+
+        GF_ASSERT (key);
+
+        vmep = _gd_get_vmep (key);
+        if (vmep && (vmep->flags & OPT_FLAG_CLIENT_OPT))
+                return _gf_true;
+
+        return _gf_false;
+}
+
+gf_boolean_t
+gd_is_xlator_option (char *key)
+{
+        struct volopt_map_entry *vmep = NULL;
+
+        GF_ASSERT (key);
+
+        vmep = _gd_get_vmep (key);
+        if (vmep && (vmep->flags & OPT_FLAG_XLATOR_OPT))
+                return _gf_true;
+
+        return _gf_false;
+}
+
+volume_option_type_t
+_gd_get_option_type (char *key)
+{
+        struct volopt_map_entry *vmep = NULL;
+        void                    *dl_handle = NULL;
+        volume_opt_list_t       vol_opt_list = {{0},};
+        int                     ret = -1;
+        volume_option_t         *opt = NULL;
+        char                    *xlopt_key = NULL;
+        volume_option_type_t    opt_type = GF_OPTION_TYPE_MAX;
+
+        GF_ASSERT (key);
+
+        vmep = _gd_get_vmep (key);
+
+        if (vmep) {
+                INIT_LIST_HEAD (&vol_opt_list.list);
+                ret = xlator_volopt_dynload (vmep->voltype, &dl_handle,
+                                             &vol_opt_list);
+                if (ret)
+                        goto out;
+
+                if (_get_xlator_opt_key_from_vme (vmep, &xlopt_key))
+                        goto out;
+
+                opt = xlator_volume_option_get_list (&vol_opt_list, xlopt_key);
+                _free_xlator_opt_key (xlopt_key);
+
+                if (opt)
+                        opt_type = opt->type;
+        }
+
+out:
+        if (dl_handle) {
+                dlclose (dl_handle);
+                dl_handle = NULL;
+        }
+
+        return opt_type;
+}
+
+gf_boolean_t
+gd_is_boolean_option (char *key)
+{
+        GF_ASSERT (key);
+
+        if (GF_OPTION_TYPE_BOOL == _gd_get_option_type (key))
+                return _gf_true;
+
+        return _gf_false;
+}
+
+/* This function will restore origin volume to it's snap.
+ * The restore operation will simply replace the Gluster origin
+ * volume with the snap volume.
+ * TODO: Multi-volume delete to be done.
+ *       Cleanup in case of restore failure is pending.
+ *
+ * @param orig_vol      volinfo of origin volume
+ * @param snap_vol      volinfo of snapshot volume
+ *
+ * @return 0 on success and negative value on error
+ */
 int
-glusterd_delete_volfile (glusterd_volinfo_t *volinfo,
-                         glusterd_brickinfo_t *brickinfo)
+gd_restore_snap_volume (dict_t *rsp_dict,
+                        glusterd_volinfo_t *orig_vol,
+                        glusterd_volinfo_t *snap_vol)
 {
-        char                    *filename  = NULL;
+        int                     ret             = -1;
+        glusterd_volinfo_t      *new_volinfo    = NULL;
+        glusterd_snap_t         *snap           = NULL;
+        xlator_t                *this           = NULL;
+        glusterd_conf_t         *conf           = NULL;
+        glusterd_volinfo_t      *temp_volinfo   = NULL;
+        glusterd_volinfo_t      *voliter        = NULL;
 
-        GF_ASSERT (volinfo);
-        GF_ASSERT (brickinfo);
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (rsp_dict);
+        conf = this->private;
+        GF_ASSERT (conf);
 
-        filename = get_brick_filename (volinfo, brickinfo);
+        GF_VALIDATE_OR_GOTO (this->name, orig_vol, out);
+        GF_VALIDATE_OR_GOTO (this->name, snap_vol, out);
+        snap = snap_vol->snapshot;
+        GF_VALIDATE_OR_GOTO (this->name, snap, out);
 
-        if (filename)
-                unlink (filename);
+        /* Snap volume must be stoped before performing the
+         * restore operation.
+         */
+        ret = glusterd_stop_volume (snap_vol);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to stop "
+                        "snap volume");
+                goto out;
+        }
 
-        if (filename)
-                GF_FREE (filename);
-        return 0;
+        /* Create a new volinfo for the restored volume */
+        ret = glusterd_volinfo_dup (snap_vol, &new_volinfo, _gf_true);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to create volinfo");
+                goto out;
+        }
+
+        /* Following entries need to be derived from origin volume. */
+        strcpy (new_volinfo->volname, orig_vol->volname);
+        uuid_copy (new_volinfo->volume_id, orig_vol->volume_id);
+        new_volinfo->snap_count = orig_vol->snap_count;
+        new_volinfo->snap_max_hard_limit = orig_vol->snap_max_hard_limit;
+        new_volinfo->is_volume_restored = _gf_true;
+
+        /* Bump the version of the restored volume, so that nodes *
+         * which are done can sync during handshake */
+        new_volinfo->version = orig_vol->version;
+
+        list_for_each_entry_safe (voliter, temp_volinfo,
+                         &orig_vol->snap_volumes, snapvol_list) {
+                list_add_tail (&voliter->snapvol_list,
+                               &new_volinfo->snap_volumes);
+        }
+        /* Copy the snap vol info to the new_volinfo.*/
+        ret = glusterd_snap_volinfo_restore (rsp_dict, new_volinfo, snap_vol);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to restore snap");
+                (void)glusterd_volinfo_delete (new_volinfo);
+                goto out;
+        }
+
+        /* If the orig_vol is already restored then we should delete
+         * the backend LVMs */
+        if (orig_vol->is_volume_restored) {
+                ret = glusterd_lvm_snapshot_remove (rsp_dict, orig_vol);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to remove "
+                                "LVM backend");
+                        (void)glusterd_volinfo_delete (new_volinfo);
+                        goto out;
+                }
+        }
+
+        /* Once the new_volinfo is completely constructed then delete
+         * the orinal volinfo
+         */
+        ret = glusterd_volinfo_delete (orig_vol);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to delete volinfo");
+                (void)glusterd_volinfo_delete (new_volinfo);
+                goto out;
+        }
+
+        /* New volinfo always shows the status as created. Therefore
+         * set the status to stop. */
+        glusterd_set_volume_status (new_volinfo, GLUSTERD_STATUS_STOPPED);
+
+        list_add_tail (&new_volinfo->vol_list, &conf->volumes);
+
+        /* Now delete the snap entry. As a first step delete the snap
+         * volume information stored in store. */
+        ret = glusterd_snap_remove (rsp_dict, snap, _gf_false, _gf_true);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "Failed to delete "
+                        "snap %s", snap->snapname);
+                goto out;
+        }
+
+        ret = glusterd_store_volinfo (new_volinfo,
+                                      GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to store volinfo");
+                goto out;
+        }
+
+        ret = 0;
+out:
+
+        return ret;
 }
diff --git a/xlators/mgmt/glusterd/src/glusterd-volgen.h b/xlators/mgmt/glusterd/src/glusterd-volgen.h
index a93bac675..fcbaaf93e 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volgen.h
+++ b/xlators/mgmt/glusterd/src/glusterd-volgen.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _GLUSTERD_VOLGEN_H_
 #define _GLUSTERD_VOLGEN_H_
 
@@ -24,124 +14,163 @@
 #define _CONFIG_H
 #include "config.h"
 #endif
-#include "xlator.h"
-#include "protocol-common.h"
+
 #include "glusterd.h"
-#include "defaults.h"
-#include "list.h"
-#include "dict.h"
-#include "compat.h"
-#include "compat-errno.h"
-#include "glusterd-sm.h"
-#include "glusterd-op-sm.h"
-#include "cli1.h"
-#include "glusterd-mem-types.h"
-
-#define VOLGEN_GET_NFS_DIR(path)                                      \
-        do {                                                            \
-                glusterd_conf_t *priv = THIS->private;                  \
-                snprintf (path, PATH_MAX, "%s/nfs", priv->workdir);\
-        } while (0);                                                    \
 
-#define VOLGEN_GET_VOLUME_DIR(path, volinfo)                            \
-        do {                                                            \
-                glusterd_conf_t *priv = THIS->private;                  \
-                snprintf (path, PATH_MAX, "%s/vols/%s", priv->workdir,  \
-                          volinfo->volname);                            \
-        } while (0);                                                    \
+/* volopt map key name definitions */
 
+#define VKEY_DIAG_CNT_FOP_HITS    "diagnostics.count-fop-hits"
+#define VKEY_DIAG_LAT_MEASUREMENT "diagnostics.latency-measurement"
+#define VKEY_FEATURES_LIMIT_USAGE "features.limit-usage"
+#define VKEY_MARKER_XTIME         GEOREP".indexing"
+#define VKEY_MARKER_XTIME_FORCE   GEOREP".ignore-pid-check"
+#define VKEY_CHANGELOG            "changelog.changelog"
+#define VKEY_FEATURES_QUOTA       "features.quota"
 
-#define VOLGEN_GET_BRICK_DIR(path, volinfo)                             \
-        do {                                                            \
-                glusterd_conf_t *priv = THIS->private;                  \
-                snprintf (path, PATH_MAX, "%s/%s/%s/%s", priv->workdir, \
-                          GLUSTERD_VOLUME_DIR_PREFIX, volinfo->volname, \
-                          GLUSTERD_BRICK_INFO_DIR);                     \
-        } while (0);                                                    \
-
-
-#define VOLGEN_GENERATE_VOLNAME(str, volname, subvol)             \
-        do {                                                      \
-                snprintf (str, 2048, "%s-%s", volname, subvol);   \
-        } while (0);                                              \
-
-#define VOLGEN_POSIX_OPTION_ODIRECT "volgen_posix_option_statfssize"
-#define VOLGEN_POSIX_OPTION_STATFSSIZE "volgen_posix_option_statfssize"
-#define VOLGEN_POSIX_OPTION_MANDATTR "volgen_posix_option_mandattr"
-#define VOLGEN_POSIX_OPTION_SPANDEVICES "volgen_posix_option_spandevices"
-#define VOLGEN_POSIX_OPTION_BCKUNLINK "volgen_posix_option_bckunlink"
-
-#define VOLGEN_LOCKS_OPTION_TRACE "volgen_locks_option_trace"
-#define VOLGEN_LOCKS_OPTION_MAND "volgen_locks_option_mand"
-
-#define VOLGEN_CLIENT_OPTION_TRANSTYPE "volgen_client_option_transtype"
-#define VOLGEN_CLIENT_OPTION_NODELAY "volgen_client_option_nodelay"
-
-#define VOLGEN_IOT_OPTION_THREADCOUNT "volgen_iot_option_threadcount"
-#define VOLGEN_IOT_OPTION_AUTOSCALING "volgen_iot_option_autoscaling"
-#define VOLGEN_IOT_OPTION_MINTHREADS "volgen_iot_option_minthreads"
-#define VOLGEN_IOT_OPTION_MAXTHREADS "volgen_iot_option_maxthreads"
-
-#define VOLGEN_SERVER_OPTION_TRANSTYPE "volgen_server_option_transtype"
-#define VOLGEN_SERVER_OPTION_NODELAY "volgen_server_option_nodelay"
-
-#define VOLGEN_REPLICATE_OPTION_READSUBVOL "volgen_replicate_option_readsubvol"
-#define VOLGEN_REPLICATE_OPTION_FAVCHILD "volgen_replicate_option_favchild"
-#define VOLGEN_REPLICATE_OPTION_BCKSHCOUNT "volgen_replicate_option_bckshcount"
-#define VOLGEN_REPLICATE_OPTION_DATASH "volgen_replicate_option_datash"
-#define VOLGEN_REPLICATE_OPTION_DATASHALGO "volgen_replicate_option_datashalgo"
-#define VOLGEN_REPLICATE_OPTION_SHWINDOWSIZE "volgen_replicate_option_shwindowsize"
-#define VOLGEN_REPLICATE_OPTION_METASH "volgen_replicate_option_metash"
-#define VOLGEN_REPLICATE_OPTION_ENTRYSH "volgen_replicate_option_entrysh"
-#define VOLGEN_REPLICATE_OPTION_DATACHANGELOG "volgen_replicate_option_datachangelog"
-#define VOLGEN_REPLICATE_OPTION_METADATACHANGELOG "volgen_replicate_option_metadatachangelog"
-#define VOLGEN_REPLICATE_OPTION_ENTRYCHANGELOG "volgen_replicate_option_entrychangelog"
-#define VOLGEN_REPLICATE_OPTION_STRICTREADDIR "volgen_replicate_option_strictreaddir"
-
-#define VOLGEN_STRIPE_OPTION_BLOCKSIZE "volgen_stripe_option_blocksize"
-#define VOLGEN_STRIPE_OPTION_USEXATTR  "volgen_stripe_option_usexattr"
-
-#define VOLGEN_DHT_OPTION_LOOKUPUNHASH "volgen_dht_option_lookupunhash"
-#define VOLGEN_DHT_OPTION_MINFREEDISK "volgen_dht_option_minfreedisk"
-#define VOLGEN_DHT_OPTION_UNHASHSTICKY "volgen_dht_option_unhashsticky"
-
-#define VOLGEN_WB_OPTION_FLUSHBEHIND "volgen_wb_option_flushbehind"
-#define VOLGEN_WB_OPTION_CACHESIZE "volgen_wb_option_cachesize"
-#define VOLGEN_WB_OPTION_DISABLENBYTES "volgen_wb_option_disablenbytes"
-#define VOLGEN_WB_OPTION_OSYNC "volgen_wb_option_osync"
-#define VOLGEN_WB_OPTION_TRICKLINGWRITES "volgen_wb_option_tricklingwrites"
-
-#define VOLGEN_RA_OPTION_ATIME "volgen_ra_option_atime"
-#define VOLGEN_RA_OPTION_PAGECOUNT "volgen_ra_option_pagecount"
-
-#define VOLGEN_IOCACHE_OPTION_PRIORITY "volgen_iocache_option_priority"
-#define VOLGEN_IOCACHE_OPTION_TIMEOUT "volgen_iocache_option_timeout"
-#define VOLGEN_IOCACHE_OPTION_CACHESIZE "volgen_iocache_option_cachesize"
-#define VOLGEN_IOCACHE_OPTION_MINFILESIZE "volgen_iocache_option_minfilesize"
-#define VOLGEN_IOCACHE_OPTION_MAXFILESIZE "volgen_iocache_option_maxfilesize"
-
-#define VOLGEN_QR_OPTION_PRIORITY "volgen_qr_option_priority"
-#define VOLGEN_QR_OPTION_TIMEOUT "volgen_qr_option_timeout"
-#define VOLGEN_QR_OPTION_CACHESIZE "volgen_qr_option_cachesize"
-#define VOLGEN_QR_OPTION_MAXFILESIZE "volgen_qr_option_maxfilesize"
+#define AUTH_ALLOW_MAP_KEY "auth.allow"
+#define AUTH_REJECT_MAP_KEY "auth.reject"
+#define NFS_DISABLE_MAP_KEY "nfs.disable"
+#define AUTH_ALLOW_OPT_KEY "auth.addr.*.allow"
+#define AUTH_REJECT_OPT_KEY "auth.addr.*.reject"
+#define NFS_DISABLE_OPT_KEY "nfs.*.disable"
 
 
-int
-glusterd_create_volfiles (glusterd_volinfo_t *volinfo);
+typedef enum {
+        GF_CLIENT_TRUSTED,
+        GF_CLIENT_OTHER
+} glusterd_client_type_t;
 
+#define COMPLETE_OPTION(key, completion, ret)                           \
+        do {                                                            \
+                if (!strchr (key, '.')) {                               \
+                        ret = option_complete (key, &completion);       \
+                        if (ret) {                                      \
+                                gf_log ("", GF_LOG_ERROR, "Out of memory"); \
+                                return _gf_false;                       \
+                        }                                               \
+                                                                        \
+                        if (!completion) {                              \
+                                gf_log ("", GF_LOG_ERROR, "option %s does not" \
+                                        "exist", key);                  \
+                                return _gf_false;                       \
+                        }                                               \
+                }                                                       \
+                                                                        \
+                if (completion)                                         \
+                        GF_FREE (completion);                           \
+        } while (0);
+
+typedef enum gd_volopt_flags_ {
+        OPT_FLAG_NONE,
+        OPT_FLAG_FORCE = 0x01,      // option needs force to be reset
+        OPT_FLAG_XLATOR_OPT = 0x02, // option enables/disables xlators
+        OPT_FLAG_CLIENT_OPT = 0x04, // option affects clients
+} gd_volopt_flags_t;
+
+typedef enum {
+        GF_XLATOR_POSIX = 0,
+        GF_XLATOR_ACL,
+        GF_XLATOR_LOCKS,
+        GF_XLATOR_IOT,
+        GF_XLATOR_INDEX,
+        GF_XLATOR_MARKER,
+        GF_XLATOR_IO_STATS,
+        GF_XLATOR_BD,
+        GF_XLATOR_NONE,
+} glusterd_server_xlator_t;
+
+/* As of now debug xlators can be loaded only below fuse in the client
+ * graph via cli. More xlators can be added below when the cli option
+ * for adding debug xlators anywhere in the client graph has to be made
+ * available.
+ */
+typedef enum {
+        GF_CLNT_XLATOR_FUSE = 0,
+        GF_CLNT_XLATOR_NONE,
+} glusterd_client_xlator_t;
+
+typedef enum  { DOC, NO_DOC, GLOBAL_DOC, GLOBAL_NO_DOC } option_type_t;
+
+typedef int (*vme_option_validation) (dict_t *dict, char *key, char *value,
+                                      char **op_errstr);
+
+struct volopt_map_entry {
+        char *key;
+        char *voltype;
+        char *option;
+        char *value;
+        option_type_t type;
+        uint32_t flags;
+        uint32_t op_version;
+        char *description;
+        vme_option_validation validate_fn;
+        /* If client_option is true, the option affects clients.
+         * this is used to calculate client-op-version of volumes
+         */
+        //gf_boolean_t client_option;
+};
+
+int glusterd_create_rb_volfiles (glusterd_volinfo_t *volinfo,
+                                 glusterd_brickinfo_t *brickinfo);
+
+int glusterd_create_volfiles_and_notify_services (glusterd_volinfo_t *volinfo);
+
+void glusterd_get_nfs_filepath (char *filename);
+
+void glusterd_get_shd_filepath (char *filename);
+
+int glusterd_create_nfs_volfile ();
+int glusterd_create_shd_volfile ();
+
+int glusterd_delete_volfile (glusterd_volinfo_t *volinfo,
+                             glusterd_brickinfo_t *brickinfo);
+int
+glusterd_delete_snap_volfile (glusterd_volinfo_t *volinfo,
+                              glusterd_volinfo_t *snap_volinfo,
+                              glusterd_brickinfo_t *brickinfo);
+
+int glusterd_volinfo_get (glusterd_volinfo_t *volinfo, char *key, char **value);
+int glusterd_volinfo_get_boolean (glusterd_volinfo_t *volinfo, char *key);
+
+int glusterd_validate_globalopts (glusterd_volinfo_t *volinfo, dict_t *val_dict, char **op_errstr);
+
+int glusterd_validate_localopts (dict_t *val_dict, char **op_errstr);
+gf_boolean_t glusterd_check_globaloption (char *key);
+gf_boolean_t
+glusterd_check_voloption_flags (char *key, int32_t flags);
+gf_boolean_t
+glusterd_is_valid_volfpath (char *volname, char *brick);
+int generate_brick_volfiles (glusterd_volinfo_t *volinfo);
+int generate_snap_brick_volfiles (glusterd_volinfo_t *volinfo,
+                                  glusterd_volinfo_t *snap_volinfo);
+int generate_client_volfiles (glusterd_volinfo_t *volinfo,
+                              glusterd_client_type_t client_type);
+int
+generate_snap_client_volfiles (glusterd_volinfo_t *actual_volinfo,
+                               glusterd_volinfo_t *snap_volinfo,
+                               glusterd_client_type_t client_type,
+                               gf_boolean_t vol_restore);
+int glusterd_get_volopt_content (dict_t *dict, gf_boolean_t xml_out);
+char*
+glusterd_get_trans_type_rb (gf_transport_type ttype);
+int
+glusterd_check_nfs_volfile_identical (gf_boolean_t *identical);
 int
-glusterd_delete_volfile (glusterd_volinfo_t *volinfo,
-                         glusterd_brickinfo_t *brickinfo);
+glusterd_check_nfs_topology_identical (gf_boolean_t *identical);
 
-int32_t
-glusterd_default_xlator_options (glusterd_volinfo_t *volinfo);
+uint32_t
+glusterd_get_op_version_for_key (char *key);
 
-char *
-glusterd_get_nfs_filepath ();
+gf_boolean_t
+gd_is_client_option (char *key);
 
-int
-volgen_generate_nfs_volfile (glusterd_volinfo_t *volinfo);
+gf_boolean_t
+gd_is_xlator_option (char *key);
 
-int
-uncomment_option( char *opt_str,char *comment_str);
+gf_boolean_t
+gd_is_boolean_option (char *key);
+
+int gd_restore_snap_volume (dict_t *rsp_dict,
+                            glusterd_volinfo_t *orig_vol,
+                            glusterd_volinfo_t *snap_vol);
 #endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
new file mode 100644
index 000000000..0d322b9ad
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
@@ -0,0 +1,2225 @@
+/*
+   Copyright (c) 2011-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#ifdef HAVE_BD_XLATOR
+#include <lvm2app.h>
+#endif
+
+#include "common-utils.h"
+#include "syscall.h"
+#include "cli1-xdr.h"
+#include "xdr-generic.h"
+#include "glusterd.h"
+#include "glusterd-op-sm.h"
+#include "glusterd-store.h"
+#include "glusterd-utils.h"
+#include "glusterd-volgen.h"
+#include "run.h"
+
+#define glusterd_op_start_volume_args_get(dict, volname, flags) \
+        glusterd_op_stop_volume_args_get (dict, volname, flags)
+
+
+int
+__glusterd_handle_create_volume (rpcsvc_request_t *req)
+{
+        int32_t                 ret         = -1;
+        gf_cli_req              cli_req     = {{0,}};
+        dict_t                 *dict        = NULL;
+        char                   *bricks      = NULL;
+        char                   *volname     = NULL;
+        int                    brick_count  = 0;
+        void                   *cli_rsp     = NULL;
+        char                    err_str[2048] = {0,};
+        gf_cli_rsp              rsp         = {0,};
+        xlator_t               *this        = NULL;
+        char                   *free_ptr    = NULL;
+        char                   *trans_type  = NULL;
+        uuid_t                  volume_id   = {0,};
+        uuid_t                  tmp_uuid    = {0};
+        int32_t                 type        = 0;
+        char                   *username    = NULL;
+        char                   *password    = NULL;
+
+        GF_ASSERT (req);
+
+        this = THIS;
+        GF_ASSERT(this);
+
+        ret = -1;
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                req->rpc_err = GARBAGE_ARGS;
+                snprintf (err_str, sizeof (err_str), "Failed to decode request "
+                          "received from cli");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "Received create volume req");
+
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
+                        goto out;
+                } else {
+                        dict->extra_stdfree = cli_req.dict.dict_val;
+                }
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get volume "
+                          "name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        if ((ret = glusterd_check_volume_exists (volname))) {
+                snprintf (err_str, sizeof (err_str), "Volume %s already exists",
+                          volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "count", &brick_count);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get brick count"
+                          " for volume %s", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "type", &type);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get type of "
+                          "volume %s", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "transport", &trans_type);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get "
+                          "transport-type of volume %s", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+        ret = dict_get_str (dict, "bricks", &bricks);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get bricks for "
+                          "volume %s", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        if (!dict_get (dict, "force")) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get 'force' flag");
+                goto out;
+        }
+
+        uuid_generate (volume_id);
+        free_ptr = gf_strdup (uuid_utoa (volume_id));
+        ret = dict_set_dynstr (dict, "volume-id", free_ptr);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to set volume "
+                          "id of volume %s", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+        free_ptr = NULL;
+
+        /* generate internal username and password */
+
+        uuid_generate (tmp_uuid);
+        username = gf_strdup (uuid_utoa (tmp_uuid));
+        ret = dict_set_dynstr (dict, "internal-username", username);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set username for "
+                        "volume %s", volname);
+                goto out;
+        }
+
+        uuid_generate (tmp_uuid);
+        password = gf_strdup (uuid_utoa (tmp_uuid));
+        ret = dict_set_dynstr (dict, "internal-password", password);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set password for "
+                        "volume %s", volname);
+                goto out;
+        }
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_CREATE_VOLUME, dict);
+
+out:
+        if (ret) {
+                rsp.op_ret = -1;
+                rsp.op_errno = 0;
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                rsp.op_errstr = err_str;
+                cli_rsp = &rsp;
+                glusterd_to_cli (req, cli_rsp, NULL, 0, NULL,
+                                 (xdrproc_t)xdr_gf_cli_rsp, dict);
+                ret = 0; //Client response sent, prevent second response
+        }
+
+        GF_FREE(free_ptr);
+
+        return ret;
+}
+
+int
+glusterd_handle_create_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_create_volume);
+}
+
+int
+__glusterd_handle_cli_start_volume (rpcsvc_request_t *req)
+{
+        int32_t                         ret = -1;
+        gf_cli_req                      cli_req = {{0,}};
+        char                            *volname = NULL;
+        dict_t                          *dict = NULL;
+        glusterd_op_t                   cli_op = GD_OP_START_VOLUME;
+        char                            errstr[2048] = {0,};
+        xlator_t                        *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                snprintf (errstr, sizeof (errstr), "Failed to decode message "
+                        "received from cli");
+                req->rpc_err = GARBAGE_ARGS;
+                gf_log (this->name, sizeof (errstr), "%s", errstr);
+                goto out;
+        }
+
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (errstr, sizeof (errstr), "Unable to decode "
+                                  "the command");
+                        goto out;
+                }
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (errstr, sizeof (errstr), "Unable to get volume name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", errstr);
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "Received start vol req"
+                " for volume %s", volname);
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_START_VOLUME, dict);
+
+out:
+        free (cli_req.dict.dict_val); //its malloced by xdr
+
+        if (ret) {
+                if(errstr[0] == '\0')
+                        snprintf (errstr, sizeof (errstr), "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, errstr);
+        }
+
+        return ret;
+}
+
+int
+glusterd_handle_cli_start_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_start_volume);
+}
+
+int
+__glusterd_handle_cli_stop_volume (rpcsvc_request_t *req)
+{
+        int32_t                         ret = -1;
+        gf_cli_req                      cli_req = {{0,}};
+        char                            *dup_volname = NULL;
+        dict_t                          *dict = NULL;
+        glusterd_op_t                   cli_op = GD_OP_STOP_VOLUME;
+        xlator_t                        *this = NULL;
+        char                            err_str[2048] = {0,};
+
+        this = THIS;
+        GF_ASSERT (this);
+        GF_ASSERT (req);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                snprintf (err_str, sizeof (err_str), "Failed to decode message "
+                          "received from cli");
+                req->rpc_err = GARBAGE_ARGS;
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
+                        goto out;
+                }
+        }
+
+        ret = dict_get_str (dict, "volname", &dup_volname);
+
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Failed to get volume "
+                          "name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "Received stop vol req "
+                "for volume %s", dup_volname);
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_STOP_VOLUME, dict);
+
+out:
+        free (cli_req.dict.dict_val); //its malloced by xdr
+
+        if (ret) {
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
+        }
+
+        return ret;
+}
+
+int
+glusterd_handle_cli_stop_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_stop_volume);
+}
+
+int
+__glusterd_handle_cli_delete_volume (rpcsvc_request_t *req)
+{
+        int32_t        ret         = -1;
+        gf_cli_req     cli_req     = {{0,},};
+        glusterd_op_t  cli_op      = GD_OP_DELETE_VOLUME;
+        dict_t        *dict        = NULL;
+        char          *volname     = NULL;
+        char          err_str[2048]= {0,};
+        xlator_t      *this        = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        GF_ASSERT (req);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                snprintf (err_str, sizeof (err_str), "Failed to decode request "
+                          "received from cli");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to decode "
+                                  "the command");
+                        goto out;
+                }
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Failed to get volume "
+                          "name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "Received delete vol req"
+                "for volume %s", volname);
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_DELETE_VOLUME, dict);
+
+out:
+        free (cli_req.dict.dict_val); //its malloced by xdr
+
+        if (ret) {
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
+        }
+
+        return ret;
+}
+
+int
+glusterd_handle_cli_delete_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_delete_volume);
+}
+
+int
+__glusterd_handle_cli_heal_volume (rpcsvc_request_t *req)
+{
+        int32_t                         ret = -1;
+        gf_cli_req                      cli_req = {{0,}};
+        dict_t                          *dict = NULL;
+        glusterd_op_t                   cli_op = GD_OP_HEAL_VOLUME;
+        char                            *volname = NULL;
+        glusterd_volinfo_t              *volinfo = NULL;
+        xlator_t                        *this = NULL;
+        char                            op_errstr[2048] = {0,};
+
+        GF_ASSERT (req);
+
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                //failed to decode msg;
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (op_errstr, sizeof (op_errstr),
+                        "Unable to decode the command");
+                        goto out;
+                } else {
+                        dict->extra_stdfree = cli_req.dict.dict_val;
+                }
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (op_errstr, sizeof (op_errstr), "Unable to find "
+                          "volume name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", op_errstr);
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_INFO, "Received heal vol req "
+                "for volume %s", volname);
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (op_errstr, sizeof (op_errstr),
+                          "Volume %s does not exist", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", op_errstr);
+                goto out;
+        }
+
+        ret = glusterd_add_bricks_hname_path_to_dict (dict, volinfo);
+        if (ret)
+                goto out;
+
+        ret = dict_set_int32 (dict, "count", volinfo->brick_count);
+        if (ret)
+                goto out;
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_HEAL_VOLUME, dict);
+
+out:
+        if (ret) {
+                if (op_errstr[0] == '\0')
+                        snprintf (op_errstr, sizeof (op_errstr),
+                                  "operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, op_errstr);
+        }
+
+        return ret;
+}
+
+int
+glusterd_handle_cli_heal_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_heal_volume);
+}
+
+int
+__glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req)
+{
+        int32_t                         ret = -1;
+        gf_cli_req                      cli_req = {{0,}};
+        char                            *volname = NULL;
+        char                            *options = NULL;
+        dict_t                          *dict = NULL;
+        int32_t                         option_cnt = 0;
+        glusterd_op_t                   cli_op = GD_OP_STATEDUMP_VOLUME;
+        char                            err_str[2048] = {0,};
+        xlator_t                        *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        GF_ASSERT (req);
+
+        ret = -1;
+        ret = xdr_to_generic (req->msg[0], &cli_req, (xdrproc_t)xdr_gf_cli_req);
+        if (ret < 0) {
+                req->rpc_err = GARBAGE_ARGS;
+                goto out;
+        }
+        if (cli_req.dict.dict_len) {
+                /* Unserialize the dictionary */
+                dict  = dict_new ();
+
+                ret = dict_unserialize (cli_req.dict.dict_val,
+                                        cli_req.dict.dict_len,
+                                        &dict);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to "
+                                "unserialize req-buffer to dictionary");
+                        snprintf (err_str, sizeof (err_str), "Unable to "
+                                  "decode the command");
+                        goto out;
+                }
+        }
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get the volume "
+                          "name");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "options", &options);
+        if (ret) {
+                snprintf (err_str, sizeof (err_str), "Unable to get options");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "option_cnt", &option_cnt);
+        if (ret) {
+                snprintf (err_str , sizeof (err_str), "Unable to get option "
+                          "count");
+                gf_log (this->name, GF_LOG_ERROR, "%s", err_str);
+                goto out;
+        }
+
+
+        gf_log (this->name, GF_LOG_INFO, "Received statedump request for "
+                "volume %s with options %s", volname, options);
+
+        ret = glusterd_op_begin_synctask (req, GD_OP_STATEDUMP_VOLUME, dict);
+
+out:
+        if (ret) {
+                if (err_str[0] == '\0')
+                        snprintf (err_str, sizeof (err_str),
+                                  "Operation failed");
+                ret = glusterd_op_send_cli_response (cli_op, ret, 0, req,
+                                                     dict, err_str);
+        }
+        free (cli_req.dict.dict_val);
+
+        return ret;
+}
+
+int
+glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req)
+{
+        return glusterd_big_locked_handler (req,
+                                            __glusterd_handle_cli_statedump_volume);
+}
+
+#ifdef HAVE_BD_XLATOR
+/*
+ * Validates if given VG in the brick exists or not. Also checks if VG has
+ * GF_XATTR_VOL_ID_KEY tag set to avoid using same VG for multiple bricks.
+ * Tag is checked only during glusterd_op_stage_create_volume. Tag is set during
+ * glusterd_validate_and_create_brickpath().
+ * @brick - brick info, @check_tag - check for VG tag or not
+ * @msg - Error message to return to caller
+ */
+int
+glusterd_is_valid_vg (glusterd_brickinfo_t *brick, int check_tag, char *msg)
+{
+        lvm_t                     handle     = NULL;
+        vg_t                      vg         = NULL;
+        char                     *vg_name    = NULL;
+        int                       retval     = 0;
+        char                     *p          = NULL;
+        char                     *ptr        = NULL;
+        struct dm_list           *dm_lvlist = NULL;
+        struct dm_list           *dm_seglist = NULL;
+        struct lvm_lv_list       *lv_list    = NULL;
+        struct lvm_property_value prop       = {0, };
+        struct lvm_lvseg_list    *seglist    = NULL;
+        struct dm_list           *taglist    = NULL;
+        struct lvm_str_list      *strl       = NULL;
+
+        handle = lvm_init (NULL);
+        if (!handle) {
+                sprintf (msg, "lvm_init failed, could not validate vg");
+                return -1;
+        }
+        if (*brick->vg == '\0') { /* BD xlator has vg in brick->path */
+                p = gf_strdup (brick->path);
+                vg_name = strtok_r (p, "/", &ptr);
+        } else
+                vg_name = brick->vg;
+
+        vg = lvm_vg_open (handle, vg_name, "r", 0);
+        if (!vg) {
+                sprintf (msg, "no such vg: %s", vg_name);
+                retval = -1;
+                goto out;
+        }
+        if (!check_tag)
+                goto next;
+
+        taglist = lvm_vg_get_tags (vg);
+        if (!taglist)
+                goto next;
+
+        dm_list_iterate_items (strl, taglist) {
+                if (!strncmp(strl->str, GF_XATTR_VOL_ID_KEY,
+                             strlen (GF_XATTR_VOL_ID_KEY))) {
+                            sprintf (msg, "VG %s is already part of"
+                                    " a brick", vg_name);
+                            retval = -1;
+                            goto out;
+                    }
+        }
+next:
+
+        brick->caps = CAPS_BD | CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT;
+
+        dm_lvlist = lvm_vg_list_lvs (vg);
+        if (!dm_lvlist)
+                goto out;
+
+        dm_list_iterate_items (lv_list, dm_lvlist) {
+                dm_seglist = lvm_lv_list_lvsegs (lv_list->lv);
+                dm_list_iterate_items (seglist, dm_seglist) {
+                        prop = lvm_lvseg_get_property (seglist->lvseg,
+                                                       "segtype");
+                        if (!prop.is_valid || !prop.value.string)
+                                continue;
+                        if (!strcmp (prop.value.string, "thin-pool")) {
+                                brick->caps |= CAPS_THIN;
+                                gf_log (THIS->name, GF_LOG_INFO, "Thin Pool "
+                                        "\"%s\" will be used for thin LVs",
+                                        lvm_lv_get_name (lv_list->lv));
+                                break;
+                        }
+                }
+        }
+
+        retval = 0;
+out:
+        if (vg)
+                lvm_vg_close (vg);
+        lvm_quit (handle);
+        if (p)
+                GF_FREE (p);
+        return retval;
+}
+#endif
+
+/* op-sm */
+int
+glusterd_op_stage_create_volume (dict_t *dict, char **op_errstr)
+{
+        int                                     ret = 0;
+        char                                    *volname = NULL;
+        gf_boolean_t                            exists = _gf_false;
+        char                                    *bricks = NULL;
+        char                                    *brick_list = NULL;
+        char                                    *free_ptr = NULL;
+        glusterd_brickinfo_t                    *brick_info = NULL;
+        int32_t                                 brick_count = 0;
+        int32_t                                 i = 0;
+        char                                    *brick = NULL;
+        char                                    *tmpptr = NULL;
+        xlator_t                                *this = NULL;
+        glusterd_conf_t                         *priv = NULL;
+        char                                    msg[2048] = {0};
+        uuid_t                                  volume_uuid;
+        char                                    *volume_uuid_str;
+        gf_boolean_t                             is_force = _gf_false;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        exists = glusterd_check_volume_exists (volname);
+        if (exists) {
+                snprintf (msg, sizeof (msg), "Volume %s already exists",
+                          volname);
+                ret = -1;
+                goto out;
+        } else {
+                ret = 0;
+        }
+
+        ret = dict_get_int32 (dict, "count", &brick_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get brick count "
+                        "for volume %s", volname);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "volume-id", &volume_uuid_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume id of "
+                        "volume %s", volname);
+                goto out;
+        }
+
+        ret = uuid_parse (volume_uuid_str, volume_uuid);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to parse volume id of"
+                        " volume %s", volname);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "bricks", &bricks);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get bricks for "
+                        "volume %s", volname);
+                goto out;
+        }
+
+        is_force = dict_get_str_boolean (dict, "force", _gf_false);
+
+        if (bricks) {
+                brick_list = gf_strdup (bricks);
+                if (!brick_list) {
+                        ret = -1;
+                        goto out;
+                } else {
+                        free_ptr = brick_list;
+                }
+        }
+
+        while ( i < brick_count) {
+                i++;
+                brick= strtok_r (brick_list, " \n", &tmpptr);
+                brick_list = tmpptr;
+
+                if (!glusterd_store_is_valid_brickpath (volname, brick) ||
+                        !glusterd_is_valid_volfpath (volname, brick)) {
+                        snprintf (msg, sizeof (msg), "brick path %s is too "
+                                  "long.", brick);
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = glusterd_brickinfo_new_from_brick (brick, &brick_info);
+                if (ret)
+                        goto out;
+
+                ret = glusterd_new_brick_validate (brick, brick_info, msg,
+                                                   sizeof (msg));
+                if (ret)
+                        goto out;
+
+                ret = glusterd_resolve_brick (brick_info);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK,
+                                brick_info->hostname, brick_info->path);
+                        goto out;
+                }
+
+                if (!uuid_compare (brick_info->uuid, MY_UUID)) {
+
+#ifdef HAVE_BD_XLATOR
+                        if (brick_info->vg[0]) {
+                                ret = glusterd_is_valid_vg (brick_info, 1, msg);
+                                if (ret)
+                                        goto out;
+                        }
+#endif
+                        ret = glusterd_validate_and_create_brickpath (brick_info,
+                                                          volume_uuid, op_errstr,
+                                                          is_force);
+                        if (ret)
+                                goto out;
+                        brick_list = tmpptr;
+                }
+                glusterd_brickinfo_delete (brick_info);
+                brick_info = NULL;
+        }
+out:
+        GF_FREE (free_ptr);
+        if (brick_info)
+                glusterd_brickinfo_delete (brick_info);
+
+        if (msg[0] != '\0') {
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+        }
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+int
+glusterd_op_stop_volume_args_get (dict_t *dict, char** volname, int *flags)
+{
+        int ret = -1;
+        xlator_t *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        if (!dict || !volname || !flags)
+                goto out;
+
+        ret = dict_get_str (dict, "volname", volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "flags", flags);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get flags");
+                goto out;
+        }
+out:
+        return ret;
+}
+
+int
+glusterd_op_statedump_volume_args_get (dict_t *dict, char **volname,
+                                       char **options, int *option_cnt)
+{
+        int ret = -1;
+
+        if (!dict || !volname || !options || !option_cnt)
+                goto out;
+
+        ret = dict_get_str (dict, "volname", volname);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volname");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "options", options);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get options");
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "option_cnt", option_cnt);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get option count");
+                goto out;
+        }
+
+out:
+        return ret;
+}
+
+int
+glusterd_op_stage_start_volume (dict_t *dict, char **op_errstr)
+{
+        int                                     ret = 0;
+        char                                    *volname = NULL;
+        int                                     flags = 0;
+        gf_boolean_t                            exists = _gf_false;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        glusterd_brickinfo_t                    *brickinfo = NULL;
+        char                                    msg[2048];
+        glusterd_conf_t                         *priv = NULL;
+        xlator_t                                *this = NULL;
+        uuid_t                                  volume_id = {0,};
+        char                                    volid[50] = {0,};
+        char                                    xattr_volid[50] = {0,};
+        int                                     caps = 0;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = glusterd_op_start_volume_args_get (dict, &volname, &flags);
+        if (ret)
+                goto out;
+
+        exists = glusterd_check_volume_exists (volname);
+
+        if (!exists) {
+                snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname);
+                ret = -1;
+                goto out;
+        }
+
+        ret  = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS,
+                        volname);
+                goto out;
+        }
+
+        ret = glusterd_validate_volume_id (dict, volinfo);
+        if (ret)
+                goto out;
+
+        if (!(flags & GF_CLI_FLAG_OP_FORCE)) {
+                if (glusterd_is_volume_started (volinfo)) {
+                        snprintf (msg, sizeof (msg), "Volume %s already "
+                                  "started", volname);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                ret = glusterd_resolve_brick (brickinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK,
+                                brickinfo->hostname, brickinfo->path);
+                        goto out;
+                }
+
+                if (uuid_compare (brickinfo->uuid, MY_UUID))
+                        continue;
+
+                ret = gf_lstat_dir (brickinfo->path, NULL);
+                if (ret && (flags & GF_CLI_FLAG_OP_FORCE)) {
+                        continue;
+                } else if (ret) {
+                        snprintf (msg, sizeof (msg), "Failed to find "
+                                          "brick directory %s for volume %s. "
+                                          "Reason : %s", brickinfo->path,
+                                          volname, strerror (errno));
+                        goto out;
+                }
+                ret = sys_lgetxattr (brickinfo->path, GF_XATTR_VOL_ID_KEY,
+                                     volume_id, 16);
+                if (ret < 0 && (!(flags & GF_CLI_FLAG_OP_FORCE))) {
+                        snprintf (msg, sizeof (msg), "Failed to get "
+                                  "extended attribute %s for brick dir %s. "
+                                  "Reason : %s", GF_XATTR_VOL_ID_KEY,
+                                  brickinfo->path, strerror (errno));
+                        ret = -1;
+                        goto out;
+                } else if (ret < 0) {
+                        ret = sys_lsetxattr (brickinfo->path,
+                                             GF_XATTR_VOL_ID_KEY,
+                                             volinfo->volume_id, 16,
+                                             XATTR_CREATE);
+                        if (ret) {
+                                snprintf (msg, sizeof (msg), "Failed to set "
+                                        "extended attribute %s on %s. Reason: "
+                                        "%s", GF_XATTR_VOL_ID_KEY,
+                                        brickinfo->path, strerror (errno));
+                                goto out;
+                        } else {
+                                continue;
+                        }
+                }
+                if (uuid_compare (volinfo->volume_id, volume_id)) {
+                        snprintf (msg, sizeof (msg), "Volume id mismatch for "
+                                  "brick %s:%s. Expected volume id %s, "
+                                  "volume id %s found", brickinfo->hostname,
+                                  brickinfo->path,
+                                  uuid_utoa_r (volinfo->volume_id, volid),
+                                  uuid_utoa_r (volume_id, xattr_volid));
+                        ret = -1;
+                        goto out;
+                }
+#ifdef HAVE_BD_XLATOR
+                if (brickinfo->vg[0])
+                        caps = CAPS_BD | CAPS_THIN |
+                                CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT;
+                /* Check for VG/thin pool if its BD volume */
+                if (brickinfo->vg[0]) {
+                        ret = glusterd_is_valid_vg (brickinfo, 0, msg);
+                        if (ret)
+                                goto out;
+                        /* if anyone of the brick does not have thin support,
+                           disable it for entire volume */
+                        caps &= brickinfo->caps;
+                } else
+                        caps = 0;
+#endif
+        }
+
+        volinfo->caps = caps;
+        ret = 0;
+out:
+        if (ret && (msg[0] != '\0')) {
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+        }
+        return ret;
+}
+
+int
+glusterd_op_stage_stop_volume (dict_t *dict, char **op_errstr)
+{
+        int                                     ret = -1;
+        char                                    *volname = NULL;
+        int                                     flags = 0;
+        gf_boolean_t                            exists = _gf_false;
+        gf_boolean_t                            is_run = _gf_false;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        char                                    msg[2048] = {0};
+        xlator_t                                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = glusterd_op_stop_volume_args_get (dict, &volname, &flags);
+        if (ret)
+                goto out;
+
+        exists = glusterd_check_volume_exists (volname);
+
+        if (!exists) {
+                snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                ret = -1;
+                goto out;
+        }
+
+        ret  = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        ret = glusterd_validate_volume_id (dict, volinfo);
+        if (ret)
+                goto out;
+
+        /* If 'force' flag is given, no check is required */
+        if (flags & GF_CLI_FLAG_OP_FORCE)
+                goto out;
+
+        if (_gf_false == glusterd_is_volume_started (volinfo)) {
+                snprintf (msg, sizeof(msg), "Volume %s "
+                          "is not in the started state", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                ret = -1;
+                goto out;
+        }
+        ret = glusterd_check_gsync_running (volinfo, &is_run);
+        if (ret && (is_run == _gf_false))
+                gf_log (this->name, GF_LOG_WARNING, "Unable to get the status"
+                        " of active "GEOREP" session");
+        if (is_run) {
+                gf_log (this->name, GF_LOG_WARNING, GEOREP" sessions active"
+                        "for the volume %s ", volname);
+                snprintf (msg, sizeof(msg), GEOREP" sessions are active "
+                          "for the volume '%s'.\nUse 'volume "GEOREP" "
+                          "status' command for more info. Use 'force' "
+                          "option to ignore and stop the volume.",
+                          volname);
+                ret = -1;
+                goto out;
+        }
+
+        if (glusterd_is_rb_ongoing (volinfo)) {
+                snprintf (msg, sizeof (msg), "Replace brick is in progress on "
+                          "volume %s. Please retry after replace-brick "
+                          "operation is committed or aborted", volname);
+                gf_log (this->name, GF_LOG_WARNING, "replace-brick in progress "
+                        "on volume %s", volname);
+                ret = -1;
+                goto out;
+        }
+
+        if (glusterd_is_defrag_on (volinfo)) {
+                snprintf (msg, sizeof(msg), "rebalance session is "
+                          "in progress for the volume '%s'", volname);
+                gf_log (this->name, GF_LOG_WARNING, "%s", msg);
+                ret = -1;
+                goto out;
+        }
+        if (volinfo->rep_brick.rb_status != GF_RB_STATUS_NONE) {
+                snprintf (msg, sizeof(msg), "replace-brick session is "
+                          "in progress for the volume '%s'", volname);
+                gf_log (this->name, GF_LOG_WARNING, "%s", msg);
+                ret = -1;
+                goto out;
+        }
+
+out:
+        if (msg[0] != 0)
+                *op_errstr = gf_strdup (msg);
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+int
+glusterd_op_stage_delete_volume (dict_t *dict, char **op_errstr)
+{
+        int                                     ret = 0;
+        char                                    *volname = NULL;
+        gf_boolean_t                            exists = _gf_false;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        char                                    msg[2048] = {0};
+        xlator_t                                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        exists = glusterd_check_volume_exists (volname);
+        if (!exists) {
+                snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname);
+                ret = -1;
+                goto out;
+        } else {
+                ret = 0;
+        }
+
+        ret  = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (msg, sizeof (msg), FMTSTR_CHECK_VOL_EXISTS, volname);
+                goto out;
+        }
+
+        ret = glusterd_validate_volume_id (dict, volinfo);
+        if (ret)
+                goto out;
+
+        if (glusterd_is_volume_started (volinfo)) {
+                snprintf (msg, sizeof (msg), "Volume %s has been started."
+                          "Volume needs to be stopped before deletion.",
+                          volname);
+                ret = -1;
+                goto out;
+        }
+
+        if (volinfo->snap_count > 0 || !list_empty(&volinfo->snap_volumes)) {
+                snprintf (msg, sizeof (msg), "Cannot delete Volume %s ,"
+                        "as it has %ld snapshots. "
+                        "To delete the volume, "
+                        "first delete all the snapshots under it.",
+                          volname, volinfo->snap_count);
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        if (msg[0] != '\0') {
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+        }
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+int
+glusterd_op_stage_heal_volume (dict_t *dict, char **op_errstr)
+{
+        int                                     ret = 0;
+        char                                    *volname = NULL;
+        gf_boolean_t                            enabled = _gf_false;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        char                                    msg[2048];
+        glusterd_conf_t                         *priv = NULL;
+        dict_t                                  *opt_dict = NULL;
+        gf_xl_afr_op_t                          heal_op = GF_AFR_OP_INVALID;
+        xlator_t                                *this = NULL;
+
+        this = THIS;
+        priv = this->private;
+        if (!priv) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "priv is NULL");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log ("", GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret  = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                ret = -1;
+                snprintf (msg, sizeof (msg), "Volume %s does not exist", volname);
+                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = glusterd_validate_volume_id (dict, volinfo);
+        if (ret)
+                goto out;
+
+        if (!glusterd_is_volume_replicate (volinfo)) {
+                ret = -1;
+                snprintf (msg, sizeof (msg), "Volume %s is not of type "
+                          "replicate", volname);
+                *op_errstr = gf_strdup (msg);
+                gf_log (this->name, GF_LOG_WARNING, "%s", msg);
+                goto out;
+        }
+
+        if (!glusterd_is_volume_started (volinfo)) {
+                ret = -1;
+                snprintf (msg, sizeof (msg), "Volume %s is not started.",
+                          volname);
+                gf_log (THIS->name, GF_LOG_WARNING, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        opt_dict = volinfo->dict;
+        if (!opt_dict) {
+                ret = 0;
+                goto out;
+        }
+
+        enabled = dict_get_str_boolean (opt_dict, "cluster.self-heal-daemon",
+                                        1);
+        if (!enabled) {
+                ret = -1;
+                snprintf (msg, sizeof (msg), "Self-heal-daemon is "
+                          "disabled. Heal will not be triggered on volume %s",
+                          volname);
+                gf_log (this->name, GF_LOG_WARNING, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "heal-op", (int32_t*)&heal_op);
+        if (ret || (heal_op == GF_AFR_OP_INVALID)) {
+                ret = -1;
+                *op_errstr = gf_strdup("Invalid heal-op");
+                gf_log (this->name, GF_LOG_WARNING, "%s", "Invalid heal-op");
+                goto out;
+        }
+
+        switch (heal_op) {
+                case GF_AFR_OP_INDEX_SUMMARY:
+                case GF_AFR_OP_STATISTICS_HEAL_COUNT:
+                case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA:
+                        break;
+                default:
+                        if (!glusterd_is_nodesvc_online("glustershd")){
+                                ret = -1;
+                                *op_errstr = gf_strdup ("Self-heal daemon is "
+                                                "not running. Check self-heal "
+                                                "daemon log file.");
+                                gf_log (this->name, GF_LOG_WARNING, "%s",
+                                        "Self-heal daemon is not running."
+                                        "Check self-heal daemon log file.");
+                                goto out;
+                        }
+        }
+
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+int
+glusterd_op_stage_statedump_volume (dict_t *dict, char **op_errstr)
+{
+        int                     ret = -1;
+        char                    *volname = NULL;
+        char                    *options = NULL;
+        int                     option_cnt = 0;
+        gf_boolean_t            is_running = _gf_false;
+        glusterd_volinfo_t      *volinfo = NULL;
+        char                    msg[2408] = {0,};
+
+        ret = glusterd_op_statedump_volume_args_get (dict, &volname, &options,
+                                                     &option_cnt);
+        if (ret)
+                goto out;
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (msg, sizeof(msg), "Volume %s does not exist",
+                          volname);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = glusterd_validate_volume_id (dict, volinfo);
+        if (ret)
+                goto out;
+
+        is_running = glusterd_is_volume_started (volinfo);
+        if (!is_running) {
+                snprintf (msg, sizeof(msg), "Volume %s is not in a started"
+                          " state", volname);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                ret = -1;
+                goto out;
+        }
+
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_stage_clearlocks_volume (dict_t *dict, char **op_errstr)
+{
+        int                     ret = -1;
+        char                    *volname = NULL;
+        char                    *path    = NULL;
+        char                    *type    = NULL;
+        char                    *kind    = NULL;
+        glusterd_volinfo_t      *volinfo = NULL;
+        char                    msg[2048] = {0,};
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                snprintf (msg, sizeof(msg), "Failed to get volume name");
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "path", &path);
+        if (ret) {
+                snprintf (msg, sizeof(msg), "Failed to get path");
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "kind", &kind);
+        if (ret) {
+                snprintf (msg, sizeof(msg), "Failed to get kind");
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "type", &type);
+        if (ret) {
+                snprintf (msg, sizeof(msg), "Failed to get type");
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (msg, sizeof(msg), "Volume %s does not exist",
+                          volname);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = glusterd_validate_volume_id (dict, volinfo);
+        if (ret)
+                goto out;
+
+        if (!glusterd_is_volume_started (volinfo)) {
+                snprintf (msg, sizeof(msg), "Volume %s is not started",
+                          volname);
+                gf_log ("", GF_LOG_ERROR, "%s", msg);
+                *op_errstr = gf_strdup (msg);
+                goto out;
+        }
+
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_create_volume (dict_t *dict, char **op_errstr)
+{
+        int                   ret        = 0;
+        char                 *volname    = NULL;
+        glusterd_conf_t      *priv       = NULL;
+        glusterd_volinfo_t   *volinfo    = NULL;
+        gf_boolean_t          vol_added = _gf_false;
+        glusterd_brickinfo_t *brickinfo  = NULL;
+        xlator_t             *this       = NULL;
+        char                 *brick      = NULL;
+        int32_t               count      = 0;
+        int32_t               i          = 1;
+        char                 *bricks     = NULL;
+        char                 *brick_list = NULL;
+        char                 *free_ptr   = NULL;
+        char                 *saveptr    = NULL;
+        char                 *trans_type = NULL;
+        char                 *str        = NULL;
+        char                 *username   = NULL;
+        char                 *password   = NULL;
+        int                   caps       = 0;
+        char                  msg[1024] __attribute__((unused)) = {0, };
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = glusterd_volinfo_new (&volinfo);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to allocate memory for volinfo");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "volname", &volname);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        strncpy (volinfo->volname, volname, GLUSTERD_MAX_VOLUME_NAME);
+        GF_ASSERT (volinfo->volname);
+
+        ret = dict_get_int32 (dict, "type", &volinfo->type);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get type of volume"
+                        " %s", volname);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "count", &volinfo->brick_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get brick count of"
+                        " volume %s", volname);
+                goto out;
+        }
+
+        ret = dict_get_int32 (dict, "port", &volinfo->port);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get port");
+                goto out;
+        }
+
+        count = volinfo->brick_count;
+
+        ret = dict_get_str (dict, "bricks", &bricks);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get bricks for "
+                        "volume %s", volname);
+                goto out;
+        }
+
+        /* replica-count 1 means, no replication, file is in one brick only */
+        volinfo->replica_count = 1;
+        /* stripe-count 1 means, no striping, file is present as a whole */
+        volinfo->stripe_count = 1;
+
+        if (GF_CLUSTER_TYPE_REPLICATE == volinfo->type) {
+                ret = dict_get_int32 (dict, "replica-count",
+                                      &volinfo->replica_count);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                 "replica count for volume %s", volname);
+                        goto out;
+                }
+        } else if (GF_CLUSTER_TYPE_STRIPE == volinfo->type) {
+                ret = dict_get_int32 (dict, "stripe-count",
+                                      &volinfo->stripe_count);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get stripe"
+                                " count for volume %s", volname);
+                        goto out;
+                }
+        } else if (GF_CLUSTER_TYPE_STRIPE_REPLICATE == volinfo->type) {
+                ret = dict_get_int32 (dict, "stripe-count",
+                                      &volinfo->stripe_count);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get stripe"
+                                " count for volume %s", volname);
+                        goto out;
+                }
+                ret = dict_get_int32 (dict, "replica-count",
+                                      &volinfo->replica_count);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                "replica count for volume %s", volname);
+                        goto out;
+                }
+        }
+
+        /* dist-leaf-count is the count of brick nodes for a given
+           subvolume of distribute */
+        volinfo->dist_leaf_count = glusterd_get_dist_leaf_count (volinfo);
+
+        /* subvol_count is the count of number of subvolumes present
+           for a given distribute volume */
+        volinfo->subvol_count = (volinfo->brick_count /
+                                 volinfo->dist_leaf_count);
+
+        /* Keep sub-count same as earlier, for the sake of backward
+           compatibility */
+        if (volinfo->dist_leaf_count > 1)
+                volinfo->sub_count = volinfo->dist_leaf_count;
+
+        ret = dict_get_str (dict, "transport", &trans_type);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to get transport type of volume %s", volname);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "volume-id", &str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to get volume-id of volume %s", volname);
+                goto out;
+        }
+        ret = uuid_parse (str, volinfo->volume_id);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "unable to parse uuid %s of volume %s", str, volname);
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "internal-username", &username);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "unable to get internal username of volume %s",
+                        volname);
+                goto out;
+        }
+        glusterd_auth_set_username (volinfo, username);
+
+        ret = dict_get_str (dict, "internal-password", &password);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "unable to get internal password of volume %s",
+                        volname);
+                goto out;
+        }
+        glusterd_auth_set_password (volinfo, password);
+
+        if (strcasecmp (trans_type, "rdma") == 0) {
+                volinfo->transport_type = GF_TRANSPORT_RDMA;
+                volinfo->nfs_transport_type = GF_TRANSPORT_RDMA;
+        } else if (strcasecmp (trans_type, "tcp") == 0) {
+                volinfo->transport_type = GF_TRANSPORT_TCP;
+                volinfo->nfs_transport_type = GF_TRANSPORT_TCP;
+        } else {
+                volinfo->transport_type = GF_TRANSPORT_BOTH_TCP_RDMA;
+                volinfo->nfs_transport_type = GF_DEFAULT_NFS_TRANSPORT;
+        }
+
+        if (bricks) {
+                brick_list = gf_strdup (bricks);
+                free_ptr = brick_list;
+        }
+
+        if (count)
+                brick = strtok_r (brick_list+1, " \n", &saveptr);
+        caps = CAPS_BD | CAPS_THIN | CAPS_OFFLOAD_COPY | CAPS_OFFLOAD_SNAPSHOT;
+
+        while ( i <= count) {
+                ret = glusterd_brickinfo_new_from_brick (brick, &brickinfo);
+                if (ret)
+                        goto out;
+
+                ret = glusterd_resolve_brick (brickinfo);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, FMTSTR_RESOLVE_BRICK,
+                                brickinfo->hostname, brickinfo->path);
+                        goto out;
+                }
+
+#ifdef HAVE_BD_XLATOR
+                if (!uuid_compare (brickinfo->uuid, MY_UUID)) {
+                        if (brickinfo->vg[0]) {
+                                ret = glusterd_is_valid_vg (brickinfo, 0, msg);
+                                if (ret) {
+                                        gf_log (this->name, GF_LOG_ERROR, "%s",
+                                                msg);
+                                        goto out;
+                                }
+
+                                /* if anyone of the brick does not have thin
+                                   support, disable it for entire volume */
+                                caps &= brickinfo->caps;
+
+
+                        } else
+                                caps = 0;
+                }
+#endif
+
+                list_add_tail (&brickinfo->brick_list, &volinfo->bricks);
+                brick = strtok_r (NULL, " \n", &saveptr);
+                i++;
+        }
+
+        gd_update_volume_op_versions (volinfo);
+
+        volinfo->caps = caps;
+
+        ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret) {
+                glusterd_store_delete_volume (volinfo);
+                *op_errstr = gf_strdup ("Failed to store the Volume information");
+                goto out;
+        }
+
+        ret = glusterd_create_volfiles_and_notify_services (volinfo);
+        if (ret) {
+                *op_errstr = gf_strdup ("Failed to create volume files");
+                goto out;
+        }
+
+        volinfo->rebal.defrag_status = 0;
+        list_add_tail (&volinfo->vol_list, &priv->volumes);
+        vol_added = _gf_true;
+
+out:
+        GF_FREE(free_ptr);
+        if (!vol_added && volinfo)
+                glusterd_volinfo_delete (volinfo);
+        return ret;
+}
+
+int
+glusterd_op_start_volume (dict_t *dict, char **op_errstr)
+{
+        int                                     ret = 0;
+        char                                    *volname = NULL;
+        int                                     flags = 0;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        glusterd_brickinfo_t                    *brickinfo = NULL;
+        xlator_t                                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = glusterd_op_start_volume_args_get (dict, &volname, &flags);
+        if (ret)
+                goto out;
+
+        ret  = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS,
+                        volname);
+                goto out;
+        }
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                ret = glusterd_brick_start (volinfo, brickinfo, _gf_true);
+                /* If 'force' try to start all bricks regardless of success or
+                 * failure
+                 */
+                if (!(flags & GF_CLI_FLAG_OP_FORCE) && ret)
+                        goto out;
+        }
+
+        glusterd_set_volume_status (volinfo, GLUSTERD_STATUS_STARTED);
+
+        ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret)
+                goto out;
+
+        ret = glusterd_nodesvcs_handle_graph_change (volinfo);
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d ", ret);
+        return ret;
+}
+
+int
+glusterd_stop_volume (glusterd_volinfo_t *volinfo)
+{
+        int                     ret             = -1;
+        glusterd_brickinfo_t    *brickinfo      = NULL;
+        xlator_t                *this           = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        GF_VALIDATE_OR_GOTO (this->name, volinfo, out);
+
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                ret = glusterd_brick_stop (volinfo, brickinfo, _gf_false);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to stop "
+                                "brick (%s)", brickinfo->path);
+                        goto out;
+                }
+        }
+
+        glusterd_set_volume_status (volinfo, GLUSTERD_STATUS_STOPPED);
+
+        ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to store volinfo of "
+                        "%s volume", volinfo->volname);
+                goto out;
+        }
+
+        ret = glusterd_nodesvcs_handle_graph_change (volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to notify graph "
+                        "change for %s volume", volinfo->volname);
+                goto out;
+        }
+
+out:
+        return ret;
+}
+
+
+int
+glusterd_op_stop_volume (dict_t *dict)
+{
+        int                                     ret = 0;
+        int                                     flags = 0;
+        char                                    *volname = NULL;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        xlator_t                                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        ret = glusterd_op_stop_volume_args_get (dict, &volname, &flags);
+        if (ret)
+                goto out;
+
+        ret  = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS,
+                        volname);
+                goto out;
+        }
+
+        ret = glusterd_stop_volume (volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to stop %s volume",
+                        volname);
+                goto out;
+        }
+out:
+        return ret;
+}
+
+int
+glusterd_op_delete_volume (dict_t *dict)
+{
+        int                                     ret = 0;
+        char                                    *volname = NULL;
+        glusterd_conf_t                         *priv = NULL;
+        glusterd_volinfo_t                      *volinfo = NULL;
+        xlator_t                                *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret  = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, FMTSTR_CHECK_VOL_EXISTS,
+                        volname);
+                goto out;
+        }
+
+        ret = glusterd_delete_volume (volinfo);
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+int
+glusterd_op_heal_volume (dict_t *dict, char **op_errstr)
+{
+        int                                     ret = 0;
+        /* Necessary subtasks of heal are completed in brick op */
+
+        return ret;
+}
+
+int
+glusterd_op_statedump_volume (dict_t *dict, char **op_errstr)
+{
+        int                     ret = 0;
+        char                    *volname = NULL;
+        char                    *options = NULL;
+        int                     option_cnt = 0;
+        glusterd_volinfo_t      *volinfo = NULL;
+        glusterd_brickinfo_t    *brickinfo = NULL;
+
+        ret = glusterd_op_statedump_volume_args_get (dict, &volname, &options,
+                                                     &option_cnt);
+        if (ret)
+                goto out;
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret)
+                goto out;
+        gf_log ("", GF_LOG_DEBUG, "Performing statedump on volume %s", volname);
+        if (strstr (options, "nfs") != NULL) {
+                ret = glusterd_nfs_statedump (options, option_cnt, op_errstr);
+                if (ret)
+                        goto out;
+        } else {
+                list_for_each_entry (brickinfo, &volinfo->bricks,
+                                                        brick_list) {
+                        ret = glusterd_brick_statedump (volinfo, brickinfo,
+                                                        options, option_cnt,
+                                                        op_errstr);
+                        /* Let us take the statedump of other bricks instead of
+                         * exiting, if statedump of this brick fails.
+                         */
+                        if (ret)
+                                gf_log (THIS->name, GF_LOG_WARNING, "could not "
+                                        "take the statedump of the brick %s:%s."
+                                        " Proceeding to other bricks",
+                                        brickinfo->hostname, brickinfo->path);
+                }
+        }
+
+out:
+        return ret;
+}
+
+int
+glusterd_clearlocks_send_cmd (glusterd_volinfo_t *volinfo, char *cmd,
+                              char *path, char *result, char *errstr,
+                              int err_len, char *mntpt)
+{
+        int               ret                   = -1;
+        glusterd_conf_t  *priv                  = NULL;
+        char             abspath[PATH_MAX]      = {0, };
+
+        priv = THIS->private;
+
+        snprintf (abspath, sizeof (abspath), "%s/%s", mntpt, path);
+        ret = sys_lgetxattr (abspath, cmd, result, PATH_MAX);
+        if (ret < 0) {
+                snprintf (errstr, err_len, "clear-locks getxattr command "
+                          "failed. Reason: %s", strerror (errno));
+                gf_log (THIS->name, GF_LOG_DEBUG, "%s", errstr);
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_clearlocks_rmdir_mount (glusterd_volinfo_t *volinfo, char *mntpt)
+{
+        int              ret               = -1;
+        glusterd_conf_t *priv              = NULL;
+
+        priv = THIS->private;
+
+        ret = rmdir (mntpt);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_DEBUG, "rmdir failed");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+void
+glusterd_clearlocks_unmount (glusterd_volinfo_t *volinfo, char *mntpt)
+{
+        glusterd_conf_t  *priv               = NULL;
+        runner_t          runner             = {0,};
+        int               ret                = 0;
+
+        priv = THIS->private;
+
+        /*umount failures are ignored. Using stat we could have avoided
+         * attempting to unmount a non-existent filesystem. But a failure of
+         * stat() on mount can be due to network failures.*/
+
+        runinit (&runner);
+        runner_add_args (&runner, "/bin/umount", "-f", NULL);
+        runner_argprintf (&runner, "%s", mntpt);
+
+        synclock_unlock (&priv->big_lock);
+        ret = runner_run (&runner);
+        synclock_lock (&priv->big_lock);
+        if (ret) {
+                ret = 0;
+                gf_log ("", GF_LOG_DEBUG,
+                        "umount failed on maintenance client");
+        }
+
+        return;
+}
+
+int
+glusterd_clearlocks_create_mount (glusterd_volinfo_t *volinfo, char **mntpt)
+{
+        int              ret                    = -1;
+        glusterd_conf_t *priv                   = NULL;
+        char             template[PATH_MAX]     = {0,};
+        char            *tmpl                   = NULL;
+
+        priv = THIS->private;
+
+        snprintf (template, sizeof (template), "/tmp/%s.XXXXXX",
+                  volinfo->volname);
+        tmpl = mkdtemp (template);
+        if (!tmpl) {
+                gf_log (THIS->name, GF_LOG_DEBUG, "Couldn't create temporary "
+                        "mount directory. Reason %s", strerror (errno));
+                goto out;
+        }
+
+        *mntpt = gf_strdup (tmpl);
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_clearlocks_mount (glusterd_volinfo_t *volinfo, char **xl_opts,
+                           char *mntpt)
+{
+        int             ret                             = -1;
+        int             i                               = 0;
+        glusterd_conf_t *priv                           = NULL;
+        runner_t        runner                          = {0,};
+        char            client_volfpath[PATH_MAX]       = {0,};
+        char            self_heal_opts[3][1024]      = {"*replicate*.data-self-heal=off",
+                                                        "*replicate*.metadata-self-heal=off",
+                                                        "*replicate*.entry-self-heal=off"};
+
+        priv = THIS->private;
+
+        runinit (&runner);
+        glusterd_get_trusted_client_filepath (client_volfpath, volinfo,
+                                      volinfo->transport_type);
+        runner_add_args (&runner, SBIN_DIR"/glusterfs", "-f", NULL);
+        runner_argprintf (&runner, "%s", client_volfpath);
+        runner_add_arg (&runner, "-l");
+        runner_argprintf (&runner, DEFAULT_LOG_FILE_DIRECTORY
+                          "/%s-clearlocks-mnt.log", volinfo->volname);
+        if (volinfo->memory_accounting)
+                runner_add_arg (&runner, "--mem-accounting");
+
+        for (i = 0; i < volinfo->brick_count && xl_opts[i]; i++) {
+                runner_add_arg (&runner, "--xlator-option");
+                runner_argprintf (&runner, "%s", xl_opts[i]);
+        }
+
+        for (i = 0; i < 3; i++) {
+                runner_add_args (&runner, "--xlator-option",
+                                 self_heal_opts[i], NULL);
+        }
+
+        runner_argprintf (&runner, "%s", mntpt);
+        synclock_unlock (&priv->big_lock);
+        ret = runner_run (&runner);
+        synclock_lock (&priv->big_lock);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_DEBUG,
+                        "Could not start glusterfs");
+                goto out;
+        }
+        gf_log (THIS->name, GF_LOG_DEBUG,
+                "Started glusterfs successfully");
+
+out:
+        return ret;
+}
+
+int
+glusterd_clearlocks_get_local_client_ports (glusterd_volinfo_t *volinfo,
+                                            char **xl_opts)
+{
+        glusterd_brickinfo_t    *brickinfo      = NULL;
+        glusterd_conf_t         *priv           = NULL;
+        int                     index           = 0;
+        int                     ret             = -1;
+        int                     i               = 0;
+        int                     port            = 0;
+
+        GF_ASSERT (xl_opts);
+        if (!xl_opts) {
+                gf_log (THIS->name, GF_LOG_DEBUG, "Should pass non-NULL "
+                        "xl_opts");
+                goto out;
+        }
+
+        priv = THIS->private;
+
+        index = -1;
+        list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) {
+                index++;
+                if (uuid_compare (brickinfo->uuid, MY_UUID))
+                        continue;
+
+                port = pmap_registry_search (THIS, brickinfo->path,
+                                             GF_PMAP_PORT_BRICKSERVER);
+                if (!port) {
+                        ret = -1;
+                        gf_log (THIS->name, GF_LOG_DEBUG, "Couldn't get port "
+                                " for brick %s:%s", brickinfo->hostname,
+                                brickinfo->path);
+                        goto out;
+                }
+
+                ret = gf_asprintf (&xl_opts[i], "%s-client-%d.remote-port=%d",
+                                   volinfo->volname, index, port);
+                if (ret == -1) {
+                        xl_opts[i] = NULL;
+                        goto out;
+                }
+                i++;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+glusterd_op_clearlocks_volume (dict_t *dict, char **op_errstr, dict_t *rsp_dict)
+{
+        int32_t                         ret                 = -1;
+        int                             i                   = 0;
+        char                            *volname            = NULL;
+        char                            *path               = NULL;
+        char                            *kind               = NULL;
+        char                            *type               = NULL;
+        char                            *opts               = NULL;
+        char                            *cmd_str            = NULL;
+        char                            *free_ptr           = NULL;
+        char                            msg[PATH_MAX]       = {0,};
+        char                            result[PATH_MAX]    = {0,};
+        char                            *mntpt              = NULL;
+        char                            **xl_opts           = NULL;
+        glusterd_volinfo_t              *volinfo            = NULL;
+
+        ret = dict_get_str (dict, "volname", &volname);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to get volume name");
+                goto out;
+        }
+        gf_log ("", GF_LOG_DEBUG, "Performing clearlocks on volume %s", volname);
+
+        ret = dict_get_str (dict, "path", &path);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to get path");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "kind", &kind);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to get kind");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "type", &type);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "Failed to get type");
+                goto out;
+        }
+
+        ret = dict_get_str (dict, "opts", &opts);
+        if (ret)
+                ret = 0;
+
+        gf_log (THIS->name, GF_LOG_INFO, "Received clear-locks request for "
+                "volume %s with kind %s type %s and options %s", volname,
+                kind, type, opts);
+
+        if (opts)
+                ret = gf_asprintf (&cmd_str, GF_XATTR_CLRLK_CMD".t%s.k%s.%s",
+                                   type, kind, opts);
+        else
+                ret = gf_asprintf (&cmd_str, GF_XATTR_CLRLK_CMD".t%s.k%s",
+                                   type, kind);
+        if (ret == -1)
+                goto out;
+
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Volume %s doesn't exist.",
+                          volname);
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        xl_opts = GF_CALLOC (volinfo->brick_count+1, sizeof (char*),
+                             gf_gld_mt_charptr);
+        if (!xl_opts)
+                goto out;
+
+        ret = glusterd_clearlocks_get_local_client_ports (volinfo, xl_opts);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Couldn't get port numbers of "
+                          "local bricks");
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        ret = glusterd_clearlocks_create_mount (volinfo, &mntpt);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Creating mount directory "
+                          "for clear-locks failed.");
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        ret = glusterd_clearlocks_mount (volinfo, xl_opts, mntpt);
+        if (ret) {
+                snprintf (msg, sizeof (msg), "Failed to mount clear-locks "
+                          "maintenance client.");
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                goto out;
+        }
+
+        ret = glusterd_clearlocks_send_cmd (volinfo, cmd_str, path, result,
+                                            msg, sizeof (msg), mntpt);
+        if (ret) {
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+                goto umount;
+        }
+
+        free_ptr = gf_strdup(result);
+        if (dict_set_dynstr (rsp_dict, "lk-summary", free_ptr)) {
+                GF_FREE (free_ptr);
+                snprintf (msg, sizeof (msg), "Failed to set clear-locks "
+                          "result");
+                gf_log (THIS->name, GF_LOG_ERROR, "%s", msg);
+        }
+
+umount:
+        glusterd_clearlocks_unmount (volinfo, mntpt);
+
+        if (glusterd_clearlocks_rmdir_mount (volinfo, mntpt))
+                gf_log (THIS->name, GF_LOG_WARNING, "Couldn't unmount "
+                        "clear-locks mount point");
+
+out:
+        if (ret)
+                *op_errstr = gf_strdup (msg);
+
+        if (xl_opts) {
+                for (i = 0; i < volinfo->brick_count && xl_opts[i]; i++)
+                        GF_FREE (xl_opts[i]);
+                GF_FREE (xl_opts);
+        }
+
+        GF_FREE (cmd_str);
+
+        GF_FREE (mntpt);
+
+        return ret;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-set.c b/xlators/mgmt/glusterd/src/glusterd-volume-set.c
new file mode 100644
index 000000000..665a8b298
--- /dev/null
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-set.c
@@ -0,0 +1,1452 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterd-volgen.h"
+#include "glusterd-utils.h"
+
+static int
+check_dict_key_value (dict_t *dict, char *key, char *value)
+{
+        glusterd_conf_t     *priv          = NULL;
+        int                  ret           = 0;
+        xlator_t            *this          = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        if (!dict) {
+                gf_log (this->name, GF_LOG_ERROR, "Received Empty Dict.");
+                ret = -1;
+                goto out;
+        }
+
+        if (!key) {
+                gf_log (this->name, GF_LOG_ERROR, "Received Empty Key.");
+                ret = -1;
+                goto out;
+        }
+
+        if (!value) {
+                gf_log (this->name, GF_LOG_ERROR, "Received Empty Value.");
+                ret = -1;
+                goto out;
+        }
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+static int
+get_volname_volinfo (dict_t *dict, char **volname, glusterd_volinfo_t **volinfo)
+{
+        glusterd_conf_t     *priv          = NULL;
+        int                  ret           = 0;
+        xlator_t            *this          = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = dict_get_str (dict, "volname", volname);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name");
+                goto out;
+        }
+
+        ret = glusterd_volinfo_find (*volname, volinfo);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to allocate memory");
+                goto out;
+        }
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+static int
+validate_cache_max_min_size (dict_t *dict, char *key, char *value,
+                             char **op_errstr)
+{
+        char                *current_max_value = NULL;
+        char                *current_min_value = NULL;
+        char                 errstr[2048]  = "";
+        char                *volname       = NULL;
+        glusterd_conf_t     *priv          = NULL;
+        glusterd_volinfo_t  *volinfo       = NULL;
+        int                  ret           = 0;
+        uint64_t             max_value     = 0;
+        uint64_t             min_value     = 0;
+        xlator_t            *this          = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = check_dict_key_value (dict, key, value);
+        if (ret)
+                goto out;
+
+        ret = get_volname_volinfo (dict, &volname, &volinfo);
+        if (ret)
+                goto out;
+
+        if ((!strcmp (key, "performance.cache-min-file-size")) ||
+            (!strcmp (key, "cache-min-file-size"))) {
+                glusterd_volinfo_get (volinfo,
+                                      "performance.cache-max-file-size",
+                                      &current_max_value);
+                if (current_max_value) {
+                        gf_string2bytesize (current_max_value, &max_value);
+                        gf_string2bytesize (value, &min_value);
+                        current_min_value = value;
+                }
+        } else  if ((!strcmp (key, "performance.cache-max-file-size")) ||
+                    (!strcmp (key, "cache-max-file-size"))) {
+                glusterd_volinfo_get (volinfo,
+                                      "performance.cache-min-file-size",
+                                      &current_min_value);
+                if (current_min_value) {
+                        gf_string2bytesize (current_min_value, &min_value);
+                        gf_string2bytesize (value, &max_value);
+                        current_max_value = value;
+                }
+        }
+
+        if (min_value > max_value) {
+                snprintf (errstr, sizeof (errstr),
+                          "cache-min-file-size (%s) is greater than "
+                          "cache-max-file-size (%s)",
+                          current_min_value, current_max_value);
+                gf_log (this->name, GF_LOG_ERROR, "%s", errstr);
+                *op_errstr = gf_strdup (errstr);
+                ret = -1;
+                goto out;
+        }
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+static int
+validate_quota (dict_t *dict, char *key, char *value,
+                char **op_errstr)
+{
+        char                 errstr[2048] = "";
+        char                *volname      = NULL;
+        glusterd_conf_t     *priv         = NULL;
+        glusterd_volinfo_t  *volinfo      = NULL;
+        int                  ret          = 0;
+        xlator_t            *this         = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = check_dict_key_value (dict, key, value);
+        if (ret)
+                goto out;
+
+        ret = get_volname_volinfo (dict, &volname, &volinfo);
+        if (ret)
+                goto out;
+
+        ret = glusterd_volinfo_get_boolean (volinfo, VKEY_FEATURES_QUOTA);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get the quota status");
+                goto out;
+        }
+
+        if (ret == _gf_false) {
+                snprintf (errstr, sizeof (errstr),
+                          "Cannot set %s. Enable quota first.", key);
+                gf_log (this->name, GF_LOG_ERROR, "%s", errstr);
+                *op_errstr = gf_strdup (errstr);
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+static int
+validate_stripe (dict_t *dict, char *key, char *value, char **op_errstr)
+{
+        char                 errstr[2048]  = "";
+        char                *volname       = NULL;
+        glusterd_conf_t     *priv          = NULL;
+        glusterd_volinfo_t  *volinfo       = NULL;
+        int                  ret           = 0;
+        xlator_t            *this          = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = check_dict_key_value (dict, key, value);
+        if (ret)
+                goto out;
+
+        ret = get_volname_volinfo (dict, &volname, &volinfo);
+        if (ret)
+                goto out;
+
+        if (volinfo->stripe_count == 1) {
+                snprintf (errstr, sizeof (errstr),
+                          "Cannot set %s for a non-stripe volume.", key);
+                gf_log (this->name, GF_LOG_ERROR, "%s", errstr);
+                *op_errstr = gf_strdup (errstr);
+                ret = -1;
+               goto out;
+        }
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+static int
+validate_subvols_per_directory (dict_t *dict, char *key, char *value,
+                                char **op_errstr)
+{
+        char                 errstr[2048]  = "";
+        char                *volname       = NULL;
+        glusterd_conf_t     *priv          = NULL;
+        glusterd_volinfo_t  *volinfo       = NULL;
+        int                  ret           = 0;
+        int                  subvols       = 0;
+        xlator_t            *this          = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        ret = check_dict_key_value (dict, key, value);
+        if (ret)
+                goto out;
+
+        ret = get_volname_volinfo (dict, &volname, &volinfo);
+        if (ret)
+                goto out;
+
+        subvols = atoi(value);
+
+        /* Checking if the subvols-per-directory exceed the total
+           number of subvolumes. */
+        if (subvols > volinfo->subvol_count) {
+                snprintf (errstr, sizeof(errstr),
+                          "subvols-per-directory(%d) is greater "
+                          "than the number of subvolumes(%d).",
+                          subvols, volinfo->subvol_count);
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s.", errstr);
+                *op_errstr = gf_strdup (errstr);
+                ret = -1;
+                goto out;
+        }
+
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret);
+
+        return ret;
+}
+
+
+/* dispatch table for VOLUME SET
+ * -----------------------------
+ *
+ * Format of entries:
+ *
+ * First field is the <key>, for the purpose of looking it up
+ * in volume dictionary. Each <key> is of the format "<domain>.<specifier>".
+ *
+ * Second field is <voltype>.
+ *
+ * Third field is <option>, if its unset, it's assumed to be
+ * the same as <specifier>.
+ *
+ * Fourth field is <value>. In this context they are used to specify
+ * a default. That is, even the volume dict doesn't have a value,
+ * we procced as if the default value were set for it.
+ *
+ * Fifth field is <doctype>, which decides if the option is public and available
+ * in "set help" or not. "NO_DOC" entries are not part of the public interface
+ * and are subject to change at any time. This also decides if an option is
+ * global (apllies to all volumes) or normal (applies to only specified volume).
+ *
+ * Sixth field is <flags>.
+ *
+ * Seventh field is <op-version>.
+ *
+ * Eight field is description of option: If NULL, tried to fetch from
+ * translator code's xlator_options table.
+ *
+ * Nineth field is validation function: If NULL, xlator's option specific
+ * validation will be tried, otherwise tried at glusterd code itself.
+ *
+ * There are two type of entries: basic and special.
+ *
+ * - Basic entries are the ones where the <option> does _not_ start with
+ *   the bang! character ('!').
+ *
+ *   In their case, <option> is understood as an option for an xlator of
+ *   type <voltype>. Their effect is to copy over the volinfo->dict[<key>]
+ *   value to all graph nodes of type <voltype> (if such a value is set).
+ *
+ *   You are free to add entries of this type, they will become functional
+ *   just by being present in the table.
+ *
+ * - Special entries where the <option> starts with the bang!.
+ *
+ *   They are not applied to all graphs during generation, and you cannot
+ *   extend them in a trivial way which could be just picked up. Better
+ *   not touch them unless you know what you do.
+ *
+ *
+ * Another kind of grouping for options, according to visibility:
+ *
+ * - Exported: one which is used in the code. These are characterized by
+ *   being used a macro as <key> (of the format VKEY_..., defined in
+ *   glusterd-volgen.h
+ *
+ * - Non-exported: the rest; these have string literal <keys>.
+ *
+ * Adhering to this policy, option name changes shall be one-liners.
+ *
+ */
+
+struct volopt_map_entry glusterd_volopt_map[] = {
+        /* DHT xlator options */
+        { .key        = "cluster.lookup-unhashed",
+          .voltype    = "cluster/distribute",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.min-free-disk",
+          .voltype    = "cluster/distribute",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.min-free-inodes",
+          .voltype    = "cluster/distribute",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.rebalance-stats",
+          .voltype    = "cluster/distribute",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key         = "cluster.subvols-per-directory",
+          .voltype     = "cluster/distribute",
+          .option      = "directory-layout-spread",
+          .op_version  = 2,
+          .validate_fn = validate_subvols_per_directory,
+          .flags       = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.readdir-optimize",
+          .voltype    = "cluster/distribute",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.rsync-hash-regex",
+          .voltype    = "cluster/distribute",
+          .type       = NO_DOC,
+          .op_version = 3,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.extra-hash-regex",
+          .voltype    = "cluster/distribute",
+          .type       = NO_DOC,
+          .op_version = 3,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.dht-xattr-name",
+          .voltype    = "cluster/distribute",
+          .option     = "xattr-name",
+          .type       = NO_DOC,
+          .op_version = 3,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+
+        /* NUFA xlator options (Distribute special case) */
+        { .key        = "cluster.nufa",
+          .voltype    = "cluster/distribute",
+          .option     = "!nufa",
+          .type       = NO_DOC,
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.local-volume-name",
+          .voltype    = "cluster/nufa",
+          .option     = "local-volume-name",
+          .type       = NO_DOC,
+          .op_version = 3,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+
+        /* Switch xlator options (Distribute special case) */
+        { .key        = "cluster.switch",
+          .voltype    = "cluster/distribute",
+          .option     = "!switch",
+          .type       = NO_DOC,
+          .op_version = 3,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.switch-pattern",
+          .voltype    = "cluster/switch",
+          .option     = "pattern.switch.case",
+          .type       = NO_DOC,
+          .op_version = 3,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+
+        /* AFR xlator options */
+        { .key        = "cluster.entry-change-log",
+          .voltype    = "cluster/replicate",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.read-subvolume",
+          .voltype    = "cluster/replicate",
+          .op_version = 1,
+          .flags     = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.read-subvolume-index",
+          .voltype    = "cluster/replicate",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.read-hash-mode",
+          .voltype    = "cluster/replicate",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.background-self-heal-count",
+          .voltype    = "cluster/replicate",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.metadata-self-heal",
+          .voltype    = "cluster/replicate",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.data-self-heal",
+          .voltype    = "cluster/replicate",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.entry-self-heal",
+          .voltype    = "cluster/replicate",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key           = "cluster.self-heal-daemon",
+          .voltype       = "cluster/replicate",
+          .option        = "!self-heal-daemon",
+          .op_version    = 1
+        },
+        { .key        = "cluster.heal-timeout",
+          .voltype    = "cluster/replicate",
+          .option     = "!heal-timeout",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.strict-readdir",
+          .voltype    = "cluster/replicate",
+          .type       = NO_DOC,
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.self-heal-window-size",
+          .voltype    = "cluster/replicate",
+          .option     = "data-self-heal-window-size",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.data-change-log",
+          .voltype    = "cluster/replicate",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.metadata-change-log",
+          .voltype    = "cluster/replicate",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.data-self-heal-algorithm",
+          .voltype    = "cluster/replicate",
+          .option     = "data-self-heal-algorithm",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.eager-lock",
+          .voltype    = "cluster/replicate",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.quorum-type",
+          .voltype    = "cluster/replicate",
+          .option     = "quorum-type",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.quorum-count",
+          .voltype    = "cluster/replicate",
+          .option     = "quorum-count",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.choose-local",
+          .voltype    = "cluster/replicate",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.self-heal-readdir-size",
+          .voltype    = "cluster/replicate",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.post-op-delay-secs",
+          .voltype    = "cluster/replicate",
+          .type       = NO_DOC,
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.readdir-failover",
+          .voltype    = "cluster/replicate",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.ensure-durability",
+          .voltype    = "cluster/replicate",
+          .op_version = 3,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+
+        /* Stripe xlator options */
+        { .key         = "cluster.stripe-block-size",
+          .voltype     = "cluster/stripe",
+          .option      = "block-size",
+          .op_version  = 1,
+          .validate_fn = validate_stripe,
+          .flags       = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "cluster.stripe-coalesce",
+          .voltype    = "cluster/stripe",
+          .option     = "coalesce",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+
+        /* IO-stats xlator options */
+        { .key         = VKEY_DIAG_LAT_MEASUREMENT,
+          .voltype     = "debug/io-stats",
+          .option      = "latency-measurement",
+          .value       = "off",
+          .op_version  = 1
+        },
+        { .key         = "diagnostics.dump-fd-stats",
+          .voltype     = "debug/io-stats",
+          .op_version  = 1
+        },
+        { .key         = VKEY_DIAG_CNT_FOP_HITS,
+          .voltype     = "debug/io-stats",
+          .option      = "count-fop-hits",
+          .value       = "off",
+          .type        = NO_DOC,
+          .op_version  = 1
+        },
+        { .key         = "diagnostics.brick-log-level",
+          .voltype     = "debug/io-stats",
+          .option      = "!brick-log-level",
+          .op_version  = 1
+        },
+        { .key        = "diagnostics.client-log-level",
+          .voltype    = "debug/io-stats",
+          .option     = "!client-log-level",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key         = "diagnostics.brick-sys-log-level",
+          .voltype     = "debug/io-stats",
+          .option      = "!sys-log-level",
+          .op_version  = 1
+        },
+        { .key        = "diagnostics.client-sys-log-level",
+          .voltype    = "debug/io-stats",
+          .option     = "!sys-log-level",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+
+        /* IO-cache xlator options */
+        { .key         = "performance.cache-max-file-size",
+          .voltype     = "performance/io-cache",
+          .option      = "max-file-size",
+          .op_version  = 1,
+          .validate_fn = validate_cache_max_min_size,
+          .flags       = OPT_FLAG_CLIENT_OPT
+        },
+        { .key         = "performance.cache-min-file-size",
+          .voltype     = "performance/io-cache",
+          .option      = "min-file-size",
+          .op_version  = 1,
+          .validate_fn = validate_cache_max_min_size,
+          .flags       = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "performance.cache-refresh-timeout",
+          .voltype    = "performance/io-cache",
+          .option     = "cache-timeout",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "performance.cache-priority",
+          .voltype    = "performance/io-cache",
+          .option     = "priority",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "performance.cache-size",
+          .voltype    = "performance/io-cache",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+
+        /* IO-threads xlator options */
+        { .key         = "performance.io-thread-count",
+          .voltype     = "performance/io-threads",
+          .option      = "thread-count",
+          .op_version  = 1
+        },
+        { .key         = "performance.high-prio-threads",
+          .voltype     = "performance/io-threads",
+          .op_version  = 1
+        },
+        { .key         = "performance.normal-prio-threads",
+          .voltype     = "performance/io-threads",
+          .op_version  = 1
+        },
+        { .key         = "performance.low-prio-threads",
+          .voltype     = "performance/io-threads",
+          .op_version  = 1
+        },
+        { .key         = "performance.least-prio-threads",
+          .voltype     = "performance/io-threads",
+          .op_version  = 1
+        },
+        { .key         = "performance.enable-least-priority",
+          .voltype     = "performance/io-threads",
+          .op_version  = 1
+        },
+        { .key         = "performance.least-rate-limit",
+          .voltype     = "performance/io-threads",
+          .op_version  = 2
+        },
+
+        /* Other perf xlators' options */
+        { .key        = "performance.cache-size",
+          .voltype    = "performance/quick-read",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "performance.flush-behind",
+          .voltype    = "performance/write-behind",
+          .option     = "flush-behind",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "performance.write-behind-window-size",
+          .voltype    = "performance/write-behind",
+          .option     = "cache-size",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "performance.strict-o-direct",
+          .voltype    = "performance/write-behind",
+          .option     = "strict-O_DIRECT",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "performance.strict-write-ordering",
+          .voltype    = "performance/write-behind",
+          .option     = "strict-write-ordering",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "performance.lazy-open",
+          .voltype    = "performance/open-behind",
+          .option     = "lazy-open",
+          .op_version = 3,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "performance.read-ahead-page-count",
+          .voltype    = "performance/read-ahead",
+          .option     = "page-count",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "performance.md-cache-timeout",
+          .voltype    = "performance/md-cache",
+          .option     = "md-cache-timeout",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+
+ 	/* Crypt xlator options */
+
+	{ .key         = "features.encryption",
+	  .voltype     = "encryption/crypt",
+	  .option      = "!feat",
+	  .value       = "off",
+	  .op_version  = 3,
+	  .description = "enable/disable client-side encryption for "
+                         "the volume.",
+	  .flags       = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT
+	},
+
+        { .key         = "encryption.master-key",
+          .voltype     = "encryption/crypt",
+          .op_version  = 3,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key         = "encryption.data-key-size",
+          .voltype     = "encryption/crypt",
+          .op_version  = 3,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key         = "encryption.block-size",
+          .voltype     = "encryption/crypt",
+          .op_version  = 3,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+
+        /* Client xlator options */
+        { .key        = "network.frame-timeout",
+          .voltype    = "protocol/client",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "network.ping-timeout",
+          .voltype    = "protocol/client",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "network.tcp-window-size",
+          .voltype    = "protocol/client",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "features.lock-heal",
+          .voltype    = "protocol/client",
+          .option     = "lk-heal",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "features.grace-timeout",
+          .voltype    = "protocol/client",
+          .option     = "grace-timeout",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "client.ssl",
+          .voltype    = "protocol/client",
+          .option     = "transport.socket.ssl-enabled",
+          .type       = NO_DOC,
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key        = "network.remote-dio",
+          .voltype    = "protocol/client",
+          .option     = "filter-O_DIRECT",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+
+        /* Server xlator options */
+        { .key         = "network.tcp-window-size",
+          .voltype     = "protocol/server",
+          .op_version  = 1
+        },
+        { .key         = "network.inode-lru-limit",
+          .voltype     = "protocol/server",
+          .op_version  = 1
+        },
+        { .key         = AUTH_ALLOW_MAP_KEY,
+          .voltype     = "protocol/server",
+          .option      = "!server-auth",
+          .value       = "*",
+          .op_version  = 1
+        },
+        { .key         = AUTH_REJECT_MAP_KEY,
+          .voltype     = "protocol/server",
+          .option      = "!server-auth",
+          .op_version  = 1
+        },
+        { .key         = "transport.keepalive",
+          .voltype     = "protocol/server",
+          .option      = "transport.socket.keepalive",
+          .type        = NO_DOC,
+          .op_version  = 1
+        },
+        { .key         = "server.allow-insecure",
+          .voltype     = "protocol/server",
+          .option      = "rpc-auth-allow-insecure",
+          .type        = NO_DOC,
+          .op_version  = 1
+        },
+        { .key         = "server.root-squash",
+          .voltype     = "protocol/server",
+          .option      = "root-squash",
+          .op_version  = 2
+        },
+        { .key         = "server.statedump-path",
+          .voltype     = "protocol/server",
+          .option      = "statedump-path",
+          .op_version  = 1
+        },
+        { .key         = "server.outstanding-rpc-limit",
+          .voltype     = "protocol/server",
+          .option      = "rpc.outstanding-rpc-limit",
+          .type        = GLOBAL_DOC,
+          .op_version  = 3
+        },
+        { .key         = "features.lock-heal",
+          .voltype     = "protocol/server",
+          .option      = "lk-heal",
+          .type        = NO_DOC,
+          .op_version  = 1
+        },
+        { .key         = "features.grace-timeout",
+          .voltype     = "protocol/server",
+          .option      = "grace-timeout",
+          .type        = NO_DOC,
+          .op_version  = 1
+        },
+        { .key         = "server.ssl",
+          .voltype     = "protocol/server",
+          .option      = "transport.socket.ssl-enabled",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+
+        /* Performance xlators enable/disbable options */
+        { .key         = "performance.write-behind",
+          .voltype     = "performance/write-behind",
+          .option      = "!perf",
+          .value       = "on",
+          .op_version  = 1,
+          .description = "enable/disable write-behind translator in the "
+                         "volume.",
+          .flags       = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT
+        },
+        { .key         = "performance.read-ahead",
+          .voltype     = "performance/read-ahead",
+          .option      = "!perf",
+          .value       = "on",
+          .op_version  = 1,
+          .description = "enable/disable read-ahead translator in the volume.",
+          .flags       = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT
+        },
+        { .key         = "performance.readdir-ahead",
+          .voltype     = "performance/readdir-ahead",
+          .option      = "!perf",
+          .value       = "off",
+          .op_version  = 3,
+          .description = "enable/disable readdir-ahead translator in the volume.",
+          .flags       = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT
+        },
+
+        { .key         = "performance.io-cache",
+          .voltype     = "performance/io-cache",
+          .option      = "!perf",
+          .value       = "on",
+          .op_version  = 1,
+          .description = "enable/disable io-cache translator in the volume.",
+          .flags       = OPT_FLAG_CLIENT_OPT
+        },
+        { .key         = "performance.quick-read",
+          .voltype     = "performance/quick-read",
+          .option      = "!perf",
+          .value       = "on",
+          .op_version  = 1,
+          .description = "enable/disable quick-read translator in the volume.",
+          .flags       = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT
+
+        },
+        { .key         = "performance.open-behind",
+          .voltype     = "performance/open-behind",
+          .option      = "!perf",
+          .value       = "on",
+          .op_version  = 2,
+          .description = "enable/disable open-behind translator in the volume.",
+          .flags       = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT
+
+        },
+        { .key         = "performance.stat-prefetch",
+          .voltype     = "performance/md-cache",
+          .option      = "!perf",
+          .value       = "on",
+          .op_version  = 1,
+          .description = "enable/disable meta-data caching translator in the "
+                         "volume.",
+          .flags       = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT
+        },
+        { .key         = "performance.client-io-threads",
+          .voltype     = "performance/io-threads",
+          .option      = "!perf",
+          .value       = "off",
+          .op_version  = 1,
+          .description = "enable/disable io-threads translator in the client "
+                         "graph of volume.",
+          .flags       = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT
+        },
+        { .key        = "performance.nfs.write-behind",
+          .voltype    = "performance/write-behind",
+          .option     = "!nfsperf",
+          .value      = "on",
+          .type       = NO_DOC,
+          .op_version = 1,
+          .flags      = OPT_FLAG_XLATOR_OPT
+        },
+        { .key        = "performance.nfs.read-ahead",
+          .voltype    = "performance/read-ahead",
+          .option     = "!nfsperf",
+          .value      = "off",
+          .type       = NO_DOC,
+          .op_version = 1,
+          .flags      = OPT_FLAG_XLATOR_OPT
+        },
+        { .key        = "performance.nfs.io-cache",
+          .voltype    = "performance/io-cache",
+          .option     = "!nfsperf",
+          .value      = "off",
+          .type       = NO_DOC,
+          .op_version = 1,
+          .flags      = OPT_FLAG_XLATOR_OPT
+        },
+        { .key        = "performance.nfs.quick-read",
+          .voltype    = "performance/quick-read",
+          .option     = "!nfsperf",
+          .value      = "off",
+          .type       = NO_DOC,
+          .op_version = 1,
+          .flags      = OPT_FLAG_XLATOR_OPT
+        },
+        { .key        = "performance.nfs.stat-prefetch",
+          .voltype    = "performance/md-cache",
+          .option     = "!nfsperf",
+          .value      = "off",
+          .type       = NO_DOC,
+          .op_version = 1,
+          .flags      = OPT_FLAG_XLATOR_OPT
+        },
+        { .key        = "performance.nfs.io-threads",
+          .voltype    = "performance/io-threads",
+          .option     = "!nfsperf",
+          .value      = "off",
+          .type       = NO_DOC,
+          .op_version = 1,
+          .flags      = OPT_FLAG_XLATOR_OPT
+        },
+        { .key        = "performance.force-readdirp",
+          .voltype    = "performance/md-cache",
+          .option     = "force-readdirp",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+
+	/* Feature translators */
+        { .key         = "features.file-snapshot",
+          .voltype     = "features/qemu-block",
+          .option      = "!feat",
+          .value       = "off",
+          .op_version  = 3,
+          .description = "enable/disable file-snapshot feature in the "
+                         "volume.",
+          .flags       = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT
+        },
+
+#ifdef HAVE_LIB_Z
+        /* Compressor-decompressor xlator options
+         * defaults used from xlator/feature/compress/src/cdc.h
+         */
+        { .key         = "features.compress",
+          .voltype     = "features/cdc",
+          .option      = "!compress",
+          .value       = "off",
+          .type        = NO_DOC,
+          .op_version  = 2,
+          .description = "enable/disable compression translator"
+        },
+        { .key         = "compress.mode",
+          .voltype     = "features/cdc",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "compress.window-size",
+          .voltype     = "features/cdc",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "compress.mem-level",
+          .voltype     = "features/cdc",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "compress.min-size",
+          .voltype     = "features/cdc",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "compress.compression-level",
+          .voltype     = "features/cdc",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "compress.debug",
+          .voltype     = "features/cdc",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+ #endif
+
+        /* Quota xlator options */
+        { .key        = VKEY_FEATURES_LIMIT_USAGE,
+          .voltype    = "features/quota",
+          .option     = "limit-set",
+          .type       = NO_DOC,
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key         = "features.quota-timeout",
+          .voltype     = "features/quota",
+          .option      = "timeout",
+          .value       = "0",
+          .op_version  = 1,
+          .validate_fn = validate_quota,
+          .flags       = OPT_FLAG_CLIENT_OPT
+        },
+        { .key         = "features.quota-deem-statfs",
+          .voltype     = "features/quota",
+          .option      = "deem-statfs",
+          .value       = "off",
+          .type        = DOC,
+          .op_version  = 3,
+          .validate_fn = validate_quota,
+          .flags       = OPT_FLAG_CLIENT_OPT
+        },
+
+        /* Marker xlator options */
+        { .key         = VKEY_MARKER_XTIME,
+          .voltype     = "features/marker",
+          .option      = "xtime",
+          .value       = "off",
+          .type        = NO_DOC,
+          .flags       = OPT_FLAG_FORCE,
+          .op_version  = 1
+        },
+        { .key         = VKEY_MARKER_XTIME,
+          .voltype     = "features/marker",
+          .option      = "!xtime",
+          .value       = "off",
+          .type        = NO_DOC,
+          .flags       = OPT_FLAG_FORCE,
+          .op_version  = 1
+        },
+        { .key         = VKEY_MARKER_XTIME_FORCE,
+          .voltype     = "features/marker",
+          .option      = "gsync-force-xtime",
+          .value       = "off",
+          .type        = NO_DOC,
+          .flags       = OPT_FLAG_FORCE,
+          .op_version  = 2
+        },
+        { .key         = VKEY_MARKER_XTIME_FORCE,
+          .voltype     = "features/marker",
+          .option      = "!gsync-force-xtime",
+          .value       = "off",
+          .type        = NO_DOC,
+          .flags       = OPT_FLAG_FORCE,
+          .op_version  = 2
+        },
+        { .key         = VKEY_FEATURES_QUOTA,
+          .voltype     = "features/marker",
+          .option      = "quota",
+          .value       = "off",
+          .type        = NO_DOC,
+          .flags       = OPT_FLAG_FORCE,
+          .op_version  = 1
+        },
+
+        /* Debug xlators options */
+        { .key        = "debug.trace",
+          .voltype    = "debug/trace",
+          .option     = "!debug",
+          .value      = "off",
+          .type       = NO_DOC,
+          .op_version = 1,
+          .flags      = OPT_FLAG_XLATOR_OPT
+        },
+        { .key         = "debug.log-history",
+          .voltype     = "debug/trace",
+          .option      = "log-history",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "debug.log-file",
+          .voltype     = "debug/trace",
+          .option      = "log-file",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "debug.exclude-ops",
+          .voltype     = "debug/trace",
+          .option      = "exclude-ops",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "debug.include-ops",
+          .voltype     = "debug/trace",
+          .option      = "include-ops",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key        = "debug.error-gen",
+          .voltype    = "debug/error-gen",
+          .option     = "!debug",
+          .value      = "off",
+          .type       = NO_DOC,
+          .op_version = 1,
+          .flags      = OPT_FLAG_XLATOR_OPT
+        },
+        { .key         = "debug.error-failure",
+          .voltype     = "debug/error-gen",
+          .option      = "failure",
+          .type        = NO_DOC,
+          .op_version  = 3
+        },
+        { .key         = "debug.error-number",
+          .voltype     = "debug/error-gen",
+          .option      = "error-no",
+          .type        = NO_DOC,
+          .op_version  = 3
+        },
+        { .key         = "debug.random-failure",
+          .voltype     = "debug/error-gen",
+          .option      = "random-failure",
+          .type        = NO_DOC,
+          .op_version  = 3
+        },
+        { .key         = "debug.error-fops",
+          .voltype     = "debug/error-gen",
+          .option      = "enable",
+          .type        = NO_DOC,
+          .op_version  = 3
+        },
+
+
+        /* NFS xlator options */
+        { .key         = "nfs.enable-ino32",
+          .voltype     = "nfs/server",
+          .option      = "nfs.enable-ino32",
+          .type        = GLOBAL_DOC,
+          .op_version  = 1
+        },
+        { .key         = "nfs.mem-factor",
+          .voltype     = "nfs/server",
+          .option      = "nfs.mem-factor",
+          .type        = GLOBAL_DOC,
+          .op_version  = 1
+        },
+        { .key         = "nfs.export-dirs",
+          .voltype     = "nfs/server",
+          .option      = "nfs3.export-dirs",
+          .type        = GLOBAL_DOC,
+          .op_version  = 1
+        },
+        { .key         = "nfs.export-volumes",
+          .voltype     = "nfs/server",
+          .option      = "nfs3.export-volumes",
+          .type        = GLOBAL_DOC,
+          .op_version  = 1
+        },
+        { .key         = "nfs.addr-namelookup",
+          .voltype     = "nfs/server",
+          .option      = "rpc-auth.addr.namelookup",
+          .type        = GLOBAL_DOC,
+          .op_version  = 1
+        },
+        { .key         = "nfs.dynamic-volumes",
+          .voltype     = "nfs/server",
+          .option      = "nfs.dynamic-volumes",
+          .type        = GLOBAL_DOC,
+          .op_version  = 1
+        },
+        { .key         = "nfs.register-with-portmap",
+          .voltype     = "nfs/server",
+          .option      = "rpc.register-with-portmap",
+          .type        = GLOBAL_DOC,
+          .op_version  = 1
+        },
+        { .key         = "nfs.outstanding-rpc-limit",
+          .voltype     = "nfs/server",
+          .option      = "rpc.outstanding-rpc-limit",
+          .type        = GLOBAL_DOC,
+          .op_version  = 3
+        },
+        { .key         = "nfs.port",
+          .voltype     = "nfs/server",
+          .option      = "nfs.port",
+          .type        = GLOBAL_DOC,
+          .op_version  = 1
+        },
+        { .key         = "nfs.rpc-auth-unix",
+          .voltype     = "nfs/server",
+          .option      = "!rpc-auth.auth-unix.*",
+          .op_version  = 1
+        },
+        { .key         = "nfs.rpc-auth-null",
+          .voltype     = "nfs/server",
+          .option      = "!rpc-auth.auth-null.*",
+          .op_version  = 1
+        },
+        { .key         = "nfs.rpc-auth-allow",
+          .voltype     = "nfs/server",
+          .option      = "!rpc-auth.addr.*.allow",
+          .op_version  = 1
+        },
+        { .key         = "nfs.rpc-auth-reject",
+          .voltype     = "nfs/server",
+          .option      = "!rpc-auth.addr.*.reject",
+          .op_version  = 1
+        },
+        { .key         = "nfs.ports-insecure",
+          .voltype     = "nfs/server",
+          .option      = "!rpc-auth.ports.*.insecure",
+          .op_version  = 1
+        },
+        { .key         = "nfs.transport-type",
+          .voltype     = "nfs/server",
+          .option      = "!nfs.transport-type",
+          .value       = "tcp",
+          .op_version  = 1,
+          .description = "Specifies the nfs transport type. Valid "
+                         "transport types are 'tcp' and 'rdma'."
+        },
+        { .key         = "nfs.trusted-sync",
+          .voltype     = "nfs/server",
+          .option      = "!nfs3.*.trusted-sync",
+          .op_version  = 1
+        },
+        { .key         = "nfs.trusted-write",
+          .voltype     = "nfs/server",
+          .option      = "!nfs3.*.trusted-write",
+          .op_version  = 1
+        },
+        { .key         = "nfs.volume-access",
+          .voltype     = "nfs/server",
+          .option      = "!nfs3.*.volume-access",
+          .op_version  = 1
+        },
+        { .key         = "nfs.export-dir",
+          .voltype     = "nfs/server",
+          .option      = "!nfs3.*.export-dir",
+          .op_version  = 1
+        },
+        { .key         = NFS_DISABLE_MAP_KEY,
+          .voltype     = "nfs/server",
+          .option      = "!nfs-disable",
+          .op_version  = 1
+        },
+        { .key         = "nfs.nlm",
+          .voltype     = "nfs/server",
+          .option      = "nfs.nlm",
+          .type        = GLOBAL_DOC,
+          .op_version  = 1
+        },
+        { .key         = "nfs.acl",
+          .voltype     = "nfs/server",
+          .option      = "nfs.acl",
+          .type        = GLOBAL_DOC,
+          .op_version  = 3
+        },
+        { .key         = "nfs.mount-udp",
+          .voltype     = "nfs/server",
+          .option      = "nfs.mount-udp",
+          .type        = GLOBAL_DOC,
+          .op_version  = 1
+        },
+        { .key         = "nfs.mount-rmtab",
+          .voltype     = "nfs/server",
+          .option      = "nfs.mount-rmtab",
+          .type        = GLOBAL_DOC,
+          .op_version  = 1
+        },
+        { .key         = "nfs.server-aux-gids",
+          .voltype     = "nfs/server",
+          .option      = "nfs.server-aux-gids",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "nfs.drc",
+          .voltype     = "nfs/server",
+          .option      = "nfs.drc",
+          .type        = GLOBAL_DOC,
+          .op_version  = 3
+        },
+        { .key         = "nfs.drc-size",
+          .voltype     = "nfs/server",
+          .option      = "nfs.drc-size",
+          .type        = GLOBAL_DOC,
+          .op_version  = 3
+        },
+        { .key         = "nfs.read-size",
+          .voltype     = "nfs/server",
+          .option      = "nfs3.read-size",
+          .type        = GLOBAL_DOC,
+          .op_version  = 3
+        },
+        { .key         = "nfs.write-size",
+          .voltype     = "nfs/server",
+          .option      = "nfs3.write-size",
+          .type        = GLOBAL_DOC,
+          .op_version  = 3
+        },
+        { .key         = "nfs.readdir-size",
+          .voltype     = "nfs/server",
+          .option      = "nfs3.readdir-size",
+          .type        = GLOBAL_DOC,
+          .op_version  = 3
+        },
+
+        /* Other options which don't fit any place above */
+        { .key        = "features.read-only",
+          .voltype    = "features/read-only",
+          .option     = "!read-only",
+          .value      = "off",
+          .op_version = 1,
+          .flags      = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT
+        },
+        { .key        = "features.worm",
+          .voltype    = "features/worm",
+          .option     = "!worm",
+          .value      = "off",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT | OPT_FLAG_XLATOR_OPT
+        },
+        { .key         = "storage.linux-aio",
+          .voltype     = "storage/posix",
+          .op_version  = 1
+        },
+        { .key         = "storage.batch-fsync-mode",
+          .voltype     = "storage/posix",
+          .op_version  = 3
+        },
+        { .key         = "storage.batch-fsync-delay-usec",
+          .voltype     = "storage/posix",
+          .op_version  = 3
+        },
+        { .key         = "storage.owner-uid",
+          .voltype     = "storage/posix",
+          .option      = "brick-uid",
+          .op_version  = 1
+        },
+        { .key         = "storage.owner-gid",
+          .voltype     = "storage/posix",
+          .option      = "brick-gid",
+          .op_version  = 1
+        },
+        { .key         = "storage.node-uuid-pathinfo",
+          .voltype     = "storage/posix",
+          .op_version  = 3
+        },
+        { .key         = "storage.health-check-interval",
+          .voltype     = "storage/posix",
+          .op_version  = 3
+        },
+        { .key         = "storage.bd-aio",
+          .voltype     = "storage/bd",
+          .op_version  = 3
+        },
+        { .key        = "config.memory-accounting",
+          .voltype    = "configuration",
+          .option     = "!config",
+          .op_version = 2,
+          .flags      = OPT_FLAG_CLIENT_OPT
+        },
+        { .key         = "config.transport",
+          .voltype     = "configuration",
+          .option      = "!config",
+          .op_version  = 2
+        },
+        { .key         = GLUSTERD_QUORUM_TYPE_KEY,
+          .voltype     = "mgmt/glusterd",
+          .value       = "off",
+          .op_version  = 2
+        },
+        { .key         = GLUSTERD_QUORUM_RATIO_KEY,
+          .voltype     = "mgmt/glusterd",
+          .value       = "0",
+          .op_version  = 2
+        },
+        /* changelog translator - global tunables */
+        { .key         = "changelog.changelog",
+          .voltype     = "features/changelog",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "changelog.changelog-dir",
+          .voltype     = "features/changelog",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "changelog.encoding",
+          .voltype     = "features/changelog",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "changelog.rollover-time",
+          .voltype     = "features/changelog",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = "changelog.fsync-interval",
+          .voltype     = "features/changelog",
+          .type        = NO_DOC,
+          .op_version  = 2
+        },
+        { .key         = NULL
+        }
+};
diff --git a/xlators/mgmt/glusterd/src/glusterd.c b/xlators/mgmt/glusterd/src/glusterd.c
index da40c285c..59288ada0 100644
--- a/xlators/mgmt/glusterd/src/glusterd.c
+++ b/xlators/mgmt/glusterd/src/glusterd.c
@@ -1,28 +1,19 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is GF_FREE software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
 #endif
 #include <time.h>
+#include <grp.h>
 #include <sys/uio.h>
 #include <sys/resource.h>
 
@@ -43,14 +34,29 @@
 #include "glusterd-sm.h"
 #include "glusterd-op-sm.h"
 #include "glusterd-store.h"
+#include "glusterd-hooks.h"
+#include "glusterd-utils.h"
+#include "glusterd-locks.h"
+#include "common-utils.h"
+#include "run.h"
 
+#include "syncop.h"
+
+#include "glusterd-mountbroker.h"
 
-static uuid_t glusterd_uuid;
-extern struct rpcsvc_program glusterd1_mop_prog;
 extern struct rpcsvc_program gluster_handshake_prog;
-extern struct rpc_clnt_program glusterd3_1_mgmt_prog;
+extern struct rpcsvc_program gluster_cli_getspec_prog;
 extern struct rpcsvc_program gluster_pmap_prog;
 extern glusterd_op_info_t opinfo;
+extern struct rpcsvc_program gd_svc_mgmt_prog;
+extern struct rpcsvc_program gd_svc_mgmt_v3_prog;
+extern struct rpcsvc_program gd_svc_peer_prog;
+extern struct rpcsvc_program gd_svc_cli_prog;
+extern struct rpcsvc_program gd_svc_cli_prog_ro;
+extern struct rpc_clnt_program gd_brick_prog;
+extern struct rpcsvc_program glusterd_mgmt_hndsk_prog;
+
+extern char snap_mount_folder[PATH_MAX];
 
 rpcsvc_cbk_program_t glusterd_cbk_prog = {
         .progname  = "Gluster Callback",
@@ -58,49 +64,92 @@ rpcsvc_cbk_program_t glusterd_cbk_prog = {
         .progver   = GLUSTER_CBK_VERSION,
 };
 
+struct rpcsvc_program *gd_inet_programs[] = {
+        &gd_svc_peer_prog,
+        &gd_svc_cli_prog_ro,
+        &gd_svc_mgmt_prog,
+        &gd_svc_mgmt_v3_prog,
+        &gluster_pmap_prog,
+        &gluster_handshake_prog,
+        &glusterd_mgmt_hndsk_prog,
+};
+int gd_inet_programs_count = (sizeof (gd_inet_programs) /
+                              sizeof (gd_inet_programs[0]));
+
+struct rpcsvc_program *gd_uds_programs[] = {
+        &gd_svc_cli_prog,
+        &gluster_cli_getspec_prog,
+};
+int gd_uds_programs_count = (sizeof (gd_uds_programs) /
+                             sizeof (gd_uds_programs[0]));
+
+const char *gd_op_list[GD_OP_MAX + 1] = {
+        [GD_OP_NONE]                    = "Invalid op",
+        [GD_OP_CREATE_VOLUME]           = "Create",
+        [GD_OP_START_BRICK]             = "Start Brick",
+        [GD_OP_STOP_BRICK]              = "Stop Brick",
+        [GD_OP_DELETE_VOLUME]           = "Delete",
+        [GD_OP_START_VOLUME]            = "Start",
+        [GD_OP_STOP_VOLUME]             = "Stop",
+        [GD_OP_DEFRAG_VOLUME]           = "Rebalance",
+        [GD_OP_ADD_BRICK]               = "Add brick",
+        [GD_OP_REMOVE_BRICK]            = "Remove brick",
+        [GD_OP_REPLACE_BRICK]           = "Replace brick",
+        [GD_OP_SET_VOLUME]              = "Set",
+        [GD_OP_RESET_VOLUME]            = "Reset",
+        [GD_OP_SYNC_VOLUME]             = "Sync",
+        [GD_OP_LOG_ROTATE]              = "Log rotate",
+        [GD_OP_GSYNC_SET]               = "Geo-replication",
+        [GD_OP_PROFILE_VOLUME]          = "Profile",
+        [GD_OP_QUOTA]                   = "Quota",
+        [GD_OP_STATUS_VOLUME]           = "Status",
+        [GD_OP_REBALANCE]               = "Rebalance",
+        [GD_OP_HEAL_VOLUME]             = "Heal",
+        [GD_OP_STATEDUMP_VOLUME]        = "Statedump",
+        [GD_OP_LIST_VOLUME]             = "Lists",
+        [GD_OP_CLEARLOCKS_VOLUME]       = "Clear locks",
+        [GD_OP_DEFRAG_BRICK_VOLUME]     = "Rebalance",
+        [GD_OP_COPY_FILE]               = "Copy File",
+        [GD_OP_SYS_EXEC]                = "Execute system commands",
+        [GD_OP_GSYNC_CREATE]            = "Geo-replication Create",
+        [GD_OP_SNAP]                    = "Snapshot",
+        [GD_OP_MAX]                     = "Invalid op"
+};
 
 static int
 glusterd_opinfo_init ()
 {
         int32_t ret = -1;
 
-        ret = pthread_mutex_init (&opinfo.lock, NULL);
+        opinfo.op = GD_OP_NONE;
 
         return ret;
 }
 
-static int
-glusterd_uuid_init (int flag)
+
+int
+glusterd_uuid_init ()
 {
         int             ret = -1;
-        char            str[50] = {0,};
+        xlator_t        *this = NULL;
         glusterd_conf_t *priv = NULL;
 
-        priv = THIS->private;
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
 
-        if (!flag) {
-                ret = glusterd_retrieve_uuid ();
-                if (!ret) {
-                        uuid_unparse (priv->uuid, str);
-                        uuid_copy (glusterd_uuid, priv->uuid);
-                        gf_log ("glusterd", GF_LOG_NORMAL,
-                                "retrieved UUID: %s", str);
-                        return 0;
-                }
+        ret = glusterd_retrieve_uuid ();
+        if (ret == 0) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "retrieved UUID: %s", uuid_utoa (priv->uuid));
+                return 0;
         }
 
-        uuid_generate (glusterd_uuid);
-        uuid_unparse (glusterd_uuid, str);
-
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                        "generated UUID: %s",str);
-        uuid_copy (priv->uuid, glusterd_uuid);
-
-        ret = glusterd_store_uuid ();
+        ret = glusterd_uuid_generate_save ();
 
         if (ret) {
                 gf_log ("glusterd", GF_LOG_ERROR,
-                          "Unable to store generated UUID");
+                          "Unable to generate and save new UUID");
                 return ret;
         }
 
@@ -108,6 +157,63 @@ glusterd_uuid_init (int flag)
 }
 
 int
+glusterd_uuid_generate_save ()
+{
+        int               ret = -1;
+        glusterd_conf_t   *priv = NULL;
+        xlator_t          *this = NULL;
+
+        this = THIS;
+        GF_ASSERT (this);
+        priv = this->private;
+        GF_ASSERT (priv);
+
+        uuid_generate (priv->uuid);
+
+        gf_log (this->name, GF_LOG_INFO, "generated UUID: %s",
+                uuid_utoa (priv->uuid));
+
+        ret = glusterd_store_global_info (this);
+
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unable to store the generated uuid %s",
+                        uuid_utoa (priv->uuid));
+
+        return ret;
+}
+
+int
+glusterd_options_init (xlator_t *this)
+{
+        int             ret = -1;
+        glusterd_conf_t *priv = NULL;
+        char            *initial_version = "0";
+
+        priv = this->private;
+
+        priv->opts = dict_new ();
+        if (!priv->opts)
+                goto out;
+
+        ret = glusterd_store_retrieve_options (this);
+        if (ret == 0)
+                goto out;
+
+        ret = dict_set_str (priv->opts, GLUSTERD_GLOBAL_OPT_VERSION,
+                            initial_version);
+        if (ret)
+                goto out;
+        ret = glusterd_store_options (this, priv->opts);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Unable to store version");
+                return ret;
+        }
+out:
+
+        return 0;
+}
+int
 glusterd_fetchspec_notify (xlator_t *this)
 {
         int              ret   = -1;
@@ -116,10 +222,15 @@ glusterd_fetchspec_notify (xlator_t *this)
 
         priv = this->private;
 
-        list_for_each_entry (trans, &priv->xprt_list, list) {
-                rpcsvc_callback_submit (priv->rpc, trans, &glusterd_cbk_prog,
-                                        GF_CBK_FETCHSPEC, NULL, 0);
+        pthread_mutex_lock (&priv->xprt_lock);
+        {
+                list_for_each_entry (trans, &priv->xprt_list, list) {
+                        rpcsvc_callback_submit (priv->rpc, trans,
+                                                &glusterd_cbk_prog,
+                                                GF_CBK_FETCHSPEC, NULL, 0);
+                }
         }
+        pthread_mutex_unlock (&priv->xprt_lock);
 
         ret = 0;
 
@@ -177,12 +288,16 @@ glusterd_rpcsvc_notify (rpcsvc_t *rpc, void *xl, rpcsvc_event_t event,
         {
                 INIT_LIST_HEAD (&xprt->list);
 
+                pthread_mutex_lock (&priv->xprt_lock);
                 list_add_tail (&xprt->list, &priv->xprt_list);
+                pthread_mutex_unlock (&priv->xprt_lock);
                 break;
         }
         case RPCSVC_EVENT_DISCONNECT:
         {
+                pthread_mutex_lock (&priv->xprt_lock);
                 list_del (&xprt->list);
+                pthread_mutex_unlock (&priv->xprt_lock);
                 pmap_registry_remove (this, 0, NULL, GF_PMAP_PORT_NONE, xprt);
                 break;
         }
@@ -215,6 +330,810 @@ out:
         return ret;
 }
 
+int
+glusterd_rpcsvc_options_build (dict_t *options)
+{
+        int             ret = 0;
+        uint32_t        backlog = 0;
+
+        ret = dict_get_uint32 (options, "transport.socket.listen-backlog",
+                               &backlog);
+
+        if (ret) {
+                backlog = GLUSTERD_SOCKET_LISTEN_BACKLOG;
+                ret = dict_set_uint32 (options,
+                                      "transport.socket.listen-backlog",
+                                      backlog);
+                if (ret)
+                        goto out;
+        }
+
+        gf_log ("", GF_LOG_DEBUG, "listen-backlog value: %d", backlog);
+
+out:
+        return ret;
+}
+
+#if SYNCDAEMON_COMPILE
+static int
+glusterd_check_gsync_present (int *valid_state)
+{
+        char                buff[PATH_MAX] = {0, };
+        runner_t            runner = {0,};
+        char               *ptr = NULL;
+        int                 ret = 0;
+
+        runinit (&runner);
+        runner_add_args (&runner, GSYNCD_PREFIX"/gsyncd", "--version", NULL);
+        runner_redir (&runner, STDOUT_FILENO, RUN_PIPE);
+        ret = runner_start (&runner);
+        if (ret == -1) {
+                if (errno == ENOENT) {
+                        gf_log ("glusterd", GF_LOG_INFO, GEOREP
+                                 " module not installed in the system");
+                        *valid_state = 0;
+                }
+                else {
+                        gf_log ("glusterd", GF_LOG_ERROR, GEOREP
+                                  " module not working as desired");
+                        *valid_state = -1;
+                }
+                goto out;
+        }
+
+        ptr = fgets(buff, sizeof(buff), runner_chio (&runner, STDOUT_FILENO));
+        if (ptr) {
+                if (!strstr (buff, "gsyncd")) {
+                        ret = -1;
+                        gf_log ("glusterd", GF_LOG_ERROR, GEOREP" module not "
+                                 "working as desired");
+                        *valid_state = -1;
+                        goto out;
+                }
+        } else {
+                ret = -1;
+                gf_log ("glusterd", GF_LOG_ERROR, GEOREP" module not "
+                         "working as desired");
+                *valid_state = -1;
+                goto out;
+        }
+
+        ret = 0;
+ out:
+
+        runner_end (&runner);
+
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+
+}
+
+static int
+group_write_allow (char *path, gid_t gid)
+{
+        struct stat st = {0,};
+        int ret        = 0;
+
+        ret = stat (path, &st);
+        if (ret == -1)
+                goto out;
+        GF_ASSERT (S_ISDIR (st.st_mode));
+
+        ret = chown (path, -1, gid);
+        if (ret == -1)
+                goto out;
+
+        ret = chmod (path, (st.st_mode & ~S_IFMT) | S_IWGRP|S_IXGRP|S_ISVTX);
+
+ out:
+        if (ret == -1)
+                gf_log ("", GF_LOG_CRITICAL,
+                        "failed to set up write access to %s for group %d (%s)",
+                        path, gid, strerror (errno));
+        return ret;
+}
+
+static int
+glusterd_crt_georep_folders (char *georepdir, glusterd_conf_t *conf)
+{
+        char *greplg_s   = NULL;
+        struct group *gr = NULL;
+        int ret          = 0;
+
+        GF_ASSERT (georepdir);
+        GF_ASSERT (conf);
+
+        if (strlen (conf->workdir)+2 > PATH_MAX-strlen(GEOREP)) {
+                ret = -1;
+                gf_log ("glusterd", GF_LOG_CRITICAL,
+                        "directory path %s/"GEOREP" is longer than PATH_MAX",
+                        conf->workdir);
+                goto out;
+        }
+
+        snprintf (georepdir, PATH_MAX, "%s/"GEOREP, conf->workdir);
+        ret = mkdir_p (georepdir, 0777, _gf_true);
+        if (-1 == ret) {
+                gf_log ("glusterd", GF_LOG_CRITICAL,
+                        "Unable to create "GEOREP" directory %s",
+                        georepdir);
+                goto out;
+        }
+
+        if (strlen (DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP) >= PATH_MAX) {
+                ret = -1;
+                gf_log ("glusterd", GF_LOG_CRITICAL,
+                        "directory path "DEFAULT_LOG_FILE_DIRECTORY"/"
+                        GEOREP" is longer than PATH_MAX");
+                goto out;
+        }
+        ret = mkdir_p (DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP, 0777, _gf_true);
+        if (-1 == ret) {
+                gf_log ("glusterd", GF_LOG_CRITICAL,
+                        "Unable to create "GEOREP" log directory");
+                goto out;
+        }
+
+        /* Slave log file directory */
+        if (strlen(DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves") >= PATH_MAX) {
+                ret = -1;
+                gf_log ("glusterd", GF_LOG_CRITICAL,
+                        "directory path "DEFAULT_LOG_FILE_DIRECTORY"/"
+                        GEOREP"-slaves"" is longer than PATH_MAX");
+                goto out;
+        }
+        ret = mkdir_p (DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves", 0777,
+                      _gf_true);
+        if (-1 == ret) {
+                gf_log ("glusterd", GF_LOG_CRITICAL,
+                        "Unable to create "GEOREP" slave log directory");
+                goto out;
+        }
+
+        /* MountBroker log file directory */
+        if (strlen(DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/mbr") >= PATH_MAX) {
+                ret = -1;
+                gf_log ("glusterd", GF_LOG_CRITICAL,
+                        "directory path "DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP
+                        "-slaves/mbr"" is longer than PATH_MAX");
+                goto out;
+        }
+        ret = mkdir_p (DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/mbr", 0777,
+                      _gf_true);
+        if (-1 == ret) {
+                gf_log ("glusterd", GF_LOG_CRITICAL,
+                        "Unable to create "GEOREP" mountbroker slave log directory");
+                goto out;
+        }
+
+        ret = dict_get_str (THIS->options, GEOREP"-log-group", &greplg_s);
+        if (ret)
+                ret = 0;
+        else {
+                gr = getgrnam (greplg_s);
+                if (!gr) {
+                        gf_log ("glusterd", GF_LOG_CRITICAL,
+                                "group "GEOREP"-log-group %s does not exist", greplg_s);
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = group_write_allow (DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP,
+                                         gr->gr_gid);
+                if (ret == 0)
+                        ret = group_write_allow (DEFAULT_LOG_FILE_DIRECTORY"/"
+                                                 GEOREP"-slaves", gr->gr_gid);
+                if (ret == 0)
+                        ret = group_write_allow (DEFAULT_LOG_FILE_DIRECTORY"/"
+                                                 GEOREP"-slaves/mbr", gr->gr_gid);
+        }
+
+ out:
+        gf_log("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+static void
+runinit_gsyncd_setrx (runner_t *runner, glusterd_conf_t *conf)
+{
+        runinit (runner);
+        runner_add_args (runner, GSYNCD_PREFIX"/gsyncd", "-c", NULL);
+        runner_argprintf (runner, "%s/"GSYNC_CONF_TEMPLATE, conf->workdir);
+        runner_add_arg (runner, "--config-set-rx");
+}
+
+static int
+configure_syncdaemon (glusterd_conf_t *conf)
+#define RUN_GSYNCD_CMD do {                                                          \
+        ret = runner_run_reuse (&runner);                                            \
+        if (ret == -1) {                                                             \
+                runner_log (&runner, "glusterd", GF_LOG_ERROR, "command failed");    \
+                runner_end (&runner);                                                \
+                goto out;                                                            \
+        }                                                                            \
+        runner_end (&runner);                                                        \
+} while (0)
+{
+        int ret = 0;
+        runner_t runner = {0,};
+        char georepdir[PATH_MAX] = {0,};
+        int valid_state = 0;
+
+        ret = setenv ("_GLUSTERD_CALLED_", "1", 1);
+        if (ret < 0) {
+                ret = 0;
+                goto out;
+        }
+        valid_state = -1;
+        ret = glusterd_check_gsync_present (&valid_state);
+        if (-1 == ret) {
+                ret = valid_state;
+                goto out;
+        }
+
+        glusterd_crt_georep_folders (georepdir, conf);
+        if (ret) {
+                ret = 0;
+                goto out;
+        }
+
+        /************
+         * master pre-configuration
+         ************/
+
+        /* remote-gsyncd */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner, "remote-gsyncd", GSYNCD_PREFIX"/gsyncd", ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner, "remote-gsyncd", "/nonexistent/gsyncd",
+                         ".", "^ssh:", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* gluster-command-dir */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner, "gluster-command-dir", SBIN_DIR"/",
+                         ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* gluster-params */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner, "gluster-params",
+                         "aux-gfid-mount xlator-option=*-dht.assert-no-child-down=true",
+                         ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* ssh-command */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_arg (&runner, "ssh-command");
+        runner_argprintf (&runner,
+                          "ssh -oPasswordAuthentication=no "
+                           "-oStrictHostKeyChecking=no "
+                           "-i %s/secret.pem", georepdir);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* pid-file */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_arg (&runner, "pid-file");
+        runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.pid", georepdir);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* state-file */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_arg (&runner, "state-file");
+        runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}.status", georepdir);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* state-detail-file */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_arg (&runner, "state-detail-file");
+        runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}-detail.status",
+                          georepdir);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* state-detail-file */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_arg (&runner, "state-detail-file");
+        runner_argprintf (&runner, "%s/${mastervol}_${remotehost}_${slavevol}/${eSlave}-detail.status",
+                          georepdir);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* state-socket */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_arg (&runner, "state-socket-unencoded");
+        runner_argprintf (&runner, "%s/${mastervol}/${eSlave}.socket", georepdir);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* socketdir */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner, "socketdir", GLUSTERD_SOCK_DIR, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* log-file */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner,
+                         "log-file",
+                         DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"/${mastervol}/${eSlave}.log",
+                         ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* gluster-log-file */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner,
+                         "gluster-log-file",
+                         DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"/${mastervol}/${eSlave}${local_id}.gluster.log",
+                         ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* ignore-deletes */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner, "ignore-deletes", "true", ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* special-sync-mode */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner, "special-sync-mode", "partial", ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* change-detector == changelog */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args(&runner, "change-detector", "changelog", ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_arg(&runner, "working-dir");
+        runner_argprintf(&runner, "%s/${mastervol}/${eSlave}",
+                         DEFAULT_VAR_RUN_DIRECTORY);
+        runner_add_args (&runner, ".", ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /************
+         * slave pre-configuration
+         ************/
+
+        /* gluster-command-dir */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner, "gluster-command-dir", SBIN_DIR"/",
+                         ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* gluster-params */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner, "gluster-params",
+                         "aux-gfid-mount xlator-option=*-dht.assert-no-child-down=true",
+                         ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* log-file */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner,
+                         "log-file",
+                         DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/${session_owner}:${eSlave}.log",
+                         ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* MountBroker log-file */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner,
+                         "log-file-mbr",
+                         DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/mbr/${session_owner}:${eSlave}.log",
+                         ".", NULL);
+        RUN_GSYNCD_CMD;
+
+        /* gluster-log-file */
+        runinit_gsyncd_setrx (&runner, conf);
+        runner_add_args (&runner,
+                         "gluster-log-file",
+                         DEFAULT_LOG_FILE_DIRECTORY"/"GEOREP"-slaves/${session_owner}:${eSlave}.gluster.log",
+                         ".", NULL);
+        RUN_GSYNCD_CMD;
+
+ out:
+        return ret ? -1 : 0;
+}
+#undef RUN_GSYNCD_CMD
+#else /* SYNCDAEMON_COMPILE */
+static int
+configure_syncdaemon (glusterd_conf_t *conf)
+{
+        return 0;
+}
+#endif /* !SYNCDAEMON_COMPILE */
+
+
+static int
+check_prepare_mountbroker_root (char *mountbroker_root)
+{
+        int dfd0        = -1;
+        int dfd         = -1;
+        int dfd2        = -1;
+        struct stat st  = {0,};
+        struct stat st2 = {0,};
+        int ret         = 0;
+
+        ret = open (mountbroker_root, O_RDONLY);
+        if (ret != -1) {
+                dfd = ret;
+                ret = fstat (dfd, &st);
+        }
+        if (ret == -1 || !S_ISDIR (st.st_mode)) {
+                gf_log ("", GF_LOG_ERROR,
+                        "cannot access mountbroker-root directory %s",
+                        mountbroker_root);
+                ret = -1;
+                goto out;
+        }
+        if (st.st_uid != 0 ||
+            (st.st_mode & (S_IWGRP|S_IWOTH))) {
+                gf_log ("", GF_LOG_ERROR,
+                        "permissions on mountbroker-root directory %s are "
+                        "too liberal", mountbroker_root);
+                ret = -1;
+                goto out;
+        }
+        if (!(st.st_mode & (S_IXGRP|S_IXOTH))) {
+                gf_log ("", GF_LOG_WARNING,
+                        "permissions on mountbroker-root directory %s are "
+                        "probably too strict", mountbroker_root);
+        }
+
+        dfd0 = dup (dfd);
+
+        for (;;) {
+                ret = openat (dfd, "..", O_RDONLY);
+                if (ret != -1) {
+                        dfd2 = ret;
+                        ret = fstat (dfd2, &st2);
+                }
+                if (ret == -1) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "error while checking mountbroker-root ancestors "
+                                "%d (%s)", errno, strerror (errno));
+                        goto out;
+                }
+
+                if (st2.st_ino == st.st_ino)
+                        break; /* arrived to root */
+
+                if (st2.st_uid != 0 ||
+                    ((st2.st_mode & (S_IWGRP|S_IWOTH)) &&
+                     !(st2.st_mode & S_ISVTX))) {
+                        gf_log ("", GF_LOG_ERROR,
+                                "permissions on ancestors of mountbroker-root "
+                                "directory are too liberal");
+                        ret = -1;
+                        goto out;
+                }
+                if (!(st.st_mode & (S_IXGRP|S_IXOTH))) {
+                        gf_log ("", GF_LOG_WARNING,
+                                "permissions on ancestors of mountbroker-root "
+                                "directory are probably too strict");
+                }
+
+                close (dfd);
+                dfd = dfd2;
+                st = st2;
+        }
+
+        ret = mkdirat (dfd0, MB_HIVE, 0711);
+        if (ret == -1 && errno == EEXIST)
+                ret = 0;
+        if (ret != -1)
+                ret = fstatat (dfd0, MB_HIVE, &st, AT_SYMLINK_NOFOLLOW);
+        if (ret == -1 || st.st_mode != (S_IFDIR|0711)) {
+                gf_log ("", GF_LOG_ERROR,
+                        "failed to set up mountbroker-root directory %s",
+                        mountbroker_root);
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+
+ out:
+        if (dfd0 != -1)
+                close (dfd0);
+        if (dfd != -1)
+                close (dfd);
+        if (dfd2 != -1)
+                close (dfd2);
+
+        return ret;
+}
+
+static int
+_install_mount_spec (dict_t *opts, char *key, data_t *value, void *data)
+{
+        glusterd_conf_t *priv           = THIS->private;
+        char            *label          = NULL;
+        gf_boolean_t     georep         = _gf_false;
+        gf_boolean_t     ghadoop        = _gf_false;
+        char            *pdesc          = value->data;
+        char            *volname        = NULL;
+        int              rv             = 0;
+        gf_mount_spec_t *mspec          = NULL;
+        char            *user           = NULL;
+        char            *volfile_server = NULL;
+
+        label = strtail (key, "mountbroker.");
+
+        /* check for presence of geo-rep/hadoop label */
+        if (!label) {
+                label = strtail (key, "mountbroker-"GEOREP".");
+                if (label)
+                        georep = _gf_true;
+                else {
+                        label = strtail (key, "mountbroker-"GHADOOP".");
+                        if (label)
+                                ghadoop = _gf_true;
+                }
+        }
+
+        if (!label)
+                return 0;
+
+        mspec = GF_CALLOC (1, sizeof (*mspec), gf_gld_mt_mount_spec);
+        if (!mspec)
+                goto err;
+        mspec->label = label;
+
+        if (georep || ghadoop) {
+                volname = gf_strdup (pdesc);
+                if (!volname)
+                        goto err;
+                user = strchr (volname, ':');
+                if (user) {
+                        *user = '\0';
+                        user++;
+                } else
+                        user = label;
+
+                if (georep)
+                        rv = make_georep_mountspec (mspec, volname, user);
+
+                if (ghadoop) {
+                        volfile_server = strchr (user, ':');
+                        if (volfile_server)
+                                *volfile_server++ = '\0';
+                        else
+                                volfile_server = "localhost";
+
+                        rv = make_ghadoop_mountspec (mspec, volname, user, volfile_server);
+                }
+
+                GF_FREE (volname);
+                if (rv != 0)
+                        goto err;
+        } else if (parse_mount_pattern_desc (mspec, pdesc) != 0)
+                goto err;
+
+        list_add_tail (&mspec->speclist, &priv->mount_specs);
+
+        return 0;
+ err:
+
+        gf_log ("", GF_LOG_ERROR,
+                "adding %smount spec failed: label: %s desc: %s",
+                georep ? GEOREP" " : "", label, pdesc);
+
+        return -1;
+}
+
+
+static int
+gd_default_synctask_cbk (int ret, call_frame_t *frame, void *opaque)
+{
+        glusterd_conf_t     *priv = THIS->private;
+        synclock_unlock (&priv->big_lock);
+        return ret;
+}
+
+static void
+glusterd_launch_synctask (synctask_fn_t fn, void *opaque)
+{
+        xlator_t        *this = NULL;
+        glusterd_conf_t *priv = NULL;
+        int             ret   = -1;
+
+        this = THIS;
+        priv = this->private;
+
+        synclock_lock (&priv->big_lock);
+        ret = synctask_new (this->ctx->env, fn, gd_default_synctask_cbk, NULL,
+                            opaque);
+        if (ret)
+                gf_log (this->name, GF_LOG_CRITICAL, "Failed to spawn bricks"
+                        " and other volume related services");
+}
+
+int
+glusterd_uds_rpcsvc_notify (rpcsvc_t *rpc, void *xl, rpcsvc_event_t event,
+                            void *data)
+{
+        /* glusterd_rpcsvc_notify() does stuff that calls coming in from the
+         * unix domain socket don't need. This is just an empty function to be
+         * used for the uds listener. This will be used later if required.
+         */
+        return 0;
+}
+
+/* The glusterd unix domain socket listener only listens for cli */
+rpcsvc_t *
+glusterd_init_uds_listener (xlator_t *this)
+{
+        int             ret = -1;
+        dict_t          *options = NULL;
+        rpcsvc_t        *rpc = NULL;
+        data_t          *sock_data = NULL;
+        char            sockfile[PATH_MAX+1] = {0,};
+        int             i = 0;
+
+
+        GF_ASSERT (this);
+
+        sock_data = dict_get (this->options, "glusterd-sockfile");
+        if (!sock_data) {
+                strncpy (sockfile, DEFAULT_GLUSTERD_SOCKFILE, PATH_MAX);
+        } else {
+                strncpy (sockfile, sock_data->data, PATH_MAX);
+        }
+
+        options = dict_new ();
+        if (!options)
+                goto out;
+
+        ret = rpcsvc_transport_unix_options_build (&options, sockfile);
+        if (ret)
+                goto out;
+
+        rpc = rpcsvc_init (this, this->ctx, options, 8);
+        if (rpc == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = rpcsvc_register_notify (rpc, glusterd_uds_rpcsvc_notify,
+                                      this);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Failed to register notify function");
+                goto out;
+        }
+
+        ret = rpcsvc_create_listeners (rpc, options, this->name);
+        if (ret != 1) {
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to create listener");
+                goto out;
+        }
+        ret = 0;
+
+        for (i = 0; i < gd_uds_programs_count; i++) {
+                ret = glusterd_program_register (this, rpc, gd_uds_programs[i]);
+                if (ret) {
+                        i--;
+                        for (; i >= 0; i--)
+                                rpcsvc_program_unregister (rpc,
+                                                           gd_uds_programs[i]);
+
+                        goto out;
+                }
+        }
+
+out:
+        if (options)
+                dict_unref (options);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to start glusterd "
+                        "unix domain socket listener.");
+                if (rpc) {
+                        GF_FREE (rpc);
+                        rpc = NULL;
+                }
+        }
+        return rpc;
+}
+
+void
+glusterd_stop_uds_listener (xlator_t *this)
+{
+        glusterd_conf_t         *conf = NULL;
+        rpcsvc_listener_t       *listener = NULL;
+        rpcsvc_listener_t       *next = NULL;
+
+        GF_ASSERT (this);
+        conf = this->private;
+
+        (void) rpcsvc_program_unregister (conf->uds_rpc, &gd_svc_cli_prog);
+        (void) rpcsvc_program_unregister (conf->uds_rpc, &gluster_handshake_prog);
+
+        list_for_each_entry_safe (listener, next, &conf->uds_rpc->listeners,
+                                  list) {
+                rpcsvc_listener_destroy (listener);
+        }
+
+        (void) rpcsvc_unregister_notify (conf->uds_rpc, glusterd_rpcsvc_notify,
+                                         this);
+
+        unlink (DEFAULT_GLUSTERD_SOCKFILE);
+
+        GF_FREE (conf->uds_rpc);
+        conf->uds_rpc = NULL;
+
+        return;
+}
+
+static int
+glusterd_init_snap_folder (xlator_t *this)
+{
+        int             ret = -1;
+        struct stat     buf = {0,};
+
+        GF_ASSERT (this);
+
+        /* Snapshot volumes are mounted under /var/run/gluster/snaps folder.
+         * But /var/run is normally a symbolic link to /run folder, which
+         * creates problems as the entry point in the mtab for the mount point
+         * and glusterd maintained entry point will be different. Therefore
+         * identify the correct run folder and use it for snap volume mounting.
+         */
+        ret = lstat (GLUSTERD_VAR_RUN_DIR, &buf);
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "stat fails on %s, exiting. (errno = %d)",
+                        GLUSTERD_VAR_RUN_DIR, errno);
+                goto out;
+        }
+
+        /* If /var/run is symlink then use /run folder */
+        if (S_ISLNK (buf.st_mode)) {
+                strcpy (snap_mount_folder, GLUSTERD_RUN_DIR);
+        } else {
+                strcpy (snap_mount_folder, GLUSTERD_VAR_RUN_DIR);
+        }
+
+        strcat (snap_mount_folder, GLUSTERD_DEFAULT_SNAPS_BRICK_DIR);
+
+        ret = stat (snap_mount_folder, &buf);
+        if ((ret != 0) && (ENOENT != errno)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "stat fails on %s, exiting. (errno = %d)",
+                        snap_mount_folder, errno);
+                ret = -1;
+                goto out;
+        }
+
+        if ((!ret) && (!S_ISDIR(buf.st_mode))) {
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "Provided snap path %s is not a directory,"
+                        "exiting", snap_mount_folder);
+                ret = -1;
+                goto out;
+        }
+
+        if ((-1 == ret) && (ENOENT == errno)) {
+                /* Create missing folders */
+                ret = mkdir_p (snap_mount_folder, 0777, _gf_false);
+
+                if (-1 == ret) {
+                        gf_log (this->name, GF_LOG_CRITICAL,
+                                "Unable to create directory %s"
+                                " ,errno = %d", snap_mount_folder, errno);
+                        goto out;
+                }
+        }
+
+out:
+        return ret;
+}
+
 /*
  * init - called during glusterd initialization
  *
@@ -226,57 +1145,70 @@ init (xlator_t *this)
 {
         int32_t            ret               = -1;
         rpcsvc_t          *rpc               = NULL;
+        rpcsvc_t          *uds_rpc           = NULL;
         glusterd_conf_t   *conf              = NULL;
         data_t            *dir_data          = NULL;
         struct stat        buf               = {0,};
-        char              *port_str          = NULL;
-        int                port_num          = 0;
-        char               voldir [PATH_MAX] = {0,};
-        char               dirname [PATH_MAX];
+        char               storedir [PATH_MAX] = {0,};
+        char               workdir [PATH_MAX] = {0,};
+        char               hooks_dir [PATH_MAX] = {0,};
         char               cmd_log_filename [PATH_MAX] = {0,};
         int                first_time        = 0;
+        char              *mountbroker_root  = NULL;
+        int                i                 = 0;
+        char              *valgrind_str      = NULL;
 
         dir_data = dict_get (this->options, "working-directory");
 
         if (!dir_data) {
                 //Use default working dir
-                strncpy (dirname, GLUSTERD_DEFAULT_WORKDIR, PATH_MAX);
+                strncpy (workdir, GLUSTERD_DEFAULT_WORKDIR, PATH_MAX);
         } else {
-                strncpy (dirname, dir_data->data, PATH_MAX);
+                strncpy (workdir, dir_data->data, PATH_MAX);
         }
 
-        ret = stat (dirname, &buf);
+        ret = stat (workdir, &buf);
         if ((ret != 0) && (ENOENT != errno)) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "stat fails on %s, exiting. (errno = %d)",
-			dirname, errno);
+                        workdir, errno);
                 exit (1);
         }
 
         if ((!ret) && (!S_ISDIR(buf.st_mode))) {
                 gf_log (this->name, GF_LOG_CRITICAL,
                         "Provided working area %s is not a directory,"
-                        "exiting", dirname);
+                        "exiting", workdir);
                 exit (1);
         }
 
 
         if ((-1 == ret) && (ENOENT == errno)) {
-                ret = mkdir (dirname, 0777);
+                ret = mkdir (workdir, 0777);
 
                 if (-1 == ret) {
                         gf_log (this->name, GF_LOG_CRITICAL,
                                 "Unable to create directory %s"
-                                " ,errno = %d", dirname, errno);
+                                " ,errno = %d", workdir, errno);
                         exit (1);
                 }
+
                 first_time = 1;
         }
 
-        gf_log (this->name, GF_LOG_NORMAL, "Using %s as working directory",
-                dirname);
+        setenv ("GLUSTERD_WORKING_DIR", workdir, 1);
+        gf_log (this->name, GF_LOG_INFO, "Using %s as working directory",
+                workdir);
+
+        ret = glusterd_init_snap_folder (this);
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL, "Unable to create "
+                        "snap backend folder");
+                 exit (1);
+         }
 
-        snprintf (cmd_log_filename, PATH_MAX,"%s/.cmd_log_history",dirname);
+        snprintf (cmd_log_filename, PATH_MAX,"%s/.cmd_log_history",
+                  DEFAULT_LOG_FILE_DIRECTORY);
         ret = gf_cmd_log_init (cmd_log_filename);
 
         if (ret == -1) {
@@ -285,56 +1217,68 @@ init (xlator_t *this)
                 exit (1);
         }
 
-        snprintf (voldir, PATH_MAX, "%s/vols", dirname);
+        snprintf (storedir, PATH_MAX, "%s/vols", workdir);
 
-        ret = mkdir (voldir, 0777);
+        ret = mkdir (storedir, 0777);
 
         if ((-1 == ret) && (errno != EEXIST)) {
                 gf_log (this->name, GF_LOG_CRITICAL,
                         "Unable to create volume directory %s"
-                        " ,errno = %d", voldir, errno);
+                        " ,errno = %d", storedir, errno);
                 exit (1);
         }
 
-        snprintf (voldir, PATH_MAX, "%s/peers", dirname);
+        snprintf (storedir, PATH_MAX, "%s/peers", workdir);
 
-        ret = mkdir (voldir, 0777);
+        ret = mkdir (storedir, 0777);
 
         if ((-1 == ret) && (errno != EEXIST)) {
                 gf_log (this->name, GF_LOG_CRITICAL,
                         "Unable to create peers directory %s"
-                        " ,errno = %d", voldir, errno);
+                        " ,errno = %d", storedir, errno);
                 exit (1);
         }
 
-        snprintf (voldir, PATH_MAX, "%s/logs", dirname);
-        ret = mkdir (voldir, 0777);
+        snprintf (storedir, PATH_MAX, "%s/bricks", DEFAULT_LOG_FILE_DIRECTORY);
+        ret = mkdir (storedir, 0777);
         if ((-1 == ret) && (errno != EEXIST)) {
                 gf_log (this->name, GF_LOG_CRITICAL,
                         "Unable to create logs directory %s"
-                        " ,errno = %d", voldir, errno);
+                        " ,errno = %d", storedir, errno);
                 exit (1);
         }
 
-        snprintf (voldir, PATH_MAX, "%s/logs/bricks", dirname);
-        ret = mkdir (voldir, 0777);
+        snprintf (storedir, PATH_MAX, "%s/nfs", workdir);
+        ret = mkdir (storedir, 0777);
         if ((-1 == ret) && (errno != EEXIST)) {
                 gf_log (this->name, GF_LOG_CRITICAL,
-                        "Unable to create logs directory %s"
-                        " ,errno = %d", voldir, errno);
+                        "Unable to create nfs directory %s"
+                        " ,errno = %d", storedir, errno);
                 exit (1);
         }
 
-        snprintf (voldir, PATH_MAX, "%s/nfs", dirname);
-        ret = mkdir (voldir, 0777);
+        snprintf (storedir, PATH_MAX, "%s/glustershd", workdir);
+        ret = mkdir (storedir, 0777);
         if ((-1 == ret) && (errno != EEXIST)) {
                 gf_log (this->name, GF_LOG_CRITICAL,
-                        "Unable to create nfs directory %s"
-                        " ,errno = %d", voldir, errno);
+                        "Unable to create glustershd directory %s"
+                        " ,errno = %d", storedir, errno);
+                exit (1);
+        }
+
+        snprintf (storedir, PATH_MAX, "%s/groups", workdir);
+        ret = mkdir (storedir, 0777);
+        if ((-1 == ret) && (errno != EEXIST)) {
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "Unable to create glustershd directory %s"
+                        " ,errno = %d", storedir, errno);
                 exit (1);
         }
 
-        rpc = rpcsvc_init (this->ctx, this->options);
+        ret = glusterd_rpcsvc_options_build (this->options);
+        if (ret)
+                goto out;
+        rpc = rpcsvc_init (this, this->ctx, this->options, 64);
         if (rpc == NULL) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "failed to init rpc");
@@ -348,13 +1292,6 @@ init (xlator_t *this)
                 goto out;
         }
 
-        glusterd1_mop_prog.options = this->options;
-        port_str = getenv ("GLUSTERD_LOCAL_PORT");
-        if (port_str) {
-                port_num = atoi (port_str);
-                glusterd1_mop_prog.progport = port_num;
-        }
-
         /*
          * only one (atmost a pair - rdma and socket) listener for
          * glusterd1_mop_prog, gluster_pmap_prog and gluster_handshake_prog.
@@ -367,61 +1304,148 @@ init (xlator_t *this)
                 goto out;
         }
 
-        ret = glusterd_program_register (this, rpc, &glusterd1_mop_prog);
-        if (ret) {
-                goto out;
-        }
+        for (i = 0; i < gd_inet_programs_count; i++) {
+                ret = glusterd_program_register (this, rpc,
+                                                 gd_inet_programs[i]);
+                if (ret) {
+                        i--;
+                        for (; i >= 0; i--)
+                                rpcsvc_program_unregister (rpc,
+                                                           gd_inet_programs[i]);
 
-        ret = glusterd_program_register (this, rpc, &gluster_pmap_prog);
-        if (ret) {
-                rpcsvc_program_unregister (rpc, &glusterd1_mop_prog);
-                goto out;
+                        goto out;
+                }
         }
 
-        gluster_handshake_prog.options = this->options;
-        ret = glusterd_program_register (this, rpc, &gluster_handshake_prog);
-        if (ret) {
-                rpcsvc_program_unregister (rpc, &glusterd1_mop_prog);
-                rpcsvc_program_unregister (rpc, &gluster_handshake_prog);
+        /* Start a unix domain socket listener just for cli commands
+         * This should prevent ports from being wasted by being in TIMED_WAIT
+         * when cli commands are done continuously
+         */
+        uds_rpc = glusterd_init_uds_listener (this);
+        if (uds_rpc == NULL) {
+                ret = -1;
                 goto out;
         }
 
         conf = GF_CALLOC (1, sizeof (glusterd_conf_t),
                           gf_gld_mt_glusterd_conf_t);
         GF_VALIDATE_OR_GOTO(this->name, conf, out);
+        conf->shd = GF_CALLOC (1, sizeof (nodesrv_t),
+                               gf_gld_mt_nodesrv_t);
+        GF_VALIDATE_OR_GOTO(this->name, conf->shd, out);
+        conf->nfs = GF_CALLOC (1, sizeof (nodesrv_t),
+                               gf_gld_mt_nodesrv_t);
+        GF_VALIDATE_OR_GOTO(this->name, conf->nfs, out);
+
         INIT_LIST_HEAD (&conf->peers);
         INIT_LIST_HEAD (&conf->volumes);
-        INIT_LIST_HEAD (&conf->hostnames);
+        INIT_LIST_HEAD (&conf->snapshots);
+        INIT_LIST_HEAD (&conf->missed_snaps_list);
+
         pthread_mutex_init (&conf->mutex, NULL);
         conf->rpc = rpc;
-        conf->mgmt = &glusterd3_1_mgmt_prog;
-        strncpy (conf->workdir, dirname, PATH_MAX);
+        conf->uds_rpc = uds_rpc;
+        conf->gfs_mgmt = &gd_brick_prog;
+        strncpy (conf->workdir, workdir, PATH_MAX);
 
+        synclock_init (&conf->big_lock);
+        pthread_mutex_init (&conf->xprt_lock, NULL);
         INIT_LIST_HEAD (&conf->xprt_list);
 
+        glusterd_friend_sm_init ();
+        glusterd_op_sm_init ();
+        glusterd_opinfo_init ();
+        glusterd_mgmt_v3_lock_init ();
+        glusterd_txn_opinfo_dict_init ();
+        ret = glusterd_sm_tr_log_init (&conf->op_sm_log,
+                                       glusterd_op_sm_state_name_get,
+                                       glusterd_op_sm_event_name_get,
+                                       GLUSTERD_TR_LOG_SIZE);
+        if (ret)
+                goto out;
+
+         conf->base_port = GF_IANA_PRIV_PORTS_START;
+         if (dict_get_uint32(this->options, "base-port", &conf->base_port) == 0) {
+                 gf_log (this->name, GF_LOG_INFO,
+                         "base-port override: %d", conf->base_port);
+         }
+
+        /* Set option to run bricks on valgrind if enabled in glusterd.vol */
+        conf->valgrind = _gf_false;
+        ret = dict_get_str (this->options, "run-with-valgrind", &valgrind_str);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "cannot get run-with-valgrind value");
+        }
+        if (valgrind_str) {
+                if (gf_string2boolean (valgrind_str, &(conf->valgrind))) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "run-with-valgrind value not a boolean string");
+                }
+        }
+
         this->private = conf;
-        //this->ctx->top = this;
+        (void) glusterd_nodesvc_set_online_status ("glustershd", _gf_false);
 
-        ret = glusterd_uuid_init (first_time);
-        if (ret < 0)
+        GLUSTERD_GET_HOOKS_DIR (hooks_dir, GLUSTERD_HOOK_VER, conf);
+        if (stat (hooks_dir, &buf)) {
+                ret = glusterd_hooks_create_hooks_directory (conf->workdir);
+                if (-1 == ret) {
+                        gf_log (this->name, GF_LOG_CRITICAL,
+                                "Unable to create hooks directory ");
+                        exit (1);
+                }
+        }
+
+        INIT_LIST_HEAD (&conf->mount_specs);
+
+        ret = dict_foreach (this->options, _install_mount_spec, NULL);
+        if (ret)
+                goto out;
+        ret = dict_get_str (this->options, "mountbroker-root",
+                            &mountbroker_root);
+        if (ret)
+                ret = 0;
+        else
+                ret = check_prepare_mountbroker_root (mountbroker_root);
+        if (ret)
+                goto out;
+
+        ret = configure_syncdaemon (conf);
+        if (ret)
                 goto out;
 
         ret = glusterd_restore ();
         if (ret < 0)
                 goto out;
 
-        glusterd_friend_sm_init ();
-        glusterd_op_sm_init ();
-        glusterd_opinfo_init ();
+        /* If there are no 'friends', this would be the best time to
+         * spawn process/bricks that may need (re)starting since last
+         * time (this) glusterd was up.*/
+
+        if (list_empty (&conf->peers)) {
+                glusterd_launch_synctask (glusterd_spawn_daemons, NULL);
+        }
+        ret = glusterd_options_init (this);
+        if (ret < 0)
+                goto out;
+
+        ret = glusterd_handle_upgrade_downgrade (this->options, conf);
+        if (ret)
+                goto out;
 
+        ret = glusterd_hooks_spawn_worker (this);
+        if (ret)
+                goto out;
 
         ret = 0;
 out:
-        if (ret == -1) {
+        if (ret < 0) {
                 if (this->private != NULL) {
                         GF_FREE (this->private);
                         this->private = NULL;
                 }
+
         }
 
         return ret;
@@ -445,11 +1469,17 @@ fini (xlator_t *this)
                 goto out;
 
         conf = this->private;
-        if (conf->pmap)
-                FREE (conf->pmap);
+
+        glusterd_stop_uds_listener (this);
+
+        FREE (conf->pmap);
         if (conf->handle)
-                glusterd_store_handle_destroy (conf->handle);
+                gf_store_handle_destroy (conf->handle);
+        glusterd_sm_tr_log_delete (&conf->op_sm_log);
+        glusterd_mgmt_v3_lock_fini ();
+        glusterd_txn_opinfo_dict_fini ();
         GF_FREE (conf);
+
         this->private = NULL;
 out:
         return;
@@ -487,11 +1517,9 @@ notify (xlator_t *this, int32_t event, void *data, ...)
 }
 
 
-struct xlator_fops fops = {
-};
+struct xlator_fops fops;
 
-struct xlator_cbks cbks = {
-};
+struct xlator_cbks cbks;
 
 struct xlator_dumpops dumpops = {
         .priv  = glusterd_priv,
@@ -511,6 +1539,59 @@ struct volume_options options[] = {
         { .key   = {"rpc-auth.*"},
           .type  = GF_OPTION_TYPE_ANY,
         },
-
+        { .key   = {"rpc-auth-allow-insecure"},
+          .type  = GF_OPTION_TYPE_BOOL,
+        },
+        { .key  = {"upgrade"},
+          .type = GF_OPTION_TYPE_BOOL,
+        },
+        { .key  = {"downgrade"},
+          .type = GF_OPTION_TYPE_BOOL,
+        },
+        { .key = {"bind-insecure"},
+          .type = GF_OPTION_TYPE_BOOL,
+        },
+        { .key  = {"mountbroker-root"},
+          .type = GF_OPTION_TYPE_PATH,
+        },
+        { .key  = {"mountbroker.*"},
+          .type = GF_OPTION_TYPE_ANY,
+        },
+        { .key  = {"mountbroker-"GEOREP".*"},
+          .type = GF_OPTION_TYPE_ANY,
+        },
+        { .key  = {"mountbroker-"GHADOOP".*"},
+          .type = GF_OPTION_TYPE_ANY,
+        },
+        { .key = {GEOREP"-log-group"},
+          .type = GF_OPTION_TYPE_ANY,
+        },
+        { .key = {"run-with-valgrind"},
+          .type = GF_OPTION_TYPE_BOOL,
+        },
+        { .key = {"server-quorum-type"},
+          .type = GF_OPTION_TYPE_STR,
+          .value = { "none", "server"},
+          .description = "If set to server, enables the specified "
+          "volume to participate in quorum."
+        },
+        { .key = {"server-quorum-ratio"},
+          .type = GF_OPTION_TYPE_PERCENT,
+          .description = "Sets the quorum percentage for the trusted "
+          "storage pool."
+        },
+        { .key = {"glusterd-sockfile"},
+          .type = GF_OPTION_TYPE_PATH,
+          .description = "The socket file on which glusterd should listen for "
+                        "cli requests. Default is "DEFAULT_GLUSTERD_SOCKFILE "."
+        },
+        { .key = {"base-port"},
+          .type = GF_OPTION_TYPE_INT,
+          .description = "Sets the base port for portmap query"
+        },
+        { .key = {"snap-brick-path"},
+          .type = GF_OPTION_TYPE_STR,
+          .description = "directory where the bricks for the snapshots will be created"
+        },
         { .key   = {NULL} },
 };
diff --git a/xlators/mgmt/glusterd/src/glusterd.h b/xlators/mgmt/glusterd/src/glusterd.h
index b9890760f..3aa395ebc 100644
--- a/xlators/mgmt/glusterd/src/glusterd.h
+++ b/xlators/mgmt/glusterd/src/glusterd.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _GLUSTERD_H_
 #define _GLUSTERD_H_
 
@@ -45,44 +35,164 @@
 #include "glusterd1-xdr.h"
 #include "protocol-common.h"
 #include "glusterd-pmap.h"
-
+#include "cli1-xdr.h"
+#include "syncop.h"
+#include "store.h"
 
 #define GLUSTERD_MAX_VOLUME_NAME        1000
+#define GLUSTERD_TR_LOG_SIZE            50
+#define GLUSTERD_NAME                   "glusterd"
+#define GLUSTERD_SOCKET_LISTEN_BACKLOG  128
+#define GLUSTERD_QUORUM_TYPE_KEY        "cluster.server-quorum-type"
+#define GLUSTERD_QUORUM_RATIO_KEY       "cluster.server-quorum-ratio"
+#define GLUSTERD_GLOBAL_OPT_VERSION     "global-option-version"
+#define GLUSTERD_COMMON_PEM_PUB_FILE    "/geo-replication/common_secret.pem.pub"
+#define GEO_CONF_MAX_OPT_VALS           5
+#define GLUSTERD_CREATE_HOOK_SCRIPT     "/hooks/1/gsync-create/post/" \
+                                        "S56glusterd-geo-rep-create-post.sh"
+
+
+#define GLUSTERD_SNAPS_MAX_HARD_LIMIT 256
+#define GLUSTERD_SNAPS_DEF_SOFT_LIMIT_PERCENT 90
+#define GLUSTERD_SNAPS_MAX_SOFT_LIMIT_PERCENT 100
+#define GLUSTERD_SERVER_QUORUM "server"
+
+#define FMTSTR_CHECK_VOL_EXISTS "Volume %s does not exist"
+#define FMTSTR_RESOLVE_BRICK "Could not find peer on which brick %s:%s resides"
+
+#define LOGSTR_FOUND_BRICK  "Found brick %s:%s in volume %s"
+#define LOGSTR_BUILD_PAYLOAD "Failed to build payload for operation 'Volume %s'"
+#define LOGSTR_STAGE_FAIL "Staging of operation 'Volume %s' failed on %s %s %s"
+#define LOGSTR_COMMIT_FAIL "Commit of operation 'Volume %s' failed on %s %s %s"
+
+#define OPERRSTR_BUILD_PAYLOAD "Failed to build payload. Please check the log "\
+                               "file for more details."
+#define OPERRSTR_STAGE_FAIL "Staging failed on %s. Please check the log file " \
+                            "for more details."
+#define OPERRSTR_COMMIT_FAIL "Commit failed on %s. Please check the log file "\
+                             "for more details."
+
+struct glusterd_volinfo_;
+typedef struct glusterd_volinfo_ glusterd_volinfo_t;
 
-
-struct glusterd_store_iter_ {
-        int     fd;
-        FILE    *file;
+struct glusterd_snap_;
+typedef struct glusterd_snap_ glusterd_snap_t;
+
+typedef enum glusterd_op_ {
+        GD_OP_NONE = 0,
+        GD_OP_CREATE_VOLUME,
+        GD_OP_START_BRICK,
+        GD_OP_STOP_BRICK,
+        GD_OP_DELETE_VOLUME,
+        GD_OP_START_VOLUME,
+        GD_OP_STOP_VOLUME,
+        GD_OP_DEFRAG_VOLUME,
+        GD_OP_ADD_BRICK,
+        GD_OP_REMOVE_BRICK,
+        GD_OP_REPLACE_BRICK,
+        GD_OP_SET_VOLUME,
+        GD_OP_RESET_VOLUME,
+        GD_OP_SYNC_VOLUME,
+        GD_OP_LOG_ROTATE,
+        GD_OP_GSYNC_SET,
+        GD_OP_PROFILE_VOLUME,
+        GD_OP_QUOTA,
+        GD_OP_STATUS_VOLUME,
+        GD_OP_REBALANCE,
+        GD_OP_HEAL_VOLUME,
+        GD_OP_STATEDUMP_VOLUME,
+        GD_OP_LIST_VOLUME,
+        GD_OP_CLEARLOCKS_VOLUME,
+        GD_OP_DEFRAG_BRICK_VOLUME,
+        GD_OP_COPY_FILE,
+        GD_OP_SYS_EXEC,
+        GD_OP_GSYNC_CREATE,
+        GD_OP_SNAP,
+        GD_OP_MAX,
+} glusterd_op_t;
+
+extern const char * gd_op_list[];
+
+struct glusterd_volgen {
+        dict_t *dict;
 };
 
-typedef struct glusterd_store_iter_     glusterd_store_iter_t;
+typedef struct {
+        struct rpc_clnt  *rpc;
+        gf_boolean_t      online;
+} nodesrv_t;
 
 typedef struct {
-        struct _volfile_ctx *volfile;
-	pthread_mutex_t   mutex;
-	struct list_head  peers;
-//	struct list_head  pending_peers;
-        gf_boolean_t      verify_volfile_checksum;
-        gf_boolean_t      trace;
-        uuid_t            uuid;
-        char              workdir[PATH_MAX];
-        rpcsvc_t          *rpc;
-        rpc_clnt_prog_t   *mgmt;
-        struct pmap_registry *pmap;
-        struct list_head  volumes;
-        struct list_head  hostnames;
-        struct list_head  xprt_list;
-        glusterd_store_handle_t *handle;
+        gf_boolean_t    quorum;
+        double          quorum_ratio;
+        uint64_t        gl_opt_version;
+} gd_global_opts_t;
+
+typedef struct {
+        struct _volfile_ctx      *volfile;
+        pthread_mutex_t           mutex;
+        struct list_head          peers;
+        struct list_head          xaction_peers;
+        gf_boolean_t              verify_volfile_checksum;
+        gf_boolean_t              trace;
+        uuid_t                    uuid;
+        char                      workdir[PATH_MAX];
+        rpcsvc_t                 *rpc;
+        nodesrv_t                *shd;
+        nodesrv_t                *nfs;
+        struct pmap_registry     *pmap;
+        struct list_head          volumes;
+        struct list_head          snapshots; /*List of snap volumes */
+        pthread_mutex_t           xprt_lock;
+        struct list_head          xprt_list;
+        gf_store_handle_t        *handle;
+        gf_timer_t               *timer;
+        glusterd_sm_tr_log_t      op_sm_log;
+        struct rpc_clnt_program  *gfs_mgmt;
+
+        struct list_head           mount_specs;
+        gf_boolean_t               valgrind;
+        pthread_t                  brick_thread;
+        void                      *hooks_priv;
+        /* need for proper handshake_t */
+        int                        op_version; /* Starts with 1 for 3.3.0 */
+        xlator_t                  *xl;  /* Should be set to 'THIS' before creating thread */
+        gf_boolean_t               pending_quorum_action;
+        dict_t                    *opts;
+        synclock_t                 big_lock;
+        gf_boolean_t               restart_done;
+        rpcsvc_t                  *uds_rpc; /* RPCSVC for the unix domain socket */
+        uint32_t                   base_port;
+        uint64_t                   snap_max_hard_limit;
+        uint64_t                   snap_max_soft_limit;
+        char                      *snap_bricks_directory;
+        gf_store_handle_t         *missed_snaps_list_shandle;
+        struct list_head           missed_snaps_list;
 } glusterd_conf_t;
 
+
+typedef enum gf_brick_status {
+        GF_BRICK_STOPPED,
+        GF_BRICK_STARTED,
+} gf_brick_status_t;
+
 struct glusterd_brickinfo {
-        char    hostname[1024];
-        char    path[PATH_MAX];
-        struct list_head  brick_list;
-        uuid_t  uuid;
-        int     port;
-        char   *logfile;
-        glusterd_store_handle_t *shandle;
+        char               hostname[1024];
+        char               path[PATH_MAX];
+        char               device_path[PATH_MAX];
+        struct list_head   brick_list;
+        uuid_t             uuid;
+        int                port;
+        int                rdma_port;
+        char              *logfile;
+        gf_boolean_t       signed_in;
+        gf_store_handle_t *shandle;
+        gf_brick_status_t  status;
+        struct rpc_clnt   *rpc;
+        int                decommissioned;
+        char vg[PATH_MAX]; /* FIXME: Use max size for length of vg */
+        int     caps; /* Capability */
+        int32_t            snap_status;
 };
 
 typedef struct glusterd_brickinfo glusterd_brickinfo_t;
@@ -93,23 +203,25 @@ struct gf_defrag_brickinfo_ {
         int   size;
 };
 
-typedef enum gf_defrag_status_ {
-        GF_DEFRAG_STATUS_NOT_STARTED,
-        GF_DEFRAG_STATUS_STARTED,
-        GF_DEFRAG_STATUS_STOPED,
-        GF_DEFRAG_STATUS_COMPLETE,
-        GF_DEFRAG_STATUS_FAILED,
-} gf_defrag_status_t;
+typedef int (*defrag_cbk_fn_t) (glusterd_volinfo_t *volinfo,
+                                gf_defrag_status_t status);
 
 struct glusterd_defrag_info_ {
         uint64_t                     total_files;
         uint64_t                     total_data;
         uint64_t                     num_files_lookedup;
+        uint64_t                     total_failures;
         gf_lock_t                    lock;
+        int                          cmd;
         pthread_t                    th;
+        gf_defrag_status_t           defrag_status;
+        struct rpc_clnt             *rpc;
+        uint32_t                     connected;
         char                         mount[1024];
         char                         databuf[131072];
         struct gf_defrag_brickinfo_ *bricks; /* volinfo->brick_count */
+
+        defrag_cbk_fn_t              cbk_fn;
 };
 
 
@@ -118,37 +230,192 @@ typedef struct glusterd_defrag_info_ glusterd_defrag_info_t;
 typedef enum gf_transport_type_ {
         GF_TRANSPORT_TCP,       //DEFAULT
         GF_TRANSPORT_RDMA,
+        GF_TRANSPORT_BOTH_TCP_RDMA,
 } gf_transport_type;
 
+#define GF_DEFAULT_NFS_TRANSPORT  GF_TRANSPORT_RDMA
+
+typedef enum gf_rb_status_ {
+        GF_RB_STATUS_NONE,
+        GF_RB_STATUS_STARTED,
+        GF_RB_STATUS_PAUSED,
+} gf_rb_status_t;
+
+struct _auth {
+        char       *username;
+        char       *password;
+};
+
+typedef struct _auth auth_t;
+
+/* Capabilities of xlator */
+#define CAPS_BD               0x00000001
+#define CAPS_THIN             0x00000002
+#define CAPS_OFFLOAD_COPY     0x00000004
+#define CAPS_OFFLOAD_SNAPSHOT 0x00000008
+
+struct glusterd_rebalance_ {
+        gf_defrag_status_t       defrag_status;
+        uint64_t                 rebalance_files;
+        uint64_t                 rebalance_data;
+        uint64_t                 lookedup_files;
+        uint64_t                 skipped_files;
+        glusterd_defrag_info_t  *defrag;
+        gf_cli_defrag_type       defrag_cmd;
+        uint64_t                 rebalance_failures;
+        uuid_t                   rebalance_id;
+        double                   rebalance_time;
+        glusterd_op_t            op;
+        dict_t                  *dict; /* Dict to store misc information
+                                        * like list of bricks being removed */
+};
+
+typedef struct glusterd_rebalance_ glusterd_rebalance_t;
+
+struct glusterd_replace_brick_ {
+        gf_rb_status_t          rb_status;
+        glusterd_brickinfo_t   *src_brick;
+        glusterd_brickinfo_t   *dst_brick;
+        uuid_t                  rb_id;
+};
+
+typedef struct glusterd_replace_brick_ glusterd_replace_brick_t;
+
 struct glusterd_volinfo_ {
-        char                    volname[GLUSTERD_MAX_VOLUME_NAME];
-        int                     type;
-        int                     brick_count;
-        struct list_head        vol_list;
-        struct list_head        bricks;
-        glusterd_volume_status  status;
-        int                     sub_count;
-        int                     port;
-        glusterd_store_handle_t *shandle;
+        gf_lock_t                 lock;
+        char                      volname[GLUSTERD_MAX_VOLUME_NAME];
+        gf_boolean_t              is_snap_volume;
+        glusterd_snap_t          *snapshot;
+        gf_boolean_t              is_volume_restored;
+        char                      parent_volname[GLUSTERD_MAX_VOLUME_NAME];
+                                         /* In case of a snap volume
+                                            i.e (is_snap_volume == TRUE) this
+                                            field will contain the name of
+                                            the volume which is snapped. In
+                                            case of a non-snap volume, this
+                                            field will be initialized as N/A */
+        int                       type;
+        int                       brick_count;
+        uint64_t                  snap_count;
+        uint64_t                  snap_max_hard_limit;
+        struct list_head          vol_list;
+                                      /* In case of a snap volume
+                                         i.e (is_snap_volume == TRUE) this
+                                         is linked to glusterd_snap_t->volumes.
+                                         In case of a non-snap volume, this is
+                                         linked to glusterd_conf_t->volumes */
+        struct list_head          snapvol_list;
+                                      /* This is a current pointer for
+                                         glusterd_volinfo_t->snap_volumes */
+        struct list_head          bricks;
+        struct list_head          snap_volumes;
+                                      /* TODO : Need to remove this, as this
+                                       * is already part of snapshot object.
+                                       */
+        glusterd_volume_status    status;
+        int                       sub_count;  /* backward compatibility */
+        int                       stripe_count;
+        int                       replica_count;
+        int                       subvol_count; /* Number of subvolumes in a
+                                                 distribute volume */
+        int                       dist_leaf_count; /* Number of bricks in one
+                                                    distribute subvolume */
+        int                       port;
+        gf_store_handle_t        *shandle;
+        gf_store_handle_t        *rb_shandle;
+        gf_store_handle_t        *node_state_shandle;
 
         /* Defrag/rebalance related */
-        gf_defrag_status_t      defrag_status;
-        uint64_t                rebalance_files;
-        uint64_t                rebalance_data;
-        uint64_t                lookedup_files;
-        glusterd_defrag_info_t  *defrag;
+        glusterd_rebalance_t      rebal;
+
+        /* Replace brick status */
+        glusterd_replace_brick_t  rep_brick;
 
-        int                     version;
-        uint32_t                cksum;
-        gf_transport_type       transport_type;
+        int                       version;
+        uint32_t                  cksum;
+        gf_transport_type         transport_type;
+        gf_transport_type         nfs_transport_type;
 
-        /* All xlator options */
-        dict_t                  *dict;
+        dict_t                   *dict;
 
-        uuid_t                  volume_id;
+        uuid_t                    volume_id;
+        auth_t                    auth;
+        char                     *logdir;
+
+        dict_t                   *gsync_slaves;
+
+        int                       decommission_in_progress;
+        xlator_t                 *xl;
+
+        gf_boolean_t              memory_accounting;
+        int                      caps; /* Capability */
+
+        int                       op_version;
+        int                       client_op_version;
 };
 
-typedef struct glusterd_volinfo_ glusterd_volinfo_t;
+typedef enum gd_snap_status_ {
+        GD_SNAP_STATUS_NONE,
+        GD_SNAP_STATUS_INIT,
+        GD_SNAP_STATUS_IN_USE,
+        GD_SNAP_STATUS_DECOMMISSION,
+        GD_SNAP_STATUS_RESTORED,
+} gd_snap_status_t;
+
+struct glusterd_snap_ {
+        gf_lock_t                lock;
+        struct  list_head        volumes;
+        struct  list_head        snap_list;
+        char                     snapname[GLUSTERD_MAX_SNAP_NAME];
+        uuid_t                   snap_id;
+        char                    *description;
+        time_t                   time_stamp;
+        gf_boolean_t             snap_restored;
+        gd_snap_status_t         snap_status;
+        gf_store_handle_t       *shandle;
+};
+
+typedef struct glusterd_snap_op_ {
+        int32_t                brick_num;
+        char                  *brick_path;
+        int32_t                op;
+        int32_t                status;
+        struct list_head       snap_ops_list;
+} glusterd_snap_op_t;
+
+typedef struct glusterd_missed_snap_ {
+        char                   *node_snap_info;
+        struct list_head        missed_snaps;
+        struct list_head        snap_ops;
+} glusterd_missed_snap_info;
+
+typedef enum gd_node_type_ {
+        GD_NODE_NONE,
+        GD_NODE_BRICK,
+        GD_NODE_SHD,
+        GD_NODE_REBALANCE,
+        GD_NODE_NFS,
+} gd_node_type;
+
+typedef enum missed_snap_stat {
+        GD_MISSED_SNAP_NONE,
+        GD_MISSED_SNAP_PENDING,
+        GD_MISSED_SNAP_DONE,
+} missed_snap_stat;
+
+typedef struct glusterd_pending_node_ {
+        struct list_head list;
+        void   *node;
+        gd_node_type type;
+        int32_t index;
+} glusterd_pending_node_t;
+
+struct gsync_config_opt_vals_ {
+        char           *op_name;
+        int             no_of_pos_vals;
+        gf_boolean_t    case_sensitive;
+        char           *values[GEO_CONF_MAX_OPT_VALS];
+};
 
 enum glusterd_op_ret {
         GLUSTERD_CONNECTION_AWAITED = 100,
@@ -161,46 +428,62 @@ enum glusterd_vol_comp_status_ {
         GLUSTERD_VOL_COMP_RJT,
 };
 
-#define GLUSTERD_DEFAULT_WORKDIR "/etc/glusterd"
-#define GLUSTERD_DEFAULT_PORT   6969
+#define GLUSTERD_DEFAULT_WORKDIR "/var/lib/glusterd"
+#define GLUSTERD_DEFAULT_PORT    GF_DEFAULT_BASE_PORT
 #define GLUSTERD_INFO_FILE      "glusterd.info"
 #define GLUSTERD_VOLUME_DIR_PREFIX "vols"
 #define GLUSTERD_PEER_DIR_PREFIX "peers"
 #define GLUSTERD_VOLUME_INFO_FILE "info"
+#define GLUSTERD_SNAP_INFO_FILE "info"
+#define GLUSTERD_VOLUME_RBSTATE_FILE "rbstate"
 #define GLUSTERD_BRICK_INFO_DIR "bricks"
 #define GLUSTERD_CKSUM_FILE "cksum"
+#define GLUSTERD_TRASH "trash"
+#define GLUSTERD_NODE_STATE_FILE "node_state.info"
+#define GLUSTERD_MISSED_SNAPS_LIST_FILE "missed_snaps_list"
+#define GLUSTERD_VOL_SNAP_DIR_PREFIX "snaps"
 
-/*All definitions related to replace brick */
-#define RB_PUMP_START_CMD       "trusted.glusterfs.pump.start"
-#define RB_PUMP_PAUSE_CMD       "trusted.glusterfs.pump.pause"
-#define RB_PUMP_ABORT_CMD       "trusted.glusterfs.pump.abort"
-#define RB_PUMP_STATUS_CMD      "trusted.glusterfs.pump.status"
+#define GLUSTERD_DEFAULT_SNAPS_BRICK_DIR        "/gluster/snaps"
+#define GLUSTERD_VAR_RUN_DIR                    "/var/run"
+#define GLUSTERD_RUN_DIR                        "/run"
+
+/* definitions related to replace brick */
 #define RB_CLIENT_MOUNTPOINT    "rb_mount"
 #define RB_CLIENTVOL_FILENAME   "rb_client.vol"
 #define RB_DSTBRICK_PIDFILE     "rb_dst_brick.pid"
 #define RB_DSTBRICKVOL_FILENAME "rb_dst_brick.vol"
+#define RB_PUMP_DEF_ARG         "default"
 
 #define GLUSTERD_UUID_LEN 50
 
 typedef ssize_t (*gd_serialize_t) (struct iovec outmsg, void *args);
 
-#define GLUSTERD_GET_NFS_DIR(path, priv)                                \
-        do {                                                            \
-                snprintf (path, PATH_MAX, "%s/nfs", priv->workdir);\
-        } while (0);                                                    \
-
-#define GLUSTERD_GET_VOLUME_DIR(path, volinfo, priv) \
-        snprintf (path, PATH_MAX, "%s/vols/%s", priv->workdir,\
-                  volinfo->volname);
+#define GLUSTERD_GET_VOLUME_DIR(path, volinfo, priv)                       \
+        if (volinfo->is_snap_volume) {                                     \
+                snprintf (path, PATH_MAX, "%s/snaps/%s/%s", priv->workdir, \
+                          volinfo->snapshot->snapname, volinfo->volname);  \
+        } else {                                                           \
+                snprintf (path, PATH_MAX, "%s/vols/%s", priv->workdir,     \
+                          volinfo->volname);                               \
+        }
 
-#define GLUSTERD_GET_BRICK_DIR(path, volinfo, priv) \
-        snprintf (path, PATH_MAX, "%s/%s/%s/%s", priv->workdir,\
-                  GLUSTERD_VOLUME_DIR_PREFIX, volinfo->volname, \
-                  GLUSTERD_BRICK_INFO_DIR);
+#define GLUSTERD_GET_SNAP_DIR(path, snap, priv)                           \
+                snprintf (path, PATH_MAX, "%s/snaps/%s", priv->workdir,   \
+                          snap->snapname);
+
+#define GLUSTERD_GET_BRICK_DIR(path, volinfo, priv)                           \
+        if (volinfo->is_snap_volume) {                                        \
+                snprintf (path, PATH_MAX, "%s/snaps/%s/%s/%s", priv->workdir, \
+                          volinfo->snapshot->snapname, volinfo->volname,      \
+                          GLUSTERD_BRICK_INFO_DIR);                           \
+        } else {                                                              \
+                snprintf (path, PATH_MAX, "%s/%s/%s/%s", priv->workdir,       \
+                          GLUSTERD_VOLUME_DIR_PREFIX, volinfo->volname,       \
+                          GLUSTERD_BRICK_INFO_DIR);                           \
+        }
 
-#define GLUSTERD_GET_NFS_PIDFILE(pidfile)                               \
-                snprintf (pidfile, PATH_MAX, "%s/nfs/run/nfs.pid", \
-                          priv->workdir);                               \
+#define GLUSTERD_GET_NFS_DIR(path, priv) \
+        snprintf (path, PATH_MAX, "%s/nfs", priv->workdir);
 
 #define GLUSTERD_REMOVE_SLASH_FROM_PATH(path,string) do {               \
                 int i = 0;                                              \
@@ -211,22 +494,96 @@ typedef ssize_t (*gd_serialize_t) (struct iovec outmsg, void *args);
                 }                                                       \
         } while (0)
 
-#define GLUSTERD_GET_BRICK_PIDFILE(pidfile,volpath,hostname,brickpath) { \
-                char exp_path[PATH_MAX] = {0,};                         \
-                GLUSTERD_REMOVE_SLASH_FROM_PATH (brickpath, exp_path);  \
-                snprintf (pidfile, PATH_MAX, "%s/run/%s-%s.pid",        \
-                          volpath, hostname, exp_path);                 \
+#define GLUSTERD_GET_BRICK_PIDFILE(pidfile,volinfo,brickinfo, priv) do {      \
+                char exp_path[PATH_MAX] = {0,};                               \
+                char volpath[PATH_MAX]  = {0,};                               \
+                GLUSTERD_GET_VOLUME_DIR (volpath, volinfo, priv);             \
+                GLUSTERD_REMOVE_SLASH_FROM_PATH (brickinfo->path, exp_path);  \
+                snprintf (pidfile, PATH_MAX, "%s/run/%s-%s.pid",              \
+                          volpath, brickinfo->hostname, exp_path);      \
+        } while (0)
+
+#define GLUSTERD_GET_NFS_PIDFILE(pidfile,nfspath) {                     \
+                snprintf (pidfile, PATH_MAX, "%s/run/nfs.pid",          \
+                          nfspath);                                     \
         }
 
+#define GLUSTERD_STACK_DESTROY(frame) do {\
+                frame->local = NULL;                                    \
+                STACK_DESTROY (frame->root);                            \
+        } while (0)
+
+#define GLUSTERD_GET_DEFRAG_DIR(path, volinfo, priv) do {               \
+                char vol_path[PATH_MAX];                                \
+                GLUSTERD_GET_VOLUME_DIR(vol_path, volinfo, priv);       \
+                snprintf (path, PATH_MAX, "%s/rebalance",vol_path);     \
+        } while (0)
+
+#define GLUSTERD_GET_DEFRAG_SOCK_FILE(path, volinfo, priv) do {         \
+                char defrag_path[PATH_MAX];                             \
+                GLUSTERD_GET_DEFRAG_DIR(defrag_path, volinfo, priv);    \
+                snprintf (path, PATH_MAX, "%s/%s.sock", defrag_path,    \
+                           uuid_utoa(MY_UUID));                         \
+        } while (0)
+
+#define GLUSTERD_GET_DEFRAG_PID_FILE(path, volinfo, priv) do {          \
+                char defrag_path[PATH_MAX];                             \
+                GLUSTERD_GET_DEFRAG_DIR(defrag_path, volinfo, priv);    \
+                snprintf (path, PATH_MAX, "%s/%s.pid", defrag_path,     \
+                           uuid_utoa(MY_UUID));                         \
+        } while (0)
+
+#define GLUSTERD_GET_UUID_NOHYPHEN(ret_string, uuid) do {               \
+                char *snap_volname_ptr = ret_string;                    \
+                char *snap_volid_ptr = uuid_utoa(uuid);                 \
+                while (*snap_volid_ptr) {                               \
+                        if (*snap_volid_ptr == '-') {                   \
+                                snap_volid_ptr++;                       \
+                        } else {                                        \
+                                (*snap_volname_ptr++) =                 \
+                                (*snap_volid_ptr++);                    \
+                        }                                               \
+                }                                                       \
+                *snap_volname_ptr = '\0';                               \
+        } while (0)
+
+int glusterd_uuid_init();
+
+int glusterd_uuid_generate_save ();
+
+#define MY_UUID (__glusterd_uuid())
+
+static inline unsigned char *
+__glusterd_uuid()
+{
+        glusterd_conf_t *priv = THIS->private;
+
+        if (uuid_is_null (priv->uuid))
+                glusterd_uuid_init();
+        return &priv->uuid[0];
+}
+
+int glusterd_big_locked_notify (struct rpc_clnt *rpc, void *mydata,
+                                rpc_clnt_event_t event,
+                                void *data, rpc_clnt_notify_t notify_fn);
+
+int
+glusterd_big_locked_cbk (struct rpc_req *req, struct iovec *iov,
+                         int count, void *myframe, fop_cbk_fn_t fn);
+
+int glusterd_big_locked_handler (rpcsvc_request_t *req, rpcsvc_actor actor_fn);
+
 int32_t
 glusterd_brick_from_brickinfo (glusterd_brickinfo_t *brickinfo,
                                char **new_brick);
 int
-glusterd_probe_begin (rpcsvc_request_t *req, const char *hoststr, int port);
+glusterd_probe_begin (rpcsvc_request_t *req, const char *hoststr, int port,
+                      dict_t *dict);
 
 int
-glusterd_xfer_friend_add_resp (rpcsvc_request_t *req, char *hostname,
-                               int port, int32_t op_ret, int32_t op_errno);
+glusterd_xfer_friend_add_resp (rpcsvc_request_t *req, char *myhostname,
+                               char *remote_hostname, int port, int32_t op_ret,
+                               int32_t op_errno);
 
 int
 glusterd_friend_find (uuid_t uuid, char *hostname,
@@ -235,11 +592,13 @@ glusterd_friend_find (uuid_t uuid, char *hostname,
 int
 glusterd_friend_add (const char *hoststr, int port,
                      glusterd_friend_sm_state_t state,
-                     uuid_t *uuid, struct rpc_clnt    *rpc,
-                     glusterd_peerinfo_t **friend,
-                     gf_boolean_t restore);
+                     uuid_t *uuid, glusterd_peerinfo_t **friend,
+                     gf_boolean_t restore, glusterd_peerctx_args_t *args);
 
 int
+glusterd_friend_rpc_create (xlator_t *this, glusterd_peerinfo_t *peerinfo,
+                            glusterd_peerctx_args_t *args);
+int
 glusterd_friend_remove (uuid_t uuid, char *hostname);
 
 int
@@ -249,8 +608,17 @@ int
 glusterd_op_unlock_send_resp (rpcsvc_request_t *req, int32_t status);
 
 int
+glusterd_op_mgmt_v3_lock_send_resp (rpcsvc_request_t *req,
+                                    uuid_t *txn_id, int32_t status);
+
+int
+glusterd_op_mgmt_v3_unlock_send_resp (rpcsvc_request_t *req,
+                                      uuid_t *txn_id, int32_t status);
+
+int
 glusterd_op_stage_send_resp (rpcsvc_request_t *req,
-                             int32_t op, int32_t status, char *op_errstr);
+                             int32_t op, int32_t status,
+                             char *op_errstr, dict_t *rsp_dict);
 
 int
 glusterd_op_commmit_send_resp (rpcsvc_request_t *req,
@@ -260,10 +628,6 @@ int32_t
 glusterd_create_volume (rpcsvc_request_t *req, dict_t *dict);
 
 int
-glusterd_rpc_notify (struct rpc_clnt *rpc, void *mydata,
-                     rpc_clnt_event_t event,
-                     void *data);
-int
 glusterd_handle_incoming_friend_req (rpcsvc_request_t *req);
 
 int
@@ -291,8 +655,12 @@ int
 glusterd_handle_defrag_volume (rpcsvc_request_t *req);
 
 int
+glusterd_handle_defrag_volume_v2 (rpcsvc_request_t *req);
+
+int
 glusterd_xfer_cli_probe_resp (rpcsvc_request_t *req, int32_t op_ret,
-                              int32_t op_errno, char *hostname, int port);
+                              int32_t op_errno, char *op_errstr, char *hostname,
+                              int port, dict_t *dict);
 
 int
 glusterd_op_commit_send_resp (rpcsvc_request_t *req,
@@ -303,7 +671,8 @@ int
 glusterd_xfer_friend_remove_resp (rpcsvc_request_t *req, char *hostname, int port);
 
 int
-glusterd_deprobe_begin (rpcsvc_request_t *req, const char *hoststr, int port);
+glusterd_deprobe_begin (rpcsvc_request_t *req, const char *hoststr, int port,
+                        uuid_t uuid, dict_t *dict);
 
 int
 glusterd_handle_cli_deprobe (rpcsvc_request_t *req);
@@ -320,9 +689,6 @@ glusterd_handle_cli_list_friends (rpcsvc_request_t *req);
 int
 glusterd_handle_cli_start_volume (rpcsvc_request_t *req);
 
-int32_t
-glusterd_start_volume (rpcsvc_request_t *req, char *volname, int flags);
-
 int
 glusterd_handle_friend_update (rpcsvc_request_t *req);
 
@@ -330,12 +696,6 @@ int
 glusterd_handle_cli_stop_volume (rpcsvc_request_t *req);
 
 int
-glusterd_stop_volume (rpcsvc_request_t *req, char *volname, int flags);
-
-int32_t
-glusterd_delete_volume (rpcsvc_request_t *req, char *volname, int flags);
-
-int
 glusterd_handle_cli_delete_volume (rpcsvc_request_t *req);
 
 int
@@ -344,15 +704,9 @@ glusterd_handle_cli_get_volume (rpcsvc_request_t *req);
 int32_t
 glusterd_get_volumes (rpcsvc_request_t *req, dict_t *dict, int32_t flags);
 
-int32_t
-glusterd_add_brick (rpcsvc_request_t *req, dict_t *dict);
-
 int
 glusterd_handle_add_brick (rpcsvc_request_t *req);
 
-int32_t
-glusterd_replace_brick (rpcsvc_request_t *req, dict_t *dict);
-
 int
 glusterd_handle_replace_brick (rpcsvc_request_t *req);
 
@@ -360,10 +714,6 @@ int
 glusterd_handle_remove_brick (rpcsvc_request_t *req);
 
 int
-glusterd_handle_log_filename (rpcsvc_request_t *req);
-int
-glusterd_handle_log_locate (rpcsvc_request_t *req);
-int
 glusterd_handle_log_rotate (rpcsvc_request_t *req);
 
 int
@@ -381,19 +731,222 @@ glusterd_remove_brick (rpcsvc_request_t *req, dict_t *dict);
 int32_t
 glusterd_set_volume (rpcsvc_request_t *req, dict_t *dict);
 
+int32_t
+glusterd_reset_volume (rpcsvc_request_t *req, dict_t *dict);
+
+int32_t
+glusterd_gsync_set (rpcsvc_request_t *req, dict_t *dict);
+
+int32_t
+glusterd_quota (rpcsvc_request_t *req, dict_t *dict);
+
 int
 glusterd_handle_set_volume (rpcsvc_request_t *req);
 
 int
+glusterd_handle_reset_volume (rpcsvc_request_t *req);
+
+int
+glusterd_handle_copy_file (rpcsvc_request_t *req);
+
+int
+glusterd_handle_sys_exec (rpcsvc_request_t *req);
+
+int
+glusterd_handle_gsync_set (rpcsvc_request_t *req);
+
+int
+glusterd_handle_quota (rpcsvc_request_t *req);
+
+int
+glusterd_handle_fsm_log (rpcsvc_request_t *req);
+
+int
 glusterd_xfer_cli_deprobe_resp (rpcsvc_request_t *req, int32_t op_ret,
-                                int32_t op_errno, char *hostname);
+                                int32_t op_errno, char *op_errstr,
+                                char *hostname, dict_t *dict);
 
 int
 glusterd_fetchspec_notify (xlator_t *this);
 
-int32_t
-glusterd_sync_volume (rpcsvc_request_t *req, dict_t *ctx);
 int
 glusterd_add_volume_detail_to_dict (glusterd_volinfo_t *volinfo,
                                     dict_t  *volumes, int   count);
+
+int
+glusterd_restart_bricks (glusterd_conf_t *conf);
+
+int32_t
+glusterd_volume_txn (rpcsvc_request_t *req, char *volname, int flags,
+                     glusterd_op_t op);
+
+int
+glusterd_peer_dump_version (xlator_t *this, struct rpc_clnt *rpc,
+                            glusterd_peerctx_t *peerctx);
+
+int
+glusterd_validate_reconfopts (glusterd_volinfo_t *volinfo, dict_t *val_dict, char **op_errstr);
+int
+glusterd_handle_cli_profile_volume (rpcsvc_request_t *req);
+
+int
+glusterd_handle_getwd (rpcsvc_request_t *req);
+
+int32_t
+glusterd_set_volume (rpcsvc_request_t *req, dict_t *dict);
+int
+glusterd_peer_rpc_notify (struct rpc_clnt *rpc, void *mydata,
+                          rpc_clnt_event_t event,
+                          void *data);
+int
+glusterd_brick_rpc_notify (struct rpc_clnt *rpc, void *mydata,
+                          rpc_clnt_event_t event, void *data);
+
+int
+glusterd_nodesvc_rpc_notify (struct rpc_clnt *rpc, void *mydata,
+                             rpc_clnt_event_t event, void *data);
+
+int
+glusterd_rpc_create (struct rpc_clnt **rpc, dict_t *options,
+                     rpc_clnt_notify_t notify_fn, void *notify_data);
+
+
+/* handler functions */
+int32_t glusterd_op_begin (rpcsvc_request_t *req, glusterd_op_t op, void *ctx,
+                           char *err_str, size_t size);
+
+/* removed other definitions as they have been defined elsewhere in this file*/
+
+int glusterd_handle_cli_statedump_volume (rpcsvc_request_t *req);
+int glusterd_handle_cli_clearlocks_volume (rpcsvc_request_t *req);
+
+int glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr,
+                                  size_t len, int cmd, defrag_cbk_fn_t cbk,
+                                  glusterd_op_t op);
+int
+glusterd_rebalance_rpc_create (glusterd_volinfo_t *volinfo,
+                               glusterd_conf_t *priv, int cmd);
+
+int glusterd_handle_cli_heal_volume (rpcsvc_request_t *req);
+
+int glusterd_handle_cli_list_volume (rpcsvc_request_t *req);
+
+int
+glusterd_handle_snapshot (rpcsvc_request_t *req);
+
+/* op-sm functions */
+int glusterd_op_stage_heal_volume (dict_t *dict, char **op_errstr);
+int glusterd_op_heal_volume (dict_t *dict, char **op_errstr);
+int glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr);
+int glusterd_op_gsync_set (dict_t *dict, char **op_errstr, dict_t *rsp_dict);
+int glusterd_op_stage_copy_file (dict_t *dict, char **op_errstr);
+int glusterd_op_copy_file (dict_t *dict, char **op_errstr);
+int glusterd_op_stage_sys_exec (dict_t *dict, char **op_errstr);
+int glusterd_op_sys_exec (dict_t *dict, char **op_errstr, dict_t *rsp_dict);
+int glusterd_op_stage_gsync_create (dict_t *dict, char **op_errstr);
+int glusterd_op_gsync_create (dict_t *dict, char **op_errstr, dict_t *rsp_dict);
+int glusterd_op_quota (dict_t *dict, char **op_errstr, dict_t *rsp_dict);
+int glusterd_op_stage_quota (dict_t *dict, char **op_errstr);
+int glusterd_op_stage_replace_brick (dict_t *dict, char **op_errstr,
+                                     dict_t *rsp_dict);
+int glusterd_op_replace_brick (dict_t *dict, dict_t *rsp_dict);
+int glusterd_op_log_rotate (dict_t *dict);
+int glusterd_op_stage_log_rotate (dict_t *dict, char **op_errstr);
+int glusterd_op_stage_create_volume (dict_t *dict, char **op_errstr);
+int glusterd_op_stage_start_volume (dict_t *dict, char **op_errstr);
+int glusterd_op_stage_stop_volume (dict_t *dict, char **op_errstr);
+int glusterd_op_stage_delete_volume (dict_t *dict, char **op_errstr);
+int glusterd_op_create_volume (dict_t *dict, char **op_errstr);
+int glusterd_op_start_volume (dict_t *dict, char **op_errstr);
+int glusterd_op_stop_volume (dict_t *dict);
+int glusterd_op_delete_volume (dict_t *dict);
+
+int glusterd_op_add_brick (dict_t *dict, char **op_errstr);
+int glusterd_op_remove_brick (dict_t *dict, char **op_errstr);
+int glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr);
+int glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr);
+
+int glusterd_op_stage_rebalance (dict_t *dict, char **op_errstr);
+int glusterd_op_rebalance (dict_t *dict, char **op_errstr, dict_t *rsp_dict);
+
+int glusterd_op_stage_statedump_volume (dict_t *dict, char **op_errstr);
+int glusterd_op_statedump_volume (dict_t *dict, char **op_errstr);
+
+int glusterd_op_stage_clearlocks_volume (dict_t *dict, char **op_errstr);
+int glusterd_op_clearlocks_volume (dict_t *dict, char **op_errstr,
+                                   dict_t *rsp_dict);
+
+/* misc */
+void glusterd_do_replace_brick (void *data);
+int glusterd_op_perform_remove_brick (glusterd_volinfo_t  *volinfo, char *brick,
+                                      int force, int *need_migrate);
+int glusterd_op_stop_volume_args_get (dict_t *dict, char** volname, int *flags);
+int glusterd_op_statedump_volume_args_get (dict_t *dict, char **volname,
+                                           char **options, int *option_cnt);
+
+int glusterd_op_gsync_args_get (dict_t *dict, char **op_errstr,
+                                char **master, char **slave, char **host_uuid);
+int glusterd_stop_volume (glusterd_volinfo_t *volinfo);
+
+/* Synctask part */
+int32_t glusterd_op_begin_synctask (rpcsvc_request_t *req, glusterd_op_t op,
+                                    void *dict);
+int32_t
+glusterd_defrag_event_notify_handle (dict_t *dict);
+
+int32_t
+glusterd_txn_opinfo_dict_init ();
+
+void
+glusterd_txn_opinfo_dict_fini ();
+
+void
+glusterd_txn_opinfo_init ();
+
+/* snapshot */
+glusterd_snap_t*
+glusterd_new_snap_object();
+
+int32_t
+glusterd_list_add_snapvol (glusterd_volinfo_t *origin_vol,
+                           glusterd_volinfo_t *snap_vol);
+
+glusterd_snap_t*
+glusterd_remove_snap_by_id (uuid_t snap_id);
+
+glusterd_snap_t*
+glusterd_remove_snap_by_name (char *snap_name);
+
+glusterd_snap_t*
+glusterd_find_snap_by_name (char *snap_name);
+
+glusterd_snap_t*
+glusterd_find_snap_by_id (uuid_t snap_id);
+
+int
+glusterd_snapshot_prevalidate (dict_t *dict, char **op_errstr,
+                               dict_t *rsp_dict);
+int
+glusterd_snapshot_brickop (dict_t *dict, char **op_errstr, dict_t *rsp_dict);
+int
+glusterd_snapshot (dict_t *dict, char **op_errstr, dict_t *rsp_dict);
+int
+glusterd_snapshot_postvalidate (dict_t *dict, int32_t op_ret, char **op_errstr,
+                                dict_t *rsp_dict);
+char *
+glusterd_build_snap_device_path (char *device, char *snapname);
+int32_t
+glusterd_snap_remove (dict_t *rsp_dict, glusterd_snap_t *snap,
+                      gf_boolean_t remove_lvm, gf_boolean_t force);
+int32_t
+glusterd_snapshot_cleanup (dict_t *dict, char **op_errstr, dict_t *rsp_dict);
+
+int32_t
+glusterd_add_missed_snaps_to_list (dict_t *dict, int32_t missed_snap_count);
+
+int32_t
+glusterd_store_missed_snaps_list (char *missed_info, int32_t brick_num,
+                                  char *brick_path, int32_t snap_op,
+                                  int32_t snap_status);
+
 #endif
diff --git a/xlators/mgmt/glusterd/src/glusterd3_1-mops.c b/xlators/mgmt/glusterd/src/glusterd3_1-mops.c
deleted file mode 100644
index 2d0b223a9..000000000
--- a/xlators/mgmt/glusterd/src/glusterd3_1-mops.c
+++ /dev/null
@@ -1,1479 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "rpc-clnt.h"
-#include "glusterd1-xdr.h"
-#include "glusterd1.h"
-
-#include "compat-errno.h"
-#include "glusterd-op-sm.h"
-#include "glusterd-sm.h"
-#include "glusterd.h"
-#include "protocol-common.h"
-#include "glusterd-utils.h"
-#include <sys/uio.h>
-
-
-#define SERVER_PATH_MAX  (16 * 1024)
-
-
-extern glusterd_op_info_t    opinfo;
-
-int
-glusterd_null (rpcsvc_request_t *req)
-{
-
-        return 0;
-}
-
-int
-glusterd3_1_probe_cbk (struct rpc_req *req, struct iovec *iov,
-                        int count, void *myframe)
-{
-        gd1_mgmt_probe_rsp    rsp   = {{0},};
-        glusterd_conf_t       *conf = NULL;
-        int                   ret   = 0;
-        char                  str[50] = {0,};
-        glusterd_peerinfo_t           *peerinfo = NULL;
-        glusterd_peerinfo_t           *dup_peerinfo = NULL;
-        glusterd_friend_sm_event_t    *event = NULL;
-        glusterd_peer_hostname_t      *name = NULL;
-        glusterd_probe_ctx_t          *ctx = NULL;
-
-        conf  = THIS->private;
-
-        if (-1 == req->rpc_status) {
-                goto out;
-        }
-
-        ret = gd_xdr_to_mgmt_probe_rsp (*iov, &rsp);
-        if (ret < 0) {
-                gf_log ("", GF_LOG_ERROR, "error");
-                //rsp.op_ret   = -1;
-                //rsp.op_errno = EINVAL;
-                goto out;
-        }
-        uuid_unparse (rsp.uuid, str);
-
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received probe resp from uuid: %s, host: %s",
-                str, rsp.hostname);
-        if (rsp.op_ret != 0) {
-                ctx = ((call_frame_t *)myframe)->local;
-                ((call_frame_t *)myframe)->local = NULL;
-
-                GF_ASSERT (ctx);
-
-                if (ctx->req) {
-                        glusterd_xfer_cli_probe_resp (ctx->req, rsp.op_ret,
-                                                      rsp.op_errno,
-                                                      ctx->hostname, ctx->port);
-                }
-
-                ret = rsp.op_ret;
-                (void) glusterd_friend_remove (rsp.uuid, rsp.hostname);
-                goto out;
-        }
-        ret = glusterd_friend_find (rsp.uuid, rsp.hostname, &peerinfo);
-        if (ret) {
-                GF_ASSERT (0);
-        }
-
-        if (list_empty (&peerinfo->hostnames)) {
-                glusterd_friend_find (NULL, rsp.hostname, &dup_peerinfo);
-                GF_ASSERT (dup_peerinfo);
-                peerinfo->hostname = gf_strdup (rsp.hostname);
-                glusterd_peer_hostname_new (rsp.hostname, &name);
-                list_add_tail (&name->hostname_list, &peerinfo->hostnames);
-                peerinfo->rpc = dup_peerinfo->rpc;
-                peerinfo->connected = dup_peerinfo->connected;
-                glusterd_peer_destroy  (dup_peerinfo);
-        }
-        if (!peerinfo->hostname)
-                peerinfo->hostname = gf_strdup (rsp.hostname);
-        uuid_copy (peerinfo->uuid, rsp.uuid);
-
-        ret = glusterd_friend_sm_new_event
-                        (GD_FRIEND_EVENT_INIT_FRIEND_REQ, &event);
-
-        if (ret) {
-                gf_log ("glusterd", GF_LOG_ERROR,
-                         "Unable to get event");
-                goto out;
-        }
-
-        event->peerinfo = peerinfo;
-        event->ctx      = ((call_frame_t *)myframe)->local;
-        ((call_frame_t *)myframe)->local = NULL;
-        ret = glusterd_friend_sm_inject_event (event);
-
-
-        if (!ret) {
-                glusterd_friend_sm ();
-                glusterd_op_sm ();
-        }
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Received resp to probe req");
-
-out:
-        if (rsp.hostname)
-                free (rsp.hostname);//malloced by xdr
-        return ret;
-}
-
-int
-glusterd3_1_friend_add_cbk (struct rpc_req * req, struct iovec *iov,
-                            int count, void *myframe)
-{
-        gd1_mgmt_friend_rsp           rsp   = {{0},};
-        glusterd_conf_t               *conf = NULL;
-        int                           ret   = -1;
-        glusterd_friend_sm_event_t        *event = NULL;
-        glusterd_friend_sm_event_type_t    event_type = GD_FRIEND_EVENT_NONE;
-        glusterd_peerinfo_t           *peerinfo = NULL;
-        char                          str[50] = {0,};
-        int32_t                       op_ret = -1;
-        int32_t                       op_errno = -1;
-        glusterd_probe_ctx_t          *ctx = NULL;
-        glusterd_friend_update_ctx_t  *ev_ctx = NULL;
-
-        conf  = THIS->private;
-
-        if (-1 == req->rpc_status) {
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                goto out;
-        }
-
-        ret = gd_xdr_to_mgmt_friend_rsp (*iov, &rsp);
-        if (ret < 0) {
-                gf_log ("", GF_LOG_ERROR, "error");
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                goto out;
-        }
-        uuid_unparse (rsp.uuid, str);
-
-        op_ret = rsp.op_ret;
-        op_errno = rsp.op_errno;
-
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received %s from uuid: %s, host: %s, port: %d",
-                (op_ret)?"RJT":"ACC", str, rsp.hostname, rsp.port);
-
-        ret = glusterd_friend_find (rsp.uuid, rsp.hostname, &peerinfo);
-
-        if (ret) {
-                GF_ASSERT (0);
-        }
-
-        if (op_ret)
-                event_type = GD_FRIEND_EVENT_RCVD_RJT;
-        else
-                event_type = GD_FRIEND_EVENT_RCVD_ACC;
-
-        ret = glusterd_friend_sm_new_event (event_type, &event);
-
-        if (ret) {
-                gf_log ("glusterd", GF_LOG_ERROR,
-                         "Unable to get event");
-                goto out;
-        }
-        event->peerinfo = peerinfo;
-        ev_ctx = GF_CALLOC (1, sizeof (*ev_ctx),
-                                gf_gld_mt_friend_update_ctx_t);
-        if (!ev_ctx) {
-                ret = -1;
-                goto out;
-        }
-
-        uuid_copy (ev_ctx->uuid, rsp.uuid);
-        ev_ctx->hostname = gf_strdup (rsp.hostname);
-
-        event->ctx = ev_ctx;
-        ret = glusterd_friend_sm_inject_event (event);
-
-        if (ret)
-                goto out;
-
-        ctx = ((call_frame_t *)myframe)->local;
-        ((call_frame_t *)myframe)->local = NULL;
-
-        GF_ASSERT (ctx);
-
-        if (ctx->req)//reverse probe doesnt have req
-                ret = glusterd_xfer_cli_probe_resp (ctx->req, op_ret, op_errno,
-                                                    ctx->hostname, ctx->port);
-        if (!ret) {
-                glusterd_friend_sm ();
-                glusterd_op_sm ();
-        }
-        if (ctx)
-                glusterd_destroy_probe_ctx (ctx);
-out:
-        if (rsp.hostname)
-                free (rsp.hostname);//malloced by xdr
-        return ret;
-}
-
-int
-glusterd3_1_friend_remove_cbk (struct rpc_req * req, struct iovec *iov,
-                               int count, void *myframe)
-{
-        gd1_mgmt_friend_rsp             rsp   = {{0},};
-        glusterd_conf_t                 *conf = NULL;
-        int                             ret   = -1;
-        glusterd_friend_sm_event_t      *event = NULL;
-        glusterd_friend_sm_event_type_t event_type = GD_FRIEND_EVENT_NONE;
-        glusterd_peerinfo_t             *peerinfo = NULL;
-        char                            str[50] = {0,};
-        int32_t                         op_ret = -1;
-        int32_t                         op_errno = -1;
-        glusterd_probe_ctx_t            *ctx = NULL;
-
-        conf  = THIS->private;
-        GF_ASSERT (conf);
-
-        ctx = ((call_frame_t *)myframe)->local;
-        ((call_frame_t *)myframe)->local = NULL;
-        GF_ASSERT (ctx);
-
-        if (-1 == req->rpc_status) {
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                goto inject;
-        }
-
-        ret = gd_xdr_to_mgmt_friend_rsp (*iov, &rsp);
-        if (ret < 0) {
-                gf_log ("", GF_LOG_ERROR, "error");
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                goto respond;
-        }
-        uuid_unparse (rsp.uuid, str);
-
-        op_ret = rsp.op_ret;
-        op_errno = rsp.op_errno;
-
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received %s from uuid: %s, host: %s, port: %d",
-                (op_ret)?"RJT":"ACC", str, rsp.hostname, rsp.port);
-
-        if (op_ret)
-                goto respond;
-
-inject:
-        ret = glusterd_friend_find (rsp.uuid, ctx->hostname, &peerinfo);
-
-        if (ret) {
-                //can happen as part of rpc clnt connection cleanup
-                //when the frame timeout happens after 30 minutes
-                goto respond;
-        }
-
-        event_type = GD_FRIEND_EVENT_REMOVE_FRIEND;
-
-        ret = glusterd_friend_sm_new_event (event_type, &event);
-
-        if (ret) {
-                gf_log ("glusterd", GF_LOG_ERROR,
-                         "Unable to get event");
-                goto respond;
-        }
-        event->peerinfo = peerinfo;
-
-        ret = glusterd_friend_sm_inject_event (event);
-
-        if (ret)
-                goto respond;
-
-        glusterd_friend_sm ();
-        glusterd_op_sm ();
-
-        op_ret = 0;
-
-
-respond:
-        ret = glusterd_xfer_cli_deprobe_resp (ctx->req, op_ret, op_errno,
-                                              ctx->hostname);
-        if (!ret) {
-                glusterd_friend_sm ();
-                glusterd_op_sm ();
-        }
-
-        if (ctx)
-                glusterd_destroy_probe_ctx (ctx);
-
-        if (rsp.hostname)
-                free (rsp.hostname);//malloced by xdr
-        return ret;
-}
-
-int32_t
-glusterd3_1_friend_update_cbk (struct rpc_req *req, struct iovec *iov,
-                              int count, void *myframe)
-{
-        gd1_mgmt_cluster_lock_rsp     rsp   = {{0},};
-        int                           ret   = -1;
-        int32_t                       op_ret = -1;
-        char                          str[50] = {0,};
-
-        GF_ASSERT (req);
-
-        if (-1 == req->rpc_status) {
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                goto out;
-        }
-
-/*        ret = gd_xdr_to_mgmt_friend_update_rsp (*iov, &rsp);
-        if (ret < 0) {
-                gf_log ("", GF_LOG_ERROR, "error");
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                goto out;
-        }
-        uuid_unparse (rsp.uuid, str);
-
-        op_ret = rsp.op_ret;
-*/
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received %s from uuid: %s",
-                (op_ret)?"RJT":"ACC", str);
-
-out:
-        return ret;
-}
-int32_t
-glusterd3_1_cluster_lock_cbk (struct rpc_req *req, struct iovec *iov,
-                              int count, void *myframe)
-{
-        gd1_mgmt_cluster_lock_rsp     rsp   = {{0},};
-        int                           ret   = -1;
-        int32_t                       op_ret = -1;
-        glusterd_op_sm_event_type_t   event_type = GD_OP_EVENT_NONE;
-        glusterd_peerinfo_t           *peerinfo = NULL;
-        char                          str[50] = {0,};
-
-        GF_ASSERT (req);
-
-        if (-1 == req->rpc_status) {
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                goto out;
-        }
-
-        ret = gd_xdr_to_mgmt_cluster_lock_rsp (*iov, &rsp);
-        if (ret < 0) {
-                gf_log ("", GF_LOG_ERROR, "error");
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                goto out;
-        }
-        uuid_unparse (rsp.uuid, str);
-
-        op_ret = rsp.op_ret;
-
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received %s from uuid: %s",
-                (op_ret)?"RJT":"ACC", str);
-
-        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
-
-        if (ret) {
-                gf_log ("", GF_LOG_CRITICAL, "Lock response received from "
-                        "unknown peer: %s", str);
-        }
-
-        if (op_ret) {
-                event_type = GD_OP_EVENT_RCVD_RJT;
-                opinfo.op_ret = op_ret;
-        } else {
-                event_type = GD_OP_EVENT_RCVD_ACC;
-        }
-
-        ret = glusterd_op_sm_inject_event (event_type, NULL);
-
-        if (!ret) {
-                glusterd_friend_sm ();
-                glusterd_op_sm ();
-        }
-
-out:
-        return ret;
-}
-
-int32_t
-glusterd3_1_cluster_unlock_cbk (struct rpc_req *req, struct iovec *iov,
-                                 int count, void *myframe)
-{
-        gd1_mgmt_cluster_lock_rsp     rsp   = {{0},};
-        int                           ret   = -1;
-        int32_t                       op_ret = -1;
-        glusterd_op_sm_event_type_t   event_type = GD_OP_EVENT_NONE;
-        glusterd_peerinfo_t           *peerinfo = NULL;
-        char                          str[50] = {0,};
-
-
-        GF_ASSERT (req);
-
-        if (-1 == req->rpc_status) {
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                goto out;
-        }
-
-        ret = gd_xdr_to_mgmt_cluster_unlock_rsp (*iov, &rsp);
-        if (ret < 0) {
-                gf_log ("", GF_LOG_ERROR, "error");
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                goto out;
-        }
-        uuid_unparse (rsp.uuid, str);
-
-        op_ret = rsp.op_ret;
-
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received %s from uuid: %s",
-                (op_ret)?"RJT":"ACC", str);
-
-        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
-
-        if (ret) {
-                gf_log ("", GF_LOG_CRITICAL, "Unlock response received from "
-                        "unknown peer %s", str);
-        }
-
-        if (op_ret) {
-                event_type = GD_OP_EVENT_RCVD_RJT;
-                opinfo.op_ret = op_ret;
-        } else {
-                event_type = GD_OP_EVENT_RCVD_ACC;
-        }
-
-        ret = glusterd_op_sm_inject_event (event_type, NULL);
-
-        if (!ret) {
-                glusterd_friend_sm ();
-                glusterd_op_sm ();
-        }
-
-out:
-        return ret;
-}
-
-int32_t
-glusterd3_1_stage_op_cbk (struct rpc_req *req, struct iovec *iov,
-                          int count, void *myframe)
-{
-        gd1_mgmt_stage_op_rsp         rsp   = {{0},};
-        int                           ret   = -1;
-        int32_t                       op_ret = -1;
-        glusterd_op_sm_event_type_t   event_type = GD_OP_EVENT_NONE;
-        glusterd_peerinfo_t           *peerinfo = NULL;
-        char                          str[50] = {0,};
-
-        GF_ASSERT (req);
-
-        if (-1 == req->rpc_status) {
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                rsp.op_errstr = "";
-                goto out;
-        }
-
-        ret = gd_xdr_to_mgmt_stage_op_rsp (*iov, &rsp);
-        if (ret < 0) {
-                gf_log ("", GF_LOG_ERROR, "error");
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                rsp.op_errstr = "";
-                goto out;
-        }
-        uuid_unparse (rsp.uuid, str);
-
-        op_ret = rsp.op_ret;
-
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received %s from uuid: %s",
-                (op_ret)?"RJT":"ACC", str);
-
-        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
-
-        if (ret) {
-                gf_log ("", GF_LOG_CRITICAL, "Stage response received from "
-                        "unknown peer: %s", str);
-        }
-
-        if (op_ret) {
-                event_type = GD_OP_EVENT_RCVD_RJT;
-                opinfo.op_ret = op_ret;
-                opinfo.op_errstr = gf_strdup(rsp.op_errstr);
-                if (!opinfo.op_errstr) {
-                        gf_log ("", GF_LOG_ERROR, "memory allocation failed");
-                        ret = -1;
-                        goto out;
-                }
-        } else {
-                event_type = GD_OP_EVENT_RCVD_ACC;
-        }
-
-        ret = glusterd_op_sm_inject_event (event_type, NULL);
-
-        if (!ret) {
-                glusterd_friend_sm ();
-                glusterd_op_sm ();
-        }
-
-out:
-        if (rsp.op_errstr && strcmp (rsp.op_errstr, ""))
-                free (rsp.op_errstr); //malloced by xdr
-        return ret;
-}
-
-static int32_t
-glusterd_sync_use_rsp_dict (dict_t *rsp_dict)
-{
-        int      ret      = 0;
-
-        GF_ASSERT (rsp_dict);
-
-        if (!rsp_dict) {
-                goto out;
-        }
-
-        ret = glusterd_import_friend_volumes (rsp_dict);
-out:
-        return ret;
-
-}
-
-static int32_t
-glusterd_rb_use_rsp_dict (dict_t *rsp_dict)
-{
-        int32_t  src_port = 0;
-        int      ret      = 0;
-        dict_t  *ctx      = NULL;
-
-        GF_ASSERT (rsp_dict);
-
-        if (rsp_dict) {
-                ret = dict_get_int32 (rsp_dict, "src-brick-port", &src_port);
-                if (ret) {
-                        gf_log ("", GF_LOG_DEBUG,
-                                "src-brick-port not present");
-                        ret = 0;
-                        goto out;
-                }
-
-                if (src_port) {
-                        ctx = glusterd_op_get_ctx (GD_OP_REPLACE_BRICK);
-                        if (!ctx) {
-                                gf_log ("", GF_LOG_ERROR,
-                                        "Operation Context is not present");
-                                ret = 0;
-                                goto out;
-                        }
-
-                        ret = dict_set_int32 (ctx, "src-brick-port",
-                                              src_port);
-                        if (ret) {
-                                gf_log ("", GF_LOG_DEBUG,
-                                        "Could not set src-brick");
-                                goto out;
-                        }
-                }
-        }
-
-out:
-        return ret;
-
-}
-
-int32_t
-glusterd3_1_commit_op_cbk (struct rpc_req *req, struct iovec *iov,
-                          int count, void *myframe)
-{
-        gd1_mgmt_commit_op_rsp         rsp   = {{0},};
-        int                           ret   = -1;
-        int32_t                       op_ret = -1;
-        glusterd_op_sm_event_type_t   event_type = GD_OP_EVENT_NONE;
-        glusterd_peerinfo_t           *peerinfo = NULL;
-        char                          str[50] = {0,};
-        dict_t                        *dict = NULL;
-
-
-        GF_ASSERT (req);
-
-        if (-1 == req->rpc_status) {
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                rsp.op_errstr = "";
-                goto out;
-        }
-
-        ret = gd_xdr_to_mgmt_commit_op_rsp (*iov, &rsp);
-        if (ret < 0) {
-                gf_log ("", GF_LOG_ERROR, "error");
-                rsp.op_ret   = -1;
-                rsp.op_errno = EINVAL;
-                rsp.op_errstr = "";
-                goto out;
-        }
-        uuid_unparse (rsp.uuid, str);
-
-        if (rsp.dict.dict_len) {
-                /* Unserialize the dictionary */
-                dict  = dict_new ();
-
-                ret = dict_unserialize (rsp.dict.dict_val,
-                                        rsp.dict.dict_len,
-                                        &dict);
-                if (ret < 0) {
-                        gf_log ("glusterd", GF_LOG_ERROR,
-                                "failed to "
-                                "unserialize rsp-buffer to dictionary");
-                        goto out;
-                }
-        }
-
-        op_ret = rsp.op_ret;
-
-        gf_log ("glusterd", GF_LOG_NORMAL,
-                "Received %s from uuid: %s",
-                (op_ret)?"RJT":"ACC", str);
-
-        ret = glusterd_friend_find (rsp.uuid, NULL, &peerinfo);
-
-        if (ret) {
-                gf_log ("", GF_LOG_CRITICAL, "Commit response received from "
-                        "unknown peer: %s", str);
-        }
-
-        if (op_ret) {
-                event_type = GD_OP_EVENT_RCVD_RJT;
-                opinfo.op_ret = op_ret;
-                opinfo.op_errstr = gf_strdup(rsp.op_errstr);
-                if (!opinfo.op_errstr) {
-                        gf_log ("", GF_LOG_ERROR, "memory allocation failed");
-                        ret = -1;
-                        goto out;
-                }
-        } else {
-                switch (rsp.op) {
-                case GD_OP_REPLACE_BRICK:
-                        ret = glusterd_rb_use_rsp_dict (dict);
-                        if (ret)
-                                goto out;
-                        break;
-                case GD_OP_SYNC_VOLUME:
-                        ret = glusterd_sync_use_rsp_dict (dict);
-                        if (ret)
-                                goto out;
-                        break;
-                default:
-                        break;
-                }
-                event_type = GD_OP_EVENT_RCVD_ACC;
-        }
-
-        ret = glusterd_op_sm_inject_event (event_type, NULL);
-
-        if (!ret) {
-                glusterd_friend_sm ();
-                glusterd_op_sm ();
-        }
-
-        return ret;
-
-
-out:
-        if (dict)
-                dict_unref (dict);
-        if (rsp.op_errstr && strcmp (rsp.op_errstr, ""))
-                free (rsp.op_errstr); //malloced by xdr
-        return ret;
-}
-
-
-
-int32_t
-glusterd3_1_probe (call_frame_t *frame, xlator_t *this,
-                   void *data)
-{
-        gd1_mgmt_probe_req      req = {{0},};
-        int                     ret = 0;
-        int                     port = 0;
-        char                    *hostname = NULL;
-        glusterd_peerinfo_t     *peerinfo = NULL;
-        glusterd_conf_t         *priv = NULL;
-        dict_t                  *dict = NULL;
-
-        if (!frame || !this ||  !data) {
-                ret = -1;
-                goto out;
-        }
-
-        dict = data;
-        priv = this->private;
-
-        GF_ASSERT (priv);
-        ret = dict_get_str (dict, "hostname", &hostname);
-        if (ret)
-                goto out;
-        ret = dict_get_int32 (dict, "port", &port);
-        if (ret)
-                port = 6969;
-
-        ret = glusterd_friend_find (NULL, hostname, &peerinfo);
-
-        if (ret) {
-                //We should not reach this state ideally
-                GF_ASSERT (0);
-                goto out;
-        }
-
-        uuid_copy (req.uuid, priv->uuid);
-        req.hostname = gf_strdup (hostname);
-        req.port = port;
-
-        ret = glusterd_submit_request (peerinfo, &req, frame, priv->mgmt,
-                                       GD_MGMT_PROBE_QUERY,
-                                       NULL, gd_xdr_from_mgmt_probe_req,
-                                       this, glusterd3_1_probe_cbk);
-
-out:
-        if (req.hostname)
-                GF_FREE (req.hostname);
-        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
-        return ret;
-}
-
-
-int32_t
-glusterd3_1_friend_add (call_frame_t *frame, xlator_t *this,
-                        void *data)
-{
-        gd1_mgmt_friend_req     req = {{0},};
-        int                     ret = 0;
-        glusterd_peerinfo_t     *peerinfo = NULL;
-        glusterd_conf_t         *priv = NULL;
-        glusterd_friend_sm_event_t     *event = NULL;
-        glusterd_friend_req_ctx_t *ctx = NULL;
-        dict_t                  *vols = NULL;
-
-
-        if (!frame || !this || !data) {
-                ret = -1;
-                goto out;
-        }
-
-        event = data;
-        priv = this->private;
-
-        GF_ASSERT (priv);
-
-        ctx = event->ctx;
-
-        peerinfo = event->peerinfo;
-
-        ret = glusterd_build_volume_dict (&vols);
-        if (ret)
-                goto out;
-
-        uuid_copy (req.uuid, priv->uuid);
-        req.hostname = gf_strdup (peerinfo->hostname);
-        req.port = peerinfo->port;
-
-        ret = dict_allocate_and_serialize (vols, &req.vols.vols_val,
-                                           (size_t *)&req.vols.vols_len);
-        if (ret)
-                goto out;
-
-        ret = glusterd_submit_request (peerinfo, &req, frame, priv->mgmt,
-                                       GD_MGMT_FRIEND_ADD,
-                                       NULL, gd_xdr_from_mgmt_friend_req,
-                                       this, glusterd3_1_friend_add_cbk);
-
-
-out:
-        if (req.vols.vols_val)
-                GF_FREE (req.vols.vols_val);
-
-        if (vols)
-                dict_unref (vols);
-
-        if (req.hostname)
-                GF_FREE (req.hostname);
-
-        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
-        return ret;
-}
-
-int32_t
-glusterd3_1_friend_remove (call_frame_t *frame, xlator_t *this,
-                           void *data)
-{
-        gd1_mgmt_friend_req             req = {{0},};
-        int                             ret = 0;
-        glusterd_peerinfo_t             *peerinfo = NULL;
-        glusterd_conf_t                 *priv = NULL;
-        glusterd_friend_sm_event_t      *event = NULL;
-        glusterd_friend_req_ctx_t       *ctx = NULL;
-
-
-        if (!frame || !this || !data) {
-                ret = -1;
-                goto out;
-        }
-
-        event = data;
-        priv = this->private;
-
-        GF_ASSERT (priv);
-
-        ctx = event->ctx;
-
-        peerinfo = event->peerinfo;
-
-        uuid_copy (req.uuid, priv->uuid);
-        req.hostname = peerinfo->hostname;
-        req.port = peerinfo->port;
-        ret = glusterd_submit_request (peerinfo, &req, frame, priv->mgmt,
-                                       GD_MGMT_FRIEND_REMOVE,
-                                       NULL, gd_xdr_from_mgmt_friend_req,
-                                       this, glusterd3_1_friend_remove_cbk);
-        //Override ret
-        ret = 0;
-
-out:
-        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
-        return ret;
-}
-
-
-int32_t
-glusterd3_1_friend_update (call_frame_t *frame, xlator_t *this,
-                           void *data)
-{
-        gd1_mgmt_friend_update          req = {{0},};
-        int                             ret = 0;
-        glusterd_peerinfo_t             *peerinfo = NULL;
-        glusterd_conf_t                 *priv = NULL;
-        glusterd_friend_sm_event_t      *event = NULL;
-        glusterd_friend_update_ctx_t     *ctx = NULL;
-        dict_t                          *friends = NULL;
-        char                            key[100] = {0,};
-        char                            uuid_buf[50] = {0,};
-        char                            *dup_buf = NULL;
-        int32_t                         count = 0;
-        char                            *dict_buf = NULL;
-        size_t                         len = -1;
-        call_frame_t                    *dummy_frame = NULL;
-
-
-        if ( !this || !data) {
-                ret = -1;
-                goto out;
-        }
-
-        friends = dict_new ();
-        if (!friends)
-                goto out;
-
-        event = data;
-        priv = this->private;
-
-        GF_ASSERT (priv);
-
-        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
-                count++;
-                uuid_unparse (peerinfo->uuid, uuid_buf);
-                snprintf (key, sizeof (key), "friend%d.uuid", count);
-                dup_buf = gf_strdup (uuid_buf);
-                ret = dict_set_dynstr (friends, key, dup_buf);
-                if (ret)
-                        goto out;
-                snprintf (key, sizeof (key), "friend%d.hostname", count);
-                ret = dict_set_str (friends, key, peerinfo->hostname);
-                if (ret)
-                        goto out;
-                gf_log ("", GF_LOG_NORMAL, "Added uuid: %s, host: %s",
-                        dup_buf, peerinfo->hostname);
-        }
-
-        ret = dict_set_int32 (friends, "count", count);
-        if (ret)
-                goto out;
-
-        ctx = event->ctx;
-
-        ret = dict_allocate_and_serialize (friends, &dict_buf, (size_t *)&len);
-
-        if (ret)
-                goto out;
-
-        req.friends.friends_val = dict_buf;
-        req.friends.friends_len = len;
-
-        peerinfo = event->peerinfo;
-        uuid_copy (req.uuid, priv->uuid);
-
-        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
-                if (!peerinfo->connected)
-                        continue;
-                dummy_frame = create_frame (this, this->ctx->pool);
-                ret = glusterd_submit_request (peerinfo, &req, dummy_frame,
-                                               priv->mgmt,
-                                               GD_MGMT_FRIEND_UPDATE,
-                                               NULL, gd_xdr_from_mgmt_friend_update,
-                                               this, glusterd3_1_friend_update_cbk);
-        }
-
-out:
-        if (friends)
-                dict_unref (friends);
-        if (req.friends.friends_val)
-                GF_FREE (req.friends.friends_val);
-        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
-        return ret;
-}
-
-int32_t
-glusterd3_1_cluster_lock (call_frame_t *frame, xlator_t *this,
-                           void *data)
-{
-        gd1_mgmt_cluster_lock_req       req = {{0},};
-        int                             ret = 0;
-        glusterd_peerinfo_t             *peerinfo = NULL;
-        glusterd_conf_t                 *priv = NULL;
-        call_frame_t                    *dummy_frame = NULL;
-        int32_t                         pending_lock = 0;
-
-        if (!this) {
-                ret = -1;
-                goto out;
-        }
-
-        priv = this->private;
-        glusterd_get_uuid (&req.uuid);
-
-        GF_ASSERT (priv);
-        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
-                GF_ASSERT (peerinfo);
-
-                if ((peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED) &&
-                    (glusterd_op_get_op() != GD_OP_SYNC_VOLUME))
-                        continue;
-
-                dummy_frame = create_frame (this, this->ctx->pool);
-
-                if (!dummy_frame)
-                        continue;
-
-                ret = glusterd_submit_request (peerinfo, &req, dummy_frame,
-                                        priv->mgmt, GD_MGMT_CLUSTER_LOCK,
-                                        NULL,
-                                        gd_xdr_from_mgmt_cluster_lock_req,
-                                        this, glusterd3_1_cluster_lock_cbk);
-                if (!ret)
-                        pending_lock++;
-                //TODO: Instead of keeping count, maintain a list of locked
-                //UUIDs.
-        }
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Sent lock req to %d peers",
-                                            pending_lock);
-        opinfo.pending_count = pending_lock;
-out:
-        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
-        return ret;
-}
-
-int32_t
-glusterd3_1_cluster_unlock (call_frame_t *frame, xlator_t *this,
-                            void *data)
-{
-        gd1_mgmt_cluster_lock_req       req = {{0},};
-        int                             ret = 0;
-        glusterd_peerinfo_t             *peerinfo = NULL;
-        glusterd_conf_t                 *priv = NULL;
-        int32_t                         pending_unlock = 0;
-        call_frame_t                    *dummy_frame = NULL;
-
-        if (!this ) {
-                ret = -1;
-                goto out;
-        }
-
-        priv = this->private;
-
-        glusterd_get_uuid (&req.uuid);
-
-        GF_ASSERT (priv);
-        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
-                GF_ASSERT (peerinfo);
-
-                if ((peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED) &&
-                    (glusterd_op_get_op() != GD_OP_SYNC_VOLUME))
-                        continue;
-
-                dummy_frame = create_frame (this, this->ctx->pool);
-
-                if (!dummy_frame)
-                        continue;
-
-                ret = glusterd_submit_request (peerinfo, &req, dummy_frame,
-                                        priv->mgmt, GD_MGMT_CLUSTER_UNLOCK,
-                                        NULL,
-                                        gd_xdr_from_mgmt_cluster_unlock_req,
-                                        this, glusterd3_1_cluster_unlock_cbk);
-                if (!ret)
-                        pending_unlock++;
-                //TODO: Instead of keeping count, maintain a list of locked
-                //UUIDs.
-        }
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Sent unlock req to %d peers",
-                                            pending_unlock);
-        opinfo.pending_count = pending_unlock;
-
-out:
-        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
-        return ret;
-}
-
-int32_t
-glusterd3_1_stage_op (call_frame_t *frame, xlator_t *this,
-                      void *data)
-{
-        gd1_mgmt_stage_op_req           *req = NULL;
-        int                             ret = 0;
-        glusterd_peerinfo_t             *peerinfo = NULL;
-        glusterd_conf_t                 *priv = NULL;
-        int32_t                         pending_peer = 0;
-        int                             i = 0;
-        call_frame_t                    *dummy_frame = NULL;
-        char                            *op_errstr = NULL;
-
-        if (!this) {
-                ret = -1;
-                goto out;
-        }
-
-        priv = this->private;
-
-        GF_ASSERT (priv);
-
-        for ( i = GD_OP_NONE; i < GD_OP_MAX; i++) {
-                if (opinfo.pending_op[i])
-                        break;
-        }
-
-        if (GD_OP_MAX == i) {
-
-                //No pending ops, inject stage_acc
-
-                ret = glusterd_op_sm_inject_event
-                        (GD_OP_EVENT_STAGE_ACC, NULL);
-
-                return ret;
-        }
-
-        glusterd_op_clear_pending_op (i);
-
-
-        ret = glusterd_op_build_payload (i, &req);
-
-        if (ret)
-                goto out;
-
-        ret = glusterd_op_stage_validate (req, &op_errstr);
-
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Staging failed");
-                opinfo.op_errstr = op_errstr;
-                goto out;
-        }
-
-        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
-                GF_ASSERT (peerinfo);
-
-                if ((peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED) &&
-                    (glusterd_op_get_op() != GD_OP_SYNC_VOLUME))
-                        continue;
-
-                dummy_frame = create_frame (this, this->ctx->pool);
-
-                if (!dummy_frame)
-                        continue;
-
-                ret = glusterd_submit_request (peerinfo, req, dummy_frame,
-                                                priv->mgmt, GD_MGMT_STAGE_OP,
-                                                NULL,
-                                                gd_xdr_from_mgmt_stage_op_req,
-                                                this, glusterd3_1_stage_op_cbk);
-                if (!ret)
-                        pending_peer++;
-                //TODO: Instead of keeping count, maintain a list of pending
-                //UUIDs.
-        }
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Sent op req to %d peers",
-                                            pending_peer);
-        opinfo.pending_count = pending_peer;
-
-out:
-        if (ret) {
-                glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL);
-                opinfo.op_ret = ret;
-        }
-        if (req) {
-                GF_FREE (req->buf.buf_val);
-                GF_FREE (req);
-        }
-        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
-        return ret;
-}
-
-int32_t
-glusterd3_1_commit_op (call_frame_t *frame, xlator_t *this,
-                      void *data)
-{
-        gd1_mgmt_commit_op_req          *req = NULL;
-        int                             ret = 0;
-        glusterd_peerinfo_t             *peerinfo = NULL;
-        glusterd_conf_t                 *priv = NULL;
-        int32_t                         pending_peer = 0;
-        int                             i = 0;
-        call_frame_t                    *dummy_frame = NULL;
-        char                            *op_errstr = NULL;
-
-        if (!this) {
-                ret = -1;
-                goto out;
-        }
-
-        priv = this->private;
-
-        GF_ASSERT (priv);
-
-        for ( i = GD_OP_NONE; i < GD_OP_MAX; i++) {
-                if (opinfo.commit_op[i])
-                        break;
-        }
-
-        if (GD_OP_MAX == i) {
-
-                //No pending ops, inject stage_acc
-
-                ret = glusterd_op_sm_inject_event
-                        (GD_OP_EVENT_COMMIT_ACC, NULL);
-
-                return ret;
-        }
-
-        glusterd_op_clear_commit_op (i);
-
-        ret = glusterd_op_build_payload (i, (gd1_mgmt_stage_op_req **)&req);
-
-        if (ret)
-                goto out;
-
-        ret = glusterd_op_commit_perform ((gd1_mgmt_stage_op_req *)req, &op_errstr,
-                                          NULL);//rsp_dict invalid for source
-
-        if (ret) {
-                gf_log ("", GF_LOG_ERROR, "Commit failed");
-                opinfo.op_errstr = op_errstr;
-                goto out;
-        }
-
-        list_for_each_entry (peerinfo, &priv->peers, uuid_list) {
-                GF_ASSERT (peerinfo);
-
-                if ((peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED) &&
-                    (glusterd_op_get_op() != GD_OP_SYNC_VOLUME))
-                        continue;
-
-                dummy_frame = create_frame (this, this->ctx->pool);
-
-                if (!dummy_frame)
-                        continue;
-
-                ret = glusterd_submit_request (peerinfo, req, dummy_frame,
-                                                priv->mgmt, GD_MGMT_COMMIT_OP,
-                                                NULL,
-                                                gd_xdr_from_mgmt_commit_op_req,
-                                                this, glusterd3_1_commit_op_cbk);
-                if (!ret)
-                        pending_peer++;
-                //TODO: Instead of keeping count, maintain a list of pending
-                //UUIDs.
-        }
-
-        gf_log ("glusterd", GF_LOG_NORMAL, "Sent op req to %d peers",
-                                            pending_peer);
-        opinfo.pending_count = pending_peer;
-
-out:
-        if (ret) {
-                glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL);
-                opinfo.op_ret = ret;
-        }
-        if (req) {
-                GF_FREE (req->buf.buf_val);
-                GF_FREE (req);
-        }
-        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
-        return ret;
-}
-
-
-int
-glusterd_handle_rpc_msg (rpcsvc_request_t *req)
-{
-        int             ret = -1;
-        gf_boolean_t    is_cli_req = _gf_false;
-        char            *op_errstr = NULL;
-
-        GF_ASSERT (req);
-
-        is_cli_req = glusterd_is_cli_op_req (req->procnum);
-
-        if (is_cli_req) {
-                ret = glusterd_op_set_cli_op (req->procnum);
-
-                if (ret) {
-                        gf_log ("", GF_LOG_ERROR, "Unable to set cli op: %d",
-                                ret);
-                        goto out;
-                }
-        }
-
-        switch (req->procnum) {
-                case GD_MGMT_PROBE_QUERY:
-                        ret = glusterd_handle_probe_query (req);
-                        break;
-
-                case GD_MGMT_FRIEND_ADD:
-                        ret = glusterd_handle_incoming_friend_req (req);
-                        break;
-
-                case GD_MGMT_CLUSTER_LOCK:
-                        ret = glusterd_handle_cluster_lock (req);
-                        break;
-
-                case GD_MGMT_CLUSTER_UNLOCK:
-                        ret = glusterd_handle_cluster_unlock (req);
-                        break;
-
-                case GD_MGMT_STAGE_OP:
-                        ret = glusterd_handle_stage_op (req);
-                        break;
-
-                case GD_MGMT_COMMIT_OP:
-                        ret = glusterd_handle_commit_op (req);
-                        break;
-
-                case GD_MGMT_CLI_PROBE:
-                        ret = glusterd_handle_cli_probe (req);
-                        if (ret == GLUSTERD_CONNECTION_AWAITED)
-                                return 0;
-                        break;
-
-                case GD_MGMT_CLI_CREATE_VOLUME:
-                        ret = glusterd_handle_create_volume (req);
-                        break;
-
-                case GD_MGMT_CLI_DEPROBE:
-                        ret = glusterd_handle_cli_deprobe (req);
-                        break;
-
-                case GD_MGMT_FRIEND_REMOVE:
-                        ret = glusterd_handle_incoming_unfriend_req (req);
-                        break;
-
-                case GD_MGMT_CLI_LIST_FRIENDS:
-                        ret = glusterd_handle_cli_list_friends (req);
-                        break;
-
-                case GD_MGMT_CLI_START_VOLUME:
-                        ret = glusterd_handle_cli_start_volume (req);
-                        break;
-
-                case GD_MGMT_CLI_STOP_VOLUME:
-                        ret = glusterd_handle_cli_stop_volume (req);
-                        break;
-
-                case GD_MGMT_CLI_DELETE_VOLUME:
-                        ret = glusterd_handle_cli_delete_volume (req);
-                        break;
-
-                case GD_MGMT_FRIEND_UPDATE:
-                        ret = glusterd_handle_friend_update (req);
-                        break;
-
-                case GD_MGMT_CLI_GET_VOLUME:
-                        ret = glusterd_handle_cli_get_volume (req);
-                        break;
-
-                case GD_MGMT_CLI_DEFRAG_VOLUME:
-                        ret = glusterd_handle_defrag_volume (req);
-                        break;
-
-                case GD_MGMT_CLI_ADD_BRICK:
-                        ret = glusterd_handle_add_brick (req);
-                        break;
-
-                case GD_MGMT_CLI_REPLACE_BRICK:
-                        ret = glusterd_handle_replace_brick (req);
-                        break;
-
-                case GD_MGMT_CLI_REMOVE_BRICK:
-                        ret = glusterd_handle_remove_brick (req);
-                        break;
-
-                case GD_MGMT_CLI_LOG_FILENAME:
-                        ret = glusterd_handle_log_filename (req);
-                        break;
-
-                case GD_MGMT_CLI_LOG_LOCATE:
-                        ret = glusterd_handle_log_locate (req);
-                        break;
-
-                case GD_MGMT_CLI_LOG_ROTATE:
-                        ret = glusterd_handle_log_rotate (req);
-                        break;
-
-                case GD_MGMT_CLI_SET_VOLUME:
-                        ret = glusterd_handle_set_volume (req);
-			break;
-
-                case GD_MGMT_CLI_SYNC_VOLUME:
-                        ret = glusterd_handle_sync_volume (req);
-                        break;
-
-                default:
-			gf_log("", GF_LOG_ERROR, "Recieved Invalid procnum:%d",
-			       req->procnum);
-                        GF_ASSERT (0);
-        }
-
-        glusterd_friend_sm ();
-        glusterd_op_sm ();
-
-out:
-        if (ret && is_cli_req) {
-                /* if we are sending a reply here, then return value should
-                   be 0, and we should not point to any RPC errors, because
-                   otherwise rpcsvc.c will send an error reply for the same
-                   request, which causes double replies */
-                ret = glusterd_op_send_cli_response (req->procnum, ret, 0, req, NULL, op_errstr);
-                if (!ret)
-                        req->rpc_err = SUCCESS;
-        }
-        if (op_errstr && (strcmp (op_errstr, "")))
-                GF_FREE (op_errstr);
-        if (!ret)
-                gf_log ("", GF_LOG_WARNING, "Returning %d", ret);
-
-        return ret;
-}
-
-
-rpcsvc_actor_t glusterd1_mgmt_actors[] = {
-        [GD_MGMT_NULL]        = { "NULL",       GD_MGMT_NULL, glusterd_null, NULL, NULL},
-        [GD_MGMT_PROBE_QUERY] = { "PROBE_QUERY", GD_MGMT_PROBE_QUERY, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_FRIEND_ADD] = { "FRIEND_ADD", GD_MGMT_FRIEND_ADD, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_FRIEND_REMOVE] = { "FRIEND_REMOVE", GD_MGMT_FRIEND_REMOVE, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_FRIEND_UPDATE] = { "FRIEND_UPDATE", GD_MGMT_FRIEND_UPDATE, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLUSTER_LOCK] = { "CLUSTER_LOCK", GD_MGMT_CLUSTER_LOCK, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLUSTER_UNLOCK] = { "CLUSTER_UNLOCK", GD_MGMT_CLUSTER_UNLOCK, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_STAGE_OP] = { "STAGE_OP", GD_MGMT_STAGE_OP, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_COMMIT_OP] = { "COMMIT_OP", GD_MGMT_COMMIT_OP, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_PROBE] = { "CLI_PROBE", GD_MGMT_CLI_PROBE, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_CREATE_VOLUME] = { "CLI_CREATE_VOLUME", GD_MGMT_CLI_CREATE_VOLUME, glusterd_handle_rpc_msg, NULL,NULL},
-        [GD_MGMT_CLI_DEFRAG_VOLUME] = { "CLI_DEFRAG_VOLUME", GD_MGMT_CLI_DEFRAG_VOLUME, glusterd_handle_defrag_volume, NULL,NULL},
-        [GD_MGMT_CLI_DEPROBE] = { "FRIEND_REMOVE", GD_MGMT_CLI_DEPROBE, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_LIST_FRIENDS] = { "LIST_FRIENDS", GD_MGMT_CLI_LIST_FRIENDS, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_START_VOLUME] = { "START_VOLUME", GD_MGMT_CLI_START_VOLUME, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_STOP_VOLUME] = { "STOP_VOLUME", GD_MGMT_CLI_STOP_VOLUME, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_DELETE_VOLUME] = { "DELETE_VOLUME", GD_MGMT_CLI_DELETE_VOLUME, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_GET_VOLUME] = { "GET_VOLUME", GD_MGMT_CLI_GET_VOLUME, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_ADD_BRICK] = { "ADD_BRICK", GD_MGMT_CLI_ADD_BRICK, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_REPLACE_BRICK] = { "REPLACE_BRICK", GD_MGMT_CLI_REPLACE_BRICK, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_REMOVE_BRICK] = { "REMOVE_BRICK", GD_MGMT_CLI_REMOVE_BRICK, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_LOG_FILENAME] = { "LOG FILENAME", GD_MGMT_CLI_LOG_FILENAME, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_LOG_LOCATE] = { "LOG LOCATE", GD_MGMT_CLI_LOG_LOCATE, glusterd_handle_log_locate, NULL, NULL},
-        [GD_MGMT_CLI_LOG_ROTATE] = { "LOG FILENAME", GD_MGMT_CLI_LOG_ROTATE, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_SET_VOLUME] = { "SET_VOLUME", GD_MGMT_CLI_SET_VOLUME, glusterd_handle_rpc_msg, NULL, NULL},
-        [GD_MGMT_CLI_SYNC_VOLUME] = { "SYNC_VOLUME", GD_MGMT_CLI_SYNC_VOLUME, glusterd_handle_rpc_msg, NULL, NULL},
-};
-
-/*rpcsvc_actor_t glusterd1_mgmt_actors[] = {
-        [GD_MGMT_NULL]        = { "NULL",       GD_MGMT_NULL, glusterd_null, NULL, NULL},
-        [GD_MGMT_PROBE_QUERY] = { "PROBE_QUERY", GD_MGMT_PROBE_QUERY, glusterd_handle_probe_query, NULL, NULL},
-        [GD_MGMT_FRIEND_ADD] = { "FRIEND_ADD", GD_MGMT_FRIEND_ADD, glusterd_handle_incoming_friend_req, NULL, NULL},
-        [GD_MGMT_CLUSTER_LOCK] = { "CLUSTER_LOCK", GD_MGMT_CLUSTER_LOCK, glusterd_handle_cluster_lock, NULL, NULL},
-        [GD_MGMT_CLUSTER_UNLOCK] = { "CLUSTER_UNLOCK", GD_MGMT_CLUSTER_UNLOCK, glusterd_handle_cluster_unlock, NULL, NULL},
-        [GD_MGMT_STAGE_OP] = { "STAGE_OP", GD_MGMT_STAGE_OP, glusterd_handle_stage_op, NULL, NULL},
-        [GD_MGMT_COMMIT_OP] = { "COMMIT_OP", GD_MGMT_COMMIT_OP, glusterd_handle_commit_op, NULL, NULL},
-        [GD_MGMT_CLI_PROBE] = { "CLI_PROBE", GD_MGMT_CLI_PROBE, glusterd_handle_cli_probe, NULL, NULL},
-};*/
-
-
-struct rpcsvc_program glusterd1_mop_prog = {
-        .progname  = "GlusterD0.0.1",
-        .prognum   = GLUSTERD1_MGMT_PROGRAM,
-        .progver   = GLUSTERD1_MGMT_VERSION,
-        .numactors = GLUSTERD1_MGMT_PROCCNT,
-        .actors    = glusterd1_mgmt_actors,
-};
-
-
-struct rpc_clnt_procedure glusterd3_1_clnt_mgmt_actors[GD_MGMT_MAXVALUE] = {
-        [GD_MGMT_NULL]        = {"NULL", NULL },
-        [GD_MGMT_PROBE_QUERY]  = { "PROBE_QUERY",  glusterd3_1_probe},
-        [GD_MGMT_FRIEND_ADD]  = { "FRIEND_ADD",  glusterd3_1_friend_add },
-        [GD_MGMT_CLUSTER_LOCK] = {"CLUSTER_LOCK", glusterd3_1_cluster_lock},
-        [GD_MGMT_CLUSTER_UNLOCK] = {"CLUSTER_UNLOCK", glusterd3_1_cluster_unlock},
-        [GD_MGMT_STAGE_OP] = {"STAGE_OP", glusterd3_1_stage_op},
-        [GD_MGMT_COMMIT_OP] = {"COMMIT_OP", glusterd3_1_commit_op},
-        [GD_MGMT_FRIEND_REMOVE]  = { "FRIEND_REMOVE",  glusterd3_1_friend_remove},
-        [GD_MGMT_FRIEND_UPDATE]  = { "FRIEND_UPDATE",  glusterd3_1_friend_update},
-};
-
-
-
-struct rpc_clnt_program glusterd3_1_mgmt_prog = {
-        .progname = "Mgmt 3.1",
-        .prognum  = GLUSTERD1_MGMT_PROGRAM,
-        .progver  = GLUSTERD1_MGMT_VERSION,
-        .proctable    = glusterd3_1_clnt_mgmt_actors,
-        .numproc  = GLUSTERD1_MGMT_PROCCNT,
-};
diff --git a/xlators/mount/fuse/src/Makefile.am b/xlators/mount/fuse/src/Makefile.am
index 66d044e32..653121d18 100644
--- a/xlators/mount/fuse/src/Makefile.am
+++ b/xlators/mount/fuse/src/Makefile.am
@@ -1,25 +1,36 @@
-noinst_HEADERS = $(CONTRIBDIR)/fuse-include/fuse_kernel.h\
-	$(CONTRIBDIR)/fuse-include/fuse-mount.h\
+noinst_HEADERS_linux = $(CONTRIBDIR)/fuse-include/fuse_kernel.h\
+	$(CONTRIBDIR)/fuse-include/mount_util.h\
+	$(CONTRIBDIR)/fuse-lib/mount-gluster-compat.h
+noinst_HEADERS_darwin =	$(CONTRIBDIR)/fuse-include/fuse_kernel_macfuse.h
+noinst_HEADERS_common = $(CONTRIBDIR)/fuse-include/fuse-mount.h\
 	$(CONTRIBDIR)/fuse-include/fuse-misc.h fuse-mem-types.h \
 	fuse-bridge.h
 
+if GF_DARWIN_HOST_OS
+    noinst_HEADERS = $(noinst_HEADERS_common) $(noinst_HEADERS_darwin)
+else
+    noinst_HEADERS = $(noinst_HEADERS_common) $(noinst_HEADERS_linux)
+endif
+
 xlator_LTLIBRARIES = fuse.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/mount
 
 if GF_DARWIN_HOST_OS
     mount_source=$(CONTRIBDIR)/macfuse/mount_darwin.c
 else
-    mount_source=$(CONTRIBDIR)/fuse-lib/mount.c
+    mount_source=$(CONTRIBDIR)/fuse-lib/mount.c $(CONTRIBDIR)/fuse-lib/mount-common.c
 endif
 
 fuse_la_SOURCES = fuse-helpers.c fuse-resolve.c fuse-bridge.c \
 	$(CONTRIBDIR)/fuse-lib/misc.c $(mount_source)
 
-fuse_la_LDFLAGS = -module -avoidversion -shared -nostartfiles
+fuse_la_LDFLAGS = -module -avoid-version
+fuse_la_LIBADD = @GF_FUSE_LDADD@
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -D$(GF_HOST_OS) -Wall \
+AM_CPPFLAGS = $(GF_CPPFLAGS) \
 	-I$(top_srcdir)/libglusterfs/src -I$(CONTRIBDIR)/fuse-include \
-	-I$(top_srcdir)/glusterfsd/src $(GF_CFLAGS) $(GF_FUSE_CFLAGS)
+	-I$(CONTRIBDIR)/fuse-lib $(GF_FUSE_CFLAGS)
 
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES =
diff --git a/xlators/mount/fuse/src/fuse-bridge.c b/xlators/mount/fuse/src/fuse-bridge.c
index 122304446..6a5587c2d 100644
--- a/xlators/mount/fuse/src/fuse-bridge.c
+++ b/xlators/mount/fuse/src/fuse-bridge.c
@@ -1,35 +1,164 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
-*/
+  Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
 
-/*
- * TODO:
- * Need to free_state() when fuse_reply_err() + return.
- * Check loc->path for "" after fuse_loc_fill in all fops
- * (now being done in getattr, lookup) or better - make
- * fuse_loc_fill() and inode_path() return success/failure.
- */
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
 
+#include <sys/wait.h>
 #include "fuse-bridge.h"
+#include "mount-gluster-compat.h"
+#include "glusterfs.h"
+#include "glusterfs-acl.h"
+
+#ifdef __NetBSD__
+#undef open /* in perfuse.h, pulled from mount-gluster-compat.h */
+#endif
 
 static int gf_fuse_conn_err_log;
 static int gf_fuse_xattr_enotsup_log;
 
+void fini (xlator_t *this_xl);
+
+static void fuse_invalidate_inode(xlator_t *this, uint64_t fuse_ino);
+
+/*
+ * Send an invalidate notification up to fuse to purge the file from local
+ * page cache.
+ */
+static int32_t
+fuse_invalidate(xlator_t *this, inode_t *inode)
+{
+        fuse_private_t *priv = this->private;
+        uint64_t nodeid;
+
+        /*
+         * NOTE: We only invalidate at the moment if fopen_keep_cache is
+         * enabled because otherwise this is a departure from default
+         * behavior. Specifically, the performance/write-behind xlator
+         * causes unconditional invalidations on write requests.
+         */
+        if (!priv->fopen_keep_cache)
+                return 0;
+
+        nodeid = inode_to_fuse_nodeid(inode);
+        gf_log(this->name, GF_LOG_DEBUG, "Invalidate inode id %lu.", nodeid);
+        fuse_log_eh (this, "Sending invalidate inode id: %lu gfid: %s", nodeid,
+                     uuid_utoa (inode->gfid));
+        fuse_invalidate_inode(this, nodeid);
+
+        return 0;
+}
+
+static int32_t
+fuse_forget_cbk (xlator_t *this, inode_t *inode)
+{
+        //Nothing to free in inode ctx, hence return.
+        return 0;
+}
+
+void
+fuse_inode_set_need_lookup (inode_t *inode, xlator_t *this)
+{
+	uint64_t          need_lookup = 1;
+
+	if (!inode || !this)
+		return;
+
+	inode_ctx_set (inode, this, &need_lookup);
+
+	return;
+}
+
+
+gf_boolean_t
+fuse_inode_needs_lookup (inode_t *inode, xlator_t *this)
+{
+	uint64_t          need_lookup = 0;
+	gf_boolean_t      ret = _gf_false;
+
+	if (!inode || !this)
+		return ret;
+
+	inode_ctx_get (inode, this, &need_lookup);
+	if (need_lookup)
+		ret = _gf_true;
+	need_lookup = 0;
+	inode_ctx_set (inode, this, &need_lookup);
+
+	return ret;
+}
+
+
+fuse_fd_ctx_t *
+__fuse_fd_ctx_check_n_create (xlator_t *this, fd_t *fd)
+{
+        uint64_t       val    = 0;
+        int32_t        ret    = 0;
+        fuse_fd_ctx_t *fd_ctx = NULL;
+
+        ret = __fd_ctx_get (fd, this, &val);
+
+        fd_ctx = (fuse_fd_ctx_t *)(unsigned long) val;
+
+        if (fd_ctx == NULL) {
+                fd_ctx = GF_CALLOC (1, sizeof (*fd_ctx),
+                                    gf_fuse_mt_fd_ctx_t);
+                if (!fd_ctx) {
+                    goto out;
+                }
+                ret = __fd_ctx_set (fd, this,
+                                    (uint64_t)(unsigned long)fd_ctx);
+                if (ret < 0) {
+                        gf_log ("glusterfs-fuse", GF_LOG_DEBUG,
+                                "fd-ctx-set failed");
+                        GF_FREE (fd_ctx);
+                        fd_ctx = NULL;
+                }
+        }
+out:
+        return fd_ctx;
+}
+
+fuse_fd_ctx_t *
+fuse_fd_ctx_check_n_create (xlator_t *this, fd_t *fd)
+{
+        fuse_fd_ctx_t *fd_ctx = NULL;
+
+        if ((fd == NULL) || (this == NULL)) {
+                goto out;
+        }
+
+        LOCK (&fd->lock);
+        {
+                fd_ctx = __fuse_fd_ctx_check_n_create (this, fd);
+        }
+        UNLOCK (&fd->lock);
+
+out:
+        return fd_ctx;
+}
+
+fuse_fd_ctx_t *
+fuse_fd_ctx_get (xlator_t *this, fd_t *fd)
+{
+        fuse_fd_ctx_t *fdctx = NULL;
+        uint64_t       value = 0;
+        int            ret   = 0;
+
+        ret = fd_ctx_get (fd, this, &value);
+        if (ret < 0) {
+                goto out;
+        }
+
+        fdctx = (fuse_fd_ctx_t *) (unsigned long)value;
+
+out:
+        return fdctx;
+}
 
 /*
  * iov_out should contain a fuse_out_header at zeroth position.
@@ -43,6 +172,10 @@ send_fuse_iov (xlator_t *this, fuse_in_header_t *finh, struct iovec *iov_out,
         struct fuse_out_header *fouh = NULL;
         int res, i;
 
+        if (!this || !finh || !iov_out) {
+                gf_log ("send_fuse_iov", GF_LOG_ERROR,"Invalid arguments");
+                return EINVAL;
+        }
         priv = this->private;
 
         fouh = iov_out[0].iov_base;
@@ -94,45 +227,198 @@ send_fuse_data (xlator_t *this, fuse_in_header_t *finh, void *data, size_t size)
 #define send_fuse_obj(this, finh, obj) \
         send_fuse_data (this, finh, obj, sizeof (*(obj)))
 
+
+static void
+fuse_invalidate_entry (xlator_t *this, uint64_t fuse_ino)
+{
+        struct fuse_out_header             *fouh   = NULL;
+        struct fuse_notify_inval_entry_out *fnieo  = NULL;
+        fuse_private_t                     *priv   = NULL;
+        dentry_t                           *dentry = NULL;
+        inode_t                            *inode  = NULL;
+        size_t                              nlen   = 0;
+        int                                 rv     = 0;
+        char inval_buf[INVAL_BUF_SIZE]             = {0,};
+
+        fouh  = (struct fuse_out_header *)inval_buf;
+        fnieo = (struct fuse_notify_inval_entry_out *)(fouh + 1);
+
+        priv = this->private;
+        if (priv->revchan_out == -1)
+                return;
+
+        fouh->unique = 0;
+        fouh->error = FUSE_NOTIFY_INVAL_ENTRY;
+
+        inode = fuse_ino_to_inode (fuse_ino, this);
+
+        list_for_each_entry (dentry, &inode->dentry_list, inode_list) {
+                nlen = strlen (dentry->name);
+                fouh->len = sizeof (*fouh) + sizeof (*fnieo) + nlen + 1;
+                fnieo->parent = inode_to_fuse_nodeid (dentry->parent);
+
+                fnieo->namelen = nlen;
+                strcpy (inval_buf + sizeof (*fouh) + sizeof (*fnieo), dentry->name);
+
+                rv = write (priv->revchan_out, inval_buf, fouh->len);
+                if (rv != fouh->len) {
+                        gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                                "kernel notification daemon defunct");
+
+                        close (priv->fd);
+                        break;
+                }
+
+                gf_log ("glusterfs-fuse", GF_LOG_TRACE, "INVALIDATE entry: "
+                        "%"PRIu64"/%s", fnieo->parent, dentry->name);
+
+                if (dentry->parent) {
+                        fuse_log_eh (this, "Invalidated entry %s (parent: %s)",
+                                     dentry->name,
+                                     uuid_utoa (dentry->parent->gfid));
+                } else {
+                        fuse_log_eh (this, "Invalidated entry %s(nodeid: %ld)",
+                                     dentry->name, fnieo->parent);
+                }
+        }
+
+        if (inode)
+                inode_unref (inode);
+}
+
+/*
+ * Send an inval inode notification to fuse. This causes an invalidation of the
+ * entire page cache mapping on the inode.
+ */
+static void
+fuse_invalidate_inode(xlator_t *this, uint64_t fuse_ino)
+{
+        struct fuse_out_header *fouh = NULL;
+        struct fuse_notify_inval_inode_out *fniio = NULL;
+        fuse_private_t *priv = NULL;
+        int rv = 0;
+        char inval_buf[INVAL_BUF_SIZE] = {0};
+        inode_t    *inode = NULL;
+
+        fouh = (struct fuse_out_header *) inval_buf;
+        fniio = (struct fuse_notify_inval_inode_out *) (fouh + 1);
+
+        priv = this->private;
+
+        if (priv->revchan_out < 0)
+                return;
+
+        fouh->unique = 0;
+        fouh->error = FUSE_NOTIFY_INVAL_INODE;
+        fouh->len = sizeof(struct fuse_out_header) +
+                sizeof(struct fuse_notify_inval_inode_out);
+
+        /* inval the entire mapping until we learn how to be more granular */
+        fniio->ino = fuse_ino;
+        fniio->off = 0;
+        fniio->len = -1;
+
+        inode = fuse_ino_to_inode (fuse_ino, this);
+
+        rv = write(priv->revchan_out, inval_buf, fouh->len);
+        if (rv != fouh->len) {
+                gf_log("glusterfs-fuse", GF_LOG_ERROR, "kernel notification "
+                        "daemon defunct");
+                close(priv->fd);
+        }
+
+        gf_log("glusterfs-fuse", GF_LOG_TRACE, "INVALIDATE inode: %lu", fuse_ino);
+
+        if (inode) {
+                fuse_log_eh (this, "Invalidated inode %lu (gfid: %s)",
+                             fuse_ino, uuid_utoa (inode->gfid));
+        } else {
+                fuse_log_eh (this, "Invalidated inode %lu ", fuse_ino);
+        }
+
+        if (inode)
+                inode_unref (inode);
+}
+
 int
 send_fuse_err (xlator_t *this, fuse_in_header_t *finh, int error)
 {
         struct fuse_out_header fouh = {0, };
         struct iovec iov_out;
+        inode_t  *inode = NULL;
 
         fouh.error = -error;
         iov_out.iov_base = &fouh;
 
+        inode = fuse_ino_to_inode (finh->nodeid, this);
+
+        // filter out ENOENT
+        if (error != ENOENT) {
+                if (inode) {
+                        fuse_log_eh (this,"Sending %s for operation %d on "
+                                     "inode %s", strerror (error), finh->opcode,
+                                     uuid_utoa (inode->gfid));
+                } else {
+                        fuse_log_eh (this, "Sending %s for operation %d on "
+                                     "inode %ld", strerror (error),
+                                     finh->opcode, finh->nodeid);
+                }
+        }
+
+        if (inode)
+                inode_unref (inode);
+
         return send_fuse_iov (this, finh, &iov_out, 1);
 }
 
 static int
 fuse_entry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 int32_t op_ret, int32_t op_errno,
-                inode_t *inode, struct iatt *buf)
+                inode_t *inode, struct iatt *buf, dict_t *xdata)
 {
-        fuse_state_t            *state = NULL;
-        fuse_in_header_t        *finh = NULL;
-        struct fuse_entry_out    feo = {0, };
-        fuse_private_t          *priv = NULL;
-        inode_t                 *linked_inode = NULL;
+        fuse_state_t          *state        = NULL;
+        fuse_in_header_t      *finh         = NULL;
+        struct fuse_entry_out  feo          = {0, };
+        fuse_private_t        *priv         = NULL;
+        inode_t               *linked_inode = NULL;
 
         priv = this->private;
         state = frame->root->state;
         finh = state->finh;
 
-        if (!op_ret && state->loc.ino == 1) {
-                buf->ia_ino = 1;
+        if (op_ret == 0) {
+                if (__is_root_gfid (state->loc.inode->gfid))
+                        buf->ia_ino = 1;
+                if (uuid_is_null (buf->ia_gfid)) {
+                        /* With a NULL gfid inode linking is
+                           not possible. Let's not pretend this
+                           call was a "success".
+                        */
+                        gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                                "Received NULL gfid for %s. Forcing EIO",
+                                state->loc.path);
+                        op_ret = -1;
+                        op_errno = EIO;
+                }
         }
 
+        /* log into the event-history after the null uuid check is done, since
+         * the op_ret and op_errno are being changed if the gfid is NULL.
+         */
+        fuse_log_eh (this, "op_ret: %d op_errno: %d "
+                     "%"PRIu64": %s() %s => %s", op_ret, op_errno,
+                     frame->root->unique, gf_fop_list[frame->root->op],
+                     state->loc.path, (op_ret == 0)?
+                     uuid_utoa(buf->ia_gfid):uuid_utoa(state->loc.gfid));
+
         if (op_ret == 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                        "%"PRIu64": %s() %s => %"PRId64" (%"PRId64")",
+                        "%"PRIu64": %s() %s => %"PRIu64,
                         frame->root->unique, gf_fop_list[frame->root->op],
-                        state->loc.path, buf->ia_ino, state->loc.ino);
+                        state->loc.path, buf->ia_ino);
 
                 buf->ia_blksize = this->ctx->page_size;
-                gf_fuse_stat2attr (buf, &feo.attr);
+                gf_fuse_stat2attr (buf, &feo.attr, priv->enable_ino32);
 
                 if (!buf->ia_ino) {
                         gf_log ("glusterfs-fuse", GF_LOG_WARNING,
@@ -149,8 +435,6 @@ fuse_entry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
                 inode_lookup (linked_inode);
 
-                /* TODO: make these timeouts configurable (via meta?) */
-                /* should we do linked_node or inode */
                 feo.nodeid = inode_to_fuse_nodeid (linked_inode);
 
                 inode_unref (linked_inode);
@@ -178,7 +462,16 @@ fuse_entry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         "%"PRIu64": %s() %s => -1 (%s)", frame->root->unique,
                         gf_fop_list[frame->root->op], state->loc.path,
                         strerror (op_errno));
-                send_fuse_err (this, state->finh, op_errno);
+
+		if ((op_errno == ENOENT) && (priv->negative_timeout != 0)) {
+			feo.entry_valid =
+				calc_timeout_sec (priv->negative_timeout);
+			feo.entry_valid_nsec =
+				calc_timeout_nsec (priv->negative_timeout);
+			send_fuse_obj (this, finh, &feo);
+		} else {
+			send_fuse_err (this, state->finh, op_errno);
+                }
         }
 
         free_fuse_state (state);
@@ -186,18 +479,17 @@ fuse_entry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         return 0;
 }
 
-
 static int
 fuse_newentry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno,
                    inode_t *inode, struct iatt *buf, struct iatt *preparent,
-                   struct iatt *postparent)
+                   struct iatt *postparent, dict_t *xdata)
 {
-        fuse_entry_cbk (frame, cookie, this, op_ret, op_errno, inode, buf);
+        fuse_entry_cbk (frame, cookie, this, op_ret, op_errno, inode, buf,
+                        xdata);
         return 0;
 }
 
-
 static int
 fuse_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno,
@@ -212,18 +504,29 @@ fuse_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         prev  = cookie;
 
         if (op_ret == -1 && state->is_revalidate == 1) {
-                itable = state->loc.inode->table;
+                itable = state->itable;
+		/*
+		 * A stale mapping might exist for a dentry/inode that has been
+		 * removed from another client.
+		 */
+		if (op_errno == ENOENT)
+			inode_unlink(state->loc.inode, state->loc.parent,
+				     state->loc.name);
                 inode_unref (state->loc.inode);
                 state->loc.inode = inode_new (itable);
                 state->is_revalidate = 2;
+                if (uuid_is_null (state->gfid))
+                        uuid_generate (state->gfid);
+                fuse_gfid_set (state);
 
                 STACK_WIND (frame, fuse_lookup_cbk,
                             prev->this, prev->this->fops->lookup,
-                            &state->loc, state->dict);
+                            &state->loc, state->xdata);
                 return 0;
         }
 
-        fuse_entry_cbk (frame, cookie, this, op_ret, op_errno, inode, stat);
+        fuse_entry_cbk (frame, cookie, this, op_ret, op_errno, inode, stat,
+                        dict);
         return 0;
 }
 
@@ -237,77 +540,106 @@ fuse_lookup_resume (fuse_state_t *state)
                 free_fuse_state (state);
                 return;
         }
-        if (!state->loc.inode)
+
+        /* parent was resolved, entry could not, may be a missing gfid?
+         * Hence try to do a regular lookup
+         */
+        if ((state->resolve.op_ret == -1)
+            && (state->resolve.op_errno == ENODATA)) {
+                state->resolve.op_ret = 0;
+        }
+
+        if (state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                        "%"PRIu64": LOOKUP %s(%s)", state->finh->unique,
+                        state->loc.path, uuid_utoa (state->loc.inode->gfid));
+                state->is_revalidate = 1;
+        } else {
+                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                        "%"PRIu64": LOOKUP %s", state->finh->unique,
+                        state->loc.path);
                 state->loc.inode = inode_new (state->loc.parent->table);
+                if (uuid_is_null (state->gfid))
+                        uuid_generate (state->gfid);
+                fuse_gfid_set (state);
+        }
 
         FUSE_FOP (state, fuse_lookup_cbk, GF_FOP_LOOKUP,
-                  lookup, &state->loc, state->dict);
+                  lookup, &state->loc, state->xdata);
 }
 
 static void
 fuse_lookup (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
-        char         *name      = msg;
-        fuse_state_t *state     = NULL;
-        int32_t       ret       = -1;
+        char           *name     = msg;
+        fuse_state_t   *state    = NULL;
 
         GET_STATE (this, finh, state);
 
-        uuid_generate (state->gfid);
+        (void) fuse_resolve_entry_init (state, &state->resolve,
+                                        finh->nodeid, name);
 
-        ret = fuse_loc_fill (&state->loc, state, 0, finh->nodeid, name);
+        fuse_resolve_and_resume (state, fuse_lookup_resume);
 
-        if (ret < 0) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": LOOKUP %"PRIu64"/%s (fuse_loc_fill() failed)",
-                        finh->unique, finh->nodeid, name);
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+        return;
+}
 
-        if (state->loc.inode) {
-                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                        "%"PRIu64": LOOKUP %s(%"PRId64")", finh->unique,
-                        state->loc.path, state->loc.inode->ino);
-                state->is_revalidate = 1;
-        } else {
-                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                        "%"PRIu64": LOOKUP %s", finh->unique,
-                        state->loc.path);
-        }
+static inline void
+do_forget(xlator_t *this, uint64_t unique, uint64_t nodeid, uint64_t nlookup)
+{
+	inode_t *fuse_inode = fuse_ino_to_inode(nodeid, this);
 
-        fuse_resolve_and_resume (state, fuse_lookup_resume);
+	fuse_log_eh(this, "%"PRIu64": FORGET %"PRIu64"/%"PRIu64" gfid: (%s)",
+		    unique, nodeid, nlookup, uuid_utoa(fuse_inode->gfid));
 
+	inode_forget(fuse_inode, nlookup);
+	inode_unref(fuse_inode);
 }
 
-
 static void
 fuse_forget (xlator_t *this, fuse_in_header_t *finh, void *msg)
 
 {
-        struct fuse_forget_in *ffi = msg;
-
-        inode_t      *fuse_inode;
+        struct fuse_forget_in *ffi        = msg;
 
         if (finh->nodeid == 1) {
                 GF_FREE (finh);
                 return;
         }
 
-        fuse_inode = fuse_ino_to_inode (finh->nodeid, this);
+        gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                "%"PRIu64": FORGET %"PRIu64"/%"PRIu64,
+                finh->unique, finh->nodeid, ffi->nlookup);
 
-        inode_forget (fuse_inode, ffi->nlookup);
-        inode_unref (fuse_inode);
+	do_forget(this, finh->unique, finh->nodeid, ffi->nlookup);
 
         GF_FREE (finh);
 }
 
+static void
+fuse_batch_forget(xlator_t *this, fuse_in_header_t *finh, void *msg)
+{
+	struct fuse_batch_forget_in *fbfi = msg;
+	struct fuse_forget_one *ffo = (struct fuse_forget_one *) (fbfi + 1);
+	int i;
+
+	gf_log("glusterfs-fuse", GF_LOG_TRACE,
+		"%"PRIu64": BATCH_FORGET %"PRIu64"/%"PRIu32,
+		finh->unique, finh->nodeid, fbfi->count);
+
+	for (i = 0; i < fbfi->count; i++) {
+                if (ffo[i].nodeid == 1)
+                        continue;
+		do_forget(this, finh->unique, ffo[i].nodeid, ffo[i].nlookup);
+        }
+
+	GF_FREE(finh);
+}
 
 static int
 fuse_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                   struct iatt *postbuf)
+                   struct iatt *postbuf, dict_t *xdata)
 {
         fuse_state_t     *state;
         fuse_in_header_t *finh;
@@ -318,17 +650,17 @@ fuse_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         state = frame->root->state;
         finh  = state->finh;
 
+        fuse_log_eh_fop(this, state, frame, op_ret, op_errno);
+
         if (op_ret == 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                        "%"PRIu64": %s() %s => %"PRId64, frame->root->unique,
+                        "%"PRIu64": %s() %s => %"PRIu64, frame->root->unique,
                         gf_fop_list[frame->root->op],
                         state->loc.path ? state->loc.path : "ERR",
                         prebuf->ia_ino);
 
-                /* TODO: make these timeouts configurable via meta */
-                /* TODO: what if the inode number has changed by now */
                 postbuf->ia_blksize = this->ctx->page_size;
-                gf_fuse_stat2attr (postbuf, &fao.attr);
+                gf_fuse_stat2attr (postbuf, &fao.attr, priv->enable_ino32);
 
                 fao.attr_valid = calc_timeout_sec (priv->attribute_timeout);
                 fao.attr_valid_nsec =
@@ -358,10 +690,9 @@ fuse_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         return 0;
 }
 
-
 static int
 fuse_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-               int32_t op_ret, int32_t op_errno, struct iatt *buf)
+               int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
         fuse_state_t     *state;
         fuse_in_header_t *finh;
@@ -372,17 +703,19 @@ fuse_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         state = frame->root->state;
         finh  = state->finh;
 
+        fuse_log_eh (this, "op_ret: %d, op_errno: %d, %"PRIu64": %s() %s => "
+                   "gfid: %s", op_ret, op_errno, frame->root->unique,
+                   gf_fop_list[frame->root->op], state->loc.path,
+                   state->loc.inode ? uuid_utoa (state->loc.inode->gfid) : "");
         if (op_ret == 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                        "%"PRIu64": %s() %s => %"PRId64, frame->root->unique,
+                        "%"PRIu64": %s() %s => %"PRIu64, frame->root->unique,
                         gf_fop_list[frame->root->op],
                         state->loc.path ? state->loc.path : "ERR",
                         buf->ia_ino);
 
-                /* TODO: make these timeouts configurable via meta */
-                /* TODO: what if the inode number has changed by now */
                 buf->ia_blksize = this->ctx->page_size;
-                gf_fuse_stat2attr (buf, &fao.attr);
+                gf_fuse_stat2attr (buf, &fao.attr, priv->enable_ino32);
 
                 fao.attr_valid = calc_timeout_sec (priv->attribute_timeout);
                 fao.attr_valid_nsec =
@@ -397,9 +730,9 @@ fuse_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 send_fuse_obj (this, finh, &fao);
 #endif
         } else {
-                GF_LOG_OCCASIONALLY ( gf_fuse_conn_err_log, "glusterfs-fuse", 
-                                      GF_LOG_WARNING, 
-                                      "%"PRIu64": %s() %s => -1 (%s)", 
+                GF_LOG_OCCASIONALLY ( gf_fuse_conn_err_log, "glusterfs-fuse",
+                                      GF_LOG_WARNING,
+                                      "%"PRIu64": %s() %s => -1 (%s)",
                                       frame->root->unique,
                                       gf_fop_list[frame->root->op],
                                       state->loc.path ? state->loc.path : "ERR",
@@ -414,14 +747,13 @@ fuse_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         return 0;
 }
 
-
 static int
 fuse_root_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                       int32_t op_ret, int32_t op_errno,
                       inode_t *inode, struct iatt *stat, dict_t *dict,
                       struct iatt *postparent)
 {
-        fuse_attr_cbk (frame, cookie, this, op_ret, op_errno, stat);
+        fuse_attr_cbk (frame, cookie, this, op_ret, op_errno, stat, dict);
 
         return 0;
 }
@@ -429,14 +761,28 @@ fuse_root_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 void
 fuse_getattr_resume (fuse_state_t *state)
 {
-        if (!state->fd || IA_ISDIR (state->loc.inode->ia_type)) {
+        if (!state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "%"PRIu64": GETATTR %"PRIu64" (%s) resolution failed",
+                        state->finh->unique, state->finh->nodeid,
+                        uuid_utoa (state->resolve.gfid));
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
+
+        if (!IA_ISDIR (state->loc.inode->ia_type)) {
+                state->fd = fd_lookup (state->loc.inode, state->finh->pid);
+        }
+
+        if (!state->fd) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                         "%"PRIu64": GETATTR %"PRIu64" (%s)",
                         state->finh->unique, state->finh->nodeid,
                         state->loc.path);
 
                 FUSE_FOP (state, fuse_attr_cbk, GF_FOP_STAT,
-                          stat, &state->loc);
+                          stat, &state->loc, state->xdata);
         } else {
 
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
@@ -445,7 +791,7 @@ fuse_getattr_resume (fuse_state_t *state)
                         state->loc.path, state->fd);
 
                 FUSE_FOP (state, fuse_attr_cbk, GF_FOP_FSTAT,
-                          fstat, state->fd);
+                          fstat, state->fd, state->xdata);
         }
 }
 
@@ -453,7 +799,6 @@ static void
 fuse_getattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
         fuse_state_t *state;
-        fd_t         *fd = NULL;
         int32_t       ret = -1;
 
         GET_STATE (this, finh, state);
@@ -474,59 +819,72 @@ fuse_getattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
                 fuse_gfid_set (state);
 
                 FUSE_FOP (state, fuse_root_lookup_cbk, GF_FOP_LOOKUP,
-                          lookup, &state->loc, state->dict);
+                          lookup, &state->loc, state->xdata);
                 return;
         }
 
-        ret = fuse_loc_fill (&state->loc, state, state->finh->nodeid, 0, NULL);
+        fuse_resolve_inode_init (state, &state->resolve, state->finh->nodeid);
 
-        if (!state->loc.inode) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": GETATTR %"PRIu64" (%s) (fuse_loc_fill() returned NULL inode)",
-                        state->finh->unique, state->finh->nodeid, state->loc.path);
-                send_fuse_err (state->this, state->finh, ENOENT);
-                free_fuse_state (state);
-                return;
+        fuse_resolve_and_resume (state, fuse_getattr_resume);
+}
+
+static int32_t
+fuse_fd_inherit_directio (xlator_t *this, fd_t *fd, struct fuse_open_out *foo)
+{
+        int32_t        ret    = 0;
+        fuse_fd_ctx_t *fdctx  = NULL, *tmp_fdctx = NULL;
+        fd_t          *tmp_fd = NULL;
+
+        GF_VALIDATE_OR_GOTO_WITH_ERROR ("glusterfs-fuse", this, out, ret,
+                                        -EINVAL);
+        GF_VALIDATE_OR_GOTO_WITH_ERROR ("glusterfs-fuse", fd, out, ret,
+                                        -EINVAL);
+        GF_VALIDATE_OR_GOTO_WITH_ERROR ("glusterfs-fuse", foo, out, ret,
+                                        -EINVAL);
+
+        fdctx = fuse_fd_ctx_get (this, fd);
+        if (!fdctx) {
+                ret = -ENOMEM;
+                goto out;
         }
 
-        fd = fd_lookup (state->loc.inode, state->finh->pid);
-        state->fd = fd;
-        if (!fd || IA_ISDIR (state->loc.inode->ia_type)) {
-                /* this is the @ret of fuse_loc_fill, checked here
-                   to permit fstat() to happen even when fuse_loc_fill fails
-                */
-                if (ret < 0) {
-                        gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                                "%"PRIu64": GETATTR %"PRIu64" (fuse_loc_fill() failed)",
-                                state->finh->unique, state->finh->nodeid);
-                        send_fuse_err (state->this, state->finh, ENOENT);
-                        free_fuse_state (state);
-                        return;
+        tmp_fd = fd_lookup (fd->inode, 0);
+        if (tmp_fd) {
+                tmp_fdctx = fuse_fd_ctx_get (this, tmp_fd);
+                if (tmp_fdctx) {
+                        foo->open_flags &= ~FOPEN_DIRECT_IO;
+                        foo->open_flags |= (tmp_fdctx->open_flags
+                                            & FOPEN_DIRECT_IO);
                 }
+        }
 
-                if (state->fd)
-                        fd_unref (state->fd);
+        fdctx->open_flags |= (foo->open_flags & FOPEN_DIRECT_IO);
 
-                state->fd = NULL;
+        if (tmp_fd != NULL) {
+                fd_unref (tmp_fd);
         }
 
-        fuse_resolve_and_resume (state, fuse_getattr_resume);
+        ret = 0;
+out:
+        return ret;
 }
 
-
 static int
 fuse_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-             int32_t op_ret, int32_t op_errno, fd_t *fd)
+             int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
-        fuse_state_t          *state;
-        fuse_in_header_t      *finh;
-        fuse_private_t        *priv = NULL;
-        struct fuse_open_out   foo = {0, };
+        fuse_state_t         *state  = NULL;
+        fuse_in_header_t     *finh   = NULL;
+        fuse_private_t       *priv   = NULL;
+        int32_t               ret    = 0;
+        struct fuse_open_out  foo    = {0, };
 
         priv = this->private;
         state = frame->root->state;
         finh = state->finh;
 
+        fuse_log_eh_fop(this, state, frame, op_ret, op_errno);
+
         if (op_ret >= 0) {
                 foo.fh = (uintptr_t) fd;
                 foo.open_flags = 0;
@@ -537,17 +895,27 @@ fuse_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                             || (priv->direct_io_mode == 1))
                                 foo.open_flags |= FOPEN_DIRECT_IO;
 #ifdef GF_DARWIN_HOST_OS
-                                /* In Linux: by default, buffer cache
-                                 * is purged upon open, setting
-                                 * FOPEN_KEEP_CACHE implies no-purge
-                                 *
-                                 * In MacFUSE: by default, buffer cache
-                                 * is left intact upon open, setting
-                                 * FOPEN_PURGE_UBC implies purge
-                                 *
-                                 * [[Innnnteresting...]]
-                                 */
+                        /* In Linux: by default, buffer cache
+                         * is purged upon open, setting
+                         * FOPEN_KEEP_CACHE implies no-purge
+                         *
+                         * In MacFUSE: by default, buffer cache
+                         * is left intact upon open, setting
+                         * FOPEN_PURGE_UBC implies purge
+                         *
+                         * [[Interesting...]]
+                         */
+                        if (!priv->fopen_keep_cache)
                                 foo.open_flags |= FOPEN_PURGE_UBC;
+#else
+                        /*
+                         * If fopen-keep-cache is enabled, we set the associated
+                         * flag here such that files are not invalidated on open.
+                         * File invalidations occur either in fuse or explicitly
+                         * when the cache is set invalid on the inode.
+                         */
+                        if (priv->fopen_keep_cache)
+                                foo.open_flags |= FOPEN_KEEP_CACHE;
 #endif
                 }
 
@@ -555,22 +923,33 @@ fuse_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         "%"PRIu64": %s() %s => %p", frame->root->unique,
                         gf_fop_list[frame->root->op], state->loc.path, fd);
 
-                fd_ref (fd);
+                ret = fuse_fd_inherit_directio (this, fd, &foo);
+                if (ret < 0) {
+                        op_errno = -ret;
+                        gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                                "cannot inherit direct-io values for fd "
+                                "(ptr:%p inode-gfid:%s) from fds already "
+                                "opened", fd, uuid_utoa (fd->inode->gfid));
+                        goto err;
+                }
+
                 if (send_fuse_obj (this, finh, &foo) == ENOENT) {
                         gf_log ("glusterfs-fuse", GF_LOG_DEBUG,
                                 "open(%s) got EINTR", state->loc.path);
-                        fd_unref (fd);
-                                goto out;
+                        gf_fd_put (priv->fdtable, state->fd_no);
+                        goto out;
                 }
 
                 fd_bind (fd);
         } else {
+        err:
                 gf_log ("glusterfs-fuse", GF_LOG_WARNING,
                         "%"PRIu64": %s() %s => -1 (%s)", frame->root->unique,
                         gf_fop_list[frame->root->op], state->loc.path,
                         strerror (op_errno));
 
                 send_fuse_err (this, finh, op_errno);
+                gf_fd_put (priv->fdtable, state->fd_no);
         }
 out:
         free_fuse_state (state);
@@ -578,26 +957,24 @@ out:
         return 0;
 }
 
-
 static void
 fuse_do_truncate (fuse_state_t *state, size_t size)
 {
         if (state->fd) {
                 FUSE_FOP (state, fuse_truncate_cbk, GF_FOP_FTRUNCATE,
-                          ftruncate, state->fd, size);
+                          ftruncate, state->fd, size, state->xdata);
         } else {
                 FUSE_FOP (state, fuse_truncate_cbk, GF_FOP_TRUNCATE,
-                          truncate, &state->loc, size);
+                          truncate, &state->loc, size, state->xdata);
         }
 
         return;
 }
 
-
 static int
 fuse_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno,
-                  struct iatt *statpre, struct iatt *statpost)
+                  struct iatt *statpre, struct iatt *statpost, dict_t *xdata)
 {
         fuse_state_t     *state;
         fuse_in_header_t *finh;
@@ -610,19 +987,20 @@ fuse_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         state = frame->root->state;
         finh  = state->finh;
 
+        fuse_log_eh(this, "op_ret: %d, op_errno: %d, %"PRIu64", %s() %s => "
+                    "gfid: %s", op_ret, op_errno, frame->root->unique,
+                    gf_fop_list[frame->root->op], state->loc.path,
+                    state->loc.inode ? uuid_utoa (state->loc.inode->gfid) : "");
+
         if (op_ret == 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                        "%"PRIu64": %s() %s => %"PRId64, frame->root->unique,
+                        "%"PRIu64": %s() %s => %"PRIu64, frame->root->unique,
                         gf_fop_list[frame->root->op],
                         state->loc.path ? state->loc.path : "ERR",
                         statpost->ia_ino);
 
-                /* TODO: make these timeouts configurable via meta */
-                /* TODO: what if the inode number has changed by now */
-
                 statpost->ia_blksize = this->ctx->page_size;
-
-                gf_fuse_stat2attr (statpost, &fao.attr);
+                gf_fuse_stat2attr (statpost, &fao.attr, priv->enable_ino32);
 
                 fao.attr_valid = calc_timeout_sec (priv->attribute_timeout);
                 fao.attr_valid_nsec =
@@ -661,7 +1039,6 @@ fuse_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         return 0;
 }
 
-
 static int32_t
 fattr_to_gf_set_attr (int32_t valid)
 {
@@ -688,7 +1065,6 @@ fattr_to_gf_set_attr (int32_t valid)
         return gf_valid;
 }
 
-
 #define FATTR_MASK   (FATTR_SIZE                        \
                       | FATTR_UID | FATTR_GID           \
                       | FATTR_ATIME | FATTR_MTIME       \
@@ -697,6 +1073,26 @@ fattr_to_gf_set_attr (int32_t valid)
 void
 fuse_setattr_resume (fuse_state_t *state)
 {
+        if (!state->fd && !state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "%"PRIu64": SETATTR %"PRIu64" (%s) resolution failed",
+                        state->finh->unique, state->finh->nodeid,
+                        uuid_utoa (state->resolve.gfid));
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
+
+        gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                "%"PRIu64": SETATTR (%"PRIu64")%s", state->finh->unique,
+                state->finh->nodeid, state->loc.path);
+
+#ifdef GF_TEST_FFOP
+        /* this is for calls like 'fchmod()' */
+        if (!state->fd)
+                state->fd = fd_lookup (state->loc.inode, state->finh->pid);
+#endif /* GF_TEST_FFOP */
+
         if ((state->valid & (FATTR_MASK)) != FATTR_SIZE) {
                 if (state->fd &&
                     !((state->valid & FATTR_ATIME) ||
@@ -708,11 +1104,13 @@ fuse_setattr_resume (fuse_state_t *state)
 
                         FUSE_FOP (state, fuse_setattr_cbk, GF_FOP_FSETATTR,
                                   fsetattr, state->fd, &state->attr,
-                                  fattr_to_gf_set_attr (state->valid));
+                                  fattr_to_gf_set_attr (state->valid),
+                                  state->xdata);
                 } else {
                         FUSE_FOP (state, fuse_setattr_cbk, GF_FOP_SETATTR,
                                   setattr, &state->loc, &state->attr,
-                                  fattr_to_gf_set_attr (state->valid));
+                                  fattr_to_gf_set_attr (state->valid),
+                                  state->xdata);
                 }
         } else {
                 fuse_do_truncate (state, state->size);
@@ -727,18 +1125,18 @@ fuse_setattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
 
         fuse_private_t  *priv = NULL;
         fuse_state_t *state = NULL;
-        int32_t       ret   = -1;
 
         GET_STATE (this, finh, state);
 
         if (fsi->valid & FATTR_FH &&
-            !(fsi->valid & (FATTR_ATIME|FATTR_MTIME)))
+            !(fsi->valid & (FATTR_ATIME|FATTR_MTIME))) {
                 /* We need no loc if kernel sent us an fd and
                  * we are not fiddling with times */
-                ret = 1;
-        else
-                ret = fuse_loc_fill (&state->loc, state, finh->nodeid, 0,
-                                     NULL);
+                state->fd = FH_TO_FD (fsi->fh);
+                fuse_resolve_fd_init (state, &state->resolve, state->fd);
+        } else {
+                fuse_resolve_inode_init (state, &state->resolve, finh->nodeid);
+        }
 
         /*
          * This is just stub code demonstrating how to retrieve
@@ -759,29 +1157,8 @@ fuse_setattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
                 state->lk_owner = fsi->lock_owner;
 #endif
 
-        if ((state->loc.inode == NULL && ret == 0) ||
-            (ret < 0)) {
-
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": SETATTR %s (fuse_loc_fill() failed)",
-                        finh->unique, state->loc.path);
-
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-
-                return;
-        }
-
-        gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": SETATTR (%"PRIu64")%s", finh->unique,
-                finh->nodeid, state->loc.path);
-
         state->valid = fsi->valid;
 
-        if (fsi->valid & FATTR_FH) {
-                state->fd = FH_TO_FD (fsi->fh);
-        }
-
         if ((fsi->valid & (FATTR_MASK)) != FATTR_SIZE) {
                 if (fsi->valid & FATTR_SIZE) {
                         state->size            = fsi->size;
@@ -804,14 +1181,15 @@ fuse_setattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
         fuse_resolve_and_resume (state, fuse_setattr_resume);
 }
 
-
 static int
 fuse_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-              int32_t op_ret, int32_t op_errno)
+              int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         fuse_state_t *state = frame->root->state;
         fuse_in_header_t *finh = state->finh;
 
+        fuse_log_eh_fop(this, state, frame, op_ret, op_errno);
+
         if (op_ret == 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                         "%"PRIu64": %s() %s => 0", frame->root->unique,
@@ -836,19 +1214,17 @@ fuse_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         return 0;
 }
 
-
 static int
 fuse_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                struct iatt *postbuf)
+                struct iatt *postbuf, dict_t *xdata)
 {
-        return fuse_err_cbk (frame, cookie, this, op_ret, op_errno);
+        return fuse_err_cbk (frame, cookie, this, op_ret, op_errno, xdata);
 }
 
-
 static int
 fuse_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno)
+                   int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         if (op_ret == -1 && op_errno == ENOTSUP)
                 GF_LOG_OCCASIONALLY (gf_fuse_xattr_enotsup_log,
@@ -856,14 +1232,13 @@ fuse_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                      "extended attribute not supported "
                                      "by the backend storage");
 
-        return fuse_err_cbk (frame, cookie, this, op_ret, op_errno);
+        return fuse_err_cbk (frame, cookie, this, op_ret, op_errno, xdata);
 }
 
-
 static int
 fuse_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                 struct iatt *postparent)
+                 struct iatt *postparent, dict_t *xdata)
 {
         fuse_state_t     *state = NULL;
         fuse_in_header_t *finh = NULL;
@@ -871,11 +1246,14 @@ fuse_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         state = frame->root->state;
         finh = state->finh;
 
-        if (op_ret == 0)
-                inode_unlink (state->loc.inode, state->loc.parent,
-                              state->loc.name);
+        fuse_log_eh (this, "op_ret: %d, op_errno: %d, %"PRIu64": %s() %s => "
+                     "gfid: %s", op_ret, op_errno, frame->root->unique,
+                     gf_fop_list[frame->root->op], state->loc.path,
+                     state->loc.inode ? uuid_utoa (state->loc.inode->gfid) : "");
 
         if (op_ret == 0) {
+                inode_unlink (state->loc.inode, state->loc.parent,
+                              state->loc.name);
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                         "%"PRIu64": %s() %s => 0", frame->root->unique,
                         gf_fop_list[frame->root->op], state->loc.path);
@@ -900,48 +1278,46 @@ fuse_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 void
 fuse_access_resume (fuse_state_t *state)
 {
+        if (!state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "%"PRIu64": ACCESS %"PRIu64" (%s) resolution failed",
+                        state->finh->unique, state->finh->nodeid,
+                        uuid_utoa (state->resolve.gfid));
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
+
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                 "%"PRIu64" ACCESS %s/%"PRIu64" mask=%d",
                 state->finh->unique, state->loc.path,
                 state->finh->nodeid, state->mask);
 
         FUSE_FOP (state, fuse_err_cbk, GF_FOP_ACCESS, access,
-                  &state->loc, state->mask);
-
+                  &state->loc, state->mask, state->xdata);
 }
 
 static void
 fuse_access (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
         struct fuse_access_in *fai = msg;
-
         fuse_state_t *state = NULL;
-        int32_t       ret = -1;
 
         GET_STATE (this, finh, state);
 
-        ret = fuse_loc_fill (&state->loc, state, finh->nodeid, 0, NULL);
-        if ((state->loc.inode == NULL) ||
-            (ret < 0)) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": ACCESS %"PRIu64" (%s) (fuse_loc_fill() failed)",
-                        finh->unique, finh->nodeid, state->loc.path);
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+        fuse_resolve_inode_init (state, &state->resolve, finh->nodeid);
 
         state->mask = fai->mask;
 
         fuse_resolve_and_resume (state, fuse_access_resume);
+
         return;
 }
 
-
 static int
 fuse_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, const char *linkname,
-                   struct iatt *buf)
+                   struct iatt *buf, dict_t *xdata)
 {
         fuse_state_t     *state = NULL;
         fuse_in_header_t *finh = NULL;
@@ -949,6 +1325,12 @@ fuse_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         state = frame->root->state;
         finh = state->finh;
 
+        fuse_log_eh (this, "op_ret: %d, op_errno: %d %"PRIu64": %s() => %s"
+                     " linkname: %s, gfid: %s", op_ret, op_errno,
+                     frame->root->unique, gf_fop_list[frame->root->op],
+                     state->loc.gfid, linkname,
+                     uuid_utoa (state->loc.gfid));
+
         if (op_ret > 0) {
                 ((char *)linkname)[op_ret] = '\0';
 
@@ -974,32 +1356,31 @@ fuse_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 void
 fuse_readlink_resume (fuse_state_t *state)
 {
+        if (!state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "READLINK %"PRIu64" (%s) resolution failed",
+                        state->finh->unique, uuid_utoa (state->resolve.gfid));
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
+
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64" READLINK %s/%"PRId64, state->finh->unique,
-                state->loc.path, state->loc.inode->ino);
+                "%"PRIu64" READLINK %s/%s", state->finh->unique,
+                state->loc.path, uuid_utoa (state->loc.inode->gfid));
 
         FUSE_FOP (state, fuse_readlink_cbk, GF_FOP_READLINK,
-                  readlink, &state->loc, 4096);
-
+                  readlink, &state->loc, 4096, state->xdata);
 }
 
 static void
 fuse_readlink (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
         fuse_state_t *state = NULL;
-        int32_t       ret = -1;
 
         GET_STATE (this, finh, state);
-        ret = fuse_loc_fill (&state->loc, state, finh->nodeid, 0, NULL);
-        if ((state->loc.inode == NULL) ||
-            (ret < 0)) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64" READLINK %s (fuse_loc_fill() returned NULL inode)",
-                        finh->unique, state->loc.path);
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+
+        fuse_resolve_inode_init (state, &state->resolve, finh->nodeid);
 
         fuse_resolve_and_resume (state, fuse_readlink_resume);
 
@@ -1010,25 +1391,35 @@ void
 fuse_mknod_resume (fuse_state_t *state)
 {
         if (!state->loc.parent) {
-                gf_log ("fuse", GF_LOG_ERROR, "failed to resolve path %s",
-                        state->loc.path);
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "MKNOD %"PRIu64"/%s (%s/%s) resolution failed",
+                        state->finh->nodeid, state->resolve.bname,
+                        uuid_utoa (state->resolve.gfid), state->resolve.bname);
                 send_fuse_err (state->this, state->finh, ENOENT);
                 free_fuse_state (state);
                 return;
         }
 
-        if (!state->loc.inode)
-                state->loc.inode = inode_new (state->loc.parent->table);
-        else
-                gf_log (state->this->name, GF_LOG_ERROR, "inode already present");
+        if (state->resolve.op_errno == ENOENT) {
+                state->resolve.op_ret = 0;
+                state->resolve.op_errno = 0;
+        }
+
+        if (state->loc.inode) {
+                gf_log (state->this->name, GF_LOG_DEBUG, "inode already present");
+                inode_unref (state->loc.inode);
+                state->loc.inode = NULL;
+        }
+
+        state->loc.inode = inode_new (state->loc.parent->table);
 
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                 "%"PRIu64": MKNOD %s", state->finh->unique,
                 state->loc.path);
 
         FUSE_FOP (state, fuse_newentry_cbk, GF_FOP_MKNOD,
-                  mknod, &state->loc, state->mode, state->rdev, state->dict);
-
+                  mknod, &state->loc, state->mode, state->rdev, state->umask,
+                  state->xdata);
 }
 
 static void
@@ -1051,19 +1442,16 @@ fuse_mknod (xlator_t *this, fuse_in_header_t *finh, void *msg)
 
         uuid_generate (state->gfid);
 
-        ret = fuse_loc_fill (&state->loc, state, 0, finh->nodeid, name);
-        if (ret < 0) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64" MKNOD %s (fuse_loc_fill() failed)",
-                        finh->unique, state->loc.path);
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+        fuse_resolve_entry_init (state, &state->resolve, finh->nodeid, name);
 
         state->mode = fmi->mode;
         state->rdev = fmi->rdev;
 
+        priv = this->private;
+#if FUSE_KERNEL_MINOR_VERSION >=12
+        FUSE_ENTRY_CREATE(this, priv, finh, state, fmi, "MKNOD");
+#endif
+
         fuse_resolve_and_resume (state, fuse_mknod_resume);
 
         return;
@@ -1073,24 +1461,34 @@ void
 fuse_mkdir_resume (fuse_state_t *state)
 {
         if (!state->loc.parent) {
-                gf_log ("fuse", GF_LOG_ERROR, "failed to resolve path %s",
-                        state->loc.path);
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "MKDIR %"PRIu64" (%s/%s) resolution failed",
+                        state->finh->nodeid, uuid_utoa (state->resolve.gfid),
+                        state->resolve.bname);
                 send_fuse_err (state->this, state->finh, ENOENT);
                 free_fuse_state (state);
                 return;
         }
 
-        if (!state->loc.inode)
-                state->loc.inode = inode_new (state->loc.parent->table);
-        else
-                gf_log (state->this->name, GF_LOG_ERROR, "inode already present");
+        if (state->resolve.op_errno == ENOENT) {
+                state->resolve.op_ret = 0;
+                state->resolve.op_errno = 0;
+        }
+
+        if (state->loc.inode) {
+                gf_log (state->this->name, GF_LOG_DEBUG, "inode already present");
+                inode_unref (state->loc.inode);
+                state->loc.inode = NULL;
+        }
+
+        state->loc.inode = inode_new (state->loc.parent->table);
 
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                 "%"PRIu64": MKDIR %s", state->finh->unique,
                 state->loc.path);
 
         FUSE_FOP (state, fuse_newentry_cbk, GF_FOP_MKDIR,
-                  mkdir, &state->loc, state->mode, state->dict);
+                  mkdir, &state->loc, state->mode, state->umask, state->xdata);
 }
 
 static void
@@ -1098,6 +1496,7 @@ fuse_mkdir (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
         struct fuse_mkdir_in *fmi = msg;
         char *name = (char *)(fmi + 1);
+        fuse_private_t *priv = NULL;
 
         fuse_state_t *state;
         int32_t ret = -1;
@@ -1106,18 +1505,15 @@ fuse_mkdir (xlator_t *this, fuse_in_header_t *finh, void *msg)
 
         uuid_generate (state->gfid);
 
-        ret = fuse_loc_fill (&state->loc, state, 0, finh->nodeid, name);
-        if (ret < 0) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64" MKDIR %s (fuse_loc_fill() failed)",
-                        finh->unique, state->loc.path);
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+        fuse_resolve_entry_init (state, &state->resolve, finh->nodeid, name);
 
         state->mode = fmi->mode;
 
+        priv = this->private;
+#if FUSE_KERNEL_MINOR_VERSION >=12
+        FUSE_ENTRY_CREATE(this, priv, finh, state, fmi, "MKDIR");
+#endif
+
         fuse_resolve_and_resume (state, fuse_mkdir_resume);
 
         return;
@@ -1126,39 +1522,33 @@ fuse_mkdir (xlator_t *this, fuse_in_header_t *finh, void *msg)
 void
 fuse_unlink_resume (fuse_state_t *state)
 {
-        if (!state->loc.inode) {
-                gf_log ("fuse", GF_LOG_WARNING, "path resolving failed");
+        if (!state->loc.parent || !state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "UNLINK %"PRIu64" (%s/%s) resolution failed",
+                        state->finh->nodeid, uuid_utoa (state->resolve.gfid),
+                        state->resolve.bname);
                 send_fuse_err (state->this, state->finh, ENOENT);
                 free_fuse_state (state);
                 return;
         }
+
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                 "%"PRIu64": UNLINK %s", state->finh->unique,
                 state->loc.path);
 
         FUSE_FOP (state, fuse_unlink_cbk, GF_FOP_UNLINK,
-                  unlink, &state->loc);
-
+                  unlink, &state->loc, 0, state->xdata);
 }
 
 static void
 fuse_unlink (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
         char         *name = msg;
-
         fuse_state_t *state = NULL;
-        int32_t       ret = -1;
 
         GET_STATE (this, finh, state);
-        ret = fuse_loc_fill (&state->loc, state, 0, finh->nodeid, name);
-        if (ret < 0) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": UNLINK %s (fuse_loc_fill() returned NULL inode)",
-                        finh->unique, state->loc.path);
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+
+        fuse_resolve_entry_init (state, &state->resolve, finh->nodeid, name);
 
         fuse_resolve_and_resume (state, fuse_unlink_resume);
 
@@ -1168,8 +1558,11 @@ fuse_unlink (xlator_t *this, fuse_in_header_t *finh, void *msg)
 void
 fuse_rmdir_resume (fuse_state_t *state)
 {
-        if (!state->loc.inode) {
-                gf_log ("fuse", GF_LOG_WARNING, "path resolving failed");
+        if (!state->loc.parent || !state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "RMDIR %"PRIu64" (%s/%s) resolution failed",
+                        state->finh->nodeid, uuid_utoa (state->resolve.gfid),
+                        state->resolve.bname);
                 send_fuse_err (state->this, state->finh, ENOENT);
                 free_fuse_state (state);
                 return;
@@ -1180,29 +1573,21 @@ fuse_rmdir_resume (fuse_state_t *state)
                 state->loc.path);
 
         FUSE_FOP (state, fuse_unlink_cbk, GF_FOP_RMDIR,
-                  rmdir, &state->loc);
+                  rmdir, &state->loc, 0, state->xdata);
 }
 
 static void
 fuse_rmdir (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
         char         *name = msg;
-
         fuse_state_t *state = NULL;
-        int32_t       ret = -1;
 
         GET_STATE (this, finh, state);
-        ret = fuse_loc_fill (&state->loc, state, 0, finh->nodeid, name);
-        if (ret < 0) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": RMDIR %s (fuse_loc_fill() failed)",
-                        finh->unique, state->loc.path);
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+
+        fuse_resolve_entry_init (state, &state->resolve, finh->nodeid, name);
 
         fuse_resolve_and_resume (state, fuse_rmdir_resume);
+
         return;
 }
 
@@ -1210,60 +1595,62 @@ void
 fuse_symlink_resume (fuse_state_t *state)
 {
         if (!state->loc.parent) {
-                gf_log ("fuse", GF_LOG_ERROR, "failed to resolve path %s",
-                        state->loc.path);
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "SYMLINK %"PRIu64" (%s/%s) -> %s resolution failed",
+                        state->finh->nodeid, uuid_utoa (state->resolve.gfid),
+                        state->resolve.bname, state->name);
                 send_fuse_err (state->this, state->finh, ENOENT);
                 free_fuse_state (state);
                 return;
         }
 
-        if (!state->loc.inode)
-                state->loc.inode = inode_new (state->loc.parent->table);
-        else
-                gf_log (state->this->name, GF_LOG_ERROR, "inode already present");
+        if (state->resolve.op_errno == ENOENT) {
+                state->resolve.op_ret = 0;
+                state->resolve.op_errno = 0;
+        }
+
+        if (state->loc.inode) {
+                gf_log (state->this->name, GF_LOG_DEBUG, "inode already present");
+                inode_unref (state->loc.inode);
+                state->loc.inode = NULL;
+        }
+
+        state->loc.inode = inode_new (state->loc.parent->table);
 
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                 "%"PRIu64": SYMLINK %s -> %s", state->finh->unique,
                 state->loc.path, state->name);
 
         FUSE_FOP (state, fuse_newentry_cbk, GF_FOP_SYMLINK,
-                  symlink, state->name, &state->loc, state->dict);
+                  symlink, state->name, &state->loc, state->umask, state->xdata);
 }
 
 static void
 fuse_symlink (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
-        char *name = msg;
-        char *linkname = name + strlen (name) + 1;
-
+        char         *name = msg;
+        char         *linkname = name + strlen (name) + 1;
         fuse_state_t *state = NULL;
-        int32_t       ret = -1;
 
         GET_STATE (this, finh, state);
 
         uuid_generate (state->gfid);
 
-        ret = fuse_loc_fill (&state->loc, state, 0, finh->nodeid, name);
-        if (ret < 0) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64" SYMLINK %s -> %s (fuse_loc_fill() failed)",
-                        finh->unique, state->loc.path, linkname);
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+        fuse_resolve_entry_init (state, &state->resolve, finh->nodeid, name);
 
         state->name = gf_strdup (linkname);
+
         fuse_resolve_and_resume (state, fuse_symlink_resume);
+
         return;
 }
 
-
 int
 fuse_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno, struct iatt *buf,
                  struct iatt *preoldparent, struct iatt *postoldparent,
-                 struct iatt *prenewparent, struct iatt *postnewparent)
+                 struct iatt *prenewparent, struct iatt *postnewparent,
+                 dict_t *xdata)
 {
         fuse_state_t     *state = NULL;
         fuse_in_header_t *finh = NULL;
@@ -1271,17 +1658,25 @@ fuse_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         state = frame->root->state;
         finh  = state->finh;
 
+        fuse_log_eh (this, "op_ret: %d, op_errno: %d, %"PRIu64": %s() "
+                     "path: %s parent: %s ==> path: %s parent: %s"
+                     "gfid: %s", op_ret, op_errno, frame->root->unique,
+                     gf_fop_list[frame->root->op], state->loc.path,
+                     state->loc.parent?uuid_utoa (state->loc.parent->gfid):"",
+                     state->loc2.path,
+                     state->loc2.parent?uuid_utoa (state->loc2.parent->gfid):"",
+                     state->loc.inode?uuid_utoa (state->loc.inode->gfid):"");
+
         if (op_ret == 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                        "%"PRIu64": %s -> %s => 0 (buf->ia_ino=%"PRId64" , loc->ino=%"PRId64")",
+                        "%"PRIu64": %s -> %s => 0 (buf->ia_ino=%"PRIu64")",
                         frame->root->unique, state->loc.path, state->loc2.path,
-                        buf->ia_ino, state->loc.ino);
+                        buf->ia_ino);
 
                 {
                         /* ugly ugly - to stay blind to situation where
                            rename happens on a new inode
                         */
-                        buf->ia_ino = state->loc.ino;
                         buf->ia_type = state->loc.inode->ia_type;
                 }
                 buf->ia_blksize = this->ctx->page_size;
@@ -1308,19 +1703,47 @@ fuse_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 void
 fuse_rename_resume (fuse_state_t *state)
 {
-        if (!state->loc.inode) {
+        char loc_uuid[64]  = {0,};
+        char loc2_uuid[64] = {0,};
+
+        if (!state->loc.parent || !state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "RENAME %"PRIu64" %s/%s -> %s/%s src resolution failed",
+                        state->finh->unique,
+                        uuid_utoa_r (state->resolve.gfid, loc_uuid),
+                        state->resolve.bname,
+                        uuid_utoa_r (state->resolve2.gfid, loc2_uuid),
+                        state->resolve2.bname);
+
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
+
+        if (!state->loc2.parent) {
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "RENAME %"PRIu64" %s/%s -> %s/%s dst resolution failed",
+                        state->finh->unique,
+                        uuid_utoa_r (state->resolve.gfid, loc_uuid),
+                        state->resolve.bname,
+                        uuid_utoa_r (state->resolve2.gfid, loc2_uuid),
+                        state->resolve2.bname);
+
                 send_fuse_err (state->this, state->finh, ENOENT);
                 free_fuse_state (state);
                 return;
         }
 
+        state->resolve.op_ret = 0;
+        state->resolve2.op_ret = 0;
+
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": RENAME `%s (%"PRId64")' -> `%s (%"PRId64")'",
-                state->finh->unique, state->loc.path, state->loc.ino,
-                state->loc2.path, state->loc2.ino);
+                "%"PRIu64": RENAME `%s (%s)' -> `%s (%s)'",
+                state->finh->unique, state->loc.path, loc_uuid,
+                state->loc2.path, loc2_uuid);
 
         FUSE_FOP (state, fuse_rename_cbk, GF_FOP_RENAME,
-                  rename, &state->loc, &state->loc2);
+                  rename, &state->loc, &state->loc2, state->xdata);
 }
 
 static void
@@ -1329,35 +1752,13 @@ fuse_rename (xlator_t *this, fuse_in_header_t *finh, void *msg)
         struct fuse_rename_in  *fri = msg;
         char *oldname = (char *)(fri + 1);
         char *newname = oldname + strlen (oldname) + 1;
-
         fuse_state_t *state = NULL;
-        int32_t       ret = -1;
 
         GET_STATE (this, finh, state);
 
-        ret = fuse_loc_fill (&state->loc, state, 0, finh->nodeid, oldname);
-        if (ret < 0) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "for %s %"PRIu64": RENAME `%s' -> `%s' (fuse_loc_fill() failed)",
-                        state->loc.path, finh->unique, state->loc.path,
-                        state->loc2.path);
+        fuse_resolve_entry_init (state, &state->resolve, finh->nodeid, oldname);
 
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
-
-        ret = fuse_loc_fill (&state->loc2, state, 0, fri->newdir, newname);
-        if (ret < 0) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "for %s %"PRIu64": RENAME `%s' -> `%s' (fuse_loc_fill() failed)",
-                        state->loc.path, finh->unique, state->loc.path,
-                        state->loc2.path);
-
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+        fuse_resolve_entry_init (state, &state->resolve2, fri->newdir, newname);
 
         fuse_resolve_and_resume (state, fuse_rename_resume);
 
@@ -1367,18 +1768,31 @@ fuse_rename (xlator_t *this, fuse_in_header_t *finh, void *msg)
 void
 fuse_link_resume (fuse_state_t *state)
 {
-        if (!state->loc.inode)
-                state->loc.inode = inode_ref (state->loc2.inode);
-        else
-                gf_log (state->this->name, GF_LOG_ERROR, "inode already present");
+        if (!state->loc2.inode || !state->loc.parent) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "fuse_loc_fill() failed %"PRIu64": LINK %s %s",
+                        state->finh->unique, state->loc2.path, state->loc.path);
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
+
+        state->resolve.op_ret = 0;
+        state->resolve2.op_ret = 0;
+
+        if (state->loc.inode) {
+                inode_unref (state->loc.inode);
+                state->loc.inode = NULL;
+        }
+        state->loc.inode = inode_ref (state->loc2.inode);
 
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": LINK() %s (%"PRId64") -> %s (%"PRId64")",
-                state->finh->unique, state->loc2.path, state->loc2.ino,
-                state->loc.path, state->loc.ino);
+                "%"PRIu64": LINK() %s -> %s",
+                state->finh->unique, state->loc2.path,
+                state->loc.path);
 
         FUSE_FOP (state, fuse_newentry_cbk, GF_FOP_LINK,
-                  link, &state->loc2, &state->loc);
+                  link, &state->loc2, &state->loc, state->xdata);
 }
 
 static void
@@ -1386,54 +1800,41 @@ fuse_link (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
         struct fuse_link_in *fli = msg;
         char         *name = (char *)(fli + 1);
-
         fuse_state_t *state = NULL;
-        int32_t       ret = -1;
 
         GET_STATE (this, finh, state);
 
-        ret = fuse_loc_fill (&state->loc, state, 0, finh->nodeid, name);
-        if (ret == 0)
-                ret = fuse_loc_fill (&state->loc2, state, fli->oldnodeid, 0,
-                                     NULL);
+        fuse_resolve_inode_init (state, &state->resolve2, fli->oldnodeid);
 
-        if ((state->loc2.inode == NULL) ||
-            (ret < 0)) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "fuse_loc_fill() failed for %s %"PRIu64": LINK %s %s",
-                        state->loc2.path, finh->unique,
-                        state->loc2.path, state->loc.path);
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+        fuse_resolve_entry_init (state, &state->resolve, finh->nodeid, name);
 
         fuse_resolve_and_resume (state, fuse_link_resume);
 
         return;
 }
 
-
 static int
 fuse_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno,
                  fd_t *fd, inode_t *inode, struct iatt *buf,
-                 struct iatt *preparent, struct iatt *postparent)
+                 struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-        fuse_state_t            *state = NULL;
-        fuse_in_header_t        *finh = NULL;
-        fuse_private_t          *priv = NULL;
-        struct fuse_out_header   fouh = {0, };
-        struct fuse_entry_out    feo = {0, };
-        struct fuse_open_out     foo = {0, };
-        struct iovec             iov_out[3];
-        inode_t                 *linked_inode = NULL;
+        fuse_state_t           *state        = NULL;
+        fuse_in_header_t       *finh         = NULL;
+        fuse_private_t         *priv         = NULL;
+        struct fuse_out_header  fouh         = {0, };
+        struct fuse_entry_out   feo          = {0, };
+        struct fuse_open_out    foo          = {0, };
+        struct iovec            iov_out[3];
+        inode_t                *linked_inode = NULL;
 
         state    = frame->root->state;
         priv     = this->private;
         finh     = state->finh;
         foo.open_flags = 0;
 
+        fuse_log_eh_fop(this, state, frame, op_ret, op_errno);
+
         if (op_ret >= 0) {
                 foo.fh = (uintptr_t) fd;
 
@@ -1443,12 +1844,12 @@ fuse_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         foo.open_flags |= FOPEN_DIRECT_IO;
 
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                        "%"PRIu64": %s() %s => %p (ino=%"PRId64")",
+                        "%"PRIu64": %s() %s => %p (ino=%"PRIu64")",
                         frame->root->unique, gf_fop_list[frame->root->op],
                         state->loc.path, fd, buf->ia_ino);
 
                 buf->ia_blksize = this->ctx->page_size;
-                gf_fuse_stat2attr (buf, &feo.attr);
+                gf_fuse_stat2attr (buf, &feo.attr, priv->enable_ino32);
 
                 linked_inode = inode_link (inode, state->loc.parent,
                                            state->loc.name, buf);
@@ -1464,11 +1865,8 @@ fuse_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
                 inode_lookup (linked_inode);
 
-                inode_ref (state->loc.parent);
                 inode_unref (linked_inode);
 
-                fd_ref (fd);
-
                 feo.nodeid = inode_to_fuse_nodeid (linked_inode);
 
                 feo.entry_valid = calc_timeout_sec (priv->entry_timeout);
@@ -1489,11 +1887,12 @@ fuse_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 #endif
                 iov_out[2].iov_base = &foo;
                 iov_out[2].iov_len = sizeof (foo);
+
                 if (send_fuse_iov (this, finh, iov_out, 3) == ENOENT) {
                         gf_log ("glusterfs-fuse", GF_LOG_DEBUG,
                                 "create(%s) got EINTR", state->loc.path);
                         inode_forget (inode, 1);
-                        fd_unref (fd);
+                        gf_fd_put (priv->fdtable, state->fd_no);
                         goto out;
                 }
 
@@ -1503,6 +1902,7 @@ fuse_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         "%"PRIu64": %s => -1 (%s)", finh->unique,
                         state->loc.path, strerror (op_errno));
                 send_fuse_err (this, finh, op_errno);
+                gf_fd_put (priv->fdtable, state->fd_no);
         }
 out:
         free_fuse_state (state);
@@ -1514,23 +1914,59 @@ out:
 void
 fuse_create_resume (fuse_state_t *state)
 {
-        fd_t *fd = NULL;
+        fd_t           *fd    = NULL;
+        fuse_private_t *priv  = NULL;
+        fuse_fd_ctx_t  *fdctx = NULL;
 
         if (!state->loc.parent) {
-                gf_log ("fuse", GF_LOG_ERROR, "failed to resolve path %s",
-                        state->loc.path);
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "%"PRIu64" CREATE %s/%s resolution failed",
+                        state->finh->unique, uuid_utoa (state->resolve.gfid),
+                        state->resolve.bname);
                 send_fuse_err (state->this, state->finh, ENOENT);
                 free_fuse_state (state);
                 return;
         }
 
-        if (!state->loc.inode)
-                state->loc.inode = inode_new (state->loc.parent->table);
-        else
-                gf_log (state->this->name, GF_LOG_ERROR, "inode already present");
+        if (state->resolve.op_errno == ENOENT) {
+                state->resolve.op_ret = 0;
+                state->resolve.op_errno = 0;
+        }
+
+        if (state->loc.inode) {
+                gf_log (state->this->name, GF_LOG_DEBUG,
+                        "inode already present");
+                inode_unref (state->loc.inode);
+        }
+
+        state->loc.inode = inode_new (state->loc.parent->table);
 
         fd = fd_create (state->loc.inode, state->finh->pid);
-        state->fd = fd;
+        if (fd == NULL) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "%"PRIu64" CREATE cannot create a new fd",
+                        state->finh->unique);
+                send_fuse_err (state->this, state->finh, ENOMEM);
+                free_fuse_state (state);
+                return;
+        }
+
+        fdctx = fuse_fd_ctx_check_n_create (state->this, fd);
+        if (fdctx == NULL) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "%"PRIu64" CREATE creation of fdctx failed",
+                        state->finh->unique);
+                fd_unref (fd);
+                send_fuse_err (state->this, state->finh, ENOMEM);
+                free_fuse_state (state);
+                return;
+        }
+
+        priv = state->this->private;
+
+        state->fd_no = gf_fd_unused_get (priv->fdtable, fd);
+
+        state->fd = fd_ref (fd);
         fd->flags = state->flags;
 
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
@@ -1539,7 +1975,7 @@ fuse_create_resume (fuse_state_t *state)
 
         FUSE_FOP (state, fuse_create_cbk, GF_FOP_CREATE,
                   create, &state->loc, state->flags, state->mode,
-                  fd, state->dict);
+                  state->umask, fd, state->xdata);
 
 }
 
@@ -1567,19 +2003,15 @@ fuse_create (xlator_t *this, fuse_in_header_t *finh, void *msg)
 
         uuid_generate (state->gfid);
 
-        ret = fuse_loc_fill (&state->loc, state, 0, finh->nodeid, name);
-        if (ret < 0) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64" CREATE %s (fuse_loc_fill() failed)",
-                        finh->unique, state->loc.path);
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+        fuse_resolve_entry_init (state, &state->resolve, finh->nodeid, name);
 
         state->mode = fci->mode;
         state->flags = fci->flags;
 
+        priv = this->private;
+#if FUSE_KERNEL_MINOR_VERSION >=12
+        FUSE_ENTRY_CREATE(this, priv, finh, state, fci, "CREATE");
+#endif
         fuse_resolve_and_resume (state, fuse_create_resume);
 
         return;
@@ -1588,10 +2020,44 @@ fuse_create (xlator_t *this, fuse_in_header_t *finh, void *msg)
 void
 fuse_open_resume (fuse_state_t *state)
 {
-        fd_t *fd = NULL;
+        fd_t           *fd    = NULL;
+        fuse_private_t *priv  = NULL;
+        fuse_fd_ctx_t  *fdctx = NULL;
+
+        if (!state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                        "%"PRIu64": OPEN %s resolution failed",
+                        state->finh->unique, uuid_utoa (state->resolve.gfid));
+
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
 
         fd = fd_create (state->loc.inode, state->finh->pid);
-        state->fd = fd;
+        if (!fd) {
+                gf_log ("fuse", GF_LOG_ERROR,
+                        "fd is NULL");
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
+
+        fdctx = fuse_fd_ctx_check_n_create (state->this, fd);
+        if (fdctx == NULL) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "%"PRIu64": OPEN creation of fdctx failed",
+                        state->finh->unique);
+                fd_unref (fd);
+                send_fuse_err (state->this, state->finh, ENOMEM);
+                free_fuse_state (state);
+                return;
+        }
+
+        priv = state->this->private;
+
+        state->fd_no = gf_fd_unused_get (priv->fdtable, fd);
+        state->fd = fd_ref (fd);
         fd->flags = state->flags;
 
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
@@ -1599,43 +2065,31 @@ fuse_open_resume (fuse_state_t *state)
                 state->loc.path);
 
         FUSE_FOP (state, fuse_fd_cbk, GF_FOP_OPEN,
-                  open, &state->loc, state->flags, fd, 0);
+                  open, &state->loc, state->flags, fd, state->xdata);
 }
 
 static void
 fuse_open (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
         struct fuse_open_in *foi = msg;
-
         fuse_state_t *state = NULL;
-        int32_t       ret = -1;
 
         GET_STATE (this, finh, state);
 
-        ret = fuse_loc_fill (&state->loc, state, finh->nodeid, 0, NULL);
-        if ((state->loc.inode == NULL) ||
-            (ret < 0)) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": OPEN %s (fuse_loc_fill() failed)",
-                        finh->unique, state->loc.path);
-
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+        fuse_resolve_inode_init (state, &state->resolve, finh->nodeid);
 
         state->flags = foi->flags;
+
         fuse_resolve_and_resume (state, fuse_open_resume);
 
         return;
 }
 
-
 static int
 fuse_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 int32_t op_ret, int32_t op_errno,
                 struct iovec *vector, int32_t count,
-                struct iatt *stbuf, struct iobref *iobref)
+                struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
 {
         fuse_state_t *state = NULL;
         fuse_in_header_t *finh = NULL;
@@ -1645,9 +2099,11 @@ fuse_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         state = frame->root->state;
         finh = state->finh;
 
+        fuse_log_eh_fop(this, state, frame, op_ret, op_errno);
+
         if (op_ret >= 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                        "%"PRIu64": READ => %d/%"GF_PRI_SIZET",%"PRId64"/%"PRId64,
+                        "%"PRIu64": READ => %d/%"GF_PRI_SIZET",%"PRId64"/%"PRIu64,
                         frame->root->unique,
                         op_ret, state->size, state->off, stbuf->ia_size);
 
@@ -1682,8 +2138,8 @@ fuse_readv_resume (fuse_state_t *state)
                 "%"PRIu64": READ (%p, size=%zu, offset=%"PRIu64")",
                 state->finh->unique, state->fd, state->size, state->off);
 
-        FUSE_FOP (state, fuse_readv_cbk, GF_FOP_READ,
-                  readv, state->fd, state->size, state->off);
+        FUSE_FOP (state, fuse_readv_cbk, GF_FOP_READ, readv, state->fd,
+                  state->size, state->off, state->io_flags, state->xdata);
 }
 
 static void
@@ -1697,10 +2153,11 @@ fuse_readv (xlator_t *this, fuse_in_header_t *finh, void *msg)
 
         GET_STATE (this, finh, state);
 
-
         fd = FH_TO_FD (fri->fh);
         state->fd = fd;
 
+        fuse_resolve_fd_init (state, &state->resolve, fd);
+
         /* See comment by similar code in fuse_settatr */
         priv = this->private;
 #if FUSE_KERNEL_MINOR_VERSION >= 9
@@ -1710,15 +2167,16 @@ fuse_readv (xlator_t *this, fuse_in_header_t *finh, void *msg)
 
         state->size = fri->size;
         state->off = fri->offset;
+        /* lets ignore 'fri->read_flags', but just consider 'fri->flags' */
+        state->io_flags = fri->flags;
 
         fuse_resolve_and_resume (state, fuse_readv_resume);
 }
 
-
 static int
 fuse_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno,
-                 struct iatt *stbuf, struct iatt *postbuf)
+                 struct iatt *stbuf, struct iatt *postbuf, dict_t *xdata)
 {
         fuse_state_t *state = NULL;
         fuse_in_header_t *finh = NULL;
@@ -1727,9 +2185,11 @@ fuse_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         state = frame->root->state;
         finh = state->finh;
 
+        fuse_log_eh_fop(this, state, frame, op_ret, op_errno);
+
         if (op_ret >= 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                        "%"PRIu64": WRITE => %d/%"GF_PRI_SIZET",%"PRId64"/%"PRId64,
+                        "%"PRIu64": WRITE => %d/%"GF_PRI_SIZET",%"PRId64"/%"PRIu64,
                         frame->root->unique,
                         op_ret, state->size, state->off, stbuf->ia_size);
 
@@ -1755,11 +2215,6 @@ fuse_write_resume (fuse_state_t *state)
         struct iobref *iobref = NULL;
         struct iobuf  *iobuf = NULL;
 
-        if (!state->fd || !state->fd->inode) {
-                send_fuse_err (state->this, state->finh, EBADFD);
-                free_fuse_state (state);
-                return;
-        }
 
         iobref = iobref_new ();
         if (!iobref) {
@@ -1775,8 +2230,13 @@ fuse_write_resume (fuse_state_t *state)
         iobuf = ((fuse_private_t *) (state->this->private))->iobuf;
         iobref_add (iobref, iobuf);
 
+        gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                "%"PRIu64": WRITE (%p, size=%"GF_PRI_SIZET", offset=%"PRId64")",
+                state->finh->unique, state->fd, state->size, state->off);
+
         FUSE_FOP (state, fuse_writev_cbk, GF_FOP_WRITE, writev, state->fd,
-                  &state->vector, 1, state->off, iobref);
+                  &state->vector, 1, state->off, state->io_flags, iobref,
+                  state->xdata);
 
         iobref_unref (iobref);
 }
@@ -1802,6 +2262,15 @@ fuse_write (xlator_t *this, fuse_in_header_t *finh, void *msg)
         state->size = fwi->size;
         state->off  = fwi->offset;
 
+        /* lets ignore 'fwi->write_flags', but just consider 'fwi->flags' */
+        state->io_flags = fwi->flags;
+        /* TODO: may need to handle below flag
+           (fwi->write_flags & FUSE_WRITE_CACHE);
+        */
+
+
+        fuse_resolve_fd_init (state, &state->resolve, fd);
+
         /* See comment by similar code in fuse_settatr */
         priv = this->private;
 #if FUSE_KERNEL_MINOR_VERSION >= 9
@@ -1809,10 +2278,6 @@ fuse_write (xlator_t *this, fuse_in_header_t *finh, void *msg)
                 state->lk_owner = fwi->lock_owner;
 #endif
 
-        gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": WRITE (%p, size=%"PRIu32", offset=%"PRId64")",
-                finh->unique, fd, fwi->size, fwi->offset);
-
         state->vector.iov_base = msg;
         state->vector.iov_len  = fwi->size;
 
@@ -1825,7 +2290,7 @@ void
 fuse_flush_resume (fuse_state_t *state)
 {
         FUSE_FOP (state, fuse_err_cbk, GF_FOP_FLUSH,
-                  flush, state->fd);
+                  flush, state->fd, state->xdata);
 }
 
 static void
@@ -1839,8 +2304,8 @@ fuse_flush (xlator_t *this, fuse_in_header_t *finh, void *msg)
         GET_STATE (this, finh, state);
         fd = FH_TO_FD (ffi->fh);
         state->fd = fd;
-        if (fd)
-                fd->flush_unique = finh->unique;
+
+        fuse_resolve_fd_init (state, &state->resolve, fd);
 
         state->lk_owner = ffi->lock_owner;
 
@@ -1852,104 +2317,64 @@ fuse_flush (xlator_t *this, fuse_in_header_t *finh, void *msg)
         return;
 }
 
-static int
-fuse_release_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno)
-{
-        STACK_DESTROY (frame->root);
-
-        return 0;
-}
-
 static void
 fuse_release (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
-        struct fuse_release_in *fri        = msg;
-        fd_t                   *new_fd     = NULL;
-        fd_t                   *fd         = NULL;
-        int                     do_flush   = 0;
-        uint64_t                tmp_fd_ctx = 0;
-        int                     ret        = 0;
-        fuse_state_t           *state      = NULL;
+        struct fuse_release_in *fri       = msg;
+        fd_t                   *activefd = NULL;
+        fd_t                   *fd        = NULL;
+        uint64_t                val       = 0;
+        int                     ret       = 0;
+        fuse_state_t           *state     = NULL;
+        fuse_fd_ctx_t          *fdctx     = NULL;
+        fuse_private_t         *priv      = NULL;
 
         GET_STATE (this, finh, state);
         fd = FH_TO_FD (fri->fh);
         state->fd = fd;
 
-#ifdef  GF_LINUX_HOST_OS
-        /* This is an ugly Linux specific hack, relying on subtle
-         * implementation details.
-         *
-         * The self-heal algorithm of replicate relies on being
-         * notified by means of a flush fop whenever a consumer
-         * of a file is done with that file. If this happens
-         * from userspace by means of close(2) or process termination,
-         * the kernel sends us a FLUSH message which we can handle with
-         * the flush fop (nb. this mechanism itself is Linux specific!!).
-         *
-         * However, if it happens from a kernel context, we get no FLUSH,
-         * just the final RELEASE when all references to the file are gone.
-         * We try to guess that this is the case by checking if the last FLUSH
-         * on the file was just the previous message. If not, we conjecture
-         * that this release is from a kernel context and call the flush fop
-         * here.
-         *
-         * Note #1: we check the above condition by means of looking at
-         * the "unique" values of the FUSE messages, relying on which is
-         * a big fat NO NO NO in any sane code.
-         *
-         * Note #2: there is no guarantee against false positives (in theory
-         * it's possible that the scheduler arranges an unrelated FUSE message
-         * in between FLUSH and RELEASE, although it seems to be unlikely), but
-         * extra flushes are not a problem.
-         *
-         * Note #3: cf. Bug #223.
-         */
+        priv = this->private;
 
-        if (fd && fd->flush_unique + 1 != finh->unique)
-                do_flush = 1;
-#endif
+        fuse_log_eh (this, "RELEASE(): %"PRIu64":, fd: %p, gfid: %s",
+                     finh->unique, fd, uuid_utoa (fd->inode->gfid));
 
-        /* TODO */
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": RELEASE %p%s", finh->unique, fd,
-                do_flush ? " (FLUSH implied)" : "");
-
-        if (do_flush) {
-                state->lk_owner = (uint64_t)-1;
-                ret = fd_ctx_get (fd, this, &tmp_fd_ctx);
-                if (!ret) {
-                        new_fd = (fd_t *)(long)tmp_fd_ctx;
-                        FUSE_FOP (state, fuse_release_cbk, GF_FOP_FLUSH, flush,
-                                  new_fd);
-                        fd_unref (new_fd);
-                }
+                "%"PRIu64": RELEASE %p", finh->unique, state->fd);
 
-                FUSE_FOP (state, fuse_err_cbk, GF_FOP_FLUSH, flush, fd);
+        ret = fd_ctx_del (fd, this, &val);
+        if (!ret) {
+                fdctx = (fuse_fd_ctx_t *)(unsigned long)val;
+                if (fdctx) {
+                        activefd = fdctx->activefd;
+                        if (activefd) {
+                                fd_unref (activefd);
+                        }
 
-                fd_unref (fd);
-        } else {
-                ret = fd_ctx_get (fd, this, &tmp_fd_ctx);
-                if (!ret) {
-                        new_fd = (fd_t *)(long)tmp_fd_ctx;
-                        fd_unref (new_fd);
+                        GF_FREE (fdctx);
                 }
-                fd_unref (fd);
+        }
+        fd_unref (fd);
 
-                send_fuse_err (this, finh, 0);
+        state->fd = NULL;
 
-                free_fuse_state (state);
-        }
+        gf_fdptr_put (priv->fdtable, fd);
+
+        send_fuse_err (this, finh, 0);
 
+        free_fuse_state (state);
         return;
 }
 
 void
 fuse_fsync_resume (fuse_state_t *state)
 {
+        gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                "%"PRIu64": FSYNC %p", state->finh->unique,
+                state->fd);
+
         /* fsync_flags: 1 means "datasync" (no defines for this) */
         FUSE_FOP (state, fuse_fsync_cbk, GF_FOP_FSYNC,
-                  fsync, state->fd, state->flags & 1);
+                  fsync, state->fd, (state->flags & 1), state->xdata);
 }
 
 static void
@@ -1964,8 +2389,7 @@ fuse_fsync (xlator_t *this, fuse_in_header_t *finh, void *msg)
         fd = FH_TO_FD (fsi->fh);
         state->fd = fd;
 
-        gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": FSYNC %p", finh->unique, fd);
+        fuse_resolve_fd_init (state, &state->resolve, fd);
 
         state->flags = fsi->fsync_flags;
         fuse_resolve_and_resume (state, fuse_fsync_resume);
@@ -1975,17 +2399,50 @@ fuse_fsync (xlator_t *this, fuse_in_header_t *finh, void *msg)
 void
 fuse_opendir_resume (fuse_state_t *state)
 {
-        fd_t *fd = NULL;
+        fd_t           *fd    = NULL;
+        fuse_private_t *priv  = NULL;
+        fuse_fd_ctx_t  *fdctx = NULL;
+
+        priv = state->this->private;
+
+        if (!state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "%"PRIu64": OPENDIR (%s) resolution failed",
+                        state->finh->unique, uuid_utoa (state->resolve.gfid));
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
 
         fd = fd_create (state->loc.inode, state->finh->pid);
-        state->fd = fd;
+        if (fd == NULL) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "%"PRIu64": OPENDIR fd creation failed",
+                        state->finh->unique);
+                send_fuse_err (state->this, state->finh, ENOMEM);
+                free_fuse_state (state);
+        }
+
+        fdctx = fuse_fd_ctx_check_n_create (state->this, fd);
+        if (fdctx == NULL) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "%"PRIu64": OPENDIR creation of fdctx failed",
+                        state->finh->unique);
+                fd_unref (fd);
+                send_fuse_err (state->this, state->finh, ENOMEM);
+                free_fuse_state (state);
+                return;
+        }
+
+        state->fd = fd_ref (fd);
+        state->fd_no = gf_fd_unused_get (priv->fdtable, fd);
 
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                 "%"PRIu64": OPENDIR %s", state->finh->unique,
                 state->loc.path);
 
         FUSE_FOP (state, fuse_fd_cbk, GF_FOP_OPENDIR,
-                  opendir, &state->loc, fd);
+                  opendir, &state->loc, fd, state->xdata);
 }
 
 static void
@@ -1996,24 +2453,14 @@ fuse_opendir (xlator_t *this, fuse_in_header_t *finh, void *msg)
          */
 
         fuse_state_t *state = NULL;
-        int32_t       ret = -1;
 
         GET_STATE (this, finh, state);
-        ret = fuse_loc_fill (&state->loc, state, finh->nodeid, 0, NULL);
-        if ((state->loc.inode == NULL) || (ret < 0)) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": OPENDIR %s (fuse_loc_fill() failed)",
-                        finh->unique, state->loc.path);
 
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+        fuse_resolve_inode_init (state, &state->resolve, finh->nodeid);
 
         fuse_resolve_and_resume (state, fuse_opendir_resume);
 }
 
-
 unsigned char
 d_type_from_stat (struct iatt *buf)
 {
@@ -2047,10 +2494,10 @@ d_type_from_stat (struct iatt *buf)
         return d_type;
 }
 
-
 static int
 fuse_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno, gf_dirent_t *entries)
+                  int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                  dict_t *xdata)
 {
         fuse_state_t *state = NULL;
         fuse_in_header_t *finh = NULL;
@@ -2058,9 +2505,13 @@ fuse_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         char         *buf = NULL;
         gf_dirent_t  *entry = NULL;
         struct fuse_dirent *fde = NULL;
+        fuse_private_t *priv = NULL;
 
         state = frame->root->state;
         finh  = state->finh;
+        priv = state->this->private;
+
+        fuse_log_eh_fop(this, state, frame, op_ret, op_errno);
 
         if (op_ret < 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_WARNING,
@@ -2080,6 +2531,11 @@ fuse_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                            strlen (entry->d_name));
         }
 
+	if (size <= 0) {
+		send_fuse_data (this, finh, 0, 0);
+		goto out;
+	}
+
         buf = GF_CALLOC (1, size, gf_fuse_mt_char);
         if (!buf) {
                 gf_log ("glusterfs-fuse", GF_LOG_DEBUG,
@@ -2092,20 +2548,19 @@ fuse_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         size = 0;
         list_for_each_entry (entry, &entries->list, list) {
                 fde = (struct fuse_dirent *)(buf + size);
-                fde->ino = entry->d_ino;
-                fde->off = entry->d_off;
-                fde->namelen = strlen (entry->d_name);
-                strncpy (fde->name, entry->d_name, fde->namelen);
+                gf_fuse_fill_dirent (entry, fde, priv->enable_ino32);
                 size += FUSE_DIRENT_SIZE (fde);
         }
 
         send_fuse_data (this, finh, buf, size);
 
+        /* TODO: */
+        /* gf_link_inodes_from_dirent (this, state->fd->inode, entries); */
+
 out:
         free_fuse_state (state);
         STACK_DESTROY (frame->root);
-        if (buf)
-                GF_FREE (buf);
+        GF_FREE (buf);
         return 0;
 
 }
@@ -2114,11 +2569,11 @@ void
 fuse_readdir_resume (fuse_state_t *state)
 {
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": READDIR (%p, size=%zu, offset=%"PRId64")",
+                "%"PRIu64": READDIR (%p, size=%"GF_PRI_SIZET", offset=%"PRId64")",
                 state->finh->unique, state->fd, state->size, state->off);
 
         FUSE_FOP (state, fuse_readdir_cbk, GF_FOP_READDIR,
-                  readdir, state->fd, state->size, state->off);
+                  readdir, state->fd, state->size, state->off, state->xdata);
 }
 
 static void
@@ -2135,34 +2590,230 @@ fuse_readdir (xlator_t *this, fuse_in_header_t *finh, void *msg)
         fd = FH_TO_FD (fri->fh);
         state->fd = fd;
 
+        fuse_resolve_fd_init (state, &state->resolve, fd);
+
         fuse_resolve_and_resume (state, fuse_readdir_resume);
 }
 
 
+static int
+fuse_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		   int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+		   dict_t *xdata)
+{
+	fuse_state_t *state = NULL;
+	fuse_in_header_t *finh = NULL;
+	int           size = 0;
+	char         *buf = NULL;
+	gf_dirent_t  *entry = NULL;
+	struct fuse_direntplus *fde = NULL;
+	struct fuse_entry_out *feo = NULL;
+	fuse_private_t         *priv   = NULL;
+
+	state = frame->root->state;
+	finh  = state->finh;
+	priv = this->private;
+
+	if (op_ret < 0) {
+		gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+			"%"PRIu64": READDIRP => -1 (%s)", frame->root->unique,
+			strerror (op_errno));
+
+		send_fuse_err (this, finh, op_errno);
+		goto out;
+	}
+
+	gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+		"%"PRIu64": READDIRP => %d/%"GF_PRI_SIZET",%"PRId64,
+		frame->root->unique, op_ret, state->size, state->off);
+
+	list_for_each_entry (entry, &entries->list, list) {
+		size += FUSE_DIRENT_ALIGN (FUSE_NAME_OFFSET_DIRENTPLUS +
+					   strlen (entry->d_name));
+	}
+
+	if (size <= 0) {
+		send_fuse_data (this, finh, 0, 0);
+		goto out;
+	}
+
+	buf = GF_CALLOC (1, size, gf_fuse_mt_char);
+	if (!buf) {
+		gf_log ("glusterfs-fuse", GF_LOG_DEBUG,
+			"%"PRIu64": READDIRP => -1 (%s)", frame->root->unique,
+			strerror (ENOMEM));
+		send_fuse_err (this, finh, ENOMEM);
+		goto out;
+	}
+
+	size = 0;
+	list_for_each_entry (entry, &entries->list, list) {
+		inode_t *linked_inode;
+
+		fde = (struct fuse_direntplus *)(buf + size);
+		feo = &fde->entry_out;
+		fde->dirent.ino = entry->d_ino;
+		fde->dirent.off = entry->d_off;
+		fde->dirent.type = entry->d_type;
+		fde->dirent.namelen = strlen (entry->d_name);
+		strncpy (fde->dirent.name, entry->d_name, fde->dirent.namelen);
+		size += FUSE_DIRENTPLUS_SIZE (fde);
+
+		if (!entry->inode)
+			continue;
+
+		entry->d_stat.ia_blksize = this->ctx->page_size;
+		gf_fuse_stat2attr (&entry->d_stat, &feo->attr, priv->enable_ino32);
+
+		linked_inode = inode_link (entry->inode, state->fd->inode,
+					   entry->d_name, &entry->d_stat);
+		if (!linked_inode)
+			continue;
+
+		inode_lookup (linked_inode);
+
+		feo->nodeid = inode_to_fuse_nodeid (linked_inode);
+
+		fuse_inode_set_need_lookup (linked_inode, this);
+
+		inode_unref (linked_inode);
+
+		feo->entry_valid =
+			calc_timeout_sec (priv->entry_timeout);
+		feo->entry_valid_nsec =
+			calc_timeout_nsec (priv->entry_timeout);
+		feo->attr_valid =
+			calc_timeout_sec (priv->attribute_timeout);
+		feo->attr_valid_nsec =
+			calc_timeout_nsec (priv->attribute_timeout);
+	}
+
+	send_fuse_data (this, finh, buf, size);
+out:
+	free_fuse_state (state);
+	STACK_DESTROY (frame->root);
+	GF_FREE (buf);
+	return 0;
+
+}
+
+
+void
+fuse_readdirp_resume (fuse_state_t *state)
+{
+	gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+		"%"PRIu64": READDIRP (%p, size=%"GF_PRI_SIZET", offset=%"PRId64")",
+		state->finh->unique, state->fd, state->size, state->off);
+
+	FUSE_FOP (state, fuse_readdirp_cbk, GF_FOP_READDIRP,
+		  readdirp, state->fd, state->size, state->off, state->xdata);
+}
+
+
+static void
+fuse_readdirp (xlator_t *this, fuse_in_header_t *finh, void *msg)
+{
+	struct fuse_read_in *fri = msg;
+
+	fuse_state_t *state = NULL;
+	fd_t         *fd = NULL;
+
+	GET_STATE (this, finh, state);
+	state->size = fri->size;
+	state->off = fri->offset;
+	fd = FH_TO_FD (fri->fh);
+	state->fd = fd;
+
+	fuse_resolve_fd_init (state, &state->resolve, fd);
+
+	fuse_resolve_and_resume (state, fuse_readdirp_resume);
+}
+
+static int
+fuse_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+		   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+		   struct iatt *postbuf, dict_t *xdata)
+{
+	return fuse_err_cbk(frame, cookie, this, op_ret, op_errno, xdata);
+}
+
+static void
+fuse_fallocate_resume(fuse_state_t *state)
+{
+	gf_log("glusterfs-fuse", GF_LOG_TRACE,
+	       "%"PRIu64": FALLOCATE (%p, flags=%d, size=%zu, offset=%"PRId64")",
+	       state->finh->unique, state->fd, state->flags, state->size,
+	       state->off);
+
+	if (state->flags & FALLOC_FL_PUNCH_HOLE)
+		FUSE_FOP(state, fuse_fallocate_cbk, GF_FOP_DISCARD, discard,
+			 state->fd, state->off, state->size, state->xdata);
+	else
+		FUSE_FOP(state, fuse_fallocate_cbk, GF_FOP_FALLOCATE, fallocate,
+			 state->fd, (state->flags & FALLOC_FL_KEEP_SIZE),
+			 state->off, state->size, state->xdata);
+}
+
+static void
+fuse_fallocate(xlator_t *this, fuse_in_header_t *finh, void *msg)
+{
+	struct fuse_fallocate_in *ffi = msg;
+	fuse_state_t *state = NULL;
+
+	GET_STATE(this, finh, state);
+	state->off = ffi->offset;
+	state->size = ffi->length;
+	state->flags = ffi->mode;
+	state->fd = FH_TO_FD(ffi->fh);
+
+	fuse_resolve_fd_init(state, &state->resolve, state->fd);
+	fuse_resolve_and_resume(state, fuse_fallocate_resume);
+}
+
+
 static void
 fuse_releasedir (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
-        struct fuse_release_in *fri        = msg;
-        fd_t                   *new_fd     = NULL;
-        uint64_t                tmp_fd_ctx = 0;
-        int                     ret        = 0;
-        fuse_state_t           *state      = NULL;
+        struct fuse_release_in *fri       = msg;
+        fd_t                   *activefd = NULL;
+        uint64_t                val       = 0;
+        int                     ret       = 0;
+        fuse_state_t           *state     = NULL;
+        fuse_fd_ctx_t          *fdctx     = NULL;
+        fuse_private_t         *priv      = NULL;
 
         GET_STATE (this, finh, state);
         state->fd = FH_TO_FD (fri->fh);
 
-        /* TODO: free 'new-fd' in the ctx */
+        priv = this->private;
+
+        fuse_log_eh (this, "RELEASEDIR (): %"PRIu64": fd: %p, gfid: %s",
+                     finh->unique, state->fd,
+                     uuid_utoa (state->fd->inode->gfid));
+
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                 "%"PRIu64": RELEASEDIR %p", finh->unique, state->fd);
 
-        ret = fd_ctx_get (state->fd, this, &tmp_fd_ctx);
+        ret = fd_ctx_del (state->fd, this, &val);
+
         if (!ret) {
-                new_fd = (fd_t *)(long)tmp_fd_ctx;
-                fd_unref (new_fd);
+                fdctx = (fuse_fd_ctx_t *)(unsigned long)val;
+                if (fdctx) {
+                        activefd = fdctx->activefd;
+                        if (activefd) {
+                                fd_unref (activefd);
+                        }
+
+                        GF_FREE (fdctx);
+                }
         }
 
         fd_unref (state->fd);
 
+        gf_fdptr_put (priv->fdtable, state->fd);
+
+        state->fd = NULL;
+
         send_fuse_err (this, finh, 0);
 
         free_fuse_state (state);
@@ -2174,7 +2825,7 @@ void
 fuse_fsyncdir_resume (fuse_state_t *state)
 {
         FUSE_FOP (state, fuse_err_cbk, GF_FOP_FSYNCDIR,
-                  fsyncdir, state->fd, state->flags & 1);
+                  fsyncdir, state->fd, (state->flags & 1), state->xdata);
 
 }
 
@@ -2191,16 +2842,18 @@ fuse_fsyncdir (xlator_t *this, fuse_in_header_t *finh, void *msg)
         GET_STATE (this, finh, state);
         state->fd = fd;
 
+        fuse_resolve_fd_init (state, &state->resolve, fd);
+
         state->flags = fsi->fsync_flags;
         fuse_resolve_and_resume (state, fuse_fsyncdir_resume);
 
         return;
 }
 
-
 static int
 fuse_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int32_t op_ret, int32_t op_errno, struct statvfs *buf)
+                 int32_t op_ret, int32_t op_errno, struct statvfs *buf,
+                 dict_t *xdata)
 {
         fuse_state_t *state = NULL;
         fuse_in_header_t *finh = NULL;
@@ -2210,19 +2863,10 @@ fuse_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         state = frame->root->state;
         priv  = this->private;
         finh  = state->finh;
-        /*
-          Filesystems (like ZFS on solaris) reports
-          different ->f_frsize and ->f_bsize. Old coreutils
-          df tools use statfs() and do not see ->f_frsize.
-          the ->f_blocks, ->f_bavail and ->f_bfree are
-          w.r.t ->f_frsize and not ->f_bsize which makes the
-          df tools report wrong values.
-
-          Scale the block counts to match ->f_bsize.
-        */
-        /* TODO: with old coreutils, f_bsize is taken from stat()'s ia_blksize
-         * so the df with old coreutils this wont work :(
-         */
+
+        fuse_log_eh (this, "op_ret: %d, op_errno: %d, %"PRIu64": %s()",
+                     op_ret, op_errno, frame->root->unique,
+                     gf_fop_list[frame->root->op]);
 
         if (op_ret == 0) {
 #ifndef GF_DARWIN_HOST_OS
@@ -2263,53 +2907,92 @@ fuse_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         return 0;
 }
 
-static void
-fuse_statfs (xlator_t *this, fuse_in_header_t *finh, void *msg)
+void
+fuse_statfs_resume (fuse_state_t *state)
 {
-        fuse_state_t *state = NULL;
-        int32_t       ret = -1;
-
-        GET_STATE (this, finh, state);
-        ret = fuse_loc_fill (&state->loc, state, 1, 0, NULL);
-        if ((state->loc.inode == NULL) ||
-            (ret < 0)) {
+        if (!state->loc.inode) {
                 gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": STATFS (fuse_loc_fill() fail)",
-                        finh->unique);
+                        "%"PRIu64": STATFS (%s) resolution fail",
+                        state->finh->unique, uuid_utoa (state->resolve.gfid));
 
-                send_fuse_err (this, finh, ENOENT);
+                send_fuse_err (state->this, state->finh, ENOENT);
                 free_fuse_state (state);
                 return;
         }
 
         gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": STATFS", finh->unique);
+                "%"PRIu64": STATFS", state->finh->unique);
 
         FUSE_FOP (state, fuse_statfs_cbk, GF_FOP_STATFS,
-                  statfs, &state->loc);
+                  statfs, &state->loc, state->xdata);
 }
 
+
+static void
+fuse_statfs (xlator_t *this, fuse_in_header_t *finh, void *msg)
+{
+        fuse_state_t *state = NULL;
+
+        GET_STATE (this, finh, state);
+
+        fuse_resolve_inode_init (state, &state->resolve, finh->nodeid);
+
+        fuse_resolve_and_resume (state, fuse_statfs_resume);
+}
+
+
 void
 fuse_setxattr_resume (fuse_state_t *state)
 {
-        gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": SETXATTR %s/%"PRIu64" (%s)", state->finh->unique,
-                state->loc.path, state->finh->nodeid, state->name);
+        if (!state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "%"PRIu64": SETXATTR %s/%"PRIu64" (%s) "
+                        "resolution failed",
+                        state->finh->unique, uuid_utoa (state->resolve.gfid),
+                        state->finh->nodeid, state->name);
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
+
+#ifdef GF_TEST_FFOP
+        state->fd = fd_lookup (state->loc.inode, state->finh->pid);
+#endif /* GF_TEST_FFOP */
+
+        if (state->fd) {
+                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                        "%"PRIu64": SETXATTR %p/%"PRIu64" (%s)", state->finh->unique,
+                        state->fd, state->finh->nodeid, state->name);
+
+                FUSE_FOP (state, fuse_setxattr_cbk, GF_FOP_FSETXATTR,
+                          fsetxattr, state->fd, state->xattr, state->flags,
+                          state->xdata);
+        } else {
+                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                        "%"PRIu64": SETXATTR %s/%"PRIu64" (%s)", state->finh->unique,
+                        state->loc.path, state->finh->nodeid, state->name);
 
-        FUSE_FOP (state, fuse_setxattr_cbk, GF_FOP_SETXATTR,
-                  setxattr, &state->loc, state->dict, state->flags);
+                FUSE_FOP (state, fuse_setxattr_cbk, GF_FOP_SETXATTR,
+                          setxattr, &state->loc, state->xattr, state->flags,
+                          state->xdata);
+        }
 }
 
+
 static void
 fuse_setxattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
         struct fuse_setxattr_in *fsi = msg;
         char         *name = (char *)(fsi + 1);
         char         *value = name + strlen (name) + 1;
+        struct fuse_private *priv = NULL;
 
         fuse_state_t *state = NULL;
         char         *dict_value = NULL;
         int32_t       ret = -1;
+        char *newkey = NULL;
+
+        priv = this->private;
 
 #ifdef GF_DARWIN_HOST_OS
         if (fsi->position) {
@@ -2323,13 +3006,27 @@ fuse_setxattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
         }
 #endif
 
-#ifdef DISABLE_POSIX_ACL
-        if (!strncmp (name, "system.", 7)) {
-                send_fuse_err (this, finh, EOPNOTSUPP);
-                GF_FREE (finh);
+        if (fuse_ignore_xattr_set (priv, name)) {
+                (void) send_fuse_err (this, finh, 0);
                 return;
         }
-#endif
+
+        if (!priv->acl) {
+                if ((strcmp (name, POSIX_ACL_ACCESS_XATTR) == 0) ||
+                    (strcmp (name, POSIX_ACL_DEFAULT_XATTR) == 0)) {
+                        send_fuse_err (this, finh, EOPNOTSUPP);
+                        GF_FREE (finh);
+                        return;
+                }
+        }
+
+        if (!priv->selinux) {
+                if (strncmp (name, "security.", 9) == 0) {
+                        send_fuse_err (this, finh, EOPNOTSUPP);
+                        GF_FREE (finh);
+                        return;
+                }
+        }
 
         /* Check if the command is for changing the log
            level of process or specific xlator */
@@ -2340,23 +3037,28 @@ fuse_setxattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
                 return;
         }
 
-        GET_STATE (this, finh, state);
-        state->size = fsi->size;
-        ret = fuse_loc_fill (&state->loc, state, finh->nodeid, 0, NULL);
-        if ((state->loc.inode == NULL) ||
-            (ret < 0)) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": SETXATTR %s/%"PRIu64" (%s) (fuse_loc_fill() failed)",
-                        finh->unique,
-                        state->loc.path, finh->nodeid, name);
+        if (!strcmp ("inode-invalidate", name)) {
+                gf_log ("fuse", GF_LOG_TRACE,
+                        "got request to invalidate %"PRIu64, finh->nodeid);
+                send_fuse_err (this, finh, 0);
+                fuse_invalidate_entry (this, finh->nodeid);
+                GF_FREE (finh);
+                return;
+        }
 
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
+        if (!strcmp (GFID_XATTR_KEY, name)) {
+                send_fuse_err (this, finh, EPERM);
+                GF_FREE (finh);
                 return;
         }
 
-        state->dict = get_new_dict ();
-        if (!state->dict) {
+        GET_STATE (this, finh, state);
+        state->size = fsi->size;
+
+        fuse_resolve_inode_init (state, &state->resolve, finh->nodeid);
+
+        state->xattr = get_new_dict ();
+        if (!state->xattr) {
                 gf_log ("glusterfs-fuse", GF_LOG_ERROR,
                         "%"PRIu64": SETXATTR dict allocation failed",
                         finh->unique);
@@ -2366,19 +3068,32 @@ fuse_setxattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
                 return;
         }
 
-        dict_value = memdup (value, fsi->size);
-        dict_set (state->dict, (char *)name,
+        ret = fuse_flip_xattr_ns (priv, name, &newkey);
+        if (ret) {
+                send_fuse_err (this, finh, ENOMEM);
+                free_fuse_state (state);
+                return;
+        }
+
+        if (fsi->size > 0) {
+                dict_value = memdup (value, fsi->size);
+        } else {
+                gf_log (THIS->name, GF_LOG_ERROR, "value size zero");
+                dict_value = NULL;
+        }
+        dict_set (state->xattr, newkey,
                   data_from_dynptr ((void *)dict_value, fsi->size));
-        dict_ref (state->dict);
+        dict_ref (state->xattr);
 
         state->flags = fsi->flags;
-        state->name = gf_strdup (name);
+        state->name = newkey;
 
         fuse_resolve_and_resume (state, fuse_setxattr_resume);
 
         return;
 }
 
+
 static void
 send_fuse_xattr (xlator_t *this, fuse_in_header_t *finh, const char *value,
                  size_t size, size_t expected)
@@ -2402,9 +3117,28 @@ send_fuse_xattr (xlator_t *this, fuse_in_header_t *finh, const char *value,
         }
 }
 
+/* filter out xattrs that need not be visible on the
+ * mount point. this is _specifically_ for geo-rep
+ * as of now, to prevent Rsync from crying out loud
+ * when it tries to setxattr() for selinux xattrs
+ */
+static int
+fuse_filter_xattr(char *key)
+{
+        int need_filter = 0;
+        struct fuse_private *priv = THIS->private;
+
+        if ((priv->client_pid == GF_CLIENT_PID_GSYNCD)
+            && fnmatch ("*.selinux*", key, FNM_PERIOD) == 0)
+                need_filter = 1;
+
+        return need_filter;
+}
+
+
 static int
 fuse_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                int32_t op_ret, int32_t op_errno, dict_t *dict)
+                int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
 {
         int             need_to_free_dict = 0;
         char           *value = "";
@@ -2413,11 +3147,13 @@ fuse_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         data_t         *value_data = NULL;
         int             ret = -1;
         int32_t         len = 0;
-        data_pair_t    *trav = NULL;
+        int32_t         len_next = 0;
 
         state = frame->root->state;
         finh  = state->finh;
 
+        fuse_log_eh_fop(this, state, frame, op_ret, op_errno);
+
         if (op_ret >= 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                         "%"PRIu64": %s() %s => %d", frame->root->unique,
@@ -2439,22 +3175,24 @@ fuse_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         } /* if(value_data)...else */
                 } else {
                         /* if callback for listxattr */
-                        trav = dict->members_list;
-                        while (trav) {
-                                len += strlen (trav->key) + 1;
-                                trav = trav->next;
-                        } /* while(trav) */
+                        /* we need to invoke fuse_filter_xattr() twice. Once
+                         * while counting size and then while filling buffer
+                         */
+                        len = dict_keys_join (NULL, 0, dict, fuse_filter_xattr);
+                        if (len < 0)
+                                goto out;
+
                         value = alloca (len + 1);
                         if (!value)
                                 goto out;
-                        len = 0;
-                        trav = dict->members_list;
-                        while (trav) {
-                                strcpy (value + len, trav->key);
-                                value[len + strlen (trav->key)] = '\0';
-                                len += strlen (trav->key) + 1;
-                                trav = trav->next;
-                        } /* while(trav) */
+
+                        len_next = dict_keys_join (value, len, dict,
+                                                   fuse_filter_xattr);
+                        if (len_next != len)
+                                gf_log (THIS->name, GF_LOG_ERROR,
+                                        "sizes not equal %d != %d",
+                                        len, len_next);
+
                         send_fuse_xattr (this, finh, value, len, state->size);
                 } /* if(state->name)...else */
         } else {
@@ -2469,9 +3207,9 @@ fuse_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                                      "storage");
                         } else {
                                 gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                                        "%"PRIu64": %s() %s => -1 (%s)",
+                                        "%"PRIu64": %s(%s) %s => -1 (%s)",
                                         frame->root->unique,
-                                        gf_fop_list[frame->root->op],
+                                        gf_fop_list[frame->root->op], state->name,
                                         state->loc.path, strerror (op_errno));
                         }
                 } else {
@@ -2495,25 +3233,100 @@ out:
         return 0;
 }
 
+
 void
 fuse_getxattr_resume (fuse_state_t *state)
 {
-        gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": GETXATTR %s/%"PRIu64" (%s)", state->finh->unique,
-                state->loc.path, state->finh->nodeid, state->name);
+        char *value  = NULL;
+
+        if (!state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "%"PRIu64": GETXATTR %s/%"PRIu64" (%s) "
+                        "resolution failed",
+                        state->finh->unique,
+                        uuid_utoa (state->resolve.gfid),
+                        state->finh->nodeid, state->name);
+
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
+
+#ifdef GF_TEST_FFOP
+        state->fd = fd_lookup (state->loc.inode, state->finh->pid);
+#endif /* GF_TEST_FFOP */
+
+        if (state->name &&
+            (strcmp (state->name, VIRTUAL_GFID_XATTR_KEY) == 0)) {
+                /* send glusterfs gfid in binary form */
+
+                value = GF_CALLOC (16 + 1, sizeof(char),
+                                   gf_common_mt_char);
+                if (!value) {
+                        send_fuse_err (state->this, state->finh, ENOMEM);
+                        goto internal_out;
+                }
+                memcpy (value, state->loc.inode->gfid, 16);
+
+                send_fuse_xattr (THIS, state->finh, value, 16, state->size);
+                GF_FREE (value);
+        internal_out:
+                free_fuse_state (state);
+                return;
+        }
+
+        if (state->name &&
+            (strcmp (state->name, VIRTUAL_GFID_XATTR_KEY_STR) == 0)) {
+                /* transform binary gfid to canonical form */
+
+                value = GF_CALLOC (UUID_CANONICAL_FORM_LEN + 1, sizeof(char),
+                                   gf_common_mt_char);
+                if (!value) {
+                        send_fuse_err (state->this, state->finh, ENOMEM);
+                        goto internal_out1;
+                }
+                uuid_utoa_r (state->loc.inode->gfid, value);
+
+                send_fuse_xattr (THIS, state->finh, value,
+                                 UUID_CANONICAL_FORM_LEN, state->size);
+                GF_FREE (value);
+        internal_out1:
+                free_fuse_state (state);
+                return;
+        }
+
+
+        if (state->fd) {
+                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                        "%"PRIu64": GETXATTR %p/%"PRIu64" (%s)", state->finh->unique,
+                        state->fd, state->finh->nodeid, state->name);
+
+                FUSE_FOP (state, fuse_xattr_cbk, GF_FOP_FGETXATTR,
+                          fgetxattr, state->fd, state->name, state->xdata);
+        } else {
+                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                        "%"PRIu64": GETXATTR %s/%"PRIu64" (%s)", state->finh->unique,
+                        state->loc.path, state->finh->nodeid, state->name);
 
-        FUSE_FOP (state, fuse_xattr_cbk, GF_FOP_GETXATTR,
-                  getxattr, &state->loc, state->name);
+                FUSE_FOP (state, fuse_xattr_cbk, GF_FOP_GETXATTR,
+                          getxattr, &state->loc, state->name, state->xdata);
+        }
 }
 
+
 static void
 fuse_getxattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
-        struct fuse_getxattr_in *fgxi = msg;
-        char         *name = (char *)(fgxi + 1);
+        struct fuse_getxattr_in *fgxi     = msg;
+        char                    *name     = (char *)(fgxi + 1);
+        fuse_state_t            *state    = NULL;
+        struct fuse_private     *priv     = NULL;
+        int                      rv       = 0;
+        int                     op_errno  = EINVAL;
+        char                    *newkey   = NULL;
 
-        fuse_state_t *state = NULL;
-        int32_t       ret = -1;
+        priv = this->private;
+        GET_STATE (this, finh, state);
 
 #ifdef GF_DARWIN_HOST_OS
         if (fgxi->position) {
@@ -2529,114 +3342,171 @@ fuse_getxattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
                         "%"PRIu64": GETXATTR %s/%"PRIu64" (%s):"
                         "refusing positioned getxattr",
                         finh->unique, state->loc.path, finh->nodeid, name);
-                send_fuse_err (this, finh, EINVAL);
-                FREE (finh);
-                return;
+                op_errno = EINVAL;
+                goto err;
         }
 #endif
 
-#ifdef DISABLE_POSIX_ACL
-        if (!strncmp (name, "system.", 7)) {
-                send_fuse_err (this, finh, ENODATA);
-                GF_FREE (finh);
-                return;
+        if (!priv->acl) {
+                if ((strcmp (name, POSIX_ACL_ACCESS_XATTR) == 0) ||
+                    (strcmp (name, POSIX_ACL_DEFAULT_XATTR) == 0)) {
+                        op_errno = ENOTSUP;
+                        goto err;
+                }
+        }
+
+        if (!priv->selinux) {
+                if (strncmp (name, "security.", 9) == 0) {
+                        op_errno = ENODATA;
+                        goto err;
+                }
         }
-#endif
 
         GET_STATE (this, finh, state);
 
-        ret = fuse_loc_fill (&state->loc, state, finh->nodeid, 0, NULL);
-        if ((state->loc.inode == NULL) ||
-            (ret < 0)) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": GETXATTR %s/%"PRIu64" (%s) (fuse_loc_fill() failed)",
-                        finh->unique, state->loc.path, finh->nodeid, name);
+        fuse_resolve_inode_init (state, &state->resolve, finh->nodeid);
 
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
+        rv = fuse_flip_xattr_ns (priv, name, &newkey);
+        if (rv) {
+                op_errno = ENOMEM;
+                goto err;
         }
 
         state->size = fgxi->size;
-        state->name = gf_strdup (name);
+        state->name = newkey;
 
         fuse_resolve_and_resume (state, fuse_getxattr_resume);
+
+        return;
+ err:
+        send_fuse_err (this, finh, op_errno);
+        free_fuse_state (state);
         return;
 }
 
+
 void
 fuse_listxattr_resume (fuse_state_t *state)
 {
-        gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": LISTXATTR %s/%"PRIu64, state->finh->unique,
-                state->loc.path, state->finh->nodeid);
+        if (!state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "%"PRIu64": LISTXATTR %s/%"PRIu64
+                        "resolution failed", state->finh->unique,
+                        uuid_utoa (state->resolve.gfid), state->finh->nodeid);
 
-        FUSE_FOP (state, fuse_xattr_cbk, GF_FOP_GETXATTR,
-                  getxattr, &state->loc, NULL);
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
+
+#ifdef GF_TEST_FFOP
+        state->fd = fd_lookup (state->loc.inode, state->finh->pid);
+#endif /* GF_TEST_FFOP */
+
+        if (state->fd) {
+                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                        "%"PRIu64": LISTXATTR %p/%"PRIu64, state->finh->unique,
+                        state->fd, state->finh->nodeid);
+
+                FUSE_FOP (state, fuse_xattr_cbk, GF_FOP_FGETXATTR,
+                          fgetxattr, state->fd, NULL, state->xdata);
+        } else {
+                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                        "%"PRIu64": LISTXATTR %s/%"PRIu64, state->finh->unique,
+                        state->loc.path, state->finh->nodeid);
+
+                FUSE_FOP (state, fuse_xattr_cbk, GF_FOP_GETXATTR,
+                          getxattr, &state->loc, NULL, state->xdata);
+        }
 }
 
+
 static void
 fuse_listxattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
         struct fuse_getxattr_in *fgxi = msg;
-
         fuse_state_t *state = NULL;
-        int32_t       ret = -1;
 
         GET_STATE (this, finh, state);
 
-        ret = fuse_loc_fill (&state->loc, state, finh->nodeid, 0, NULL);
-        if ((state->loc.inode == NULL) ||
-            (ret < 0)) {
-                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
-                        "%"PRIu64": LISTXATTR %s/%"PRIu64" (fuse_loc_fill() failed)",
-                        finh->unique, state->loc.path, finh->nodeid);
-
-                send_fuse_err (this, finh, ENOENT);
-                free_fuse_state (state);
-                return;
-        }
+        fuse_resolve_inode_init (state, &state->resolve, finh->nodeid);
 
         state->size = fgxi->size;
+
         fuse_resolve_and_resume (state, fuse_listxattr_resume);
 
         return;
 }
 
+
 void
 fuse_removexattr_resume (fuse_state_t *state)
 {
-        gf_log ("glusterfs-fuse", GF_LOG_TRACE,
-                "%"PRIu64": REMOVEXATTR %s/%"PRIu64" (%s)", state->finh->unique,
-                state->loc.path, state->finh->nodeid, state->name);
+         if (!state->loc.inode) {
+                gf_log ("glusterfs-fuse", GF_LOG_DEBUG,
+                        "%"PRIu64": REMOVEXATTR %s/%"PRIu64" (%s) "
+                        "resolution failed",
+                        state->finh->unique, uuid_utoa (state->resolve.gfid),
+                        state->finh->nodeid, state->name);
+
+                send_fuse_err (state->this, state->finh, ENOENT);
+                free_fuse_state (state);
+                return;
+        }
+
+#ifdef GF_TEST_FFOP
+        state->fd = fd_lookup (state->loc.inode, state->finh->pid);
+#endif /* GF_TEST_FFOP */
 
-        FUSE_FOP (state, fuse_err_cbk, GF_FOP_REMOVEXATTR,
-                  removexattr, &state->loc, state->name);
+        if (state->fd) {
+                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                        "%"PRIu64": REMOVEXATTR %p/%"PRIu64" (%s)", state->finh->unique,
+                        state->fd, state->finh->nodeid, state->name);
+
+                FUSE_FOP (state, fuse_err_cbk, GF_FOP_FREMOVEXATTR,
+                          fremovexattr, state->fd, state->name, state->xdata);
+        } else {
+                gf_log ("glusterfs-fuse", GF_LOG_TRACE,
+                        "%"PRIu64": REMOVEXATTR %s/%"PRIu64" (%s)", state->finh->unique,
+                        state->loc.path, state->finh->nodeid, state->name);
+
+                FUSE_FOP (state, fuse_err_cbk, GF_FOP_REMOVEXATTR,
+                          removexattr, &state->loc, state->name, state->xdata);
+        }
 }
 
+
 static void
 fuse_removexattr (xlator_t *this, fuse_in_header_t *finh, void *msg)
-
 {
         char *name = msg;
 
         fuse_state_t *state = NULL;
+        fuse_private_t *priv = NULL;
         int32_t       ret = -1;
+        char *newkey = NULL;
+
+        if (!strcmp (GFID_XATTR_KEY, name)) {
+                send_fuse_err (this, finh, EPERM);
+                GF_FREE (finh);
+                return;
+        }
+
+        priv = this->private;
 
         GET_STATE (this, finh, state);
-        ret = fuse_loc_fill (&state->loc, state, finh->nodeid, 0, NULL);
-        if ((state->loc.inode == NULL) ||
-            (ret < 0)) {
-                gf_log ("glusterfs-fuse", GF_LOG_DEBUG,
-                        "%"PRIu64": REMOVEXATTR %s/%"PRIu64" (%s) (fuse_loc_fill() failed)",
-                        finh->unique, state->loc.path, finh->nodeid, name);
 
-                send_fuse_err (this, finh, ENOENT);
+        fuse_resolve_inode_init (state, &state->resolve, finh->nodeid);
+
+        ret = fuse_flip_xattr_ns (priv, name, &newkey);
+        if (ret) {
+                send_fuse_err (this, finh, ENOMEM);
                 free_fuse_state (state);
                 return;
         }
 
-        state->name = gf_strdup (name);
+        state->name = newkey;
+
         fuse_resolve_and_resume (state, fuse_removexattr_resume);
         return;
 }
@@ -2646,13 +3516,16 @@ static int gf_fuse_lk_enosys_log;
 
 static int
 fuse_getlk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                int32_t op_ret, int32_t op_errno, struct flock *lock)
+                int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+                dict_t *xdata)
 {
         fuse_state_t *state = NULL;
 
         state = frame->root->state;
         struct fuse_lk_out flo = {{0, }, };
 
+        fuse_log_eh_fop(this, state, frame, op_ret, op_errno);
+
         if (op_ret == 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                         "%"PRIu64": ERR => 0", frame->root->unique);
@@ -2690,6 +3563,7 @@ fuse_getlk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         return 0;
 }
 
+
 void
 fuse_getlk_resume (fuse_state_t *state)
 {
@@ -2697,9 +3571,10 @@ fuse_getlk_resume (fuse_state_t *state)
                 "%"PRIu64": GETLK %p", state->finh->unique, state->fd);
 
         FUSE_FOP (state, fuse_getlk_cbk, GF_FOP_LK,
-                  lk, state->fd, F_GETLK, &state->lk_lock);
+                  lk, state->fd, F_GETLK, &state->lk_lock, state->xdata);
 }
 
+
 static void
 fuse_getlk (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
@@ -2711,7 +3586,11 @@ fuse_getlk (xlator_t *this, fuse_in_header_t *finh, void *msg)
         fd = FH_TO_FD (fli->fh);
         GET_STATE (this, finh, state);
         state->fd = fd;
-        convert_fuse_file_lock (&fli->lk, &state->lk_lock);
+
+        fuse_resolve_fd_init (state, &state->resolve, fd);
+
+        convert_fuse_file_lock (&fli->lk, &state->lk_lock,
+                                fli->owner);
 
         state->lk_owner = fli->owner;
 
@@ -2723,15 +3602,24 @@ fuse_getlk (xlator_t *this, fuse_in_header_t *finh, void *msg)
 
 static int
 fuse_setlk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                int32_t op_ret, int32_t op_errno, struct flock *lock)
+                int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+                dict_t *xdata)
 {
+        uint32_t      op    = 0;
         fuse_state_t *state = NULL;
 
         state = frame->root->state;
+        op    = state->finh->opcode;
+
+        fuse_log_eh_fop(this, state, frame, op_ret, op_errno);
 
         if (op_ret == 0) {
                 gf_log ("glusterfs-fuse", GF_LOG_TRACE,
                         "%"PRIu64": ERR => 0", frame->root->unique);
+                fd_lk_insert_and_merge (state->fd,
+                                        (op == FUSE_SETLK) ? F_SETLK : F_SETLKW,
+                                        &state->lk_lock);
+
                 send_fuse_err (this, state->finh, 0);
         } else {
                 if (op_errno == ENOSYS) {
@@ -2745,12 +3633,11 @@ fuse_setlk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 } else if (op_errno == EAGAIN) {
                         gf_log ("glusterfs-fuse", GF_LOG_DEBUG,
                                 "Returning EAGAIN Flock: "
-                                "start=%llu, len=%llu, pid=%llu, lk-owner=%llu",
-                                (unsigned long long) lock->l_start,
-                                (unsigned long long) lock->l_len,
-                                (unsigned long long) lock->l_pid,
-                                (unsigned long long) frame->root->lk_owner);
-
+                                "start=%llu, len=%llu, pid=%llu, lk-owner=%s",
+                                (unsigned long long) state->lk_lock.l_start,
+                                (unsigned long long) state->lk_lock.l_len,
+                                (unsigned long long) state->lk_lock.l_pid,
+                                lkowner_utoa (&frame->root->lk_owner));
                 } else  {
                         gf_log ("glusterfs-fuse", GF_LOG_WARNING,
                                 "%"PRIu64": ERR => -1 (%s)",
@@ -2776,9 +3663,10 @@ fuse_setlk_resume (fuse_state_t *state)
 
         FUSE_FOP (state, fuse_setlk_cbk, GF_FOP_LK, lk, state->fd,
                   state->finh->opcode == FUSE_SETLK ? F_SETLK : F_SETLKW,
-                  &state->lk_lock);
+                  &state->lk_lock, state->xdata);
 }
 
+
 static void
 fuse_setlk (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
@@ -2791,7 +3679,11 @@ fuse_setlk (xlator_t *this, fuse_in_header_t *finh, void *msg)
         GET_STATE (this, finh, state);
         state->finh = finh;
         state->fd = fd;
-        convert_fuse_file_lock (&fli->lk, &state->lk_lock);
+
+        fuse_resolve_fd_init (state, &state->resolve, fd);
+
+        convert_fuse_file_lock (&fli->lk, &state->lk_lock,
+                                fli->owner);
 
         state->lk_owner = fli->owner;
 
@@ -2800,14 +3692,53 @@ fuse_setlk (xlator_t *this, fuse_in_header_t *finh, void *msg)
         return;
 }
 
+
+static void *
+notify_kernel_loop (void *data)
+{
+        xlator_t               *this = NULL;
+        fuse_private_t         *priv = NULL;
+        struct fuse_out_header *fouh = NULL;
+        int                     rv   = 0;
+
+        char inval_buf[INVAL_BUF_SIZE] = {0,};
+
+        this = data;
+        priv = this->private;
+
+        for (;;) {
+                rv = read (priv->revchan_in, inval_buf, sizeof (*fouh));
+                if (rv != sizeof (*fouh))
+                        break;
+                fouh = (struct fuse_out_header *)inval_buf;
+                rv = read (priv->revchan_in, inval_buf + sizeof (*fouh),
+                           fouh->len - sizeof (*fouh));
+                if (rv != fouh->len - sizeof (*fouh))
+                        break;
+                rv = write (priv->fd, inval_buf, fouh->len);
+                if (rv != fouh->len && !(rv == -1 && errno == ENOENT))
+                        break;
+        }
+
+        close (priv->revchan_in);
+        close (priv->revchan_out);
+
+        gf_log ("glusterfs-fuse", GF_LOG_INFO,
+                "kernel notifier loop terminated");
+
+        return NULL;
+}
+
+
 static void
 fuse_init (xlator_t *this, fuse_in_header_t *finh, void *msg)
 {
-        struct fuse_init_in *fini = msg;
-
-        struct fuse_init_out fino;
-        fuse_private_t *priv = NULL;
-        int ret;
+        struct fuse_init_in  *fini      = msg;
+        struct fuse_init_out  fino      = {0,};
+        fuse_private_t       *priv      = NULL;
+        int                   ret       = 0;
+        int                   pfd[2]    = {0,};
+        pthread_t             messenger;
 
         priv = this->private;
 
@@ -2836,6 +3767,17 @@ fuse_init (xlator_t *this, fuse_in_header_t *finh, void *msg)
         fino.max_readahead = 1 << 17;
         fino.max_write = 1 << 17;
         fino.flags = FUSE_ASYNC_READ | FUSE_POSIX_LOCKS;
+#if FUSE_KERNEL_MINOR_VERSION >= 17
+	if (fini->minor >= 17)
+		fino.flags |= FUSE_FLOCK_LOCKS;
+#endif
+#if FUSE_KERNEL_MINOR_VERSION >= 12
+        if (fini->minor >= 12) {
+            /* let fuse leave the umask processing to us, so that it does not
+             * break extended POSIX ACL defaults on server */
+            fino.flags |= FUSE_DONT_MASK;
+        }
+#endif
 #if FUSE_KERNEL_MINOR_VERSION >= 9
         if (fini->minor >= 6 /* fuse_init_in has flags */ &&
             fini->flags & FUSE_BIG_WRITES) {
@@ -2844,15 +3786,94 @@ fuse_init (xlator_t *this, fuse_in_header_t *finh, void *msg)
                         priv->direct_io_mode = 0;
                 fino.flags |= FUSE_BIG_WRITES;
         }
-        if (fini->minor >= 13) {
-                /* these values seemed to work fine during testing */
 
-                fino.max_background = 64;
-                fino.congestion_threshold = 48;
+        /* Used for 'reverse invalidation of inode' */
+        if (fini->minor >= 12) {
+                if (pipe(pfd) == -1) {
+                        gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                                "cannot create pipe pair (%s)",
+                                strerror(errno));
+
+                        close (priv->fd);
+                        goto out;
+                }
+                priv->revchan_in  = pfd[0];
+                priv->revchan_out = pfd[1];
+                ret = gf_thread_create (&messenger, NULL, notify_kernel_loop,
+					this);
+                if (ret != 0) {
+                        gf_log ("glusterfs-fuse", GF_LOG_ERROR,
+                                "failed to start messenger daemon (%s)",
+                                strerror(errno));
+
+                        close (priv->fd);
+                        goto out;
+                }
+                priv->reverse_fuse_thread_started = _gf_true;
+        } else {
+                /*
+                 * FUSE minor < 12 does not implement invalidate notifications.
+                 * This mechanism is required for fopen-keep-cache to operate
+                 * correctly. Disable and warn the user.
+                 */
+                if (priv->fopen_keep_cache) {
+                        gf_log("glusterfs-fuse", GF_LOG_WARNING, "FUSE version "
+                                "%d.%d does not support inval notifications. "
+                                "fopen-keep-cache disabled.", fini->major,
+                                fini->minor);
+                        priv->fopen_keep_cache = 0;
+                }
+        }
+
+        if (fini->minor >= 13) {
+                fino.max_background = priv->background_qlen;
+                fino.congestion_threshold = priv->congestion_threshold;
         }
         if (fini->minor < 9)
                 *priv->msg0_len_p = sizeof(*finh) + FUSE_COMPAT_WRITE_IN_SIZE;
 #endif
+        if (priv->use_readdirp) {
+                if (fini->flags & FUSE_DO_READDIRPLUS)
+                        fino.flags |= FUSE_DO_READDIRPLUS;
+        }
+
+	if (priv->fopen_keep_cache == 2) {
+		/* If user did not explicitly set --fopen-keep-cache[=off],
+		   then check if kernel support FUSE_AUTO_INVAL_DATA and ...
+		*/
+		if (fini->flags & FUSE_AUTO_INVAL_DATA) {
+			/* ... enable fopen_keep_cache mode if supported.
+			*/
+			gf_log ("glusterfs-fuse", GF_LOG_DEBUG, "Detected "
+				"support for FUSE_AUTO_INVAL_DATA. Enabling "
+				"fopen_keep_cache automatically.");
+			fino.flags |= FUSE_AUTO_INVAL_DATA;
+			priv->fopen_keep_cache = 1;
+		} else {
+			gf_log ("glusterfs-fuse", GF_LOG_DEBUG, "No support "
+				"for FUSE_AUTO_INVAL_DATA. Disabling "
+				"fopen_keep_cache.");
+			/* ... else disable. */
+			priv->fopen_keep_cache = 0;
+		}
+	} else if (priv->fopen_keep_cache == 1) {
+		/* If user explicitly set --fopen-keep-cache[=on],
+		   then enable FUSE_AUTO_INVAL_DATA if possible.
+		*/
+		if (fini->flags & FUSE_AUTO_INVAL_DATA) {
+			gf_log ("glusterfs-fuse", GF_LOG_DEBUG, "fopen_keep_cache "
+				"is explicitly set. Enabling FUSE_AUTO_INVAL_DATA");
+			fino.flags |= FUSE_AUTO_INVAL_DATA;
+		} else {
+			gf_log ("glusterfs-fuse", GF_LOG_WARNING, "fopen_keep_cache "
+				"is explicitly set. Support for "
+				"FUSE_AUTO_INVAL_DATA is missing");
+		}
+	}
+
+	if (fini->flags & FUSE_ASYNC_DIO)
+		fino.flags |= FUSE_ASYNC_DIO;
+
         ret = send_fuse_obj (this, finh, &fino);
         if (ret == 0)
                 gf_log ("glusterfs-fuse", GF_LOG_INFO,
@@ -2942,8 +3963,8 @@ fuse_first_lookup (xlator_t *this)
 
         loc.path = "/";
         loc.name = "";
-        loc.ino = 1;
         loc.inode = fuse_ino_to_inode (1, this);
+        uuid_copy (loc.gfid, loc.inode->gfid);
         loc.parent = NULL;
 
         dict = dict_new ();
@@ -2961,6 +3982,8 @@ fuse_first_lookup (xlator_t *this)
         memset (gfid, 0, 16);
         gfid[15] = 1;
         ret = dict_set_static_bin (dict, "gfid-req", gfid, 16);
+        if (ret)
+                gf_log (xl->name, GF_LOG_ERROR, "failed to set 'gfid-req'");
 
         STACK_WIND (frame, fuse_first_lookup_cbk, xl, xl->fops->lookup,
                     &loc, dict);
@@ -2985,13 +4008,507 @@ fuse_first_lookup (xlator_t *this)
 
 
 int
+fuse_nameless_lookup (xlator_t *xl, uuid_t gfid, loc_t *loc)
+{
+        int          ret          = -1;
+        dict_t      *xattr_req    = NULL;
+        struct iatt  iatt         = {0, };
+        inode_t     *linked_inode = NULL;
+
+        if ((loc == NULL) || (xl == NULL)) {
+                goto out;
+        }
+
+        if (loc->inode == NULL) {
+                loc->inode = inode_new (xl->itable);
+                if (loc->inode == NULL) {
+                        goto out;
+                }
+        }
+
+        uuid_copy (loc->gfid, gfid);
+
+        xattr_req = dict_new ();
+        if (xattr_req == NULL) {
+                goto out;
+        }
+
+        ret = syncop_lookup (xl, loc, xattr_req, &iatt, NULL, NULL);
+        if (ret < 0) {
+                goto out;
+        }
+
+        linked_inode = inode_link (loc->inode, NULL, NULL, &iatt);
+        inode_unref (loc->inode);
+        loc->inode = linked_inode;
+
+        ret = 0;
+out:
+        if (xattr_req != NULL) {
+                dict_unref (xattr_req);
+        }
+
+        return ret;
+}
+
+
+int
+fuse_migrate_fd_open (xlator_t *this, fd_t *basefd, fd_t *oldfd,
+                      xlator_t *old_subvol, xlator_t *new_subvol)
+{
+        loc_t          loc       = {0, };
+        fd_t          *newfd     = NULL, *old_activefd = NULL;
+        fuse_fd_ctx_t *basefd_ctx = NULL;
+        fuse_fd_ctx_t *newfd_ctx = NULL;
+        int            ret       = 0, flags = 0;
+
+        ret = inode_path (basefd->inode, NULL, (char **)&loc.path);
+        if (ret < 0) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "cannot construct path of gfid (%s) failed"
+                        "(old-subvolume:%s-%d new-subvolume:%s-%d)",
+                        uuid_utoa (basefd->inode->gfid),
+                        old_subvol->name, old_subvol->graph->id,
+                        new_subvol->name, new_subvol->graph->id);
+                goto out;
+        }
+
+        uuid_copy (loc.gfid, basefd->inode->gfid);
+
+        loc.inode = inode_find (new_subvol->itable, basefd->inode->gfid);
+
+        if (loc.inode == NULL) {
+                ret = fuse_nameless_lookup (new_subvol, basefd->inode->gfid,
+                                            &loc);
+                if (ret < 0) {
+                        gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                                "name-less lookup of gfid (%s) failed (%s)"
+                                "(old-subvolume:%s-%d new-subvolume:%s-%d)",
+                                uuid_utoa (basefd->inode->gfid),
+                                strerror (errno),
+                                old_subvol->name, old_subvol->graph->id,
+                                new_subvol->name, new_subvol->graph->id);
+                        goto out;
+                }
+
+        }
+
+        basefd_ctx = fuse_fd_ctx_get (this, basefd);
+        GF_VALIDATE_OR_GOTO ("glusterfs-fuse", basefd_ctx, out);
+
+        newfd = fd_create (loc.inode, basefd->pid);
+        if (newfd == NULL) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "cannot create new fd, hence not migrating basefd "
+                        "(ptr:%p inode-gfid:%s) "
+                        "(old-subvolume:%s-%d new-subvolume:%s-%d)", basefd,
+                        uuid_utoa (loc.inode->gfid),
+                        old_subvol->name, old_subvol->graph->id,
+                        new_subvol->name, new_subvol->graph->id);
+                goto out;
+        }
+
+        newfd->flags = basefd->flags;
+	if (newfd->lk_ctx)
+		fd_lk_ctx_unref (newfd->lk_ctx);
+
+        newfd->lk_ctx = fd_lk_ctx_ref (oldfd->lk_ctx);
+
+        newfd_ctx = fuse_fd_ctx_check_n_create (this, newfd);
+        GF_VALIDATE_OR_GOTO ("glusterfs-fuse", newfd_ctx, out);
+
+        if (IA_ISDIR (basefd->inode->ia_type)) {
+                ret = syncop_opendir (new_subvol, &loc, newfd);
+        } else {
+                flags = basefd->flags & ~(O_CREAT | O_EXCL | O_TRUNC);
+                ret = syncop_open (new_subvol, &loc, flags, newfd);
+        }
+
+        if (ret < 0) {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "open on basefd (ptr:%p inode-gfid:%s) failed (%s)"
+                        "(old-subvolume:%s-%d new-subvolume:%s-%d)", basefd,
+                        uuid_utoa (basefd->inode->gfid), strerror (errno),
+                        old_subvol->name, old_subvol->graph->id,
+                        new_subvol->name, new_subvol->graph->id);
+                goto out;
+        }
+
+        fd_bind (newfd);
+
+        LOCK (&basefd->lock);
+        {
+                if (basefd_ctx->activefd != NULL) {
+                        old_activefd = basefd_ctx->activefd;
+                }
+
+                basefd_ctx->activefd = newfd;
+        }
+        UNLOCK (&basefd->lock);
+
+        if (old_activefd != NULL) {
+                fd_unref (old_activefd);
+        }
+
+        gf_log ("glusterfs-fuse", GF_LOG_INFO,
+                "migrated basefd (%p) to newfd (%p) (inode-gfid:%s)"
+                "(old-subvolume:%s-%d new-subvolume:%s-%d)", basefd, newfd,
+                uuid_utoa (basefd->inode->gfid),
+		old_subvol->name, old_subvol->graph->id,
+                new_subvol->name, new_subvol->graph->id);
+
+        ret = 0;
+
+out:
+        loc_wipe (&loc);
+
+        return ret;
+}
+
+int
+fuse_migrate_locks (xlator_t *this, fd_t *basefd, fd_t *oldfd,
+                    xlator_t *old_subvol, xlator_t *new_subvol)
+{
+        int	 ret      = -1;
+        dict_t	*lockinfo = NULL;
+        void    *ptr      = NULL;
+	fd_t    *newfd    = NULL;
+        fuse_fd_ctx_t *basefd_ctx         = NULL;
+
+
+	if (!oldfd->lk_ctx || fd_lk_ctx_empty (oldfd->lk_ctx))
+		return 0;
+
+        basefd_ctx = fuse_fd_ctx_get (this, basefd);
+        GF_VALIDATE_OR_GOTO ("glusterfs-fuse", basefd_ctx, out);
+
+        LOCK (&basefd->lock);
+        {
+                newfd = fd_ref (basefd_ctx->activefd);
+        }
+        UNLOCK (&basefd->lock);
+
+        ret = syncop_fgetxattr (old_subvol, oldfd, &lockinfo,
+                                GF_XATTR_LOCKINFO_KEY);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+			"getting lockinfo failed while migrating locks"
+			"(oldfd:%p newfd:%p inode-gfid:%s)"
+			"(old-subvol:%s-%d new-subvol:%s-%d)",
+			oldfd, newfd, uuid_utoa (newfd->inode->gfid),
+			old_subvol->name, old_subvol->graph->id,
+			new_subvol->name, new_subvol->graph->id);
+                goto out;
+        }
+
+        ret = dict_get_ptr (lockinfo, GF_XATTR_LOCKINFO_KEY, &ptr);
+        if (ptr == NULL) {
+                ret = 0;
+                gf_log (this->name, GF_LOG_INFO,
+                        "No lockinfo present on any of the bricks "
+			"(oldfd: %p newfd:%p inode-gfid:%s) "
+                        "(old-subvol:%s-%d new-subvol:%s-%d)",
+			oldfd, newfd, uuid_utoa (newfd->inode->gfid),
+			old_subvol->name, old_subvol->graph->id,
+			new_subvol->name, new_subvol->graph->id);
+
+                goto out;
+        }
+
+        ret = syncop_fsetxattr (new_subvol, newfd, lockinfo, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+			"migrating locks failed (oldfd:%p newfd:%p "
+			"inode-gfid:%s) (old-subvol:%s-%d new-subvol:%s-%d)",
+			oldfd, newfd, uuid_utoa (newfd->inode->gfid),
+			old_subvol->name, old_subvol->graph->id,
+			new_subvol->name, new_subvol->graph->id);
+                goto out;
+        }
+
+out:
+	if (newfd)
+		fd_unref (newfd);
+
+        if (lockinfo != NULL) {
+                dict_unref (lockinfo);
+        }
+
+        return ret;
+}
+
+
+int
+fuse_migrate_fd (xlator_t *this, fd_t *basefd, xlator_t *old_subvol,
+                 xlator_t *new_subvol)
+{
+        int            ret                = -1;
+        char           create_in_progress = 0;
+        fuse_fd_ctx_t *basefd_ctx         = NULL;
+        fd_t          *oldfd              = NULL;
+
+        basefd_ctx = fuse_fd_ctx_get (this, basefd);
+        GF_VALIDATE_OR_GOTO ("glusterfs-fuse", basefd_ctx, out);
+
+        LOCK (&basefd->lock);
+        {
+                oldfd = basefd_ctx->activefd ? basefd_ctx->activefd
+                        : basefd;
+                fd_ref (oldfd);
+        }
+        UNLOCK (&basefd->lock);
+
+        LOCK (&oldfd->inode->lock);
+        {
+                if (uuid_is_null (oldfd->inode->gfid)) {
+                        create_in_progress = 1;
+                } else {
+                        create_in_progress = 0;
+                }
+        }
+        UNLOCK (&oldfd->inode->lock);
+
+        if (create_in_progress) {
+                gf_log ("glusterfs-fuse", GF_LOG_INFO,
+                        "create call on fd (%p) is in progress "
+                        "(basefd-ptr:%p basefd-inode.gfid:%s), "
+                        "hence deferring migration till application does an "
+                        "fd based operation on this fd"
+                        "(old-subvolume:%s-%d, new-subvolume:%s-%d)",
+                        oldfd, basefd, uuid_utoa (basefd->inode->gfid),
+                        old_subvol->name, old_subvol->graph->id,
+                        new_subvol->name, new_subvol->graph->id);
+
+                ret = 0;
+                goto out;
+        }
+
+        if (oldfd->inode->table->xl == old_subvol) {
+                ret = syncop_fsync (old_subvol, oldfd, 0);
+                if (ret < 0) {
+                        gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                                "syncop_fsync failed (%s) on fd (%p)"
+                                "(basefd:%p basefd-inode.gfid:%s) "
+                                "(old-subvolume:%s-%d new-subvolume:%s-%d)",
+                                strerror (errno), oldfd, basefd,
+                                uuid_utoa (basefd->inode->gfid),
+                                old_subvol->name, old_subvol->graph->id,
+                                new_subvol->name, new_subvol->graph->id);
+                }
+        } else {
+                gf_log ("glusterfs-fuse", GF_LOG_WARNING,
+                        "basefd (ptr:%p inode-gfid:%s) was not "
+                        "migrated during previous graph switch"
+                        "(old-subvolume:%s-%d new-subvolume: %s-%d)", basefd,
+                        basefd->inode->gfid,
+                        old_subvol->name, old_subvol->graph->id,
+                        new_subvol->name, new_subvol->graph->id);
+        }
+
+        ret = fuse_migrate_fd_open (this, basefd, oldfd, old_subvol,
+                                    new_subvol);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "open corresponding to "
+                        "basefd (ptr:%p inode-gfid:%s) in new graph failed "
+                        "(old-subvolume:%s-%d new-subvolume:%s-%d)", basefd,
+                        uuid_utoa (basefd->inode->gfid), old_subvol->name,
+                        old_subvol->graph->id, new_subvol->name,
+                        new_subvol->graph->id);
+                goto out;
+        }
+
+        ret = fuse_migrate_locks (this, basefd, oldfd, old_subvol,
+                                  new_subvol);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "migrating locks from old-subvolume (%s-%d) to "
+                        "new-subvolume (%s-%d) failed (inode-gfid:%s oldfd:%p "
+                        "basefd:%p)", old_subvol->name, old_subvol->graph->id,
+                        new_subvol->name, new_subvol->graph->id,
+                        uuid_utoa (basefd->inode->gfid), oldfd, basefd);
+
+        }
+out:
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "migration of basefd "
+                        "(ptr:%p inode-gfid:%s) failed"
+                        "(old-subvolume:%s-%d new-subvolume:%s-%d)", basefd,
+                        oldfd ? uuid_utoa (oldfd->inode->gfid) : NULL,
+                        old_subvol->name, old_subvol->graph->id,
+                        new_subvol->name, new_subvol->graph->id);
+        }
+
+        fd_unref (oldfd);
+
+        return ret;
+}
+
+
+int
+fuse_handle_opened_fds (xlator_t *this, xlator_t *old_subvol,
+                        xlator_t *new_subvol)
+{
+        fuse_private_t *priv      = NULL;
+        fdentry_t      *fdentries = NULL;
+        uint32_t        count     = 0;
+        fdtable_t      *fdtable   = NULL;
+        int             i         = 0;
+        fd_t           *fd        = NULL;
+        int32_t         ret       = 0;
+        fuse_fd_ctx_t  *fdctx     = NULL;
+
+        priv = this->private;
+
+        fdtable = priv->fdtable;
+
+        fdentries = gf_fd_fdtable_copy_all_fds (fdtable, &count);
+        if (fdentries != NULL) {
+                for (i = 0; i < count; i++) {
+                        fd = fdentries[i].fd;
+                        if (fd == NULL)
+                                continue;
+
+                        ret = fuse_migrate_fd (this, fd, old_subvol,
+                                               new_subvol);
+
+                        fdctx = fuse_fd_ctx_get (this, fd);
+                        if (fdctx) {
+                                LOCK (&fd->lock);
+                                {
+                                        if (ret < 0) {
+                                                fdctx->migration_failed = 1;
+                                        } else {
+                                                fdctx->migration_failed = 0;
+                                        }
+                                }
+                                UNLOCK (&fd->lock);
+                        }
+                }
+
+                for (i = 0; i < count ; i++) {
+                        fd = fdentries[i].fd;
+                        if (fd)
+                                fd_unref (fd);
+                }
+
+                GF_FREE (fdentries);
+        }
+
+        return 0;
+}
+
+
+static int
+fuse_handle_blocked_locks (xlator_t *this, xlator_t *old_subvol,
+                           xlator_t *new_subvol)
+{
+        return 0;
+}
+
+
+static int
+fuse_graph_switch_task (void *data)
+{
+        fuse_graph_switch_args_t *args = NULL;
+
+        args = data;
+        if (args == NULL) {
+                goto out;
+        }
+
+        /* don't change the order of handling open fds and blocked locks, since
+         * the act of opening files also reacquires granted locks in new graph.
+         */
+        fuse_handle_opened_fds (args->this, args->old_subvol, args->new_subvol);
+
+        fuse_handle_blocked_locks (args->this, args->old_subvol,
+                                   args->new_subvol);
+
+out:
+        return 0;
+}
+
+
+fuse_graph_switch_args_t *
+fuse_graph_switch_args_alloc (void)
+{
+        fuse_graph_switch_args_t *args = NULL;
+
+        args = GF_CALLOC (1, sizeof (*args), gf_fuse_mt_graph_switch_args_t);
+        if (args == NULL) {
+                goto out;
+        }
+
+out:
+        return args;
+}
+
+
+void
+fuse_graph_switch_args_destroy (fuse_graph_switch_args_t *args)
+{
+        if (args == NULL) {
+                goto out;
+        }
+
+        GF_FREE (args);
+out:
+        return;
+}
+
+
+int
+fuse_handle_graph_switch (xlator_t *this, xlator_t *old_subvol,
+                          xlator_t *new_subvol)
+{
+        call_frame_t             *frame = NULL;
+        int32_t                   ret   = -1;
+        fuse_graph_switch_args_t *args  = NULL;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (frame == NULL) {
+                goto out;
+        }
+
+        args = fuse_graph_switch_args_alloc ();
+        if (args == NULL) {
+                goto out;
+        }
+
+        args->this = this;
+        args->old_subvol = old_subvol;
+        args->new_subvol = new_subvol;
+
+        ret = synctask_new (this->ctx->env, fuse_graph_switch_task, NULL, frame,
+                            args);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING, "starting sync-task to "
+                        "handle graph switch failed");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if (args != NULL) {
+                fuse_graph_switch_args_destroy (args);
+        }
+
+        if (frame != NULL) {
+                STACK_DESTROY (frame->root);
+        }
+
+        return ret;
+}
+
+
+int
 fuse_graph_sync (xlator_t *this)
 {
-        fuse_private_t   *priv = NULL;
-        int               need_first_lookup = 0;
-        struct timeval    now = {0, };
-        struct timespec   timeout = {0, };
-        int               ret = 0;
+        fuse_private_t *priv                = NULL;
+        int             need_first_lookup   = 0;
+        int             ret                 = 0;
+        xlator_t       *old_subvol          = NULL, *new_subvol = NULL;
+        uint64_t        winds_on_old_subvol = 0;
 
         priv = this->private;
 
@@ -3000,24 +4517,20 @@ fuse_graph_sync (xlator_t *this)
                 if (!priv->next_graph)
                         goto unlock;
 
-                priv->active_subvol = priv->next_graph->top;
+                old_subvol = priv->active_subvol;
+                new_subvol = priv->active_subvol = priv->next_graph->top;
                 priv->next_graph = NULL;
                 need_first_lookup = 1;
 
-                gettimeofday (&now, NULL);
-                timeout.tv_sec = now.tv_sec + MAX_FUSE_PROC_DELAY;
-                timeout.tv_nsec = now.tv_usec * 1000;
-
-                while (!priv->child_up) {
-                        ret = pthread_cond_timedwait (&priv->sync_cond,
-                                                      &priv->sync_mutex,
-                                                      &timeout);
+                while (!priv->event_recvd) {
+                        ret = pthread_cond_wait (&priv->sync_cond,
+                                                 &priv->sync_mutex);
                         if (ret != 0) {
-			          gf_log (this->name, GF_LOG_DEBUG,
-					  "timedwait returned non zero value "
-					  "ret: %d errno: %d", ret, errno);
-				  break;
-			}
+                                  gf_log (this->name, GF_LOG_DEBUG,
+                                          "timedwait returned non zero value "
+                                          "ret: %d errno: %d", ret, errno);
+                                  break;
+                        }
                 }
         }
 unlock:
@@ -3027,23 +4540,57 @@ unlock:
                 fuse_first_lookup (this);
         }
 
+        if ((old_subvol != NULL) && (new_subvol != NULL)) {
+                fuse_handle_graph_switch (this, old_subvol, new_subvol);
+
+                pthread_mutex_lock (&priv->sync_mutex);
+                {
+                        old_subvol->switched = 1;
+                        winds_on_old_subvol = old_subvol->winds;
+                }
+                pthread_mutex_unlock (&priv->sync_mutex);
+
+                if (winds_on_old_subvol == 0) {
+                        xlator_notify (old_subvol, GF_EVENT_PARENT_DOWN,
+                                       old_subvol, NULL);
+                }
+        }
+
         return 0;
 }
 
+int
+fuse_get_mount_status (xlator_t *this)
+{
+        int             kid_status = -1;
+        fuse_private_t *priv = this->private;
+
+        if (read(priv->status_pipe[0],&kid_status, sizeof(kid_status)) < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "could not get mount status");
+                kid_status = -1;
+        }
+        gf_log (this->name, GF_LOG_DEBUG, "mount status is %d", kid_status);
+
+        close(priv->status_pipe[0]);
+        close(priv->status_pipe[1]);
+        return kid_status;
+}
 
 static void *
 fuse_thread_proc (void *data)
 {
-        char           *mount_point = NULL;
-        xlator_t       *this = NULL;
-        fuse_private_t *priv = NULL;
-        ssize_t         res = 0;
-        struct iobuf   *iobuf = NULL;
-        fuse_in_header_t *finh;
-        struct iovec iov_in[2];
-        void *msg = NULL;
-        const size_t msg0_size = sizeof (*finh) + 128;
-        fuse_handler_t **fuse_ops = NULL;
+        char                     *mount_point = NULL;
+        xlator_t                 *this = NULL;
+        fuse_private_t           *priv = NULL;
+        ssize_t                   res = 0;
+        struct iobuf             *iobuf = NULL;
+        fuse_in_header_t         *finh;
+        struct iovec              iov_in[2];
+        void                     *msg = NULL;
+        const size_t              msg0_size = sizeof (*finh) + 128;
+        fuse_handler_t          **fuse_ops = NULL;
+        struct pollfd             pfd[2] = {{0,}};
+        gf_boolean_t              mount_finished = _gf_false;
 
         this = data;
         priv = this->private;
@@ -3053,19 +4600,59 @@ fuse_thread_proc (void *data)
 
         iov_in[0].iov_len = sizeof (*finh) + sizeof (struct fuse_write_in);
         iov_in[1].iov_len = ((struct iobuf_pool *)this->ctx->iobuf_pool)
-                              ->page_size;
+                              ->default_page_size;
         priv->msg0_len_p = &iov_in[0].iov_len;
 
         for (;;) {
                 /* THIS has to be reset here */
                 THIS = this;
 
+                if (!mount_finished) {
+                        memset(pfd,0,sizeof(pfd));
+                        pfd[0].fd = priv->status_pipe[0];
+                        pfd[0].events = POLLIN | POLLHUP | POLLERR;
+                        pfd[1].fd = priv->fd;
+                        pfd[1].events = POLLIN | POLLHUP | POLLERR;
+                        if (poll(pfd,2,-1) < 0) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "poll error %s", strerror(errno));
+                                break;
+                        }
+                        if (pfd[0].revents & POLLIN) {
+                                if (fuse_get_mount_status(this) != 0) {
+                                        break;
+                                }
+                                mount_finished = _gf_true;
+                        }
+                        else if (pfd[0].revents) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "mount pipe closed without status");
+                                break;
+                        }
+                        if (!pfd[1].revents) {
+                                continue;
+                        }
+                }
+
+                /*
+                 * We don't want to block on readv while we're still waiting
+                 * for mount status.  That means we only want to get here if
+                 * mount_status is true (meaning that our wait completed
+                 * already) or if we already called poll(2) on priv->fd to
+                 * make sure it's ready.
+                 */
+
                 if (priv->init_recvd)
                         fuse_graph_sync (this);
 
+                /* TODO: This place should always get maximum supported buffer
+                   size from 'fuse', which is as of today 128KB. If we bring in
+                   support for higher block sizes support, then we should be
+                   changing this one too */
                 iobuf = iobuf_get (this->ctx->iobuf_pool);
+
                 /* Add extra 128 byte to the first iov so that it can
-                 * accomodate "ordinary" non-write requests. It's not
+                 * accommodate "ordinary" non-write requests. It's not
                  * guaranteed to be big enough, as SETXATTR and namespace
                  * operations with very long names may grow behind it,
                  * but it's good enough in most cases (and we can handle
@@ -3090,17 +4677,24 @@ fuse_thread_proc (void *data)
 
                 if (res == -1) {
                         if (errno == ENODEV || errno == EBADF) {
-                                gf_log ("glusterfs-fuse", GF_LOG_NORMAL,
+                                gf_log ("glusterfs-fuse", GF_LOG_DEBUG,
                                         "terminating upon getting %s when "
                                         "reading /dev/fuse",
                                         errno == ENODEV ? "ENODEV" : "EBADF");
-
+                                fuse_log_eh (this, "glusterfs-fuse: terminating"
+                                             " upon getting %s when "
+                                             "reading /dev/fuse",
+                                             errno == ENODEV ? "ENODEV":
+                                             "EBADF");
                                 break;
                         }
                         if (errno != EINTR) {
                                 gf_log ("glusterfs-fuse", GF_LOG_WARNING,
                                         "read from /dev/fuse returned -1 (%s)",
                                         strerror (errno));
+                                fuse_log_eh (this, "glusterfs-fuse: read from "
+                                             "/dev/fuse returned -1 (%s)",
+                                             strerror (errno));
                         }
 
                         goto cont_err;
@@ -3108,6 +4702,8 @@ fuse_thread_proc (void *data)
                 if (res < sizeof (finh)) {
                         gf_log ("glusterfs-fuse", GF_LOG_WARNING,
                                 "short read on /dev/fuse");
+                        fuse_log_eh (this, "glusterfs-fuse: short read on "
+                                     "/dev/fuse");
                         break;
                 }
 
@@ -3126,6 +4722,8 @@ fuse_thread_proc (void *data)
                    ) {
                         gf_log ("glusterfs-fuse", GF_LOG_WARNING,
                                 "inconsistent read on /dev/fuse");
+                        fuse_log_eh (this, "glusterfs-fuse: inconsistent read "
+                                     "on /dev/fuse");
                         break;
                 }
 
@@ -3135,11 +4733,12 @@ fuse_thread_proc (void *data)
                         msg = iov_in[1].iov_base;
                 else {
                         if (res > msg0_size) {
-                                iov_in[0].iov_base =
-                                  GF_REALLOC (iov_in[0].iov_base, res);
-                                if (iov_in[0].iov_base)
+                                void *b = GF_REALLOC (iov_in[0].iov_base, res);
+                                if (b) {
+                                        iov_in[0].iov_base = b;
                                         finh = (fuse_in_header_t *)
                                                  iov_in[0].iov_base;
+                                }
                                 else {
                                         gf_log ("glusterfs-fuse", GF_LOG_ERROR,
                                                 "Out of memory");
@@ -3156,13 +4755,15 @@ fuse_thread_proc (void *data)
 
                         msg = finh + 1;
                 }
-#ifdef GF_DARWIN_HOST_OS
+                if (priv->uid_map_root &&
+                    finh->uid == priv->uid_map_root)
+                        finh->uid = 0;
+
                 if (finh->opcode >= FUSE_OP_HIGH)
                         /* turn down MacFUSE specific messages */
                         fuse_enosys (this, finh, msg);
                 else
-#endif
-                fuse_ops[finh->opcode] (this, finh, msg);
+                        fuse_ops[finh->opcode] (this, finh, msg);
 
                 iobuf_unref (iobuf);
                 continue;
@@ -3172,23 +4773,26 @@ fuse_thread_proc (void *data)
                 GF_FREE (iov_in[0].iov_base);
         }
 
-        iobuf_unref (iobuf);
-        GF_FREE (iov_in[0].iov_base);
+        /*
+         * We could be in all sorts of states with respect to iobuf and iov_in
+         * by the time we get here, and it's just not worth untangling them if
+         * we're about to kill ourselves anyway.
+         */
 
         if (dict_get (this->options, ZR_MOUNTPOINT_OPT))
                 mount_point = data_to_str (dict_get (this->options,
                                                      ZR_MOUNTPOINT_OPT));
         if (mount_point) {
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "unmounting %s", mount_point);
-                dict_del (this->options, ZR_MOUNTPOINT_OPT);
         }
 
+        /* Kill the whole process, not just this thread. */
         kill (getpid(), SIGTERM);
-
         return NULL;
 }
 
+
 int32_t
 fuse_itable_dump (xlator_t  *this)
 {
@@ -3198,7 +4802,7 @@ fuse_itable_dump (xlator_t  *this)
         gf_proc_dump_add_section("xlator.mount.fuse.itable");
         inode_table_dump(this->itable, "xlator.mount.fuse.itable");
 
-	return 0;
+        return 0;
 }
 
 int32_t
@@ -3216,33 +4820,76 @@ fuse_priv_dump (xlator_t  *this)
 
         gf_proc_dump_add_section("xlator.mount.fuse.priv");
 
-        gf_proc_dump_write("xlator.mount.fuse.priv.fd", "%d", private->fd);
-        gf_proc_dump_write("xlator.mount.fuse.priv.proto_minor", "%u",
+        gf_proc_dump_write("fd", "%d", private->fd);
+        gf_proc_dump_write("proto_minor", "%u",
                             private->proto_minor);
-        gf_proc_dump_write("xlator.mount.fuse.priv.volfile", "%s",
+        gf_proc_dump_write("volfile", "%s",
                             private->volfile?private->volfile:"None");
-        gf_proc_dump_write("xlator.mount.fuse.volfile_size", "%d",
+        gf_proc_dump_write("volfile_size", "%d",
                             private->volfile_size);
-        gf_proc_dump_write("xlator.mount.fuse.mount_point", "%s",
+        gf_proc_dump_write("mount_point", "%s",
                             private->mount_point);
-        gf_proc_dump_write("xlator.mount.fuse.iobuf", "%u",
+        gf_proc_dump_write("iobuf", "%u",
                             private->iobuf);
-        gf_proc_dump_write("xlator.mount.fuse.fuse_thread_started", "%d",
+        gf_proc_dump_write("fuse_thread_started", "%d",
                             (int)private->fuse_thread_started);
-        gf_proc_dump_write("xlator.mount.fuse.direct_io_mode", "%d",
+        gf_proc_dump_write("direct_io_mode", "%d",
                             private->direct_io_mode);
-        gf_proc_dump_write("xlator.mount.fuse.entry_timeout", "%lf",
+        gf_proc_dump_write("entry_timeout", "%lf",
                             private->entry_timeout);
-        gf_proc_dump_write("xlator.mount.fuse.attribute_timeout", "%lf",
+        gf_proc_dump_write("attribute_timeout", "%lf",
                             private->attribute_timeout);
-        gf_proc_dump_write("xlator.mount.fuse.init_recvd", "%d",
+        gf_proc_dump_write("init_recvd", "%d",
                             (int)private->init_recvd);
-        gf_proc_dump_write("xlator.mount.fuse.strict_volfile_check", "%d",
+        gf_proc_dump_write("strict_volfile_check", "%d",
                             (int)private->strict_volfile_check);
+        gf_proc_dump_write("reverse_thread_started", "%d",
+                           (int)private->reverse_fuse_thread_started);
+        gf_proc_dump_write("use_readdirp", "%d", private->use_readdirp);
 
         return 0;
 }
 
+int
+fuse_history_dump (xlator_t *this)
+{
+        int      ret    = -1;
+        char key_prefix[GF_DUMP_MAX_BUF_LEN] = {0,};
+
+        GF_VALIDATE_OR_GOTO ("fuse", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, this->history, out);
+
+        gf_proc_dump_build_key (key_prefix, "xlator.mount.fuse",
+                                "history");
+        gf_proc_dump_add_section (key_prefix);
+        eh_dump (this->history, NULL, dump_history_fuse);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+dump_history_fuse (circular_buffer_t *cb, void *data)
+{
+        char       *string   =  NULL;
+        struct tm  *tm       =  NULL;
+        char       timestr[256] = {0,};
+
+        string = (char *)cb->data;
+        tm = localtime (&cb->tv.tv_sec);
+
+        if (tm) {
+                strftime (timestr, 256, "%Y-%m-%d %H:%M:%S", tm);
+                snprintf (timestr + strlen (timestr), 256 - strlen (timestr),
+                          ".%"GF_PRI_SUSECONDS, cb->tv.tv_usec);
+                gf_proc_dump_write ("TIME", "%s", timestr);
+        }
+
+        gf_proc_dump_write ("message", "%s\n", string);
+
+        return 0;
+}
 
 int
 fuse_graph_setup (xlator_t *this, glusterfs_graph_t *graph)
@@ -3257,6 +4904,11 @@ fuse_graph_setup (xlator_t *this, glusterfs_graph_t *graph)
         if (priv->active_subvol == graph->top)
                 return 0; /* This is a valid case */
 
+        if (graph->used)
+                return 0;
+
+        graph->used = 1;
+
         itable = inode_table_new (0, graph->top);
         if (!itable)
                 return -1;
@@ -3266,12 +4918,15 @@ fuse_graph_setup (xlator_t *this, glusterfs_graph_t *graph)
         pthread_mutex_lock (&priv->sync_mutex);
         {
                 priv->next_graph = graph;
-                priv->child_up = 0;
+                priv->event_recvd = 0;
 
                 pthread_cond_signal (&priv->sync_cond);
         }
         pthread_mutex_unlock (&priv->sync_mutex);
 
+        gf_log ("fuse", GF_LOG_INFO, "switched to graph %d",
+                ((graph) ? graph->id : 0));
+
         return ret;
 }
 
@@ -3285,54 +4940,42 @@ notify (xlator_t *this, int32_t event, void *data, ...)
 
         private = this->private;
 
+        graph = data;
+
+        gf_log ("fuse", GF_LOG_DEBUG, "got event %d on graph %d",
+                event, ((graph) ? graph->id : 0));
+
         switch (event)
         {
         case GF_EVENT_GRAPH_NEW:
-                graph = data;
-                /* TODO: */
-
                 break;
 
+        case GF_EVENT_CHILD_UP:
+        case GF_EVENT_CHILD_DOWN:
         case GF_EVENT_CHILD_CONNECTING:
         {
-                if (!private->fuse_thread_started) {
-                        private->fuse_thread_started = 1;
-
-                        ret = pthread_create (&private->fuse_thread, NULL,
-                                              fuse_thread_proc, this);
-                        if (ret != 0) {
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "pthread_create() failed (%s)",
-                                        strerror (errno));
-                                break;
-                        }
-                }
-
-                if (data) {
-                        graph = data;
+                if (graph) {
                         ret = fuse_graph_setup (this, graph);
                         if (ret)
                                 gf_log (this->name, GF_LOG_WARNING,
                                         "failed to setup the graph");
                 }
 
-                break;
-        }
-        case GF_EVENT_CHILD_UP:
-        {
-                /* set priv->active_subvol */
-                /* set priv->first_lookup = 1 */
-                graph = data;
-                ret = fuse_graph_setup (this, graph);
-                if (ret)
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "failed to setup the graph");
+                if ((event == GF_EVENT_CHILD_UP)
+                    || (event == GF_EVENT_CHILD_DOWN)) {
+                        pthread_mutex_lock (&private->sync_mutex);
+                        {
+                                private->event_recvd = 1;
+                                pthread_cond_broadcast (&private->sync_cond);
+                        }
+                        pthread_mutex_unlock (&private->sync_mutex);
+                }
 
                 if (!private->fuse_thread_started) {
                         private->fuse_thread_started = 1;
 
-                        ret = pthread_create (&private->fuse_thread, NULL,
-                                              fuse_thread_proc, this);
+                        ret = gf_thread_create (&private->fuse_thread, NULL,
+						fuse_thread_proc, this);
                         if (ret != 0) {
                                 gf_log (this->name, GF_LOG_DEBUG,
                                         "pthread_create() failed (%s)",
@@ -3341,16 +4984,17 @@ notify (xlator_t *this, int32_t event, void *data, ...)
                         }
                 }
 
-                pthread_mutex_lock (&private->sync_mutex);
-                {
-                        private->child_up = 1;
-                        pthread_cond_broadcast (&private->sync_cond);
-                }
-                pthread_mutex_unlock (&private->sync_mutex);
-
                 break;
         }
 
+        case GF_EVENT_AUTH_FAILED:
+        {
+                /* Authentication failure is an error and glusterfs should stop */
+               gf_log (this->name, GF_LOG_ERROR, "Server authenication failed. Shutting down.");
+               fini (this);
+               break;
+        }
+
         default:
                 break;
         }
@@ -3379,45 +5023,52 @@ mem_acct_init (xlator_t *this)
 
 
 static fuse_handler_t *fuse_std_ops[FUSE_OP_HIGH] = {
-        [FUSE_INIT]        = fuse_init,
-        [FUSE_DESTROY]     = fuse_destroy,
         [FUSE_LOOKUP]      = fuse_lookup,
         [FUSE_FORGET]      = fuse_forget,
         [FUSE_GETATTR]     = fuse_getattr,
         [FUSE_SETATTR]     = fuse_setattr,
-        [FUSE_OPENDIR]     = fuse_opendir,
-        [FUSE_READDIR]     = fuse_readdir,
-        [FUSE_RELEASEDIR]  = fuse_releasedir,
-        [FUSE_ACCESS]      = fuse_access,
         [FUSE_READLINK]    = fuse_readlink,
+        [FUSE_SYMLINK]     = fuse_symlink,
         [FUSE_MKNOD]       = fuse_mknod,
         [FUSE_MKDIR]       = fuse_mkdir,
         [FUSE_UNLINK]      = fuse_unlink,
         [FUSE_RMDIR]       = fuse_rmdir,
-        [FUSE_SYMLINK]     = fuse_symlink,
         [FUSE_RENAME]      = fuse_rename,
         [FUSE_LINK]        = fuse_link,
-        [FUSE_CREATE]      = fuse_create,
         [FUSE_OPEN]        = fuse_open,
         [FUSE_READ]        = fuse_readv,
         [FUSE_WRITE]       = fuse_write,
-        [FUSE_FLUSH]       = fuse_flush,
+        [FUSE_STATFS]      = fuse_statfs,
         [FUSE_RELEASE]     = fuse_release,
         [FUSE_FSYNC]       = fuse_fsync,
-        [FUSE_FSYNCDIR]    = fuse_fsyncdir,
-        [FUSE_STATFS]      = fuse_statfs,
         [FUSE_SETXATTR]    = fuse_setxattr,
         [FUSE_GETXATTR]    = fuse_getxattr,
         [FUSE_LISTXATTR]   = fuse_listxattr,
         [FUSE_REMOVEXATTR] = fuse_removexattr,
+        [FUSE_FLUSH]       = fuse_flush,
+        [FUSE_INIT]        = fuse_init,
+        [FUSE_OPENDIR]     = fuse_opendir,
+        [FUSE_READDIR]     = fuse_readdir,
+        [FUSE_RELEASEDIR]  = fuse_releasedir,
+        [FUSE_FSYNCDIR]    = fuse_fsyncdir,
         [FUSE_GETLK]       = fuse_getlk,
         [FUSE_SETLK]       = fuse_setlk,
         [FUSE_SETLKW]      = fuse_setlk,
+        [FUSE_ACCESS]      = fuse_access,
+        [FUSE_CREATE]      = fuse_create,
+     /* [FUSE_INTERRUPT] */
+     /* [FUSE_BMAP] */
+        [FUSE_DESTROY]     = fuse_destroy,
+     /* [FUSE_IOCTL] */
+     /* [FUSE_POLL] */
+     /* [FUSE_NOTIFY_REPLY] */
+	[FUSE_BATCH_FORGET]= fuse_batch_forget,
+	[FUSE_FALLOCATE]   = fuse_fallocate,
+	[FUSE_READDIRPLUS] = fuse_readdirp,
 };
 
 
-static fuse_handler_t *fuse_dump_ops[FUSE_OP_HIGH] = {
-};
+static fuse_handler_t *fuse_dump_ops[FUSE_OP_HIGH];
 
 
 static void
@@ -3445,7 +5096,7 @@ fuse_dumper (xlator_t *this, fuse_in_header_t *finh, void *msg)
                         "failed to dump fuse message (R): %s",
                         strerror (errno));
 
-        return priv->fuse_ops0[finh->opcode] (this, finh, msg);
+        priv->fuse_ops0[finh->opcode] (this, finh, msg);
 }
 
 
@@ -3462,6 +5113,12 @@ init (xlator_t *this_xl)
         int                i = 0;
         int                xl_name_allocated = 0;
         int                fsname_allocated = 0;
+        glusterfs_ctx_t   *ctx = NULL;
+        gf_boolean_t       sync_to_mount = _gf_false;
+        gf_boolean_t       fopen_keep_cache = _gf_false;
+        unsigned long      mntflags = 0;
+        char              *mnt_args = NULL;
+        eh_t              *event = NULL;
 
         if (this_xl == NULL)
                 return -1;
@@ -3469,6 +5126,10 @@ init (xlator_t *this_xl)
         if (this_xl->options == NULL)
                 return -1;
 
+        ctx = this_xl->ctx;
+        if (!ctx)
+                return -1;
+
         options = this_xl->options;
 
         if (this_xl->name == NULL) {
@@ -3492,6 +5153,8 @@ init (xlator_t *this_xl)
         this_xl->private = (void *) priv;
         priv->mount_point = NULL;
         priv->fd = -1;
+        priv->revchan_in = -1;
+        priv->revchan_out = -1;
 
         /* get options from option dictionary */
         ret = dict_get_str (options, ZR_MOUNTPOINT_OPT, &value_string);
@@ -3534,31 +5197,44 @@ init (xlator_t *this_xl)
                 goto cleanup_exit;
         }
 
-        ret = dict_get_double (options, "attribute-timeout",
-                               &priv->attribute_timeout);
-        if (ret != 0)
-                priv->attribute_timeout = 1.0; /* default */
+        GF_OPTION_INIT ("attribute-timeout", priv->attribute_timeout, double,
+                        cleanup_exit);
+
+        GF_OPTION_INIT ("entry-timeout", priv->entry_timeout, double,
+                        cleanup_exit);
 
-        ret = dict_get_double (options, "entry-timeout",
-                               &priv->entry_timeout);
-        if (ret != 0)
-                priv->entry_timeout = 1.0; /* default */
+        GF_OPTION_INIT ("negative-timeout", priv->negative_timeout, double,
+                        cleanup_exit);
 
+        GF_OPTION_INIT ("client-pid", priv->client_pid, int32, cleanup_exit);
+        /* have to check & register the presence of client-pid manually */
+        priv->client_pid_set = !!dict_get (this_xl->options, "client-pid");
+
+        GF_OPTION_INIT ("uid-map-root", priv->uid_map_root, uint32,
+                        cleanup_exit);
 
         priv->direct_io_mode = 2;
         ret = dict_get_str (options, ZR_DIRECT_IO_OPT, &value_string);
         if (ret == 0) {
                 ret = gf_string2boolean (value_string, &priv->direct_io_mode);
-                assert (ret == 0);
+                GF_ASSERT (ret == 0);
         }
 
-        priv->strict_volfile_check = 0;
-        ret = dict_get_str (options, ZR_STRICT_VOLFILE_CHECK, &value_string);
-        if (ret == 0) {
-                ret = gf_string2boolean (value_string,
-                                         &priv->strict_volfile_check);
-                assert (ret == 0);
-        }
+        GF_OPTION_INIT (ZR_STRICT_VOLFILE_CHECK, priv->strict_volfile_check,
+                        bool, cleanup_exit);
+
+        GF_OPTION_INIT ("acl", priv->acl, bool, cleanup_exit);
+
+        if (priv->uid_map_root)
+                priv->acl = 1;
+
+        GF_OPTION_INIT ("selinux", priv->selinux, bool, cleanup_exit);
+
+        GF_OPTION_INIT ("read-only", priv->read_only, bool, cleanup_exit);
+
+        GF_OPTION_INIT ("enable-ino32", priv->enable_ino32, bool, cleanup_exit);
+
+        GF_OPTION_INIT ("use-readdirp", priv->use_readdirp, bool, cleanup_exit);
 
         priv->fuse_dump_fd = -1;
         ret = dict_get_str (options, "dump-fuse", &value_string);
@@ -3577,6 +5253,58 @@ init (xlator_t *this_xl)
                 priv->fuse_dump_fd = ret;
         }
 
+        sync_to_mount = _gf_false;
+        ret = dict_get_str (options, "sync-to-mount", &value_string);
+        if (ret == 0) {
+                ret = gf_string2boolean (value_string,
+                                         &sync_to_mount);
+                GF_ASSERT (ret == 0);
+        }
+
+	priv->fopen_keep_cache = 2;
+	if (dict_get (options, "fopen-keep-cache")) {
+		GF_OPTION_INIT("fopen-keep-cache", fopen_keep_cache, bool,
+			       cleanup_exit);
+		priv->fopen_keep_cache = fopen_keep_cache;
+	}
+
+        GF_OPTION_INIT("gid-timeout", priv->gid_cache_timeout, int32,
+                cleanup_exit);
+
+        GF_OPTION_INIT ("fuse-mountopts", priv->fuse_mountopts, str, cleanup_exit);
+
+        if (gid_cache_init(&priv->gid_cache, priv->gid_cache_timeout) < 0) {
+                gf_log("glusterfs-fuse", GF_LOG_ERROR, "Failed to initialize "
+                        "group cache.");
+                goto cleanup_exit;
+        }
+
+        /* default values seemed to work fine during testing */
+        GF_OPTION_INIT ("background-qlen", priv->background_qlen, int32,
+                        cleanup_exit);
+        GF_OPTION_INIT ("congestion-threshold", priv->congestion_threshold,
+                        int32, cleanup_exit);
+
+        /* user has set only background-qlen, not congestion-threshold,
+           use the fuse kernel driver formula to set congestion. ie, 75% */
+        if (dict_get (this_xl->options, "background-qlen") &&
+            !dict_get (this_xl->options, "congestion-threshold")) {
+                priv->congestion_threshold = (priv->background_qlen * 3) / 4;
+                gf_log (this_xl->name, GF_LOG_INFO,
+                        "setting congestion control as 75%% of "
+                        "background-queue length (ie, (.75 * %d) = %d",
+                        priv->background_qlen, priv->congestion_threshold);
+        }
+
+        /* congestion should not be higher than background queue length */
+        if (priv->congestion_threshold > priv->background_qlen) {
+                gf_log (this_xl->name, GF_LOG_INFO,
+                        "setting congestion control same as "
+                        "background-queue length (%d)",
+                        priv->background_qlen);
+                priv->congestion_threshold = priv->background_qlen;
+        }
+
         cmd_args = &this_xl->ctx->cmd_args;
         fsname = cmd_args->volfile;
         if (!fsname && cmd_args->volfile_server) {
@@ -3600,17 +5328,46 @@ init (xlator_t *this_xl)
         if (!fsname)
                 fsname = "glusterfs";
 
+        priv->fdtable = gf_fd_fdtable_alloc ();
+        if (priv->fdtable == NULL) {
+                gf_log ("glusterfs-fuse", GF_LOG_ERROR, "Out of memory");
+                goto cleanup_exit;
+        }
+
+        if (priv->read_only)
+                mntflags |= MS_RDONLY;
+        gf_asprintf (&mnt_args, "%s%s%sallow_other,max_read=131072",
+                     priv->acl ? "" : "default_permissions,",
+                     priv->fuse_mountopts ? priv->fuse_mountopts : "",
+                     priv->fuse_mountopts ? "," : "");
+        if (!mnt_args)
+                goto cleanup_exit;
+
+        if (pipe(priv->status_pipe) < 0) {
+                gf_log (this_xl->name, GF_LOG_ERROR,
+                        "could not create pipe to separate mount process");
+                goto cleanup_exit;
+        }
 
-        priv->fd = gf_fuse_mount (priv->mount_point, fsname,
-                                  "allow_other,default_permissions,"
-                                  "max_read=131072");
+        priv->fd = gf_fuse_mount (priv->mount_point, fsname, mntflags, mnt_args,
+                                  sync_to_mount ? &ctx->mnt_pid : NULL,
+                                  priv->status_pipe[1]);
         if (priv->fd == -1)
                 goto cleanup_exit;
 
+        event = eh_new (FUSE_EVENT_HISTORY_SIZE, _gf_false, NULL);
+        if (!event) {
+                gf_log (this_xl->name, GF_LOG_ERROR,
+                        "could not create a new event history");
+                goto cleanup_exit;
+        }
+
+        this_xl->history = event;
+
         pthread_mutex_init (&priv->fuse_dump_mutex, NULL);
         pthread_cond_init (&priv->sync_cond, NULL);
         pthread_mutex_init (&priv->sync_mutex, NULL);
-        priv->child_up = 0;
+        priv->event_recvd = 0;
 
         for (i = 0; i < FUSE_OP_HIGH; i++) {
                 if (!fuse_std_ops[i])
@@ -3624,6 +5381,9 @@ init (xlator_t *this_xl)
                 priv->fuse_ops  = fuse_dump_ops;
         }
 
+        if (fsname_allocated)
+                GF_FREE (fsname);
+        GF_FREE (mnt_args);
         return 0;
 
 cleanup_exit:
@@ -3633,10 +5393,13 @@ cleanup_exit:
                 GF_FREE (fsname);
         if (priv) {
                 GF_FREE (priv->mount_point);
-                close (priv->fd);
-                close (priv->fuse_dump_fd);
+                if (priv->fd != -1)
+                        close (priv->fd);
+                if (priv->fuse_dump_fd != -1)
+                        close (priv->fuse_dump_fd);
                 GF_FREE (priv);
         }
+        GF_FREE (mnt_args);
         return -1;
 }
 
@@ -3657,25 +5420,31 @@ fini (xlator_t *this_xl)
                 mount_point = data_to_str (dict_get (this_xl->options,
                                                      ZR_MOUNTPOINT_OPT));
         if (mount_point != NULL) {
-                gf_log (this_xl->name, GF_LOG_NORMAL,
+                gf_log (this_xl->name, GF_LOG_INFO,
                         "Unmounting '%s'.", mount_point);
 
-                dict_del (this_xl->options, ZR_MOUNTPOINT_OPT);
                 gf_fuse_unmount (mount_point, priv->fd);
                 close (priv->fuse_dump_fd);
+                dict_del (this_xl->options, ZR_MOUNTPOINT_OPT);
         }
+        /* Process should terminate once fuse xlator is finished.
+         * Required for AUTH_FAILED event.
+         */
+        kill (getpid (), SIGTERM);
 }
 
-struct xlator_fops fops = {
-};
+struct xlator_fops fops;
 
 struct xlator_cbks cbks = {
+        .invalidate = fuse_invalidate,
+        .forget     = fuse_forget_cbk,
 };
 
 
 struct xlator_dumpops dumpops = {
         .priv  = fuse_priv_dump,
         .inode = fuse_itable_dump,
+        .history = fuse_history_dump,
 };
 
 struct volume_options options[] = {
@@ -3689,13 +5458,71 @@ struct volume_options options[] = {
           .type = GF_OPTION_TYPE_PATH
         },
         { .key  = {ZR_ATTR_TIMEOUT_OPT},
-          .type = GF_OPTION_TYPE_DOUBLE
+          .type = GF_OPTION_TYPE_DOUBLE,
+          .default_value = "1.0"
         },
         { .key  = {ZR_ENTRY_TIMEOUT_OPT},
-          .type = GF_OPTION_TYPE_DOUBLE
+          .type = GF_OPTION_TYPE_DOUBLE,
+          .default_value = "1.0"
+        },
+        { .key  = {ZR_NEGATIVE_TIMEOUT_OPT},
+          .type = GF_OPTION_TYPE_DOUBLE,
+          .default_value = "0.0"
         },
         { .key  = {ZR_STRICT_VOLFILE_CHECK},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "false"
+        },
+        { .key  = {"client-pid"},
+          .type = GF_OPTION_TYPE_INT
+        },
+        { .key  = {"uid-map-root"},
+          .type = GF_OPTION_TYPE_INT
+        },
+        { .key  = {"sync-to-mount"},
+          .type = GF_OPTION_TYPE_BOOL
+        },
+        { .key = {"read-only"},
           .type = GF_OPTION_TYPE_BOOL
         },
+        { .key = {"fopen-keep-cache"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "false"
+        },
+        { .key = {"gid-timeout"},
+          .type = GF_OPTION_TYPE_INT,
+          .default_value = "2"
+        },
+        { .key = {"acl"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "false"
+        },
+        { .key = {"selinux"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "false"
+        },
+        { .key = {"enable-ino32"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "false"
+        },
+        { .key  = {"background-qlen"},
+          .type = GF_OPTION_TYPE_INT,
+          .default_value = "64",
+          .min = 16,
+          .max = (64 * GF_UNIT_KB),
+        },
+        { .key  = {"congestion-threshold"},
+          .type = GF_OPTION_TYPE_INT,
+          .default_value = "48",
+          .min = 12,
+          .max = (64 * GF_UNIT_KB),
+        },
+        { .key = {"fuse-mountopts"},
+          .type = GF_OPTION_TYPE_STR
+        },
+        { .key = {"use-readdirp"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "yes"
+        },
         { .key = {NULL} },
 };
diff --git a/xlators/mount/fuse/src/fuse-bridge.h b/xlators/mount/fuse/src/fuse-bridge.h
index 53a2cc598..34794b6ea 100644
--- a/xlators/mount/fuse/src/fuse-bridge.h
+++ b/xlators/mount/fuse/src/fuse-bridge.h
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef _GF_FUSE_BRIDGE_H_
 #define _GF_FUSE_BRIDGE_H_
 
@@ -39,7 +29,6 @@
 #include "xlator.h"
 #include "defaults.h"
 #include "common-utils.h"
-#include "glusterfsd-common.h"
 #include "statedump.h"
 
 #ifdef GF_DARWIN_HOST_OS
@@ -55,12 +44,11 @@
 
 #include "list.h"
 #include "dict.h"
+#include "syncop.h"
+#include "gidcache.h"
 
-/* TODO: when supporting posix acl, remove this definition */
-#define DISABLE_POSIX_ACL
-
-#ifdef GF_LINUX_HOST_OS
-#define FUSE_OP_HIGH (FUSE_POLL + 1)
+#if defined(GF_LINUX_HOST_OS) || defined(__NetBSD__)
+#define FUSE_OP_HIGH (FUSE_READDIRPLUS + 1)
 #endif
 #ifdef GF_DARWIN_HOST_OS
 #define FUSE_OP_HIGH (FUSE_DESTROY + 1)
@@ -88,11 +76,12 @@ struct fuse_private {
         size_t              *msg0_len_p;
 
         double               entry_timeout;
+        double               negative_timeout;
         double               attribute_timeout;
 
         pthread_cond_t       sync_cond;
         pthread_mutex_t      sync_mutex;
-        char                 child_up;
+        char                 event_recvd;
 
         char                 init_recvd;
 
@@ -105,17 +94,61 @@ struct fuse_private {
 
         glusterfs_graph_t   *next_graph;
         xlator_t            *active_subvol;
+
+        pid_t                client_pid;
+        gf_boolean_t         client_pid_set;
+        unsigned             uid_map_root;
+        gf_boolean_t         acl;
+        gf_boolean_t         selinux;
+        gf_boolean_t         read_only;
+	int32_t	             fopen_keep_cache;
+	int32_t		     gid_cache_timeout;
+        gf_boolean_t         enable_ino32;
+        fdtable_t           *fdtable;
+	gid_cache_t	     gid_cache;
+        char                *fuse_mountopts;
+
+        /* For fuse-reverse-validation */
+        int                  revchan_in;
+        int                  revchan_out;
+        gf_boolean_t         reverse_fuse_thread_started;
+
+        /* For communicating with separate mount thread. */
+        int                  status_pipe[2];
+
+        /* for fuse queue length and congestion threshold */
+        int background_qlen;
+        int congestion_threshold;
+
+        /* for using fuse-kernel readdirp*/
+        gf_boolean_t use_readdirp;
 };
 typedef struct fuse_private fuse_private_t;
 
+struct fuse_graph_switch_args {
+        xlator_t        *this;
+        xlator_t        *old_subvol;
+        xlator_t        *new_subvol;
+};
+typedef struct fuse_graph_switch_args fuse_graph_switch_args_t;
+
+#define INVAL_BUF_SIZE (sizeof (struct fuse_out_header) +               \
+                        max (sizeof (struct fuse_notify_inval_inode_out), \
+                             sizeof (struct fuse_notify_inval_entry_out) + \
+                             NAME_MAX + 1))
+
+#define FUSE_EVENT_HISTORY_SIZE 1024
+
 #define _FH_TO_FD(fh) ((fd_t *)(uintptr_t)(fh))
 
 #define FH_TO_FD(fh) ((_FH_TO_FD (fh))?(fd_ref (_FH_TO_FD (fh))):((fd_t *) 0))
 
 #define FUSE_FOP(state, ret, op_num, fop, args ...)                     \
         do {                                                            \
-                call_frame_t *frame = NULL;                             \
-                xlator_t *xl = NULL;                                    \
+                call_frame_t   *frame   = NULL;                         \
+                xlator_t       *xl      = NULL;                         \
+                int32_t         op_ret  = 0, op_errno = 0;              \
+                fuse_resolve_t *resolve = NULL;                         \
                                                                         \
                 frame = get_call_frame_for_req (state);                 \
                 if (!frame) {                                           \
@@ -125,7 +158,7 @@ typedef struct fuse_private fuse_private_t;
                           * better than trying to go on with a NULL     \
                           * frame ...                                   \
                           */                                            \
-                        gf_log ("glusterfs-fuse",                       \
+                        gf_log_callingfn ("glusterfs-fuse",             \
                                 GF_LOG_ERROR,                           \
                                 "FUSE message"                          \
                                 " unique %"PRIu64" opcode %d:"          \
@@ -133,29 +166,86 @@ typedef struct fuse_private fuse_private_t;
                                 state->finh->unique,                    \
                                 state->finh->opcode);                   \
                         free_fuse_state (state);                        \
-                        return;                                         \
+                        /* ideally, need to 'return', but let the */    \
+                        /* calling function take care of it */          \
+                        break;                                          \
                 }                                                       \
                                                                         \
                 frame->root->state = state;                             \
                 frame->root->op    = op_num;                            \
                 frame->op          = op_num;                            \
                                                                         \
-                xl = fuse_state_subvol (state);                         \
+                if ( state->resolve_now ) {                             \
+                        resolve = state->resolve_now;                   \
+                } else {                                                \
+                        resolve = &(state->resolve);                    \
+                }                                                       \
+                                                                        \
+                xl = state->active_subvol;				\
                 if (!xl) {                                              \
-                        gf_log ("glusterfs-fuse", GF_LOG_ERROR,         \
-                                "xl is NULL");                          \
-                        send_fuse_err (state->this, state->finh, ENOENT); \
+                        gf_log_callingfn ("glusterfs-fuse", GF_LOG_ERROR, \
+                                          "xl is NULL");                \
+                        op_errno = ENOENT;                              \
+                        op_ret = -1;                                    \
+                } else if (resolve->op_ret < 0) {                       \
+                        op_errno = resolve->op_errno;                   \
+                        op_ret = -1;                                    \
+                        if (op_num == GF_FOP_LOOKUP) {                  \
+                                gf_log ("glusterfs-fuse",               \
+                                        (op_errno == ENOENT ? GF_LOG_TRACE \
+                                         : GF_LOG_WARNING),             \
+                                        "%"PRIu64": %s() %s => -1 (%s)", \
+                                        frame->root->unique,            \
+                                        gf_fop_list[frame->root->op],   \
+                                        resolve->resolve_loc.path,      \
+                                        strerror (op_errno));           \
+                        } else {                                        \
+                                gf_log ("glusterfs-fuse",               \
+                                        GF_LOG_WARNING,                 \
+                                        "%"PRIu64": %s() inode "        \
+                                        "migration of %s failed (%s)",  \
+                                        frame->root->unique,            \
+                                        gf_fop_list[frame->root->op],   \
+                                        resolve->resolve_loc.path,      \
+                                        strerror (op_errno));           \
+                        }                                               \
+                } else if (state->resolve2.op_ret < 0) {                \
+                        op_errno = state->resolve2.op_errno;            \
+                        op_ret = -1;                                    \
+                        gf_log ("glusterfs-fuse",                       \
+                                GF_LOG_WARNING,                         \
+                                "%"PRIu64": %s() inode "                \
+                                "migration of %s failed (%s)",          \
+                                frame->root->unique,                    \
+                                gf_fop_list[frame->root->op],           \
+                                state->resolve2.resolve_loc.path,       \
+                                strerror (op_errno));                   \
+                }                                                       \
+                                                                        \
+                if (op_ret < 0) {                                       \
+                        send_fuse_err (state->this, state->finh, op_errno); \
                         free_fuse_state (state);                        \
                         STACK_DESTROY (frame->root);                    \
                 } else {                                                \
+                        if (state->this->history)                       \
+                                gf_log_eh ("%"PRIu64", %s, path: (%s), gfid: " \
+                                           "(%s)", frame->root->unique, \
+                                           gf_fop_list[frame->root->op], \
+                                           state->loc.path,             \
+                                           (state->fd == NULL)?         \
+                                           uuid_utoa (state->loc.gfid): \
+                                           uuid_utoa (state->fd->inode->gfid));\
                         STACK_WIND (frame, ret, xl, xl->fops->fop, args); \
                 }                                                       \
+                                                                        \
         } while (0)
 
 
 #define FUSE_FOP_COOKIE(state, xl, ret, cky, op_num, fop, args ...)     \
         do {                                                            \
-                call_frame_t *frame = NULL;                             \
+                call_frame_t *frame  = NULL;                            \
+                xlator_t     *xl     = NULL;                            \
+                int32_t       op_ret = 0, op_errno = 0;                 \
                                                                         \
                 frame = get_call_frame_for_req (state);                 \
                 if (!frame) {                                           \
@@ -173,7 +263,62 @@ typedef struct fuse_private fuse_private_t;
                 frame->root->state = state;                             \
                 frame->root->op    = op_num;                            \
 		frame->op          = op_num;				\
-                STACK_WIND_COOKIE (frame, ret, cky, xl, xl->fops->fop, args); \
+                                                                        \
+                xl = state->active_subvol;				\
+                if (!xl) {                                              \
+                        gf_log_callingfn ("glusterfs-fuse", GF_LOG_ERROR, \
+                                          "xl is NULL");                \
+                        op_errno = ENOENT;                              \
+                        op_ret = -1;                                    \
+                } else if (state->resolve.op_ret < 0) {                 \
+                        op_errno = state->resolve.op_errno;             \
+                        op_ret = -1;                                    \
+                        if (op_num == GF_FOP_LOOKUP) {                  \
+                                gf_log ("glusterfs-fuse",               \
+                                        (op_errno == ENOENT ? GF_LOG_TRACE \
+                                         : GF_LOG_WARNING),             \
+                                        "%"PRIu64": %s() %s => -1 (%s)", \
+                                        frame->root->unique,            \
+                                        gf_fop_list[frame->root->op],   \
+                                        state->resolve.resolve_loc.path, \
+                                        strerror (op_errno));           \
+                        } else {                                        \
+                                gf_log ("glusterfs-fuse",               \
+                                        GF_LOG_WARNING,                 \
+                                        "%"PRIu64": %s() inode "        \
+                                        "migration of %s failed (%s)",  \
+                                        frame->root->unique,            \
+                                        gf_fop_list[frame->root->op],   \
+                                        state->resolve.resolve_loc.path, \
+                                        strerror (op_errno));           \
+                        }                                               \
+                } else if (state->resolve2.op_ret < 0) {                \
+                        op_errno = state->resolve2.op_errno;            \
+                        op_ret = -1;                                    \
+                        gf_log ("glusterfs-fuse",                       \
+                                GF_LOG_WARNING,                         \
+                                "%"PRIu64": %s() inode "                \
+                                "migration of %s failed (%s)",          \
+                                frame->root->unique,                    \
+                                gf_fop_list[frame->root->op],           \
+                                state->resolve2.resolve_loc.path,       \
+                                strerror (op_errno));                   \
+                }                                                       \
+                                                                        \
+                if (op_ret < 0) {                                       \
+                        send_fuse_err (state->this, state->finh, op_errno); \
+                        free_fuse_state (state);                        \
+                        STACK_DESTROY (frame->root);                    \
+                } else {                                                \
+                        if (xl->history)                                \
+                                gf_log_eh ("%"PRIu64", %s, path: (%s), gfid: " \
+                                           "(%s)", frame->root->unique, \
+                                           gf_fop_list[frame->root->op], \
+                                           state->loc.path,             \
+                                           uuid_utoa (state->loc.gfid)); \
+                        STACK_WIND_COOKIE (frame, ret, cky, xl, xl->fops->fop, \
+                                           args);                       \
+                }                                                       \
         } while (0)
 
 #define GF_SELECT_LOG_LEVEL(_errno)                     \
@@ -197,6 +342,90 @@ typedef struct fuse_private fuse_private_t;
                 }                                                          \
         } while (0)
 
+#define FUSE_ENTRY_CREATE(this, priv, finh, state, fci, op)             \
+                do {                                                    \
+                        if (priv->proto_minor >= 12)                    \
+                                state->mode &= ~fci->umask;             \
+                        if (priv->proto_minor >= 12 && priv->acl) {     \
+                                state->xdata = dict_new ();             \
+                                if (!state->xdata) {                    \
+                                        gf_log ("glusterfs-fuse",       \
+                                                GF_LOG_WARNING,         \
+                                                "%s failed to allocate " \
+                                                "a param dictionary", op); \
+                                        send_fuse_err (this, finh, ENOMEM); \
+                                        free_fuse_state (state);        \
+                                        return;                         \
+                                }                                       \
+                                state->umask = fci->umask;              \
+                                                                        \
+/* TODO: remove this after 3.4.0 release. keeping it for the            \
+   sake of backward compatibility with old (3.3.[01])                   \
+   releases till then. */                                               \
+                                ret = dict_set_int16 (state->xdata, "umask", \
+                                                      fci->umask);      \
+                                if (ret < 0) {                          \
+                                        gf_log ("glusterfs-fuse",       \
+                                                GF_LOG_WARNING,         \
+                                                "%s Failed adding umask"\
+                                                " to request", op);     \
+                                        dict_destroy (state->xdata);    \
+                                        send_fuse_err (this, finh, ENOMEM); \
+                                        free_fuse_state (state);        \
+                                        return;                         \
+                                }                                       \
+                                ret = dict_set_int16 (state->xdata, "mode", \
+                                                      fci->mode);       \
+                                if (ret < 0) {                          \
+                                        gf_log ("glusterfs-fuse",       \
+                                                GF_LOG_WARNING,         \
+                                                "%s Failed adding mode " \
+                                                "to request", op);         \
+                                        dict_destroy (state->xdata);    \
+                                        send_fuse_err (this, finh, ENOMEM); \
+                                        free_fuse_state (state);        \
+                                        return;                         \
+                                }                                       \
+                        }                                               \
+                } while (0)
+
+#define fuse_log_eh_fop(this, state, frame, op_ret, op_errno)               \
+        do {                                                            \
+                if (this->history) {                                    \
+                        if (state->fd)                                  \
+                                gf_log_eh ("op_ret: %d, op_errno: %d, " \
+                                           "%"PRIu64", %s () => %p, gfid: %s", \
+                                           op_ret, op_errno,            \
+                                           frame->root->unique,         \
+                                           gf_fop_list[frame->root->op], \
+                                           state->fd,                   \
+                                           uuid_utoa (state->fd->inode->gfid)); \
+                        else                                            \
+                                gf_log_eh ("op_ret: %d, op_errno: %d, " \
+                                           "%"PRIu64", %s () => %s, gfid: %s", \
+                                           op_ret, op_errno,            \
+                                           frame->root->unique,         \
+                                           gf_fop_list[frame->root->op], \
+                                           state->loc.path,             \
+                                           uuid_utoa (state->loc.gfid)); \
+                }                                                       \
+        } while(0)
+
+#define fuse_log_eh(this, args...)              \
+        do {                                    \
+                if (this->history)              \
+                        gf_log_eh(args);        \
+        } while (0)
+
+static inline xlator_t *
+fuse_active_subvol (xlator_t *fuse)
+{
+        fuse_private_t *priv = NULL;
+
+        priv = fuse->private;
+
+        return priv->active_subvol;
+}
 
 
 typedef enum {
@@ -205,35 +434,29 @@ typedef enum {
         RESOLVE_MAY,
         RESOLVE_DONTCARE,
         RESOLVE_EXACT
-} gf_resolve_type_t;
+} fuse_resolve_type_t;
 
-struct gf_resolve_comp {
-        char      *basename;
-        ino_t      ino;
-        uint64_t   gen;
-        inode_t   *inode;
-};
 
 typedef struct {
-        gf_resolve_type_t      type;
-        ino_t                  ino;
-        uint64_t               gen;
-        ino_t                  par;
+        fuse_resolve_type_t    type;
         fd_t                  *fd;
         char                  *path;
         char                  *bname;
+        u_char                 gfid[16];
+	inode_t               *hint;
+        u_char                 pargfid[16];
+	inode_t               *parhint;
 	char                  *resolved;
         int                    op_ret;
         int                    op_errno;
-        loc_t                  deep_loc;
-        struct gf_resolve_comp *components;
-        int                    comp_count;
-} gf_resolve_t;
+        loc_t                  resolve_loc;
+} fuse_resolve_t;
 
 
 typedef struct {
         void             *pool;
         xlator_t         *this;
+	xlator_t         *active_subvol;
         inode_table_t    *itable;
         loc_t             loc;
         loc_t             loc2;
@@ -243,7 +466,8 @@ typedef struct {
         size_t            size;
         unsigned long     nlookup;
         fd_t             *fd;
-        dict_t           *dict;
+        dict_t           *xattr;
+        dict_t           *xdata;
         char             *name;
         char              is_revalidate;
         gf_boolean_t      truncate_needed;
@@ -252,11 +476,11 @@ typedef struct {
 
         /* used within resolve_and_resume */
         /* */
-        gf_resolve_t resolve;
-        gf_resolve_t resolve2;
+        fuse_resolve_t resolve;
+        fuse_resolve_t resolve2;
 
         loc_t        *loc_now;
-        gf_resolve_t *resolve_now;
+        fuse_resolve_t *resolve_now;
 
         void *resume_fn;
 
@@ -264,13 +488,22 @@ typedef struct {
         int            mask;
         dev_t          rdev;
         mode_t         mode;
+        mode_t         umask;
         struct iatt    attr;
-        struct flock   lk_lock;
+        struct gf_flock   lk_lock;
         struct iovec   vector;
 
         uuid_t         gfid;
+        uint32_t       io_flags;
+        int32_t        fd_no;
 } fuse_state_t;
 
+typedef struct {
+        uint32_t  open_flags;
+        char      migration_failed;
+        fd_t     *activefd;
+} fuse_fd_ctx_t;
+
 typedef void (*fuse_resume_fn_t) (fuse_state_t *state);
 
 GF_MUST_CHECK int32_t
@@ -279,13 +512,26 @@ fuse_loc_fill (loc_t *loc, fuse_state_t *state, ino_t ino,
 call_frame_t *get_call_frame_for_req (fuse_state_t *state);
 fuse_state_t *get_fuse_state (xlator_t *this, fuse_in_header_t *finh);
 void free_fuse_state (fuse_state_t *state);
-void gf_fuse_stat2attr (struct iatt *st, struct fuse_attr *fa);
+void gf_fuse_stat2attr (struct iatt *st, struct fuse_attr *fa,
+                        gf_boolean_t enable_ino32);
+void gf_fuse_fill_dirent (gf_dirent_t *entry, struct fuse_dirent *fde,
+                          gf_boolean_t enable_ino32);
 uint64_t inode_to_fuse_nodeid (inode_t *inode);
-xlator_t *fuse_state_subvol (fuse_state_t *state);
 xlator_t *fuse_active_subvol (xlator_t *fuse);
 inode_t *fuse_ino_to_inode (uint64_t ino, xlator_t *fuse);
-int fuse_resolve_and_resume (fuse_state_t *state, fuse_resume_fn_t fn);
-int is_gf_log_command (xlator_t *this, const char *name, char *value);
 int send_fuse_err (xlator_t *this, fuse_in_header_t *finh, int error);
 int fuse_gfid_set (fuse_state_t *state);
+int fuse_flip_xattr_ns (struct fuse_private *priv, char *okey, char **nkey);
+fuse_fd_ctx_t * __fuse_fd_ctx_check_n_create (xlator_t *this, fd_t *fd);
+fuse_fd_ctx_t * fuse_fd_ctx_check_n_create (xlator_t *this, fd_t *fd);
+
+int fuse_resolve_and_resume (fuse_state_t *state, fuse_resume_fn_t fn);
+int fuse_resolve_inode_init (fuse_state_t *state, fuse_resolve_t *resolve,
+			     ino_t ino);
+int fuse_resolve_entry_init (fuse_state_t *state, fuse_resolve_t *resolve,
+			     ino_t par, char *name);
+int fuse_resolve_fd_init (fuse_state_t *state, fuse_resolve_t *resolve,
+			  fd_t *fd);
+int fuse_ignore_xattr_set (fuse_private_t *priv, char *key);
+int dump_history_fuse (circular_buffer_t *cb, void *data);
 #endif /* _GF_FUSE_BRIDGE_H_ */
diff --git a/xlators/mount/fuse/src/fuse-helpers.c b/xlators/mount/fuse/src/fuse-helpers.c
index 9a6b13514..4d478b919 100644
--- a/xlators/mount/fuse/src/fuse-helpers.c
+++ b/xlators/mount/fuse/src/fuse-helpers.c
@@ -1,96 +1,73 @@
 /*
-   Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
+#ifdef __NetBSD__
+#define _KMEMUSER
+#endif
 
 #include "fuse-bridge.h"
-
-xlator_t *
-fuse_state_subvol (fuse_state_t *state)
-{
-        xlator_t *subvol = NULL;
-
-        if (!state)
-                return NULL;
-
-        if (state->loc.inode)
-                subvol = state->loc.inode->table->xl;
-
-        if (state->fd)
-                subvol = state->fd->inode->table->xl;
-
-        return subvol;
-}
-
-
-xlator_t *
-fuse_active_subvol (xlator_t *fuse)
-{
-        fuse_private_t *priv = NULL;
-
-        priv = fuse->private;
-
-        return priv->active_subvol;
-}
-
+#if defined(GF_SOLARIS_HOST_OS)
+#include <sys/procfs.h>
+#else
+#include <sys/sysctl.h>
+#endif
 
 
 static void
-fuse_resolve_wipe (gf_resolve_t *resolve)
+fuse_resolve_wipe (fuse_resolve_t *resolve)
 {
-        struct gf_resolve_comp *comp = NULL;
+        GF_FREE ((void *)resolve->path);
 
-        if (resolve->path)
-                GF_FREE ((void *)resolve->path);
+        GF_FREE ((void *)resolve->bname);
 
-        if (resolve->bname)
-                GF_FREE ((void *)resolve->bname);
+        GF_FREE ((void *)resolve->resolved);
 
-        if (resolve->resolved)
-                GF_FREE ((void *)resolve->resolved);
+        if (resolve->fd)
+                fd_unref (resolve->fd);
 
-        loc_wipe (&resolve->deep_loc);
+        loc_wipe (&resolve->resolve_loc);
 
-        comp = resolve->components;
+	if (resolve->hint) {
+		inode_unref (resolve->hint);
+		resolve->hint = 0;
+	}
 
-        if (comp) {
-/*
-                int                  i = 0;
-
-                for (i = 0; comp[i].basename; i++) {
-                        if (comp[i].inode)
-                                inode_unref (comp[i].inode);
-                }
-*/
-                GF_FREE ((void *)resolve->components);
-        }
+	if (resolve->parhint) {
+		inode_unref (resolve->parhint);
+		resolve->parhint = 0;
+	}
 }
 
+
 void
 free_fuse_state (fuse_state_t *state)
 {
+        xlator_t       *this     = NULL;
+        fuse_private_t *priv     = NULL;
+        uint64_t        winds    = 0;
+        char            switched = 0;
+
+        this = state->this;
+
+        priv = this->private;
+
         loc_wipe (&state->loc);
 
         loc_wipe (&state->loc2);
 
-        if (state->dict) {
-                dict_unref (state->dict);
-                state->dict = (void *)0xaaaaeeee;
+        if (state->xdata) {
+                dict_unref (state->xdata);
+                state->xdata = (void *)0xaaaaeeee;
         }
+        if (state->xattr)
+                dict_unref (state->xattr);
+
         if (state->name) {
                 GF_FREE (state->name);
                 state->name = NULL;
@@ -107,6 +84,18 @@ free_fuse_state (fuse_state_t *state)
         fuse_resolve_wipe (&state->resolve);
         fuse_resolve_wipe (&state->resolve2);
 
+        pthread_mutex_lock (&priv->sync_mutex);
+        {
+                winds = --state->active_subvol->winds;
+                switched = state->active_subvol->switched;
+        }
+        pthread_mutex_unlock (&priv->sync_mutex);
+
+        if ((winds == 0) && (switched)) {
+                xlator_notify (state->active_subvol, GF_EVENT_PARENT_DOWN,
+                               state->active_subvol, NULL);
+        }
+
 #ifdef DEBUG
         memset (state, 0x90, sizeof (*state));
 #endif
@@ -118,12 +107,28 @@ free_fuse_state (fuse_state_t *state)
 fuse_state_t *
 get_fuse_state (xlator_t *this, fuse_in_header_t *finh)
 {
-        fuse_state_t *state = NULL;
+        fuse_state_t   *state         = NULL;
+	xlator_t       *active_subvol = NULL;
+        fuse_private_t *priv          = NULL;
 
         state = (void *)GF_CALLOC (1, sizeof (*state),
                                    gf_fuse_mt_fuse_state_t);
         if (!state)
                 return NULL;
+
+	state->this = THIS;
+        priv = this->private;
+
+        pthread_mutex_lock (&priv->sync_mutex);
+        {
+                active_subvol = fuse_active_subvol (state->this);
+                active_subvol->winds++;
+        }
+        pthread_mutex_unlock (&priv->sync_mutex);
+
+	state->active_subvol = active_subvol;
+	state->itable = active_subvol->itable;
+
         state->pool = this->ctx->pool;
         state->finh = finh;
         state->this = this;
@@ -134,6 +139,160 @@ get_fuse_state (xlator_t *this, fuse_in_header_t *finh)
 }
 
 
+#define FUSE_MAX_AUX_GROUPS 32 /* We can get only up to 32 aux groups from /proc */
+void
+frame_fill_groups (call_frame_t *frame)
+{
+#if defined(GF_LINUX_HOST_OS)
+        char         filename[32];
+        char         line[4096];
+        char        *ptr = NULL;
+        FILE        *fp = NULL;
+        int          idx = 0;
+        long int     id = 0;
+        char        *saveptr = NULL;
+        char        *endptr = NULL;
+        int          ret = 0;
+
+        ret = snprintf (filename, sizeof filename, "/proc/%d/status", frame->root->pid);
+        if (ret >= sizeof filename)
+                goto out;
+
+        fp = fopen (filename, "r");
+        if (!fp)
+                goto out;
+
+	if (call_stack_alloc_groups (frame->root, FUSE_MAX_AUX_GROUPS) != 0)
+		goto out;
+
+        while ((ptr = fgets (line, sizeof line, fp))) {
+                if (strncmp (ptr, "Groups:", 7) != 0)
+                        continue;
+
+                ptr = line + 8;
+
+                for (ptr = strtok_r (ptr, " \t\r\n", &saveptr);
+                     ptr;
+                     ptr = strtok_r (NULL, " \t\r\n", &saveptr)) {
+                        errno = 0;
+                        id = strtol (ptr, &endptr, 0);
+                        if (errno == ERANGE)
+                                break;
+                        if (!endptr || *endptr)
+                                break;
+                        frame->root->groups[idx++] = id;
+                        if (idx == FUSE_MAX_AUX_GROUPS)
+                                break;
+                }
+
+                frame->root->ngrps = idx;
+                break;
+        }
+out:
+        if (fp)
+                fclose (fp);
+#elif defined(GF_SOLARIS_HOST_OS)
+        char         filename[32];
+        char         scratch[128];
+        prcred_t    *prcred = (prcred_t *) scratch;
+        FILE        *fp = NULL;
+        int          ret = 0;
+	int          ngrps;
+
+        ret = snprintf (filename, sizeof filename,
+                        "/proc/%d/cred", frame->root->pid);
+
+        if (ret < sizeof filename) {
+                fp = fopen (filename, "r");
+                if (fp != NULL) {
+                        if (fgets (scratch, sizeof scratch, fp) != NULL) {
+                                ngrps = MIN(prcred->pr_ngroups,
+					    GF_MAX_AUX_GROUPS);
+				if (call_stack_alloc_groups (frame->root,
+							     ngrps) != 0)
+					return;
+                        }
+                        fclose (fp);
+                 }
+         }
+#elif defined(CTL_KERN) /* DARWIN and *BSD */
+        /* 
+           N.B. CTL_KERN is an enum on Linux. (Meaning, if it's not
+           obvious, that it's not subject to preprocessor directives 
+           like '#if defined'.)
+           Unlike Linux, on Mac OS and the BSDs it is a #define. We
+           could test to see that KERN_PROC is defined, but, barring any 
+           evidence to the contrary, I think that's overkill.
+           We might also test that GF_DARWIN_HOST_OS is defined, why
+           limit this to just Mac OS. It's equally valid for the BSDs
+           and we do have people building on NetBSD and FreeBSD.
+        */
+        int name[] = { CTL_KERN, KERN_PROC, KERN_PROC_PID, frame->root->pid };
+        size_t namelen = sizeof name / sizeof name[0];
+        struct kinfo_proc kp;
+        size_t kplen = sizeof(kp);
+        int i, ngroups;
+
+        if (sysctl(name, namelen, &kp, &kplen, NULL, 0) != 0)
+                return;
+        ngroups = MIN(kp.kp_eproc.e_ucred.cr_ngroups, GF_MAX_AUX_GROUPS);
+	if (call_stack_alloc_groups (frame->root, ngroups) != 0)
+		return;
+        for (i = 0; i < ngroups; i++)
+                frame->root->groups[i] = kp.kp_eproc.e_ucred.cr_groups[i];
+        frame->root->ngrps = ngroups;
+#else
+        frame->root->ngrps = 0;
+#endif /* GF_LINUX_HOST_OS */
+}
+
+/*
+ * Get the groups for the PID associated with this frame. If enabled,
+ * use the gid cache to reduce group list collection.
+ */
+static void get_groups(fuse_private_t *priv, call_frame_t *frame)
+{
+	int i;
+	const gid_list_t *gl;
+	gid_list_t agl;
+
+        if (-1 == priv->gid_cache_timeout) {
+                frame->root->ngrps = 0;
+                return;
+        }
+
+	if (!priv->gid_cache_timeout) {
+		frame_fill_groups(frame);
+		return;
+	}
+
+	gl = gid_cache_lookup(&priv->gid_cache, frame->root->pid);
+	if (gl) {
+		if (call_stack_alloc_groups (frame->root, gl->gl_count) != 0)
+			return;
+		frame->root->ngrps = gl->gl_count;
+		for (i = 0; i < gl->gl_count; i++)
+			frame->root->groups[i] = gl->gl_list[i];
+		gid_cache_release(&priv->gid_cache, gl);
+		return;
+	}
+
+	frame_fill_groups (frame);
+
+	agl.gl_id = frame->root->pid;
+	agl.gl_count = frame->root->ngrps;
+	agl.gl_list = GF_CALLOC(frame->root->ngrps, sizeof(gid_t),
+			gf_fuse_mt_gids_t);
+	if (!agl.gl_list)
+		return;
+
+	for (i = 0; i < frame->root->ngrps; i++)
+		agl.gl_list[i] = frame->root->groups[i];
+
+	if (gid_cache_add(&priv->gid_cache, &agl) != 1)
+		GF_FREE(agl.gl_list);
+}
+
 call_frame_t *
 get_call_frame_for_req (fuse_state_t *state)
 {
@@ -141,10 +300,12 @@ get_call_frame_for_req (fuse_state_t *state)
         fuse_in_header_t      *finh = NULL;
         call_frame_t          *frame = NULL;
         xlator_t              *this = NULL;
+        fuse_private_t        *priv = NULL;
 
         pool = state->pool;
         finh = state->finh;
         this = state->this;
+        priv = this->private;
 
         frame = create_frame (this, pool);
         if (!frame)
@@ -154,10 +315,16 @@ get_call_frame_for_req (fuse_state_t *state)
                 frame->root->uid      = finh->uid;
                 frame->root->gid      = finh->gid;
                 frame->root->pid      = finh->pid;
-                frame->root->lk_owner = state->lk_owner;
                 frame->root->unique   = finh->unique;
+                set_lk_owner_from_uint64 (&frame->root->lk_owner,
+                                          state->lk_owner);
         }
 
+	get_groups(priv, frame);
+
+        if (priv && priv->client_pid_set)
+                frame->root->pid = priv->client_pid;
+
         frame->root->type = GF_OP_TYPE_FOP;
 
         return frame;
@@ -172,7 +339,8 @@ fuse_ino_to_inode (uint64_t ino, xlator_t *fuse)
 
         if (ino == 1) {
                 active_subvol = fuse_active_subvol (fuse);
-                inode = active_subvol->itable->root;
+                if (active_subvol)
+                        inode = active_subvol->itable->root;
         } else {
                 inode = (inode_t *) (unsigned long) ino;
                 inode_ref (inode);
@@ -184,7 +352,9 @@ fuse_ino_to_inode (uint64_t ino, xlator_t *fuse)
 uint64_t
 inode_to_fuse_nodeid (inode_t *inode)
 {
-        if (!inode || inode->ino == 1)
+        if (!inode)
+		return 0;
+	if (__is_root_gfid (inode->gfid))
                 return 1;
 
         return (unsigned long) inode;
@@ -199,6 +369,7 @@ fuse_loc_fill (loc_t *loc, fuse_state_t *state, ino_t ino,
         inode_t  *parent = NULL;
         int32_t   ret = -1;
         char     *path = NULL;
+        uuid_t    null_gfid = {0,};
 
         /* resistance against multiple invocation of loc_fill not to get
            reference leaks via inode_search() */
@@ -208,6 +379,8 @@ fuse_loc_fill (loc_t *loc, fuse_state_t *state, ino_t ino,
                 if (!parent) {
                         parent = fuse_ino_to_inode (par, state->this);
                         loc->parent = parent;
+                        if (parent)
+                                uuid_copy (loc->pargfid, parent->gfid);
                 }
 
                 inode = loc->inode;
@@ -219,8 +392,8 @@ fuse_loc_fill (loc_t *loc, fuse_state_t *state, ino_t ino,
                 ret = inode_path (parent, name, &path);
                 if (ret <= 0) {
                         gf_log ("glusterfs-fuse", GF_LOG_DEBUG,
-                                "inode_path failed for %"PRId64"/%s",
-                                parent->ino, name);
+                                "inode_path failed for %s/%s",
+                                (parent)?uuid_utoa (parent->gfid):"0", name);
                         goto fail;
                 }
                 loc->path = path;
@@ -229,27 +402,29 @@ fuse_loc_fill (loc_t *loc, fuse_state_t *state, ino_t ino,
                 if (!inode) {
                         inode = fuse_ino_to_inode (ino, state->this);
                         loc->inode = inode;
+                        if (inode)
+                                uuid_copy (loc->gfid, inode->gfid);
                 }
 
                 parent = loc->parent;
                 if (!parent) {
-                        parent = inode_parent (inode, par, name);
+                        parent = inode_parent (inode, null_gfid, NULL);
                         loc->parent = parent;
+                        if (parent)
+                                uuid_copy (loc->pargfid, parent->gfid);
+
                 }
 
                 ret = inode_path (inode, NULL, &path);
                 if (ret <= 0) {
                         gf_log ("glusterfs-fuse", GF_LOG_DEBUG,
-                                "inode_path failed for %"PRId64,
-                                inode->ino);
+                                "inode_path failed for %s",
+                                (inode) ? uuid_utoa (inode->gfid) : "0");
                         goto fail;
                 }
                 loc->path = path;
         }
 
-        if (inode)
-                loc->ino = inode->ino;
-
         if (loc->path) {
                 loc->name = strrchr (loc->path, '/');
                 if (loc->name)
@@ -267,122 +442,164 @@ fuse_loc_fill (loc_t *loc, fuse_state_t *state, ino_t ino,
         }
         ret = 0;
 fail:
+        /* this should not happen as inode_path returns -1 when buf is NULL
+           for sure */
+        if (path && !loc->path)
+                GF_FREE (path);
         return ret;
 }
 
+/* Use the same logic as the Linux NFS-client */
+#define GF_FUSE_SQUASH_INO(ino) ((uint32_t) ino) ^ (ino >> 32)
 
 /* courtesy of folly */
 void
-gf_fuse_stat2attr (struct iatt *st, struct fuse_attr *fa)
+gf_fuse_stat2attr (struct iatt *st, struct fuse_attr *fa, gf_boolean_t enable_ino32)
 {
-        fa->ino        = st->ia_ino;
-        fa->size       = st->ia_size;
-        fa->blocks     = st->ia_blocks;
-        fa->atime      = st->ia_atime;
-        fa->mtime      = st->ia_mtime;
-        fa->ctime      = st->ia_ctime;
-        fa->atimensec  = st->ia_atime_nsec;
-        fa->mtimensec  = st->ia_mtime_nsec;
-        fa->ctimensec  = st->ia_ctime_nsec;
-        fa->mode       = st_mode_from_ia (st->ia_prot, st->ia_type);
-        fa->nlink      = st->ia_nlink;
-        fa->uid        = st->ia_uid;
-        fa->gid        = st->ia_gid;
-        fa->rdev       = st->ia_rdev;
+        if (enable_ino32)
+                fa->ino = GF_FUSE_SQUASH_INO(st->ia_ino);
+        else
+                fa->ino = st->ia_ino;
+
+        fa->size        = st->ia_size;
+        fa->blocks      = st->ia_blocks;
+        fa->atime       = st->ia_atime;
+        fa->mtime       = st->ia_mtime;
+        fa->ctime       = st->ia_ctime;
+        fa->atimensec   = st->ia_atime_nsec;
+        fa->mtimensec   = st->ia_mtime_nsec;
+        fa->ctimensec   = st->ia_ctime_nsec;
+        fa->mode        = st_mode_from_ia (st->ia_prot, st->ia_type);
+        fa->nlink       = st->ia_nlink;
+        fa->uid         = st->ia_uid;
+        fa->gid         = st->ia_gid;
+        fa->rdev        = makedev (ia_major (st->ia_rdev),
+                                   ia_minor (st->ia_rdev));
 #if FUSE_KERNEL_MINOR_VERSION >= 9
-        fa->blksize    = st->ia_blksize;
+        fa->blksize     = st->ia_blksize;
 #endif
 #ifdef GF_DARWIN_HOST_OS
-        fa->crtime     = (uint64_t)-1;
-        fa->crtimensec = (uint32_t)-1;
-        fa->flags      = 0;
+        fa->crtime      = (uint64_t)-1;
+        fa->crtimensec  = (uint32_t)-1;
+        fa->flags       = 0;
 #endif
 }
 
-
-int
-is_gf_log_command (xlator_t *this, const char *name, char *value)
+void
+gf_fuse_fill_dirent (gf_dirent_t *entry, struct fuse_dirent *fde, gf_boolean_t enable_ino32)
 {
-        fuse_private_t *priv        = NULL;
-        xlator_t       *trav        = NULL;
-        char            key[1024]   = {0,};
-        int             ret         = -1;
-        int             log_level   = -1;
-        gf_boolean_t    syslog_flag = 0;
+        if (enable_ino32)
+                fde->ino = GF_FUSE_SQUASH_INO(entry->d_ino);
+        else
+                fde->ino = entry->d_ino;
+
+        fde->off         = entry->d_off;
+        fde->type        = entry->d_type;
+        fde->namelen     = strlen (entry->d_name);
+        strncpy (fde->name, entry->d_name, fde->namelen);
+}
 
-        priv = this->private;
+static int
+fuse_do_flip_xattr_ns (char *okey, const char *nns, char **nkey)
+{
+        int   ret = 0;
+        char *key = NULL;
 
-        if (!strcmp ("trusted.glusterfs.syslog", name)) {
-                ret = gf_string2boolean (value, &syslog_flag);
-                if (ret) {
-                        ret = EOPNOTSUPP;
-                        goto out;
-                }
-                if (syslog_flag)
-                        gf_log_enable_syslog ();
-                else
-                        gf_log_disable_syslog ();
+        okey = strchr (okey, '.');
+        GF_ASSERT (okey);
 
+        key = GF_CALLOC (1, strlen (nns) + strlen(okey) + 1,
+                         gf_common_mt_char);
+        if (!key) {
+                ret = -1;
                 goto out;
         }
 
-        if (fnmatch ("trusted.glusterfs*set-log-level", name, FNM_NOESCAPE))
-                goto out;
+        strcpy (key, nns);
+        strcat (key, okey);
+
+        *nkey = key;
+
+ out:
+        return ret;
+}
 
-        if (!strcasecmp (value, "CRITICAL")) {
-                log_level = GF_LOG_CRITICAL;
-        } else if (!strcasecmp (value, "ERROR")) {
-                log_level = GF_LOG_ERROR;
-        } else if (!strcasecmp (value, "WARNING")) {
-                log_level = GF_LOG_WARNING;
-        } else if (!strcasecmp (value, "INFO")) {
-                log_level = GF_LOG_INFO;
-        } else if (!strcasecmp (value, "DEBUG")) {
-                log_level = GF_LOG_DEBUG;
-        } else if (!strcasecmp (value, "TRACE")) {
-                log_level = GF_LOG_TRACE;
-        } else if (!strcasecmp (value, "NONE")) {
-                log_level = GF_LOG_NONE;
+static int
+fuse_xattr_alloc_default (char *okey, char **nkey)
+{
+        int ret = 0;
+
+        *nkey = gf_strdup (okey);
+        if (!*nkey)
+                ret = -1;
+        return ret;
+}
+
+#define PRIV_XA_NS   "trusted"
+#define UNPRIV_XA_NS "system"
+
+int
+fuse_flip_xattr_ns (fuse_private_t *priv, char *okey, char **nkey)
+{
+        int             ret       = 0;
+        gf_boolean_t    need_flip = _gf_false;
+
+        switch (priv->client_pid) {
+        case GF_CLIENT_PID_GSYNCD:
+                /* valid xattr(s): *xtime, volume-mark* */
+                gf_log("glusterfs-fuse", GF_LOG_DEBUG, "PID: %d, checking xattr(s): "
+                       "volume-mark*, *xtime", priv->client_pid);
+                if ( (strcmp (okey, UNPRIV_XA_NS".glusterfs.volume-mark") == 0)
+                     || (fnmatch (UNPRIV_XA_NS".glusterfs.volume-mark.*", okey, FNM_PERIOD) == 0)
+                     || (fnmatch (UNPRIV_XA_NS".glusterfs.*.xtime", okey, FNM_PERIOD) == 0) )
+                        need_flip = _gf_true;
+                break;
+
+        case GF_CLIENT_PID_HADOOP:
+                /* valid xattr(s): pathinfo */
+                gf_log("glusterfs-fuse", GF_LOG_DEBUG, "PID: %d, checking xattr(s): "
+                       "pathinfo", priv->client_pid);
+                if (strcmp (okey, UNPRIV_XA_NS".glusterfs.pathinfo") == 0)
+                        need_flip = _gf_true;
+                break;
         }
 
-        if (log_level == -1) {
-                ret = EOPNOTSUPP;
-                goto out;
+        if (need_flip) {
+                gf_log ("glusterfs-fuse", GF_LOG_DEBUG, "flipping %s to "PRIV_XA_NS" equivalent",
+                        okey);
+                ret = fuse_do_flip_xattr_ns (okey, PRIV_XA_NS, nkey);
+        } else {
+                /* if we cannot match, continue with what we got */
+                ret = fuse_xattr_alloc_default (okey, nkey);
         }
 
-        /* Some crude way to change the log-level of process */
-        if (!strcmp (name, "trusted.glusterfs.set-log-level")) {
-                /* */
-                gf_log ("glusterfs", gf_log_get_loglevel(),
-                        "setting log level to %d (old-value=%d)",
-                        log_level, gf_log_get_loglevel());
-                gf_log_set_loglevel (log_level);
-                ret = 0;
+        return ret;
+}
+
+int
+fuse_ignore_xattr_set (fuse_private_t *priv, char *key)
+{
+        int ret = 0;
+
+        /* don't mess with user namespace */
+        if (fnmatch ("user.*", key, FNM_PERIOD) == 0)
                 goto out;
-        }
-        if (!strcmp (name, "trusted.glusterfs.fuse.set-log-level")) {
-                /* */
-                gf_log (this->name, gf_log_get_xl_loglevel (this),
-                        "setting log level to %d (old-value=%d)",
-                        log_level, gf_log_get_xl_loglevel (this));
-                gf_log_set_xl_loglevel (this, log_level);
-                ret = 0;
+
+        if (priv->client_pid != GF_CLIENT_PID_GSYNCD)
                 goto out;
-        }
 
-        trav = priv->active_subvol;
-        while (trav) {
-                snprintf (key, 1024, "trusted.glusterfs.%s.set-log-level",
-                          trav->name);
-                if (fnmatch (name, key, FNM_NOESCAPE) == 0) {
-                        gf_log (trav->name, gf_log_get_xl_loglevel (trav),
-                                "setting log level to %d (old-value=%d)",
-                                log_level, gf_log_get_xl_loglevel (trav));
-                        gf_log_set_xl_loglevel (trav, log_level);
-                        ret = 0;
-                }
-                trav = trav->next;
-        }
-out:
+        /* trusted NS check */
+        if (!((fnmatch ("*.glusterfs.*.xtime", key, FNM_PERIOD) == 0)
+              || (fnmatch ("*.glusterfs.volume-mark",
+                           key, FNM_PERIOD) == 0)
+              || (fnmatch ("*.glusterfs.volume-mark.*",
+                           key, FNM_PERIOD) == 0)))
+                ret = -1;
+
+ out:
+        gf_log ("glusterfs-fuse", GF_LOG_DEBUG, "%s setxattr: key [%s], "
+                " client pid [%d]", (ret ? "disallowing" : "allowing"), key,
+                priv->client_pid);
+
         return ret;
 }
diff --git a/xlators/mount/fuse/src/fuse-mem-types.h b/xlators/mount/fuse/src/fuse-mem-types.h
index b0eb816cd..28b4dfbdd 100644
--- a/xlators/mount/fuse/src/fuse-mem-types.h
+++ b/xlators/mount/fuse/src/fuse-mem-types.h
@@ -1,23 +1,13 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
 
-
 #ifndef __FUSE_MEM_TYPES_H__
 #define __FUSE_MEM_TYPES_H__
 
@@ -29,6 +19,9 @@ enum gf_fuse_mem_types_ {
         gf_fuse_mt_char,
         gf_fuse_mt_iov_base,
         gf_fuse_mt_fuse_state_t,
+        gf_fuse_mt_fd_ctx_t,
+        gf_fuse_mt_graph_switch_args_t,
+	gf_fuse_mt_gids_t,
         gf_fuse_mt_end
 };
 #endif
diff --git a/xlators/mount/fuse/src/fuse-resolve.c b/xlators/mount/fuse/src/fuse-resolve.c
index dd248f3d9..8565ce0e4 100644
--- a/xlators/mount/fuse/src/fuse-resolve.c
+++ b/xlators/mount/fuse/src/fuse-resolve.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -25,618 +15,638 @@
 #include "fuse-bridge.h"
 
 static int
-gf_resolve_all (fuse_state_t *state);
-static int
-resolve_entry_simple (fuse_state_t *state);
-static int
-resolve_inode_simple (fuse_state_t *state);
-static int
-resolve_path_simple (fuse_state_t *state);
-
-static int
-component_count (const char *path)
-{
-        int         count = 0;
-        const char *trav = NULL;
-
-        for (trav = path; *trav; trav++) {
-                if (*trav == '/')
-                        count++;
-        }
+fuse_resolve_all (fuse_state_t *state);
 
-        return count + 2;
-}
+int fuse_resolve_continue (fuse_state_t *state);
+int fuse_resolve_entry_simple (fuse_state_t *state);
+int fuse_resolve_inode_simple (fuse_state_t *state);
+int fuse_migrate_fd (xlator_t *this, fd_t *fd, xlator_t *old_subvol,
+                     xlator_t *new_subvol);
 
+fuse_fd_ctx_t *
+fuse_fd_ctx_get (xlator_t *this, fd_t *fd);
 
-static int
-prepare_components (fuse_state_t *state)
-{
-        xlator_t               *active_xl  = NULL;
-        gf_resolve_t           *resolve    = NULL;
-        char                   *resolved   = NULL;
-        struct gf_resolve_comp *components = NULL;
-        char                   *trav       = NULL;
-        int                     count      = 0;
-        int                     i          = 0;
-
-        resolve = state->resolve_now;
-
-        resolved = gf_strdup (resolve->path);
-        resolve->resolved = resolved;
-
-        count = component_count (resolve->path);
-        components = GF_CALLOC (sizeof (*components), count, 0); //TODO
-        if (!components)
-                goto out;
-        resolve->components = components;
-
-        active_xl = fuse_active_subvol (state->this);
-
-        components[0].basename = "";
-        components[0].ino      = 1;
-        components[0].gen      = 0;
-        components[0].inode    = inode_ref (active_xl->itable->root);
-
-        i = 1;
-        for (trav = resolved; *trav; trav++) {
-                if (*trav == '/') {
-                        components[i].basename = trav + 1;
-                        *trav = 0;
-                        i++;
-                }
-        }
-out:
-        return 0;
-}
-
+gf_boolean_t fuse_inode_needs_lookup (inode_t *inode, xlator_t *this);
 
 static int
-resolve_loc_touchup (fuse_state_t *state)
+fuse_resolve_loc_touchup (fuse_state_t *state)
 {
-        gf_resolve_t *resolve = NULL;
-        loc_t        *loc     = NULL;
-        char         *path    = NULL;
-        int           ret     = 0;
+        fuse_resolve_t *resolve = NULL;
+        loc_t          *loc     = NULL;
+        char           *path    = NULL;
+        int             ret     = 0;
 
         resolve = state->resolve_now;
         loc     = state->loc_now;
 
         if (!loc->path) {
-                if (loc->parent) {
+                if (loc->parent && resolve->bname) {
                         ret = inode_path (loc->parent, resolve->bname, &path);
+			uuid_copy (loc->pargfid, loc->parent->gfid);
+			loc->name = resolve->bname;
                 } else if (loc->inode) {
                         ret = inode_path (loc->inode, NULL, &path);
+			uuid_copy (loc->gfid, loc->inode->gfid);
                 }
                 if (ret)
-                        gf_log ("", GF_LOG_TRACE,
+                        gf_log (THIS->name, GF_LOG_TRACE,
                                 "return value inode_path %d", ret);
-
-                if (!path)
-                        path = gf_strdup (resolve->path);
-
                 loc->path = path;
         }
 
-        loc->name = strrchr (loc->path, '/');
-        if (loc->name)
-                loc->name++;
-
-        if (!loc->parent && loc->inode) {
-                loc->parent = inode_parent (loc->inode, 0, NULL);
-        }
-
         return 0;
 }
 
-static int
-fuse_resolve_newfd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno, fd_t *fd)
+
+int
+fuse_resolve_entry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			int op_ret, int op_errno, inode_t *inode,
+			struct iatt *buf, dict_t *xattr,
+			struct iatt *postparent)
 {
-        fuse_state_t *state      = NULL;
-        gf_resolve_t *resolve    = NULL;
-        fd_t         *old_fd     = NULL;
-        fd_t         *tmp_fd     = NULL;
-        uint64_t      tmp_fd_ctx = 0;
-        int           ret        = 0;
+        fuse_state_t   *state      = NULL;
+        fuse_resolve_t *resolve    = NULL;
+        inode_t        *link_inode = NULL;
+        loc_t          *resolve_loc   = NULL;
 
         state = frame->root->state;
         resolve = state->resolve_now;
+        resolve_loc = &resolve->resolve_loc;
 
         STACK_DESTROY (frame->root);
 
         if (op_ret == -1) {
-                resolve->op_ret   = -1;
+                gf_log (this->name, (op_errno == ENOENT)
+                        ? GF_LOG_DEBUG : GF_LOG_WARNING,
+                        "%s/%s: failed to resolve (%s)",
+                        uuid_utoa (resolve_loc->pargfid), resolve_loc->name,
+                        strerror (op_errno));
+                resolve->op_ret = -1;
                 resolve->op_errno = op_errno;
                 goto out;
         }
 
-        old_fd = resolve->fd;
-
-        state->fd = fd_ref (fd);
+        link_inode = inode_link (inode, resolve_loc->parent,
+                                 resolve_loc->name, buf);
 
-        fd_bind (fd);
+	state->loc_now->inode = link_inode;
 
-        resolve->fd = NULL;
-        ret = fd_ctx_del (old_fd, state->this, &tmp_fd_ctx);
-        if (!ret) {
-                tmp_fd = (fd_t *)(long)tmp_fd_ctx;
-                fd_unref (tmp_fd);
-        }
-        ret = fd_ctx_set (old_fd, state->this, (uint64_t)(long)fd);
-        if (ret)
-                gf_log ("resolve", GF_LOG_WARNING,
-                        "failed to set the fd ctx with resolved fd");
 out:
-        gf_resolve_all (state);
+        loc_wipe (resolve_loc);
+
+        fuse_resolve_continue (state);
         return 0;
 }
 
-static void
-gf_resolve_new_fd (fuse_state_t *state)
+
+int
+fuse_resolve_entry (fuse_state_t *state)
 {
-        gf_resolve_t *resolve = NULL;
-        fd_t         *new_fd  = NULL;
-        fd_t         *fd      = NULL;
+	fuse_resolve_t   *resolve = NULL;
+	loc_t            *resolve_loc = NULL;
 
-        resolve = state->resolve_now;
-        fd = resolve->fd;
+	resolve = state->resolve_now;
+	resolve_loc = &resolve->resolve_loc;
+
+	resolve_loc->parent = inode_ref (state->loc_now->parent);
+	uuid_copy (resolve_loc->pargfid, state->loc_now->pargfid);
+        resolve_loc->name = resolve->bname;
+        resolve_loc->inode = inode_new (state->itable);
 
-        new_fd = fd_create (state->loc.inode, state->finh->pid);
-        new_fd->flags = (fd->flags & ~O_TRUNC);
+        inode_path (resolve_loc->parent, resolve_loc->name,
+                    (char **) &resolve_loc->path);
 
-        gf_log ("resolve", GF_LOG_DEBUG,
-                "%"PRIu64": OPEN %s", state->finh->unique,
-                state->loc.path);
+        FUSE_FOP (state, fuse_resolve_entry_cbk, GF_FOP_LOOKUP,
+                  lookup, resolve_loc, NULL);
 
-        FUSE_FOP (state, fuse_resolve_newfd_cbk, GF_FOP_OPEN,
-                  open, &state->loc, new_fd->flags, new_fd, 0);
+	return 0;
 }
 
-static int
-resolve_deep_continue (fuse_state_t *state)
+
+int
+fuse_resolve_gfid_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int op_ret, int op_errno, inode_t *inode,
+                       struct iatt *buf, dict_t *xattr, struct iatt *postparent)
 {
-        gf_resolve_t     *resolve = NULL;
-        int               ret = 0;
+        fuse_state_t   *state      = NULL;
+        fuse_resolve_t *resolve    = NULL;
+        inode_t        *link_inode = NULL;
+        loc_t          *loc_now   = NULL;
 
+        state = frame->root->state;
         resolve = state->resolve_now;
+	loc_now = state->loc_now;
+
+        STACK_DESTROY (frame->root);
+
+        if (op_ret == -1) {
+                gf_log (this->name, (op_errno == ENOENT)
+                        ? GF_LOG_DEBUG : GF_LOG_WARNING,
+                        "%s: failed to resolve (%s)",
+                        uuid_utoa (resolve->resolve_loc.gfid),
+			strerror (op_errno));
+                loc_wipe (&resolve->resolve_loc);
+
+                /* resolve->op_ret can have 3 values: 0, -1, -2.
+                 * 0 : resolution was successful.
+                 * -1: parent inode could not be resolved.
+                 * -2: entry (inode corresponding to path) could not be resolved
+                 */
+
+                if (uuid_is_null (resolve->gfid)) {
+                        resolve->op_ret = -1;
+                } else {
+                        resolve->op_ret = -2;
+                }
+
+                resolve->op_errno = op_errno;
+                goto out;
+        }
 
-        resolve->op_ret   = 0;
-        resolve->op_errno = 0;
+        loc_wipe (&resolve->resolve_loc);
 
-        if (resolve->par)
-                ret = resolve_entry_simple (state);
-        else if (resolve->ino)
-                ret = resolve_inode_simple (state);
-        else if (resolve->path)
-                ret = resolve_path_simple (state);
-        if (ret)
-                gf_log ("resolve", GF_LOG_TRACE,
-                        "return value of resolve_*_simple %d", ret);
+        link_inode = inode_link (inode, NULL, NULL, buf);
+
+        if (!link_inode)
+                goto out;
 
-        resolve_loc_touchup (state);
+	if (!uuid_is_null (resolve->gfid)) {
+		loc_now->inode = link_inode;
+		goto out;
+	}
 
-        /* This function is called by either fd resolve or inode resolve */
-        if (!resolve->fd)
-                gf_resolve_all (state);
-        else
-                gf_resolve_new_fd (state);
+	loc_now->parent = link_inode;
+        uuid_copy (loc_now->pargfid, link_inode->gfid);
 
+	fuse_resolve_entry (state);
+
+        return 0;
+out:
+        fuse_resolve_continue (state);
         return 0;
 }
 
 
-static int
-resolve_deep_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int op_ret, int op_errno, inode_t *inode, struct iatt *buf,
-                  dict_t *xattr, struct iatt *postparent)
+int
+fuse_resolve_gfid (fuse_state_t *state)
 {
-        xlator_t               *active_xl = NULL;
-        fuse_state_t           *state      = NULL;
-        gf_resolve_t           *resolve    = NULL;
-        struct gf_resolve_comp *components = NULL;
-        inode_t                *link_inode = NULL;
-        int                     i          = 0;
+        fuse_resolve_t *resolve  = NULL;
+        loc_t          *resolve_loc = NULL;
+        int             ret      = 0;
 
-        state = frame->root->state;
         resolve = state->resolve_now;
-        components = resolve->components;
-
-        i = (long) cookie;
+        resolve_loc = &resolve->resolve_loc;
 
-        STACK_DESTROY (frame->root);
-
-        if (op_ret == -1) {
-                goto get_out_of_here;
+        if (!uuid_is_null (resolve->pargfid)) {
+                uuid_copy (resolve_loc->gfid, resolve->pargfid);
+        } else if (!uuid_is_null (resolve->gfid)) {
+                uuid_copy (resolve_loc->gfid, resolve->gfid);
         }
 
-        if (i != 0) {
-                inode_ref (inode);
-                /* no linking for root inode */
-                link_inode = inode_link (inode, resolve->deep_loc.parent,
-                                         resolve->deep_loc.name, buf);
-                components[i].inode  = inode_ref (link_inode);
-                link_inode = NULL;
+	/* inode may already exist in case we are looking up an inode which was
+	   linked through readdirplus */
+	resolve_loc->inode = inode_find (state->itable, resolve_loc->gfid);
+	if (!resolve_loc->inode)
+		resolve_loc->inode = inode_new (state->itable);
+	ret = loc_path (resolve_loc, NULL);
+
+        if (ret <= 0) {
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to get the path for inode %s",
+                        uuid_utoa (resolve->gfid));
         }
-        inode_ref (resolve->deep_loc.parent);
-        inode_ref (inode);
-        loc_wipe (&resolve->deep_loc);
-        i++; /* next component */
-
-        if (!components[i].basename) {
-                /* all components of the path are resolved */
-                goto get_out_of_here;
-        }
-
-        /* join the current component with the path resolved until now */
-        *(components[i].basename - 1) = '/';
 
-        active_xl = fuse_active_subvol (state->this);
+        FUSE_FOP (state, fuse_resolve_gfid_cbk, GF_FOP_LOOKUP,
+                  lookup, resolve_loc, NULL);
 
-        resolve->deep_loc.path   = gf_strdup (resolve->resolved);
-        resolve->deep_loc.parent = inode_ref (components[i-1].inode);
-        resolve->deep_loc.inode  = inode_new (active_xl->itable);
-        resolve->deep_loc.name   = components[i].basename;
-
-        FUSE_FOP_COOKIE (state, active_xl, resolve_deep_cbk, (void *)(long)i,
-                         GF_FOP_LOOKUP, lookup, &resolve->deep_loc, NULL);
-        return 0;
-
-get_out_of_here:
-        resolve_deep_continue (state);
         return 0;
 }
 
 
-static int
-resolve_path_deep (fuse_state_t *state)
+/*
+ * Return value:
+ * 0 - resolved parent and entry (as necessary)
+ * -1 - resolved parent but not entry (though necessary)
+ * 1 - resolved neither parent nor entry
+ */
+
+int
+fuse_resolve_parent_simple (fuse_state_t *state)
 {
-        xlator_t               *active_xl  = NULL;
-        gf_resolve_t           *resolve    = NULL;
-        struct gf_resolve_comp *components = NULL;
-        inode_t                *inode      = NULL;
-        long                    i          = 0;
+        fuse_resolve_t *resolve   = NULL;
+	loc_t          *loc       = NULL;
+        inode_t        *parent    = NULL;
+        inode_t        *inode     = NULL;
 
         resolve = state->resolve_now;
+	loc = state->loc_now;
+
+	loc->name = resolve->bname;
+
+	parent = resolve->parhint;
+	if (parent->table == state->itable) {
+		if (fuse_inode_needs_lookup (parent, THIS))
+			return 1;
+
+		/* no graph switches since */
+		loc->parent = inode_ref (parent);
+		uuid_copy (loc->pargfid, parent->gfid);
+		loc->inode = inode_grep (state->itable, parent, loc->name);
+
+                /* nodeid for root is 1 and we blindly take the latest graph's
+                 * table->root as the parhint and because of this there is
+                 * ambiguity whether the entry should have existed or not, and
+                 * we took the conservative approach of assuming entry should
+                 * have been there even though it need not have (bug #804592).
+                 */
+                if ((loc->inode == NULL)
+                    && __is_root_gfid (parent->gfid)) {
+                        /* non decisive result - entry missing */
+                        return -1;
+                }
 
-        prepare_components (state);
+		/* decisive result - resolution success */
+		return 0;
+	}
+
+        parent = inode_find (state->itable, resolve->pargfid);
+	if (!parent) {
+		/* non decisive result - parent missing */
+		return 1;
+	}
+	if (fuse_inode_needs_lookup (parent, THIS)) {
+		inode_unref (parent);
+		return 1;
+	}
+
+	loc->parent = parent;
+        uuid_copy (loc->pargfid, resolve->pargfid);
+
+	inode = inode_grep (state->itable, parent, loc->name);
+	if (inode) {
+		loc->inode = inode;
+		/* decisive result - resolution success */
+		return 0;
+	}
+
+	/* non decisive result - entry missing */
+        return -1;
+}
 
-        components = resolve->components;
 
-        /* start from the root */
-        active_xl = fuse_active_subvol (state->this);
-        resolve->deep_loc.inode = inode_ref (active_xl->itable->root);
-        resolve->deep_loc.path  = gf_strdup ("/");
-        resolve->deep_loc.name  = "";
+int
+fuse_resolve_parent (fuse_state_t *state)
+{
+        int    ret = 0;
 
-        for (i = 1; components[i].basename; i++) {
-                *(components[i].basename - 1) = '/';
-                inode = inode_grep (active_xl->itable, components[i-1].inode,
-                                    components[i].basename);
-                if (!inode)
-                        break;
-                components[i].inode = inode_ref (inode);
+        ret = fuse_resolve_parent_simple (state);
+        if (ret > 0) {
+                fuse_resolve_gfid (state);
+                return 0;
         }
 
-        if (!components[i].basename)
-                goto resolved;
-
-        resolve->deep_loc.path   = gf_strdup (resolve->resolved);
-        resolve->deep_loc.parent = inode_ref (components[i-1].inode);
-        resolve->deep_loc.inode  = inode_new (active_xl->itable);
-        resolve->deep_loc.name   = components[i].basename;
+	if (ret < 0) {
+		fuse_resolve_entry (state);
+		return 0;
+	}
 
-        FUSE_FOP_COOKIE (state, active_xl, resolve_deep_cbk, (void *)(long)i,
-                         GF_FOP_LOOKUP, lookup, &resolve->deep_loc, NULL);
+        fuse_resolve_continue (state);
 
         return 0;
-resolved:
-        resolve_deep_continue (state);
-        return 0;
 }
 
 
-static int
-resolve_path_simple (fuse_state_t *state)
+int
+fuse_resolve_inode_simple (fuse_state_t *state)
 {
-        gf_resolve_t           *resolve    = NULL;
-        struct gf_resolve_comp *components = NULL;
-        int                     ret        = -1;
-        int                     par_idx    = 0;
-        int                     ino_idx    = 0;
-        int                     i          = 0;
+        fuse_resolve_t *resolve   = NULL;
+	loc_t          *loc = NULL;
+        inode_t        *inode     = NULL;
 
         resolve = state->resolve_now;
-        components = resolve->components;
-
-        if (!components) {
-                resolve->op_ret   = -1;
-                resolve->op_errno = ENOENT;
-                goto out;
-        }
+	loc = state->loc_now;
+
+	inode = resolve->hint;
+	if (inode->table == state->itable)
+		inode_ref (inode);
+	else
+		inode = inode_find (state->itable, resolve->gfid);
+
+        if (inode) {
+		if (!fuse_inode_needs_lookup (inode, THIS))
+			goto found;
+		/* inode was linked through readdirplus */
+		inode_unref (inode);
+	}
+
+        return 1;
+found:
+	loc->inode = inode;
+	return 0;
+}
 
-        for (i = 0; components[i].basename; i++) {
-                par_idx = ino_idx;
-                ino_idx = i;
-        }
 
-        if (!components[par_idx].inode) {
-                resolve->op_ret    = -1;
-                resolve->op_errno  = ENOENT;
-                goto out;
-        }
+int
+fuse_resolve_inode (fuse_state_t *state)
+{
+        int                 ret = 0;
 
-        if (!components[ino_idx].inode &&
-            (resolve->type == RESOLVE_MUST || resolve->type == RESOLVE_EXACT)) {
-                resolve->op_ret    = -1;
-                resolve->op_errno  = ENOENT;
-                goto out;
-        }
+        ret = fuse_resolve_inode_simple (state);
 
-        if (components[ino_idx].inode && resolve->type == RESOLVE_NOT) {
-                resolve->op_ret    = -1;
-                resolve->op_errno  = EEXIST;
-                goto out;
+        if (ret > 0) {
+                fuse_resolve_gfid (state);
+                return 0;
         }
 
-        if (components[ino_idx].inode)
-                state->loc_now->inode  = inode_ref (components[ino_idx].inode);
-        state->loc_now->parent = inode_ref (components[par_idx].inode);
-
-        ret = 0;
+        fuse_resolve_continue (state);
 
-out:
-        return ret;
+        return 0;
 }
 
-/*
-  Check if the requirements are fulfilled by entries in the inode cache itself
-  Return value:
-  <= 0 - simple resolution was decisive and complete (either success or failure)
-  > 0  - indecisive, need to perform deep resolution
-*/
 
-static int
-resolve_entry_simple (fuse_state_t *state)
+int
+fuse_migrate_fd_task (void *data)
 {
-        xlator_t     *this    = NULL;
-        xlator_t     *active_xl = NULL;
-        gf_resolve_t *resolve = NULL;
-        inode_t      *parent  = NULL;
-        inode_t      *inode   = NULL;
-        int           ret     = 0;
-
-        this  = state->this;
-        resolve = state->resolve_now;
-
-        active_xl = fuse_active_subvol (this);
-
-        parent = inode_get (active_xl->itable, resolve->par, 0);
-        if (!parent) {
-                /* simple resolution is indecisive. need to perform
-                   deep resolution */
-                resolve->op_ret   = -1;
-                resolve->op_errno = ENOENT;
-                ret = 1;
-
-                inode = inode_grep (active_xl->itable, parent, resolve->bname);
-                if (inode != NULL) {
-                        gf_log (this->name, GF_LOG_DEBUG, "%"PRId64": inode "
-                                "(pointer:%p ino: %"PRIu64") present but parent"
-                                " is NULL for path (%s)", 0L,
-                                inode, inode->ino, resolve->path);
-                        inode_unref (inode);
-                }
+        int            ret        = -1;
+        fuse_state_t  *state      = NULL;
+        fd_t          *basefd     = NULL, *oldfd = NULL;
+        fuse_fd_ctx_t *basefd_ctx = NULL;
+        xlator_t      *old_subvol = NULL;
+
+        state = data;
+        if (state == NULL) {
                 goto out;
         }
 
-        /* expected @parent was found from the inode cache */
-        state->loc_now->parent = inode_ref (parent);
-
-        inode = inode_grep (active_xl->itable, parent, resolve->bname);
-        if (!inode) {
-                switch (resolve->type) {
-                case RESOLVE_DONTCARE:
-                case RESOLVE_NOT:
-                        ret = 0;
-                        break;
-                case RESOLVE_MAY:
-                        ret = 1;
-                        break;
-                default:
-                        resolve->op_ret   = -1;
-                        resolve->op_errno = ENOENT;
-                        ret = 1;
-                        break;
-                }
+        basefd = state->fd;
 
-                goto out;
+        basefd_ctx = fuse_fd_ctx_get (state->this, basefd);
+
+        LOCK (&basefd->lock);
+        {
+                oldfd = basefd_ctx->activefd ? basefd_ctx->activefd : basefd;
+                fd_ref (oldfd);
         }
+        UNLOCK (&basefd->lock);
 
-        if (resolve->type == RESOLVE_NOT) {
-                gf_log (this->name, GF_LOG_DEBUG, "inode (pointer: %p ino:%"
-                        PRIu64") found for path (%s) while type is RESOLVE_NOT",
-                        inode, inode->ino, resolve->path);
-                resolve->op_ret   = -1;
-                resolve->op_errno = EEXIST;
-                ret = -1;
-                goto out;
+        old_subvol = oldfd->inode->table->xl;
+
+        ret = fuse_migrate_fd (state->this, basefd, old_subvol,
+                               state->active_subvol);
+
+        LOCK (&basefd->lock);
+        {
+                if (ret < 0) {
+                        basefd_ctx->migration_failed = 1;
+                } else {
+                        basefd_ctx->migration_failed = 0;
+                }
         }
+        UNLOCK (&basefd->lock);
 
         ret = 0;
 
-        state->loc_now->inode  = inode_ref (inode);
-
 out:
-        if (parent)
-                inode_unref (parent);
-
-        if (inode)
-                inode_unref (inode);
+        if (oldfd)
+                fd_unref (oldfd);
 
         return ret;
 }
 
 
-static int
-gf_resolve_entry (fuse_state_t *state)
+static inline int
+fuse_migrate_fd_error (xlator_t *this, fd_t *fd)
 {
-        int                 ret = 0;
-        loc_t              *loc = NULL;
-
-        loc  = state->loc_now;
+        fuse_fd_ctx_t *fdctx = NULL;
+        char           error = 0;
 
-        ret = resolve_entry_simple (state);
-
-        if (ret > 0) {
-                loc_wipe (loc);
-                resolve_path_deep (state);
-                return 0;
+        fdctx = fuse_fd_ctx_get (this, fd);
+        if (fdctx != NULL) {
+                if (fdctx->migration_failed) {
+                        error = 1;
+                }
         }
 
-        if (ret == 0)
-                resolve_loc_touchup (state);
-
-        gf_resolve_all (state);
-
-        return 0;
+        return error;
 }
 
+#define FUSE_FD_GET_ACTIVE_FD(activefd, basefd)                 \
+        do {                                                    \
+                LOCK (&basefd->lock);                           \
+                {                                               \
+                        activefd = basefd_ctx->activefd ?       \
+                                basefd_ctx->activefd : basefd;  \
+                        if (activefd != basefd) {               \
+                                fd_ref (activefd);              \
+                        }                                       \
+                }                                               \
+                UNLOCK (&basefd->lock);                         \
+                                                                \
+                if (activefd == basefd) {                       \
+                        fd_ref (activefd);                      \
+                }                                               \
+        } while (0);
+
 
 static int
-resolve_inode_simple (fuse_state_t *state)
+fuse_resolve_fd (fuse_state_t *state)
 {
-        xlator_t     *active_xl = NULL;
-        gf_resolve_t *resolve = NULL;
-        inode_t      *inode   = NULL;
-        int           ret     = 0;
+        fuse_resolve_t *resolve            = NULL;
+	fd_t           *basefd             = NULL, *activefd = NULL;
+	xlator_t       *active_subvol      = NULL, *this = NULL;
+        int             ret                = 0;
+        char            fd_migration_error = 0;
+        fuse_fd_ctx_t  *basefd_ctx         = NULL;
 
         resolve = state->resolve_now;
 
-        active_xl = fuse_active_subvol (state->this);
+        this = state->this;
 
-        if (resolve->type == RESOLVE_EXACT) {
-                inode = inode_get (active_xl->itable, resolve->ino,
-                                   resolve->gen);
-        } else {
-                inode = inode_get (active_xl->itable, resolve->ino, 0);
+        basefd = resolve->fd;
+        basefd_ctx = fuse_fd_ctx_get (this, basefd);
+        if (basefd_ctx == NULL) {
+                gf_log (state->this->name, GF_LOG_WARNING,
+                        "fdctx is NULL for basefd (ptr:%p inode-gfid:%s), "
+                        "resolver erroring out with errno EINVAL",
+                        basefd, uuid_utoa (basefd->inode->gfid));
+                resolve->op_ret = -1;
+                resolve->op_errno = EINVAL;
+                goto resolve_continue;
         }
 
-        if (!inode) {
-                resolve->op_ret   = -1;
-                resolve->op_errno = ENOENT;
-                ret = 1;
-                goto out;
+        FUSE_FD_GET_ACTIVE_FD (activefd, basefd);
+
+        active_subvol = activefd->inode->table->xl;
+
+        fd_migration_error = fuse_migrate_fd_error (state->this, basefd);
+        if (fd_migration_error) {
+                resolve->op_ret = -1;
+                resolve->op_errno = EBADF;
+        } else if (state->active_subvol != active_subvol) {
+                ret = synctask_new (state->this->ctx->env, fuse_migrate_fd_task,
+                                    NULL, NULL, state);
+
+                fd_migration_error = fuse_migrate_fd_error (state->this,
+                                                            basefd);
+                fd_unref (activefd);
+
+                FUSE_FD_GET_ACTIVE_FD (activefd, basefd);
+                active_subvol = activefd->inode->table->xl;
+
+                if ((ret == -1) || fd_migration_error
+                    || (state->active_subvol != active_subvol)) {
+                        if (ret == -1) {
+                                gf_log (state->this->name, GF_LOG_WARNING,
+                                        "starting sync-task to migrate "
+                                        "basefd (ptr:%p inode-gfid:%s) failed "
+                                        "(old-subvolume:%s-%d "
+                                        "new-subvolume:%s-%d)",
+                                        basefd,
+                                        uuid_utoa (basefd->inode->gfid),
+                                        active_subvol->name,
+                                        active_subvol->graph->id,
+                                        state->active_subvol->name,
+                                        state->active_subvol->graph->id);
+                        } else {
+                                gf_log (state->this->name, GF_LOG_WARNING,
+                                        "fd migration of basefd "
+                                        "(ptr:%p inode-gfid:%s) failed "
+                                        "(old-subvolume:%s-%d "
+                                        "new-subvolume:%s-%d)",
+                                        basefd,
+                                        uuid_utoa (basefd->inode->gfid),
+                                        active_subvol->name,
+                                        active_subvol->graph->id,
+                                        state->active_subvol->name,
+                                        state->active_subvol->graph->id);
+                        }
+
+                        resolve->op_ret = -1;
+                        resolve->op_errno = EBADF;
+                } else {
+                        gf_log (state->this->name, GF_LOG_DEBUG,
+                                "basefd (ptr:%p inode-gfid:%s) migrated "
+                                "successfully in resolver "
+                                "(old-subvolume:%s-%d new-subvolume:%s-%d)",
+                                basefd, uuid_utoa (basefd->inode->gfid),
+                                active_subvol->name, active_subvol->graph->id,
+                                state->active_subvol->name,
+                                state->active_subvol->graph->id);
+                }
         }
 
-        ret = 0;
+        if ((resolve->op_ret == -1) && (resolve->op_errno == EBADF)) {
+                gf_log ("fuse-resolve", GF_LOG_WARNING,
+                        "migration of basefd (ptr:%p inode-gfid:%s) "
+                        "did not complete, failing fop with EBADF "
+                        "(old-subvolume:%s-%d new-subvolume:%s-%d)", basefd,
+                        uuid_utoa (basefd->inode->gfid),
+                        active_subvol->name, active_subvol->graph->id,
+                        state->active_subvol->name,
+                        state->active_subvol->graph->id);
+        }
 
-        state->loc_now->inode = inode_ref (inode);
+        if (activefd != basefd) {
+                state->fd = fd_ref (activefd);
+                fd_unref (basefd);
+        }
 
-out:
-        if (inode)
-                inode_unref (inode);
+	/* state->active_subvol = active_subvol; */
 
-        return ret;
+resolve_continue:
+        if (activefd != NULL) {
+                fd_unref (activefd);
+        }
+
+        fuse_resolve_continue (state);
+
+        return 0;
 }
 
 
-static int
-gf_resolve_inode (fuse_state_t *state)
+int
+fuse_gfid_set (fuse_state_t *state)
 {
-        int                 ret = 0;
-        loc_t              *loc = NULL;
+        int   ret = 0;
 
-        loc  = state->loc_now;
+        if (uuid_is_null (state->gfid))
+                goto out;
 
-        ret = resolve_inode_simple (state);
+        if (!state->xdata)
+                state->xdata = dict_new ();
 
-        if (ret > 0) {
-                loc_wipe (loc);
-                resolve_path_deep (state);
-                return 0;
+        if (!state->xdata) {
+                ret = -1;
+                goto out;
         }
 
-        if (ret == 0)
-                resolve_loc_touchup (state);
-
-        gf_resolve_all (state);
-
-        return 0;
+        ret = dict_set_static_bin (state->xdata, "gfid-req",
+                                   state->gfid, sizeof (state->gfid));
+out:
+        return ret;
 }
 
 
-static int
-gf_resolve_fd (fuse_state_t *state)
+int
+fuse_resolve_entry_init (fuse_state_t *state, fuse_resolve_t *resolve,
+			 ino_t par, char *name)
 {
-        gf_resolve_t  *resolve    = NULL;
-        fd_t          *fd         = NULL;
-        int            ret        = 0;
-        uint64_t       tmp_fd_ctx = 0;
-        char          *path       = NULL;
-        char          *name       = NULL;
+	inode_t       *parent = NULL;
 
-        resolve = state->resolve_now;
+	parent = fuse_ino_to_inode (par, state->this);
+	uuid_copy (resolve->pargfid, parent->gfid);
+	resolve->parhint = parent;
+	resolve->bname = gf_strdup (name);
 
-        fd = resolve->fd;
+	return 0;
+}
 
-        ret = fd_ctx_get (fd, state->this, &tmp_fd_ctx);
-        if (!ret) {
-                state->fd = (fd_t *)(long)tmp_fd_ctx;
-                fd_ref (state->fd);
-                gf_resolve_all (state);
-                goto out;
-        }
 
-        ret = inode_path (fd->inode, 0, &path);
-        if (!ret || !path)
-                gf_log ("", GF_LOG_WARNING,
-                        "failed to do inode-path on fd %d %s", ret, path);
+int
+fuse_resolve_inode_init (fuse_state_t *state, fuse_resolve_t *resolve,
+			 ino_t ino)
+{
+	inode_t       *inode = NULL;
 
-        name = strrchr (path, '/');
-        if (name)
-                name++;
+	inode = fuse_ino_to_inode (ino, state->this);
+	uuid_copy (resolve->gfid, inode->gfid);
+	resolve->hint = inode;
 
-        resolve->path = path;
-        resolve->bname = gf_strdup (name);
+	return 0;
+}
 
-        state->loc_now     = &state->loc;
 
-        resolve_path_deep (state);
+int
+fuse_resolve_fd_init (fuse_state_t *state, fuse_resolve_t *resolve,
+		      fd_t *fd)
+{
+	resolve->fd = fd_ref (fd);
 
-out:
-        return 0;
+	return 0;
 }
 
 
 static int
-gf_resolve (fuse_state_t *state)
- {
-        gf_resolve_t   *resolve = NULL;
+fuse_resolve (fuse_state_t *state)
+{
+        fuse_resolve_t   *resolve = NULL;
 
         resolve = state->resolve_now;
 
         if (resolve->fd) {
 
-                gf_resolve_fd (state);
+                fuse_resolve_fd (state);
 
-        } else if (resolve->par) {
+        } else if (!uuid_is_null (resolve->pargfid)) {
 
-                gf_resolve_entry (state);
+                fuse_resolve_parent (state);
 
-        } else if (resolve->ino) {
+        } else if (!uuid_is_null (resolve->gfid)) {
 
-                gf_resolve_inode (state);
+                fuse_resolve_inode (state);
 
-        } else if (resolve->path) {
-
-                resolve_path_deep (state);
-
-        } else  {
-
-                resolve->op_ret = 0;
-                resolve->op_errno = EINVAL;
-
-                gf_resolve_all (state);
+        } else {
+                fuse_resolve_all (state);
         }
 
         return 0;
@@ -644,21 +654,14 @@ gf_resolve (fuse_state_t *state)
 
 
 static int
-gf_resolve_done (fuse_state_t *state)
+fuse_resolve_done (fuse_state_t *state)
 {
         fuse_resume_fn_t fn = NULL;
 
-        if (state->resolve.op_ret || state->resolve2.op_ret) {
-                send_fuse_err (state->this, state->finh,
-                               state->resolve.op_errno);
-                free_fuse_state (state);
-                goto out;
-        }
         fn = state->resume_fn;
-        if (fn)
-                fn (state);
 
-out:
+	fn (state);
+
         return 0;
 }
 
@@ -668,25 +671,25 @@ out:
  * state->resolve_now is used to decide which location/fd is to be resolved now
  */
 static int
-gf_resolve_all (fuse_state_t *state)
+fuse_resolve_all (fuse_state_t *state)
 {
         if (state->resolve_now == NULL) {
 
                 state->resolve_now = &state->resolve;
                 state->loc_now     = &state->loc;
 
-                gf_resolve (state);
+                fuse_resolve (state);
 
         } else if (state->resolve_now == &state->resolve) {
 
                 state->resolve_now = &state->resolve2;
                 state->loc_now     = &state->loc2;
 
-                gf_resolve (state);
+                fuse_resolve (state);
 
         } else if (state->resolve_now == &state->resolve2) {
 
-                gf_resolve_done (state);
+                fuse_resolve_done (state);
 
         } else {
                 gf_log ("fuse-resolve", GF_LOG_ERROR,
@@ -698,82 +701,24 @@ gf_resolve_all (fuse_state_t *state)
 
 
 int
-fuse_gfid_set (fuse_state_t *state)
+fuse_resolve_continue (fuse_state_t *state)
 {
-        int   ret = 0;
+        fuse_resolve_loc_touchup (state);
 
-        if (uuid_is_null (state->gfid))
-                goto out;
-
-        if (!state->dict)
-                state->dict = dict_new ();
-
-        if (!state->dict) {
-                ret = -1;
-                goto out;
-        }
+        fuse_resolve_all (state);
 
-        ret = dict_set_static_bin (state->dict, "gfid-req",
-                                   state->gfid, sizeof (state->gfid));
-out:
-        return ret;
+        return 0;
 }
 
 
 int
 fuse_resolve_and_resume (fuse_state_t *state, fuse_resume_fn_t fn)
 {
-        xlator_t *inode_xl = NULL;
-        xlator_t *active_xl = NULL;
-
         fuse_gfid_set (state);
 
         state->resume_fn = fn;
 
-        active_xl = fuse_active_subvol (state->this);
-        inode_xl = fuse_state_subvol (state);
-        if (!inode_xl && state->loc.parent)
-                inode_xl = state->loc.parent->table->xl;
-
-        /* If inode or fd is already in new graph, goto resume */
-        if (inode_xl == active_xl)
-                goto resume;
-
-        /* If the resolve is for 'fd' and its open with 'write' flag
-           set, don't switch to new graph yet */
-        if (state->fd && ((state->fd->flags & O_RDWR) ||
-                          (state->fd->flags & O_WRONLY)))
-                goto resume;
-
-        if (state->loc.path) {
-                state->resolve.path = gf_strdup (state->loc.path);
-                state->resolve.bname = gf_strdup (state->loc.name);
-                /* TODO: make sure there is no leaks in inode refs */
-                //loc_wipe (&state->loc);
-                state->loc.inode = NULL;
-                state->loc.parent = NULL;
-        }
-
-        /* Needed for rename and link */
-        if (state->loc2.path) {
-                state->resolve2.path = gf_strdup (state->loc2.path);
-                state->resolve2.bname = gf_strdup (state->loc2.name);
-                //loc_wipe (&state->loc2);
-                state->loc2.inode = NULL;
-                state->loc2.parent = NULL;
-        }
-
-        if (state->fd) {
-                state->resolve.fd = state->fd;
-                /* TODO: check if its a leak, if yes, then do 'unref' */
-                state->fd = NULL;
-        }
-
-        gf_resolve_all (state);
-
-        return 0;
-resume:
-        fn (state);
+        fuse_resolve_all (state);
 
         return 0;
 }
diff --git a/xlators/mount/fuse/utils/mount.glusterfs.in b/xlators/mount/fuse/utils/mount.glusterfs.in
index 5fbdd1472..a192d6059 100755
--- a/xlators/mount/fuse/utils/mount.glusterfs.in
+++ b/xlators/mount/fuse/utils/mount.glusterfs.in
@@ -1,4 +1,4 @@
-#!/bin/bash
+#!/bin/sh
 # (C) 2006, 2007, 2008 Gluster Inc. <http://www.gluster.com>
 #
 # This program is free software; you can redistribute it and/or
@@ -23,113 +23,244 @@ _init ()
     LOG_CRITICAL=CRITICAL;
     LOG_ERROR=ERROR;
     LOG_WARNING=WARNING;
-    LOG_NORMAL=NORMAL
+    LOG_INFO=INFO
     LOG_DEBUG=DEBUG;
     LOG_TRACE=TRACE;
 
-    # set default log level to NORMAL
-    log_level=$LOG_NORMAL;
+    HOST_NAME_MAX=64;
+
     prefix="@prefix@";
     exec_prefix=@exec_prefix@;
     cmd_line=$(echo "@sbindir@/glusterfs");
 
+    case `uname -s` in
+    NetBSD)
+        getinode="stat -f %i"
+        getdev="stat -f %d"
+        lgetinode="${getinode} -L"
+        lgetdev="${getdev} -L"
+
+        mounttab=/proc/mounts
+        ;;
+    Linux)
+        getinode="stat -c %i $i"
+        getdev="stat -c %d $d"
+        lgetinode="${getinode} -L"
+        lgetdev="${getdev} -L"
+
+        mounttab=/etc/mtab
+        ;;
+    esac
+
     UPDATEDBCONF=/etc/updatedb.conf
 }
 
+parse_backup_volfile_servers ()
+{
+    local server_list=$1
+    local servers=""
+    local new_servers=""
+
+    servers=$(echo ${server_list} | sed 's/\:/ /g')
+    for server in ${servers}; do
+        length=$(echo $server | wc -c)
+        if [ ${length} -gt ${HOST_NAME_MAX} ]; then
+            echo "Hostname:${server} provided is too long.. skipping"
+            continue
+        fi
+        new_servers=$(echo "$new_servers $server")
+    done
+    echo ${new_servers}
+}
+
 start_glusterfs ()
 {
+    # lets the comparsion be case insensitive for all strings
+
     if [ -n "$log_level_str" ]; then
-	case "$log_level_str" in
-	    "ERROR")
-		log_level=$LOG_ERROR;
-		;;
-            "NORMAL")
-                log_level=$LOG_NORMAL
+        case "$( echo $log_level_str | tr '[a-z]' '[A-Z]')" in
+            "ERROR")
+                log_level=$LOG_ERROR;
+                ;;
+            "INFO")
+                log_level=$LOG_INFO
+                ;;
+            "DEBUG")
+                log_level=$LOG_DEBUG;
+                ;;
+            "CRITICAL")
+                log_level=$LOG_CRITICAL;
+                ;;
+            "WARNING")
+                log_level=$LOG_WARNING;
+                ;;
+            "TRACE")
+                log_level=$LOG_TRACE;
                 ;;
-	    "DEBUG")
-		log_level=$LOG_DEBUG;
-		;;
-	    "CRITICAL")
-		log_level=$LOG_CRITICAL;
-		;;
-	    "WARNING")
-		log_level=$LOG_WARNING;
-		;;
-	    "TRACE")
-		log_level=$LOG_TRACE;
-		;;
-	    "NONE")
-		log_level=$LOG_NONE;
-		;;
-	    *)
-		echo "invalid log level $log_level_str, using NORMAL";
-		log_level=$LOG_NORMAL;
-		;;
-	esac
-    fi
-    cmd_line=$(echo "$cmd_line --log-level=$log_level");
+            "NONE")
+                log_level=$LOG_NONE;
+                ;;
+            *)
+                echo "invalid log level $log_level_str, using INFO";
+                log_level=$LOG_INFO;
+                ;;
+        esac
+    fi
 
+#options without values start here
     if [ -n "$read_only" ]; then
-	cmd_line=$(echo "$cmd_line --read-only");
+        cmd_line=$(echo "$cmd_line --read-only");
     fi
 
-    if [ -n "$log_file" ]; then
-	cmd_line=$(echo "$cmd_line --log-file=$log_file");
+    if [ -n "$acl" ]; then
+        cmd_line=$(echo "$cmd_line --acl");
+    fi
+
+    if [ -n "$selinux" ]; then
+        cmd_line=$(echo "$cmd_line --selinux");
+    fi
+
+    if [ -n "$enable_ino32" ]; then
+        cmd_line=$(echo "$cmd_line --enable-ino32");
+    fi
+
+    if [ -n "$worm" ]; then
+        cmd_line=$(echo "$cmd_line --worm");
+    fi
+
+    if [ -n "$fopen_keep_cache" ]; then
+        cmd_line=$(echo "$cmd_line --fopen-keep-cache");
     fi
 
     if [ -n "$volfile_check" ]; then
-	cmd_line=$(echo "$cmd_line --volfile-check");
+        cmd_line=$(echo "$cmd_line --volfile-check");
+    fi
+
+    if [ -n "$mem_accounting" ]; then
+        cmd_line=$(echo "$cmd_line --mem-accounting");
+    fi
+
+    if [ -n "$aux_gfid_mount" ]; then
+        cmd_line=$(echo "$cmd_line --aux-gfid-mount");
+    fi
+
+#options with values start here
+    if [ -n "$log_level" ]; then
+        cmd_line=$(echo "$cmd_line --log-level=$log_level");
+    fi
+
+    if [ -n "$log_file" ]; then
+        cmd_line=$(echo "$cmd_line --log-file=$log_file");
     fi
 
     if [ -n "$direct_io_mode" ]; then
-	cmd_line=$(echo "$cmd_line --disable-direct-io-mode");
+        cmd_line=$(echo "$cmd_line --direct-io-mode=$direct_io_mode");
+    fi
+
+    if [ -n "$use_readdirp" ]; then
+        cmd_line=$(echo "$cmd_line --use-readdirp=$use_readdirp");
     fi
 
     if [ -n "$volume_name" ]; then
         cmd_line=$(echo "$cmd_line --volume-name=$volume_name");
     fi
 
-    if [ -n "$log_server" ]; then
-        if [ -n "$log_server_port" ]; then
-            cmd_line=$(echo "$cmd_line \
---log-server=$log_server \
---log-server-port=$log_server_port");
-        fi
+    if [ -n "$attribute_timeout" ]; then
+        cmd_line=$(echo "$cmd_line --attribute-timeout=$attribute_timeout");
+    fi
+
+    if [ -n "$entry_timeout" ]; then
+        cmd_line=$(echo "$cmd_line --entry-timeout=$entry_timeout");
+    fi
+
+    if [ -n "$negative_timeout" ]; then
+        cmd_line=$(echo "$cmd_line --negative-timeout=$negative_timeout");
+    fi
+
+    if [ -n "$gid_timeout" ]; then
+        cmd_line=$(echo "$cmd_line --gid-timeout=$gid_timeout");
+    fi
+
+    if [ -n "$bg_qlen" ]; then
+        cmd_line=$(echo "$cmd_line --background-qlen=$bg_qlen");
+    fi
+
+    if [ -n "$cong_threshold" ]; then
+        cmd_line=$(echo "$cmd_line --congestion-threshold=$cong_threshold");
+    fi
+
+    if [ -n "$fuse_mountopts" ]; then
+        cmd_line=$(echo "$cmd_line --fuse-mountopts=$fuse_mountopts");
+    fi
+
+    if [ -n "$xlator_option" ]; then
+        xlator_option=$(echo $xlator_option | sed s/"xlator-option="/"--xlator-option "/g)
+        cmd_line=$(echo "$cmd_line $xlator_option");
     fi
 
+    # for rdma volume, we have to fetch volfile with '.rdma' added
+    # to volume name, so that it fetches the right client vol file
+    volume_id_rdma="";
+
     if [ -z "$volfile_loc" ]; then
         if  [ -n "$server_ip" ]; then
+
+            cmd_line=$(echo "$cmd_line --volfile-server=$server_ip");
+
+            if [ -n "$backup_volfile_servers" ]; then
+                servers=$(parse_backup_volfile_servers ${backup_volfile_servers})
+                for i in $(echo ${servers}); do
+                    cmd_line=$(echo "$cmd_line --volfile-server=$i");
+                done
+            fi
+
             if [ -n "$server_port" ]; then
                 cmd_line=$(echo "$cmd_line --volfile-server-port=$server_port");
             fi
-	    if [ -n "$transport" ]; then
+
+            if [ -n "$transport" ]; then
                 cmd_line=$(echo "$cmd_line --volfile-server-transport=$transport");
+                if [ "$transport" = "rdma" ]; then
+                    volume_id_rdma=".rdma";
+                fi
             fi
+
             if [ -n "$volume_id" ]; then
+                if [ -n "$volume_id_rdma" ]; then
+                    volume_id="$volume_id$volume_id_rdma";
+                fi
                 cmd_line=$(echo "$cmd_line --volfile-id=$volume_id");
             fi
-
-            if [ -n "$backupvolfile_server" ]; then
-                cmd_line1=$(echo "$cmd_line --volfile-server=$backupvolfile_server");
-            fi
-
-	    cmd_line=$(echo "$cmd_line --volfile-server=$server_ip");
         fi
     else
         cmd_line=$(echo "$cmd_line --volfile=$volfile_loc");
     fi
 
+    if [ -n "$fuse_mountopts" ]; then
+        cmd_line=$(echo "$cmd_line --fuse-mountopts=$fuse_mountopts");
+    fi
+
     cmd_line=$(echo "$cmd_line $mount_point");
+    err=0;
     $cmd_line;
 
-    # retry the failover
-    if [ $? != "0" ]; then
-        if [ -n "$cmd_line1" ]; then
-            cmd_line1=$(echo "$cmd_line1 $mount_point");
-            $cmd_line1
-        fi
+
+    inode=$( ${getinode} $mount_point 2>/dev/null);
+
+    # this is required if the stat returns error
+    if [ -z "$inode" ]; then
+        inode="0";
     fi
 
+    if [ $inode -ne 1 ]; then
+        err=1;
+    fi
+
+    if [ $err -eq "1" ]; then
+            echo "Mount failed. Please check the log file for more details."
+            umount $mount_point > /dev/null 2>&1;
+            exit 1;
+    fi
 }
 
 usage ()
@@ -144,63 +275,157 @@ mount.glusterfs --version"
 
 }
 
-main ()
+# check for recursive mounts. i.e, mounting over an existing brick
+check_recursive_mount ()
 {
-    helper=$(echo "$@" | sed -n 's/.*\--[ ]*\([^ ]*\).*/\1/p');
-
-    options=$(echo "$@" | sed -n 's/.*\-o[ ]*\([^ ]*\).*/\1/p');
-
-    new_log_level=$(echo "$options" | sed -n 's/.*log-level=\([^,]*\).*/\1/p');
-
-    [ -n "$new_log_level" ] && {
-	log_level_str="$new_log_level";
-    }
-
-    log_file=$(echo "$options" | sed -n 's/.*log-file=\([^,]*\).*/\1/p');
-
-    read_only=$(echo "$options" | sed -n 's/.*\(ro\)[^,]*.*/\1/p');
-
-    transport=$(echo "$options" | sed -n 's/.*transport=\([^,]*\).*/\1/p');
-
-    direct_io_mode=$(echo "$options" | sed -n 's/.*direct-io-mode=\([^,]*\).*/\1/p');
-
-    volume_name=$(echo "$options" | sed -n 's/.*volume-name=\([^,]*\).*/\1/p');
-
-    volume_id=$(echo "$options" | sed -n 's/.*volume_id=\([^,]*\).*/\1/p');
+    if [ $1 = "/" ]; then
+        echo Cannot mount over root;
+        exit 2;
+    fi
+    # GFID check first
+    # remove trailing / from mount point
+    mnt_dir=${1%/};
+
+    export PATH;
+    # check whether getfattr exists
+    which getfattr > /dev/null 2>&1;
+    if [ $? -ne 0 ]; then
+        return;
+    fi
 
-    volfile_check=$(echo "$options" | sed -n 's/.*volfile-check=\([^,]*\).*/\1/p');
+    getfattr -n trusted.gfid $mnt_dir 2>/dev/null | grep -iq "trusted.gfid=";
+    if [ $? -eq 0 ]; then
+        echo "ERROR: $mnt_dir is in use as a brick of a gluster volume";
+        exit 2;
+    fi
 
-    server_port=$(echo "$options" | sed -n 's/.*server-port=\([^,]*\).*/\1/p');
-    backupvolfile_server=$(echo "$options" | sed -n 's/.*backupvolfile-server=\([^,]*\).*/\1/p');
+    # check if the mount point is a brick's parent directory
+    GLUSTERD_WORKDIR="/var/lib/glusterd";
 
-    log_server=$(echo "$options" | sed -n 's/.*log-server=\([^,]*\).*/\1/p');
+    ls -L "$GLUSTERD_WORKDIR"/vols/*/bricks/* > /dev/null 2>&1;
+    if [ $? -ne 0 ]; then
+        return;
+    fi
 
-    log_server_port=$(echo "$options" | sed -n 's/.*log-server-port=\([^,]*\).*/\1/p');
+    brick_path=`grep ^path "$GLUSTERD_WORKDIR"/vols/*/bricks/* | cut -d "=" -f 2`;
+    root_inode=`${lgetinode} /`;
+    root_dev=`${lgetdev} /`;
+    mnt_inode=`${lgetinode} $mnt_dir`;
+    mnt_dev=`${lgetdev} $mnt_dir`;
+    for brick in "$brick_path";
+    do
+        # evaluate brick path to see if this is local, if non-local, skip iteration
+        ls $brick > /dev/null 2>&1;
+        if [ $? -ne 0 ]; then
+            continue;
+        fi
+        getfattr -n trusted.gfid "$brick" 2>/dev/null | grep -iq "trusted.gfid=";
+        if [ $? -ne 0 ]; then
+            continue;
+        else
+            # brick is local
+            while [ 1 ];
+            do
+                tmp_brick="$brick";
+                brick="$brick"/..;
+                brick_dev=`${lgetdev} $brick`;
+                brick_inode=`${lgetinode} $brick`;
+                if [ "$mnt_inode" -eq "$brick_inode" -a "$mnt_dev" -eq "$brick_dev" ]; then
+                    echo ERROR: $mnt_dir is a parent of the brick $tmp_brick;
+                    exit 2;
+                fi
+                [ "$root_inode" -ne "$brick_inode" -o "$root_dev" -ne "$brick_dev" ] || break;
+            done;
+        fi
+    done;
+}
 
-    volfile_loc="$1";
+main ()
+{
+    helper=$(echo "$@" | sed -n 's/.*\--[ ]*\([^ ]*\).*/\1/p');
+    in_opt="no"
+    pos_args=0
+    for opt in "$@"; do
+        if [ "$in_opt" = "yes" ]; then
+            for pair in $(echo "$opt" | tr "," " "); do
+                # Handle options without values.
+                case "$pair" in
+                    "ro")               read_only=1 ;;
+                    "acl")              acl=1 ;;
+                    "selinux")          selinux=1 ;;
+                    "worm")             worm=1 ;;
+                    "fopen-keep-cache") fopen_keep_cache=1 ;;
+                    "enable-ino32")     enable_ino32=1 ;;
+                    "mem-accounting")   mem_accounting=1;;
+                    "aux-gfid-mount")
+                        if [ `uname -s` = "Linux" ]; then
+                            aux_gfid_mount=1
+                        fi
+                        ;;
+                    # "mount -t glusterfs" sends this, but it's useless.
+                    "rw")       ;;
+                    # these ones are interpreted during system initialization
+                    "noauto")   ;;
+                    "_netdev")  ;;
+                    *)
+                        key=$(echo "$pair" | cut -f1 -d'=');
+                        value=$(echo "$pair" | cut -f2- -d'=');
+
+                        # Handle options with values.
+                        case "$key" in
+                            "log-level")        log_level_str=$value ;;
+                            "log-file")         log_file=$value ;;
+                            "transport")        transport=$value ;;
+                            "direct-io-mode")   direct_io_mode=$value ;;
+                            "volume-name")      volume_name=$value ;;
+                            "volume-id")        volume_id=$value ;;
+                            "volfile-check")    volfile_check=$value ;;
+                            "server-port")      server_port=$value ;;
+                            "attribute-timeout")
+                                attribute_timeout=$value ;;
+                            "entry-timeout")    entry_timeout=$value ;;
+                            "negative-timeout")    negative_timeout=$value ;;
+                            "gid-timeout")   gid_timeout=$value ;;
+                            "background-qlen") bg_qlen=$value ;;
+                            "backup-volfile-servers") backup_volfile_servers=$value ;;
+                            "congestion-threshold")  cong_threshold=$value ;;
+                            "xlator-option")  xlator_option=$xlator_option" "$pair ;;
+                            "fuse-mountopts")  fuse_mountopts=$value ;;
+                            "use-readdirp") use_readdirp=$value ;;
+                            *)
+                                # Passthru
+                                [ -z "$fuse_mountopts" ] || fuse_mountopts="$fuse_mountopts,"
+                                fuse_mountopts="$fuse_mountopts$pair"
+                                ;;
+                        esac
+                esac
+            done
+            in_opt="no"
+        elif [ "$opt" = "-o" ]; then
+            in_opt="yes"
+        else
+            case $pos_args in
+                0)      volfile_loc=$opt ;;
+                1)      mount_point=$opt ;;
+                *)      echo "extra arguments at end (ignored)" ;;
+            esac
+            pos_args=$((pos_args+1))
+        fi
+    done
+    if [ $in_opt = "yes" -o $pos_args -lt 2 ]; then
+        usage
+        exit 1
+    fi
 
     [ -r "$volfile_loc" ] || {
-	server_ip=$(echo "$volfile_loc" | sed -n 's/\([^\:]*\).*/\1/p');
+        server_ip=$(echo "$volfile_loc" | sed -n 's/\([a-zA-Z0-9:.\-]*\):.*/\1/p');
         test_str=$(echo "$volfile_loc" | sed -n 's/.*:\([^ ]*\).*/\1/p');
         [ -n "$test_str" ] && {
             volume_id="$test_str";
         }
-	volfile_loc="";
+        volfile_loc="";
     }
 
-    new_fs_options=$(echo "$options" | sed -e 's/[,]*log-file=[^,]*//' \
-        -e 's/[,]*log-level=[^,]*//' \
-        -e 's/[,]*volume-name=[^,]*//' \
-        -e 's/[,]*direct-io-mode=[^,]*//' \
-        -e 's/[,]*volfile-check=[^,]*//' \
-        -e 's/[,]*transport=[^,]*//' \
-        -e 's/[,]*backupvolfile-server=[^,]*//' \
-        -e 's/[,]*server-port=[^,]*//' \
-        -e 's/[,]*volume-id=[^,]*//' \
-        -e 's/[,]*log-server=[^,]*//' \
-        -e 's/[,]*ro[^,]*//' \
-        -e 's/[,]*log-server-port=[^,]*//');
-
     #
     [ -n "$helper" ] && {
         cmd_line=$(echo "$cmd_line --$helper");
@@ -208,27 +433,22 @@ main ()
         exit 0;
     }
 
-    mount_point=""
-    for arg in "$@"; do
-        [ -d "$arg" ] && {
-            mount_point=$arg
-        }
-    done
-
-    [ -z "$mount_point" ] && {
+    # No need to do a ! -d test, it is taken care while initializing the
+    # variable mount_point
+    [ -z "$mount_point" -o ! -d "$mount_point" ] && {
+        echo "ERROR: Mount point does not exist."
         usage;
         exit 0;
     }
 
     # Simple check to avoid multiple identical mounts
-    if grep -q " $mount_point fuse" /etc/mtab; then
+    if grep -q "[[:space:]+]${mount_point}[[:space:]+]fuse" $mounttab; then
         echo -n "$0: according to mtab, GlusterFS is already mounted on "
         echo "$mount_point"
-        sleep 1;
         exit 0;
     fi
 
-    fs_options=$(echo "$fs_options,$new_fs_options");
+    check_recursive_mount "$mount_point";
 
     # Append fuse.glusterfs to PRUNEFS variable in updatedb.conf(5). updatedb(8)
     # should not index files under GlusterFS, indexing will slow down GlusteFS
@@ -242,9 +462,6 @@ main ()
     }
 
     start_glusterfs;
-
-    sleep 3;
 }
 
 _init "$@" && main "$@";
-
diff --git a/xlators/mount/fuse/utils/mount_glusterfs.in b/xlators/mount/fuse/utils/mount_glusterfs.in
index 0f808c14b..b12b4e04e 100755
--- a/xlators/mount/fuse/utils/mount_glusterfs.in
+++ b/xlators/mount/fuse/utils/mount_glusterfs.in
@@ -25,11 +25,11 @@ _init ()
     LOG_CRITICAL=CRITICAL;
     LOG_ERROR=ERROR;
     LOG_WARNING=WARNING;
-    LOG_NORMAL=NORMAL;
+    LOG_INFO=INFO;
     LOG_DEBUG=DEBUG;
 
     # set default log level to ERROR
-    log_level=$LOG_NORMAL;
+    log_level=$LOG_INFO;
 }
 
 start_glusterfs ()
@@ -43,8 +43,8 @@ start_glusterfs ()
 	    "ERROR")
 		log_level=$LOG_ERROR;
 		;;
-            "NORMAL")
-                log_level=$LOG_NORMAL;
+            "INFO")
+                log_level=$LOG_INFO;
                 ;;
 	    "DEBUG")
 		log_level=$LOG_DEBUG;
@@ -59,8 +59,8 @@ start_glusterfs ()
 		log_level=$LOG_NONE;
 		;;
 	    *)
-		echo "invalid log level $log_level_str, using NORMAL";
-		log_level=$LOG_NORMAL;
+		echo "invalid log level $log_level_str, using INFO";
+		log_level=$LOG_INFO;
 		;;
 	esac
     fi
@@ -82,12 +82,10 @@ start_glusterfs ()
 	if [ -n "$transport" ]; then 
 	    cmd_line=$(echo "$cmd_line \
 --volfile-server=$server_ip \
---volfile-server-port=$server_port \
 --volfile-server-transport=$transport");
 	else
 	    cmd_line=$(echo "$cmd_line \
---volfile-server=$server_ip \
---volfile-server-port=$server_port");
+--volfile-server=$server_ip");
 	fi
     else
 	cmd_line=$(echo "$cmd_line --volfile=$volfile_loc");
@@ -167,19 +165,15 @@ main ()
 
     # TODO: use getopt. This is very much darwin specific
     volfile_loc="$1";
-    while [ "$volfile_loc" == "-o" ] ; do
+    while [ "$volfile_loc" = "-o" ] ; do
 	shift ;
 	shift ;
 	volfile_loc="$1";
     done
     
     [ -r "$volfile_loc" ] || {
-	server_ip=$(echo "$volfile_loc" | sed -n 's/\([^\:]*\).*/\1/p');
-	server_port=$(echo "$volfile_loc" | sed -n 's/.*:\([^ ]*\).*/\1/p');
-	[ -n "$server_port" ] || {
-	    server_port="6996";
-	}
-
+        server_ip=$(echo "$volfile_loc" | sed -n 's/\([a-zA-Z0-9:.\-]*\):.*/\1/p');
+        volume_id=$(echo "$volfile_loc" | sed -n 's/[a-zA-Z0-9:.\-]*:\(.*\)/\1/p');
 	volfile_loc="";
     }
     # following line is product of love towards sed
diff --git a/xlators/nfs/lib/src/auth-null.c b/xlators/nfs/lib/src/auth-null.c
deleted file mode 100644
index a98d47280..000000000
--- a/xlators/nfs/lib/src/auth-null.c
+++ /dev/null
@@ -1,72 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "rpcsvc.h"
-#include "list.h"
-#include "dict.h"
-
-
-int
-nfs_auth_null_request_init (rpcsvc_request_t *req, void *priv)
-{
-        if (!req)
-                return -1;
-
-        memset (req->cred.authdata, 0, RPCSVC_MAX_AUTH_BYTES);
-        req->cred.datalen = 0;
-
-        memset (req->verf.authdata, 0, RPCSVC_MAX_AUTH_BYTES);
-        req->verf.datalen = 0;
-
-        return 0;
-}
-
-int
-nfs_auth_null_authenticate (rpcsvc_request_t *req, void *priv)
-{
-        /* Always succeed. */
-        return RPCSVC_AUTH_ACCEPT;
-}
-
-rpcsvc_auth_ops_t nfs_auth_null_ops = {
-        .conn_init              = NULL,
-        .request_init           = nfs_auth_null_request_init,
-        .authenticate           = nfs_auth_null_authenticate
-};
-
-rpcsvc_auth_t nfs_rpcsvc_auth_null = {
-        .authname       = "AUTH_NULL",
-        .authnum        = AUTH_NULL,
-        .authops        = &nfs_auth_null_ops,
-        .authprivate    = NULL
-};
-
-
-rpcsvc_auth_t *
-nfs_rpcsvc_auth_null_init (rpcsvc_t *svc, dict_t *options)
-{
-        return &nfs_rpcsvc_auth_null;
-}
-
diff --git a/xlators/nfs/lib/src/auth-unix.c b/xlators/nfs/lib/src/auth-unix.c
deleted file mode 100644
index a3e9c83ba..000000000
--- a/xlators/nfs/lib/src/auth-unix.c
+++ /dev/null
@@ -1,92 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "rpcsvc.h"
-#include "list.h"
-#include "dict.h"
-#include "xdr-rpc.h"
-
-
-int
-nfs_auth_unix_request_init (rpcsvc_request_t *req, void *priv)
-{
-        if (!req)
-                return -1;
-        memset (req->verf.authdata, 0, RPCSVC_MAX_AUTH_BYTES);
-        req->verf.datalen = 0;
-        req->verf.flavour = AUTH_NULL;
-
-        return 0;
-}
-
-int
-nfs_auth_unix_authenticate (rpcsvc_request_t *req, void *priv)
-{
-        int                     ret = RPCSVC_AUTH_REJECT;
-        struct authunix_parms   aup;
-        char                    machname[MAX_MACHINE_NAME];
-
-        if (!req)
-                return ret;
-
-        ret = nfs_xdr_to_auth_unix_cred (req->cred.authdata, req->cred.datalen,
-                                         &aup, machname, req->auxgids);
-        if (ret == -1) {
-                ret = RPCSVC_AUTH_REJECT;
-                goto err;
-        }
-
-        req->uid = aup.aup_uid;
-        req->gid = aup.aup_gid;
-        req->auxgidcount = aup.aup_len;
-
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Auth Info: machine name: %s, uid: %d"
-                ", gid: %d", machname, req->uid, req->gid);
-        ret = RPCSVC_AUTH_ACCEPT;
-err:
-        return ret;
-}
-
-rpcsvc_auth_ops_t nfs_auth_unix_ops = {
-        .conn_init              = NULL,
-        .request_init           = nfs_auth_unix_request_init,
-        .authenticate           = nfs_auth_unix_authenticate
-};
-
-rpcsvc_auth_t nfs_rpcsvc_auth_unix = {
-        .authname       = "AUTH_UNIX",
-        .authnum        = AUTH_UNIX,
-        .authops        = &nfs_auth_unix_ops,
-        .authprivate    = NULL
-};
-
-
-rpcsvc_auth_t *
-nfs_rpcsvc_auth_unix_init (rpcsvc_t *svc, dict_t *options)
-{
-        return &nfs_rpcsvc_auth_unix;
-}
-
diff --git a/xlators/nfs/lib/src/msg-nfs3.c b/xlators/nfs/lib/src/msg-nfs3.c
deleted file mode 100644
index 9ad499364..000000000
--- a/xlators/nfs/lib/src/msg-nfs3.c
+++ /dev/null
@@ -1,554 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include <sys/uio.h>
-#include <rpc/rpc.h>
-#include <rpc/xdr.h>
-#include <sys/types.h>
-
-#include "xdr-nfs3.h"
-#include "msg-nfs3.h"
-#include "xdr-common.h"
-
-
-/* Decode the mount path from the network message in inmsg
- * into the memory referenced by outpath.iov_base.
- * The size allocated for outpath.iov_base is outpath.iov_len.
- * The size of the path extracted from the message is returned.
- */
-ssize_t
-xdr_to_mountpath (struct iovec outpath, struct iovec inmsg)
-{
-        XDR     xdr;
-        ssize_t ret = -1;
-        char    *mntpath = NULL;
-
-        if ((!outpath.iov_base) || (!inmsg.iov_base))
-                return -1;
-
-        xdrmem_create (&xdr, inmsg.iov_base, (unsigned int)inmsg.iov_len,
-                       XDR_DECODE);
-
-        mntpath = outpath.iov_base;
-        if (!xdr_dirpath (&xdr, (dirpath *)&mntpath)) {
-                ret = -1;
-                goto ret;
-        }
-
-        ret = nfs_xdr_decoded_length (xdr);
-
-ret:
-        return ret;
-}
-
-
-ssize_t
-nfs_xdr_serialize_generic (struct iovec outmsg, void *res, xdrproc_t proc)
-{
-        ssize_t ret = -1;
-        XDR     xdr;
-
-        if ((!outmsg.iov_base) || (!res) || (!proc))
-                return -1;
-
-        xdrmem_create (&xdr, outmsg.iov_base, (unsigned int)outmsg.iov_len,
-                       XDR_ENCODE);
-
-        if (!proc (&xdr, res)) {
-                ret = -1;
-                goto ret;
-        }
-
-        ret = nfs_xdr_encoded_length (xdr);
-
-ret:
-        return ret;
-}
-
-
-ssize_t
-nfs_xdr_to_generic (struct iovec inmsg, void *args, xdrproc_t proc)
-{
-        XDR     xdr;
-        ssize_t ret = -1;
-
-        if ((!inmsg.iov_base) || (!args) || (!proc))
-                return -1;
-
-        xdrmem_create (&xdr, inmsg.iov_base, (unsigned int)inmsg.iov_len,
-                       XDR_DECODE);
-
-        if (!proc (&xdr, args)) {
-                ret  = -1;
-                goto ret;
-        }
-
-        ret = nfs_xdr_decoded_length (xdr);
-ret:
-        return ret;
-}
-
-
-ssize_t
-nfs_xdr_to_generic_payload (struct iovec inmsg, void *args, xdrproc_t proc,
-                            struct iovec *pendingpayload)
-{
-        XDR     xdr;
-        ssize_t ret = -1;
-
-        if ((!inmsg.iov_base) || (!args) || (!proc))
-                return -1;
-
-        xdrmem_create (&xdr, inmsg.iov_base, (unsigned int)inmsg.iov_len,
-                       XDR_DECODE);
-
-        if (!proc (&xdr, args)) {
-                ret  = -1;
-                goto ret;
-        }
-
-        ret = nfs_xdr_decoded_length (xdr);
-
-        if (pendingpayload) {
-                pendingpayload->iov_base = nfs_xdr_decoded_remaining_addr (xdr);
-                pendingpayload->iov_len = nfs_xdr_decoded_remaining_len (xdr);
-        }
-
-ret:
-        return ret;
-}
-
-
-/* Translate the mountres3 structure in res into XDR format into memory
- * referenced by outmsg.iov_base.
- * Returns the number of bytes used in encoding into XDR format.
- */
-ssize_t
-xdr_serialize_mountres3 (struct iovec outmsg, mountres3 *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_mountres3);
-}
-
-
-ssize_t
-xdr_serialize_mountbody (struct iovec outmsg, mountbody *mb)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)mb,
-                                          (xdrproc_t)xdr_mountbody);
-}
-
-ssize_t
-xdr_serialize_mountlist (struct iovec outmsg, mountlist *ml)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)ml,
-                                          (xdrproc_t)xdr_mountlist);
-}
-
-
-ssize_t
-xdr_serialize_mountstat3 (struct iovec outmsg, mountstat3 *m)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)m,
-                                          (xdrproc_t)xdr_mountstat3);
-}
-
-
-ssize_t
-xdr_to_getattr3args (struct iovec inmsg, getattr3args *ga)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)ga,
-                                   (xdrproc_t)xdr_getattr3args);
-}
-
-
-ssize_t
-xdr_serialize_getattr3res (struct iovec outmsg, getattr3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_getattr3res);
-}
-
-
-ssize_t
-xdr_serialize_setattr3res (struct iovec outmsg, setattr3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_setattr3res);
-}
-
-
-ssize_t
-xdr_to_setattr3args (struct iovec inmsg, setattr3args *sa)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)sa,
-                                   (xdrproc_t)xdr_setattr3args);
-}
-
-
-ssize_t
-xdr_serialize_lookup3res (struct iovec outmsg, lookup3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_lookup3res);
-}
-
-
-ssize_t
-xdr_to_lookup3args (struct iovec inmsg, lookup3args *la)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)la,
-                                   (xdrproc_t)xdr_lookup3args);
-}
-
-
-ssize_t
-xdr_to_access3args (struct iovec inmsg, access3args *ac)
-{
-        return nfs_xdr_to_generic (inmsg,(void *)ac,
-                                   (xdrproc_t)xdr_access3args);
-}
-
-
-ssize_t
-xdr_serialize_access3res (struct iovec outmsg, access3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_access3res);
-}
-
-
-ssize_t
-xdr_to_readlink3args (struct iovec inmsg, readlink3args *ra)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)ra,
-                                   (xdrproc_t)xdr_readlink3args);
-}
-
-
-ssize_t
-xdr_serialize_readlink3res (struct iovec outmsg, readlink3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_readlink3res);
-}
-
-
-ssize_t
-xdr_to_read3args (struct iovec inmsg, read3args *ra)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)ra, (xdrproc_t)xdr_read3args);
-}
-
-
-ssize_t
-xdr_serialize_read3res (struct iovec outmsg, read3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_read3res);
-}
-
-ssize_t
-xdr_serialize_read3res_nocopy (struct iovec outmsg, read3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_read3res_nocopy);
-}
-
-
-ssize_t
-xdr_to_write3args (struct iovec inmsg, write3args *wa)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)wa,(xdrproc_t)xdr_write3args);
-}
-
-
-ssize_t
-xdr_to_write3args_nocopy (struct iovec inmsg, write3args *wa,
-                          struct iovec *payload)
-{
-        return nfs_xdr_to_generic_payload (inmsg, (void *)wa,
-                                           (xdrproc_t)xdr_write3args, payload);
-}
-
-
-ssize_t
-xdr_serialize_write3res (struct iovec outmsg, write3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_write3res);
-}
-
-
-ssize_t
-xdr_to_create3args (struct iovec inmsg, create3args *ca)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)ca,
-                                   (xdrproc_t)xdr_create3args);
-}
-
-
-ssize_t
-xdr_serialize_create3res (struct iovec outmsg, create3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_create3res);
-}
-
-
-ssize_t
-xdr_serialize_mkdir3res (struct iovec outmsg, mkdir3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_mkdir3res);
-}
-
-
-ssize_t
-xdr_to_mkdir3args (struct iovec inmsg, mkdir3args *ma)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)ma,
-                                   (xdrproc_t)xdr_mkdir3args);
-}
-
-
-ssize_t
-xdr_to_symlink3args (struct iovec inmsg, symlink3args *sa)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)sa,
-                                   (xdrproc_t)xdr_symlink3args);
-}
-
-
-ssize_t
-xdr_serialize_symlink3res (struct iovec outmsg, symlink3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_symlink3res);
-}
-
-
-ssize_t
-xdr_to_mknod3args (struct iovec inmsg, mknod3args *ma)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)ma,
-                                   (xdrproc_t)xdr_mknod3args);
-}
-
-
-ssize_t
-xdr_serialize_mknod3res (struct iovec outmsg, mknod3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_mknod3res);
-}
-
-
-ssize_t
-xdr_to_remove3args (struct iovec inmsg, remove3args *ra)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)ra,
-                                   (xdrproc_t)xdr_remove3args);
-}
-
-
-ssize_t
-xdr_serialize_remove3res (struct iovec outmsg, remove3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_remove3res);
-}
-
-
-ssize_t
-xdr_to_rmdir3args (struct iovec inmsg, rmdir3args *ra)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)ra,
-                                   (xdrproc_t)xdr_rmdir3args);
-}
-
-
-ssize_t
-xdr_serialize_rmdir3res (struct iovec outmsg, rmdir3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_rmdir3res);
-}
-
-
-ssize_t
-xdr_serialize_rename3res (struct iovec outmsg, rename3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_rename3res);
-}
-
-
-ssize_t
-xdr_to_rename3args (struct iovec inmsg, rename3args *ra)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)ra,
-                                   (xdrproc_t)xdr_rename3args);
-}
-
-
-ssize_t
-xdr_serialize_link3res (struct iovec outmsg, link3res *li)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)li,
-                                          (xdrproc_t)xdr_link3res);
-}
-
-
-ssize_t
-xdr_to_link3args (struct iovec inmsg, link3args *la)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)la, (xdrproc_t)xdr_link3args);
-}
-
-
-ssize_t
-xdr_to_readdir3args (struct iovec inmsg, readdir3args *rd)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)rd,
-                                   (xdrproc_t)xdr_readdir3args);
-}
-
-
-ssize_t
-xdr_serialize_readdir3res (struct iovec outmsg, readdir3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_readdir3res);
-}
-
-
-ssize_t
-xdr_to_readdirp3args (struct iovec inmsg, readdirp3args *rp)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)rp,
-                                   (xdrproc_t)xdr_readdirp3args);
-}
-
-
-ssize_t
-xdr_serialize_readdirp3res (struct iovec outmsg, readdirp3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_readdirp3res);
-}
-
-
-ssize_t
-xdr_to_fsstat3args (struct iovec inmsg, fsstat3args *fa)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)fa,
-                                   (xdrproc_t)xdr_fsstat3args);
-}
-
-
-ssize_t
-xdr_serialize_fsstat3res (struct iovec outmsg, fsstat3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_fsstat3res);
-}
-
-ssize_t
-xdr_to_fsinfo3args (struct iovec inmsg, fsinfo3args *fi)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)fi,
-                                   (xdrproc_t)xdr_fsinfo3args);
-}
-
-
-ssize_t
-xdr_serialize_fsinfo3res (struct iovec outmsg, fsinfo3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_fsinfo3res);
-}
-
-
-ssize_t
-xdr_to_pathconf3args (struct iovec inmsg, pathconf3args *pc)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)pc,
-                                   (xdrproc_t)xdr_pathconf3args);}
-
-
-ssize_t
-xdr_serialize_pathconf3res (struct iovec outmsg, pathconf3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_pathconf3res);
-}
-
-
-ssize_t
-xdr_to_commit3args (struct iovec inmsg, commit3args *ca)
-{
-        return nfs_xdr_to_generic (inmsg, (void *)ca,
-                                   (xdrproc_t)xdr_commit3args);
-}
-
-
-ssize_t
-xdr_serialize_commit3res (struct iovec outmsg, commit3res *res)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)res,
-                                          (xdrproc_t)xdr_commit3res);
-}
-
-
-ssize_t
-xdr_serialize_exports (struct iovec outmsg, exports *elist)
-{
-        XDR     xdr;
-        ssize_t  ret = -1;
-
-        if ((!outmsg.iov_base) || (!elist))
-                return -1;
-
-        xdrmem_create (&xdr, outmsg.iov_base, (unsigned int)outmsg.iov_len,
-                       XDR_ENCODE);
-
-        if (!xdr_exports (&xdr, elist))
-                goto ret;
-
-        ret = nfs_xdr_decoded_length (xdr);
-
-ret:
-        return ret;
-}
-
-
-ssize_t
-xdr_serialize_nfsstat3 (struct iovec outmsg, nfsstat3 *s)
-{
-        return nfs_xdr_serialize_generic (outmsg, (void *)s,
-                                          (xdrproc_t)xdr_nfsstat3);
-}
-
-
diff --git a/xlators/nfs/lib/src/msg-nfs3.h b/xlators/nfs/lib/src/msg-nfs3.h
deleted file mode 100644
index 047e8dfc8..000000000
--- a/xlators/nfs/lib/src/msg-nfs3.h
+++ /dev/null
@@ -1,186 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _MSG_NFS3_H_
-#define _MSG_NFS3_H_
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "xdr-nfs3.h"
-
-#include <sys/types.h>
-#include <sys/uio.h>
-
-extern ssize_t
-xdr_to_mountpath (struct iovec outpath, struct iovec inmsg);
-
-extern ssize_t
-xdr_serialize_mountres3 (struct iovec outmsg, mountres3 *res);
-
-extern ssize_t
-xdr_serialize_mountbody (struct iovec outmsg, mountbody *mb);
-
-extern ssize_t
-xdr_to_getattr3args (struct iovec inmsg, getattr3args *ga);
-
-extern ssize_t
-xdr_serialize_getattr3res (struct iovec outmsg, getattr3res *res);
-
-extern ssize_t
-xdr_serialize_setattr3res (struct iovec outmsg, setattr3res *res);
-
-extern ssize_t
-xdr_to_setattr3args (struct iovec inmsg, setattr3args *sa);
-
-extern ssize_t
-xdr_serialize_lookup3res (struct iovec outmsg, lookup3res *res);
-
-extern ssize_t
-xdr_to_lookup3args (struct iovec inmsg, lookup3args *la);
-
-extern ssize_t
-xdr_to_access3args (struct iovec inmsg, access3args *ac);
-
-extern ssize_t
-xdr_serialize_access3res (struct iovec outmsg, access3res *res);
-
-extern ssize_t
-xdr_to_readlink3args (struct iovec inmsg, readlink3args *ra);
-
-extern ssize_t
-xdr_serialize_readlink3res (struct iovec outmsg, readlink3res *res);
-
-extern ssize_t
-xdr_to_read3args (struct iovec inmsg, read3args *ra);
-
-extern ssize_t
-xdr_serialize_read3res (struct iovec outmsg, read3res *res);
-
-extern ssize_t
-xdr_serialize_read3res_nocopy (struct iovec outmsg, read3res *res);
-
-extern ssize_t
-xdr_to_write3args (struct iovec inmsg, write3args *wa);
-
-extern ssize_t
-xdr_to_write3args_nocopy (struct iovec inmsg, write3args *wa,
-                          struct iovec *payload);
-
-extern ssize_t
-xdr_serialize_write3res (struct iovec outmsg, write3res *res);
-
-extern ssize_t
-xdr_to_create3args (struct iovec inmsg, create3args *ca);
-
-extern ssize_t
-xdr_serialize_create3res (struct iovec outmsg, create3res *res);
-
-extern ssize_t
-xdr_serialize_mkdir3res (struct iovec outmsg, mkdir3res *res);
-
-extern ssize_t
-xdr_to_mkdir3args (struct iovec inmsg, mkdir3args *ma);
-
-extern ssize_t
-xdr_to_symlink3args (struct iovec inmsg, symlink3args *sa);
-
-extern ssize_t
-xdr_serialize_symlink3res (struct iovec outmsg, symlink3res *res);
-
-extern ssize_t
-xdr_to_mknod3args (struct iovec inmsg, mknod3args *ma);
-
-extern ssize_t
-xdr_serialize_mknod3res (struct iovec outmsg, mknod3res *res);
-
-extern ssize_t
-xdr_to_remove3args (struct iovec inmsg, remove3args *ra);
-
-extern ssize_t
-xdr_serialize_remove3res (struct iovec outmsg, remove3res *res);
-
-extern ssize_t
-xdr_to_rmdir3args (struct iovec inmsg, rmdir3args *ra);
-
-extern ssize_t
-xdr_serialize_rmdir3res (struct iovec outmsg, rmdir3res *res);
-
-extern ssize_t
-xdr_serialize_rename3res (struct iovec outmsg, rename3res *res);
-
-extern ssize_t
-xdr_to_rename3args (struct iovec inmsg, rename3args *ra);
-
-extern ssize_t
-xdr_serialize_link3res (struct iovec outmsg, link3res *li);
-
-extern ssize_t
-xdr_to_link3args (struct iovec inmsg, link3args *la);
-
-extern ssize_t
-xdr_to_readdir3args (struct iovec inmsg, readdir3args *rd);
-
-extern ssize_t
-xdr_serialize_readdir3res (struct iovec outmsg, readdir3res *res);
-
-extern ssize_t
-xdr_to_readdirp3args (struct iovec inmsg, readdirp3args *rp);
-
-extern ssize_t
-xdr_serialize_readdirp3res (struct iovec outmsg, readdirp3res *res);
-
-extern ssize_t
-xdr_to_fsstat3args (struct iovec inmsg, fsstat3args *fa);
-
-extern ssize_t
-xdr_serialize_fsstat3res (struct iovec outmsg, fsstat3res *res);
-
-extern ssize_t
-xdr_to_fsinfo3args (struct iovec inmsg, fsinfo3args *fi);
-
-extern ssize_t
-xdr_serialize_fsinfo3res (struct iovec outmsg, fsinfo3res *res);
-
-extern ssize_t
-xdr_to_pathconf3args (struct iovec inmsg, pathconf3args *pc);
-
-extern ssize_t
-xdr_serialize_pathconf3res (struct iovec outmsg, pathconf3res *res);
-
-extern ssize_t
-xdr_to_commit3args (struct iovec inmsg, commit3args *ca);
-
-extern ssize_t
-xdr_serialize_commit3res (struct iovec outmsg, commit3res *res);
-
-extern ssize_t
-xdr_serialize_exports (struct iovec outmsg, exports *elist);
-
-extern ssize_t
-xdr_serialize_mountlist (struct iovec outmsg, mountlist *ml);
-
-extern ssize_t
-xdr_serialize_mountstat3 (struct iovec outmsg, mountstat3 *m);
-
-extern ssize_t
-xdr_serialize_nfsstat3 (struct iovec outmsg, nfsstat3 *s);
-#endif
diff --git a/xlators/nfs/lib/src/rpc-socket.c b/xlators/nfs/lib/src/rpc-socket.c
deleted file mode 100644
index aacf7ff03..000000000
--- a/xlators/nfs/lib/src/rpc-socket.c
+++ /dev/null
@@ -1,360 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "rpc-socket.h"
-#include "rpcsvc.h"
-#include "dict.h"
-#include "logging.h"
-#include "byte-order.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/socket.h>
-#include <netdb.h>
-#include <netinet/in.h>
-#include <netinet/tcp.h>
-
-static int
-nfs_rpcsvc_socket_server_get_local_socket (int addrfam, char *listenhost,
-                                           uint16_t listenport,
-                                           struct sockaddr *addr,
-                                           socklen_t *addr_len)
-{
-        struct addrinfo         hints, *res = 0;
-        char                    service[NI_MAXSERV];
-        int                     ret = -1;
-
-        memset (service, 0, sizeof (service));
-        sprintf (service, "%d", listenport);
-
-        memset (&hints, 0, sizeof (hints));
-        addr->sa_family = hints.ai_family = addrfam;
-        hints.ai_socktype = SOCK_STREAM;
-        hints.ai_flags    = AI_ADDRCONFIG | AI_PASSIVE;
-
-        ret = getaddrinfo(listenhost, service, &hints, &res);
-        if (ret != 0) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR,
-                        "getaddrinfo failed for host %s, service %s (%s)",
-                        listenhost, service, gai_strerror (ret));
-                ret = -1;
-                goto err;
-        }
-
-        memcpy (addr, res->ai_addr, res->ai_addrlen);
-        *addr_len = res->ai_addrlen;
-
-        freeaddrinfo (res);
-        ret = 0;
-
-err:
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_socket_listen (int addrfam, char *listenhost, uint16_t listenport)
-{
-        int                     sock = -1;
-        struct sockaddr_storage sockaddr;
-        socklen_t               sockaddr_len;
-        int                     flags = 0;
-        int                     ret = -1;
-        int                     opt = 1;
-
-        ret = nfs_rpcsvc_socket_server_get_local_socket (addrfam, listenhost,
-                                                         listenport,
-                                                         SA (&sockaddr),
-                                                         &sockaddr_len);
-
-        if (ret == -1)
-                return ret;
-
-        sock = socket (SA (&sockaddr)->sa_family, SOCK_STREAM, 0);
-        if (sock == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "socket creation failed"
-                        " (%s)", strerror (errno));
-                goto err;
-        }
-
-        flags = fcntl (sock, F_GETFL);
-        if (flags == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "cannot get socket flags"
-                        " (%s)", strerror(errno));
-                goto close_err;
-        }
-
-        ret = fcntl (sock, F_SETFL, flags | O_NONBLOCK);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "cannot set socket "
-                        "non-blocking (%s)", strerror (errno));
-                goto close_err;
-        }
-
-        ret = setsockopt (sock, SOL_SOCKET, SO_REUSEADDR, &opt, sizeof (opt));
-        if (ret == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "setsockopt() for "
-                        "SO_REUSEADDR failed (%s)", strerror (errno));
-                goto close_err;
-        }
-
-        ret = bind (sock, (struct sockaddr *)&sockaddr, sockaddr_len);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "binding socket failed:"
-                        " %s", strerror (errno));
-                if (errno == EADDRINUSE)
-                        gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "Port is already"
-                                " in use");
-                goto close_err;
-        }
-
-        ret = listen (sock, 10);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "could not listen on"
-                        " socket (%s)", strerror (errno));
-                goto close_err;
-        }
-
-        return sock;
-
-close_err:
-        close (sock);
-        sock = -1;
-
-err:
-        return sock;
-}
-
-
-int
-nfs_rpcsvc_socket_accept (int listenfd)
-{
-        int                     new_sock = -1;
-        struct sockaddr_storage new_sockaddr = {0, };
-        socklen_t               addrlen = sizeof (new_sockaddr);
-        int                     flags = 0;
-        int                     ret = -1;
-        int                     on = 1;
-
-        new_sock = accept (listenfd, SA (&new_sockaddr), &addrlen);
-        if (new_sock == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR,"accept on socket failed");
-                goto err;
-        }
-
-        flags = fcntl (new_sock, F_GETFL);
-        if (flags == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "cannot get socket flags"
-                        " (%s)", strerror(errno));
-                goto close_err;
-        }
-
-        ret = fcntl (new_sock, F_SETFL, flags | O_NONBLOCK);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "cannot set socket "
-                        "non-blocking (%s)", strerror (errno));
-                goto close_err;
-        }
-
-#ifdef TCP_NODELAY
-        ret = setsockopt(new_sock, IPPROTO_TCP, TCP_NODELAY, &on, sizeof(on));
-        if (ret == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "cannot set no-delay "
-                        " socket option");
-        }
-#endif
-
-        return new_sock;
-
-close_err:
-        close (new_sock);
-        new_sock = -1;
-
-err:
-        return new_sock;
-}
-
-ssize_t
-nfs_rpcsvc_socket_read (int sockfd, char *readaddr, size_t readsize)
-{
-        ssize_t         dataread = 0;
-        ssize_t         readlen = -1;
-
-        if (!readaddr)
-                return -1;
-
-        while (readsize > 0) {
-                readlen = read (sockfd, readaddr, readsize);
-                if (readlen == -1) {
-                        if (errno != EAGAIN) {
-                                dataread = -1;
-                                break;
-                        } else
-                                break;
-                } else if (readlen == 0)
-                        break;
-
-                dataread += readlen;
-                readaddr += readlen;
-                readsize -= readlen;
-        }
-
-        return dataread;
-}
-
-
-ssize_t
-nfs_rpcsvc_socket_write (int sockfd, char *buffer, size_t size)
-{
-        size_t          writelen = -1;
-        ssize_t         written = 0;
-
-        if (!buffer)
-                return -1;
-
-        while (size > 0) {
-                writelen = write (sockfd, buffer, size);
-                if (writelen == -1) {
-                        if (errno != EAGAIN) {
-                                written = -1;
-                                break;
-                        } else
-                                break;
-                } else if (writelen == 0)
-                        break;
-
-                written += writelen;
-                size -= writelen;
-                buffer += writelen;
-        }
-
-        return written;
-}
-
-
-int
-nfs_rpcsvc_socket_peername (int sockfd, char *hostname, int hostlen)
-{
-        struct sockaddr         sa;
-        socklen_t               sl = sizeof (sa);
-        int                     ret = EAI_FAIL;
-
-        if (!hostname)
-                return ret;
-
-        ret = getpeername (sockfd, &sa, &sl);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "Failed to get peer name:"
-                        " %s", strerror (errno));
-                ret = EAI_FAIL;
-                goto err;
-        }
-
-        ret = getnameinfo (&sa, sl, hostname, hostlen, NULL, 0, 0);
-        if (ret != 0)
-                goto err;
-
-err:
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_socket_peeraddr (int sockfd, char *addrstr, int addrlen,
-                            struct sockaddr *returnsa, socklen_t sasize)
-{
-        struct sockaddr         sa;
-        int                     ret = EAI_FAIL;
-
-        if (returnsa)
-                ret = getpeername (sockfd, returnsa, &sasize);
-        else {
-                sasize = sizeof (sa);
-                ret = getpeername (sockfd, &sa, &sasize);
-        }
-
-        if (ret == -1) {
-                gf_log (GF_RPCSVC_SOCK, GF_LOG_ERROR, "Failed to get peer addr:"
-                        " %s", strerror (errno));
-                ret = EAI_FAIL;
-                goto err;
-        }
-
-        /* If caller did not specify a string into which the address can be
-         * stored, dont bother getting it.
-         */
-        if (!addrstr) {
-                ret = 0;
-                goto err;
-        }
-
-        if (returnsa)
-                ret = getnameinfo (returnsa, sasize, addrstr, addrlen, NULL, 0,
-                                   NI_NUMERICHOST);
-        else
-                ret = getnameinfo (&sa, sasize, addrstr, addrlen, NULL, 0,
-                                   NI_NUMERICHOST);
-
-err:
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_socket_block_tx (int sockfd)
-{
-
-        int     ret = -1;
-        int     on = 1;
-
-#ifdef TCP_CORK
-        ret = setsockopt(sockfd, IPPROTO_TCP, TCP_CORK, &on, sizeof(on));
-#endif
-
-#ifdef TCP_NOPUSH
-        ret = setsockopt(sockfd, IPPROTO_TCP, TCP_NOPUSH, &on, sizeof(on));
-#endif
-
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_socket_unblock_tx (int sockfd)
-{
-        int     ret = -1;
-        int     off = 0;
-
-#ifdef TCP_CORK
-        ret = setsockopt(sockfd, IPPROTO_TCP, TCP_CORK, &off, sizeof(off));
-#endif
-
-#ifdef TCP_NOPUSH
-        ret = setsockopt(sockfd, IPPROTO_TCP, TCP_NOPUSH, &off, sizeof(off));
-#endif
-        return ret;
-}
-
diff --git a/xlators/nfs/lib/src/rpc-socket.h b/xlators/nfs/lib/src/rpc-socket.h
deleted file mode 100644
index d95e5ddd6..000000000
--- a/xlators/nfs/lib/src/rpc-socket.h
+++ /dev/null
@@ -1,65 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _NFS_RPCSVC_SOCKET_H_
-#define _NFS_RPCSVC_SOCKET_H_
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "rpcsvc.h"
-#include "dict.h"
-#include "logging.h"
-#include "byte-order.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-
-#include <fcntl.h>
-#include <errno.h>
-#include <sys/socket.h>
-#include <netdb.h>
-
-#define SA(ptr)                 ((struct sockaddr *)ptr)
-#define GF_RPCSVC_SOCK          "rpc-socket"
-extern int
-nfs_rpcsvc_socket_listen (int addrfam, char *listenhost, uint16_t listenport);
-
-extern int
-nfs_rpcsvc_socket_accept (int listenfd);
-
-extern ssize_t
-nfs_rpcsvc_socket_read (int sockfd, char *readaddr, size_t readsize);
-
-extern ssize_t
-nfs_rpcsvc_socket_write (int sockfd, char *buffer, size_t size);
-
-extern int
-nfs_rpcsvc_socket_peername (int sockfd, char *hostname, int hostlen);
-
-extern int
-nfs_rpcsvc_socket_peeraddr (int sockfd, char *addrstr, int addrlen,
-                            struct sockaddr *returnsa, socklen_t sasize);
-extern int
-nfs_rpcsvc_socket_block_tx (int sockfd);
-
-extern int
-nfs_rpcsvc_socket_unblock_tx (int sockfd);
-#endif
diff --git a/xlators/nfs/lib/src/rpcsvc-auth.c b/xlators/nfs/lib/src/rpcsvc-auth.c
deleted file mode 100644
index 7ff2d793d..000000000
--- a/xlators/nfs/lib/src/rpcsvc-auth.c
+++ /dev/null
@@ -1,400 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#include "rpcsvc.h"
-#include "logging.h"
-#include "dict.h"
-
-extern rpcsvc_auth_t *
-nfs_rpcsvc_auth_null_init (rpcsvc_t *svc, dict_t *options);
-
-extern rpcsvc_auth_t *
-nfs_rpcsvc_auth_unix_init (rpcsvc_t *svc, dict_t *options);
-
-int
-nfs_rpcsvc_auth_add_initer (struct list_head *list, char *idfier,
-                            rpcsvc_auth_initer_t init)
-{
-        struct rpcsvc_auth_list         *new = NULL;
-
-        if ((!list) || (!init) || (!idfier))
-                return -1;
-
-        new = GF_CALLOC (1, sizeof (*new), gf_common_mt_rpcsvc_auth_list);
-        if (!new) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Memory allocation failed");
-                return -1;
-        }
-
-        new->init = init;
-        strcpy (new->name, idfier);
-        INIT_LIST_HEAD (&new->authlist);
-        list_add_tail (&new->authlist, list);
-        return 0;
-}
-
-
-
-int
-nfs_rpcsvc_auth_add_initers (rpcsvc_t *svc)
-{
-        int     ret = -1;
-
-        ret = nfs_rpcsvc_auth_add_initer (&svc->authschemes, "auth-unix",
-                                          (rpcsvc_auth_initer_t)
-                                          nfs_rpcsvc_auth_unix_init);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to add AUTH_UNIX");
-                goto err;
-        }
-
-        ret = nfs_rpcsvc_auth_add_initer (&svc->authschemes, "auth-null",
-                                          (rpcsvc_auth_initer_t)
-                                          nfs_rpcsvc_auth_null_init);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to add AUTH_NULL");
-                goto err;
-        }
-
-err:
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_auth_init_auth (rpcsvc_t *svc, dict_t *options,
-                           struct rpcsvc_auth_list *authitem)
-{
-        int             ret = -1;
-
-        if ((!svc) || (!options) || (!authitem))
-                return -1;
-
-        if (!authitem->init) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "No init function defined");
-                ret = -1;
-                goto err;
-        }
-
-        authitem->auth = authitem->init (svc, options);
-        if (!authitem->auth) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Registration of auth failed:"
-                        " %s", authitem->name);
-                ret = -1;
-                goto err;
-        }
-
-        authitem->enable = 1;
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Authentication enabled: %s",
-                authitem->auth->authname);
-
-        ret = 0;
-err:
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_auth_init_auths (rpcsvc_t *svc, dict_t *options)
-{
-        int                     ret = -1;
-        struct rpcsvc_auth_list *auth = NULL;
-        struct rpcsvc_auth_list *tmp = NULL;
-
-        if (!svc)
-                return -1;
-
-        if (list_empty (&svc->authschemes)) {
-                gf_log (GF_RPCSVC, GF_LOG_WARNING, "No authentication!");
-                ret = 0;
-                goto err;
-        }
-
-        /* If auth null and sys are not disabled by the user, we must enable
-         * it by default. This is a globally default rule, the user is still
-         * allowed to disable the two for particular subvolumes.
-         */
-        if (!dict_get (options, "rpc-auth.auth-null")) {
-                ret = dict_set_dynstr (options, "rpc-auth.auth-null", "on");
-                if (ret < 0) {
-                        gf_log (GF_RPCSVC, GF_LOG_ERROR,
-                                "Failed to set dict value.");
-                        goto err;
-                }
-        }
-
-        if (!dict_get (options, "rpc-auth.auth-unix")) {
-                ret = dict_set_dynstr (options, "rpc-auth.auth-unix", "on");
-                if (ret < 0) {
-                        gf_log (GF_RPCSVC, GF_LOG_ERROR,
-                                "Failed to set dict value.");
-                        goto err;
-                }
-        }
-
-        list_for_each_entry_safe (auth, tmp, &svc->authschemes, authlist) {
-                ret = nfs_rpcsvc_auth_init_auth (svc, options, auth);
-                if (ret == -1)
-                        goto err;
-        }
-
-        ret = 0;
-err:
-        return ret;
-
-}
-
-int
-nfs_rpcsvc_auth_init (rpcsvc_t *svc, dict_t *options)
-{
-        int             ret = -1;
-
-        if ((!svc) || (!options))
-                return -1;
-
-        ret = nfs_rpcsvc_auth_add_initers (svc);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to add initers");
-                goto out;
-        }
-
-        ret = nfs_rpcsvc_auth_init_auths (svc, options);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to init auth schemes");
-                goto out;
-        }
-
-out:
-        return ret;
-}
-
-
-rpcsvc_auth_t *
-__nfs_rpcsvc_auth_get_handler (rpcsvc_request_t *req)
-{
-        struct rpcsvc_auth_list *auth = NULL;
-        struct rpcsvc_auth_list *tmp = NULL;
-        rpcsvc_t                *svc = NULL;
-
-        if (!req)
-                return NULL;
-
-        svc = nfs_rpcsvc_request_service (req);
-        if (list_empty (&svc->authschemes)) {
-                gf_log (GF_RPCSVC, GF_LOG_WARNING, "No authentication!");
-                goto err;
-        }
-
-        list_for_each_entry_safe (auth, tmp, &svc->authschemes, authlist) {
-                if (!auth->enable)
-                        continue;
-                if (auth->auth->authnum == req->cred.flavour)
-                        goto err;
-
-        }
-
-        auth = NULL;
-err:
-        if (auth)
-                return auth->auth;
-        else
-                return NULL;
-}
-
-rpcsvc_auth_t *
-nfs_rpcsvc_auth_get_handler (rpcsvc_request_t *req)
-{
-        rpcsvc_auth_t           *auth = NULL;
-
-        auth = __nfs_rpcsvc_auth_get_handler (req);
-        if (auth)
-                goto ret;
-
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "No auth handler: %d",
-                req->cred.flavour);
-
-        /* The requested scheme was not available so fall back the to one
-         * scheme that will always be present.
-         */
-        req->cred.flavour = AUTH_NULL;
-        req->verf.flavour = AUTH_NULL;
-        auth = __nfs_rpcsvc_auth_get_handler (req);
-ret:
-        return auth;
-}
-
-
-int
-nfs_rpcsvc_auth_request_init (rpcsvc_request_t *req)
-{
-        int                     ret = -1;
-        rpcsvc_auth_t           *auth = NULL;
-
-        if (!req)
-                return -1;
-
-        auth = nfs_rpcsvc_auth_get_handler (req);
-        if (!auth)
-                goto err;
-        ret = 0;
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Auth handler: %s", auth->authname);
-        if (!auth->authops->request_init)
-                ret = auth->authops->request_init (req, auth->authprivate);
-
-err:
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_authenticate (rpcsvc_request_t *req)
-{
-        int                     ret = RPCSVC_AUTH_REJECT;
-        rpcsvc_auth_t           *auth = NULL;
-        int                     minauth = 0;
-
-        if (!req)
-                return ret;
-
-        minauth = nfs_rpcsvc_request_prog_minauth (req);
-        if (minauth > nfs_rpcsvc_request_cred_flavour (req)) {
-                gf_log (GF_RPCSVC, GF_LOG_DEBUG, "Auth too weak");
-                nfs_rpcsvc_request_set_autherr (req, AUTH_TOOWEAK);
-                goto err;
-        }
-
-        auth = nfs_rpcsvc_auth_get_handler (req);
-        if (!auth) {
-                gf_log (GF_RPCSVC, GF_LOG_DEBUG, "No auth handler found");
-                goto err;
-        }
-
-        if (auth->authops->authenticate)
-                ret = auth->authops->authenticate (req, auth->authprivate);
-
-err:
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_auth_array (rpcsvc_t *svc, char *volname, int *autharr, int arrlen)
-{
-        int             count = 0;
-        int             gen = RPCSVC_AUTH_REJECT;
-        int             spec = RPCSVC_AUTH_REJECT;
-        int             final = RPCSVC_AUTH_REJECT;
-        char            *srchstr = NULL;
-        char            *valstr = NULL;
-        gf_boolean_t    boolval = _gf_false;
-        int             ret = 0;
-
-        struct rpcsvc_auth_list *auth = NULL;
-        struct rpcsvc_auth_list *tmp = NULL;
-
-        if ((!svc) || (!autharr) || (!volname))
-                return -1;
-
-        memset (autharr, 0, arrlen * sizeof(int));
-        if (list_empty (&svc->authschemes)) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "No authentication!");
-                goto err;
-        }
-
-        list_for_each_entry_safe (auth, tmp, &svc->authschemes, authlist) {
-                if (count >= arrlen)
-                        break;
-
-                gen = gf_asprintf (&srchstr, "rpc-auth.%s", auth->name);
-                if (gen == -1) {
-                        count = -1;
-                        goto err;
-                }
-
-                gen = RPCSVC_AUTH_REJECT;
-                if (dict_get (svc->options, srchstr)) {
-                        ret = dict_get_str (svc->options, srchstr, &valstr);
-                        if (ret == 0) {
-                                ret = gf_string2boolean (valstr, &boolval);
-                                if (ret == 0) {
-                                        if (boolval == _gf_true)
-                                                gen = RPCSVC_AUTH_ACCEPT;
-                                } else
-                                        gf_log (GF_RPCSVC, GF_LOG_ERROR, "Faile"
-                                                "d to read auth val");
-                        } else
-                                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Faile"
-                                        "d to read auth val");
-                }
-
-                GF_FREE (srchstr);
-                spec = gf_asprintf (&srchstr, "rpc-auth.%s.%s", auth->name,
-                                    volname);
-                if (spec == -1) {
-                        count = -1;
-                        goto err;
-                }
-
-                spec = RPCSVC_AUTH_DONTCARE;
-                if (dict_get (svc->options, srchstr)) {
-                        ret = dict_get_str (svc->options, srchstr, &valstr);
-                        if (ret == 0) {
-                                ret = gf_string2boolean (valstr, &boolval);
-                                if (ret == 0) {
-                                        if (boolval == _gf_true)
-                                                spec = RPCSVC_AUTH_ACCEPT;
-                                        else
-                                                spec = RPCSVC_AUTH_REJECT;
-                                } else
-                                        gf_log (GF_RPCSVC, GF_LOG_ERROR, "Faile"
-                                                "d to read auth val");
-                        } else
-                                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Faile"
-                                        "d to read auth val");
-                }
-
-                GF_FREE (srchstr);
-                final = nfs_rpcsvc_combine_gen_spec_volume_checks (gen, spec);
-                if (final == RPCSVC_AUTH_ACCEPT) {
-                        autharr[count] = auth->auth->authnum;
-                        ++count;
-                }
-        }
-
-err:
-        return count;
-}
-
-
-gid_t *
-nfs_rpcsvc_auth_unix_auxgids (rpcsvc_request_t *req, int *arrlen)
-{
-        if ((!req) || (!arrlen))
-                return NULL;
-
-        if (req->cred.flavour != AUTH_UNIX)
-                return NULL;
-
-        *arrlen = req->auxgidcount;
-        if (*arrlen == 0)
-                return NULL;
-
-        return &req->auxgids[0];
-}
-
diff --git a/xlators/nfs/lib/src/rpcsvc.c b/xlators/nfs/lib/src/rpcsvc.c
deleted file mode 100644
index 7eaeb9233..000000000
--- a/xlators/nfs/lib/src/rpcsvc.c
+++ /dev/null
@@ -1,2803 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "rpcsvc.h"
-#include "rpc-socket.h"
-#include "dict.h"
-#include "logging.h"
-#include "byte-order.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-#include "list.h"
-#include "xdr-rpc.h"
-#include "iobuf.h"
-#include "globals.h"
-
-#include <errno.h>
-#include <pthread.h>
-#include <stdlib.h>
-#include <rpc/rpc.h>
-#include <rpc/pmap_clnt.h>
-#include <arpa/inet.h>
-#include <rpc/xdr.h>
-#include <fnmatch.h>
-#include <stdarg.h>
-#include <stdio.h>
-
-
-#define nfs_rpcsvc_alloc_request(con, request)                          \
-        do {                                                            \
-                request = (rpcsvc_request_t *) mem_get ((con)->rxpool); \
-                memset (request, 0, sizeof (rpcsvc_request_t));         \
-        } while (0)                                                     \
-
-/* The generic event handler for every stage */
-void *
-nfs_rpcsvc_stage_proc (void *arg)
-{
-        rpcsvc_stage_t          *stg = (rpcsvc_stage_t *)arg;
-
-        if (!stg)
-                return NULL;
-
-        event_dispatch (stg->eventpool);
-        return NULL;
-}
-
-
-rpcsvc_stage_t *
-nfs_rpcsvc_stage_init (rpcsvc_t *svc)
-{
-        rpcsvc_stage_t          *stg = NULL;
-        int                     ret = -1;
-        size_t                  stacksize = RPCSVC_THREAD_STACK_SIZE;
-        pthread_attr_t          stgattr;
-        unsigned int            eventpoolsize = 0;
-
-        if (!svc)
-                return NULL;
-
-        stg = GF_CALLOC (1, sizeof(*stg), gf_common_mt_rpcsvc_stage_t);
-        if (!stg)
-                return NULL;
-
-        eventpoolsize = svc->memfactor * RPCSVC_EVENTPOOL_SIZE_MULT;
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "event pool size: %d", eventpoolsize);
-        stg->eventpool = event_pool_new (eventpoolsize);
-        if (!stg->eventpool)
-                goto free_stg;
-
-        pthread_attr_init (&stgattr);
-        ret = pthread_attr_setstacksize (&stgattr, stacksize);
-        if (ret == EINVAL)
-                gf_log (GF_RPCSVC, GF_LOG_WARNING,
-                                "Using default thread stack size");
-
-        ret = pthread_create (&stg->tid, &stgattr, nfs_rpcsvc_stage_proc,
-                              (void *)stg);
-        if (ret != 0) {
-                ret = -1;
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Stage creation failed");
-                goto free_stg;
-        }
-
-        stg->svc = svc;
-        ret = 0;
-free_stg:
-        if (ret == -1) {
-                GF_FREE (stg);
-                stg = NULL;
-        }
-
-        return stg;
-}
-
-
-int
-nfs_rpcsvc_init_options (rpcsvc_t *svc, dict_t *options)
-{
-        svc->memfactor = RPCSVC_DEFAULT_MEMFACTOR;
-        return 0;
-}
-
-
-/* The global RPC service initializer.
- * Starts up the stages and then waits for RPC program registrations
- * to come in.
- */
-rpcsvc_t *
-nfs_rpcsvc_init (glusterfs_ctx_t *ctx, dict_t *options)
-{
-        rpcsvc_t        *svc = NULL;
-        int             ret = -1;
-
-        if ((!ctx) || (!options))
-                return NULL;
-
-        svc = GF_CALLOC (1, sizeof (*svc), gf_common_mt_rpcsvc_t);
-        if (!svc)
-                return NULL;
-
-        pthread_mutex_init (&svc->rpclock, NULL);
-        INIT_LIST_HEAD (&svc->stages);
-        INIT_LIST_HEAD (&svc->authschemes);
-
-        ret = nfs_rpcsvc_init_options (svc, options);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to init options");
-                goto free_svc;
-        }
-
-        ret = nfs_rpcsvc_auth_init (svc, options);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to init "
-                        "authentication");
-                goto free_svc;
-        }
-
-        ret = -1;
-        svc->defaultstage = nfs_rpcsvc_stage_init (svc);
-        if (!svc->defaultstage) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR,"RPC service init failed.");
-                goto free_svc;
-        }
-        svc->options = options;
-        svc->ctx = ctx;
-        gf_log (GF_RPCSVC, GF_LOG_DEBUG, "RPC service inited.");
-
-        ret = 0;
-free_svc:
-        if (ret == -1) {
-                GF_FREE (svc);
-                svc = NULL;
-        }
-
-        return svc;
-}
-
-
-/* Once multi-threaded support is complete, we'll be able to round-robin
- * the various incoming connections over the many available stages. This
- * function selects one from among all the stages.
- */
-rpcsvc_stage_t *
-nfs_rpcsvc_select_stage (rpcsvc_t *rpcservice)
-{
-        if (!rpcservice)
-                return NULL;
-
-        return rpcservice->defaultstage;
-}
-
-
-int
-nfs_rpcsvc_conn_peer_check_search (dict_t *options, char *pattern, char *clstr)
-{
-        int                     ret = -1;
-        char                    *addrtok = NULL;
-        char                    *addrstr = NULL;
-        char                    *svptr = NULL;
-
-        if ((!options) || (!clstr))
-                return -1;
-
-        if (!dict_get (options, pattern))
-                return -1;
-
-        ret = dict_get_str (options, pattern, &addrstr);
-        if (ret < 0) {
-                ret = -1;
-                goto err;
-        }
-
-        if (!addrstr) {
-                ret = -1;
-                goto err;
-        }
-
-        addrtok = strtok_r (addrstr, ",", &svptr);
-        while (addrtok) {
-
-                /* CASEFOLD not present on Solaris */
-#ifdef FNM_CASEFOLD
-                ret = fnmatch (addrtok, clstr, FNM_CASEFOLD);
-#else
-                ret = fnmatch (addrtok, clstr, 0);
-#endif
-                if (ret == 0)
-                        goto err;
-
-                addrtok = strtok_r (NULL, ",", &svptr);
-        }
-
-        ret = -1;
-err:
-
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_conn_peer_check_allow (dict_t *options, char *volname, char *clstr)
-{
-        int     ret = RPCSVC_AUTH_DONTCARE;
-        char    *srchstr = NULL;
-        char    globalrule[] = "rpc-auth.addr.allow";
-
-        if ((!options) || (!clstr))
-                return ret;
-
-        /* If volname is NULL, then we're searching for the general rule to
-         * determine the current address in clstr is allowed or not for all
-         * subvolumes.
-         */
-        if (volname) {
-                ret = gf_asprintf (&srchstr, "rpc-auth.addr.%s.allow", volname);
-                if (ret == -1) {
-                        gf_log (GF_RPCSVC, GF_LOG_ERROR, "asprintf failed");
-                        ret = RPCSVC_AUTH_DONTCARE;
-                        goto out;
-                }
-        } else
-                srchstr = globalrule;
-
-        ret = nfs_rpcsvc_conn_peer_check_search (options, srchstr, clstr);
-        if (volname)
-                GF_FREE (srchstr);
-
-        if (ret == 0)
-                ret = RPCSVC_AUTH_ACCEPT;
-        else
-                ret = RPCSVC_AUTH_DONTCARE;
-out:
-        return ret;
-}
-
-int
-nfs_rpcsvc_conn_peer_check_reject (dict_t *options, char *volname, char *clstr)
-{
-        int     ret = RPCSVC_AUTH_DONTCARE;
-        char    *srchstr = NULL;
-        char    generalrule[] = "rpc-auth.addr.reject";
-
-        if ((!options) || (!clstr))
-                return ret;
-
-        if (volname) {
-                ret = gf_asprintf (&srchstr, "rpc-auth.addr.%s.reject",
-                                   volname);
-                if (ret == -1) {
-                        gf_log (GF_RPCSVC, GF_LOG_ERROR, "asprintf failed");
-                        ret = RPCSVC_AUTH_REJECT;
-                        goto out;
-                }
-        } else
-                srchstr = generalrule;
-
-        ret = nfs_rpcsvc_conn_peer_check_search (options, srchstr, clstr);
-        if (volname)
-                GF_FREE (srchstr);
-
-        if (ret == 0)
-                ret = RPCSVC_AUTH_REJECT;
-        else
-                ret = RPCSVC_AUTH_DONTCARE;
-out:
-        return ret;
-}
-
-
-/* This function tests the results of the allow rule and the reject rule to
- * combine them into a single result that can be used to determine if the
- * connection should be allowed to proceed.
- * Heres the test matrix we need to follow in this function.
- *
- * A -  Allow, the result of the allow test. Never returns R.
- * R - Reject, result of the reject test. Never returns A.
- * Both can return D or dont care if no rule was given.
- *
- * | @allow | @reject | Result |
- * |    A   |   R     | R      |
- * |    D   |   D     | D      |
- * |    A   |   D     | A      |
- * |    D   |   R     | R      |
- */
-int
-nfs_rpcsvc_combine_allow_reject_volume_check (int allow, int reject)
-{
-        int     final = RPCSVC_AUTH_REJECT;
-
-        /* If allowed rule allows but reject rule rejects, we stay cautious
-         * and reject. */
-        if ((allow == RPCSVC_AUTH_ACCEPT) && (reject == RPCSVC_AUTH_REJECT))
-                final = RPCSVC_AUTH_REJECT;
-        /* if both are dont care, that is user did not specify for either allow
-         * or reject, we leave it up to the general rule to apply, in the hope
-         * that there is one.
-         */
-        else if ((allow == RPCSVC_AUTH_DONTCARE) &&
-                 (reject == RPCSVC_AUTH_DONTCARE))
-                final = RPCSVC_AUTH_DONTCARE;
-        /* If one is dont care, the other one applies. */
-        else if ((allow == RPCSVC_AUTH_ACCEPT) &&
-                 (reject == RPCSVC_AUTH_DONTCARE))
-                final = RPCSVC_AUTH_ACCEPT;
-        else if ((allow == RPCSVC_AUTH_DONTCARE) &&
-                 (reject == RPCSVC_AUTH_REJECT))
-                final = RPCSVC_AUTH_REJECT;
-
-        return final;
-}
-
-
-/* Combines the result of the general rule test against, the specific rule
- * to determine final permission for the client's address.
- *
- * | @gen   | @spec   | Result |
- * |    A   |   A     | A      |
- * |    A   |   R     | R      |
- * |    A   |   D     | A      |
- * |    D   |   A     | A      |
- * |    D   |   R     | R      |
- * |    D   |   D     | D      |
- * |    R   |   A     | A      |
- * |    R   |   D     | R      |
- * |    R   |   R     | R      |
- */
-int
-nfs_rpcsvc_combine_gen_spec_addr_checks (int gen, int spec)
-{
-        int     final = RPCSVC_AUTH_REJECT;
-
-        if ((gen == RPCSVC_AUTH_ACCEPT) && (spec == RPCSVC_AUTH_ACCEPT))
-                final = RPCSVC_AUTH_ACCEPT;
-        else if ((gen == RPCSVC_AUTH_ACCEPT) && (spec == RPCSVC_AUTH_REJECT))
-                final = RPCSVC_AUTH_REJECT;
-        else if ((gen == RPCSVC_AUTH_ACCEPT) && (spec == RPCSVC_AUTH_DONTCARE))
-                final = RPCSVC_AUTH_ACCEPT;
-        else if ((gen == RPCSVC_AUTH_DONTCARE) && (spec == RPCSVC_AUTH_ACCEPT))
-                final = RPCSVC_AUTH_ACCEPT;
-        else if ((gen == RPCSVC_AUTH_DONTCARE) && (spec == RPCSVC_AUTH_REJECT))
-                final = RPCSVC_AUTH_REJECT;
-        else if ((gen == RPCSVC_AUTH_DONTCARE) && (spec== RPCSVC_AUTH_DONTCARE))
-                final = RPCSVC_AUTH_DONTCARE;
-        else if ((gen == RPCSVC_AUTH_REJECT) && (spec == RPCSVC_AUTH_ACCEPT))
-                final = RPCSVC_AUTH_ACCEPT;
-        else if ((gen == RPCSVC_AUTH_REJECT) && (spec == RPCSVC_AUTH_DONTCARE))
-                final = RPCSVC_AUTH_REJECT;
-        else if ((gen == RPCSVC_AUTH_REJECT) && (spec == RPCSVC_AUTH_REJECT))
-                final = RPCSVC_AUTH_REJECT;
-
-        return final;
-}
-
-
-
-/* Combines the result of the general rule test against, the specific rule
- * to determine final test for the connection coming in for a given volume.
- *
- * | @gen   | @spec   | Result |
- * |    A   |   A     | A      |
- * |    A   |   R     | R      |
- * |    A   |   D     | A      |
- * |    D   |   A     | A      |
- * |    D   |   R     | R      |
- * |    D   |   D     | R      |, special case, we intentionally disallow this.
- * |    R   |   A     | A      |
- * |    R   |   D     | R      |
- * |    R   |   R     | R      |
- */
-int
-nfs_rpcsvc_combine_gen_spec_volume_checks (int gen, int spec)
-{
-        int     final = RPCSVC_AUTH_REJECT;
-
-        if ((gen == RPCSVC_AUTH_ACCEPT) && (spec == RPCSVC_AUTH_ACCEPT))
-                final = RPCSVC_AUTH_ACCEPT;
-        else if ((gen == RPCSVC_AUTH_ACCEPT) && (spec == RPCSVC_AUTH_REJECT))
-                final = RPCSVC_AUTH_REJECT;
-        else if ((gen == RPCSVC_AUTH_ACCEPT) && (spec == RPCSVC_AUTH_DONTCARE))
-                final = RPCSVC_AUTH_ACCEPT;
-        else if ((gen == RPCSVC_AUTH_DONTCARE) && (spec == RPCSVC_AUTH_ACCEPT))
-                final = RPCSVC_AUTH_ACCEPT;
-        else if ((gen == RPCSVC_AUTH_DONTCARE) && (spec == RPCSVC_AUTH_REJECT))
-                final = RPCSVC_AUTH_REJECT;
-        /* On no rule, we reject. */
-        else if ((gen == RPCSVC_AUTH_DONTCARE) && (spec== RPCSVC_AUTH_DONTCARE))
-                final = RPCSVC_AUTH_REJECT;
-        else if ((gen == RPCSVC_AUTH_REJECT) && (spec == RPCSVC_AUTH_ACCEPT))
-                final = RPCSVC_AUTH_ACCEPT;
-        else if ((gen == RPCSVC_AUTH_REJECT) && (spec == RPCSVC_AUTH_DONTCARE))
-                final = RPCSVC_AUTH_REJECT;
-        else if ((gen == RPCSVC_AUTH_REJECT) && (spec == RPCSVC_AUTH_REJECT))
-                final = RPCSVC_AUTH_REJECT;
-
-        return final;
-}
-
-
-int
-nfs_rpcsvc_conn_peer_check_name (dict_t *options, char *volname,
-                                 rpcsvc_conn_t *conn)
-{
-        int     ret = RPCSVC_AUTH_REJECT;
-        int     aret = RPCSVC_AUTH_REJECT;
-        int     rjret = RPCSVC_AUTH_REJECT;
-        char    clstr[RPCSVC_PEER_STRLEN];
-
-        if (!conn)
-                return ret;
-
-        ret = nfs_rpcsvc_conn_peername (conn, clstr, RPCSVC_PEER_STRLEN);
-        if (ret != 0) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to get remote addr: "
-                        "%s", gai_strerror (ret));
-                ret = RPCSVC_AUTH_REJECT;
-                goto err;
-        }
-
-        aret = nfs_rpcsvc_conn_peer_check_allow (options, volname, clstr);
-        rjret = nfs_rpcsvc_conn_peer_check_reject (options, volname, clstr);
-
-        ret = nfs_rpcsvc_combine_allow_reject_volume_check (aret, rjret);
-
-err:
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_conn_peer_check_addr (dict_t *options, char *volname,
-                                 rpcsvc_conn_t *conn)
-{
-        int     ret = RPCSVC_AUTH_REJECT;
-        int     aret = RPCSVC_AUTH_DONTCARE;
-        int     rjret = RPCSVC_AUTH_REJECT;
-        char    clstr[RPCSVC_PEER_STRLEN];
-
-        if (!conn)
-                return ret;
-
-        ret = nfs_rpcsvc_conn_peeraddr (conn, clstr, RPCSVC_PEER_STRLEN, NULL,
-                                        0);
-        if (ret != 0) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to get remote addr: "
-                        "%s", gai_strerror (ret));
-                ret = RPCSVC_AUTH_REJECT;
-                goto err;
-        }
-
-        aret = nfs_rpcsvc_conn_peer_check_allow (options, volname, clstr);
-        rjret = nfs_rpcsvc_conn_peer_check_reject (options, volname, clstr);
-
-        ret = nfs_rpcsvc_combine_allow_reject_volume_check (aret, rjret);
-err:
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_conn_check_volume_specific (dict_t *options, char *volname,
-                                       rpcsvc_conn_t *conn)
-{
-        int             namechk = RPCSVC_AUTH_REJECT;
-        int             addrchk = RPCSVC_AUTH_REJECT;
-        gf_boolean_t    namelookup = _gf_true;
-        char            *namestr = NULL;
-        int             ret = 0;
-
-        if ((!options) || (!volname) || (!conn))
-                return RPCSVC_AUTH_REJECT;
-
-        /* Enabled by default */
-        if ((dict_get (options, "rpc-auth.addr.namelookup"))) {
-                ret = dict_get_str (options, "rpc-auth.addr.namelookup"
-                                    , &namestr);
-                if (ret == 0)
-                        ret = gf_string2boolean (namestr, &namelookup);
-        }
-
-        /* We need two separate checks because the rules with addresses in them
-         * can be network addresses which can be general and names can be
-         * specific which will over-ride the network address rules.
-         */
-        if (namelookup)
-                namechk = nfs_rpcsvc_conn_peer_check_name (options, volname,
-                                                           conn);
-        addrchk = nfs_rpcsvc_conn_peer_check_addr (options, volname, conn);
-
-        if (namelookup)
-                ret = nfs_rpcsvc_combine_gen_spec_addr_checks (addrchk,
-                                                               namechk);
-        else
-                ret = addrchk;
-
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_conn_check_volume_general (dict_t *options, rpcsvc_conn_t *conn)
-{
-        int             addrchk = RPCSVC_AUTH_REJECT;
-        int             namechk = RPCSVC_AUTH_REJECT;
-        gf_boolean_t    namelookup = _gf_true;
-        char            *namestr = NULL;
-        int             ret = 0;
-
-        if ((!options) || (!conn))
-                return RPCSVC_AUTH_REJECT;
-
-        /* Enabled by default */
-        if ((dict_get (options, "rpc-auth.addr.namelookup"))) {
-                ret = dict_get_str (options, "rpc-auth.addr.namelookup"
-                                    , &namestr);
-                if (ret == 0)
-                        ret = gf_string2boolean (namestr, &namelookup);
-        }
-
-        /* We need two separate checks because the rules with addresses in them
-         * can be network addresses which can be general and names can be
-         * specific which will over-ride the network address rules.
-         */
-        if (namelookup)
-                namechk = nfs_rpcsvc_conn_peer_check_name (options, NULL, conn);
-        addrchk = nfs_rpcsvc_conn_peer_check_addr (options, NULL, conn);
-
-        if (namelookup)
-                ret = nfs_rpcsvc_combine_gen_spec_addr_checks (addrchk,
-                                                               namechk);
-        else
-                ret = addrchk;
-
-        return ret;
-}
-
-int
-nfs_rpcsvc_conn_peer_check (dict_t *options, char *volname, rpcsvc_conn_t *conn)
-{
-        int     general_chk = RPCSVC_AUTH_REJECT;
-        int     specific_chk = RPCSVC_AUTH_REJECT;
-
-        if ((!options) || (!volname) || (!conn))
-                return RPCSVC_AUTH_REJECT;
-
-        general_chk = nfs_rpcsvc_conn_check_volume_general (options, conn);
-        specific_chk = nfs_rpcsvc_conn_check_volume_specific (options, volname,
-                                                              conn);
-
-        return nfs_rpcsvc_combine_gen_spec_volume_checks (general_chk,
-                                                          specific_chk);
-}
-
-
-char *
-nfs_rpcsvc_volume_allowed (dict_t *options, char *volname)
-{
-        char    globalrule[] = "rpc-auth.addr.allow";
-        char    *srchstr = NULL;
-        char    *addrstr = NULL;
-        int     ret = -1;
-
-        if ((!options) || (!volname))
-                return NULL;
-
-        ret = gf_asprintf (&srchstr, "rpc-auth.addr.%s.allow", volname);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "asprintf failed");
-                goto out;
-        }
-
-        if (!dict_get (options, srchstr)) {
-                GF_FREE (srchstr);
-                srchstr = globalrule;
-                ret = dict_get_str (options, srchstr, &addrstr);
-        } else
-                ret = dict_get_str (options, srchstr, &addrstr);
-
-out:
-        return addrstr;
-}
-
-
-/* Initialize the core of a connection */
-rpcsvc_conn_t *
-nfs_rpcsvc_conn_init (rpcsvc_t *svc, rpcsvc_program_t *prog, int sockfd)
-{
-        rpcsvc_conn_t  *conn = NULL;
-        int             ret = -1;
-        unsigned int    poolcount = 0;
-
-        conn = GF_CALLOC (1, sizeof(*conn), gf_common_mt_rpcsvc_conn_t);
-        if (!conn) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "memory allocation failed");
-                return NULL;
-        }
-
-        conn->sockfd = sockfd;
-        conn->program = (void *)prog;
-        INIT_LIST_HEAD (&conn->txbufs);
-        poolcount = RPCSVC_POOLCOUNT_MULT * svc->memfactor;
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "tx pool: %d", poolcount);
-        conn->txpool = mem_pool_new (rpcsvc_txbuf_t, poolcount);
-        if (!conn->txpool) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "mem pool allocation failed");
-                goto free_conn;
-        }
-
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "rx pool: %d", poolcount);
-        conn->rxpool = mem_pool_new (rpcsvc_request_t, poolcount);
-        if (!conn->rxpool) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "mem pool allocation failed");
-                goto free_txp;
-        }
-
-        /* Cannot consider a connection connected unless the user of this
-         * connection decides it is ready to use. It is possible that we have
-         * to free this connection soon after. That free will not happpen
-         * unless the state is disconnected.
-         */
-        conn->connstate = RPCSVC_CONNSTATE_DISCONNECTED;
-        pthread_mutex_init (&conn->connlock, NULL);
-        conn->connref = 0;
-        gf_log (GF_RPCSVC, GF_LOG_DEBUG, "New connection inited: sockfd: %d",
-                sockfd);
-
-        ret = 0;
-free_txp:
-        if (ret == -1)
-                mem_pool_destroy (conn->txpool);
-
-free_conn:
-        if (ret == -1) {
-                GF_FREE (conn);
-                conn = NULL;
-        }
-
-        return conn;
-}
-
-
-void
-nfs_rpcsvc_conn_destroy (rpcsvc_conn_t *conn)
-{
-        mem_pool_destroy (conn->txpool);
-        mem_pool_destroy (conn->rxpool);
-
-        if (conn->program->conn_destroy)
-                conn->program->conn_destroy (conn->program->private, conn);
-
-        /* Need to destory record state, txlists etc. */
-        GF_FREE (conn);
-        gf_log (GF_RPCSVC, GF_LOG_DEBUG, "Connection destroyed");
-}
-
-
-int
-__nfs_rpcsvc_conn_unref (rpcsvc_conn_t *conn)
-{
-        --conn->connref;
-        return conn->connref;
-}
-
-
-void
-__nfs_rpcsvc_conn_deinit (rpcsvc_conn_t *conn)
-{
-        if (!conn)
-                return;
-
-        if ((conn->stage) && (conn->stage->eventpool)) {
-                event_unregister (conn->stage->eventpool, conn->sockfd,
-                                  conn->eventidx);
-        }
-
-        if (nfs_rpcsvc_conn_check_active (conn)) {
-                gf_log (GF_RPCSVC, GF_LOG_DEBUG, "Connection de-activated:"
-                        " sockfd: %d", conn->sockfd);
-                conn->connstate = RPCSVC_CONNSTATE_DISCONNECTED;
-        }
-
-        if (conn->sockfd != -1) {
-                close (conn->sockfd);
-                conn->sockfd = -1;
-        }
-}
-
-
-void
-nfs_rpcsvc_conn_deinit (rpcsvc_conn_t *conn)
-{
-        int ref = 0;
-
-        if (!conn)
-                return;
-
-        pthread_mutex_lock (&conn->connlock);
-        {
-                __nfs_rpcsvc_conn_deinit (conn);
-                ref = __nfs_rpcsvc_conn_unref (conn);
-        }
-        pthread_mutex_unlock (&conn->connlock);
-
-        if (ref == 0)
-                nfs_rpcsvc_conn_destroy (conn);
-
-        return;
-}
-
-
-void
-nfs_rpcsvc_conn_unref (rpcsvc_conn_t *conn)
-{
-        int ref = 0;
-        if (!conn)
-                return;
-
-        pthread_mutex_lock (&conn->connlock);
-        {
-                ref = __nfs_rpcsvc_conn_unref (conn);
-        }
-        pthread_mutex_unlock (&conn->connlock);
-
-        if (ref == 0)
-                nfs_rpcsvc_conn_destroy (conn);
-}
-
-
-int
-nfs_rpcsvc_conn_active (rpcsvc_conn_t *conn)
-{
-        int     status = 0;
-
-        if (!conn)
-                return 0;
-
-        pthread_mutex_lock (&conn->connlock);
-        {
-                status = nfs_rpcsvc_conn_check_active (conn);
-        }
-        pthread_mutex_unlock (&conn->connlock);
-
-        return status;
-}
-
-
-
-void
-nfs_rpcsvc_conn_ref (rpcsvc_conn_t *conn)
-{
-        if (!conn)
-                return;
-
-        pthread_mutex_lock (&conn->connlock);
-        {
-                ++conn->connref;
-        }
-        pthread_mutex_unlock (&conn->connlock);
-
-        return;
-}
-
-
-void
-nfs_rpcsvc_conn_state_init (rpcsvc_conn_t *conn)
-{
-        if (!conn)
-                return;
-
-        ++conn->connref;
-        conn->connstate = RPCSVC_CONNSTATE_CONNECTED;
-}
-
-/* Builds a rpcsvc_conn_t with the aim of listening on it.
- */
-rpcsvc_conn_t *
-nfs_rpcsvc_conn_listen_init (rpcsvc_t *svc, rpcsvc_program_t *newprog)
-{
-        rpcsvc_conn_t  *conn = NULL;
-        int             sock = -1;
-
-        if (!newprog)
-                return NULL;
-
-        sock = nfs_rpcsvc_socket_listen (newprog->progaddrfamily,
-                                         newprog->proghost, newprog->progport);
-        if (sock == -1)
-                goto err;
-
-        conn = nfs_rpcsvc_conn_init (svc, newprog, sock);
-        if (!conn)
-                goto sock_close_err;
-
-        nfs_rpcsvc_conn_state_init (conn);
-sock_close_err:
-        if (!conn)
-                close (sock);
-
-err:
-        return conn;
-}
-
-void
-nfs_rpcsvc_record_init (rpcsvc_record_state_t *rs, struct iobuf_pool *pool)
-{
-        if (!rs)
-                return;
-
-        rs->state = RPCSVC_READ_FRAGHDR;
-        rs->vecstate = 0;
-        rs->remainingfraghdr = RPCSVC_FRAGHDR_SIZE;
-        rs->remainingfrag = 0;
-        rs->fragsize = 0;
-        rs->recordsize = 0;
-        rs->islastfrag = 0;
-
-        /* If the rs preserves a ref to the iob used by the previous request,
-         * we must unref it here to prevent memory leak.
-         * If program actor wanted to keep that memory around, it should've
-         * refd it on entry into the actor.
-         */
-        if (rs->activeiob)
-                iobuf_unref (rs->activeiob);
-
-        if (rs->vectoriob) {
-                iobuf_unref (rs->vectoriob);
-                rs->vectoriob = NULL;
-        }
-
-        rs->activeiob = iobuf_get (pool);
-        rs->fragcurrent = iobuf_ptr (rs->activeiob);
-
-        memset (rs->fragheader, 0, RPCSVC_FRAGHDR_SIZE);
-        rs->hdrcurrent = &rs->fragheader[0];
-
-}
-
-
-int
-nfs_rpcsvc_conn_privport_check (rpcsvc_t *svc, char *volname,
-                                rpcsvc_conn_t *conn)
-{
-        struct sockaddr_in      sa;
-        int                     ret = RPCSVC_AUTH_REJECT;
-        socklen_t               sasize = sizeof (sa);
-        char                    *srchstr = NULL;
-        char                    *valstr = NULL;
-        int                     globalinsecure = RPCSVC_AUTH_REJECT;
-        int                     exportinsecure = RPCSVC_AUTH_DONTCARE;
-        uint16_t                port = 0;
-        gf_boolean_t            insecure = _gf_false;
-
-        if ((!svc) || (!volname) || (!conn))
-                return ret;
-
-        ret = nfs_rpcsvc_conn_peeraddr (conn, NULL, 0, (struct sockaddr *)&sa,
-                                        sasize);
-        if (ret != 0) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to get peer addr: %s",
-                        gai_strerror (ret));
-                ret = RPCSVC_AUTH_REJECT;
-                goto err;
-        }
-
-        port = ntohs (sa.sin_port);
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Client port: %d", (int)port);
-        /* If the port is already a privileged one, dont bother with checking
-         * options.
-         */
-        if (port <= 1024) {
-                ret = RPCSVC_AUTH_ACCEPT;
-                goto err;
-        }
-
-        /* Disabled by default */
-        if ((dict_get (svc->options, "rpc-auth.ports.insecure"))) {
-                ret = dict_get_str (svc->options, "rpc-auth.ports.insecure"
-                                    , &srchstr);
-                if (ret == 0) {
-                        ret = gf_string2boolean (srchstr, &insecure);
-                        if (ret == 0) {
-                                if (insecure == _gf_true)
-                                        globalinsecure = RPCSVC_AUTH_ACCEPT;
-                        } else
-                                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to"
-                                        " read rpc-auth.ports.insecure value");
-                } else
-                        gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to"
-                                " read rpc-auth.ports.insecure value");
-        }
-
-        /* Disabled by default */
-        ret = gf_asprintf (&srchstr, "rpc-auth.ports.%s.insecure", volname);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "asprintf failed");
-                ret = RPCSVC_AUTH_REJECT;
-                goto err;
-        }
-
-        if (dict_get (svc->options, srchstr)) {
-                ret = dict_get_str (svc->options, srchstr, &valstr);
-                if (ret == 0) {
-                        ret = gf_string2boolean (srchstr, &insecure);
-                        if (ret == 0) {
-                                if (insecure == _gf_true)
-                                       exportinsecure = RPCSVC_AUTH_ACCEPT;
-                                else
-                                        exportinsecure = RPCSVC_AUTH_REJECT;
-                        } else
-                                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to"
-                                        " read rpc-auth.ports.insecure value");
-                } else
-                        gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to"
-                                " read rpc-auth.ports.insecure value");
-        }
-
-        ret = nfs_rpcsvc_combine_gen_spec_volume_checks (globalinsecure,
-                                                         exportinsecure);
-        if (ret == RPCSVC_AUTH_ACCEPT)
-                gf_log (GF_RPCSVC, GF_LOG_DEBUG, "Unprivileged port allowed");
-        else
-                gf_log (GF_RPCSVC, GF_LOG_DEBUG, "Unprivileged port not"
-                        " allowed");
-
-err:
-        return ret;
-}
-
-/* Inits a rpcsvc_conn_t after accepting the connection.
- */
-rpcsvc_conn_t *
-nfs_rpcsvc_conn_accept_init (rpcsvc_t *svc, int listenfd,
-                             rpcsvc_program_t *destprog)
-{
-        rpcsvc_conn_t   *newconn = NULL;
-        int             sock = -1;
-        int             ret = -1;
-
-        sock = nfs_rpcsvc_socket_accept (listenfd);
-        if (sock == -1)
-                goto err;
-
-        newconn = nfs_rpcsvc_conn_init (svc, destprog, sock);
-        if (!newconn) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to init conn object");
-                ret = -1;
-                goto err;
-        }
-
-        nfs_rpcsvc_record_init (&newconn->rstate, svc->ctx->iobuf_pool);
-        nfs_rpcsvc_conn_state_init (newconn);
-        if (destprog->conn_init)
-                destprog->conn_init (destprog->private, newconn);
-        ret = 0;
-
-err:
-        if (ret == -1)
-                close (sock);
-
-        return newconn;
-}
-
-
-/* Once the connection has been created, we need to associate it with
- * a stage so that the selected stage will handle the event on this connection.
- * This function also allows the caller to decide which handler should
- * be executed in the context of the stage, and also which specific events
- * should be handed to the handler when running in this particular stage.
- */
-int
-nfs_rpcsvc_stage_conn_associate (rpcsvc_stage_t *stg, rpcsvc_conn_t *conn,
-                                 event_handler_t handler, void *data)
-{
-        int     ret = -1;
-
-        if ((!stg) || (!conn))
-                return -1;
-
-        conn->stage = stg;
-        conn->eventidx = event_register (stg->eventpool, conn->sockfd, handler,
-                                         data, 1, 0);
-        if (conn->eventidx == -1)
-                goto err;
-
-        ret = 0;
-err:
-        return ret;
-}
-
-
-/* Depending on the state we're in, return the size of the next read request. */
-size_t
-nfs_rpcsvc_record_read_size (rpcsvc_record_state_t *rs)
-{
-        size_t  toread = -1;
-
-        if (!rs)
-                return -1;
-
-        if (nfs_rpcsvc_record_readfraghdr (rs))
-                toread = rs->remainingfraghdr;
-        else if (nfs_rpcsvc_record_readfrag (rs))
-                toread = rs->remainingfrag;
-        else
-                toread = RPCSVC_CONN_READ;
-
-        return toread;
-}
-
-
-uint32_t
-nfs_rpcsvc_record_extract_fraghdr (char *fraghdr)
-{
-        uint32_t        hdr = 0;
-        if (!fraghdr)
-                return 0;
-
-        memcpy ((void *)&hdr, fraghdr, sizeof (hdr));
-
-        hdr = ntohl (hdr);
-        return hdr;
-}
-
-
-ssize_t
-nfs_rpcsvc_record_read_complete_fraghdr (rpcsvc_record_state_t *rs,
-                                         ssize_t dataread)
-{
-        uint32_t        remhdr = 0;
-        char            *fraghdrstart = NULL;
-        uint32_t        fraghdr = 0;
-
-        fraghdrstart = &rs->fragheader[0];
-        remhdr = rs->remainingfraghdr;
-        fraghdr = nfs_rpcsvc_record_extract_fraghdr (fraghdrstart);
-        rs->fragsize = RPCSVC_FRAGSIZE (fraghdr);
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Received fragment size: %d",
-                rs->fragsize);
-        if (nfs_rpcsvc_record_vectored (rs)) {
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored RPC header,"
-                        " remaining: %d", RPCSVC_BARERPC_MSGSZ);
-                rs->remainingfrag = RPCSVC_BARERPC_MSGSZ;
-        } else {
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Regular RPC header,"
-                        " remaining: %d", rs->fragsize);
-                rs->remainingfrag = rs->fragsize;
-        }
-
-        rs->state = RPCSVC_READ_FRAG;
-        dataread -= remhdr;
-        rs->remainingfraghdr -= remhdr;
-        rs->islastfrag = RPCSVC_LASTFRAG (fraghdr);
-
-        return dataread;
-}
-
-
-ssize_t
-nfs_rpcsvc_record_read_partial_fraghdr (rpcsvc_record_state_t *rs,
-                                        ssize_t dataread)
-{
-
-        /* In case we got less than even the remaining header size,
-         * we need to consume it all and wait for remaining frag hdr
-         * bytes to come in.
-         */
-         rs->remainingfraghdr -= dataread;
-         nfs_rpcsvc_record_update_currenthdr (rs, dataread);
-         dataread = 0;
-         gf_log (GF_RPCSVC, GF_LOG_TRACE, "Fragment header remaining: %d",
-                 rs->remainingfraghdr);
-
-         return dataread;
-}
-
-
-ssize_t
-nfs_rpcsvc_record_update_fraghdr (rpcsvc_record_state_t *rs, ssize_t dataread)
-{
-        if ((!rs) || (dataread <= 0))
-                return -1;
-
-        /* Why are we even here, we're not supposed to be in the fragment
-         * header processing state.
-         */
-        if (!nfs_rpcsvc_record_readfraghdr(rs)) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "record state inconsistent"
-                        ": request to update frag header when state is not"
-                        "RPCSVC_READ_FRAGHDR");
-                return -1;
-        }
-
-        /* Again, if header has been read then the state member above should've
-         * been different, this is crazy. We should not be here.
-         */
-        if (rs->remainingfraghdr == 0) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "record state inconsistent"
-                        ": request to update frag header when frag header"
-                        "remaining is 0.");
-                return -1;
-        }
-
-        /* We've definitely got the full header now and may be even more. */
-        if (dataread >= rs->remainingfraghdr)
-                dataread = nfs_rpcsvc_record_read_complete_fraghdr (rs,
-                                                                    dataread);
-        else
-                dataread = nfs_rpcsvc_record_read_partial_fraghdr (rs,
-                                                                   dataread);
-
-        return dataread;
-}
-
-ssize_t
-nfs_rpcsvc_record_read_complete_frag (rpcsvc_record_state_t *rs,
-                                      ssize_t dataread)
-{
-        uint32_t        remfrag;
-
-        /* Since the frag is now complete, change the state to the next
-         * one, i.e. to read the header of the next fragment.
-         */
-        remfrag = rs->remainingfrag;
-        rs->state = RPCSVC_READ_FRAGHDR;
-        dataread -= remfrag;
-
-        /* This will be 0 now. */
-        rs->remainingfrag -= remfrag;
-
-        /* Now that the fragment is complete, we must update the
-         * record size. Recall that fragsize was got from the frag
-         * header.
-         */
-        rs->recordsize += rs->fragsize;
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Fragment remaining: %d",
-                rs->remainingfrag);
-
-        return dataread;
-}
-
-
-ssize_t
-nfs_rpcsvc_record_read_partial_frag (rpcsvc_record_state_t *rs,
-                                     ssize_t dataread)
-{
-        /* Just take whatever has come through the current network buffer. */
-        rs->remainingfrag -= dataread;
-
-        nfs_rpcsvc_record_update_currentfrag (rs, dataread);
-        /* Since we know we're consuming the whole buffer from dataread
-         * simply setting to 0 zero is fine.
-         */
-        dataread = 0;
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Fragment remaining: %d",
-                rs->remainingfrag);
-        return dataread;
-}
-
-
-ssize_t
-nfs_rpcsvc_record_update_frag (rpcsvc_record_state_t *rs, ssize_t dataread)
-{
-        if ((!rs) || (dataread <= 0))
-                return -1;
-
-        if (!nfs_rpcsvc_record_readfrag (rs)) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "record state inconsistent"
-                        ": request to update fragment when record state is not"
-                        "RPCSVC_READ_FRAG.");
-                return -1;
-        }
-
-        if (rs->remainingfrag == 0) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "record state inconsistent"
-                        ": request to update fragment when there is no fragment"
-                        " data remaining to be read.");
-                return -1;
-        }
-
-        /* We've read in more data than the current fragment requires. */
-        if (dataread >= rs->remainingfrag)
-                dataread = nfs_rpcsvc_record_read_complete_frag (rs, dataread);
-        else
-                dataread = nfs_rpcsvc_record_read_partial_frag (rs, dataread);
-
-        return dataread;
-}
-
-
-/* This needs to change to returning errors, since
- * we need to return RPC specific error messages when some
- * of the pointers below are NULL.
- */
-rpcsvc_actor_t *
-nfs_rpcsvc_program_actor (rpcsvc_conn_t *conn, rpcsvc_request_t *req)
-{
-        rpcsvc_program_t        *program = NULL;
-        int                     err = SYSTEM_ERR;
-        rpcsvc_actor_t          *actor = NULL;
-
-        if ((!conn) || (!req))
-                goto err;
-
-        program = (rpcsvc_program_t *)conn->program;
-        if (!program)
-                goto err;
-
-        if (req->prognum != program->prognum) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "RPC program not available");
-                err = PROG_UNAVAIL;
-                goto err;
-        }
-
-        if (!program->actors) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "RPC System error");
-                err = SYSTEM_ERR;
-                goto err;
-        }
-
-        if (req->progver != program->progver) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "RPC program version not"
-                        " available");
-                err = PROG_MISMATCH;
-                goto err;
-        }
-
-        if ((req->procnum < 0) || (req->procnum >= program->numactors)) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "RPC Program procedure not"
-                        " available");
-                err = PROC_UNAVAIL;
-                goto err;
-        }
-
-        actor = &program->actors[req->procnum];
-        if (!actor->actor) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "RPC Program procedure not"
-                        " available");
-                err = PROC_UNAVAIL;
-                actor = NULL;
-                goto err;
-        }
-
-        err = SUCCESS;
-        gf_log (GF_RPCSVC, GF_LOG_DEBUG, "Actor found: %s - %s",
-                program->progname, actor->procname);
-err:
-        if (req)
-                req->rpc_err = err;
-
-        return actor;
-}
-
-
-rpcsvc_txbuf_t *
-nfs_rpcsvc_init_txbuf (rpcsvc_conn_t *conn, struct iovec msg, struct iobuf *iob,
-                       struct iobref *iobref, int txflags)
-{
-        rpcsvc_txbuf_t  *txbuf = NULL;
-
-        txbuf = (rpcsvc_txbuf_t *) mem_get(conn->txpool);
-        if (!txbuf) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to get txbuf");
-                return NULL;
-        }
-
-        memset (txbuf, 0, sizeof (*txbuf));
-        INIT_LIST_HEAD (&txbuf->txlist);
-        txbuf->buf = msg;
-
-        /* If it was required, this iob must've been ref'd already
-         *  so I dont have to bother here.
-         */
-        txbuf->iob = iob;
-        txbuf->iobref = iobref;
-        txbuf->offset = 0;
-        txbuf->txbehave = txflags;
-
-        return txbuf;
-}
-
-
-int
-nfs_rpcsvc_conn_append_txlist (rpcsvc_conn_t *conn, struct iovec msg,
-                               struct iobuf *iob, int txflags)
-{
-        rpcsvc_txbuf_t          *txbuf = NULL;
-
-        if ((!conn) || (!msg.iov_base) || (!iob))
-                return -1;
-
-        txbuf = nfs_rpcsvc_init_txbuf (conn, msg, iob, NULL, txflags);
-        if (!txbuf)
-                return -1;
-
-        list_add_tail (&txbuf->txlist, &conn->txbufs);
-        return 0;
-}
-
-
-void
-nfs_rpcsvc_set_lastfrag (uint32_t *fragsize) {
-        (*fragsize) |= 0x80000000U;
-}
-
-void
-nfs_rpcsvc_set_frag_header_size (uint32_t size, char *haddr)
-{
-        size = htonl (size);
-        memcpy (haddr, &size, sizeof (size));
-}
-
-void
-nfs_rpcsvc_set_last_frag_header_size (uint32_t size, char *haddr)
-{
-        nfs_rpcsvc_set_lastfrag (&size);
-        nfs_rpcsvc_set_frag_header_size (size, haddr);
-}
-
-
-/* Given the RPC reply structure and the payload handed by the RPC program,
- * encode the RPC record header into the buffer pointed by recordstart.
- */
-struct iovec
-nfs_rpcsvc_record_build_header (char *recordstart, size_t rlen,
-                                struct rpc_msg reply, size_t payload)
-{
-        struct iovec    replyhdr;
-        struct iovec    txrecord = {0, 0};
-        size_t          fraglen = 0;
-        int             ret = -1;
-
-        /* After leaving aside the 4 bytes for the fragment header, lets
-         * encode the RPC reply structure into the buffer given to us.
-         */
-        ret = nfs_rpc_reply_to_xdr (&reply,(recordstart + RPCSVC_FRAGHDR_SIZE),
-                                    rlen, &replyhdr);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to create RPC reply");
-                goto err;
-        }
-
-        fraglen = payload + replyhdr.iov_len;
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Reply fraglen %zu, payload: %zu, "
-                "rpc hdr: %zu", fraglen, payload, replyhdr.iov_len);
-
-        /* Since we're not spreading RPC records over mutiple fragments
-         * we just set this fragment as the first and last fragment for this
-         * record.
-         */
-        nfs_rpcsvc_set_last_frag_header_size (fraglen, recordstart);
-
-        /* Even though the RPC record starts at recordstart+RPCSVC_FRAGHDR_SIZE
-         * we need to transmit the record with the fragment header, which starts
-         * at recordstart.
-         */
-        txrecord.iov_base = recordstart;
-
-        /* Remember, this is only the vec for the RPC header and does not
-         * include the payload above. We needed the payload only to calculate
-         * the size of the full fragment. This size is sent in the fragment
-         * header.
-         */
-        txrecord.iov_len = RPCSVC_FRAGHDR_SIZE + replyhdr.iov_len;
-
-err:
-        return txrecord;
-}
-
-
-int
-nfs_rpcsvc_conn_submit (rpcsvc_conn_t *conn, struct iovec hdr,
-                        struct iobuf *hdriob, struct iovec msgvec,
-                        struct iobuf *msgiob)
-{
-        int     ret = -1;
-
-        if ((!conn) || (!hdr.iov_base) || (!hdriob))
-                return -1;
-
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Tx Header: %zu, payload: %zu",
-                hdr.iov_len, msgvec.iov_len);
-        /* Now that we have both the RPC and Program buffers in xdr format
-         * lets hand it to the transmission layer.
-         */
-        pthread_mutex_lock (&conn->connlock);
-        {
-                if (!nfs_rpcsvc_conn_check_active (conn)) {
-                        gf_log (GF_RPCSVC, GF_LOG_DEBUG, "Connection inactive");
-                        goto unlock_err;
-                }
-
-                ret = nfs_rpcsvc_conn_append_txlist (conn, hdr, hdriob,
-                                                     RPCSVC_TXB_FIRST);
-                if (ret == -1) {
-                        gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to append "
-                                "header to transmission list");
-                        goto unlock_err;
-                }
-
-                /* It is possible that this RPC reply is an error reply. In that
-                 * case we might not have been handed a payload.
-                 */
-                ret = 0;
-                if (msgiob)
-                        ret = nfs_rpcsvc_conn_append_txlist (conn, msgvec,
-                                                             msgiob,
-                                                             RPCSVC_TXB_LAST);
-                if (ret == -1) {
-                        gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to append"
-                                " payload to transmission list");
-                        goto unlock_err;
-                }
-        }
-unlock_err:
-        pthread_mutex_unlock (&conn->connlock);
-
-        if (ret == -1)
-                goto err;
-
-        /* Tell event pool, we're interested in poll_out to trigger flush
-         * of our tx buffers.
-         */
-        conn->eventidx = event_select_on (conn->stage->eventpool, conn->sockfd,
-                                          conn->eventidx, -1, 1);
-        ret = 0;
-err:
-
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_fill_reply (rpcsvc_request_t *req, struct rpc_msg *reply)
-{
-        rpcsvc_program_t        *prog = NULL;
-        if ((!req) || (!reply))
-                return -1;
-
-        prog = nfs_rpcsvc_request_program (req);
-        nfs_rpc_fill_empty_reply (reply, req->xid);
-
-        if (req->rpc_stat == MSG_DENIED)
-                nfs_rpc_fill_denied_reply (reply, req->rpc_err, req->auth_err);
-        else if (req->rpc_stat == MSG_ACCEPTED)
-                nfs_rpc_fill_accepted_reply (reply, req->rpc_err,
-                                             prog->proglowvers,
-                                             prog->proghighvers,
-                                             req->verf.flavour,
-                                             req->verf.datalen,
-                                             req->verf.authdata);
-        else
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Invalid rpc_stat value");
-
-        return 0;
-}
-
-
-/* Given a request and the reply payload, build a reply and encodes the reply
- * into a record header. This record header is encoded into the vector pointed
- * to be recbuf.
- * msgvec is the buffer that points to the payload of the RPC program.
- * This buffer can be NULL, if an RPC error reply is being constructed.
- * The only reason it is needed here is that in case the buffer is provided,
- * we should account for the length of that buffer in the RPC fragment header.
- */
-struct iobuf *
-nfs_rpcsvc_record_build_record (rpcsvc_request_t *req, size_t payload,
-                                struct iovec *recbuf)
-{
-        struct rpc_msg          reply;
-        struct iobuf            *replyiob = NULL;
-        char                    *record = NULL;
-        struct iovec            recordhdr = {0, };
-        size_t                  pagesize = 0;
-        rpcsvc_conn_t           *conn = NULL;
-        rpcsvc_t                *svc = NULL;
-
-        if ((!req) || (!req->conn) || (!recbuf))
-                return NULL;
-
-        /* First, try to get a pointer into the buffer which the RPC
-         * layer can use.
-         */
-        conn = req->conn;
-        svc = nfs_rpcsvc_conn_rpcsvc (conn);
-        replyiob = iobuf_get (svc->ctx->iobuf_pool);
-        pagesize = iobpool_pagesize ((struct iobuf_pool *)svc->ctx->iobuf_pool);
-        if (!replyiob) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to get iobuf");
-                goto err_exit;
-        }
-
-        record = iobuf_ptr (replyiob);  /* Now we have it. */
-
-        /* Fill the rpc structure and XDR it into the buffer got above. */
-        nfs_rpcsvc_fill_reply (req, &reply);
-        recordhdr = nfs_rpcsvc_record_build_header (record, pagesize, reply,
-                                                    payload);
-        if (!recordhdr.iov_base) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to build record "
-                        " header");
-                iobuf_unref (replyiob);
-                replyiob = NULL;
-                recbuf->iov_base = NULL;
-                goto err_exit;
-        }
-
-        recbuf->iov_base = recordhdr.iov_base;
-        recbuf->iov_len = recordhdr.iov_len;
-err_exit:
-        return replyiob;
-}
-
-
-/*
- * The function to submit a program message to the RPC service.
- * This message is added to the transmission queue of the
- * conn.
- *
- * Program callers are not expected to use the msgvec->iov_base
- * address for anything else.
- * Nor are they expected to free it once this function returns.
- * Once the transmission of the buffer is completed by the RPC service,
- * the memory area as referenced through @msg will be unrefed.
- * If a higher layer does not want anything to do with this iobuf
- * after this function returns, it should call unref on it. For keeping
- * it around till the transmission is actually complete, rpcsvc also refs it.
- *  *
- * If this function returns an error by returning -1, the
- * higher layer programs should assume that a disconnection happened
- * and should know that the conn memory area as well as the req structure
- * has been freed internally.
- *
- * For now, this function assumes that a submit is always called
- * to send a new record. Later, if there is a situation where different
- * buffers for the same record come from different sources, then we'll
- * need to change this code to account for multiple submit calls adding
- * the buffers into a single record.
- */
-
-int
-nfs_rpcsvc_submit_generic (rpcsvc_request_t *req, struct iovec msgvec,
-                           struct iobuf *msg)
-{
-        int                     ret = -1;
-        struct iobuf            *replyiob = NULL;
-        struct iovec            recordhdr = {0, };
-        rpcsvc_conn_t           *conn = NULL;
-
-        if ((!req) || (!req->conn))
-                return -1;
-
-        conn = req->conn;
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Tx message: %zu", msgvec.iov_len);
-        /* Build the buffer containing the encoded RPC reply. */
-        replyiob = nfs_rpcsvc_record_build_record (req, msgvec.iov_len,
-                                                   &recordhdr);
-        if (!replyiob) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR,"Reply record creation failed");
-                goto disconnect_exit;
-        }
-
-        /* Must ref the iobuf got from higher layer so that the higher layer
-         * can rest assured that it can unref it and leave the final freeing
-         * of the buffer to us. Note msg can be NULL if an RPC-only message
-         * was being sent. Happens when an RPC error reply is being sent.
-         */
-        if (msg)
-                iobuf_ref (msg);
-        ret = nfs_rpcsvc_conn_submit (conn, recordhdr, replyiob, msgvec, msg);
-        mem_put (conn->rxpool, req);
-
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to submit message");
-                iobuf_unref (replyiob);
-        }
-
-disconnect_exit:
-        /* Note that a unref is called everytime a reply is sent. This is in
-         * response to the ref that is performed on the conn when a request is
-         * handed to the RPC program.
-         *
-         * The catch, however, is that if the reply is an rpc error, we must
-         * not unref. This is because the ref only contains
-         * references for the actors to which the request was handed plus one
-         * reference maintained by the RPC layer. By unrefing for a case where
-         * no actor was called, we will be losing the ref held for the RPC
-         * layer.
-         */
-        if ((nfs_rpcsvc_request_accepted (req)) &&
-            (nfs_rpcsvc_request_accepted_success (req)))
-                nfs_rpcsvc_conn_unref (conn);
-
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_request_attach_vector (rpcsvc_request_t *req, struct iovec msgvec,
-                                  struct iobuf *iob, struct iobref *iobref,
-                                  int finalvector)
-{
-        rpcsvc_txbuf_t          *txb = NULL;
-        int                     txflags = 0;
-
-        if ((!req) || (!msgvec.iov_base))
-                return -1;
-
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Tx Vector: %zu", msgvec.iov_len);
-        if (finalvector)
-                txflags |= RPCSVC_TXB_LAST;
-        /* We only let the user decide whether this is the last vector for the
-         * record, since the first vector is always the RPC header.
-         */
-        txb = nfs_rpcsvc_init_txbuf (req->conn, msgvec, iob, iobref, txflags);
-        if (!txb) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Could not init tx buf");
-                return -1;
-        }
-
-        req->payloadsize += msgvec.iov_len;
-        if (iob)
-                iobuf_ref (iob);
-        if (iobref)
-                iobref_ref (iobref);
-        list_add_tail (&txb->txlist, &req->txlist);
-
-        return 0;
-}
-
-
-int
-nfs_rpcsvc_request_attach_vectors (rpcsvc_request_t *req, struct iovec *payload,
-                                   int vcount, struct iobref *piobref)
-{
-        int     c = 0;
-        int     ret = -1;
-
-        for (;c < (vcount-1); c++) {
-                ret = nfs_rpcsvc_request_attach_vector (req, payload[c], NULL,
-                                                        piobref, 0);
-                if (ret < 0) {
-                        gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to attach "
-                                "vector");
-                        goto out;
-                }
-        }
-
-        ret = nfs_rpcsvc_request_attach_vector (req, payload[vcount-1], NULL,
-                                                piobref, 1);
-        if (ret < 0)
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to attach final vec");
-
-out:
-	return ret;
-}
-
-
-int
-nfs_rpcsvc_submit_vectors (rpcsvc_request_t *req)
-{
-        int                     ret = -1;
-        struct iobuf            *replyiob = NULL;
-        struct iovec            recordhdr = {0, };
-        rpcsvc_txbuf_t          *rpctxb = NULL;
-
-        if ((!req) || (!req->conn))
-                return -1;
-
-        /* Build the buffer containing the encoded RPC reply. */
-        replyiob = nfs_rpcsvc_record_build_record (req, req->payloadsize,
-                                                   &recordhdr);
-        if (!replyiob) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR,"Reply record creation failed");
-                goto disconnect_exit;
-        }
-
-        rpctxb = nfs_rpcsvc_init_txbuf (req->conn, recordhdr, replyiob, NULL,
-                                        RPCSVC_TXB_FIRST);
-        if (!rpctxb) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to create tx buf");
-                goto disconnect_exit;
-        }
-
-        pthread_mutex_lock (&req->conn->connlock);
-        {
-                list_splice_init (&req->txlist, &req->conn->txbufs);
-                list_add (&rpctxb->txlist, &req->conn->txbufs);
-        }
-        pthread_mutex_unlock (&req->conn->connlock);
-
-        ret = 0;
-        req->conn->eventidx = event_select_on (req->conn->stage->eventpool,
-                                               req->conn->sockfd,
-                                               req->conn->eventidx, -1, 1);
-disconnect_exit:
-        /* Note that a unref is called everytime a reply is sent. This is in
-         * response to the ref that is performed on the conn when a request is
-         * handed to the RPC program.
-         */
-        nfs_rpcsvc_conn_unref (req->conn);
-        if (ret == -1)
-                iobuf_unref (replyiob);
-
-        mem_put (req->conn->rxpool, req);
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_error_reply (rpcsvc_request_t *req)
-{
-        struct iovec    dummyvec = {0, };
-
-        if (!req)
-                return -1;
-
-        /* At this point the req should already have been filled with the
-         * appropriate RPC error numbers.
-         */
-        return nfs_rpcsvc_submit_generic (req, dummyvec, NULL);
-}
-
-
-rpcsvc_request_t *
-nfs_rpcsvc_request_init (rpcsvc_conn_t *conn, struct rpc_msg *callmsg,
-                         struct iovec progmsg, rpcsvc_request_t *req)
-{
-        if ((!conn) || (!callmsg)|| (!req))
-                return NULL;
-
-
-        /* We start a RPC request as always denied. */
-        req->rpc_stat = MSG_DENIED;
-        req->xid = nfs_rpc_call_xid (callmsg);
-        req->prognum = nfs_rpc_call_program (callmsg);
-        req->progver = nfs_rpc_call_progver (callmsg);
-        req->procnum = nfs_rpc_call_progproc (callmsg);
-        req->conn = conn;
-        req->msg = progmsg;
-        req->recordiob = conn->rstate.activeiob;
-        INIT_LIST_HEAD (&req->txlist);
-        req->payloadsize = 0;
-
-        /* By this time, the data bytes for the auth scheme would have already
-         * been copied into the required sections of the req structure,
-         * we just need to fill in the meta-data about it now.
-         */
-        req->cred.flavour = nfs_rpc_call_cred_flavour (callmsg);
-        req->cred.datalen = nfs_rpc_call_cred_len (callmsg);
-        req->verf.flavour = nfs_rpc_call_verf_flavour (callmsg);
-        req->verf.datalen = nfs_rpc_call_verf_len (callmsg);
-
-        /* AUTH */
-        nfs_rpcsvc_auth_request_init (req);
-        return req;
-}
-
-
-rpcsvc_request_t *
-nfs_rpcsvc_request_create (rpcsvc_conn_t *conn)
-{
-        char                    *msgbuf = NULL;
-        struct rpc_msg          rpcmsg;
-        struct iovec            progmsg;        /* RPC Program payload */
-        rpcsvc_request_t        *req = NULL;
-        int                     ret = -1;
-
-        if (!conn)
-                return NULL;
-
-        /* We need to allocate the request before actually calling
-         * rpcsvc_request_init on the request so that we, can fill the auth
-         * data directly into the request structure from the message iobuf.
-         * This avoids a need to keep a temp buffer into which the auth data
-         * would've been copied otherwise.
-         */
-        nfs_rpcsvc_alloc_request (conn, req);
-        if (!req) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed to alloc request");
-                goto err;
-        }
-
-        msgbuf = iobuf_ptr (conn->rstate.activeiob);
-        ret = nfs_xdr_to_rpc_call (msgbuf, conn->rstate.recordsize, &rpcmsg,
-                                   &progmsg, req->cred.authdata,
-                                   req->verf.authdata);
-
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "RPC call decoding failed");
-                nfs_rpcsvc_request_seterr (req, GARBAGE_ARGS);
-                goto err;
-        }
-
-        ret = -1;
-        nfs_rpcsvc_request_init (conn, &rpcmsg, progmsg, req);
-
-        gf_log (GF_RPCSVC, GF_LOG_DEBUG, "RPC XID: %lx, Ver: %ld, Program: %ld,"
-                " ProgVers: %ld, Proc: %ld", nfs_rpc_call_xid (&rpcmsg),
-                nfs_rpc_call_rpcvers (&rpcmsg), nfs_rpc_call_program (&rpcmsg),
-                nfs_rpc_call_progver (&rpcmsg),
-                nfs_rpc_call_progproc (&rpcmsg));
-
-        if (nfs_rpc_call_rpcvers (&rpcmsg) != 2) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "RPC version not supported");
-                nfs_rpcsvc_request_seterr (req, RPC_MISMATCH);
-                goto err;
-        }
-
-        ret = nfs_rpcsvc_authenticate (req);
-        if (ret == RPCSVC_AUTH_REJECT) {
-                /* No need to set auth_err, that is the responsibility of
-                 * the authentication handler since only that know what exact
-                 * error happened.
-                 */
-                nfs_rpcsvc_request_seterr (req, AUTH_ERROR);
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Failed authentication");
-                ret = -1;
-                goto err;
-        }
-
-
-        /* If the error is not RPC_MISMATCH, we consider the call as accepted
-         * since we are not handling authentication failures for now.
-         */
-        req->rpc_stat = MSG_ACCEPTED;
-        ret = 0;
-err:
-        if (ret == -1) {
-                ret = nfs_rpcsvc_error_reply (req);
-                req = NULL;
-        }
-
-        return req;
-}
-
-
-int
-nfs_rpcsvc_handle_rpc_call (rpcsvc_conn_t *conn)
-{
-        rpcsvc_actor_t          *actor = NULL;
-        rpcsvc_request_t        *req = NULL;
-        int                     ret = -1;
-
-        if (!conn)
-                return -1;
-
-        req = nfs_rpcsvc_request_create (conn);
-        if (!req)
-                goto err;
-
-        if (!nfs_rpcsvc_request_accepted (req))
-                goto err_reply;
-
-        actor = nfs_rpcsvc_program_actor (conn, req);
-        if (!actor)
-                goto err_reply;
-
-        if ((actor) && (actor->actor)) {
-                THIS = nfs_rpcsvc_request_actorxl (req);
-                nfs_rpcsvc_conn_ref (conn);
-                ret = actor->actor (req);
-        }
-
-err_reply:
-        if (ret == RPCSVC_ACTOR_ERROR)
-                ret = nfs_rpcsvc_error_reply (req);
-
-        /* No need to propagate error beyond this function since the reply
-         * has now been queued. */
-        ret = 0;
-err:
-        return ret;
-}
-
-#define nfs_rpc_call_cred_addr(rs) (iobuf_ptr ((rs)->activeiob) + RPCSVC_BARERPC_MSGSZ - 4)
-
-uint32_t
-nfs_rpcsvc_call_credlen (rpcsvc_record_state_t *rs)
-{
-        char                    *credaddr = NULL;
-        uint32_t                credlen_nw = 0;
-        uint32_t                credlen_host = 0;
-
-        /* Position to the start of the credential length field. */
-        credaddr = nfs_rpc_call_cred_addr (rs);
-        credlen_nw = *(uint32_t *)credaddr;
-        credlen_host = ntohl (credlen_nw);
-
-        return credlen_host;
-}
-
-uint32_t
-nfs_rpcsvc_call_verflen (rpcsvc_record_state_t *rs)
-{
-        char            *verfaddr = NULL;
-        uint32_t        verflen_nw = 0;
-        uint32_t        verflen_host = 0;
-        uint32_t        credlen = 0;
-
-        /* Position to the start of the verifier length field. */
-        credlen = nfs_rpcsvc_call_credlen (rs);
-        verfaddr = (nfs_rpc_call_cred_addr (rs) + 4 + credlen);
-        verflen_nw = *(uint32_t *)verfaddr;
-        verflen_host = ntohl (verflen_nw);
-
-        return verflen_host;
-}
-
-
-void
-nfs_rpcsvc_update_vectored_verf (rpcsvc_record_state_t *rs)
-{
-        if (!rs)
-                return;
-
-        rs->recordsize += nfs_rpcsvc_call_verflen (rs);
-        return;
-}
-
-
-void
-nfs_rpcsvc_handle_vectored_prep_rpc_call (rpcsvc_conn_t *conn)
-{
-        rpcsvc_actor_t          *actor = NULL;
-        rpcsvc_request_t        *req = NULL;
-        rpcsvc_record_state_t   *rs = NULL;
-        rpcsvc_t                *svc = NULL;
-        int                     ret = -1;
-        ssize_t                 remfrag = RPCSVC_ACTOR_ERROR;
-        int                     newbuf = 0;
-
-        if (!conn)
-                return;
-
-        rs = &conn->rstate;
-
-        /* In case one of the steps below fails, we need to make sure that the
-         * remaining frag in the kernel's buffers are read-out so that the
-         * requests that follow can be served.
-         */
-        rs->remainingfrag = rs->fragsize - rs->recordsize;
-        rs->vecstate = RPCSVC_VECTOR_IGNORE;
-        req = nfs_rpcsvc_request_create (conn);
-        svc = nfs_rpcsvc_conn_rpcsvc (conn);
-        if (!req)
-                goto err;
-
-        if (!nfs_rpcsvc_request_accepted (req))
-                goto err_reply;
-
-        actor = nfs_rpcsvc_program_actor (conn, req);
-        if (!actor)
-                goto err_reply;
-
-        if (!actor->vector_sizer) {
-                ret = -1;
-                nfs_rpcsvc_request_seterr (req, PROC_UNAVAIL);
-                goto err_reply;
-        }
-
-        nfs_rpcsvc_conn_ref (conn);
-        THIS = nfs_rpcsvc_request_actorxl (req);
-        ret = actor->vector_sizer (req, &remfrag, &newbuf);
-        nfs_rpcsvc_conn_unref (conn);
-
-        if (ret == RPCSVC_ACTOR_ERROR) {
-                ret = -1;
-                nfs_rpcsvc_request_seterr (req, SYSTEM_ERR);
-                goto err_reply;
-        }
-
-        rs->remainingfrag = remfrag;
-        rs->vecstate = RPCSVC_VECTOR_READPROCHDR;
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored RPC proc header remaining:"
-                " %d", rs->remainingfrag);
-        conn->vectoredreq = req;
-
-        /* Store the reference to the current frag pointer. This is where the
-         * proc header will be read into.
-         */
-        req->msg.iov_base = rs->fragcurrent;
-        req->msg.iov_len = rs->remainingfrag;
-        ret = 0;
-
-err_reply:
-        if (ret == -1)
-                ret = nfs_rpcsvc_error_reply (req);
-
-        /* No need to propagate error beyond this function since the reply
-         * has now been queued. */
-        ret = 0;
-err:
-        return;
-}
-
-
-void
-nfs_rpcsvc_update_vectored_verfsz (rpcsvc_conn_t *conn)
-{
-        rpcsvc_record_state_t   *rs = NULL;
-        uint32_t                verflen = 0;
-
-        if (!conn)
-                return;
-
-        rs = &conn->rstate;
-
-        verflen = nfs_rpcsvc_call_verflen (rs);
-        rs->recordsize += 8;
-        if (verflen > 0) {
-                rs->remainingfrag = verflen;
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored RPC verf remaining: "
-                        " %d", rs->remainingfrag);
-                rs->vecstate = RPCSVC_VECTOR_READVERF;
-        } else {
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored RPC preparing call");
-                nfs_rpcsvc_handle_vectored_prep_rpc_call (conn);
-        }
-
-        return;
-}
-
-
-void
-nfs_rpcsvc_update_vectored_cred (rpcsvc_record_state_t *rs)
-{
-        uint32_t                credlen = 0;
-
-        if (!rs)
-                return;
-
-        credlen = nfs_rpcsvc_call_credlen (rs);
-        /* Update remainingfrag to read the 8 bytes needed for
-         * reading verf flavour and verf len.
-         */
-        rs->remainingfrag = (2 * sizeof (uint32_t));
-        rs->vecstate = RPCSVC_VECTOR_READVERFSZ;
-        rs->recordsize += credlen;
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored RPC verfsz remaining: %d",
-                rs->remainingfrag);
-
-        return;
-}
-
-void
-nfs_rpcsvc_update_vectored_barerpc (rpcsvc_record_state_t *rs)
-{
-        uint32_t                credlen = 0;
-
-        if (!rs)
-                return;
-
-        credlen = nfs_rpcsvc_call_credlen (rs);
-        rs->recordsize = RPCSVC_BARERPC_MSGSZ;
-        if (credlen == 0) {
-                rs->remainingfrag = 8;
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored RPC verfsz remaining"
-                        ": %d", rs->remainingfrag);
-                rs->vecstate = RPCSVC_VECTOR_READVERFSZ;
-        } else {
-                rs->remainingfrag = credlen;
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored RPC cred remaining: "
-                        "%d", rs->remainingfrag);
-                rs->vecstate = RPCSVC_VECTOR_READCRED;
-        }
-
-        return;
-}
-
-
-void
-nfs_rpcsvc_handle_vectored_rpc_call (rpcsvc_conn_t *conn)
-{
-        rpcsvc_actor_t          *actor = NULL;
-        rpcsvc_request_t        *req = NULL;
-        rpcsvc_record_state_t   *rs = NULL;
-        rpcsvc_t                *svc = NULL;
-        int                     ret = -1;
-        ssize_t                 remfrag = -1;
-        int                     newbuf = 0;
-
-        if (!conn)
-                return;
-
-        rs = &conn->rstate;
-
-        req = conn->vectoredreq;
-        svc = nfs_rpcsvc_conn_rpcsvc (conn);
-
-        if (!req)
-                goto err;
-
-        actor = nfs_rpcsvc_program_actor (conn, req);
-        if (!actor)
-                goto err_reply;
-
-        if (!actor->vector_sizer) {
-                ret = -1;
-                nfs_rpcsvc_request_seterr (req, PROC_UNAVAIL);
-                goto err_reply;
-        }
-
-        req->msg.iov_len = (unsigned long)((long)rs->fragcurrent - (long)req->msg.iov_base);
-        nfs_rpcsvc_conn_ref (conn);
-        THIS = nfs_rpcsvc_request_actorxl (req);
-        ret = actor->vector_sizer (req, &remfrag, &newbuf);
-        nfs_rpcsvc_conn_unref (conn);
-        if (ret == RPCSVC_ACTOR_ERROR) {
-                ret = -1;
-                nfs_rpcsvc_request_seterr (req, SYSTEM_ERR);
-                goto err_reply;
-        }
-
-        if (newbuf) {
-                rs->vectoriob = iobuf_get (svc->ctx->iobuf_pool);
-                rs->fragcurrent = iobuf_ptr (rs->vectoriob);
-                rs->vecstate = RPCSVC_VECTOR_READVEC;
-                rs->remainingfrag = remfrag;
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored RPC buf remaining:"
-                        " %d", rs->remainingfrag);
-        } else {
-                rs->remainingfrag = remfrag;
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored RPC proc remaining:"
-                        " %d", rs->remainingfrag);
-        }
-
-        ret = 0;
-err_reply:
-        if (ret == -1)
-                ret = nfs_rpcsvc_error_reply (req);
-
-        /* No need to propagate error beyond this function since the reply
-         * has now been queued. */
-        ret = 0;
-err:
-        return;
-}
-
-
-
-void
-nfs_rpcsvc_record_vectored_call_actor (rpcsvc_conn_t *conn)
-{
-        rpcsvc_actor_t          *actor = NULL;
-        rpcsvc_request_t        *req = NULL;
-        rpcsvc_record_state_t   *rs = NULL;
-        rpcsvc_t                *svc = NULL;
-        int                     ret = -1;
-
-        if (!conn)
-                return;
-
-        rs = &conn->rstate;
-        req = conn->vectoredreq;
-        svc = nfs_rpcsvc_conn_rpcsvc (conn);
-
-        if (!req)
-                goto err;
-
-        actor = nfs_rpcsvc_program_actor (conn, req);
-        if (!actor)
-                goto err_reply;
-
-        if (actor->vector_actor) {
-                nfs_rpcsvc_conn_ref (conn);
-                THIS = nfs_rpcsvc_request_actorxl (req);
-                ret = actor->vector_actor (req, rs->vectoriob);
-        } else {
-                nfs_rpcsvc_request_seterr (req, PROC_UNAVAIL);
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "No vectored handler present");
-                ret = RPCSVC_ACTOR_ERROR;
-        }
-
-err_reply:
-        if (ret == RPCSVC_ACTOR_ERROR)
-                ret = nfs_rpcsvc_error_reply (req);
-
-        /* No need to propagate error beyond this function since the reply
-         * has now been queued. */
-        ret = 0;
-err:
-        return;
-}
-
-
-
-ssize_t
-nfs_rpcsvc_update_vectored_state (rpcsvc_conn_t *conn)
-{
-        rpcsvc_record_state_t   *rs = NULL;
-        rpcsvc_t                *svc = NULL;
-
-        if (!conn)
-                return 0;
-
-        /* At this point, we can be confident that the activeiob contains
-         * exactly the first RPCSVC_BARERPC_MSGSZ bytes needed in order to
-         * determine the program and actor. So the next state will be
-         * to read the credentials.
-         *
-         * But first, try to determine how many more bytes do we need from the
-         * network to complete the RPC message including the credentials.
-         */
-
-        rs = &conn->rstate;
-        if (nfs_rpcsvc_record_vectored_baremsg (rs))
-                nfs_rpcsvc_update_vectored_barerpc (rs);
-        else if (nfs_rpcsvc_record_vectored_cred (rs))
-                nfs_rpcsvc_update_vectored_cred (rs);
-        else if (nfs_rpcsvc_record_vectored_verfsz (rs))
-                nfs_rpcsvc_update_vectored_verfsz (conn);
-        else if (nfs_rpcsvc_record_vectored_verfread (rs)) {
-                nfs_rpcsvc_update_vectored_verf (rs);
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored RPC preparing call");
-                nfs_rpcsvc_handle_vectored_prep_rpc_call (conn);
-        } else if (nfs_rpcsvc_record_vectored_readprochdr (rs))
-                nfs_rpcsvc_handle_vectored_rpc_call (conn);
-        else if (nfs_rpcsvc_record_vectored_ignore (rs)) {
-                svc = nfs_rpcsvc_conn_rpcsvc (conn);
-                nfs_rpcsvc_record_init (rs, svc->ctx->iobuf_pool);
-        } else if (nfs_rpcsvc_record_vectored_readvec (rs)) {
-                svc = nfs_rpcsvc_conn_rpcsvc (conn);
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored RPC vector read");
-                nfs_rpcsvc_record_vectored_call_actor (conn);
-                nfs_rpcsvc_record_init (rs, svc->ctx->iobuf_pool);
-        }
-
-        return 0;
-}
-
-
-ssize_t
-nfs_rpcsvc_record_read_partial_frag (rpcsvc_record_state_t *rs,
-                                     ssize_t dataread);
-
-ssize_t
-nfs_rpcsvc_update_vectored_msg (rpcsvc_conn_t *conn, ssize_t dataread)
-{
-
-        if (!conn)
-                return dataread;
-
-        /* find out how much of the bare msg is pending and set that up to be
-         * read into the updated fragcurrent along with the updated size into
-         * remainingfrag.
-         */
-
-
-        /* Incidently, the logic needed here is similar to a regular partial
-         * fragment read since we've already set the remainingfrag member in
-         * rstate to be RPCSVC_BARERPC_MSGSZ for the purpose of a vectored
-         * fragment.
-         */
-        return nfs_rpcsvc_record_read_partial_frag (&conn->rstate, dataread);
-}
-
-/* FIX: As a first version of vectored reading, I am assuming dataread will
- * always be equal to RPCSVC_BARERPC_MSGSZ for the sake of simplicity on the
- * belief that we're never actually reading more bytes than needed in each
- * poll_in.
- */
-ssize_t
-nfs_rpcsvc_handle_vectored_frag (rpcsvc_conn_t *conn, ssize_t dataread)
-{
-        if (!conn)
-                return dataread;
-
-        /* At this point we can be confident that only the frag size has been
-         * read from the network. Now it is up to us to have the remaining RPC
-         * fields given to us here.
-         */
-
-        /* Since the poll_in handler uses the remainingfrag field to determine
-         * how much to read from the network, we'll hack this scheme to tell
-         * the poll_in to read at most RPCSVC_BARERPC_MSGSZ bytes. This is done
-         * to, as a first step, identify which (program, actor) we need to call.
-         */
-
-        dataread = nfs_rpcsvc_update_vectored_msg (conn, dataread);
-
-        if (conn->rstate.remainingfrag == 0) {
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored frag complete");
-                dataread = nfs_rpcsvc_update_vectored_state (conn);
-        }
-
-        return dataread;
-}
-
-
-int
-nfs_rpcsvc_record_update_state (rpcsvc_conn_t *conn, ssize_t dataread)
-{
-        rpcsvc_record_state_t   *rs = NULL;
-        rpcsvc_t                *svc = NULL;
-
-        if (!conn)
-                return -1;
-
-        rs = &conn->rstate;
-        /* At entry into this function, fragcurrent will be pointing to the\
-         * start of the area into which dataread number of bytes were read.
-         */
-
-        if (nfs_rpcsvc_record_readfraghdr(rs))
-                dataread = nfs_rpcsvc_record_update_fraghdr (rs, dataread);
-
-        if (nfs_rpcsvc_record_readfrag(rs)) {
-                /* The minimum needed for triggering the vectored handler is
-                 * the frag size field. The fragsize member remains set to this
-                 * size till this request is completely extracted from the
-                 * network. Once all the data has been read from the network,
-                 * the request structure would've been created. The point being
-                 * that even if it takes multiple calls to network IO for
-                 * getting the vectored fragment, we can continue to use this
-                 * condition as the flag to tell us that this is a vectored
-                 * fragment.
-                 */
-                if ((dataread > 0) && (nfs_rpcsvc_record_vectored (rs))) {
-                        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Vectored frag");
-                        dataread = nfs_rpcsvc_handle_vectored_frag (conn,
-                                                                    dataread);
-                } else if (dataread > 0) {
-                        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Regular frag");
-                        dataread = nfs_rpcsvc_record_update_frag (rs, dataread);
-                }
-        }
-
-        /* This should not happen. We are never reading more than the current
-         * fragment needs. Something is seriously wrong.
-         */
-        if (dataread > 0) {
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Data Left: %zd", dataread);
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Unwanted data read from "
-                        " connection.");
-        }
-
-        /* If we're now supposed to wait for a new fragment header and if the
-         * fragment that we just completed in the previous call to
-         * rpcsvc_record_update_frag was the last fragment for the current
-         * RPC record, then, it is time to perform the translation from
-         * XDR formatted buffer in activeiob followed by the upcall to the
-         * protocol actor.
-         */
-        if ((nfs_rpcsvc_record_readfraghdr(rs)) && (rs->islastfrag)) {
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "Full Record Received.");
-                nfs_rpcsvc_handle_rpc_call (conn);
-                svc = nfs_rpcsvc_conn_rpcsvc (conn);
-                nfs_rpcsvc_record_init (rs, svc->ctx->iobuf_pool);
-        }
-
-        return 0;
-}
-
-
-char *
-nfs_rpcsvc_record_read_addr (rpcsvc_record_state_t *rs)
-{
-
-        if (nfs_rpcsvc_record_readfraghdr (rs))
-                return nfs_rpcsvc_record_currenthdr_addr (rs);
-        else if (nfs_rpcsvc_record_readfrag (rs))
-                return nfs_rpcsvc_record_currentfrag_addr (rs);
-
-        return NULL;
-}
-
-
-int
-nfs_rpcsvc_conn_data_poll_in (rpcsvc_conn_t *conn)
-{
-        ssize_t         dataread = -1;
-        size_t          readsize = 0;
-        char            *readaddr = NULL;
-        int             ret = -1;
-
-        readaddr = nfs_rpcsvc_record_read_addr (&conn->rstate);
-        if (!readaddr)
-                goto err;
-
-        readsize = nfs_rpcsvc_record_read_size (&conn->rstate);
-        if (readsize == -1)
-                goto err;
-
-        dataread = nfs_rpcsvc_socket_read (conn->sockfd, readaddr, readsize);
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "conn: 0x%lx, readsize: %zu, dataread:"
-                "%zd", (long)conn, readsize, dataread);
-
-        if (dataread > 0)
-                ret = nfs_rpcsvc_record_update_state (conn, dataread);
-
-err:
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_conn_data_poll_err (rpcsvc_conn_t *conn)
-{
-        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Received error event");
-        nfs_rpcsvc_conn_deinit (conn);
-        return 0;
-}
-
-
-int
-__nfs_rpcsvc_conn_data_poll_out (rpcsvc_conn_t *conn)
-{
-        rpcsvc_txbuf_t          *txbuf = NULL;
-        rpcsvc_txbuf_t          *tmp = NULL;
-        ssize_t                 written = -1;
-        char                    *writeaddr = NULL;
-        size_t                  writesize = -1;
-
-        if (!conn)
-                return -1;
-
-        /* Attempt transmission of each of the pending buffers */
-        list_for_each_entry_safe (txbuf, tmp, &conn->txbufs, txlist) {
-tx_remaining:
-                writeaddr = (char *)(txbuf->buf.iov_base + txbuf->offset);
-                writesize = (txbuf->buf.iov_len - txbuf->offset);
-
-                if (txbuf->txbehave & RPCSVC_TXB_FIRST) {
-                        gf_log (GF_RPCSVC, GF_LOG_TRACE, "First Tx Buf");
-                        nfs_rpcsvc_socket_block_tx (conn->sockfd);
-                }
-
-                written = nfs_rpcsvc_socket_write (conn->sockfd, writeaddr,
-                                                   writesize);
-                if (txbuf->txbehave & RPCSVC_TXB_LAST) {
-                        gf_log (GF_RPCSVC, GF_LOG_TRACE, "Last Tx Buf");
-                        nfs_rpcsvc_socket_unblock_tx (conn->sockfd);
-                }
-                gf_log (GF_RPCSVC, GF_LOG_TRACE, "conn: 0x%lx, Tx request: %zu,"
-                        " Tx sent: %zd", (long)conn, writesize, written);
-
-                /* There was an error transmitting this buffer */
-                if (written == -1)
-                        break;
-
-                if (written >= 0)
-                        txbuf->offset += written;
-
-                /* If the current buffer has been completely transmitted,
-                 * delete it from the list and move on to the next buffer.
-                 */
-                if (txbuf->offset == txbuf->buf.iov_len) {
-                        /* It doesnt matter who ref'ed this iobuf, rpcsvc for
-                         * its own header or a RPC program.
-                         */
-                        if (txbuf->iob)
-                                iobuf_unref (txbuf->iob);
-                        if (txbuf->iobref)
-                                iobref_unref (txbuf->iobref);
-
-                        list_del (&txbuf->txlist);
-                        mem_put (conn->txpool, txbuf);
-                } else
-                        /* If the current buffer is incompletely tx'd, do not
-                         * go to the head of the loop, since that moves us to
-                         * the next buffer.
-                         */
-                        goto tx_remaining;
-        }
-
-        /* If we've broken out of the loop above then we must unblock
-         * the transmission now.
-         */
-        nfs_rpcsvc_socket_unblock_tx (conn->sockfd);
-        if (list_empty (&conn->txbufs))
-                conn->eventidx = event_select_on (conn->stage->eventpool,
-                                                  conn->sockfd, conn->eventidx,
-                                                  -1, 0);
-
-        return 0;
-}
-
-
-int
-nfs_rpcsvc_conn_data_poll_out (rpcsvc_conn_t *conn)
-{
-        if (!conn)
-                return -1;
-
-
-        pthread_mutex_lock (&conn->connlock);
-        {
-                __nfs_rpcsvc_conn_data_poll_out (conn);
-        }
-        pthread_mutex_unlock (&conn->connlock);
-
-        return 0;
-}
-
-
-int
-nfs_rpcsvc_conn_data_handler (int fd, int idx, void *data, int poll_in,
-                              int poll_out, int poll_err)
-{
-        rpcsvc_conn_t   *conn = NULL;
-        int             ret = 0;
-
-        if (!data)
-                return 0;
-
-        conn = (rpcsvc_conn_t *)data;
-
-        if (poll_out)
-                ret = nfs_rpcsvc_conn_data_poll_out (conn);
-
-        if (poll_err) {
-                ret = nfs_rpcsvc_conn_data_poll_err (conn);
-                return 0;
-        }
-
-        if (poll_in) {
-                ret = 0;
-                ret = nfs_rpcsvc_conn_data_poll_in (conn);
-        }
-
-        if (ret == -1)
-                nfs_rpcsvc_conn_data_poll_err (conn);
-
-        return 0;
-}
-
-
-int
-nfs_rpcsvc_conn_listening_handler (int fd, int idx, void *data, int poll_in,
-                                   int poll_out, int poll_err)
-{
-        rpcsvc_conn_t           *newconn = NULL;
-        rpcsvc_stage_t          *selectedstage = NULL;
-        int                     ret = -1;
-        rpcsvc_conn_t           *conn = NULL;
-        rpcsvc_program_t        *prog = NULL;
-        rpcsvc_t                *svc = NULL;
-
-        if (!poll_in)
-                return 0;
-
-        conn = (rpcsvc_conn_t *)data;
-        prog = (rpcsvc_program_t *)conn->program;
-        svc = nfs_rpcsvc_conn_rpcsvc (conn);
-        newconn = nfs_rpcsvc_conn_accept_init (svc, fd, prog);
-        if (!newconn) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "failed to accept connection");
-                goto err;
-        }
-
-        selectedstage = nfs_rpcsvc_select_stage (svc);
-        if (!selectedstage)
-                goto close_err;
-
-        /* Now that we've accepted the connection, we need to associate
-         * its events to a stage.
-         */
-        ret = nfs_rpcsvc_stage_conn_associate (selectedstage, newconn,
-                                               nfs_rpcsvc_conn_data_handler,
-                                               newconn);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "could not associated stage "
-                        " with new connection");
-                goto close_err;
-        }
-        gf_log (GF_RPCSVC, GF_LOG_DEBUG, "New Connection: Program %s, Num: %d,"
-                " Ver: %d, Port: %d", prog->progname, prog->prognum,
-                prog->progver, prog->progport);
-        ret = 0;
-close_err:
-        if (ret == -1)
-                nfs_rpcsvc_conn_unref (newconn);
-
-err:
-        return ret;
-}
-
-
-/* Register the program with the local portmapper service. */
-int
-nfs_rpcsvc_program_register_portmap (rpcsvc_program_t *newprog)
-{
-        if (!newprog)
-                return -1;
-
-        if (!(pmap_set(newprog->prognum, newprog->progver, IPPROTO_TCP,
-                       newprog->progport))) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Could not register with"
-                        " portmap");
-                return -1;
-        }
-
-        return 0;
-}
-
-
-int
-nfs_rpcsvc_program_unregister_portmap (rpcsvc_program_t *prog)
-{
-        if (!prog)
-                return -1;
-
-        if (!(pmap_unset(prog->prognum, prog->progver))) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Could not unregister with"
-                        " portmap");
-                return -1;
-        }
-
-        return 0;
-}
-
-
-int
-nfs_rpcsvc_stage_program_register (rpcsvc_stage_t *stg,
-                                   rpcsvc_program_t *newprog)
-{
-        rpcsvc_conn_t           *newconn = NULL;
-        rpcsvc_t                *svc = NULL;
-
-        if ((!stg) || (!newprog))
-                return -1;
-
-        svc = nfs_rpcsvc_stage_service (stg);
-        /* Create a listening socket */
-        newconn = nfs_rpcsvc_conn_listen_init (svc, newprog);
-        if (!newconn) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "could not create listening"
-                        " connection");
-                return -1;
-        }
-
-        if ((nfs_rpcsvc_stage_conn_associate (stg, newconn,
-                                              nfs_rpcsvc_conn_listening_handler,
-                                              newconn)) == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR,"could not associate stage with"
-                        " listening connection");
-                return -1;
-        }
-
-        return 0;
-}
-
-
-int
-nfs_rpcsvc_program_register (rpcsvc_t *svc, rpcsvc_program_t program)
-{
-        rpcsvc_program_t        *newprog = NULL;
-        rpcsvc_stage_t          *selectedstage = NULL;
-        int                     ret = -1;
-
-        if (!svc)
-                return -1;
-
-        newprog = GF_CALLOC (1, sizeof(*newprog),gf_common_mt_rpcsvc_program_t);
-        if (!newprog)
-                return -1;
-
-        if (!program.actors)
-                goto free_prog;
-
-        memcpy (newprog, &program, sizeof (program));
-        selectedstage = nfs_rpcsvc_select_stage (svc);
-
-        ret = nfs_rpcsvc_stage_program_register (selectedstage, newprog);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "stage registration of program"
-                        " failed");
-                goto free_prog;
-        }
-
-        ret = nfs_rpcsvc_program_register_portmap (newprog);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "portmap registration of"
-                        " program failed");
-                goto free_prog;
-        }
-
-        ret = 0;
-        gf_log (GF_RPCSVC, GF_LOG_DEBUG, "New program registered: %s, Num: %d,"
-                " Ver: %d, Port: %d", newprog->progname, newprog->prognum,
-                newprog->progver, newprog->progport);
-
-free_prog:
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Program registration failed:"
-                        " %s, Num: %d, Ver: %d, Port: %d", newprog->progname,
-                        newprog->prognum, newprog->progver, newprog->progport);
-                GF_FREE (newprog);
-        }
-
-        return ret;
-}
-
-/* The only difference between the generic submit and this one is that the
- * generic submit is also used for submitting RPC error replies in where there
- * are no payloads so the msgvec and msgbuf can be NULL.
- * Since RPC programs should be using this function along with their payloads
- * we must perform NULL checks before calling the generic submit.
- */
-int
-nfs_rpcsvc_submit_message (rpcsvc_request_t *req, struct iovec msgvec,
-                           struct iobuf *msg)
-{
-        if ((!req) || (!req->conn) || (!msg) || (!msgvec.iov_base))
-                return -1;
-
-        return nfs_rpcsvc_submit_generic (req, msgvec, msg);
-}
-
-
-int
-nfs_rpcsvc_program_unregister (rpcsvc_t *svc, rpcsvc_program_t prog)
-{
-        int                     ret = -1;
-
-        if (!svc)
-                return -1;
-
-        /* TODO: De-init the listening connection for this program. */
-        ret = nfs_rpcsvc_program_unregister_portmap (&prog);
-        if (ret == -1) {
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "portmap unregistration of"
-                        " program failed");
-                goto err;
-        }
-
-        ret = 0;
-        gf_log (GF_RPCSVC, GF_LOG_DEBUG, "Program unregistered: %s, Num: %d,"
-                " Ver: %d, Port: %d", prog.progname, prog.prognum,
-                prog.progver, prog.progport);
-
-err:
-        if (ret == -1)
-                gf_log (GF_RPCSVC, GF_LOG_ERROR, "Program unregistration failed"
-                        ": %s, Num: %d, Ver: %d, Port: %d", prog.progname,
-                        prog.prognum, prog.progver, prog.progport);
-
-        return ret;
-}
-
-
-int
-nfs_rpcsvc_conn_peername (rpcsvc_conn_t *conn, char *hostname, int hostlen)
-{
-        if (!conn)
-                return -1;
-
-        return nfs_rpcsvc_socket_peername (conn->sockfd, hostname, hostlen);
-}
-
-
-int
-nfs_rpcsvc_conn_peeraddr (rpcsvc_conn_t *conn, char *addrstr, int addrlen,
-                          struct sockaddr *sa, socklen_t sasize)
-{
-        if (!conn)
-                return -1;
-
-        return nfs_rpcsvc_socket_peeraddr (conn->sockfd, addrstr, addrlen, sa,
-                                           sasize);
-}
-
diff --git a/xlators/nfs/lib/src/rpcsvc.h b/xlators/nfs/lib/src/rpcsvc.h
deleted file mode 100644
index 6d3834c85..000000000
--- a/xlators/nfs/lib/src/rpcsvc.h
+++ /dev/null
@@ -1,732 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _NFS_RPCSVC_H
-#define _NFS_RPCSVC_H
-
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "event.h"
-#include "logging.h"
-#include "dict.h"
-#include "mem-pool.h"
-#include "list.h"
-#include "iobuf.h"
-#include "xdr-rpc.h"
-#include "glusterfs.h"
-#include "xlator.h"
-
-#include <pthread.h>
-#include <sys/uio.h>
-
-#ifdef GF_DARWIN_HOST_OS
-#include <nfs/rpcv2.h>
-#define NGRPS RPCAUTH_UNIXGIDS
-#endif
-
-#define GF_RPCSVC       "nfsrpc"
-#define RPCSVC_THREAD_STACK_SIZE ((size_t)(1024 * GF_UNIT_KB))
-
-#define RPCSVC_DEFAULT_MEMFACTOR        15
-#define RPCSVC_EVENTPOOL_SIZE_MULT      1024
-#define RPCSVC_POOLCOUNT_MULT           35
-#define RPCSVC_CONN_READ        (128 * GF_UNIT_KB)
-#define RPCSVC_PAGE_SIZE        (128 * GF_UNIT_KB)
-
-/* Defines for RPC record and fragment assembly */
-
-#define RPCSVC_FRAGHDR_SIZE  4       /* 4-byte RPC fragment header size */
-
-/* Given the 4-byte fragment header, returns non-zero if this fragment
- * is the last fragment for the RPC record being assemebled.
- * RPC Record marking standard defines a 32 bit value as the fragment
- * header with the MSB signifying whether the fragment is the last
- * fragment for the record being asembled.
- */
-#define RPCSVC_LASTFRAG(fraghdr) ((uint32_t)(fraghdr & 0x80000000U))
-
-/* Given the 4-byte fragment header, extracts the bits that contain
- * the fragment size.
- */
-#define RPCSVC_FRAGSIZE(fraghdr) ((uint32_t)(fraghdr & 0x7fffffffU))
-
-/* RPC Record States */
-#define RPCSVC_READ_FRAGHDR     1
-#define RPCSVC_READ_FRAG        2
-/* The size in bytes, if crossed by a fragment will be handed over to the
- * vectored actor so that it can allocate its buffers the way it wants.
- * In our RPC layer, we assume that vectored RPC requests/records are never
- * spread over multiple RPC fragments since that prevents us from determining
- * whether the record should be handled in RPC layer completely or handed to
- * the vectored handler.
- */
-#define RPCSVC_VECTORED_FRAGSZ  4096
-#define RPCSVC_VECTOR_READCRED          1003
-#define RPCSVC_VECTOR_READVERFSZ        1004
-#define RPCSVC_VECTOR_READVERF          1005
-#define RPCSVC_VECTOR_IGNORE            1006
-#define RPCSVC_VECTOR_READVEC           1007
-#define RPCSVC_VECTOR_READPROCHDR       1008
-
-#define nfs_rpcsvc_record_vectored_baremsg(rs) (((rs)->state == RPCSVC_READ_FRAG) && (rs)->vecstate == 0)
-#define nfs_rpcsvc_record_vectored_cred(rs) ((rs)->vecstate == RPCSVC_VECTOR_READCRED)
-#define nfs_rpcsvc_record_vectored_verfsz(rs) ((rs)->vecstate == RPCSVC_VECTOR_READVERFSZ)
-#define nfs_rpcsvc_record_vectored_verfread(rs) ((rs)->vecstate == RPCSVC_VECTOR_READVERF)
-#define nfs_rpcsvc_record_vectored_ignore(rs) ((rs)->vecstate == RPCSVC_VECTOR_IGNORE)
-#define nfs_rpcsvc_record_vectored_readvec(rs) ((rs)->vecstate == RPCSVC_VECTOR_READVEC)
-#define nfs_rpcsvc_record_vectored_readprochdr(rs) ((rs)->vecstate == RPCSVC_VECTOR_READPROCHDR)
-#define nfs_rpcsvc_record_vectored(rs) ((rs)->fragsize > RPCSVC_VECTORED_FRAGSZ)
-/* Includes bytes up to and including the credential length field. The credlen
- * will be followed by @credlen bytes of credential data which will have to be
- * read separately by the vectored reader. After the credentials comes the
- * verifier which will also have to be read separately including the 8 bytes of
- * verf flavour and verflen.
- */
-#define RPCSVC_BARERPC_MSGSZ    32
-#define nfs_rpcsvc_record_readfraghdr(rs)   ((rs)->state == RPCSVC_READ_FRAGHDR)
-#define nfs_rpcsvc_record_readfrag(rs)      ((rs)->state == RPCSVC_READ_FRAG)
-
-#define nfs_rpcsvc_conn_rpcsvc(conn)        ((conn)->stage->svc)
-#define RPCSVC_LOWVERS  2
-#define RPCSVC_HIGHVERS 2
-
-typedef struct rpc_svc_program rpcsvc_program_t;
-/* A Stage is the event handler thread together with
- * the connections being served by this thread.
- * It is called a stage because all the actors, i.e, protocol actors,
- * defined by higher level users of the RPC layer, are executed here.
- */
-typedef struct rpc_svc_stage_context {
-        pthread_t               tid;
-        struct event_pool       *eventpool;     /* Per-stage event-pool */
-        void                    *svc;           /* Ref to the rpcsvc_t */
-} rpcsvc_stage_t;
-
-
-/* RPC Records and Fragments assembly state.
- * This is per-connection state that is used to determine
- * how much data has come in, how much more needs to be read
- * and where it needs to be read.
- *
- * All this state is then used to re-assemble network buffers into
- * RPC fragments, which are then re-assembled into RPC records.
- *
- * See RFC 1831: "RPC: Remote Procedure Call Protocol Specification Version 2",
- * particularly the section on Record Marking Standard.
- */
-typedef struct rpcsvc_record_state {
-
-        /* Pending messages storage
-         * This memory area is currently being used to assemble
-         * the latest RPC record.
-         *
-         * Note that this buffer contains the data other than the
-         * fragment headers received from the network. This is so that we can
-         * directly pass this buffer to higher layers without requiring to
-         * perform memory copies and marshalling of data.
-         */
-        struct iobuf            *activeiob;
-
-        struct iobuf            *vectoriob;
-        /* The pointer into activeiob memory, into which will go the
-         * contents from the next read from the network.
-         */
-        char                    *fragcurrent;
-
-        /* Size of the currently incomplete RPC fragment.
-         * This is filled in when the fragment header comes in.
-         * Even though only the 31 least significant bits are used from the
-         * fragment header, we use a 32 bit variable to store the size.
-         */
-        uint32_t               fragsize;
-
-        /* The fragment header is always read in here so that
-         * the RPC messages contained in a RPC records can be processed
-         * separately without copying them out of the activeiob above.
-         */
-        char                    fragheader[RPCSVC_FRAGHDR_SIZE];
-        char                    *hdrcurrent;
-
-        /* Bytes remaining to come in for the current fragment. */
-        uint32_t                remainingfrag;
-
-        /* It is possible for the frag header to be split over separate
-         * read calls, so we need to keep track of how much is left.
-         */
-        uint32_t                remainingfraghdr;
-
-        /* Record size, the total size of the RPC record, i.e. the total
-         * of all fragment sizes received till now. Does not include the size
-         * of a partial fragment which is continuing to be assembled right now.
-         */
-        int                     recordsize;
-
-        /* Current state of the record */
-        int                     state;
-
-        /* Current state of the vectored reading process. */
-        int                     vecstate;
-
-        /* Set to non-zero when the currently partial or complete fragment is
-         * the last fragment being received for the current RPC record.
-         */
-        uint32_t                islastfrag;
-
-} rpcsvc_record_state_t;
-
-
-#define RPCSVC_CONNSTATE_CONNECTED      1
-#define RPCSVC_CONNSTATE_DISCONNECTED   2
-
-#define nfs_rpcsvc_conn_check_active(conn) ((conn)->connstate==RPCSVC_CONNSTATE_CONNECTED)
-
-typedef struct rpcsvc_request rpcsvc_request_t;
-/* Contains the state for each connection that is used for transmitting and
- * receiving RPC messages.
- *
- * There is also an eventidx because each connection's fd is added to the event
- * pool of the stage to which a connection belongs.
- * Anything that can be accessed by a RPC program must be synced through
- * connlock.
- */
-typedef struct rpc_conn_state {
-
-        /* Transport or connection state */
-
-         /* Once we start working on RDMA support, this TCP specific state will
-         * have to be abstracted away.
-         */
-        int                     sockfd;
-        int                     eventidx;
-        int                     windowsize;
-
-        /* Reference to the stage which is handling this
-         * connection.
-         */
-        rpcsvc_stage_t          *stage;
-
-        /* RPC Records and Fragments assembly state.
-         * All incoming data is staged here before being
-         * called a full RPC message.
-         */
-        rpcsvc_record_state_t   rstate;
-
-         /* It is possible that a client disconnects while
-         * the higher layer RPC service is busy in a call.
-         * In this case, we cannot just free the conn
-         * structure, since the higher layer service could
-         * still have a reference to it.
-         * The refcount avoids freeing until all references
-         * have been given up, although the connection is clos()ed at the first
-         * call to unref.
-         */
-        int                     connref;
-        pthread_mutex_t         connlock;
-        int                     connstate;
-
-        /* The program that is listening for requests on this connection. */
-        rpcsvc_program_t        *program;
-
-        /* List of buffers awaiting transmission */
-        /* Accesses to txbufs between multiple threads calling
-         * rpcsvc_submit is synced through connlock. Prefer spinlock over
-         * mutex because this is a low overhead op that needs simple
-         * appending to the tx list.
-         */
-        struct list_head        txbufs;
-
-        /* Mem pool for the txbufs above. */
-        struct mem_pool         *txpool;
-
-        /* Memory pool for rpcsvc_request_t */
-        struct mem_pool         *rxpool;
-
-        /* The request which hasnt yet been handed to the RPC program because
-         * this request is being treated as a vector request and so needs some
-         * more data to be got from the network.
-         */
-        rpcsvc_request_t        *vectoredreq;
-} rpcsvc_conn_t;
-
-
-#define RPCSVC_MAX_AUTH_BYTES   400
-typedef struct rpcsvc_auth_data {
-        int             flavour;
-        int             datalen;
-        char            authdata[RPCSVC_MAX_AUTH_BYTES];
-} rpcsvc_auth_data_t;
-
-#define nfs_rpcsvc_auth_flavour(au)    ((au).flavour)
-
-/* The container for the RPC call handed up to an actor.
- * Dynamically allocated. Lives till the call reply is completely
- * transmitted.
- * */
-struct rpcsvc_request {
-        /* Connection over which this request came. */
-        rpcsvc_conn_t           *conn;
-
-        /* The identifier for the call from client.
-         * Needed to pair the reply with the call.
-         */
-        uint32_t                xid;
-
-        int                     prognum;
-
-        int                     progver;
-
-        int                     procnum;
-        /* Uid and gid filled by the rpc-auth module during the authentication
-         * phase.
-         */
-        uid_t                   uid;
-        gid_t                   gid;
-
-        /* Might want to move this to AUTH_UNIX specifix state since this array
-         * is not available for every authenticatino scheme.
-         */
-        gid_t                   auxgids[NGRPS];
-        int                     auxgidcount;
-
-
-        /* The RPC message payload, contains the data required
-         * by the program actors. This is the buffer that will need to
-         * be de-xdred by the actor.
-         */
-        struct iovec            msg;
-
-        /* The full message buffer allocated to store the RPC headers.
-         * This buffer is ref'd when allocated why RPC svc and unref'd after
-         * the buffer is handed to the actor. That means if the actor or any
-         * higher layer wants to keep this buffer around, they too must ref it
-         * right after entering the program actor.
-         */
-        struct iobuf            *recordiob;
-
-        /* Status of the RPC call, whether it was accepted or denied. */
-        int                     rpc_stat;
-
-        /* In case, the call was denied, the RPC error is stored here
-         * till the reply is sent.
-         */
-        int                     rpc_err;
-
-        /* In case the failure happened because of an authentication problem
-         * , this value needs to be assigned the correct auth error number.
-         */
-        int                     auth_err;
-
-        /* There can be cases of RPC requests where the reply needs to
-         * be built from multiple sources. For eg. where even the NFS reply can
-         * contain a payload, as in the NFSv3 read reply. Here the RPC header
-         * ,NFS header and the read data are brought together separately from
-         * different buffers, so we need to stage the buffers temporarily here
-         * before all of them get added to the connection's transmission list.
-         */
-        struct list_head        txlist;
-
-        /* While the reply record is being built, this variable keeps track
-         * of how many bytes have been added to the record.
-         */
-        size_t                  payloadsize;
-
-        /* The credentials extracted from the rpc request */
-        rpcsvc_auth_data_t      cred;
-
-        /* The verified extracted from the rpc request. In request side
-         * processing this contains the verifier sent by the client, on reply
-         * side processing, it is filled with the verified that will be
-         * sent to the client.
-         */
-        rpcsvc_auth_data_t      verf;
-
-        /* Container for a RPC program wanting to store a temp
-         * request-specific item.
-         */
-        void                    *private;
-
-};
-
-#define nfs_rpcsvc_request_program(req) ((rpcsvc_program_t *)((req)->conn->program))
-#define nfs_rpcsvc_request_program_private(req) (((rpcsvc_program_t *)((req)->conn->program))->private)
-#define nfs_rpcsvc_request_conn(req)        (req)->conn
-#define nfs_rpcsvc_program_xlator(prg)      ((prg)->actorxl)
-#define nfs_rpcsvc_request_actorxl(rq)      (nfs_rpcsvc_request_program(rq))->actorxl
-#define nfs_rpcsvc_request_accepted(req)    ((req)->rpc_stat == MSG_ACCEPTED)
-#define nfs_rpcsvc_request_accepted_success(req) ((req)->rpc_err == SUCCESS)
-#define nfs_rpcsvc_request_uid(req)         ((req)->uid)
-#define nfs_rpcsvc_request_gid(req)         ((req)->gid)
-#define nfs_rpcsvc_stage_service(stg)       ((rpcsvc_t *)((stg)->svc))
-#define nfs_rpcsvc_conn_stage(conn)         ((conn)->stage)
-#define nfs_rpcsvc_request_service(req)     (nfs_rpcsvc_stage_service(nfs_rpcsvc_conn_stage(nfs_rpcsvc_request_conn(req))))
-#define nfs_rpcsvc_request_prog_minauth(req) (nfs_rpcsvc_request_program(req)->min_auth)
-#define nfs_rpcsvc_request_cred_flavour(req) (nfs_rpcsvc_auth_flavour(req->cred))
-#define nfs_rpcsvc_request_verf_flavour(req) (nfs_rpcsvc_auth_flavour(req->verf))
-
-#define nfs_rpcsvc_request_uid(req)         ((req)->uid)
-#define nfs_rpcsvc_request_gid(req)         ((req)->gid)
-#define nfs_rpcsvc_request_private(req)     ((req)->private)
-#define nfs_rpcsvc_request_xid(req)         ((req)->xid)
-#define nfs_rpcsvc_request_set_private(req,prv)  (req)->private = (void *)(prv)
-#define nfs_rpcsvc_request_record_iob(rq)   ((rq)->recordiob)
-#define nfs_rpcsvc_request_record_ref(req)  (iobuf_ref ((req)->recordiob))
-#define nfs_rpcsvc_request_record_unref(req) (iobuf_unref ((req)->recordiob))
-
-
-#define RPCSVC_ACTOR_SUCCESS    0
-#define RPCSVC_ACTOR_ERROR      (-1)
-
-/* Functor for every type of protocol actor
- * must be defined like this.
- *
- * See the request structure for info on how to handle the request
- * in the program actor.
- *
- * On successful santify checks inside the actor, it should return
- * RPCSVC_ACTOR_SUCCESS.
- * On an error, on which the RPC layer is expected to return a reply, the actor
- * should return RPCSVC_ACTOR_ERROR.
- *
- */
-typedef int (*rpcsvc_actor) (rpcsvc_request_t *req);
-typedef int (*rpcsvc_vector_actor) (rpcsvc_request_t *req, struct iobuf *iob);
-typedef int (*rpcsvc_vector_sizer) (rpcsvc_request_t *req, ssize_t *readsize,
-                                    int *newiob);
-
-/* Every protocol actor will also need to specify the function the RPC layer
- * will use to serialize or encode the message into XDR format just before
- * transmitting on the connection.
- */
-typedef void *(*rpcsvc_encode_reply) (void *msg);
-
-/* Once the reply has been transmitted, the message will have to be de-allocated
- * , so every actor will need to provide a function that deallocates the message
- * it had allocated as a response.
- */
-typedef void (*rpcsvc_deallocate_reply) (void *msg);
-
-
-#define RPCSVC_NAME_MAX            32
-/* The descriptor for each procedure/actor that runs
- * over the RPC service.
- */
-typedef struct rpc_svc_actor_desc {
-        char                    procname[RPCSVC_NAME_MAX];
-        int                     procnum;
-        rpcsvc_actor            actor;
-
-        /* Handler for cases where the RPC requests fragments are large enough
-         * to benefit from being decoded into aligned memory addresses. While
-         * decoding the request in a non-vectored manner, due to the nature of
-         * the XDR scheme, RPC cannot guarantee memory aligned addresses for
-         * the resulting message-specific structures. Allowing a specialized
-         * handler for letting the RPC program read the data from the network
-         * directly into its alligned buffers.
-         */
-        rpcsvc_vector_actor     vector_actor;
-        rpcsvc_vector_sizer     vector_sizer;
-
-} rpcsvc_actor_t;
-
-typedef int (*rpcsvc_conn_notify_fn) (void *progpriv, rpcsvc_conn_t *conn);
-
-/* Describes a program and its version along with the function pointers
- * required to handle the procedures/actors of each program/version.
- * Never changed ever by any thread so no need for a lock.
- */
-struct rpc_svc_program {
-        char                    progname[RPCSVC_NAME_MAX];
-        int                     prognum;
-        int                     progver;
-        uint16_t                progport;       /* Registered with portmap */
-        int                     progaddrfamily; /* AF_INET or AF_INET6 */
-        char                    *proghost;      /* Bind host, can be NULL */
-        rpcsvc_actor_t          *actors;        /* All procedure handlers */
-        int                     numactors;      /* Num actors in actor array */
-        int                     proghighvers;   /* Highest ver for program
-                                                   supported by the system. */
-        int                     proglowvers;    /* Lowest ver */
-
-        /* Program specific state handed to actors */
-        void                    *private;
-
-        /* This upcall is made when a connection's refcount reaches 0 and the
-         * connection is about to be destroyed. We want to let the RPC program
-         * know that it should also now free any state it is maintaining
-         * for this connection.
-         */
-        rpcsvc_conn_notify_fn   conn_destroy;
-
-        /* Used to tell RPC program to init the state it needs to associate
-         * with the new connection.
-         */
-        rpcsvc_conn_notify_fn   conn_init;
-
-        /* An integer that identifies the min auth strength that is required
-         * by this protocol, for eg. MOUNT3 needs AUTH_UNIX at least.
-         * See RFC 1813, Section 5.2.1.
-         */
-        int                     min_auth;
-
-        /* The translator in whose context the actor must execute. This is
-         * needed to setup THIS for memory accounting to work correctly.
-         */
-        xlator_t                *actorxl;
-};
-
-
-/* Contains global state required for all the RPC services.
- */
-typedef struct rpc_svc_state {
-
-        /* Contains the list of rpcsvc_stage_t
-         * list of (program, version) handlers.
-         * other options.
-         */
-
-        /* At this point, lock is not used to protect anything. Later, it'll
-         * be used for protecting stages.
-         */
-        pthread_mutex_t         rpclock;
-
-        /* This is the first stage that is inited, so that any RPC based
-         * services that do not need multi-threaded support can just use the
-         * service right away. This is not added to the stages list
-         * declared later.
-         * This is also the stage over which all service listeners are run.
-         */
-        rpcsvc_stage_t          *defaultstage;
-
-        /* When we have multi-threaded RPC support, we'll use this to link
-         * to the multiple Stages.
-         */
-        struct list_head        stages;         /* All stages */
-
-        unsigned int            memfactor;
-
-        /* List of the authentication schemes available. */
-        struct list_head        authschemes;
-
-        /* Reference to the options */
-        dict_t                  *options;
-
-        /* Allow insecure ports. */
-        int                     allow_insecure;
-
-        glusterfs_ctx_t         *ctx;
-} rpcsvc_t;
-
-
-/* All users of RPC services should use this API to register their
- * procedure handlers.
- */
-extern int
-nfs_rpcsvc_program_register (rpcsvc_t *svc, rpcsvc_program_t program);
-
-extern int
-nfs_rpcsvc_program_unregister (rpcsvc_t *svc, rpcsvc_program_t program);
-
-/* Inits the global RPC service data structures.
- * Called in main.
- */
-extern rpcsvc_t *
-nfs_rpcsvc_init (glusterfs_ctx_t *ctx, dict_t *options);
-
-
-extern int
-nfs_rpcsvc_submit_message (rpcsvc_request_t * req, struct iovec msg,
-                           struct iobuf *iob);
-
-int
-nfs_rpcsvc_submit_generic (rpcsvc_request_t *req, struct iovec msgvec,
-                           struct iobuf *msg);
-#define nfs_rpcsvc_record_currentfrag_addr(rs) ((rs)->fragcurrent)
-#define nfs_rpcsvc_record_currenthdr_addr(rs) ((rs)->hdrcurrent)
-
-#define nfs_rpcsvc_record_update_currentfrag(rs, size)          \
-                        do {                                    \
-                                (rs)->fragcurrent += size;      \
-                        } while (0)                             \
-
-#define nfs_rpcsvc_record_update_currenthdr(rs, size)           \
-                        do {                                    \
-                                (rs)->hdrcurrent += size;       \
-                        } while (0)                             \
-
-
-/* These are used to differentiate between multiple txbufs which form
- * a single RPC record. For eg, one purpose we use these for is to
- * prevent dividing a RPC record over multiple TCP segments. Multiple
- * TCP segments are possible for a single RPC record because we generally do not
- * have control over how the kernel's TCP segments the buffers when putting
- * them on the wire. So, on Linux, we use these to set TCP_CORK to create
- * a single TCP segment from multiple txbufs that are part of the same RPC
- * record. This improves network performance by reducing tiny message
- * transmissions.
- */
-#define RPCSVC_TXB_FIRST        0x1
-#define RPCSVC_TXB_LAST         0x2
-
-/* The list of buffers appended to a connection's pending
- * transmission list.
- */
-typedef struct rpcsvc_txbuf {
-        struct list_head        txlist;
-        /* The iobuf which contains the full message to be transmitted */
-        struct iobuf            *iob;
-
-        /* For vectored messages from an RPC program, we need to be able
-         * maintain a ref to an iobuf which we do not have access to directly
-         * except through the iobref which in turn could've been passed to
-         * the RPC program by a higher layer.
-         *
-         * So either the iob is defined or iobref is defined for a reply,
-         * never both.
-         */
-        struct iobref           *iobref;
-        /* In order to handle non-blocking writes, we'll need to keep track of
-         * how much data from an iobuf has been written and where the next
-         * transmission needs to start from. This iov.base points to the base of
-         * the iobuf, iov.len is the size of iobuf being used for the message
-         * from the total size in the iobuf.
-         */
-        struct iovec            buf;
-        /* offset is the point from where the next transmission for this buffer
-         * should start.
-         */
-        size_t                  offset;
-
-        /* This is a special field that tells us what kind of transmission
-         * behaviour to provide to a particular buffer.
-         * See the RPCSVC_TXB_* defines for more info.
-         */
-        int                     txbehave;
-} rpcsvc_txbuf_t;
-
-extern int
-nfs_rpcsvc_error_reply (rpcsvc_request_t *req);
-
-#define RPCSVC_PEER_STRLEN      1024
-#define RPCSVC_AUTH_ACCEPT      1
-#define RPCSVC_AUTH_REJECT      2
-#define RPCSVC_AUTH_DONTCARE    3
-
-extern int
-nfs_rpcsvc_conn_peername (rpcsvc_conn_t *conn, char *hostname, int hostlen);
-
-extern int
-nfs_rpcsvc_conn_peeraddr (rpcsvc_conn_t *conn, char *addrstr, int addrlen,
-                          struct sockaddr *returnsa, socklen_t sasize);
-
-extern int
-nfs_rpcsvc_conn_peer_check (dict_t *options, char *volname,rpcsvc_conn_t *conn);
-
-extern int
-nfs_rpcsvc_conn_privport_check (rpcsvc_t *svc, char *volname,
-                                rpcsvc_conn_t *conn);
-#define nfs_rpcsvc_request_seterr(req, err)                 (req)->rpc_err = err
-#define nfs_rpcsvc_request_set_autherr(req, err)            (req)->auth_err = err
-
-extern void
-nfs_rpcsvc_conn_deinit (rpcsvc_conn_t *conn);
-extern void nfs_rpcsvc_conn_ref (rpcsvc_conn_t *conn);
-extern void nfs_rpcsvc_conn_unref (rpcsvc_conn_t *conn);
-
-extern int nfs_rpcsvc_submit_vectors (rpcsvc_request_t *req);
-
-extern int nfs_rpcsvc_request_attach_vector (rpcsvc_request_t *req,
-                                             struct iovec msgvec,
-                                             struct iobuf *iob,
-                                             struct iobref *ioref,
-                                             int finalvector);
-extern int
-nfs_rpcsvc_request_attach_vectors (rpcsvc_request_t *req, struct iovec *payload,
-                                   int vcount, struct iobref *piobref);
-
-typedef int (*auth_init_conn) (rpcsvc_conn_t *conn, void *priv);
-typedef int (*auth_init_request) (rpcsvc_request_t *req, void *priv);
-typedef int (*auth_request_authenticate) (rpcsvc_request_t *req, void *priv);
-
-/* This structure needs to be registered by every authentication scheme.
- * Our authentication schemes are stored per connection because
- * each connection will end up using a different authentication scheme.
- */
-typedef struct rpcsvc_auth_ops {
-        auth_init_conn                conn_init;
-        auth_init_request             request_init;
-        auth_request_authenticate     authenticate;
-} rpcsvc_auth_ops_t;
-
-typedef struct rpcsvc_auth_flavour_desc {
-        char                    authname[RPCSVC_NAME_MAX];
-        int                     authnum;
-        rpcsvc_auth_ops_t       *authops;
-        void                    *authprivate;
-} rpcsvc_auth_t;
-
-typedef void * (*rpcsvc_auth_initer_t) (rpcsvc_t *svc, dict_t *options);
-
-struct rpcsvc_auth_list {
-        struct list_head        authlist;
-        rpcsvc_auth_initer_t    init;
-        /* Should be the name with which we identify the auth scheme given
-         * in the volfile options.
-         * This should be different from the authname in rpc_auth_t
-         * in way that makes it easier to specify this scheme in the volfile.
-         * This is because the technical names of the schemes can be a bit
-         * arcane.
-         */
-        char                    name[RPCSVC_NAME_MAX];
-        rpcsvc_auth_t           *auth;
-        int                     enable;
-};
-
-extern int
-nfs_rpcsvc_auth_request_init (rpcsvc_request_t *req);
-
-extern int
-nfs_rpcsvc_auth_init (rpcsvc_t *svc, dict_t *options);
-
-extern int
-nfs_rpcsvc_auth_conn_init (rpcsvc_conn_t *conn);
-
-extern int
-nfs_rpcsvc_authenticate (rpcsvc_request_t *req);
-
-extern int
-nfs_rpcsvc_auth_array (rpcsvc_t *svc, char *volname, int *autharr, int arrlen);
-
-/* If the request has been sent using AUTH_UNIX, this function returns the
- * auxiliary gids as an array, otherwise, it returns NULL.
- * Move to auth-unix specific source file when we need to modularize the
- * authentication code even further to support mode auth schemes.
- */
-extern gid_t *
-nfs_rpcsvc_auth_unix_auxgids (rpcsvc_request_t *req, int *arrlen);
-
-extern int
-nfs_rpcsvc_combine_gen_spec_volume_checks (int gen, int spec);
-
-extern char *
-nfs_rpcsvc_volume_allowed (dict_t *options, char *volname);
-#endif
diff --git a/xlators/nfs/lib/src/xdr-common.h b/xlators/nfs/lib/src/xdr-common.h
deleted file mode 100644
index 0f60cd2c9..000000000
--- a/xlators/nfs/lib/src/xdr-common.h
+++ /dev/null
@@ -1,48 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _NFS_XDR_COMMON_H_
-#define _NFS_XDR_COMMON_H_
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include <rpc/rpc.h>
-#define NFS_XDR_BYTES_PER_UNIT      4
-
-/* Returns the address of the byte that follows the
- * last byte used for decoding the previous xdr component.
- * For eg, once the RPC call for NFS has been decoded, thie macro will return
- * the address from which the NFS header starts.
- */
-#define nfs_xdr_decoded_remaining_addr(xdr)        ((&xdr)->x_private)
-
-/* Returns the length of the remaining record after the previous decode
- * operation completed.
- */
-#define nfs_xdr_decoded_remaining_len(xdr)         ((&xdr)->x_handy)
-
-/* Returns the number of bytes used by the last encode operation. */
-#define nfs_xdr_encoded_length(xdr) (((size_t)(&xdr)->x_private) - ((size_t)(&xdr)->x_base))
-
-#define nfs_xdr_decoded_length(xdr) (((size_t)(&xdr)->x_private) - ((size_t)(&xdr)->x_base))
-
-#endif
diff --git a/xlators/nfs/lib/src/xdr-nfs3.c b/xlators/nfs/lib/src/xdr-nfs3.c
deleted file mode 100644
index 97ec1b6fe..000000000
--- a/xlators/nfs/lib/src/xdr-nfs3.c
+++ /dev/null
@@ -1,1897 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#include "xdr-nfs3.h"
-#include "mem-pool.h"
-
-#if GF_DARWIN_HOST_OS
-#define xdr_u_quad_t xdr_u_int64_t
-#define xdr_quad_t   xdr_int64_t
-#define xdr_uint32_t xdr_u_int32_t
-#endif
-
-bool_t
-xdr_uint64 (XDR *xdrs, uint64 *objp)
-{
-	 if (!xdr_uint64_t (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_int64 (XDR *xdrs, int64 *objp)
-{
-	 if (!xdr_int64_t (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_uint32 (XDR *xdrs, uint32 *objp)
-{
-	 if (!xdr_uint32_t (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_int32 (XDR *xdrs, int32 *objp)
-{
-	 if (!xdr_int32_t (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_filename3 (XDR *xdrs, filename3 *objp)
-{
-	 if (!xdr_string (xdrs, objp, ~0))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_nfspath3 (XDR *xdrs, nfspath3 *objp)
-{
-	 if (!xdr_string (xdrs, objp, ~0))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_fileid3 (XDR *xdrs, fileid3 *objp)
-{
-	 if (!xdr_uint64 (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_cookie3 (XDR *xdrs, cookie3 *objp)
-{
-	 if (!xdr_uint64 (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_cookieverf3 (XDR *xdrs, cookieverf3 objp)
-{
-	 if (!xdr_opaque (xdrs, objp, NFS3_COOKIEVERFSIZE))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_createverf3 (XDR *xdrs, createverf3 objp)
-{
-	 if (!xdr_opaque (xdrs, objp, NFS3_CREATEVERFSIZE))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_writeverf3 (XDR *xdrs, writeverf3 objp)
-{
-	 if (!xdr_opaque (xdrs, objp, NFS3_WRITEVERFSIZE))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_uid3 (XDR *xdrs, uid3 *objp)
-{
-	 if (!xdr_uint32 (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_gid3 (XDR *xdrs, gid3 *objp)
-{
-	 if (!xdr_uint32 (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_size3 (XDR *xdrs, size3 *objp)
-{
-	 if (!xdr_uint64 (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_offset3 (XDR *xdrs, offset3 *objp)
-{
-	 if (!xdr_uint64 (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mode3 (XDR *xdrs, mode3 *objp)
-{
-	 if (!xdr_uint32 (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_count3 (XDR *xdrs, count3 *objp)
-{
-	 if (!xdr_uint32 (xdrs, objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_nfsstat3 (XDR *xdrs, nfsstat3 *objp)
-{
-	 if (!xdr_enum (xdrs, (enum_t *) objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_ftype3 (XDR *xdrs, ftype3 *objp)
-{
-	 if (!xdr_enum (xdrs, (enum_t *) objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_specdata3 (XDR *xdrs, specdata3 *objp)
-{
-	 if (!xdr_uint32 (xdrs, &objp->specdata1))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->specdata2))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_nfs_fh3 (XDR *xdrs, nfs_fh3 *objp)
-{
-	 if (!xdr_bytes (xdrs, (char **)&objp->data.data_val, (u_int *) &objp->data.data_len, NFS3_FHSIZE))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_nfstime3 (XDR *xdrs, nfstime3 *objp)
-{
-	 if (!xdr_uint32 (xdrs, &objp->seconds))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->nseconds))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_fattr3 (XDR *xdrs, fattr3 *objp)
-{
-	 if (!xdr_ftype3 (xdrs, &objp->type))
-		 return FALSE;
-	 if (!xdr_mode3 (xdrs, &objp->mode))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->nlink))
-		 return FALSE;
-	 if (!xdr_uid3 (xdrs, &objp->uid))
-		 return FALSE;
-	 if (!xdr_gid3 (xdrs, &objp->gid))
-		 return FALSE;
-	 if (!xdr_size3 (xdrs, &objp->size))
-		 return FALSE;
-	 if (!xdr_size3 (xdrs, &objp->used))
-		 return FALSE;
-	 if (!xdr_specdata3 (xdrs, &objp->rdev))
-		 return FALSE;
-	 if (!xdr_uint64 (xdrs, &objp->fsid))
-		 return FALSE;
-	 if (!xdr_fileid3 (xdrs, &objp->fileid))
-		 return FALSE;
-	 if (!xdr_nfstime3 (xdrs, &objp->atime))
-		 return FALSE;
-	 if (!xdr_nfstime3 (xdrs, &objp->mtime))
-		 return FALSE;
-	 if (!xdr_nfstime3 (xdrs, &objp->ctime))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_post_op_attr (XDR *xdrs, post_op_attr *objp)
-{
-	 if (!xdr_bool (xdrs, &objp->attributes_follow))
-		 return FALSE;
-	switch (objp->attributes_follow) {
-	case TRUE:
-		 if (!xdr_fattr3 (xdrs, &objp->post_op_attr_u.attributes))
-			 return FALSE;
-		break;
-	case FALSE:
-		break;
-	default:
-		return FALSE;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_wcc_attr (XDR *xdrs, wcc_attr *objp)
-{
-	 if (!xdr_size3 (xdrs, &objp->size))
-		 return FALSE;
-	 if (!xdr_nfstime3 (xdrs, &objp->mtime))
-		 return FALSE;
-	 if (!xdr_nfstime3 (xdrs, &objp->ctime))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_pre_op_attr (XDR *xdrs, pre_op_attr *objp)
-{
-	 if (!xdr_bool (xdrs, &objp->attributes_follow))
-		 return FALSE;
-	switch (objp->attributes_follow) {
-	case TRUE:
-		 if (!xdr_wcc_attr (xdrs, &objp->pre_op_attr_u.attributes))
-			 return FALSE;
-		break;
-	case FALSE:
-		break;
-	default:
-		return FALSE;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_wcc_data (XDR *xdrs, wcc_data *objp)
-{
-	 if (!xdr_pre_op_attr (xdrs, &objp->before))
-		 return FALSE;
-	 if (!xdr_post_op_attr (xdrs, &objp->after))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_post_op_fh3 (XDR *xdrs, post_op_fh3 *objp)
-{
-	 if (!xdr_bool (xdrs, &objp->handle_follows))
-		 return FALSE;
-	switch (objp->handle_follows) {
-	case TRUE:
-		 if (!xdr_nfs_fh3 (xdrs, &objp->post_op_fh3_u.handle))
-			 return FALSE;
-		break;
-	case FALSE:
-		break;
-	default:
-		return FALSE;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_time_how (XDR *xdrs, time_how *objp)
-{
-	 if (!xdr_enum (xdrs, (enum_t *) objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_set_mode3 (XDR *xdrs, set_mode3 *objp)
-{
-	 if (!xdr_bool (xdrs, &objp->set_it))
-		 return FALSE;
-	switch (objp->set_it) {
-	case TRUE:
-		 if (!xdr_mode3 (xdrs, &objp->set_mode3_u.mode))
-			 return FALSE;
-		break;
-	default:
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_set_uid3 (XDR *xdrs, set_uid3 *objp)
-{
-	 if (!xdr_bool (xdrs, &objp->set_it))
-		 return FALSE;
-	switch (objp->set_it) {
-	case TRUE:
-		 if (!xdr_uid3 (xdrs, &objp->set_uid3_u.uid))
-			 return FALSE;
-		break;
-	default:
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_set_gid3 (XDR *xdrs, set_gid3 *objp)
-{
-	 if (!xdr_bool (xdrs, &objp->set_it))
-		 return FALSE;
-	switch (objp->set_it) {
-	case TRUE:
-		 if (!xdr_gid3 (xdrs, &objp->set_gid3_u.gid))
-			 return FALSE;
-		break;
-	default:
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_set_size3 (XDR *xdrs, set_size3 *objp)
-{
-	 if (!xdr_bool (xdrs, &objp->set_it))
-		 return FALSE;
-	switch (objp->set_it) {
-	case TRUE:
-		 if (!xdr_size3 (xdrs, &objp->set_size3_u.size))
-			 return FALSE;
-		break;
-	default:
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_set_atime (XDR *xdrs, set_atime *objp)
-{
-	 if (!xdr_time_how (xdrs, &objp->set_it))
-		 return FALSE;
-	switch (objp->set_it) {
-	case SET_TO_CLIENT_TIME:
-		 if (!xdr_nfstime3 (xdrs, &objp->set_atime_u.atime))
-			 return FALSE;
-		break;
-	default:
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_set_mtime (XDR *xdrs, set_mtime *objp)
-{
-	 if (!xdr_time_how (xdrs, &objp->set_it))
-		 return FALSE;
-	switch (objp->set_it) {
-	case SET_TO_CLIENT_TIME:
-		 if (!xdr_nfstime3 (xdrs, &objp->set_mtime_u.mtime))
-			 return FALSE;
-		break;
-	default:
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_sattr3 (XDR *xdrs, sattr3 *objp)
-{
-	 if (!xdr_set_mode3 (xdrs, &objp->mode))
-		 return FALSE;
-	 if (!xdr_set_uid3 (xdrs, &objp->uid))
-		 return FALSE;
-	 if (!xdr_set_gid3 (xdrs, &objp->gid))
-		 return FALSE;
-	 if (!xdr_set_size3 (xdrs, &objp->size))
-		 return FALSE;
-	 if (!xdr_set_atime (xdrs, &objp->atime))
-		 return FALSE;
-	 if (!xdr_set_mtime (xdrs, &objp->mtime))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_diropargs3 (XDR *xdrs, diropargs3 *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->dir))
-		 return FALSE;
-	 if (!xdr_filename3 (xdrs, &objp->name))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_getattr3args (XDR *xdrs, getattr3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->object))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_getattr3resok (XDR *xdrs, getattr3resok *objp)
-{
-	 if (!xdr_fattr3 (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_getattr3res (XDR *xdrs, getattr3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_getattr3resok (xdrs, &objp->getattr3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_sattrguard3 (XDR *xdrs, sattrguard3 *objp)
-{
-	 if (!xdr_bool (xdrs, &objp->check))
-		 return FALSE;
-	switch (objp->check) {
-	case TRUE:
-		 if (!xdr_nfstime3 (xdrs, &objp->sattrguard3_u.obj_ctime))
-			 return FALSE;
-		break;
-	case FALSE:
-		break;
-	default:
-		return FALSE;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_setattr3args (XDR *xdrs, setattr3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->object))
-		 return FALSE;
-	 if (!xdr_sattr3 (xdrs, &objp->new_attributes))
-		 return FALSE;
-	 if (!xdr_sattrguard3 (xdrs, &objp->guard))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_setattr3resok (XDR *xdrs, setattr3resok *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->obj_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_setattr3resfail (XDR *xdrs, setattr3resfail *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->obj_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_setattr3res (XDR *xdrs, setattr3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_setattr3resok (xdrs, &objp->setattr3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_setattr3resfail (xdrs, &objp->setattr3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_lookup3args (XDR *xdrs, lookup3args *objp)
-{
-	 if (!xdr_diropargs3 (xdrs, &objp->what))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_lookup3resok (XDR *xdrs, lookup3resok *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->object))
-		 return FALSE;
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	 if (!xdr_post_op_attr (xdrs, &objp->dir_attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_lookup3resfail (XDR *xdrs, lookup3resfail *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->dir_attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_lookup3res (XDR *xdrs, lookup3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_lookup3resok (xdrs, &objp->lookup3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_lookup3resfail (xdrs, &objp->lookup3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_access3args (XDR *xdrs, access3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->object))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->access))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_access3resok (XDR *xdrs, access3resok *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->access))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_access3resfail (XDR *xdrs, access3resfail *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_access3res (XDR *xdrs, access3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_access3resok (xdrs, &objp->access3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_access3resfail (xdrs, &objp->access3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_readlink3args (XDR *xdrs, readlink3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->symlink))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_readlink3resok (XDR *xdrs, readlink3resok *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->symlink_attributes))
-		 return FALSE;
-	 if (!xdr_nfspath3 (xdrs, &objp->data))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_readlink3resfail (XDR *xdrs, readlink3resfail *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->symlink_attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_readlink3res (XDR *xdrs, readlink3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_readlink3resok (xdrs, &objp->readlink3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_readlink3resfail (xdrs, &objp->readlink3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_read3args (XDR *xdrs, read3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->file))
-		 return FALSE;
-	 if (!xdr_offset3 (xdrs, &objp->offset))
-		 return FALSE;
-	 if (!xdr_count3 (xdrs, &objp->count))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_read3resok_nocopy (XDR *xdrs, read3resok *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->file_attributes))
-		 return FALSE;
-	 if (!xdr_count3 (xdrs, &objp->count))
-		 return FALSE;
-	 if (!xdr_bool (xdrs, &objp->eof))
-		 return FALSE;
-         if (!xdr_u_int (xdrs, (u_int *) &objp->data.data_len))
-                 return FALSE;
-	return TRUE;
-}
-
-
-bool_t
-xdr_read3resok (XDR *xdrs, read3resok *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->file_attributes))
-		 return FALSE;
-	 if (!xdr_count3 (xdrs, &objp->count))
-		 return FALSE;
-	 if (!xdr_bool (xdrs, &objp->eof))
-		 return FALSE;
-	 if (!xdr_bytes (xdrs, (char **)&objp->data.data_val, (u_int *) &objp->data.data_len, ~0))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_read3resfail (XDR *xdrs, read3resfail *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->file_attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-
-bool_t
-xdr_read3res_nocopy (XDR *xdrs, read3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_read3resok_nocopy (xdrs, &objp->read3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_read3resfail (xdrs, &objp->read3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-
-bool_t
-xdr_read3res (XDR *xdrs, read3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_read3resok (xdrs, &objp->read3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_read3resfail (xdrs, &objp->read3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_stable_how (XDR *xdrs, stable_how *objp)
-{
-	 if (!xdr_enum (xdrs, (enum_t *) objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_write3args (XDR *xdrs, write3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->file))
-		 return FALSE;
-	 if (!xdr_offset3 (xdrs, &objp->offset))
-		 return FALSE;
-	 if (!xdr_count3 (xdrs, &objp->count))
-		 return FALSE;
-	 if (!xdr_stable_how (xdrs, &objp->stable))
-		 return FALSE;
-
-         /* Added specifically to avoid copies from the xdr buffer into
-          * the write3args structure, which will also require an already
-          * allocated buffer. That is not optimal.
-          */
-         if (!xdr_u_int (xdrs, (u_int *) &objp->data.data_len))
-                 return FALSE;
-
-         /* The remaining bytes in the xdr buffer are the bytes that need to be
-          * written. See how these bytes are extracted in the xdr_to_write3args
-          * code path. Be careful, while using the write3args structure, since
-          * only the data.data_len has been filled. The actual data is
-          * extracted in xdr_to_write3args path.
-          */
-
-         /*	 if (!xdr_bytes (xdrs, (char **)&objp->data.data_val, (u_int *) &objp->data.data_len, ~0))
-		 return FALSE;
-                 */
-	return TRUE;
-}
-
-bool_t
-xdr_write3resok (XDR *xdrs, write3resok *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->file_wcc))
-		 return FALSE;
-	 if (!xdr_count3 (xdrs, &objp->count))
-		 return FALSE;
-	 if (!xdr_stable_how (xdrs, &objp->committed))
-		 return FALSE;
-	 if (!xdr_writeverf3 (xdrs, objp->verf))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_write3resfail (XDR *xdrs, write3resfail *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->file_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_write3res (XDR *xdrs, write3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_write3resok (xdrs, &objp->write3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_write3resfail (xdrs, &objp->write3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_createmode3 (XDR *xdrs, createmode3 *objp)
-{
-	 if (!xdr_enum (xdrs, (enum_t *) objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_createhow3 (XDR *xdrs, createhow3 *objp)
-{
-	 if (!xdr_createmode3 (xdrs, &objp->mode))
-		 return FALSE;
-	switch (objp->mode) {
-	case UNCHECKED:
-	case GUARDED:
-		 if (!xdr_sattr3 (xdrs, &objp->createhow3_u.obj_attributes))
-			 return FALSE;
-		break;
-	case EXCLUSIVE:
-		 if (!xdr_createverf3 (xdrs, objp->createhow3_u.verf))
-			 return FALSE;
-		break;
-	default:
-		return FALSE;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_create3args (XDR *xdrs, create3args *objp)
-{
-	 if (!xdr_diropargs3 (xdrs, &objp->where))
-		 return FALSE;
-	 if (!xdr_createhow3 (xdrs, &objp->how))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_create3resok (XDR *xdrs, create3resok *objp)
-{
-	 if (!xdr_post_op_fh3 (xdrs, &objp->obj))
-		 return FALSE;
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_create3resfail (XDR *xdrs, create3resfail *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_create3res (XDR *xdrs, create3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_create3resok (xdrs, &objp->create3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_create3resfail (xdrs, &objp->create3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_mkdir3args (XDR *xdrs, mkdir3args *objp)
-{
-	 if (!xdr_diropargs3 (xdrs, &objp->where))
-		 return FALSE;
-	 if (!xdr_sattr3 (xdrs, &objp->attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mkdir3resok (XDR *xdrs, mkdir3resok *objp)
-{
-	 if (!xdr_post_op_fh3 (xdrs, &objp->obj))
-		 return FALSE;
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mkdir3resfail (XDR *xdrs, mkdir3resfail *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mkdir3res (XDR *xdrs, mkdir3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_mkdir3resok (xdrs, &objp->mkdir3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_mkdir3resfail (xdrs, &objp->mkdir3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_symlinkdata3 (XDR *xdrs, symlinkdata3 *objp)
-{
-	 if (!xdr_sattr3 (xdrs, &objp->symlink_attributes))
-		 return FALSE;
-	 if (!xdr_nfspath3 (xdrs, &objp->symlink_data))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_symlink3args (XDR *xdrs, symlink3args *objp)
-{
-	 if (!xdr_diropargs3 (xdrs, &objp->where))
-		 return FALSE;
-	 if (!xdr_symlinkdata3 (xdrs, &objp->symlink))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_symlink3resok (XDR *xdrs, symlink3resok *objp)
-{
-	 if (!xdr_post_op_fh3 (xdrs, &objp->obj))
-		 return FALSE;
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_symlink3resfail (XDR *xdrs, symlink3resfail *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_symlink3res (XDR *xdrs, symlink3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_symlink3resok (xdrs, &objp->symlink3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_symlink3resfail (xdrs, &objp->symlink3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_devicedata3 (XDR *xdrs, devicedata3 *objp)
-{
-	 if (!xdr_sattr3 (xdrs, &objp->dev_attributes))
-		 return FALSE;
-	 if (!xdr_specdata3 (xdrs, &objp->spec))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mknoddata3 (XDR *xdrs, mknoddata3 *objp)
-{
-	 if (!xdr_ftype3 (xdrs, &objp->type))
-		 return FALSE;
-	switch (objp->type) {
-	case NF3CHR:
-	case NF3BLK:
-		 if (!xdr_devicedata3 (xdrs, &objp->mknoddata3_u.device))
-			 return FALSE;
-		break;
-	case NF3SOCK:
-	case NF3FIFO:
-		 if (!xdr_sattr3 (xdrs, &objp->mknoddata3_u.pipe_attributes))
-			 return FALSE;
-		break;
-	default:
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_mknod3args (XDR *xdrs, mknod3args *objp)
-{
-	 if (!xdr_diropargs3 (xdrs, &objp->where))
-		 return FALSE;
-	 if (!xdr_mknoddata3 (xdrs, &objp->what))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mknod3resok (XDR *xdrs, mknod3resok *objp)
-{
-	 if (!xdr_post_op_fh3 (xdrs, &objp->obj))
-		 return FALSE;
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mknod3resfail (XDR *xdrs, mknod3resfail *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mknod3res (XDR *xdrs, mknod3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_mknod3resok (xdrs, &objp->mknod3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_mknod3resfail (xdrs, &objp->mknod3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_remove3args (XDR *xdrs, remove3args *objp)
-{
-	 if (!xdr_diropargs3 (xdrs, &objp->object))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_remove3resok (XDR *xdrs, remove3resok *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_remove3resfail (XDR *xdrs, remove3resfail *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_remove3res (XDR *xdrs, remove3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_remove3resok (xdrs, &objp->remove3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_remove3resfail (xdrs, &objp->remove3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_rmdir3args (XDR *xdrs, rmdir3args *objp)
-{
-	 if (!xdr_diropargs3 (xdrs, &objp->object))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_rmdir3resok (XDR *xdrs, rmdir3resok *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_rmdir3resfail (XDR *xdrs, rmdir3resfail *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->dir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_rmdir3res (XDR *xdrs, rmdir3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_rmdir3resok (xdrs, &objp->rmdir3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_rmdir3resfail (xdrs, &objp->rmdir3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_rename3args (XDR *xdrs, rename3args *objp)
-{
-	 if (!xdr_diropargs3 (xdrs, &objp->from))
-		 return FALSE;
-	 if (!xdr_diropargs3 (xdrs, &objp->to))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_rename3resok (XDR *xdrs, rename3resok *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->fromdir_wcc))
-		 return FALSE;
-	 if (!xdr_wcc_data (xdrs, &objp->todir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_rename3resfail (XDR *xdrs, rename3resfail *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->fromdir_wcc))
-		 return FALSE;
-	 if (!xdr_wcc_data (xdrs, &objp->todir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_rename3res (XDR *xdrs, rename3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_rename3resok (xdrs, &objp->rename3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_rename3resfail (xdrs, &objp->rename3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_link3args (XDR *xdrs, link3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->file))
-		 return FALSE;
-	 if (!xdr_diropargs3 (xdrs, &objp->link))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_link3resok (XDR *xdrs, link3resok *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->file_attributes))
-		 return FALSE;
-	 if (!xdr_wcc_data (xdrs, &objp->linkdir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_link3resfail (XDR *xdrs, link3resfail *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->file_attributes))
-		 return FALSE;
-	 if (!xdr_wcc_data (xdrs, &objp->linkdir_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_link3res (XDR *xdrs, link3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_link3resok (xdrs, &objp->link3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_link3resfail (xdrs, &objp->link3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_readdir3args (XDR *xdrs, readdir3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->dir))
-		 return FALSE;
-	 if (!xdr_cookie3 (xdrs, &objp->cookie))
-		 return FALSE;
-	 if (!xdr_cookieverf3 (xdrs, objp->cookieverf))
-		 return FALSE;
-	 if (!xdr_count3 (xdrs, &objp->count))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_entry3 (XDR *xdrs, entry3 *objp)
-{
-	 if (!xdr_fileid3 (xdrs, &objp->fileid))
-		 return FALSE;
-	 if (!xdr_filename3 (xdrs, &objp->name))
-		 return FALSE;
-	 if (!xdr_cookie3 (xdrs, &objp->cookie))
-		 return FALSE;
-	 if (!xdr_pointer (xdrs, (char **)&objp->nextentry, sizeof (entry3), (xdrproc_t) xdr_entry3))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_dirlist3 (XDR *xdrs, dirlist3 *objp)
-{
-	 if (!xdr_pointer (xdrs, (char **)&objp->entries, sizeof (entry3), (xdrproc_t) xdr_entry3))
-		 return FALSE;
-	 if (!xdr_bool (xdrs, &objp->eof))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_readdir3resok (XDR *xdrs, readdir3resok *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->dir_attributes))
-		 return FALSE;
-	 if (!xdr_cookieverf3 (xdrs, objp->cookieverf))
-		 return FALSE;
-	 if (!xdr_dirlist3 (xdrs, &objp->reply))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_readdir3resfail (XDR *xdrs, readdir3resfail *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->dir_attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_readdir3res (XDR *xdrs, readdir3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_readdir3resok (xdrs, &objp->readdir3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_readdir3resfail (xdrs, &objp->readdir3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_readdirp3args (XDR *xdrs, readdirp3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->dir))
-		 return FALSE;
-	 if (!xdr_cookie3 (xdrs, &objp->cookie))
-		 return FALSE;
-	 if (!xdr_cookieverf3 (xdrs, objp->cookieverf))
-		 return FALSE;
-	 if (!xdr_count3 (xdrs, &objp->dircount))
-		 return FALSE;
-	 if (!xdr_count3 (xdrs, &objp->maxcount))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_entryp3 (XDR *xdrs, entryp3 *objp)
-{
-	 if (!xdr_fileid3 (xdrs, &objp->fileid))
-		 return FALSE;
-	 if (!xdr_filename3 (xdrs, &objp->name))
-		 return FALSE;
-	 if (!xdr_cookie3 (xdrs, &objp->cookie))
-		 return FALSE;
-	 if (!xdr_post_op_attr (xdrs, &objp->name_attributes))
-		 return FALSE;
-	 if (!xdr_post_op_fh3 (xdrs, &objp->name_handle))
-		 return FALSE;
-	 if (!xdr_pointer (xdrs, (char **)&objp->nextentry, sizeof (entryp3), (xdrproc_t) xdr_entryp3))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_dirlistp3 (XDR *xdrs, dirlistp3 *objp)
-{
-	 if (!xdr_pointer (xdrs, (char **)&objp->entries, sizeof (entryp3), (xdrproc_t) xdr_entryp3))
-		 return FALSE;
-	 if (!xdr_bool (xdrs, &objp->eof))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_readdirp3resok (XDR *xdrs, readdirp3resok *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->dir_attributes))
-		 return FALSE;
-	 if (!xdr_cookieverf3 (xdrs, objp->cookieverf))
-		 return FALSE;
-	 if (!xdr_dirlistp3 (xdrs, &objp->reply))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_readdirp3resfail (XDR *xdrs, readdirp3resfail *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->dir_attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_readdirp3res (XDR *xdrs, readdirp3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_readdirp3resok (xdrs, &objp->readdirp3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_readdirp3resfail (xdrs, &objp->readdirp3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_fsstat3args (XDR *xdrs, fsstat3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->fsroot))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_fsstat3resok (XDR *xdrs, fsstat3resok *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	 if (!xdr_size3 (xdrs, &objp->tbytes))
-		 return FALSE;
-	 if (!xdr_size3 (xdrs, &objp->fbytes))
-		 return FALSE;
-	 if (!xdr_size3 (xdrs, &objp->abytes))
-		 return FALSE;
-	 if (!xdr_size3 (xdrs, &objp->tfiles))
-		 return FALSE;
-	 if (!xdr_size3 (xdrs, &objp->ffiles))
-		 return FALSE;
-	 if (!xdr_size3 (xdrs, &objp->afiles))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->invarsec))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_fsstat3resfail (XDR *xdrs, fsstat3resfail *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_fsstat3res (XDR *xdrs, fsstat3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_fsstat3resok (xdrs, &objp->fsstat3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_fsstat3resfail (xdrs, &objp->fsstat3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_fsinfo3args (XDR *xdrs, fsinfo3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->fsroot))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_fsinfo3resok (XDR *xdrs, fsinfo3resok *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->rtmax))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->rtpref))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->rtmult))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->wtmax))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->wtpref))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->wtmult))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->dtpref))
-		 return FALSE;
-	 if (!xdr_size3 (xdrs, &objp->maxfilesize))
-		 return FALSE;
-	 if (!xdr_nfstime3 (xdrs, &objp->time_delta))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->properties))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_fsinfo3resfail (XDR *xdrs, fsinfo3resfail *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_fsinfo3res (XDR *xdrs, fsinfo3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_fsinfo3resok (xdrs, &objp->fsinfo3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_fsinfo3resfail (xdrs, &objp->fsinfo3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_pathconf3args (XDR *xdrs, pathconf3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->object))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_pathconf3resok (XDR *xdrs, pathconf3resok *objp)
-{
-	register int32_t *buf;
-
-
-	if (xdrs->x_op == XDR_ENCODE) {
-		 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-			 return FALSE;
-		 if (!xdr_uint32 (xdrs, &objp->linkmax))
-			 return FALSE;
-		 if (!xdr_uint32 (xdrs, &objp->name_max))
-			 return FALSE;
-		buf = XDR_INLINE (xdrs, 4 * BYTES_PER_XDR_UNIT);
-		if (buf == NULL) {
-			 if (!xdr_bool (xdrs, &objp->no_trunc))
-				 return FALSE;
-			 if (!xdr_bool (xdrs, &objp->chown_restricted))
-				 return FALSE;
-			 if (!xdr_bool (xdrs, &objp->case_insensitive))
-				 return FALSE;
-			 if (!xdr_bool (xdrs, &objp->case_preserving))
-				 return FALSE;
-		} else {
-			IXDR_PUT_BOOL(buf, objp->no_trunc);
-			IXDR_PUT_BOOL(buf, objp->chown_restricted);
-			IXDR_PUT_BOOL(buf, objp->case_insensitive);
-			IXDR_PUT_BOOL(buf, objp->case_preserving);
-		}
-		return TRUE;
-	} else if (xdrs->x_op == XDR_DECODE) {
-		 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-			 return FALSE;
-		 if (!xdr_uint32 (xdrs, &objp->linkmax))
-			 return FALSE;
-		 if (!xdr_uint32 (xdrs, &objp->name_max))
-			 return FALSE;
-		buf = XDR_INLINE (xdrs, 4 * BYTES_PER_XDR_UNIT);
-		if (buf == NULL) {
-			 if (!xdr_bool (xdrs, &objp->no_trunc))
-				 return FALSE;
-			 if (!xdr_bool (xdrs, &objp->chown_restricted))
-				 return FALSE;
-			 if (!xdr_bool (xdrs, &objp->case_insensitive))
-				 return FALSE;
-			 if (!xdr_bool (xdrs, &objp->case_preserving))
-				 return FALSE;
-		} else {
-			objp->no_trunc = IXDR_GET_BOOL(buf);
-			objp->chown_restricted = IXDR_GET_BOOL(buf);
-			objp->case_insensitive = IXDR_GET_BOOL(buf);
-			objp->case_preserving = IXDR_GET_BOOL(buf);
-		}
-	 return TRUE;
-	}
-
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->linkmax))
-		 return FALSE;
-	 if (!xdr_uint32 (xdrs, &objp->name_max))
-		 return FALSE;
-	 if (!xdr_bool (xdrs, &objp->no_trunc))
-		 return FALSE;
-	 if (!xdr_bool (xdrs, &objp->chown_restricted))
-		 return FALSE;
-	 if (!xdr_bool (xdrs, &objp->case_insensitive))
-		 return FALSE;
-	 if (!xdr_bool (xdrs, &objp->case_preserving))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_pathconf3resfail (XDR *xdrs, pathconf3resfail *objp)
-{
-	 if (!xdr_post_op_attr (xdrs, &objp->obj_attributes))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_pathconf3res (XDR *xdrs, pathconf3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_pathconf3resok (xdrs, &objp->pathconf3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_pathconf3resfail (xdrs, &objp->pathconf3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_commit3args (XDR *xdrs, commit3args *objp)
-{
-	 if (!xdr_nfs_fh3 (xdrs, &objp->file))
-		 return FALSE;
-	 if (!xdr_offset3 (xdrs, &objp->offset))
-		 return FALSE;
-	 if (!xdr_count3 (xdrs, &objp->count))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_commit3resok (XDR *xdrs, commit3resok *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->file_wcc))
-		 return FALSE;
-	 if (!xdr_writeverf3 (xdrs, objp->verf))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_commit3resfail (XDR *xdrs, commit3resfail *objp)
-{
-	 if (!xdr_wcc_data (xdrs, &objp->file_wcc))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_commit3res (XDR *xdrs, commit3res *objp)
-{
-	 if (!xdr_nfsstat3 (xdrs, &objp->status))
-		 return FALSE;
-	switch (objp->status) {
-	case NFS3_OK:
-		 if (!xdr_commit3resok (xdrs, &objp->commit3res_u.resok))
-			 return FALSE;
-		break;
-	default:
-		 if (!xdr_commit3resfail (xdrs, &objp->commit3res_u.resfail))
-			 return FALSE;
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_fhandle3 (XDR *xdrs, fhandle3 *objp)
-{
-	 if (!xdr_bytes (xdrs, (char **)&objp->fhandle3_val, (u_int *) &objp->fhandle3_len, FHSIZE3))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_dirpath (XDR *xdrs, dirpath *objp)
-{
-	 if (!xdr_string (xdrs, objp, MNTPATHLEN))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_name (XDR *xdrs, name *objp)
-{
-	 if (!xdr_string (xdrs, objp, MNTNAMLEN))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mountstat3 (XDR *xdrs, mountstat3 *objp)
-{
-	 if (!xdr_enum (xdrs, (enum_t *) objp))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mountres3_ok (XDR *xdrs, mountres3_ok *objp)
-{
-	 if (!xdr_fhandle3 (xdrs, &objp->fhandle))
-		 return FALSE;
-	 if (!xdr_array (xdrs, (char **)&objp->auth_flavors.auth_flavors_val, (u_int *) &objp->auth_flavors.auth_flavors_len, ~0,
-		sizeof (int), (xdrproc_t) xdr_int))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mountres3 (XDR *xdrs, mountres3 *objp)
-{
-	 if (!xdr_mountstat3 (xdrs, &objp->fhs_status))
-		 return FALSE;
-	switch (objp->fhs_status) {
-	case MNT3_OK:
-		 if (!xdr_mountres3_ok (xdrs, &objp->mountres3_u.mountinfo))
-			 return FALSE;
-		break;
-	default:
-		break;
-	}
-	return TRUE;
-}
-
-bool_t
-xdr_mountlist (XDR *xdrs, mountlist *objp)
-{
-	 if (!xdr_pointer (xdrs, (char **)objp, sizeof (struct mountbody), (xdrproc_t) xdr_mountbody))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_mountbody (XDR *xdrs, mountbody *objp)
-{
-	 if (!xdr_name (xdrs, &objp->ml_hostname))
-		 return FALSE;
-	 if (!xdr_dirpath (xdrs, &objp->ml_directory))
-		 return FALSE;
-	 if (!xdr_mountlist (xdrs, &objp->ml_next))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_groups (XDR *xdrs, groups *objp)
-{
-	 if (!xdr_pointer (xdrs, (char **)objp, sizeof (struct groupnode), (xdrproc_t) xdr_groupnode))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_groupnode (XDR *xdrs, groupnode *objp)
-{
-	 if (!xdr_name (xdrs, &objp->gr_name))
-		 return FALSE;
-	 if (!xdr_groups (xdrs, &objp->gr_next))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_exports (XDR *xdrs, exports *objp)
-{
-	 if (!xdr_pointer (xdrs, (char **)objp, sizeof (struct exportnode), (xdrproc_t) xdr_exportnode))
-		 return FALSE;
-	return TRUE;
-}
-
-bool_t
-xdr_exportnode (XDR *xdrs, exportnode *objp)
-{
-	 if (!xdr_dirpath (xdrs, &objp->ex_dir))
-		 return FALSE;
-	 if (!xdr_groups (xdrs, &objp->ex_groups))
-		 return FALSE;
-	 if (!xdr_exports (xdrs, &objp->ex_next))
-		 return FALSE;
-	return TRUE;
-}
-
-void
-xdr_free_exports_list (struct exportnode *first)
-{
-        struct exportnode       *elist = NULL;
-
-        if (!first)
-                return;
-
-        while (first) {
-                elist = first->ex_next;
-                if (first->ex_dir)
-                        GF_FREE (first->ex_dir);
-
-                if (first->ex_groups) {
-                        if (first->ex_groups->gr_name)
-                                GF_FREE (first->ex_groups->gr_name);
-                        GF_FREE (first->ex_groups);
-                }
-
-                GF_FREE (first);
-                first = elist;
-        }
-
-}
-
-
-void
-xdr_free_mountlist (mountlist ml)
-{
-        struct mountbody        *next = NULL;
-
-        if (!ml)
-                return;
-
-        while (ml) {
-                GF_FREE (ml->ml_hostname);
-                GF_FREE (ml->ml_directory);
-                next = ml->ml_next;
-                GF_FREE (ml);
-                ml = next;
-        }
-
-        return;
-}
-
-
-/* Free statements are based on the way sunrpc xdr decoding
- * code performs memory allocations.
- */
-void
-xdr_free_write3args_nocopy (write3args *wa)
-{
-        if (!wa)
-                return;
-
-        FREE (wa->file.data.data_val);
-}
-
-
diff --git a/xlators/nfs/lib/src/xdr-nfs3.h b/xlators/nfs/lib/src/xdr-nfs3.h
deleted file mode 100644
index 5af65e6b0..000000000
--- a/xlators/nfs/lib/src/xdr-nfs3.h
+++ /dev/null
@@ -1,1206 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _XDR_NFS3_H
-#define _XDR_NFS3_H
-
-#include <rpc/rpc.h>
-#include <sys/types.h>
-
-#define NFS3_FHSIZE             64
-#define NFS3_COOKIEVERFSIZE     8
-#define NFS3_CREATEVERFSIZE     8
-#define NFS3_WRITEVERFSIZE      8
-
-#define NFS3_ENTRY3_FIXED_SIZE  24
-#define NFS3_POSTOPATTR_SIZE    88
-#define NFS3_READDIR_RESOK_SIZE (NFS3_POSTOPATTR_SIZE + sizeof (bool_t) + NFS3_COOKIEVERFSIZE)
-
-/* In size of post_op_fh3, the length of the file handle will have to be
- * included separately since we have variable length fh. Here we only account
- * for the field for handle_follows and for the file handle length field.
- */
-#define NFS3_POSTOPFH3_FIXED_SIZE     (sizeof (bool_t) + sizeof (uint32_t))
-
-/* Similarly, the size of the entry will have to include the variable length
- * file handle and the length of the entry name.
- */
-#define NFS3_ENTRYP3_FIXED_SIZE  (NFS3_ENTRY3_FIXED_SIZE + NFS3_POSTOPATTR_SIZE + NFS3_POSTOPFH3_FIXED_SIZE)
-
-typedef uint64_t uint64;
-typedef int64_t int64;
-typedef uint32_t uint32;
-typedef int32_t int32;
-typedef char *filename3;
-typedef char *nfspath3;
-typedef uint64 fileid3;
-typedef uint64 cookie3;
-typedef char cookieverf3[NFS3_COOKIEVERFSIZE];
-typedef char createverf3[NFS3_CREATEVERFSIZE];
-typedef char writeverf3[NFS3_WRITEVERFSIZE];
-typedef uint32 uid3;
-typedef uint32 gid3;
-typedef uint64 size3;
-typedef uint64 offset3;
-typedef uint32 mode3;
-typedef uint32 count3;
-
-#define NFS3MODE_SETXUID        0x00800
-#define NFS3MODE_SETXGID        0x00400
-#define NFS3MODE_SAVESWAPTXT    0x00200
-#define NFS3MODE_ROWNER         0x00100
-#define NFS3MODE_WOWNER         0x00080
-#define NFS3MODE_XOWNER         0x00040
-#define NFS3MODE_RGROUP         0x00020
-#define NFS3MODE_WGROUP         0x00010
-#define NFS3MODE_XGROUP         0x00008
-#define NFS3MODE_ROTHER         0x00004
-#define NFS3MODE_WOTHER         0x00002
-#define NFS3MODE_XOTHER         0x00001
-
-enum nfsstat3 {
-	NFS3_OK = 0,
-	NFS3ERR_PERM = 1,
-	NFS3ERR_NOENT = 2,
-	NFS3ERR_IO = 5,
-	NFS3ERR_NXIO = 6,
-	NFS3ERR_ACCES = 13,
-	NFS3ERR_EXIST = 17,
-	NFS3ERR_XDEV = 18,
-	NFS3ERR_NODEV = 19,
-	NFS3ERR_NOTDIR = 20,
-	NFS3ERR_ISDIR = 21,
-	NFS3ERR_INVAL = 22,
-	NFS3ERR_FBIG = 27,
-	NFS3ERR_NOSPC = 28,
-	NFS3ERR_ROFS = 30,
-	NFS3ERR_MLINK = 31,
-	NFS3ERR_NAMETOOLONG = 63,
-	NFS3ERR_NOTEMPTY = 66,
-	NFS3ERR_DQUOT = 69,
-	NFS3ERR_STALE = 70,
-	NFS3ERR_REMOTE = 71,
-	NFS3ERR_BADHANDLE = 10001,
-	NFS3ERR_NOT_SYNC = 10002,
-	NFS3ERR_BAD_COOKIE = 10003,
-	NFS3ERR_NOTSUPP = 10004,
-	NFS3ERR_TOOSMALL = 10005,
-	NFS3ERR_SERVERFAULT = 10006,
-	NFS3ERR_BADTYPE = 10007,
-	NFS3ERR_JUKEBOX = 10008,
-};
-typedef enum nfsstat3 nfsstat3;
-
-enum ftype3 {
-	NF3REG = 1,
-	NF3DIR = 2,
-	NF3BLK = 3,
-	NF3CHR = 4,
-	NF3LNK = 5,
-	NF3SOCK = 6,
-	NF3FIFO = 7,
-};
-typedef enum ftype3 ftype3;
-
-struct specdata3 {
-	uint32 specdata1;
-	uint32 specdata2;
-};
-typedef struct specdata3 specdata3;
-
-struct nfs_fh3 {
-	struct {
-		u_int data_len;
-		char *data_val;
-	} data;
-};
-typedef struct nfs_fh3 nfs_fh3;
-
-struct nfstime3 {
-	uint32 seconds;
-	uint32 nseconds;
-};
-typedef struct nfstime3 nfstime3;
-
-struct fattr3 {
-	ftype3 type;
-	mode3 mode;
-	uint32 nlink;
-	uid3 uid;
-	gid3 gid;
-	size3 size;
-	size3 used;
-	specdata3 rdev;
-	uint64 fsid;
-	fileid3 fileid;
-	nfstime3 atime;
-	nfstime3 mtime;
-	nfstime3 ctime;
-};
-typedef struct fattr3 fattr3;
-
-struct post_op_attr {
-	bool_t attributes_follow;
-	union {
-		fattr3 attributes;
-	} post_op_attr_u;
-};
-typedef struct post_op_attr post_op_attr;
-
-struct wcc_attr {
-	size3 size;
-	nfstime3 mtime;
-	nfstime3 ctime;
-};
-typedef struct wcc_attr wcc_attr;
-
-struct pre_op_attr {
-	bool_t attributes_follow;
-	union {
-		wcc_attr attributes;
-	} pre_op_attr_u;
-};
-typedef struct pre_op_attr pre_op_attr;
-
-struct wcc_data {
-	pre_op_attr before;
-	post_op_attr after;
-};
-typedef struct wcc_data wcc_data;
-
-struct post_op_fh3 {
-	bool_t handle_follows;
-	union {
-		nfs_fh3 handle;
-	} post_op_fh3_u;
-};
-typedef struct post_op_fh3 post_op_fh3;
-
-enum time_how {
-	DONT_CHANGE = 0,
-	SET_TO_SERVER_TIME = 1,
-	SET_TO_CLIENT_TIME = 2,
-};
-typedef enum time_how time_how;
-
-struct set_mode3 {
-	bool_t set_it;
-	union {
-		mode3 mode;
-	} set_mode3_u;
-};
-typedef struct set_mode3 set_mode3;
-
-struct set_uid3 {
-	bool_t set_it;
-	union {
-		uid3 uid;
-	} set_uid3_u;
-};
-typedef struct set_uid3 set_uid3;
-
-struct set_gid3 {
-	bool_t set_it;
-	union {
-		gid3 gid;
-	} set_gid3_u;
-};
-typedef struct set_gid3 set_gid3;
-
-struct set_size3 {
-	bool_t set_it;
-	union {
-		size3 size;
-	} set_size3_u;
-};
-typedef struct set_size3 set_size3;
-
-struct set_atime {
-	time_how set_it;
-	union {
-		nfstime3 atime;
-	} set_atime_u;
-};
-typedef struct set_atime set_atime;
-
-struct set_mtime {
-	time_how set_it;
-	union {
-		nfstime3 mtime;
-	} set_mtime_u;
-};
-typedef struct set_mtime set_mtime;
-
-struct sattr3 {
-	set_mode3 mode;
-	set_uid3 uid;
-	set_gid3 gid;
-	set_size3 size;
-	set_atime atime;
-	set_mtime mtime;
-};
-typedef struct sattr3 sattr3;
-
-struct diropargs3 {
-	nfs_fh3 dir;
-	filename3 name;
-};
-typedef struct diropargs3 diropargs3;
-
-struct getattr3args {
-	nfs_fh3 object;
-};
-typedef struct getattr3args getattr3args;
-
-struct getattr3resok {
-	fattr3 obj_attributes;
-};
-typedef struct getattr3resok getattr3resok;
-
-struct getattr3res {
-	nfsstat3 status;
-	union {
-		getattr3resok resok;
-	} getattr3res_u;
-};
-typedef struct getattr3res getattr3res;
-
-struct sattrguard3 {
-	bool_t check;
-	union {
-		nfstime3 obj_ctime;
-	} sattrguard3_u;
-};
-typedef struct sattrguard3 sattrguard3;
-
-struct setattr3args {
-	nfs_fh3 object;
-	sattr3 new_attributes;
-	sattrguard3 guard;
-};
-typedef struct setattr3args setattr3args;
-
-struct setattr3resok {
-	wcc_data obj_wcc;
-};
-typedef struct setattr3resok setattr3resok;
-
-struct setattr3resfail {
-	wcc_data obj_wcc;
-};
-typedef struct setattr3resfail setattr3resfail;
-
-struct setattr3res {
-	nfsstat3 status;
-	union {
-		setattr3resok resok;
-		setattr3resfail resfail;
-	} setattr3res_u;
-};
-typedef struct setattr3res setattr3res;
-
-struct lookup3args {
-	diropargs3 what;
-};
-typedef struct lookup3args lookup3args;
-
-struct lookup3resok {
-	nfs_fh3 object;
-	post_op_attr obj_attributes;
-	post_op_attr dir_attributes;
-};
-typedef struct lookup3resok lookup3resok;
-
-struct lookup3resfail {
-	post_op_attr dir_attributes;
-};
-typedef struct lookup3resfail lookup3resfail;
-
-struct lookup3res {
-	nfsstat3 status;
-	union {
-		lookup3resok resok;
-		lookup3resfail resfail;
-	} lookup3res_u;
-};
-typedef struct lookup3res lookup3res;
-#define ACCESS3_READ 0x0001
-#define ACCESS3_LOOKUP 0x0002
-#define ACCESS3_MODIFY 0x0004
-#define ACCESS3_EXTEND 0x0008
-#define ACCESS3_DELETE 0x0010
-#define ACCESS3_EXECUTE 0x0020
-
-struct access3args {
-	nfs_fh3 object;
-	uint32 access;
-};
-typedef struct access3args access3args;
-
-struct access3resok {
-	post_op_attr obj_attributes;
-	uint32 access;
-};
-typedef struct access3resok access3resok;
-
-struct access3resfail {
-	post_op_attr obj_attributes;
-};
-typedef struct access3resfail access3resfail;
-
-struct access3res {
-	nfsstat3 status;
-	union {
-		access3resok resok;
-		access3resfail resfail;
-	} access3res_u;
-};
-typedef struct access3res access3res;
-
-struct readlink3args {
-	nfs_fh3 symlink;
-};
-typedef struct readlink3args readlink3args;
-
-struct readlink3resok {
-	post_op_attr symlink_attributes;
-	nfspath3 data;
-};
-typedef struct readlink3resok readlink3resok;
-
-struct readlink3resfail {
-	post_op_attr symlink_attributes;
-};
-typedef struct readlink3resfail readlink3resfail;
-
-struct readlink3res {
-	nfsstat3 status;
-	union {
-		readlink3resok resok;
-		readlink3resfail resfail;
-	} readlink3res_u;
-};
-typedef struct readlink3res readlink3res;
-
-struct read3args {
-	nfs_fh3 file;
-	offset3 offset;
-	count3 count;
-};
-typedef struct read3args read3args;
-
-struct read3resok {
-	post_op_attr file_attributes;
-	count3 count;
-	bool_t eof;
-	struct {
-		u_int data_len;
-		char *data_val;
-	} data;
-};
-typedef struct read3resok read3resok;
-
-struct read3resfail {
-	post_op_attr file_attributes;
-};
-typedef struct read3resfail read3resfail;
-
-struct read3res {
-	nfsstat3 status;
-	union {
-		read3resok resok;
-		read3resfail resfail;
-	} read3res_u;
-};
-typedef struct read3res read3res;
-
-enum stable_how {
-	UNSTABLE = 0,
-	DATA_SYNC = 1,
-	FILE_SYNC = 2,
-};
-typedef enum stable_how stable_how;
-
-struct write3args {
-	nfs_fh3 file;
-	offset3 offset;
-	count3 count;
-	stable_how stable;
-	struct {
-		u_int data_len;
-		char *data_val;
-	} data;
-};
-typedef struct write3args write3args;
-
-/* Generally, the protocol allows the file handle to be less than 64 bytes but
- * our server does not return file handles less than 64b so we can safely say
- * sizeof (nfs_fh3) rather than first trying to extract the fh size of the
- * network followed by a sized-read of the file handle.
- */
-#define NFS3_WRITE3ARGS_SIZE    (sizeof (uint32_t) + NFS3_FHSIZE + sizeof (offset3) + sizeof (count3) + sizeof (uint32_t))
-struct write3resok {
-	wcc_data file_wcc;
-	count3 count;
-	stable_how committed;
-	writeverf3 verf;
-};
-typedef struct write3resok write3resok;
-
-struct write3resfail {
-	wcc_data file_wcc;
-};
-typedef struct write3resfail write3resfail;
-
-struct write3res {
-	nfsstat3 status;
-	union {
-		write3resok resok;
-		write3resfail resfail;
-	} write3res_u;
-};
-typedef struct write3res write3res;
-
-enum createmode3 {
-	UNCHECKED = 0,
-	GUARDED = 1,
-	EXCLUSIVE = 2,
-};
-typedef enum createmode3 createmode3;
-
-struct createhow3 {
-	createmode3 mode;
-	union {
-		sattr3 obj_attributes;
-		createverf3 verf;
-	} createhow3_u;
-};
-typedef struct createhow3 createhow3;
-
-struct create3args {
-	diropargs3 where;
-	createhow3 how;
-};
-typedef struct create3args create3args;
-
-struct create3resok {
-	post_op_fh3 obj;
-	post_op_attr obj_attributes;
-	wcc_data dir_wcc;
-};
-typedef struct create3resok create3resok;
-
-struct create3resfail {
-	wcc_data dir_wcc;
-};
-typedef struct create3resfail create3resfail;
-
-struct create3res {
-	nfsstat3 status;
-	union {
-		create3resok resok;
-		create3resfail resfail;
-	} create3res_u;
-};
-typedef struct create3res create3res;
-
-struct mkdir3args {
-	diropargs3 where;
-	sattr3 attributes;
-};
-typedef struct mkdir3args mkdir3args;
-
-struct mkdir3resok {
-	post_op_fh3 obj;
-	post_op_attr obj_attributes;
-	wcc_data dir_wcc;
-};
-typedef struct mkdir3resok mkdir3resok;
-
-struct mkdir3resfail {
-	wcc_data dir_wcc;
-};
-typedef struct mkdir3resfail mkdir3resfail;
-
-struct mkdir3res {
-	nfsstat3 status;
-	union {
-		mkdir3resok resok;
-		mkdir3resfail resfail;
-	} mkdir3res_u;
-};
-typedef struct mkdir3res mkdir3res;
-
-struct symlinkdata3 {
-	sattr3 symlink_attributes;
-	nfspath3 symlink_data;
-};
-typedef struct symlinkdata3 symlinkdata3;
-
-struct symlink3args {
-	diropargs3 where;
-	symlinkdata3 symlink;
-};
-typedef struct symlink3args symlink3args;
-
-struct symlink3resok {
-	post_op_fh3 obj;
-	post_op_attr obj_attributes;
-	wcc_data dir_wcc;
-};
-typedef struct symlink3resok symlink3resok;
-
-struct symlink3resfail {
-	wcc_data dir_wcc;
-};
-typedef struct symlink3resfail symlink3resfail;
-
-struct symlink3res {
-	nfsstat3 status;
-	union {
-		symlink3resok resok;
-		symlink3resfail resfail;
-	} symlink3res_u;
-};
-typedef struct symlink3res symlink3res;
-
-struct devicedata3 {
-	sattr3 dev_attributes;
-	specdata3 spec;
-};
-typedef struct devicedata3 devicedata3;
-
-struct mknoddata3 {
-	ftype3 type;
-	union {
-		devicedata3 device;
-		sattr3 pipe_attributes;
-	} mknoddata3_u;
-};
-typedef struct mknoddata3 mknoddata3;
-
-struct mknod3args {
-	diropargs3 where;
-	mknoddata3 what;
-};
-typedef struct mknod3args mknod3args;
-
-struct mknod3resok {
-	post_op_fh3 obj;
-	post_op_attr obj_attributes;
-	wcc_data dir_wcc;
-};
-typedef struct mknod3resok mknod3resok;
-
-struct mknod3resfail {
-	wcc_data dir_wcc;
-};
-typedef struct mknod3resfail mknod3resfail;
-
-struct mknod3res {
-	nfsstat3 status;
-	union {
-		mknod3resok resok;
-		mknod3resfail resfail;
-	} mknod3res_u;
-};
-typedef struct mknod3res mknod3res;
-
-struct remove3args {
-	diropargs3 object;
-};
-typedef struct remove3args remove3args;
-
-struct remove3resok {
-	wcc_data dir_wcc;
-};
-typedef struct remove3resok remove3resok;
-
-struct remove3resfail {
-	wcc_data dir_wcc;
-};
-typedef struct remove3resfail remove3resfail;
-
-struct remove3res {
-	nfsstat3 status;
-	union {
-		remove3resok resok;
-		remove3resfail resfail;
-	} remove3res_u;
-};
-typedef struct remove3res remove3res;
-
-struct rmdir3args {
-	diropargs3 object;
-};
-typedef struct rmdir3args rmdir3args;
-
-struct rmdir3resok {
-	wcc_data dir_wcc;
-};
-typedef struct rmdir3resok rmdir3resok;
-
-struct rmdir3resfail {
-	wcc_data dir_wcc;
-};
-typedef struct rmdir3resfail rmdir3resfail;
-
-struct rmdir3res {
-	nfsstat3 status;
-	union {
-		rmdir3resok resok;
-		rmdir3resfail resfail;
-	} rmdir3res_u;
-};
-typedef struct rmdir3res rmdir3res;
-
-struct rename3args {
-	diropargs3 from;
-	diropargs3 to;
-};
-typedef struct rename3args rename3args;
-
-struct rename3resok {
-	wcc_data fromdir_wcc;
-	wcc_data todir_wcc;
-};
-typedef struct rename3resok rename3resok;
-
-struct rename3resfail {
-	wcc_data fromdir_wcc;
-	wcc_data todir_wcc;
-};
-typedef struct rename3resfail rename3resfail;
-
-struct rename3res {
-	nfsstat3 status;
-	union {
-		rename3resok resok;
-		rename3resfail resfail;
-	} rename3res_u;
-};
-typedef struct rename3res rename3res;
-
-struct link3args {
-	nfs_fh3 file;
-	diropargs3 link;
-};
-typedef struct link3args link3args;
-
-struct link3resok {
-	post_op_attr file_attributes;
-	wcc_data linkdir_wcc;
-};
-typedef struct link3resok link3resok;
-
-struct link3resfail {
-	post_op_attr file_attributes;
-	wcc_data linkdir_wcc;
-};
-typedef struct link3resfail link3resfail;
-
-struct link3res {
-	nfsstat3 status;
-	union {
-		link3resok resok;
-		link3resfail resfail;
-	} link3res_u;
-};
-typedef struct link3res link3res;
-
-struct readdir3args {
-	nfs_fh3 dir;
-	cookie3 cookie;
-	cookieverf3 cookieverf;
-	count3 count;
-};
-typedef struct readdir3args readdir3args;
-
-struct entry3 {
-	fileid3 fileid;
-	filename3 name;
-	cookie3 cookie;
-	struct entry3 *nextentry;
-};
-typedef struct entry3 entry3;
-
-struct dirlist3 {
-	entry3 *entries;
-	bool_t eof;
-};
-typedef struct dirlist3 dirlist3;
-
-struct readdir3resok {
-	post_op_attr dir_attributes;
-	cookieverf3 cookieverf;
-	dirlist3 reply;
-};
-typedef struct readdir3resok readdir3resok;
-
-struct readdir3resfail {
-	post_op_attr dir_attributes;
-};
-typedef struct readdir3resfail readdir3resfail;
-
-struct readdir3res {
-	nfsstat3 status;
-	union {
-		readdir3resok resok;
-		readdir3resfail resfail;
-	} readdir3res_u;
-};
-typedef struct readdir3res readdir3res;
-
-struct readdirp3args {
-	nfs_fh3 dir;
-	cookie3 cookie;
-	cookieverf3 cookieverf;
-	count3 dircount;
-	count3 maxcount;
-};
-typedef struct readdirp3args readdirp3args;
-
-struct entryp3 {
-	fileid3 fileid;
-	filename3 name;
-	cookie3 cookie;
-	post_op_attr name_attributes;
-	post_op_fh3 name_handle;
-	struct entryp3 *nextentry;
-};
-typedef struct entryp3 entryp3;
-
-struct dirlistp3 {
-	entryp3 *entries;
-	bool_t eof;
-};
-typedef struct dirlistp3 dirlistp3;
-
-struct readdirp3resok {
-	post_op_attr dir_attributes;
-	cookieverf3 cookieverf;
-	dirlistp3   reply;
-};
-typedef struct readdirp3resok readdirp3resok;
-
-struct readdirp3resfail {
-	post_op_attr dir_attributes;
-};
-typedef struct readdirp3resfail readdirp3resfail;
-
-struct readdirp3res {
-	nfsstat3 status;
-	union {
-		readdirp3resok resok;
-		readdirp3resfail resfail;
-	} readdirp3res_u;
-};
-typedef struct readdirp3res readdirp3res;
-
-struct fsstat3args {
-	nfs_fh3 fsroot;
-};
-typedef struct fsstat3args fsstat3args;
-
-struct fsstat3resok {
-	post_op_attr obj_attributes;
-	size3 tbytes;
-	size3 fbytes;
-	size3 abytes;
-	size3 tfiles;
-	size3 ffiles;
-	size3 afiles;
-	uint32 invarsec;
-};
-typedef struct fsstat3resok fsstat3resok;
-
-struct fsstat3resfail {
-	post_op_attr obj_attributes;
-};
-typedef struct fsstat3resfail fsstat3resfail;
-
-struct fsstat3res {
-	nfsstat3 status;
-	union {
-		fsstat3resok resok;
-		fsstat3resfail resfail;
-	} fsstat3res_u;
-};
-typedef struct fsstat3res fsstat3res;
-#define FSF3_LINK 0x0001
-#define FSF3_SYMLINK 0x0002
-#define FSF3_HOMOGENEOUS 0x0008
-#define FSF3_CANSETTIME 0x0010
-
-struct fsinfo3args {
-	nfs_fh3 fsroot;
-};
-typedef struct fsinfo3args fsinfo3args;
-
-struct fsinfo3resok {
-	post_op_attr obj_attributes;
-	uint32 rtmax;
-	uint32 rtpref;
-	uint32 rtmult;
-	uint32 wtmax;
-	uint32 wtpref;
-	uint32 wtmult;
-	uint32 dtpref;
-	size3 maxfilesize;
-	nfstime3 time_delta;
-	uint32 properties;
-};
-typedef struct fsinfo3resok fsinfo3resok;
-
-struct fsinfo3resfail {
-	post_op_attr obj_attributes;
-};
-typedef struct fsinfo3resfail fsinfo3resfail;
-
-struct fsinfo3res {
-	nfsstat3 status;
-	union {
-		fsinfo3resok resok;
-		fsinfo3resfail resfail;
-	} fsinfo3res_u;
-};
-typedef struct fsinfo3res fsinfo3res;
-
-struct pathconf3args {
-	nfs_fh3 object;
-};
-typedef struct pathconf3args pathconf3args;
-
-struct pathconf3resok {
-	post_op_attr obj_attributes;
-	uint32 linkmax;
-	uint32 name_max;
-	bool_t no_trunc;
-	bool_t chown_restricted;
-	bool_t case_insensitive;
-	bool_t case_preserving;
-};
-typedef struct pathconf3resok pathconf3resok;
-
-struct pathconf3resfail {
-	post_op_attr obj_attributes;
-};
-typedef struct pathconf3resfail pathconf3resfail;
-
-struct pathconf3res {
-	nfsstat3 status;
-	union {
-		pathconf3resok resok;
-		pathconf3resfail resfail;
-	} pathconf3res_u;
-};
-typedef struct pathconf3res pathconf3res;
-
-struct commit3args {
-	nfs_fh3 file;
-	offset3 offset;
-	count3 count;
-};
-typedef struct commit3args commit3args;
-
-struct commit3resok {
-	wcc_data file_wcc;
-	writeverf3 verf;
-};
-typedef struct commit3resok commit3resok;
-
-struct commit3resfail {
-	wcc_data file_wcc;
-};
-typedef struct commit3resfail commit3resfail;
-
-struct commit3res {
-	nfsstat3 status;
-	union {
-		commit3resok resok;
-		commit3resfail resfail;
-	} commit3res_u;
-};
-typedef struct commit3res commit3res;
-#define MNTPATHLEN 1024
-#define MNTNAMLEN 255
-#define FHSIZE3 NFS3_FHSIZE
-
-typedef struct {
-	u_int fhandle3_len;
-	char *fhandle3_val;
-} fhandle3;
-
-typedef char *dirpath;
-
-typedef char *name;
-
-enum mountstat3 {
-	MNT3_OK = 0,
-	MNT3ERR_PERM = 1,
-	MNT3ERR_NOENT = 2,
-	MNT3ERR_IO = 5,
-	MNT3ERR_ACCES = 13,
-	MNT3ERR_NOTDIR = 20,
-	MNT3ERR_INVAL = 22,
-	MNT3ERR_NAMETOOLONG = 63,
-	MNT3ERR_NOTSUPP = 10004,
-	MNT3ERR_SERVERFAULT = 10006,
-};
-typedef enum mountstat3 mountstat3;
-
-struct mountres3_ok {
-	fhandle3 fhandle;
-	struct {
-		u_int auth_flavors_len;
-		int *auth_flavors_val;
-	} auth_flavors;
-};
-typedef struct mountres3_ok mountres3_ok;
-
-struct mountres3 {
-	mountstat3 fhs_status;
-	union {
-		mountres3_ok mountinfo;
-	} mountres3_u;
-};
-typedef struct mountres3 mountres3;
-
-typedef struct mountbody *mountlist;
-
-struct mountbody {
-	name ml_hostname;
-	dirpath ml_directory;
-	mountlist ml_next;
-};
-typedef struct mountbody mountbody;
-
-typedef struct groupnode *groups;
-
-struct groupnode {
-	name gr_name;
-	groups gr_next;
-};
-typedef struct groupnode groupnode;
-
-typedef struct exportnode *exports;
-
-struct exportnode {
-	dirpath ex_dir;
-	groups ex_groups;
-	exports ex_next;
-};
-typedef struct exportnode exportnode;
-
-#define NFS_PROGRAM             100003
-#define NFS_V3                  3
-
-#define NFS3_NULL               0
-#define NFS3_GETATTR            1
-#define NFS3_SETATTR            2
-#define NFS3_LOOKUP             3
-#define NFS3_ACCESS             4
-#define NFS3_READLINK           5
-#define NFS3_READ               6
-#define NFS3_WRITE              7
-#define NFS3_CREATE             8
-#define NFS3_MKDIR              9
-#define NFS3_SYMLINK            10
-#define NFS3_MKNOD              11
-#define NFS3_REMOVE             12
-#define NFS3_RMDIR              13
-#define NFS3_RENAME             14
-#define NFS3_LINK               15
-#define NFS3_READDIR            16
-#define NFS3_READDIRP           17
-#define NFS3_FSSTAT             18
-#define NFS3_FSINFO             19
-#define NFS3_PATHCONF           20
-#define NFS3_COMMIT             21
-#define NFS3_PROC_COUNT         22
-
-#define MOUNT_PROGRAM           100005
-#define MOUNT_V3                3
-#define MOUNT_V1                1
-
-#define MOUNT3_NULL             0
-#define MOUNT3_MNT              1
-#define MOUNT3_DUMP             2
-#define MOUNT3_UMNT             3
-#define MOUNT3_UMNTALL          4
-#define MOUNT3_EXPORT           5
-#define MOUNT3_PROC_COUNT       6
-
-#define MOUNT1_NULL             0
-#define MOUNT1_DUMP             2
-#define MOUNT1_UMNT             3
-#define MOUNT1_EXPORT           5
-#define MOUNT1_PROC_COUNT       6
-/* the xdr functions */
-
-extern  bool_t xdr_uint64 (XDR *, uint64*);
-extern  bool_t xdr_int64 (XDR *, int64*);
-extern  bool_t xdr_uint32 (XDR *, uint32*);
-extern  bool_t xdr_int32 (XDR *, int32*);
-extern  bool_t xdr_filename3 (XDR *, filename3*);
-extern  bool_t xdr_nfspath3 (XDR *, nfspath3*);
-extern  bool_t xdr_fileid3 (XDR *, fileid3*);
-extern  bool_t xdr_cookie3 (XDR *, cookie3*);
-extern  bool_t xdr_cookieverf3 (XDR *, cookieverf3);
-extern  bool_t xdr_createverf3 (XDR *, createverf3);
-extern  bool_t xdr_writeverf3 (XDR *, writeverf3);
-extern  bool_t xdr_uid3 (XDR *, uid3*);
-extern  bool_t xdr_gid3 (XDR *, gid3*);
-extern  bool_t xdr_size3 (XDR *, size3*);
-extern  bool_t xdr_offset3 (XDR *, offset3*);
-extern  bool_t xdr_mode3 (XDR *, mode3*);
-extern  bool_t xdr_count3 (XDR *, count3*);
-extern  bool_t xdr_nfsstat3 (XDR *, nfsstat3*);
-extern  bool_t xdr_ftype3 (XDR *, ftype3*);
-extern  bool_t xdr_specdata3 (XDR *, specdata3*);
-extern  bool_t xdr_nfs_fh3 (XDR *, nfs_fh3*);
-extern  bool_t xdr_nfstime3 (XDR *, nfstime3*);
-extern  bool_t xdr_fattr3 (XDR *, fattr3*);
-extern  bool_t xdr_post_op_attr (XDR *, post_op_attr*);
-extern  bool_t xdr_wcc_attr (XDR *, wcc_attr*);
-extern  bool_t xdr_pre_op_attr (XDR *, pre_op_attr*);
-extern  bool_t xdr_wcc_data (XDR *, wcc_data*);
-extern  bool_t xdr_post_op_fh3 (XDR *, post_op_fh3*);
-extern  bool_t xdr_time_how (XDR *, time_how*);
-extern  bool_t xdr_set_mode3 (XDR *, set_mode3*);
-extern  bool_t xdr_set_uid3 (XDR *, set_uid3*);
-extern  bool_t xdr_set_gid3 (XDR *, set_gid3*);
-extern  bool_t xdr_set_size3 (XDR *, set_size3*);
-extern  bool_t xdr_set_atime (XDR *, set_atime*);
-extern  bool_t xdr_set_mtime (XDR *, set_mtime*);
-extern  bool_t xdr_sattr3 (XDR *, sattr3*);
-extern  bool_t xdr_diropargs3 (XDR *, diropargs3*);
-extern  bool_t xdr_getattr3args (XDR *, getattr3args*);
-extern  bool_t xdr_getattr3resok (XDR *, getattr3resok*);
-extern  bool_t xdr_getattr3res (XDR *, getattr3res*);
-extern  bool_t xdr_sattrguard3 (XDR *, sattrguard3*);
-extern  bool_t xdr_setattr3args (XDR *, setattr3args*);
-extern  bool_t xdr_setattr3resok (XDR *, setattr3resok*);
-extern  bool_t xdr_setattr3resfail (XDR *, setattr3resfail*);
-extern  bool_t xdr_setattr3res (XDR *, setattr3res*);
-extern  bool_t xdr_lookup3args (XDR *, lookup3args*);
-extern  bool_t xdr_lookup3resok (XDR *, lookup3resok*);
-extern  bool_t xdr_lookup3resfail (XDR *, lookup3resfail*);
-extern  bool_t xdr_lookup3res (XDR *, lookup3res*);
-extern  bool_t xdr_access3args (XDR *, access3args*);
-extern  bool_t xdr_access3resok (XDR *, access3resok*);
-extern  bool_t xdr_access3resfail (XDR *, access3resfail*);
-extern  bool_t xdr_access3res (XDR *, access3res*);
-extern  bool_t xdr_readlink3args (XDR *, readlink3args*);
-extern  bool_t xdr_readlink3resok (XDR *, readlink3resok*);
-extern  bool_t xdr_readlink3resfail (XDR *, readlink3resfail*);
-extern  bool_t xdr_readlink3res (XDR *, readlink3res*);
-extern  bool_t xdr_read3args (XDR *, read3args*);
-extern  bool_t xdr_read3resok (XDR *, read3resok*);
-extern  bool_t xdr_read3resfail (XDR *, read3resfail*);
-extern  bool_t xdr_read3res (XDR *, read3res*);
-extern  bool_t xdr_read3res_nocopy (XDR *xdrs, read3res *objp);
-extern  bool_t xdr_stable_how (XDR *, stable_how*);
-extern  bool_t xdr_write3args (XDR *, write3args*);
-extern  bool_t xdr_write3resok (XDR *, write3resok*);
-extern  bool_t xdr_write3resfail (XDR *, write3resfail*);
-extern  bool_t xdr_write3res (XDR *, write3res*);
-extern  bool_t xdr_createmode3 (XDR *, createmode3*);
-extern  bool_t xdr_createhow3 (XDR *, createhow3*);
-extern  bool_t xdr_create3args (XDR *, create3args*);
-extern  bool_t xdr_create3resok (XDR *, create3resok*);
-extern  bool_t xdr_create3resfail (XDR *, create3resfail*);
-extern  bool_t xdr_create3res (XDR *, create3res*);
-extern  bool_t xdr_mkdir3args (XDR *, mkdir3args*);
-extern  bool_t xdr_mkdir3resok (XDR *, mkdir3resok*);
-extern  bool_t xdr_mkdir3resfail (XDR *, mkdir3resfail*);
-extern  bool_t xdr_mkdir3res (XDR *, mkdir3res*);
-extern  bool_t xdr_symlinkdata3 (XDR *, symlinkdata3*);
-extern  bool_t xdr_symlink3args (XDR *, symlink3args*);
-extern  bool_t xdr_symlink3resok (XDR *, symlink3resok*);
-extern  bool_t xdr_symlink3resfail (XDR *, symlink3resfail*);
-extern  bool_t xdr_symlink3res (XDR *, symlink3res*);
-extern  bool_t xdr_devicedata3 (XDR *, devicedata3*);
-extern  bool_t xdr_mknoddata3 (XDR *, mknoddata3*);
-extern  bool_t xdr_mknod3args (XDR *, mknod3args*);
-extern  bool_t xdr_mknod3resok (XDR *, mknod3resok*);
-extern  bool_t xdr_mknod3resfail (XDR *, mknod3resfail*);
-extern  bool_t xdr_mknod3res (XDR *, mknod3res*);
-extern  bool_t xdr_remove3args (XDR *, remove3args*);
-extern  bool_t xdr_remove3resok (XDR *, remove3resok*);
-extern  bool_t xdr_remove3resfail (XDR *, remove3resfail*);
-extern  bool_t xdr_remove3res (XDR *, remove3res*);
-extern  bool_t xdr_rmdir3args (XDR *, rmdir3args*);
-extern  bool_t xdr_rmdir3resok (XDR *, rmdir3resok*);
-extern  bool_t xdr_rmdir3resfail (XDR *, rmdir3resfail*);
-extern  bool_t xdr_rmdir3res (XDR *, rmdir3res*);
-extern  bool_t xdr_rename3args (XDR *, rename3args*);
-extern  bool_t xdr_rename3resok (XDR *, rename3resok*);
-extern  bool_t xdr_rename3resfail (XDR *, rename3resfail*);
-extern  bool_t xdr_rename3res (XDR *, rename3res*);
-extern  bool_t xdr_link3args (XDR *, link3args*);
-extern  bool_t xdr_link3resok (XDR *, link3resok*);
-extern  bool_t xdr_link3resfail (XDR *, link3resfail*);
-extern  bool_t xdr_link3res (XDR *, link3res*);
-extern  bool_t xdr_readdir3args (XDR *, readdir3args*);
-extern  bool_t xdr_entry3 (XDR *, entry3*);
-extern  bool_t xdr_dirlist3 (XDR *, dirlist3*);
-extern  bool_t xdr_readdir3resok (XDR *, readdir3resok*);
-extern  bool_t xdr_readdir3resfail (XDR *, readdir3resfail*);
-extern  bool_t xdr_readdir3res (XDR *, readdir3res*);
-extern  bool_t xdr_readdirp3args (XDR *, readdirp3args*);
-extern  bool_t xdr_entryp3 (XDR *, entryp3*);
-extern  bool_t xdr_dirlistp3 (XDR *, dirlistp3*);
-extern  bool_t xdr_readdirp3resok (XDR *, readdirp3resok*);
-extern  bool_t xdr_readdirp3resfail (XDR *, readdirp3resfail*);
-extern  bool_t xdr_readdirp3res (XDR *, readdirp3res*);
-extern  bool_t xdr_fsstat3args (XDR *, fsstat3args*);
-extern  bool_t xdr_fsstat3resok (XDR *, fsstat3resok*);
-extern  bool_t xdr_fsstat3resfail (XDR *, fsstat3resfail*);
-extern  bool_t xdr_fsstat3res (XDR *, fsstat3res*);
-extern  bool_t xdr_fsinfo3args (XDR *, fsinfo3args*);
-extern  bool_t xdr_fsinfo3resok (XDR *, fsinfo3resok*);
-extern  bool_t xdr_fsinfo3resfail (XDR *, fsinfo3resfail*);
-extern  bool_t xdr_fsinfo3res (XDR *, fsinfo3res*);
-extern  bool_t xdr_pathconf3args (XDR *, pathconf3args*);
-extern  bool_t xdr_pathconf3resok (XDR *, pathconf3resok*);
-extern  bool_t xdr_pathconf3resfail (XDR *, pathconf3resfail*);
-extern  bool_t xdr_pathconf3res (XDR *, pathconf3res*);
-extern  bool_t xdr_commit3args (XDR *, commit3args*);
-extern  bool_t xdr_commit3resok (XDR *, commit3resok*);
-extern  bool_t xdr_commit3resfail (XDR *, commit3resfail*);
-extern  bool_t xdr_commit3res (XDR *, commit3res*);
-extern  bool_t xdr_fhandle3 (XDR *, fhandle3*);
-extern  bool_t xdr_dirpath (XDR *, dirpath*);
-extern  bool_t xdr_name (XDR *, name*);
-extern  bool_t xdr_mountstat3 (XDR *, mountstat3*);
-extern  bool_t xdr_mountres3_ok (XDR *, mountres3_ok*);
-extern  bool_t xdr_mountres3 (XDR *, mountres3*);
-extern  bool_t xdr_mountlist (XDR *, mountlist*);
-extern  bool_t xdr_mountbody (XDR *, mountbody*);
-extern  bool_t xdr_groups (XDR *, groups*);
-extern  bool_t xdr_groupnode (XDR *, groupnode*);
-extern  bool_t xdr_exports (XDR *, exports*);
-extern  bool_t xdr_exportnode (XDR *, exportnode*);
-
-extern void xdr_free_exports_list (struct exportnode *first);
-extern void xdr_free_mountlist (mountlist ml);
-
-extern void xdr_free_write3args_nocopy (write3args *wa);
-#endif
diff --git a/xlators/nfs/lib/src/xdr-rpc.c b/xlators/nfs/lib/src/xdr-rpc.c
deleted file mode 100644
index b61b4db1f..000000000
--- a/xlators/nfs/lib/src/xdr-rpc.c
+++ /dev/null
@@ -1,229 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include <string.h>
-#include <rpc/rpc.h>
-#include <rpc/pmap_clnt.h>
-#include <arpa/inet.h>
-#include <rpc/xdr.h>
-#include <sys/uio.h>
-#include <rpc/auth_unix.h>
-
-#include "mem-pool.h"
-#include "xdr-rpc.h"
-#include "xdr-common.h"
-#include "logging.h"
-
-/* Decodes the XDR format in msgbuf into rpc_msg.
- * The remaining payload is returned into payload.
- */
-int
-nfs_xdr_to_rpc_call (char *msgbuf, size_t len, struct rpc_msg *call,
-                     struct iovec *payload, char *credbytes, char *verfbytes)
-{
-        XDR                     xdr;
-        char                    opaquebytes[MAX_AUTH_BYTES];
-        struct opaque_auth      *oa = NULL;
-
-        if ((!msgbuf) || (!call))
-                return -1;
-
-        memset (call, 0, sizeof (*call));
-
-        oa = &call->rm_call.cb_cred;
-        if (!credbytes)
-                oa->oa_base = opaquebytes;
-        else
-                oa->oa_base = credbytes;
-
-        oa = &call->rm_call.cb_verf;
-        if (!verfbytes)
-                oa->oa_base = opaquebytes;
-        else
-                oa->oa_base = verfbytes;
-
-        xdrmem_create (&xdr, msgbuf, len, XDR_DECODE);
-        if (!xdr_callmsg (&xdr, call))
-                return -1;
-
-        if (payload) {
-                payload->iov_base = nfs_xdr_decoded_remaining_addr (xdr);
-                payload->iov_len = nfs_xdr_decoded_remaining_len (xdr);
-        }
-
-        return 0;
-}
-
-
-bool_t
-nfs_true_func (XDR *s, caddr_t *a)
-{
-        return TRUE;
-}
-
-
-int
-nfs_rpc_fill_empty_reply (struct rpc_msg *reply, uint32_t xid)
-{
-        if (!reply)
-                return -1;
-
-        /* Setting to 0 also results in reply verifier flavor to be
-         * set to AUTH_NULL which is what we want right now.
-         */
-        memset (reply, 0, sizeof (*reply));
-        reply->rm_xid = xid;
-        reply->rm_direction = REPLY;
-
-        return 0;
-}
-
-int
-nfs_rpc_fill_denied_reply (struct rpc_msg *reply, int rjstat, int auth_err)
-{
-        if (!reply)
-                return -1;
-
-        reply->rm_reply.rp_stat = MSG_DENIED;
-        reply->rjcted_rply.rj_stat = rjstat;
-        if (rjstat == RPC_MISMATCH) {
-                /* No problem with hardocoding
-                 * RPC version numbers. We only support
-                 * v2 anyway.
-                 */
-                reply->rjcted_rply.rj_vers.low = 2;
-                reply->rjcted_rply.rj_vers.high = 2;
-        } else if (rjstat == AUTH_ERROR)
-                reply->rjcted_rply.rj_why = auth_err;
-
-        return 0;
-}
-
-
-int
-nfs_rpc_fill_accepted_reply (struct rpc_msg *reply, int arstat, int proglow,
-                             int proghigh, int verf, int len, char *vdata)
-{
-        if (!reply)
-                return -1;
-
-        reply->rm_reply.rp_stat = MSG_ACCEPTED;
-        reply->acpted_rply.ar_stat = arstat;
-
-        reply->acpted_rply.ar_verf.oa_flavor = verf;
-        reply->acpted_rply.ar_verf.oa_length = len;
-        reply->acpted_rply.ar_verf.oa_base = vdata;
-        if (arstat == PROG_MISMATCH) {
-                reply->acpted_rply.ar_vers.low = proglow;
-                reply->acpted_rply.ar_vers.high = proghigh;
-        } else if (arstat == SUCCESS) {
-
-                /* This is a hack. I'd really like to build a custom
-                 * XDR library because Sun RPC interface is not very flexible.
-                 */
-                reply->acpted_rply.ar_results.proc = (xdrproc_t)nfs_true_func;
-                reply->acpted_rply.ar_results.where = NULL;
-        }
-
-        return 0;
-}
-
-int
-nfs_rpc_reply_to_xdr (struct rpc_msg *reply, char *dest, size_t len,
-                      struct iovec *dst)
-{
-        XDR             xdr;
-
-        if ((!dest) || (!reply) || (!dst))
-                return -1;
-
-        xdrmem_create (&xdr, dest, len, XDR_ENCODE);
-        if (!xdr_replymsg(&xdr, reply))
-                return -1;
-
-        dst->iov_base = dest;
-        dst->iov_len = nfs_xdr_encoded_length (xdr);
-
-        return 0;
-}
-
-
-int
-nfs_xdr_to_auth_unix_cred (char *msgbuf, int msglen, struct authunix_parms *au,
-                           char *machname, gid_t *gids)
-{
-        XDR             xdr;
-
-        if ((!msgbuf) || (!machname) || (!gids) || (!au))
-                return -1;
-
-        au->aup_machname = machname;
-#ifdef GF_DARWIN_HOST_OS
-        au->aup_gids = (int *)gids;
-#else
-        au->aup_gids = gids;
-#endif
-
-        xdrmem_create (&xdr, msgbuf, msglen, XDR_DECODE);
-
-        if (!xdr_authunix_parms (&xdr, au))
-                return -1;
-
-        return 0;
-}
-
-ssize_t
-nfs_xdr_length_round_up (size_t len, size_t bufsize)
-{
-        int     roundup = 0;
-
-        roundup = len % NFS_XDR_BYTES_PER_UNIT;
-        if (roundup > 0)
-                roundup = NFS_XDR_BYTES_PER_UNIT - roundup;
-
-        if ((roundup > 0) && ((roundup + len) <= bufsize))
-                len += roundup;
-
-        return len;
-}
-
-int
-nfs_xdr_bytes_round_up (struct iovec *vec, size_t bufsize)
-{
-        vec->iov_len = nfs_xdr_length_round_up (vec->iov_len, bufsize);
-        return 0;
-}
-
-void
-nfs_xdr_vector_round_up (struct iovec *vec, int vcount, uint32_t count)
-{
-        uint32_t round_count = 0;
-
-        round_count = nfs_xdr_length_round_up (count, 1048576);
-        round_count -= count;
-        if (round_count == 0)
-                return;
-
-        vec[vcount-1].iov_len += round_count;
-}
diff --git a/xlators/nfs/lib/src/xdr-rpc.h b/xlators/nfs/lib/src/xdr-rpc.h
deleted file mode 100644
index f8dbd33b7..000000000
--- a/xlators/nfs/lib/src/xdr-rpc.h
+++ /dev/null
@@ -1,82 +0,0 @@
-/*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _NFS_XDR_RPC_H
-#define _NFS_XDR_RPC_H_
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include <rpc/rpc.h>
-#include <rpc/pmap_clnt.h>
-#include <arpa/inet.h>
-#include <rpc/xdr.h>
-#include <sys/uio.h>
-
-/* Converts a given network buffer from its XDR format to a structure
- * that contains everything an RPC call needs to work.
- */
-extern int
-nfs_xdr_to_rpc_call (char *msgbuf, size_t len, struct rpc_msg *call,
-                     struct iovec *payload, char *credbytes, char *verfbytes);
-
-extern int
-nfs_rpc_fill_empty_reply (struct rpc_msg *reply, uint32_t xid);
-
-extern int
-nfs_rpc_fill_denied_reply (struct rpc_msg *reply, int rjstat, int auth_err);
-
-extern int
-nfs_rpc_fill_accepted_reply (struct rpc_msg *reply, int arstat, int proglow,
-                             int proghigh, int verf, int len, char *vdata);
-extern int
-nfs_rpc_reply_to_xdr (struct rpc_msg *reply, char *dest, size_t len,
-                      struct iovec *dst);
-
-extern int
-nfs_xdr_to_auth_unix_cred (char *msgbuf, int msglen, struct authunix_parms *au,
-                           char *machname, gid_t *gids);
-/* Macros that simplify accesing the members of an RPC call structure. */
-#define nfs_rpc_call_xid(call)              ((call)->rm_xid)
-#define nfs_rpc_call_direction(call)        ((call)->rm_direction)
-#define nfs_rpc_call_rpcvers(call)          ((call)->ru.RM_cmb.cb_rpcvers)
-#define nfs_rpc_call_program(call)          ((call)->ru.RM_cmb.cb_prog)
-#define nfs_rpc_call_progver(call)          ((call)->ru.RM_cmb.cb_vers)
-#define nfs_rpc_call_progproc(call)         ((call)->ru.RM_cmb.cb_proc)
-#define nfs_rpc_opaque_auth_flavour(oa)     ((oa)->oa_flavor)
-#define nfs_rpc_opaque_auth_len(oa)         ((oa)->oa_length)
-
-#define nfs_rpc_call_cred_flavour(call)     (nfs_rpc_opaque_auth_flavour ((&(call)->ru.RM_cmb.cb_cred)))
-#define nfs_rpc_call_cred_len(call)         (nfs_rpc_opaque_auth_len ((&(call)->ru.RM_cmb.cb_cred)))
-
-
-#define nfs_rpc_call_verf_flavour(call)     (nfs_rpc_opaque_auth_flavour ((&(call)->ru.RM_cmb.cb_verf)))
-#define nfs_rpc_call_verf_len(call)         (nfs_rpc_opaque_auth_len ((&(call)->ru.RM_cmb.cb_verf)))
-
-extern int
-nfs_xdr_bytes_round_up (struct iovec *vec, size_t bufsize);
-
-extern ssize_t
-nfs_xdr_length_round_up (size_t len, size_t bufsize);
-
-void
-nfs_xdr_vector_round_up (struct iovec *vec, int vcount, uint32_t count);
-#endif
diff --git a/xlators/nfs/server/src/Makefile.am b/xlators/nfs/server/src/Makefile.am
index f8968916c..62fbf6535 100644
--- a/xlators/nfs/server/src/Makefile.am
+++ b/xlators/nfs/server/src/Makefile.am
@@ -1,13 +1,24 @@
 xlator_LTLIBRARIES = server.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/nfs
-nfsrpclibdir = $(top_srcdir)/xlators/nfs/lib/src
-server_la_LDFLAGS = -module -avoidversion
-server_la_SOURCES = nfs.c nfs-common.c nfs-fops.c nfs-inodes.c nfs-generics.c mount3.c nfs3-fh.c nfs3.c nfs3-helpers.c $(nfsrpclibdir)/auth-null.c  $(nfsrpclibdir)/auth-unix.c $(nfsrpclibdir)/msg-nfs3.c  $(nfsrpclibdir)/rpc-socket.c  $(nfsrpclibdir)/rpcsvc-auth.c  $(nfsrpclibdir)/rpcsvc.c  $(nfsrpclibdir)/xdr-nfs3.c  $(nfsrpclibdir)/xdr-rpc.c
+nfsrpclibdir = $(top_srcdir)/rpc/rpc-lib/src
+server_la_LDFLAGS = -module -avoid-version
+server_la_SOURCES = nfs.c nfs-common.c nfs-fops.c nfs-inodes.c \
+	nfs-generics.c mount3.c nfs3-fh.c nfs3.c nfs3-helpers.c nlm4.c \
+	nlmcbk_svc.c mount3udp_svc.c acl3.c
 server_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = nfs.h nfs-common.h nfs-fops.h nfs-inodes.h nfs-generics.h mount3.h nfs3-fh.h nfs3.h nfs3-helpers.h nfs-mem-types.h $(nfsrpclibdir)/xdr-rpc.h $(nfsrpclibdir)/msg-nfs3.h $(nfsrpclibdir)/xdr-common.h $(nfsrpclibdir)/xdr-nfs3.h $(nfsrpclibdir)/rpc-socket.h $(nfsrpclibdir)/rpcsvc.h
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)\
-	-I$(nfsrpclibdir) -L$(xlatordir)/ -I$(CONTRIBDIR)/rbtree
+noinst_HEADERS = nfs.h nfs-common.h nfs-fops.h nfs-inodes.h nfs-generics.h \
+	mount3.h nfs3-fh.h nfs3.h nfs3-helpers.h nfs-mem-types.h nlm4.h \
+	acl3.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) \
+	-DLIBDIR=\"$(libdir)/glusterfs/$(PACKAGE_VERSION)/auth\" \
+	-I$(top_srcdir)/libglusterfs/src \
+	-I$(nfsrpclibdir) -I$(CONTRIBDIR)/rbtree \
+	-I$(top_srcdir)/rpc/xdr/src/
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+AM_LDFLAGS = -L$(xlatordir)
 
 CLEANFILES =
diff --git a/xlators/nfs/server/src/acl3.c b/xlators/nfs/server/src/acl3.c
new file mode 100644
index 000000000..08b099b4e
--- /dev/null
+++ b/xlators/nfs/server/src/acl3.c
@@ -0,0 +1,708 @@
+/*
+ * Copyright (c) 2012 Red Hat, Inc. <http://www.redhat.com>
+ * This file is part of GlusterFS.
+ *
+ * This file is licensed to you under your choice of the GNU Lesser
+ * General Public License, version 3 or any later version (LGPLv3 or
+ * later), or the GNU General Public License, version 2 (GPLv2), in all
+ * cases as published by the Free Software Foundation.
+ */
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "defaults.h"
+#include "rpcsvc.h"
+#include "dict.h"
+#include "xlator.h"
+#include "nfs.h"
+#include "mem-pool.h"
+#include "logging.h"
+#include "nfs-fops.h"
+#include "inode.h"
+#include "nfs3.h"
+#include "nfs-mem-types.h"
+#include "nfs3-helpers.h"
+#include "nfs3-fh.h"
+#include "nfs-generics.h"
+#include "acl3.h"
+
+
+typedef ssize_t (*acl3_serializer) (struct iovec outmsg, void *args);
+
+extern void nfs3_call_state_wipe (nfs3_call_state_t *cs);
+
+extern nfs3_call_state_t *
+nfs3_call_state_init (struct nfs3_state *s, rpcsvc_request_t *req, xlator_t *v);
+
+extern int
+nfs3_fh_validate (struct nfs3_fh *fh);
+
+extern fattr3
+nfs3_stat_to_fattr3 (struct iatt *buf);
+
+#define acl3_validate_nfs3_state(request, state, status, label, retval) \
+        do      {                                                       \
+                state = rpcsvc_request_program_private (request);       \
+                if (!state) {                                           \
+                        gf_log (GF_ACL, GF_LOG_ERROR, "NFSv3 state "    \
+                                "missing from RPC request");            \
+                        rpcsvc_request_seterr (req, SYSTEM_ERR);        \
+                        status = NFS3ERR_SERVERFAULT;                   \
+                        goto label;                                     \
+                }                                                       \
+        } while (0);                                                    \
+
+#define acl3_validate_gluster_fh(handle, status, errlabel)              \
+        do {                                                            \
+                if (!nfs3_fh_validate (handle)) {                       \
+                        status = NFS3ERR_SERVERFAULT;                   \
+                        goto errlabel;                                  \
+                }                                                       \
+        } while (0)                                                     \
+
+
+extern xlator_t *
+nfs3_fh_to_xlator (struct nfs3_state *nfs3, struct nfs3_fh *fh);
+
+#define acl3_map_fh_to_volume(nfs3state, handle, req, volume, status, label) \
+        do {                                                            \
+                char exportid[256], gfid[256];                          \
+                rpc_transport_t *trans = NULL;                          \
+                volume = nfs3_fh_to_xlator ((nfs3state), handle);      \
+                if (!volume) {                                          \
+                        uuid_unparse (handle->exportid, exportid);       \
+                        uuid_unparse (handle->gfid, gfid);               \
+                        trans = rpcsvc_request_transport (req);         \
+                        gf_log (GF_ACL, GF_LOG_ERROR, "Failed to map "  \
+                                "FH to vol: client=%s, exportid=%s, gfid=%s",\
+                                trans->peerinfo.identifier, exportid,   \
+                                gfid);                                  \
+                        gf_log (GF_ACL, GF_LOG_ERROR,                   \
+                                "Stale nfs client %s must be trying to "\
+                                "connect to a deleted volume, please "  \
+                                "unmount it.", trans->peerinfo.identifier);\
+                        status = NFS3ERR_STALE;                         \
+                        goto label;                                     \
+                } else {                                                \
+                        gf_log (GF_ACL, GF_LOG_TRACE, "FH to Volume: %s"\
+                                ,volume->name);                         \
+                        rpcsvc_request_set_private (req, volume);       \
+                }                                                       \
+        } while (0);                                                    \
+
+#define acl3_volume_started_check(nfs3state, vlm, rtval, erlbl)         \
+        do {                                                            \
+              if ((!nfs_subvolume_started (nfs_state (nfs3state->nfsx), vlm))){\
+                      gf_log (GF_ACL, GF_LOG_ERROR, "Volume is disabled: %s",\
+                              vlm->name);                               \
+                      rtval = RPCSVC_ACTOR_IGNORE;                      \
+                      goto erlbl;                                       \
+              }                                                         \
+        } while (0)                                                     \
+
+#define acl3_check_fh_resolve_status(cst, nfstat, erlabl)               \
+        do {                                                            \
+                xlator_t *xlatorp = NULL;                               \
+                char buf[256], gfid[256];                               \
+                rpc_transport_t *trans = NULL;                          \
+                if ((cst)->resolve_ret < 0) {                           \
+                        trans = rpcsvc_request_transport (cst->req);    \
+                        xlatorp = nfs3_fh_to_xlator (cst->nfs3state,    \
+                                                     &cst->resolvefh);  \
+                        uuid_unparse (cst->resolvefh.gfid, gfid);       \
+                        snprintf (buf, sizeof (buf), "(%s) %s : %s",             \
+                                  trans->peerinfo.identifier,           \
+                                  xlatorp ? xlatorp->name : "ERR",      \
+                                  gfid);                                \
+                        gf_log (GF_ACL, GF_LOG_ERROR, "Unable to resolve FH"\
+                                ": %s", buf);                           \
+                        nfstat = nfs3_errno_to_nfsstat3 (cst->resolve_errno);\
+                        goto erlabl;                                    \
+                }                                                       \
+        } while (0)                                                     \
+
+#define acl3_handle_call_state_init(nfs3state, calls, rq, v, opstat, errlabel)\
+        do {                                                            \
+                calls = nfs3_call_state_init ((nfs3state), (rq), v); \
+                if (!calls) {                                           \
+                        gf_log (GF_ACL, GF_LOG_ERROR, "Failed to "      \
+                                "init call state");                     \
+                        opstat = NFS3ERR_SERVERFAULT;                   \
+                        rpcsvc_request_seterr (req, SYSTEM_ERR);        \
+                        goto errlabel;                                  \
+                }                                                       \
+        } while (0)                                                     \
+
+
+int
+acl3svc_submit_reply (rpcsvc_request_t *req, void *arg, acl3_serializer sfunc)
+{
+        struct iovec            outmsg  = {0, };
+        struct iobuf            *iob    = NULL;
+        struct nfs3_state       *nfs3   = NULL;
+        int                     ret     = -1;
+        ssize_t                 msglen  = 0;
+        struct iobref           *iobref = NULL;
+
+        if (!req)
+                return -1;
+
+        nfs3 = (struct nfs3_state *)rpcsvc_request_program_private (req);
+        if (!nfs3) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "mount state not found");
+                goto ret;
+        }
+
+        /* First, get the io buffer into which the reply in arg will
+         * be serialized.
+         */
+        iob = iobuf_get (nfs3->iobpool);
+        if (!iob) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "Failed to get iobuf");
+                goto ret;
+        }
+
+        iobuf_to_iovec (iob, &outmsg);
+        /* Use the given serializer to translate the give C structure in arg
+         * to XDR format which will be written into the buffer in outmsg.
+         */
+        msglen = sfunc (outmsg, arg);
+        if (msglen < 0) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "Failed to encode message");
+                goto ret;
+        }
+        outmsg.iov_len = msglen;
+
+        iobref = iobref_new ();
+        if (iobref == NULL) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "Failed to get iobref");
+                goto ret;
+        }
+
+        ret = iobref_add (iobref, iob);
+        if (ret) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "Failed to add iob to iobref");
+                goto ret;
+        }
+
+        /* Then, submit the message for transmission. */
+        ret = rpcsvc_submit_message (req, &outmsg, 1, NULL, 0, iobref);
+        if (ret == -1) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "Reply submission failed");
+                goto ret;
+        }
+
+        ret = 0;
+ret:
+        if (iob)
+                iobuf_unref (iob);
+        if (iobref)
+                iobref_unref (iobref);
+
+        return ret;
+}
+
+
+int
+acl3svc_null (rpcsvc_request_t *req)
+{
+        struct iovec    dummyvec = {0, };
+
+        if (!req) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "Got NULL request!");
+                return 0;
+        }
+        rpcsvc_submit_generic (req, &dummyvec, 1,  NULL, 0, NULL);
+        return 0;
+}
+
+int
+acl3_getacl_reply (nfs3_call_state_t *cs, getaclreply *reply)
+{
+        acl3svc_submit_reply (cs->req, (void *)reply,
+                              (acl3_serializer)xdr_serialize_getaclreply);
+        return 0;
+}
+
+
+int
+acl3_getacl_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *dict,
+                 dict_t *xdata)
+{
+        nfsstat3                        stat = NFS3ERR_SERVERFAULT;
+        nfs3_call_state_t              *cs = NULL;
+        data_t                         *data = NULL;
+        int                            *p = NULL;
+        int                             i = 0;
+        getaclreply                     *getaclreply = NULL;
+
+        if (!frame->local) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "Invalid argument,"
+                       " frame->local NULL");
+                return  EINVAL;
+        }
+        cs = frame->local;
+        getaclreply = &cs->args.getaclreply;
+        if (op_ret == -1) {
+                stat = nfs3_cbk_errno_status (op_ret, op_errno);
+                goto err;
+        }
+
+        getaclreply->aclentry.aclentry_val = cs->aclentry;
+        getaclreply->daclentry.daclentry_val = cs->daclentry;
+
+        /* FIXME: use posix_acl_from_xattr() */
+        data = dict_get (dict, POSIX_ACL_ACCESS_XATTR);
+        if (data && (p = data_to_bin (data))) {
+                /* POSIX_ACL_VERSION */
+                p++;
+                while ((char *)p < (data->data + data->len)) {
+                        getaclreply->aclentry.aclentry_val[i].type = *(*(short **)&p)++;
+                        getaclreply->aclentry.aclentry_val[i].perm = *(*(short **)&p)++;
+                        getaclreply->aclentry.aclentry_val[i].uid = *(*(int **)&p)++;
+                        i++;
+                }
+                getaclreply->aclcount = getaclreply->aclentry.aclentry_len = i;
+        }
+        i = 0;
+
+        data = dict_get (dict, POSIX_ACL_DEFAULT_XATTR);
+        if (data && (p = data_to_bin (data))) {
+                /* POSIX_ACL_VERSION */
+                p++;
+                while ((char *)p < (data->data + data->len)) {
+                        getaclreply->daclentry.daclentry_val[i].type = *(*(short **)&p)++;
+                        getaclreply->daclentry.daclentry_val[i].perm = *(*(short **)&p)++;
+                        getaclreply->daclentry.daclentry_val[i].uid = *(*(int **)&p)++;
+                        i++;
+                }
+                getaclreply->daclcount = getaclreply->daclentry.daclentry_len = i;
+        }
+
+        acl3_getacl_reply (cs, getaclreply);
+        nfs3_call_state_wipe (cs);
+        return 0;
+
+err:
+        if (getaclreply)
+                getaclreply->status = stat;
+        acl3_getacl_reply (cs, getaclreply);
+        nfs3_call_state_wipe (cs);
+        return 0;
+}
+
+int
+acl3_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                  dict_t *xdata)
+{
+        nfsstat3                        stat = NFS3ERR_SERVERFAULT;
+        nfs3_call_state_t              *cs = NULL;
+        getaclreply                     *getaclreply = NULL;
+        int                             ret = -1;
+        nfs_user_t                      nfu = {0, };
+
+        if (!frame->local) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "Invalid argument,"
+                       " frame->local NULL");
+                return EINVAL;
+        }
+
+        cs = frame->local;
+        getaclreply = &cs->args.getaclreply;
+
+        if (op_ret == -1) {
+                stat = nfs3_cbk_errno_status (op_ret, op_errno);
+                goto err;
+        }
+
+        getaclreply->attr_follows = 1;
+        getaclreply->attr = nfs3_stat_to_fattr3 (buf);
+        getaclreply->mask = 0xf;
+        nfs_request_user_init (&nfu, cs->req);
+        ret = nfs_getxattr (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, NULL, NULL,
+                            acl3_getacl_cbk, cs);
+        if (ret == -1) {
+                stat = nfs3_cbk_errno_status (op_ret, op_errno);
+                goto err;
+        }
+        return 0;
+err:
+        getaclreply->status = stat;
+        acl3_getacl_reply (cs, getaclreply);
+        nfs3_call_state_wipe (cs);
+        return 0;
+}
+
+
+int
+acl3_getacl_resume (void *carg)
+{
+        int                             ret = -1;
+        nfs3_call_state_t               *cs = NULL;
+        nfsstat3                        stat = NFS3ERR_SERVERFAULT;
+        nfs_user_t                      nfu = {0, };
+
+        if (!carg)
+                return ret;
+
+        cs = (nfs3_call_state_t *)carg;
+        acl3_check_fh_resolve_status (cs, stat, acl3err);
+        nfs_request_user_init (&nfu, cs->req);
+
+        ret = nfs_stat (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc,
+                        acl3_stat_cbk, cs);
+        stat = -ret;
+acl3err:
+        if (ret < 0) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "unable to open_and_resume");
+                cs->args.getaclreply.status = nfs3_errno_to_nfsstat3 (stat);
+                acl3_getacl_reply (cs, &cs->args.getaclreply);
+                nfs3_call_state_wipe (cs);
+        }
+
+        return ret;
+}
+
+
+int
+acl3svc_getacl (rpcsvc_request_t *req)
+{
+        xlator_t                        *vol = NULL;
+        struct nfs_state               *nfs = NULL;
+        nfs3_state_t                   *nfs3 = NULL;
+        nfs3_call_state_t               *cs = NULL;
+        int                             ret = RPCSVC_ACTOR_ERROR;
+        nfsstat3                        stat = NFS3ERR_SERVERFAULT;
+        struct nfs3_fh                  fh, *fhp = NULL;
+        getaclargs                      getaclargs;
+
+        if (!req)
+                return ret;
+
+        acl3_validate_nfs3_state (req, nfs3, stat, rpcerr, ret);
+        nfs = nfs_state (nfs3->nfsx);
+        memset (&getaclargs, 0, sizeof (getaclargs));
+        getaclargs.fh.n_bytes = (char *)&fh;
+        if (xdr_to_getaclargs(req->msg[0], &getaclargs) <= 0) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "Error decoding args");
+                rpcsvc_request_seterr (req, GARBAGE_ARGS);
+                goto rpcerr;
+        }
+        fhp = &fh;
+        acl3_validate_gluster_fh (&fh, stat, acl3err);
+        acl3_map_fh_to_volume (nfs->nfs3state, fhp, req,
+                               vol, stat, acl3err);
+        acl3_handle_call_state_init (nfs->nfs3state, cs, req,
+                                     vol, stat, rpcerr);
+
+        cs->vol = vol;
+        acl3_volume_started_check (nfs3, vol, ret, acl3err);
+
+        ret = nfs3_fh_resolve_and_resume (cs, fhp,
+                                          NULL, acl3_getacl_resume);
+
+acl3err:
+        if (ret < 0) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "unable to resolve and resume");
+                if (cs) {
+                        cs->args.getaclreply.status = stat;
+                        acl3_getacl_reply (cs, &cs->args.getaclreply);
+                        nfs3_call_state_wipe (cs);
+                }
+                return 0;
+        }
+
+rpcerr:
+        return ret;
+}
+
+int
+acl3_setacl_cbk (call_frame_t *frame, void *cookie,
+                 xlator_t *this, int32_t op_ret, int32_t op_errno,
+                 dict_t *xdata)
+{
+        nfs3_call_state_t               *cs = NULL;
+        cs = frame->local;
+        if (op_ret < 0) {
+                nfsstat3 status = nfs3_cbk_errno_status (op_ret, op_errno);
+                cs->args.setaclreply.status = status;
+        }
+
+        acl3svc_submit_reply (cs->req, (void *)&cs->args.setaclreply,
+                              (acl3_serializer)xdr_serialize_setaclreply);
+        return 0;
+}
+
+int
+acl3_setacl_resume (void *carg)
+{
+        int                             ret = -1;
+        nfs3_call_state_t               *cs = NULL;
+        nfsstat3                        stat = NFS3ERR_SERVERFAULT;
+        nfs_user_t                      nfu = {0, };
+        dict_t                          *xattr = NULL;
+
+        if (!carg)
+                return ret;
+        cs = (nfs3_call_state_t *)carg;
+        acl3_check_fh_resolve_status (cs, stat, acl3err);
+        nfs_request_user_init (&nfu, cs->req);
+        xattr = dict_new();
+        if (cs->aclcount)
+        ret = dict_set_static_bin (xattr, POSIX_ACL_ACCESS_XATTR, cs->aclxattr,
+                                   cs->aclcount * 8 + 4);
+        if (cs->daclcount)
+        ret = dict_set_static_bin (xattr, POSIX_ACL_DEFAULT_XATTR,
+                                   cs->daclxattr, cs->daclcount * 8 + 4);
+
+        ret = nfs_setxattr (cs->nfsx, cs->vol, &nfu, &cs->resolvedloc, xattr,
+                            0, NULL, acl3_setacl_cbk, cs);
+        dict_unref (xattr);
+
+acl3err:
+        if (ret < 0) {
+                stat = -ret;
+                gf_log (GF_ACL, GF_LOG_ERROR, "unable to open_and_resume");
+                cs->args.setaclreply.status = nfs3_errno_to_nfsstat3 (stat);
+                acl3svc_submit_reply (cs->req, (void *)&cs->args.setaclreply,
+                                      (acl3_serializer)xdr_serialize_setaclreply);
+                nfs3_call_state_wipe (cs);
+        }
+
+        return ret;
+}
+
+
+int
+acl3svc_setacl (rpcsvc_request_t *req)
+{
+        xlator_t                        *vol = NULL;
+        struct nfs_state               *nfs = NULL;
+        nfs3_state_t                   *nfs3 = NULL;
+        nfs3_call_state_t               *cs = NULL;
+        int                             ret = RPCSVC_ACTOR_ERROR;
+        nfsstat3                        stat = NFS3ERR_SERVERFAULT;
+        struct nfs3_fh                  fh;
+        struct nfs3_fh                 *fhp = NULL;
+        setaclargs                      setaclargs;
+        aclentry                        *aclentry = NULL;
+        struct aclentry                 *daclentry = NULL;
+        int                             i = 0;
+        struct posix_acl_xattr_header   *bufheader = NULL;
+        struct posix_acl_xattr_entry    *bufentry  = NULL;
+
+        if (!req)
+                return ret;
+        aclentry =  GF_CALLOC (NFS_ACL_MAX_ENTRIES, sizeof(*aclentry),
+                               gf_nfs_mt_arr);
+        if (!aclentry) {
+                goto rpcerr;
+        }
+        daclentry = GF_CALLOC (NFS_ACL_MAX_ENTRIES, sizeof(*daclentry),
+                               gf_nfs_mt_arr);
+        if (!daclentry) {
+                goto rpcerr;
+        }
+
+        acl3_validate_nfs3_state (req, nfs3, stat, rpcerr, ret);
+        nfs = nfs_state (nfs3->nfsx);
+        memset (&setaclargs, 0, sizeof (setaclargs));
+        memset (&fh, 0, sizeof (fh));
+        setaclargs.fh.n_bytes = (char *)&fh;
+        setaclargs.aclentry.aclentry_val = aclentry;
+        setaclargs.daclentry.daclentry_val = daclentry;
+        if (xdr_to_setaclargs(req->msg[0], &setaclargs) <= 0) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "Error decoding args");
+                rpcsvc_request_seterr (req, GARBAGE_ARGS);
+                goto rpcerr;
+        }
+        fhp = &fh;
+        acl3_validate_gluster_fh (fhp, stat, acl3err);
+        acl3_map_fh_to_volume (nfs->nfs3state, fhp, req,
+                               vol, stat, acl3err);
+        acl3_handle_call_state_init (nfs->nfs3state, cs, req,
+                                     vol, stat, rpcerr);
+
+        cs->vol = vol;
+        acl3_volume_started_check (nfs3, vol, ret, rpcerr);
+
+        cs->aclcount = setaclargs.aclcount;
+        cs->daclcount = setaclargs.daclcount;
+
+        if ((cs->aclcount > NFS_ACL_MAX_ENTRIES) ||
+                        (cs->daclcount > NFS_ACL_MAX_ENTRIES))
+                goto acl3err;
+        /* FIXME: use posix_acl_to_xattr() */
+        /* Populate xattr buffer for user ACL */
+        bufheader = (struct posix_acl_xattr_header *)(cs->aclxattr);
+        bufheader->version = htole32(POSIX_ACL_VERSION);
+        bufentry  = bufheader->entries;
+        for (i = 0; i < cs->aclcount; i++) {
+                int uaceuid;
+                const struct aclentry *uace = &aclentry[i];
+                switch (uace->type) {
+                    case POSIX_ACL_USER:
+                    case POSIX_ACL_GROUP:
+                        uaceuid = uace->uid;
+                        break;
+                    default:
+                        uaceuid = POSIX_ACL_UNDEFINED_ID;
+                        break;
+                }
+                bufentry->tag  = htole16(uace->type);
+                bufentry->perm = htole16(uace->perm);
+                bufentry->id   = htole32(uaceuid);
+
+                bufentry++;
+        }
+
+        /* Populate xattr buffer for Default ACL */
+        bufheader = (struct posix_acl_xattr_header *)(cs->aclxattr);
+        bufheader->version = htole32(POSIX_ACL_VERSION);
+        bufentry  = bufheader->entries;
+        for (i = 0; i < cs->daclcount; i++) {
+                int daceuid;
+                int dacetype;
+                const struct aclentry *dace = &daclentry[i];
+                /*
+                 * For "default ACL", NFSv3 handles the 'type' differently
+                 * i.e. by logical OR'ing 'type' with NFS_ACL_DEFAULT.
+                 * Which the backend File system does not understand and
+                 * that needs to be masked OFF.
+                 */
+                dacetype = (dace->type & ~(NFS_ACL_DEFAULT));
+                switch (dacetype) {
+                    case POSIX_ACL_USER:
+                    case POSIX_ACL_GROUP:
+                        daceuid = dace->uid;
+                        break;
+                    default:
+                        daceuid = POSIX_ACL_UNDEFINED_ID;
+                        break;
+                }
+                bufentry->tag  = htole16(dacetype);
+                bufentry->perm = htole16(dace->perm);
+                bufentry->id   = htole32(daceuid);
+
+                bufentry++;
+        }
+
+
+        ret = nfs3_fh_resolve_and_resume (cs, fhp,
+                                          NULL, acl3_setacl_resume);
+
+acl3err:
+        if (ret < 0) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "unable to resolve and resume");
+                cs->args.setaclreply.status = stat;
+                acl3svc_submit_reply (cs->req, (void *)&cs->args.setaclreply,
+                                      (acl3_serializer)xdr_serialize_setaclreply);
+                nfs3_call_state_wipe (cs);
+                GF_FREE(aclentry);
+                GF_FREE(daclentry);
+                return 0;
+        }
+
+rpcerr:
+        if (ret < 0)
+                nfs3_call_state_wipe (cs);
+        if (aclentry)
+                GF_FREE (aclentry);
+        if (daclentry)
+                GF_FREE (daclentry);
+        return ret;
+}
+
+
+
+rpcsvc_actor_t  acl3svc_actors[ACL3_PROC_COUNT] = {
+        {"NULL",       ACL3_NULL,      acl3svc_null,   NULL,   0},
+        {"GETACL",     ACL3_GETACL,    acl3svc_getacl, NULL,   0},
+        {"SETACL",     ACL3_SETACL,    acl3svc_setacl, NULL,   0},
+};
+
+rpcsvc_program_t        acl3prog = {
+        .progname       = "ACL3",
+        .prognum        = ACL_PROGRAM,
+        .progver        = ACLV3_VERSION,
+        .progport       = GF_NFS3_PORT,
+        .actors         = acl3svc_actors,
+        .numactors      = ACL3_PROC_COUNT,
+        .min_auth       = AUTH_NULL,
+};
+
+rpcsvc_program_t *
+acl3svc_init(xlator_t *nfsx)
+{
+        struct nfs3_state *ns = NULL;
+        struct nfs_state *nfs = NULL;
+        dict_t *options = NULL;
+        int ret = -1;
+        char *portstr = NULL;
+        static gf_boolean_t acl3_inited = _gf_false;
+
+        /* Already inited */
+        if (acl3_inited)
+                return &acl3prog;
+
+        nfs = (struct nfs_state*)nfsx->private;
+
+        ns = nfs->nfs3state;
+        if (!ns) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "ACL3 init failed");
+                goto err;
+        }
+        acl3prog.private = ns;
+
+        options = dict_new ();
+
+        ret = gf_asprintf (&portstr, "%d", GF_ACL3_PORT);
+        if (ret == -1)
+                goto err;
+
+        ret = dict_set_dynstr (options, "transport.socket.listen-port",
+                               portstr);
+        if (ret == -1)
+                goto err;
+        ret = dict_set_str (options, "transport-type", "socket");
+        if (ret == -1) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "dict_set_str error");
+                goto err;
+        }
+
+        if (nfs->allow_insecure) {
+                ret = dict_set_str (options, "rpc-auth-allow-insecure", "on");
+                if (ret == -1) {
+                        gf_log (GF_ACL, GF_LOG_ERROR, "dict_set_str error");
+                        goto err;
+                }
+                ret = dict_set_str (options, "rpc-auth.ports.insecure", "on");
+                if (ret == -1) {
+                        gf_log (GF_ACL, GF_LOG_ERROR, "dict_set_str error");
+                        goto err;
+                }
+        }
+
+        ret = dict_set_str (options, "transport.address-family", "inet");
+        if (ret == -1) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "dict_set_str error");
+                goto err;
+        }
+
+        ret = rpcsvc_create_listeners (nfs->rpcsvc, options, "ACL");
+        if (ret == -1) {
+                gf_log (GF_ACL, GF_LOG_ERROR, "Unable to create listeners");
+                dict_unref (options);
+                goto err;
+        }
+
+        acl3_inited = _gf_true;
+        return &acl3prog;
+err:
+        return NULL;
+}
diff --git a/xlators/nfs/server/src/acl3.h b/xlators/nfs/server/src/acl3.h
new file mode 100644
index 000000000..e0e61281a
--- /dev/null
+++ b/xlators/nfs/server/src/acl3.h
@@ -0,0 +1,31 @@
+/*
+ * Copyright (c) 2012 Red Hat, Inc. <http://www.redhat.com>
+ * This file is part of GlusterFS.
+ *
+ * This file is licensed to you under your choice of the GNU Lesser
+ * General Public License, version 3 or any later version (LGPLv3 or
+ * later), or the GNU General Public License, version 2 (GPLv2), in all
+ * cases as published by the Free Software Foundation.
+ */
+
+#ifndef _ACL3_H
+#define _ACL3_H
+
+#include "glusterfs-acl.h"
+
+#define GF_ACL3_PORT            38469
+#define GF_ACL                  GF_NFS"-ACL"
+
+/*
+ * NFSv3, identifies the default ACL by NFS_ACL_DEFAULT. Gluster
+ * NFS needs to mask it OFF before sending it upto POSIX layer
+ * or File system layer.
+ */
+#define NFS_ACL_DEFAULT             0x1000
+
+#define NFS_ACL_MAX_ENTRIES         1024
+
+rpcsvc_program_t *
+acl3svc_init(xlator_t *nfsx);
+
+#endif
diff --git a/xlators/nfs/server/src/mount3.c b/xlators/nfs/server/src/mount3.c
index 84e1456fb..b0824bf10 100644
--- a/xlators/nfs/server/src/mount3.c
+++ b/xlators/nfs/server/src/mount3.c
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -37,28 +28,79 @@
 #include "locking.h"
 #include "iatt.h"
 #include "nfs-mem-types.h"
-
+#include "nfs.h"
+#include "common-utils.h"
+#include "store.h"
 
 #include <errno.h>
 #include <sys/socket.h>
 #include <sys/uio.h>
 
+
+#define IPv4_ADDR_SIZE  32
+
+/* Macro to typecast the parameter to struct sockaddr_in
+ */
+#define SA(addr) ((struct sockaddr_in*)(addr))
+
+/* Macro will mask the ip address with netmask.
+ */
+#define MASKED_IP(ipv4addr, netmask)                    \
+                (ntohl(SA(ipv4addr)->sin_addr.s_addr) & (netmask))
+
+/* Macro will compare two IP address after applying the mask
+ */
+#define COMPARE_IPv4_ADDRS(ip1, ip2, netmask)           \
+                ((MASKED_IP(ip1, netmask)) == (MASKED_IP(ip2, netmask)))
+
+/* This macro will assist in freeing up entire link list
+ * of host_auth_spec structure.
+ */
+#define FREE_HOSTSPEC(exp) do {                                 \
+                struct host_auth_spec *host= exp->hostspec;     \
+                while (NULL != host){                           \
+                        struct host_auth_spec* temp = host;     \
+                        host = host->next;                      \
+                        if (NULL != temp->host_addr) {          \
+                                GF_FREE (temp->host_addr);      \
+                        }                                       \
+                        GF_FREE (temp);                         \
+                }                                               \
+                exp->hostspec = NULL;                           \
+        } while (0)
+
 typedef ssize_t (*mnt3_serializer) (struct iovec outmsg, void *args);
 
+extern void *
+mount3udp_thread (void *argv);
+
+static inline void
+mnt3_export_free (struct mnt3_export *exp)
+{
+        if (!exp)
+                return;
+
+        if (exp->exptype == MNT3_EXPTYPE_DIR)
+                FREE_HOSTSPEC (exp);
+        GF_FREE (exp->expname);
+        GF_FREE (exp);
+}
 
 /* Generic reply function for MOUNTv3 specific replies. */
 int
 mnt3svc_submit_reply (rpcsvc_request_t *req, void *arg, mnt3_serializer sfunc)
 {
-        struct iovec            outmsg = {0, };
-        struct iobuf            *iob = NULL;
-        struct mount3_state     *ms = NULL;
-        int                     ret = -1;
+        struct iovec            outmsg  = {0, };
+        struct iobuf            *iob    = NULL;
+        struct mount3_state     *ms     = NULL;
+        int                     ret     = -1;
+        ssize_t                 msglen  = 0;
+        struct iobref           *iobref = NULL;
 
         if (!req)
                 return -1;
 
-        ms = (struct mount3_state *)nfs_rpcsvc_request_program_private (req);
+        ms = (struct mount3_state *)rpcsvc_request_program_private (req);
         if (!ms) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "mount state not found");
                 goto ret;
@@ -67,6 +109,7 @@ mnt3svc_submit_reply (rpcsvc_request_t *req, void *arg, mnt3_serializer sfunc)
         /* First, get the io buffer into which the reply in arg will
          * be serialized.
          */
+        /* TODO: use 'xdrproc_t' instead of 'sfunc' to get the xdr-size */
         iob = iobuf_get (ms->iobpool);
         if (!iob) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Failed to get iobuf");
@@ -77,11 +120,27 @@ mnt3svc_submit_reply (rpcsvc_request_t *req, void *arg, mnt3_serializer sfunc)
         /* Use the given serializer to translate the give C structure in arg
          * to XDR format which will be written into the buffer in outmsg.
          */
-        outmsg.iov_len = sfunc (outmsg, arg);
+        msglen = sfunc (outmsg, arg);
+        if (msglen < 0) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to encode message");
+                goto ret;
+        }
+        outmsg.iov_len = msglen;
+
+        iobref = iobref_new ();
+        if (iobref == NULL) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to get iobref");
+                goto ret;
+        }
+
+        ret = iobref_add (iobref, iob);
+        if (ret) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to add iob to iobref");
+                goto ret;
+        }
 
         /* Then, submit the message for transmission. */
-        ret = nfs_rpcsvc_submit_message (req, outmsg, iob);
-        iobuf_unref (iob);
+        ret = rpcsvc_submit_message (req, &outmsg, 1, NULL, 0, iobref);
         if (ret == -1) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Reply submission failed");
                 goto ret;
@@ -89,6 +148,11 @@ mnt3svc_submit_reply (rpcsvc_request_t *req, void *arg, mnt3_serializer sfunc)
 
         ret = 0;
 ret:
+        if (NULL != iob)
+                iobuf_unref (iob);
+        if (NULL != iobref)
+                iobref_unref (iobref);
+
         return ret;
 }
 
@@ -172,13 +236,278 @@ mnt3svc_set_mountres3 (mountstat3 stat, struct nfs3_fh *fh, int *authflavor,
         return res;
 }
 
+/* Read the rmtab from the store_handle and append (or not) the entries to the
+ * mountlist.
+ *
+ * Requires the store_handle to be locked.
+ */
+static int
+__mount_read_rmtab (gf_store_handle_t *sh, struct list_head *mountlist,
+                    gf_boolean_t append)
+{
+        int                     ret = 0;
+        unsigned int            idx = 0;
+        struct mountentry       *me = NULL, *tmp = NULL;
+                                /* me->hostname is a char[MNTPATHLEN] */
+        char                    key[MNTPATHLEN + 11];
 
+        GF_ASSERT (sh && mountlist);
+
+        if (!gf_store_locked_local (sh)) {
+                gf_log (GF_MNT, GF_LOG_WARNING, "Not reading unlocked %s",
+                        sh->path);
+                return -1;
+        }
+
+        if (!append) {
+                list_for_each_entry_safe (me, tmp, mountlist, mlist) {
+                        list_del (&me->mlist);
+                        GF_FREE (me);
+                }
+                me = NULL;
+        }
+
+        for (;;) {
+                char *value = NULL;
+
+                if (me && append) {
+                        /* do not add duplicates */
+                        list_for_each_entry (tmp, mountlist, mlist) {
+                                if (!strcmp(tmp->hostname, me->hostname) &&
+                                    !strcmp(tmp->exname, me->exname)) {
+                                        GF_FREE (me);
+                                        goto dont_add;
+                                }
+                        }
+                        list_add_tail (&me->mlist, mountlist);
+                } else if (me) {
+                        list_add_tail (&me->mlist, mountlist);
+                }
+
+dont_add:
+                me = GF_CALLOC (1, sizeof (*me), gf_nfs_mt_mountentry);
+                if (!me) {
+                        gf_log (GF_MNT, GF_LOG_ERROR, "Out of memory");
+                        ret = -1;
+                        goto out;
+                }
+
+                INIT_LIST_HEAD (&me->mlist);
+
+                snprintf (key, 9 + MNTPATHLEN, "hostname-%d", idx);
+                ret = gf_store_retrieve_value (sh, key, &value);
+                if (ret)
+                        break;
+                strncpy (me->hostname, value, MNTPATHLEN);
+                GF_FREE (value);
+
+                snprintf (key, 11 + MNTPATHLEN, "mountpoint-%d", idx);
+                ret = gf_store_retrieve_value (sh, key, &value);
+                if (ret)
+                        break;
+                strncpy (me->exname, value, MNTPATHLEN);
+                GF_FREE (value);
+
+                idx++;
+                gf_log (GF_MNT, GF_LOG_TRACE, "Read entries %s:%s", me->hostname, me->exname);
+        }
+        gf_log (GF_MNT, GF_LOG_DEBUG, "Read %d entries from '%s'", idx, sh->path);
+        GF_FREE (me);
+out:
+        return ret;
+}
+
+/* Overwrite the contents of the rwtab with te in-memory client list.
+ * Fail gracefully if the stora_handle is not locked.
+ */
+static void
+__mount_rewrite_rmtab(struct mount3_state *ms, gf_store_handle_t *sh)
+{
+        struct mountentry       *me = NULL;
+        char                    key[16];
+        int                     fd, ret;
+        unsigned int            idx = 0;
+
+        if (!gf_store_locked_local (sh)) {
+                gf_log (GF_MNT, GF_LOG_WARNING, "Not modifying unlocked %s",
+                        sh->path);
+                return;
+        }
+
+        fd = gf_store_mkstemp (sh);
+        if (fd == -1) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to open %s", sh->path);
+                return;
+        }
+
+        list_for_each_entry (me, &ms->mountlist, mlist) {
+                snprintf (key, 16, "hostname-%d", idx);
+                ret = gf_store_save_value (fd, key, me->hostname);
+                if (ret)
+                        goto fail;
+
+                snprintf (key, 16, "mountpoint-%d", idx);
+                ret = gf_store_save_value (fd, key, me->exname);
+                if (ret)
+                        goto fail;
+
+                idx++;
+        }
+
+        gf_log (GF_MNT, GF_LOG_DEBUG, "Updated rmtab with %d entries", idx);
+
+        close (fd);
+        if (gf_store_rename_tmppath (sh))
+                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to overwrite rwtab %s",
+                        sh->path);
+
+        return;
+
+fail:
+        gf_log (GF_MNT, GF_LOG_ERROR, "Failed to update %s", sh->path);
+        close (fd);
+        gf_store_unlink_tmppath (sh);
+}
+
+/* Read the rmtab into a clean ms->mountlist.
+ */
+static void
+mount_read_rmtab (struct mount3_state *ms)
+{
+        gf_store_handle_t       *sh = NULL;
+        struct nfs_state        *nfs = NULL;
+        int                     ret;
+
+        nfs = (struct nfs_state *)ms->nfsx->private;
+
+        ret = gf_store_handle_new (nfs->rmtab, &sh);
+        if (ret) {
+                gf_log (GF_MNT, GF_LOG_WARNING, "Failed to open '%s'",
+                        nfs->rmtab);
+                return;
+        }
+
+        if (gf_store_lock (sh)) {
+                gf_log (GF_MNT, GF_LOG_WARNING, "Failed to lock '%s'",
+                        nfs->rmtab);
+                goto out;
+        }
+
+        __mount_read_rmtab (sh, &ms->mountlist, _gf_false);
+        gf_store_unlock (sh);
+
+out:
+        gf_store_handle_destroy (sh);
+}
+
+/* Write the ms->mountlist to the rmtab.
+ *
+ * The rmtab could be empty, or it can exists and have been updated by a
+ * different storage server without our knowing.
+ *
+ * 1. takes the store_handle lock on the current rmtab
+ *    - blocks if an other storage server rewrites the rmtab at the same time
+ * 2. [if new_rmtab] takes the store_handle lock on the new rmtab
+ * 3. reads/merges the entries from the current rmtab
+ * 4. [if new_rmtab] reads/merges the entries from the new rmtab
+ * 5. [if new_rmtab] writes the new rmtab
+ * 6. [if not new_rmtab] writes the current rmtab
+ * 7  [if new_rmtab] replaces nfs->rmtab to point to the new location
+ * 8. [if new_rmtab] releases the store_handle lock of the new rmtab
+ * 9. releases the store_handle lock of the old rmtab
+ */
+void
+mount_rewrite_rmtab (struct mount3_state *ms, char *new_rmtab)
+{
+        gf_store_handle_t       *sh = NULL, *nsh = NULL;
+        struct nfs_state        *nfs = NULL;
+        int                     ret;
+        char                    *rmtab = NULL;
+
+        nfs = (struct nfs_state *)ms->nfsx->private;
+
+        ret = gf_store_handle_new (nfs->rmtab, &sh);
+        if (ret) {
+                gf_log (GF_MNT, GF_LOG_WARNING, "Failed to open '%s'",
+                        nfs->rmtab);
+                return;
+        }
+
+        if (gf_store_lock (sh)) {
+                gf_log (GF_MNT, GF_LOG_WARNING, "Not rewriting '%s'",
+                        nfs->rmtab);
+                goto free_sh;
+        }
+
+        if (new_rmtab) {
+                ret = gf_store_handle_new (new_rmtab, &nsh);
+                if (ret) {
+                        gf_log (GF_MNT, GF_LOG_WARNING, "Failed to open '%s'",
+                                new_rmtab);
+                        goto unlock_sh;
+                }
+
+                if (gf_store_lock (nsh)) {
+                        gf_log (GF_MNT, GF_LOG_WARNING, "Not rewriting '%s'",
+                                new_rmtab);
+                        goto free_nsh;
+                }
+        }
+
+        /* always read the currently used rmtab */
+        __mount_read_rmtab (sh, &ms->mountlist, _gf_true);
+
+        if (new_rmtab) {
+                /* read the new rmtab and write changes to the new location */
+                __mount_read_rmtab (nsh, &ms->mountlist, _gf_true);
+                __mount_rewrite_rmtab (ms, nsh);
+
+                /* replace the nfs->rmtab reference to the new rmtab */
+                rmtab = gf_strdup(new_rmtab);
+                if (rmtab == NULL) {
+                        gf_log (GF_MNT, GF_LOG_ERROR, "Out of memory, keeping "
+                                "%s as rmtab", nfs->rmtab);
+                } else {
+                        GF_FREE (nfs->rmtab);
+                        nfs->rmtab = new_rmtab;
+                }
+
+                gf_store_unlock (nsh);
+        } else {
+                /* rewrite the current (unchanged location) rmtab */
+                __mount_rewrite_rmtab (ms, sh);
+        }
+
+free_nsh:
+        if (new_rmtab)
+                gf_store_handle_destroy (nsh);
+unlock_sh:
+        gf_store_unlock (sh);
+free_sh:
+        gf_store_handle_destroy (sh);
+}
+
+/* Add a new NFS-client to the ms->mountlist and update the rmtab if we can.
+ *
+ * A NFS-client will only be removed from the ms->mountlist in case the
+ * NFS-client sends a unmount request. It is possible that a NFS-client
+ * crashed/rebooted had network loss or something else prevented the NFS-client
+ * to unmount cleanly. In this case, a duplicate entry would be added to the
+ * ms->mountlist, which is wrong and we should prevent.
+ *
+ * It is fully acceptible that the ms->mountlist is not 100% correct, this is a
+ * common issue for all(?) NFS-servers.
+ */
 int
 mnt3svc_update_mountlist (struct mount3_state *ms, rpcsvc_request_t *req,
                           char  *expname)
 {
         struct mountentry       *me = NULL;
+        struct mountentry       *cur = NULL;
         int                     ret = -1;
+        char                    *colon = NULL;
+        struct nfs_state        *nfs = NULL;
+        gf_store_handle_t       *sh = NULL;
 
         if ((!ms) || (!req) || (!expname))
                 return -1;
@@ -188,21 +517,62 @@ mnt3svc_update_mountlist (struct mount3_state *ms, rpcsvc_request_t *req,
         if (!me)
                 return -1;
 
-        strcpy (me->exname, expname);
+        nfs = (struct nfs_state *)ms->nfsx->private;
+
+        ret = gf_store_handle_new (nfs->rmtab, &sh);
+        if (ret) {
+                gf_log (GF_MNT, GF_LOG_WARNING, "Failed to open '%s'",
+                        nfs->rmtab);
+                goto free_err;
+        }
+
+        strncpy (me->exname, expname, MNTPATHLEN);
+
         INIT_LIST_HEAD (&me->mlist);
         /* Must get the IP or hostname of the client so we
          * can map it into the mount entry.
          */
-        ret = nfs_rpcsvc_conn_peername (req->conn, me->hostname, MNTPATHLEN);
+        ret = rpcsvc_transport_peername (req->trans, me->hostname, MNTPATHLEN);
         if (ret == -1)
-                goto free_err;
+                goto free_err2;
 
+        colon = strrchr (me->hostname, ':');
+        if (colon) {
+                *colon = '\0';
+        }
         LOCK (&ms->mountlock);
         {
+                /* in case locking fails, we just don't write the rmtab */
+                if (gf_store_lock (sh)) {
+                        gf_log (GF_MNT, GF_LOG_WARNING, "Failed to lock '%s'"
+                                ", changes will not be written", nfs->rmtab);
+                } else {
+                        __mount_read_rmtab (sh, &ms->mountlist, _gf_false);
+                }
+
+                /* do not add duplicates */
+                list_for_each_entry (cur, &ms->mountlist, mlist) {
+                        if (!strcmp(cur->hostname, me->hostname) &&
+                            !strcmp(cur->exname, me->exname)) {
+                                GF_FREE (me);
+                                goto dont_add;
+                        }
+                }
                 list_add_tail (&me->mlist, &ms->mountlist);
+
+                /* only write the rmtab in case it was locked */
+                if (gf_store_locked_local (sh))
+                        __mount_rewrite_rmtab (ms, sh);
         }
+dont_add:
+        if (gf_store_locked_local (sh))
+                gf_store_unlock (sh);
+
         UNLOCK (&ms->mountlock);
 
+free_err2:
+        gf_store_handle_destroy (sh);
+
 free_err:
         if (ret == -1)
                 GF_FREE (me);
@@ -221,6 +591,7 @@ __mnt3_get_volume_id (struct mount3_state *ms, xlator_t *mntxl,
         if ((!ms) || (!mntxl))
                 return ret;
 
+        LOCK (&ms->mountlock);
         list_for_each_entry (exp, &ms->exportlist, explist) {
                 if (exp->vol == mntxl) {
                         uuid_copy (volumeid, exp->volumeid);
@@ -230,6 +601,7 @@ __mnt3_get_volume_id (struct mount3_state *ms, xlator_t *mntxl,
         }
 
 out:
+        UNLOCK (&ms->mountlock);
         return ret;
 }
 
@@ -250,7 +622,7 @@ mnt3svc_lookup_mount_cbk (call_frame_t *frame, void  *cookie,
         rpcsvc_t                *svc = NULL;
         xlator_t                *mntxl = NULL;
         uuid_t                  volumeid = {0, };
-        char                    fhstr[1024];
+        char                    fhstr[1024], *path = NULL;
 
         req = (rpcsvc_request_t *)frame->local;
 
@@ -258,20 +630,29 @@ mnt3svc_lookup_mount_cbk (call_frame_t *frame, void  *cookie,
                 return -1;
 
         mntxl = (xlator_t *)cookie;
-        ms = (struct mount3_state *)nfs_rpcsvc_request_program_private (req);
+        ms = (struct mount3_state *)rpcsvc_request_program_private (req);
         if (!ms) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "mount state not found");
                 op_ret = -1;
                 op_errno = EINVAL;
         }
 
-        if (op_ret == -1)
+        if (op_ret == -1) {
+                gf_log (GF_NFS, GF_LOG_ERROR, "error=%s", strerror (op_errno));
                 status = mnt3svc_errno_to_mnterr (op_errno);
-
+        }
         if (status != MNT3_OK)
                 goto xmit_res;
 
-        mnt3svc_update_mountlist (ms, req, mntxl->name);
+        path = GF_CALLOC (PATH_MAX, sizeof (char), gf_nfs_mt_char);
+        if (!path) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Out of memory");
+                goto xmit_res;
+        }
+
+        snprintf (path, PATH_MAX, "/%s", mntxl->name);
+        mnt3svc_update_mountlist (ms, req, path);
+        GF_FREE (path);
         if (gf_nfs_dvm_off (nfs_state (ms->nfsx))) {
                 fh = nfs3_fh_build_indexed_root_fh (ms->nfsx->children, mntxl);
                 goto xmit_res;
@@ -281,13 +662,13 @@ mnt3svc_lookup_mount_cbk (call_frame_t *frame, void  *cookie,
         fh = nfs3_fh_build_uuid_root_fh (volumeid);
 
 xmit_res:
-        nfs3_fh_to_str (&fh, fhstr);
+        nfs3_fh_to_str (&fh, fhstr, sizeof (fhstr));
         gf_log (GF_MNT, GF_LOG_DEBUG, "MNT reply: fh %s, status: %d", fhstr,
                 status);
         if (op_ret == 0) {
-                svc = nfs_rpcsvc_request_service (req);
-                autharrlen = nfs_rpcsvc_auth_array (svc, mntxl->name, autharr,
-                                                    10);
+                svc = rpcsvc_request_service (req);
+                autharrlen = rpcsvc_auth_array (svc, mntxl->name, autharr,
+                                                10);
         }
 
         res = mnt3svc_set_mountres3 (status, &fh, autharr, autharrlen);
@@ -302,7 +683,7 @@ int
 mnt3_match_dirpath_export (char *expname, char *dirpath)
 {
         int     ret = 0;
-        int     dlen = 0;
+        size_t  dlen;
 
         if ((!expname) || (!dirpath))
                 return 0;
@@ -313,7 +694,7 @@ mnt3_match_dirpath_export (char *expname, char *dirpath)
          * compare.
          */
         dlen = strlen (dirpath);
-        if (dirpath [dlen - 1] == '/')
+        if (dlen && dirpath [dlen - 1] == '/')
                 dirpath [dlen - 1] = '\0';
 
         if (dirpath[0] != '/')
@@ -337,11 +718,11 @@ mnt3svc_mount_inode (rpcsvc_request_t *req, struct mount3_state *ms,
         if ((!req) || (!xl) || (!ms) || (!exportinode))
                 return ret;
 
-        ret = nfs_inode_loc_fill (exportinode, &exportloc);
+        ret = nfs_inode_loc_fill (exportinode, &exportloc, NFS_RESOLVE_EXIST);
         if (ret < 0) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Loc fill failed for export inode"
-                        ": ino %"PRIu64", volume: %s",
-                        exportinode->ino, xl->name);
+                        ": gfid %s, volume: %s",
+                        uuid_utoa (exportinode->gfid), xl->name);
                 goto err;
         }
 
@@ -378,7 +759,7 @@ mnt3svc_volume_mount (rpcsvc_request_t *req, struct mount3_state *ms,
         rootgfid[15] = 1;
         exportinode = inode_find (exp->vol->itable, rootgfid);
         if (!exportinode) {
-                gf_log (GF_MNT, GF_LOG_ERROR, "Faild to get root inode");
+                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to get root inode");
                 ret = -ENOENT;
                 goto err;
         }
@@ -398,9 +779,10 @@ err:
  * we need to strip out the volume name first.
  */
 char *
-__volume_subdir (char *dirpath)
+__volume_subdir (char *dirpath, char **volname)
 {
         char    *subdir = NULL;
+        int     volname_len = 0;
 
         if (!dirpath)
                 return NULL;
@@ -409,7 +791,22 @@ __volume_subdir (char *dirpath)
                 dirpath++;
 
         subdir = index (dirpath, (int)'/');
+        if (!subdir)
+                goto out;
+
+        if (!volname)
+                goto out;
 
+        if (!*volname)
+                goto out;
+
+        /* subdir points to the first / after the volume name while dirpath
+         * points to the first char of the volume name.
+         */
+        volname_len = subdir - dirpath;
+        strncpy (*volname, dirpath, volname_len);
+        *(*volname + volname_len) = '\0';
+out:
         return subdir;
 }
 
@@ -429,8 +826,8 @@ mnt3_resolve_state_wipe (mnt3_resolve_t *mres)
 /* Sets up the component argument to contain the next component in the path and
  * sets up path as an absolute path starting from the next component.
  */
-char *
-__setup_next_component (char *path, char *component)
+static char *
+setup_next_component (char *path, size_t plen, char *component, size_t clen)
 {
         char    *comp = NULL;
         char    *nextcomp = NULL;
@@ -438,7 +835,7 @@ __setup_next_component (char *path, char *component)
         if ((!path) || (!component))
                 return NULL;
 
-        strcpy (component, path);
+        strncpy (component, path, clen);
         comp = index (component, (int)'/');
         if (!comp)
                 goto err;
@@ -446,7 +843,7 @@ __setup_next_component (char *path, char *component)
         comp++;
         nextcomp = index (comp, (int)'/');
         if (nextcomp) {
-                strcpy (path, nextcomp);
+                strncpy (path, nextcomp, plen);
                 *nextcomp = '\0';
         } else
                 path[0] = '\0';
@@ -471,24 +868,26 @@ __mnt3_resolve_export_subdir_comp (mnt3_resolve_t *mres)
         char            *nextcomp = NULL;
         int             ret = -EFAULT;
         nfs_user_t      nfu = {0, };
-        char            gfidstr[512];
+        uuid_t          gfid = {0, };
 
         if (!mres)
                 return ret;
 
-        nextcomp = __setup_next_component (mres->remainingdir, dupsubdir);
+        nextcomp = setup_next_component (mres->remainingdir,
+                                         sizeof (mres->remainingdir),
+                                         dupsubdir, sizeof (dupsubdir));
         if (!nextcomp)
                 goto err;
 
         /* Wipe the contents of the previous component */
+        uuid_copy (gfid, mres->resolveloc.inode->gfid);
         nfs_loc_wipe (&mres->resolveloc);
-        ret = nfs_entry_loc_fill (mres->exp->vol->itable,
-                                  mres->resolveloc.inode->gfid, nextcomp,
+        ret = nfs_entry_loc_fill (mres->exp->vol->itable, gfid, nextcomp,
                                   &mres->resolveloc, NFS_RESOLVE_CREATE);
         if ((ret < 0) && (ret != -2)) {
-                uuid_unparse (mres->resolveloc.inode->gfid, gfidstr);
                 gf_log (GF_MNT, GF_LOG_ERROR, "Failed to resolve and create "
-                        "inode: parent gfid %s, entry %s", gfidstr, nextcomp);
+                        "inode: parent gfid %s, entry %s",
+                        uuid_utoa (gfid), nextcomp);
                 ret = -EFAULT;
                 goto err;
         }
@@ -516,10 +915,13 @@ mnt3_resolve_subdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         rpcsvc_t                *svc = NULL;
         mountres3               res = {0, };
         xlator_t                *mntxl = NULL;
+        char                    *path = NULL;
 
         mres = frame->local;
         mntxl = (xlator_t *)cookie;
         if (op_ret == -1) {
+                gf_log (GF_NFS, GF_LOG_ERROR, "path=%s (%s)",
+                        mres->resolveloc.path, strerror (op_errno));
                 mntstat = mnt3svc_errno_to_mnterr (op_errno);
                 goto err;
         }
@@ -531,22 +933,30 @@ mnt3_resolve_subdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         if (strlen (mres->remainingdir) <= 0) {
                 op_ret = -1;
                 mntstat = MNT3_OK;
+                path = GF_CALLOC (PATH_MAX, sizeof (char), gf_nfs_mt_char);
+                if (!path) {
+                        gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation "
+                                "failed");
+                        goto err;
+                }
+                snprintf (path, PATH_MAX, "/%s%s", mres->exp->vol->name,
+                         mres->resolveloc.path);
                 mnt3svc_update_mountlist (mres->mstate, mres->req,
-                                          mres->exp->expname);
-                goto err;
+                                          path);
+                GF_FREE (path);
+        } else {
+                mres->parentfh = fh;
+                op_ret = __mnt3_resolve_export_subdir_comp (mres);
+                if (op_ret < 0)
+                        mntstat = mnt3svc_errno_to_mnterr (-op_ret);
         }
-
-        mres->parentfh = fh;
-        op_ret = __mnt3_resolve_export_subdir_comp (mres);
-        if (op_ret < 0)
-                mntstat = mnt3svc_errno_to_mnterr (-op_ret);
 err:
         if (op_ret == -1) {
                 gf_log (GF_MNT, GF_LOG_DEBUG, "Mount reply status: %d",
                         mntstat);
-                svc = nfs_rpcsvc_request_service (mres->req);
-                autharrlen = nfs_rpcsvc_auth_array (svc, mntxl->name, autharr,
-                                                    10);
+                svc = rpcsvc_request_service (mres->req);
+                autharrlen = rpcsvc_auth_array (svc, mntxl->name, autharr,
+                                                10);
 
                 res = mnt3svc_set_mountres3 (mntstat, &fh, autharr, autharrlen);
                 mnt3svc_submit_reply (mres->req, (void *)&res,
@@ -571,7 +981,7 @@ err:
  * of the exported directory can be built.
  */
 int
-__mnt3_resolve_export_subdir (mnt3_resolve_t *mres)
+__mnt3_resolve_subdir (mnt3_resolve_t *mres)
 {
         char            dupsubdir[MNTPATHLEN];
         char            *firstcomp = NULL;
@@ -582,7 +992,9 @@ __mnt3_resolve_export_subdir (mnt3_resolve_t *mres)
         if (!mres)
                 return ret;
 
-        firstcomp = __setup_next_component (mres->remainingdir, dupsubdir);
+        firstcomp = setup_next_component (mres->remainingdir,
+                                          sizeof (mres->remainingdir),
+                                          dupsubdir, sizeof (dupsubdir));
         if (!firstcomp)
                 goto err;
 
@@ -605,21 +1017,152 @@ err:
 }
 
 
+/**
+ * This function will verify if the client is allowed to mount
+ * the directory or not. Client's IP address will be compared with
+ * allowed IP list or range present in mnt3_export structure.
+ *
+ * @param req - RPC request. This structure contains client's IP address.
+ * @param export - mnt3_export structure. Contains allowed IP list/range.
+ *
+ * @return 0 - on Success and -EACCES on failure.
+ */
 int
-mnt3_resolve_export_subdir (rpcsvc_request_t *req, struct mount3_state *ms,
-                            struct mnt3_export *exp)
+mnt3_verify_auth (rpcsvc_request_t *req, struct mnt3_export *export)
+{
+        int                     retvalue = -EACCES;
+        int                     ret = 0;
+        int                     shiftbits = 0;
+        uint32_t                ipv4netmask = 0;
+        uint32_t                routingprefix = 0;
+        struct host_auth_spec   *host = NULL;
+        struct sockaddr_in      *client_addr = NULL;
+        struct sockaddr_in      *allowed_addr = NULL;
+        struct addrinfo         *allowed_addrinfo = NULL;
+
+        /* Sanity check */
+        if ((NULL == req) ||
+            (NULL == req->trans) ||
+            (NULL == export) ||
+            (NULL == export->hostspec)) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Invalid argument");
+                return retvalue;
+        }
+
+        host = export->hostspec;
+
+
+        /* Client's IP address. */
+        client_addr = (struct sockaddr_in *)(&(req->trans->peerinfo.sockaddr));
+
+        /* Try to see if the client IP matches the allowed IP list.*/
+        while (NULL != host){
+                GF_ASSERT (host->host_addr);
+
+                if (NULL != allowed_addrinfo) {
+                        freeaddrinfo (allowed_addrinfo);
+                        allowed_addrinfo = NULL;
+                }
+
+                /* Get the addrinfo for the allowed host (host_addr). */
+                ret = getaddrinfo (host->host_addr,
+                                NULL,
+                                NULL,
+                                &allowed_addrinfo);
+                if (0 != ret){
+                        gf_log (GF_MNT, GF_LOG_ERROR, "getaddrinfo: %s\n",
+                                gai_strerror (ret));
+                        host = host->next;
+
+                        /* Failed to get IP addrinfo. Continue to check other
+                         * allowed IPs in the list.
+                         */
+                        continue;
+                }
+
+                allowed_addr = (struct sockaddr_in *)(allowed_addrinfo->ai_addr);
+
+                if (NULL == allowed_addr) {
+                        gf_log (GF_MNT, GF_LOG_ERROR, "Invalid structure");
+                        break;
+                }
+
+                if (AF_INET == allowed_addr->sin_family){
+                        if (IPv4_ADDR_SIZE < host->routeprefix) {
+                                gf_log (GF_MNT, GF_LOG_ERROR, "invalid IP "
+                                        "configured for export-dir AUTH");
+                                host = host->next;
+                                continue;
+                        }
+
+                        /* -1 means no route prefix is provided. In this case
+                         * the IP should be an exact match. Which is same as
+                         * providing a route prefix of IPv4_ADDR_SIZE.
+                         */
+                        if (-1 == host->routeprefix) {
+                                routingprefix = IPv4_ADDR_SIZE;
+                        } else {
+                                routingprefix = host->routeprefix;
+                        }
+
+                        /* Create a mask from the routing prefix. User provided
+                         * CIDR address is split into IP address (host_addr) and
+                         * routing prefix (routeprefix). This CIDR address may
+                         * denote a single, distinct interface address or the
+                         * beginning address of an entire network.
+                         *
+                         * e.g. the IPv4 block 192.168.100.0/24 represents the
+                         * 256 IPv4 addresses from 192.168.100.0 to
+                         * 192.168.100.255.
+                         * Therefore to check if an IP matches 192.168.100.0/24
+                         * we should mask the IP with FFFFFF00 and compare it
+                         * with host address part of CIDR.
+                         */
+                        shiftbits = IPv4_ADDR_SIZE - routingprefix;
+                        ipv4netmask = 0xFFFFFFFFUL << shiftbits;
+
+                        /* Mask both the IPs and then check if they match
+                         * or not. */
+                        if (COMPARE_IPv4_ADDRS (allowed_addr,
+                                                client_addr,
+                                                ipv4netmask)){
+                                retvalue = 0;
+                                break;
+                        }
+                }
+
+                /* Client IP didn't match the allowed IP.
+                 * Check with the next allowed IP.*/
+               host = host->next;
+        }
+
+        if (NULL != allowed_addrinfo) {
+               freeaddrinfo (allowed_addrinfo);
+        }
+
+        return retvalue;
+}
+
+int
+mnt3_resolve_subdir (rpcsvc_request_t *req, struct mount3_state *ms,
+                     struct mnt3_export *exp, char *subdir)
 {
         mnt3_resolve_t  *mres = NULL;
-        char            *volume_subdir = NULL;
         int             ret = -EFAULT;
         struct nfs3_fh  pfh = GF_NFS3FH_STATIC_INITIALIZER;
 
-        if ((!req) || (!ms) || (!exp))
+        if ((!req) || (!ms) || (!exp) || (!subdir))
                 return ret;
 
-        volume_subdir = __volume_subdir (exp->expname);
-        if (!volume_subdir)
-                goto err;
+        /* Need to check AUTH */
+        if (NULL != exp->hostspec) {
+                ret = mnt3_verify_auth (req, exp);
+                if (0 != ret) {
+                        gf_log (GF_MNT,GF_LOG_ERROR,
+                                        "AUTH verification failed");
+                        return ret;
+                }
+        }
 
         mres = GF_CALLOC (1, sizeof (mnt3_resolve_t), gf_nfs_mt_mnt3_resolve);
         if (!mres) {
@@ -630,14 +1173,14 @@ mnt3_resolve_export_subdir (rpcsvc_request_t *req, struct mount3_state *ms,
         mres->exp = exp;
         mres->mstate = ms;
         mres->req = req;
-        strcpy (mres->remainingdir, volume_subdir);
+        strncpy (mres->remainingdir, subdir, MNTPATHLEN);
         if (gf_nfs_dvm_off (nfs_state (ms->nfsx)))
                 pfh = nfs3_fh_build_indexed_root_fh (mres->mstate->nfsx->children, mres->exp->vol);
         else
                 pfh = nfs3_fh_build_uuid_root_fh (exp->volumeid);
 
         mres->parentfh = pfh;
-        ret = __mnt3_resolve_export_subdir (mres);
+        ret = __mnt3_resolve_subdir (mres);
         if (ret < 0) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Failed to resolve export dir: %s"
                         , mres->exp->expname);
@@ -650,6 +1193,32 @@ err:
 
 
 int
+mnt3_resolve_export_subdir (rpcsvc_request_t *req, struct mount3_state *ms,
+                            struct mnt3_export *exp)
+{
+        char            *volume_subdir = NULL;
+        int             ret = -EFAULT;
+
+        if ((!req) || (!ms) || (!exp))
+                return ret;
+
+        volume_subdir = __volume_subdir (exp->expname, NULL);
+        if (!volume_subdir)
+                goto err;
+
+        ret = mnt3_resolve_subdir (req, ms, exp, volume_subdir);
+        if (ret < 0) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to resolve export dir: %s"
+                        , exp->expname);
+                goto err;
+        }
+
+err:
+        return ret;
+}
+
+
+int
 mnt3svc_mount (rpcsvc_request_t *req, struct mount3_state *ms,
                struct mnt3_export *exp)
 {
@@ -679,6 +1248,7 @@ mnt3_mntpath_to_export (struct mount3_state *ms, char *dirpath)
         if ((!ms) || (!dirpath))
                 return NULL;
 
+        LOCK (&ms->mountlock);
         list_for_each_entry (exp, &ms->exportlist, explist) {
 
                 /* Search for the an exact match with the volume */
@@ -692,6 +1262,7 @@ mnt3_mntpath_to_export (struct mount3_state *ms, char *dirpath)
 
         gf_log (GF_MNT, GF_LOG_DEBUG, "Export not found");
 foundexp:
+        UNLOCK (&ms->mountlock);
         return found;
 }
 
@@ -700,24 +1271,37 @@ int
 mnt3_check_client_net (struct mount3_state *ms, rpcsvc_request_t *req,
                        xlator_t *targetxl)
 {
-        rpcsvc_t        *svc = NULL;
-        int             ret = -1;
+
+        rpcsvc_t                *svc = NULL;
+        rpc_transport_t         *trans = NULL;
+        struct sockaddr_storage sastorage = {0,};
+        char                    peer[RPCSVC_PEER_STRLEN] = {0,};
+        int                     ret = -1;
 
         if ((!ms) || (!req) || (!targetxl))
                 return -1;
 
-        svc = nfs_rpcsvc_request_service (req);
-        ret = nfs_rpcsvc_conn_peer_check (svc->options, targetxl->name,
-                                          nfs_rpcsvc_request_conn (req));
+        svc = rpcsvc_request_service (req);
+
+        trans = rpcsvc_request_transport (req);
+        ret = rpcsvc_transport_peeraddr (trans, peer, RPCSVC_PEER_STRLEN,
+                                         &sastorage, sizeof (sastorage));
+        if (ret != 0) {
+                gf_log (GF_MNT, GF_LOG_WARNING, "Failed to get peer addr: %s",
+                        gai_strerror (ret));
+        }
+
+        ret = rpcsvc_auth_check (svc, targetxl->name, trans);
         if (ret == RPCSVC_AUTH_REJECT) {
-                gf_log (GF_MNT, GF_LOG_TRACE, "Peer not allowed");
+                gf_log (GF_MNT, GF_LOG_INFO, "Peer %s  not allowed", peer);
                 goto err;
         }
 
-        ret = nfs_rpcsvc_conn_privport_check (svc, targetxl->name,
-                                              nfs_rpcsvc_request_conn (req));
+        ret = rpcsvc_transport_privport_check (svc, targetxl->name,
+                                               rpcsvc_request_transport (req));
         if (ret == RPCSVC_AUTH_REJECT) {
-                gf_log (GF_MNT, GF_LOG_TRACE, "Unprivileged port not allowed");
+                gf_log (GF_MNT, GF_LOG_INFO, "Peer %s rejected. Unprivileged "
+                        "port not allowed", peer);
                 goto err;
         }
 
@@ -728,6 +1312,80 @@ err:
 
 
 int
+mnt3_parse_dir_exports (rpcsvc_request_t *req, struct mount3_state *ms,
+                        char *subdir)
+{
+        char                    volname[1024];
+        struct mnt3_export      *exp = NULL;
+        char                    *volname_ptr = NULL;
+        int                     ret = -1;
+
+        if ((!ms) || (!subdir))
+                return -1;
+
+        volname_ptr = volname;
+        subdir = __volume_subdir (subdir, &volname_ptr);
+        if (!subdir)
+                goto err;
+
+        exp = mnt3_mntpath_to_export (ms, volname);
+        if (!exp)
+                goto err;
+
+        ret = mnt3_resolve_subdir (req, ms, exp, subdir);
+        if (ret < 0) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to resolve export dir: %s"
+                        , subdir);
+                goto err;
+        }
+
+err:
+        return ret;
+}
+
+
+int
+mnt3_find_export (rpcsvc_request_t *req, char *path, struct mnt3_export **e)
+{
+        int                     ret = -EFAULT;
+        struct mount3_state     *ms = NULL;
+        struct mnt3_export      *exp = NULL;
+
+        if ((!req) || (!path) || (!e))
+                return -1;
+
+        ms = (struct mount3_state *) rpcsvc_request_program_private (req);
+        if (!ms) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Mount state not present");
+                rpcsvc_request_seterr (req, SYSTEM_ERR);
+                goto err;
+        }
+
+        gf_log (GF_MNT, GF_LOG_DEBUG, "dirpath: %s", path);
+        exp = mnt3_mntpath_to_export (ms, path);
+        if (exp) {
+                ret = 0;
+                *e = exp;
+                goto err;
+        }
+
+        if (!gf_mnt3_export_dirs(ms)) {
+                ret = -1;
+                goto err;
+        }
+
+        ret = mnt3_parse_dir_exports (req, ms, path);
+        if (ret == 0) {
+                ret = -2;
+                goto err;
+        }
+
+err:
+        return ret;
+}
+
+
+int
 mnt3svc_mnt (rpcsvc_request_t *req)
 {
         struct iovec            pvec = {0, };
@@ -736,49 +1394,55 @@ mnt3svc_mnt (rpcsvc_request_t *req)
         struct mount3_state     *ms = NULL;
         mountstat3              mntstat = MNT3ERR_SERVERFAULT;
         struct mnt3_export      *exp = NULL;
+        struct nfs_state        *nfs = NULL;
 
         if (!req)
                 return -1;
 
         pvec.iov_base = path;
         pvec.iov_len = MNTPATHLEN;
-        ret = xdr_to_mountpath (pvec, req->msg);
+        ret = xdr_to_mountpath (pvec, req->msg[0]);
         if (ret == -1) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Failed to decode args");
-                nfs_rpcsvc_request_seterr (req, GARBAGE_ARGS);
+                rpcsvc_request_seterr (req, GARBAGE_ARGS);
                 goto rpcerr;
         }
 
-        ms = (struct mount3_state *)nfs_rpcsvc_request_program_private (req);
+        ms = (struct mount3_state *)rpcsvc_request_program_private (req);
         if (!ms) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Mount state not present");
-                nfs_rpcsvc_request_seterr (req, SYSTEM_ERR);
+                rpcsvc_request_seterr (req, SYSTEM_ERR);
                 ret = -1;
                 goto rpcerr;
         }
 
         ret = 0;
+        nfs = (struct nfs_state *)ms->nfsx->private;
         gf_log (GF_MNT, GF_LOG_DEBUG, "dirpath: %s", path);
-        exp = mnt3_mntpath_to_export (ms, path);
-        if (!exp) {
+        ret = mnt3_find_export (req, path, &exp);
+        if (ret == -2) {
+                ret = 0;
+                goto rpcerr;
+        } else if (ret < 0) {
                 ret = -1;
                 mntstat = MNT3ERR_NOENT;
                 goto mnterr;
         }
 
-        if ((is_nfs_subvolume_disabled (nfs_state (ms->nfsx), exp->vol))) {
-                gf_log (GF_MNT, GF_LOG_ERROR, "Volume is disabled: %s",
+        if (!nfs_subvolume_started (nfs, exp->vol)) {
+                gf_log (GF_MNT, GF_LOG_DEBUG, "Volume %s not started",
                         exp->vol->name);
-                ret = RPCSVC_ACTOR_ERROR;
-                mntstat = MNT3ERR_INVAL;
+                ret = -1;
+                mntstat = MNT3ERR_NOENT;
                 goto mnterr;
         }
 
         ret = mnt3_check_client_net (ms, req, exp->vol);
-        if (ret == -1) {
+        if (ret == RPCSVC_AUTH_REJECT) {
                 mntstat = MNT3ERR_ACCES;
                 gf_log (GF_MNT, GF_LOG_DEBUG, "Client mount not allowed");
-                goto rpcerr;
+                ret = -1;
+                goto mnterr;
         }
 
         ret = mnt3svc_mount (req, ms, exp);
@@ -804,8 +1468,7 @@ mnt3svc_null (rpcsvc_request_t *req)
                 gf_log (GF_MNT, GF_LOG_ERROR, "Got NULL request!");
                 return 0;
         }
-
-        nfs_rpcsvc_submit_generic (req, dummyvec, NULL);
+        rpcsvc_submit_generic (req, &dummyvec, 1,  NULL, 0, NULL);
         return 0;
 }
 
@@ -823,6 +1486,9 @@ __build_mountlist (struct mount3_state *ms, int *count)
         if ((!ms) || (!count))
                 return NULL;
 
+        /* read rmtab, other peers might have updated it */
+        mount_read_rmtab(ms);
+
         *count = 0;
         gf_log (GF_MNT, GF_LOG_DEBUG, "Building mount list:");
         list_for_each_entry (me, &ms->mountlist, mlist) {
@@ -833,6 +1499,8 @@ __build_mountlist (struct mount3_state *ms, int *count)
                                 " failed");
                         goto free_list;
                 }
+                if (!first)
+                        first = mlist;
 
                 mlist->ml_directory = GF_CALLOC (namelen + 2, sizeof (char),
                                                  gf_nfs_mt_char);
@@ -842,8 +1510,7 @@ __build_mountlist (struct mount3_state *ms, int *count)
                         goto free_list;
                 }
 
-                strcpy (mlist->ml_directory, "/");
-                strcat (mlist->ml_directory, me->exname);
+                strcpy (mlist->ml_directory, me->exname);
 
                 namelen = strlen (me->hostname);
                 mlist->ml_hostname = GF_CALLOC (namelen + 2, sizeof (char),
@@ -864,9 +1531,6 @@ __build_mountlist (struct mount3_state *ms, int *count)
                 } else
                         prev = mlist;
 
-                if (!first)
-                        first = mlist;
-
                 (*count)++;
         }
 
@@ -911,19 +1575,19 @@ mnt3svc_dump (rpcsvc_request_t *req)
         if (!req)
                 return -1;
 
-        ms = (struct mount3_state *)nfs_rpcsvc_request_program_private (req);
+        ms = (struct mount3_state *)rpcsvc_request_program_private (req);
         if (!ms) {
-                nfs_rpcsvc_request_seterr (req, SYSTEM_ERR);
+                rpcsvc_request_seterr (req, SYSTEM_ERR);
                 goto rpcerr;
         }
 
         sfunc = (mnt3_serializer)xdr_serialize_mountlist;
         mlist = mnt3svc_build_mountlist (ms, &ret);
-        arg = mlist;
-        
+        arg = &mlist;
+
         if (!mlist) {
                 if (ret != 0) {
-                        nfs_rpcsvc_request_seterr (req, SYSTEM_ERR);
+                        rpcsvc_request_seterr (req, SYSTEM_ERR);
                         ret = -1;
                         goto rpcerr;
                 } else {
@@ -943,67 +1607,71 @@ rpcerr:
 
 
 int
-__mnt3svc_umount (struct mount3_state *ms, char *dirpath, char *hostname)
+mnt3svc_umount (struct mount3_state *ms, char *dirpath, char *hostname)
 {
         struct mountentry       *me = NULL;
-        char                    *exname = NULL;
         int                     ret = -1;
+        gf_store_handle_t       *sh = NULL;
+        struct nfs_state        *nfs = NULL;
 
         if ((!ms) || (!dirpath) || (!hostname))
                 return -1;
 
-        if (list_empty (&ms->mountlist))
-                return 0;
-
-        if (dirpath[0] == '/')
-                exname = dirpath+1;
-        else
-                exname = dirpath;
+        nfs = (struct nfs_state *)ms->nfsx->private;
 
-        list_for_each_entry (me, &ms->mountlist, mlist) {
-               if ((strcmp (me->exname, exname) == 0) &&
-                    (strcmp (me->hostname, hostname) == 0)) {
-                       ret = 0;
-                       break;
-               }
+        ret = gf_store_handle_new (nfs->rmtab, &sh);
+        if (ret) {
+                gf_log (GF_MNT, GF_LOG_WARNING, "Failed to open '%s'",
+                        nfs->rmtab);
+                return 0;
         }
 
-        /* Need this check here because at the end of the search me might still
-         * be pointing to the last entry, which may not be the one we're
-         * looking for.
-         */
-        if (ret == -1)  {/* Not found in list. */
-                gf_log (GF_MNT, GF_LOG_DEBUG, "Export not found");
-                goto ret;
+        ret = gf_store_lock (sh);
+        if (ret) {
+                goto out_free;
         }
 
-        if (!me)
-                goto ret;
+        LOCK (&ms->mountlock);
+        {
+                __mount_read_rmtab (sh, &ms->mountlist, _gf_false);
+                if (list_empty (&ms->mountlist)) {
+                        ret = 0;
+                        goto out_unlock;
+                }
 
-        gf_log (GF_MNT, GF_LOG_DEBUG, "Unmounting: dir %s, host: %s",
-                me->exname, me->hostname);
-        list_del (&me->mlist);
-        GF_FREE (me);
-        ret = 0;
-ret:
-        return ret;
-}
+                ret = -1;
+                list_for_each_entry (me, &ms->mountlist, mlist) {
+                       if ((strcmp (me->exname, dirpath) == 0) &&
+                            (strcmp (me->hostname, hostname) == 0)) {
+                               ret = 0;
+                               break;
+                       }
+                }
 
+                /* Need this check here because at the end of the search me
+                 * might still be pointing to the last entry, which may not be
+                 * the one we're looking for.
+                 */
+                if (ret == -1)  {/* Not found in list. */
+                        gf_log (GF_MNT, GF_LOG_TRACE, "Export not found");
+                        goto out_unlock;
+                }
 
+                if (!me)
+                        goto out_unlock;
 
-int
-mnt3svc_umount (struct mount3_state *ms, char *dirpath, char *hostname)
-{
-        int ret = -1;
-        if ((!ms) || (!dirpath) || (!hostname))
-                return -1;
+                gf_log (GF_MNT, GF_LOG_DEBUG, "Unmounting: dir %s, host: %s",
+                        me->exname, me->hostname);
 
-        LOCK (&ms->mountlock);
-        {
-               ret = __mnt3svc_umount (ms, dirpath, hostname);
+                list_del (&me->mlist);
+                GF_FREE (me);
+                __mount_rewrite_rmtab (ms, sh);
         }
+out_unlock:
         UNLOCK (&ms->mountlock);
-
+        gf_store_unlock (sh);
+out_free:
+        gf_store_handle_destroy (sh);
         return ret;
 }
 
@@ -1017,6 +1685,7 @@ mnt3svc_umnt (rpcsvc_request_t *req)
         int                     ret = -1;
         struct mount3_state     *ms = NULL;
         mountstat3              mstat = MNT3_OK;
+        char                    *colon = NULL;
 
         if (!req)
                 return -1;
@@ -1024,56 +1693,43 @@ mnt3svc_umnt (rpcsvc_request_t *req)
         /* Remove the mount point from the exports list. */
         pvec.iov_base = dirpath;
         pvec.iov_len = MNTPATHLEN;
-        ret = xdr_to_mountpath (pvec, req->msg);;
+        ret = xdr_to_mountpath (pvec, req->msg[0]);
         if (ret == -1) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Failed decode args");
-                nfs_rpcsvc_request_seterr (req, GARBAGE_ARGS);
+                rpcsvc_request_seterr (req, GARBAGE_ARGS);
                 goto rpcerr;
         }
 
-        ms = (struct mount3_state *)nfs_rpcsvc_request_program_private (req);
+        ms = (struct mount3_state *)rpcsvc_request_program_private (req);
         if (!ms) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Mount state not present");
-                nfs_rpcsvc_request_seterr (req, SYSTEM_ERR);
+                rpcsvc_request_seterr (req, SYSTEM_ERR);
                 ret = -1;
                 goto rpcerr;
         }
 
-        ret = nfs_rpcsvc_conn_peername (req->conn, hostname, MNTPATHLEN);
+        ret = rpcsvc_transport_peername (req->trans, hostname, MNTPATHLEN);
         if (ret != 0) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Failed to get remote name: %s",
                         gai_strerror (ret));
-                goto try_umount_with_addr;
-        }
-
-        gf_log (GF_MNT, GF_LOG_DEBUG, "dirpath: %s, hostname: %s", dirpath,
-                hostname);
-        ret = mnt3svc_umount (ms, dirpath, hostname);
-
-        /* Unmount succeeded with the given hostname. */
-        if (ret == 0)
-                goto snd_reply;
-
-try_umount_with_addr:
-        if (ret != 0)
-                ret = nfs_rpcsvc_conn_peeraddr (req->conn, hostname, MNTPATHLEN,
-                                                NULL, 0);
-
-        if (ret != 0) {
-                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to get remote addr: %s",
-                        gai_strerror (ret));
-                nfs_rpcsvc_request_seterr (req, SYSTEM_ERR);
                 goto rpcerr;
         }
 
+        colon = strrchr (hostname, ':');
+        if (colon) {
+                *colon= '\0';
+        }
         gf_log (GF_MNT, GF_LOG_DEBUG, "dirpath: %s, hostname: %s", dirpath,
                 hostname);
         ret = mnt3svc_umount (ms, dirpath, hostname);
-        if (ret == -1)
-                mstat = MNT3ERR_INVAL;
 
-        ret = 0;
-snd_reply:
+        if (ret == -1) {
+                ret = 0;
+                mstat = MNT3ERR_NOENT;
+        }
+        /* FIXME: also take care of the corner case where the
+         * client was resolvable at mount but not at the umount - vice-versa.
+         */
         mnt3svc_submit_reply (req, &mstat,
                               (mnt3_serializer)xdr_serialize_mountstat3);
 
@@ -1086,11 +1742,15 @@ int
 __mnt3svc_umountall (struct mount3_state *ms)
 {
         struct mountentry       *me = NULL;
+        struct mountentry       *tmp = NULL;
 
         if (!ms)
                 return -1;
 
-        list_for_each_entry (me, &ms->mountlist, mlist) {
+        if (list_empty (&ms->mountlist))
+                return 0;
+
+        list_for_each_entry_safe (me, tmp, &ms->mountlist, mlist) {
                 list_del (&me->mlist);
                 GF_FREE (me);
         }
@@ -1119,18 +1779,17 @@ mnt3svc_umountall (struct mount3_state *ms)
 int
 mnt3svc_umntall (rpcsvc_request_t *req)
 {
-        int                     ret = -1;
+        int                     ret = RPCSVC_ACTOR_ERROR;
         struct mount3_state     *ms = NULL;
         mountstat3              mstat = MNT3_OK;
 
         if (!req)
-                return -1;
+                return ret;
 
-        ms = (struct mount3_state *)nfs_rpcsvc_request_program_private (req);
+        ms = (struct mount3_state *)rpcsvc_request_program_private (req);
         if (!ms) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Mount state not present");
-                nfs_rpcsvc_request_seterr (req, SYSTEM_ERR);
-                ret = -1;
+                rpcsvc_request_seterr (req, SYSTEM_ERR);
                 goto rpcerr;
         }
 
@@ -1138,6 +1797,7 @@ mnt3svc_umntall (rpcsvc_request_t *req)
         mnt3svc_submit_reply (req, &mstat,
                               (mnt3_serializer)xdr_serialize_mountstat3);
 
+        ret = RPCSVC_ACTOR_SUCCESS;
 rpcerr:
         return ret;
 }
@@ -1153,11 +1813,24 @@ mnt3_xlchildren_to_exports (rpcsvc_t *svc, struct mount3_state *ms)
         int                     ret = -1;
         char                    *addrstr = NULL;
         struct mnt3_export      *ent = NULL;
+        struct nfs_state        *nfs = NULL;
 
         if ((!ms) || (!svc))
                 return NULL;
 
+        nfs = (struct nfs_state *)ms->nfsx->private;
+        if (!nfs)
+                return NULL;
+
+        LOCK (&ms->mountlock);
         list_for_each_entry(ent, &ms->exportlist, explist) {
+
+                /* If volume is not started yet, do not list it for tools like
+                 * showmount.
+                 */
+                if (!nfs_subvolume_started (nfs, ent->vol))
+                        continue;
+
                 namelen = strlen (ent->expname) + 1;
                 elist = GF_CALLOC (1, sizeof (*elist), gf_nfs_mt_exportnode);
                 if (!elist) {
@@ -1165,7 +1838,8 @@ mnt3_xlchildren_to_exports (rpcsvc_t *svc, struct mount3_state *ms)
                                 " failed");
                         goto free_list;
                 }
-
+                 if (!first)
+                         first = elist;
                 elist->ex_dir = GF_CALLOC (namelen + 2, sizeof (char),
                                            gf_nfs_mt_char);
                 if (!elist->ex_dir) {
@@ -1173,16 +1847,10 @@ mnt3_xlchildren_to_exports (rpcsvc_t *svc, struct mount3_state *ms)
                                 " failed");
                         goto free_list;
                 }
-
                 strcpy (elist->ex_dir, ent->expname);
 
-                addrstr = nfs_rpcsvc_volume_allowed (svc->options,
-                                                     ent->vol->name);
-                if (addrstr)
-                        addrstr = gf_strdup (addrstr);
-                else
-                        addrstr = gf_strdup ("No Access");
-
+                addrstr = rpcsvc_volume_allowed (svc->options,
+                                                 ent->vol->name);
                 elist->ex_groups = GF_CALLOC (1, sizeof (struct groupnode),
                                               gf_nfs_mt_groupnode);
                 if (!elist->ex_groups) {
@@ -1190,21 +1858,29 @@ mnt3_xlchildren_to_exports (rpcsvc_t *svc, struct mount3_state *ms)
                                 " failed");
                         goto free_list;
                 }
+                /*This check has to be done after checking
+                 * elist->ex_groups allocation check to avoid resource leak;
+                */
+                if (addrstr)
+                        addrstr = gf_strdup (addrstr);
+                else
+                        addrstr = gf_strdup ("No Access");
 
+                if (!addrstr) {
+                        goto free_list;
+                }
                 elist->ex_groups->gr_name = addrstr;
                 if (prev) {
                         prev->ex_next = elist;
                         prev = elist;
                 } else
                         prev = elist;
-
-                if (!first)
-                        first = elist;
         }
 
         ret = 0;
 
 free_list:
+        UNLOCK (&ms->mountlock);
         if (ret == -1) {
                 xdr_free_exports_list (first);
                 first = NULL;
@@ -1224,21 +1900,25 @@ mnt3svc_export (rpcsvc_request_t *req)
         if (!req)
                 return -1;
 
-        ms = (struct mount3_state *)nfs_rpcsvc_request_program_private (req);
+        ms = (struct mount3_state *)rpcsvc_request_program_private (req);
         if (!ms) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "mount state not found");
-                nfs_rpcsvc_request_seterr (req, SYSTEM_ERR);
+                rpcsvc_request_seterr (req, SYSTEM_ERR);
                 goto err;
         }
 
         /* Using the children translator names, build the export list */
-        elist = mnt3_xlchildren_to_exports (nfs_rpcsvc_request_service (req),
+        elist = mnt3_xlchildren_to_exports (rpcsvc_request_service (req),
                                             ms);
+        /* Do not return error when exports list is empty. An exports list can
+         * be empty when no subvolumes have come up. No point returning error
+         * and confusing the user.
         if (!elist) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Failed to build exports list");
                 nfs_rpcsvc_request_seterr (req, SYSTEM_ERR);
                 goto err;
         }
+        */
 
         /* Note how the serializer is passed to the generic reply function. */
         mnt3svc_submit_reply (req, &elist,
@@ -1250,7 +1930,233 @@ err:
         return ret;
 }
 
+/* just declaring, definition is way down below */
+rpcsvc_program_t        mnt3prog;
+
+/* nfs3_rootfh used by mount3udp thread needs to access mount3prog.private
+ * directly as we don't have nfs xlator pointer to dereference it. But thats OK
+ */
+
+struct nfs3_fh *
+nfs3_rootfh (char* path)
+{
+        struct mount3_state     *ms = NULL;
+        struct nfs3_fh          *fh = NULL;
+        struct mnt3_export      *exp = NULL;
+        inode_t                 *inode = NULL;
+        char                    *tmp = NULL;
+
+        ms = mnt3prog.private;
+        exp = mnt3_mntpath_to_export (ms, path);
+        if (exp == NULL)
+                goto err;
+
+        tmp = (char *)path;
+        tmp = strchr (tmp, '/');
+        if (tmp == NULL)
+                tmp = "/";
+
+        inode = inode_from_path (exp->vol->itable, tmp);
+        if (inode == NULL)
+                goto err;
+
+        fh = GF_CALLOC (1, sizeof(*fh), gf_nfs_mt_nfs3_fh);
+        if (fh == NULL)
+                goto err;
+        nfs3_build_fh (inode, exp->volumeid, fh);
+
+err:
+        if (inode)
+                inode_unref (inode);
+        return fh;
+}
+
+int
+mount3udp_add_mountlist (char *host, dirpath *expname)
+{
+        struct mountentry       *me = NULL;
+        struct mount3_state     *ms = NULL;
+        char                    *export = NULL;
+
+        ms = mnt3prog.private;
+        me = GF_CALLOC (1, sizeof (*me), gf_nfs_mt_mountentry);
+        if (!me)
+                return -1;
+        export = (char *)expname;
+        while (*export == '/')
+                export++;
+
+        strncpy (me->exname, export, MNTPATHLEN);
+        strncpy (me->hostname, host, MNTPATHLEN);
+        INIT_LIST_HEAD (&me->mlist);
+        LOCK (&ms->mountlock);
+        {
+                list_add_tail (&me->mlist, &ms->mountlist);
+                mount_rewrite_rmtab(ms, NULL);
+        }
+        UNLOCK (&ms->mountlock);
+        return 0;
+}
+
+int
+mount3udp_delete_mountlist (char *hostname, dirpath *expname)
+{
+        struct mount3_state     *ms = NULL;
+        char                    *export = NULL;
+
+        ms = mnt3prog.private;
+        export = (char *)expname;
+        while (*export == '/')
+                export++;
+        mnt3svc_umount (ms, export, hostname);
+        return 0;
+}
+
+/**
+ * This function will parse the hostip (IP addres, IP range, or hostname)
+ * and fill the host_auth_spec structure.
+ *
+ * @param hostspec - struct host_auth_spec
+ * @param hostip   - IP address, IP range (CIDR format) or hostname
+ *
+ * @return 0 - on success and -1 on failure
+ */
+int
+mnt3_export_fill_hostspec (struct host_auth_spec* hostspec, const char* hostip)
+{
+        char *ipdupstr = NULL;
+        char *savptr = NULL;
+        char *ip = NULL;
+        char *token = NULL;
+        int  ret = -1;
+
+        /* Create copy of the string so that the source won't change
+         */
+        ipdupstr = gf_strdup (hostip);
+        if (NULL == ipdupstr) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed");
+                goto err;
+        }
+
+        ip = strtok_r (ipdupstr, "/", &savptr);
+        hostspec->host_addr = gf_strdup (ip);
+        if (NULL == hostspec->host_addr) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed");
+                goto err;
+        }
+
+        /* Check if the IP is in <IP address> / <Range> format.
+         * If yes, then strip the range and store it separately.
+         */
+        token = strtok_r (NULL, "/", &savptr);
+
+        if (NULL == token) {
+              hostspec->routeprefix = -1;
+        } else {
+              hostspec->routeprefix = atoi (token);
+        }
+
+        // success
+        ret = 0;
+err:
+        if (NULL != ipdupstr) {
+                GF_FREE (ipdupstr);
+        }
+        return ret;
+}
+
+
+/**
+ * This function will parse the AUTH parameter passed along with
+ * "export-dir" option. If AUTH parameter is present then it will be
+ * stripped from exportpath and stored in mnt3_export (exp) structure.
+ *
+ * @param exp - mnt3_export structure. Holds information needed for mount.
+ * @param exportpath - Value of "export-dir" key. Holds both export path
+ *                     and AUTH parameter for the path.
+ *                     exportpath format: <abspath>[(hostdesc[|hostspec|...])]
+ *
+ * @return This function will return 0 on success and -1 on failure.
+ */
+int
+mnt3_export_parse_auth_param (struct mnt3_export* exp, char* exportpath)
+{
+        char *token = NULL;
+        char *savPtr = NULL;
+        char *hostip = NULL;
+        struct host_auth_spec *host = NULL;
+        int ret = 0;
+
+        /* Using exportpath directly in strtok_r because we want
+         * to strip off AUTH parameter from exportpath. */
+        token = strtok_r (exportpath, "(", &savPtr);
+
+        /* Get the next token, which will be the AUTH parameter. */
+        token = strtok_r (NULL, ")", &savPtr);
+
+        if (NULL == token) {
+                /* If AUTH is not present then we should return success. */
+                return 0;
+        }
 
+        /* Free any previously allocated hostspec structure. */
+        if (NULL != exp->hostspec) {
+                GF_FREE (exp->hostspec);
+                exp->hostspec = NULL;
+        }
+
+        exp->hostspec = GF_CALLOC (1,
+                        sizeof (*(exp->hostspec)),
+                        gf_nfs_mt_auth_spec);
+        if (NULL == exp->hostspec){
+                gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed");
+                return -1;
+        }
+
+        /* AUTH parameter can have multiple entries. For each entry
+         * a host_auth_spec structure is created. */
+        host = exp->hostspec;
+
+        hostip = strtok_r (token, "|", &savPtr);
+
+        /* Parse all AUTH parameters separated by '|' */
+        while (NULL != hostip){
+                ret = mnt3_export_fill_hostspec (host, hostip);
+                if (0 != ret) {
+                        gf_log(GF_MNT, GF_LOG_WARNING,
+                                        "Failed to parse hostspec: %s", hostip);
+                        goto err;
+                }
+
+                hostip = strtok_r (NULL, "|", &savPtr);
+                if (NULL == hostip) {
+                        break;
+                }
+
+                host->next = GF_CALLOC (1, sizeof (*(host)),
+                                gf_nfs_mt_auth_spec);
+                if (NULL == host->next){
+                        gf_log (GF_MNT,GF_LOG_ERROR,
+                                        "Memory allocation failed");
+                        goto err;
+                }
+                host = host->next;
+        }
+
+        /* In case of success return from here */
+        return 0;
+err:
+        /* In case of failure free up hostspec structure.  */
+        FREE_HOSTSPEC (exp);
+
+        return -1;
+}
+
+/**
+ * exportpath will also have AUTH options (ip address, subnet address or
+ * hostname) mentioned.
+ * exportpath format: <abspath>[(hostdesc[|hostspec|...])]
+ */
 struct mnt3_export *
 mnt3_init_export_ent (struct mount3_state *ms, xlator_t *xl, char *exportpath,
                       uuid_t volumeid)
@@ -1268,6 +2174,20 @@ mnt3_init_export_ent (struct mount3_state *ms, xlator_t *xl, char *exportpath,
                 return NULL;
         }
 
+        if (NULL != exportpath) {
+                /* If exportpath is not NULL then we should check if AUTH
+                 * parameter is present or not. If AUTH parameter is present
+                 * then it will be stripped and stored in mnt3_export (exp)
+                 * structure.
+                 */
+                if (0 != mnt3_export_parse_auth_param (exp, exportpath)){
+                        gf_log (GF_MNT, GF_LOG_ERROR,
+                                                "Failed to parse auth param");
+                        goto err;
+                }
+        }
+
+
         INIT_LIST_HEAD (&exp->explist);
         if (exportpath)
                 alloclen = strlen (xl->name) + 2 + strlen (exportpath);
@@ -1277,8 +2197,6 @@ mnt3_init_export_ent (struct mount3_state *ms, xlator_t *xl, char *exportpath,
         exp->expname = GF_CALLOC (alloclen, sizeof (char), gf_nfs_mt_char);
         if (!exp->expname) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Memory allocation failed");
-                GF_FREE (exp);
-                exp = NULL;
                 goto err;
         }
 
@@ -1294,14 +2212,27 @@ mnt3_init_export_ent (struct mount3_state *ms, xlator_t *xl, char *exportpath,
                 exp->exptype = MNT3_EXPTYPE_VOLUME;
                 ret = snprintf (exp->expname, alloclen, "/%s", xl->name);
         }
-
+        if (ret < 0) {
+                gf_log (xl->name, GF_LOG_ERROR,
+                        "Failed to set the export name");
+                goto err;
+        }
         /* Just copy without discrimination, we'll determine whether to
          * actually use it when a mount request comes in and a file handle
          * needs to be built.
          */
         uuid_copy (exp->volumeid, volumeid);
         exp->vol = xl;
+
+        /* On success we should return from here*/
+        return exp;
 err:
+        /* On failure free exp and it's members.*/
+        if (NULL != exp) {
+                mnt3_export_free (exp);
+                exp = NULL;
+        }
+
         return exp;
 }
 
@@ -1341,8 +2272,7 @@ __mnt3_init_volume_direxports (struct mount3_state *ms, xlator_t *xlator,
 
         ret = 0;
 err:
-        if (dupopt)
-                GF_FREE (dupopt);
+        GF_FREE (dupopt);
 
         return ret;
 }
@@ -1463,8 +2393,11 @@ __mnt3_init_volume_export (struct mount3_state *ms, dict_t *opts)
                 goto err;
         }
 
-        gf_string2boolean (optstr, &boolt);
-        ret = 0;
+        ret = gf_string2boolean (optstr, &boolt);
+        if (ret < 0) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to convert"
+                        " string to boolean");
+        }
 
 err:
         if (boolt == _gf_false) {
@@ -1480,6 +2413,49 @@ err:
 
 
 int
+__mnt3_init_dir_export (struct mount3_state *ms, dict_t *opts)
+{
+        int                     ret = -1;
+        char                    *optstr  = NULL;
+        /* On by default. */
+        gf_boolean_t            boolt = _gf_true;
+
+        if ((!ms) || (!opts))
+                return -1;
+
+        if (!dict_get (opts, "nfs3.export-dirs")) {
+                ret = 0;
+                goto err;
+        }
+
+        ret = dict_get_str (opts, "nfs3.export-dirs", &optstr);
+        if (ret < 0) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to read option: "
+                        "nfs3.export-dirs");
+                ret = -1;
+                goto err;
+        }
+
+        ret = gf_string2boolean (optstr, &boolt);
+        if (ret < 0) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Failed to convert"
+                        " string to boolean");
+         }
+
+err:
+        if (boolt == _gf_false) {
+                gf_log (GF_MNT, GF_LOG_TRACE, "Dir exports disabled");
+                ms->export_dirs = 0;
+        } else {
+                gf_log (GF_MNT, GF_LOG_TRACE, "Dir exports enabled");
+                ms->export_dirs = 1;
+        }
+
+        return ret;
+}
+
+
+int
 mnt3_init_options (struct mount3_state *ms, dict_t *options)
 {
         xlator_list_t   *volentry = NULL;
@@ -1489,6 +2465,7 @@ mnt3_init_options (struct mount3_state *ms, dict_t *options)
                 return -1;
 
         __mnt3_init_volume_export (ms, options);
+        __mnt3_init_dir_export (ms, options);
         volentry = ms->nfsx->children;
         while (volentry) {
                 gf_log (GF_MNT, GF_LOG_TRACE, "Initing options for: %s",
@@ -1538,49 +2515,118 @@ mnt3_init_state (xlator_t *nfsx)
         return ms;
 }
 
+int
+mount_init_state (xlator_t *nfsx)
+{
+        int              ret = -1;
+        struct nfs_state *nfs = NULL;
+
+        if (!nfsx)
+                goto out;
+
+        nfs = (struct nfs_state *)nfs_state (nfsx);
+        /*Maintaining global state for MOUNT1 and MOUNT3*/
+        nfs->mstate =  mnt3_init_state (nfsx);
+        if (!nfs->mstate) {
+                gf_log (GF_NFS, GF_LOG_ERROR, "Failed to allocate"
+                        "mount state");
+                goto out;
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
 rpcsvc_actor_t  mnt3svc_actors[MOUNT3_PROC_COUNT] = {
-        {"NULL", MOUNT3_NULL, mnt3svc_null, NULL, NULL},
-        {"MNT", MOUNT3_MNT, mnt3svc_mnt, NULL, NULL},
-        {"DUMP", MOUNT3_DUMP, mnt3svc_dump, NULL, NULL},
-        {"UMNT", MOUNT3_UMNT, mnt3svc_umnt, NULL, NULL},
-        {"UMNTALL", MOUNT3_UMNTALL, mnt3svc_umntall, NULL, NULL},
-        {"EXPORT", MOUNT3_EXPORT, mnt3svc_export, NULL, NULL}
+        {"NULL",    MOUNT3_NULL,    mnt3svc_null,    NULL, 0, DRC_NA},
+        {"MNT",     MOUNT3_MNT,     mnt3svc_mnt,     NULL, 0, DRC_NA},
+        {"DUMP",    MOUNT3_DUMP,    mnt3svc_dump,    NULL, 0, DRC_NA},
+        {"UMNT",    MOUNT3_UMNT,    mnt3svc_umnt,    NULL, 0, DRC_NA},
+        {"UMNTALL", MOUNT3_UMNTALL, mnt3svc_umntall, NULL, 0, DRC_NA},
+        {"EXPORT",  MOUNT3_EXPORT,  mnt3svc_export,  NULL, 0, DRC_NA}
 };
 
 
 
 /* Static init parts are assigned here, dynamic ones are done in
  * mnt3svc_init and mnt3_init_state.
+ * Making MOUNT3 a synctask so that the blocking DNS calls during rpc auth
+ * gets offloaded to syncenv, keeping the main/poll thread unblocked
  */
 rpcsvc_program_t        mnt3prog = {
                         .progname       = "MOUNT3",
                         .prognum        = MOUNT_PROGRAM,
                         .progver        = MOUNT_V3,
                         .progport       = GF_MOUNTV3_PORT,
-                        .progaddrfamily = AF_INET,
-                        .proghost       = NULL,
                         .actors         = mnt3svc_actors,
                         .numactors      = MOUNT3_PROC_COUNT,
+                        .min_auth       = AUTH_NULL,
+                        .synctask       = _gf_true,
 };
 
 
+
 rpcsvc_program_t *
 mnt3svc_init (xlator_t *nfsx)
 {
         struct mount3_state     *mstate = NULL;
+        struct nfs_state        *nfs = NULL;
+        dict_t                  *options = NULL;
+        char                    *portstr = NULL;
+        int                      ret = -1;
+        pthread_t                udp_thread;
 
-        if (!nfsx)
+        if (!nfsx || !nfsx->private)
                 return NULL;
 
+        nfs = (struct nfs_state *)nfsx->private;
+
         gf_log (GF_MNT, GF_LOG_DEBUG, "Initing Mount v3 state");
-        mstate = mnt3_init_state (nfsx);
+        mstate = (struct mount3_state *)nfs->mstate;
         if (!mstate) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Mount v3 state init failed");
                 goto err;
         }
 
         mnt3prog.private = mstate;
+        options = dict_new ();
 
+        ret = gf_asprintf (&portstr, "%d", GF_MOUNTV3_PORT);
+        if (ret == -1)
+                goto err;
+
+        ret = dict_set_dynstr (options, "transport.socket.listen-port", portstr);
+        if (ret == -1)
+                goto err;
+        ret = dict_set_str (options, "transport-type", "socket");
+        if (ret == -1) {
+                gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error");
+                goto err;
+        }
+
+        if (nfs->allow_insecure) {
+                ret = dict_set_str (options, "rpc-auth-allow-insecure", "on");
+                if (ret == -1) {
+                        gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error");
+                        goto err;
+                }
+                ret = dict_set_str (options, "rpc-auth.ports.insecure", "on");
+                if (ret == -1) {
+                        gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error");
+                        goto err;
+                }
+        }
+
+        ret= rpcsvc_create_listeners (nfs->rpcsvc, options, nfsx->name);
+        if (ret == -1) {
+                gf_log (GF_NFS, GF_LOG_ERROR, "Unable to create listeners");
+                dict_unref (options);
+                goto err;
+        }
+
+        if (nfs->mount_udp) {
+                pthread_create (&udp_thread, NULL, mount3udp_thread, NULL);
+        }
         return &mnt3prog;
 err:
         return NULL;
@@ -1588,12 +2634,12 @@ err:
 
 
 rpcsvc_actor_t  mnt1svc_actors[MOUNT1_PROC_COUNT] = {
-        {"NULL", MOUNT1_NULL, mnt3svc_null, NULL, NULL},
-        {{0}, },
-        {"DUMP", MOUNT1_DUMP, mnt3svc_dump, NULL, NULL},
-        {"UMNT", MOUNT1_UMNT, mnt3svc_umnt, NULL, NULL},
-        {{0}, },
-        {"EXPORT", MOUNT1_EXPORT, mnt3svc_export, NULL, NULL}
+        {"NULL",    MOUNT1_NULL,    mnt3svc_null,   NULL, 0, DRC_NA},
+        {"MNT",     MOUNT1_MNT,     NULL,           NULL, 0, DRC_NA },
+        {"DUMP",    MOUNT1_DUMP,    mnt3svc_dump,   NULL, 0, DRC_NA},
+        {"UMNT",    MOUNT1_UMNT,    mnt3svc_umnt,   NULL, 0, DRC_NA},
+        {"UMNTALL", MOUNT1_UMNTALL, NULL,           NULL, 0, DRC_NA},
+        {"EXPORT",  MOUNT1_EXPORT,  mnt3svc_export, NULL, 0, DRC_NA}
 };
 
 rpcsvc_program_t        mnt1prog = {
@@ -1601,10 +2647,10 @@ rpcsvc_program_t        mnt1prog = {
                         .prognum        = MOUNT_PROGRAM,
                         .progver        = MOUNT_V1,
                         .progport       = GF_MOUNTV1_PORT,
-                        .progaddrfamily = AF_INET,
-                        .proghost       = NULL,
                         .actors         = mnt1svc_actors,
                         .numactors      = MOUNT1_PROC_COUNT,
+                        .min_auth       = AUTH_NULL,
+                        .synctask       = _gf_true,
 };
 
 
@@ -1612,12 +2658,18 @@ rpcsvc_program_t *
 mnt1svc_init (xlator_t *nfsx)
 {
         struct mount3_state     *mstate = NULL;
+        struct nfs_state        *nfs = NULL;
+        dict_t                  *options = NULL;
+        char                    *portstr = NULL;
+        int                      ret = -1;
 
-        if (!nfsx)
+        if (!nfsx || !nfsx->private)
                 return NULL;
 
+        nfs = (struct nfs_state *)nfsx->private;
+
         gf_log (GF_MNT, GF_LOG_DEBUG, "Initing Mount v1 state");
-        mstate = mnt3_init_state (nfsx);
+        mstate = (struct mount3_state *)nfs->mstate;
         if (!mstate) {
                 gf_log (GF_MNT, GF_LOG_ERROR, "Mount v3 state init failed");
                 goto err;
@@ -1625,9 +2677,83 @@ mnt1svc_init (xlator_t *nfsx)
 
         mnt1prog.private = mstate;
 
+        options = dict_new ();
+
+        ret = gf_asprintf (&portstr, "%d", GF_MOUNTV1_PORT);
+        if (ret == -1)
+                goto err;
+
+        ret = dict_set_dynstr (options, "transport.socket.listen-port", portstr);
+        if (ret == -1)
+                goto err;
+        ret = dict_set_str (options, "transport-type", "socket");
+        if (ret == -1) {
+                gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error");
+                goto err;
+        }
+
+        if (nfs->allow_insecure) {
+                ret = dict_set_str (options, "rpc-auth-allow-insecure", "on");
+                if (ret == -1) {
+                        gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error");
+                        goto err;
+                }
+                ret = dict_set_str (options, "rpc-auth.ports.insecure", "on");
+                if (ret == -1) {
+                        gf_log (GF_NFS, GF_LOG_ERROR, "dict_set_str error");
+                        goto err;
+                }
+        }
+
+        ret = rpcsvc_create_listeners (nfs->rpcsvc, options, nfsx->name);
+        if (ret == -1) {
+                gf_log (GF_NFS, GF_LOG_ERROR, "Unable to create listeners");
+                dict_unref (options);
+                goto err;
+        }
+
         return &mnt1prog;
 err:
         return NULL;
 }
 
+int
+mount_reconfigure_state (xlator_t *nfsx, dict_t *options)
+{
+        int                    ret      = -1;
+        struct nfs_state       *nfs     = NULL;
+        struct mount3_state    *ms      = NULL;
+        struct mnt3_export     *exp     = NULL;
+        struct mnt3_export     *texp  = NULL;
+
+        if ((!nfsx) || (!options))
+                return (-1);
+
+        nfs = (struct nfs_state *)nfs_state (nfsx);
+        if (!nfs)
+                return (-1);
+
+        ms = nfs->mstate;
+        if (!ms)
+                return (-1);
+
+        /*
+         * Free() up the old export list. mnt3_init_options() will
+         * rebuild the export list from scratch. Do it with locking
+         * to avoid unnecessary race conditions.
+         */
+        LOCK (&ms->mountlock);
+        list_for_each_entry_safe (exp, texp, &ms->exportlist, explist) {
+                list_del (&exp->explist);
+                mnt3_export_free (exp);
+        }
+        ret = mnt3_init_options (ms, options);
+        UNLOCK (&ms->mountlock);
 
+        if (ret < 0) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "Options reconfigure failed");
+                return (-1);
+        }
+
+        return (0);
+}
diff --git a/xlators/nfs/server/src/mount3.h b/xlators/nfs/server/src/mount3.h
index 16b89f6da..7fc16ed57 100644
--- a/xlators/nfs/server/src/mount3.h
+++ b/xlators/nfs/server/src/mount3.h
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _MOUNT3_H_
@@ -50,7 +41,16 @@ mnt3svc_init (xlator_t *nfsx);
 extern rpcsvc_program_t *
 mnt1svc_init (xlator_t *nfsx);
 
-/* Data structureused to store the list of mounts points currently
+extern int
+mount_init_state (xlator_t *nfsx);
+
+extern int
+mount_reconfigure_state (xlator_t *nfsx, dict_t *options);
+
+void
+mount_rewrite_rmtab (struct mount3_state *ms, char *new_rmtab);
+
+/* Data structure used to store the list of mounts points currently
  * in use by NFS clients.
  */
 struct mountentry {
@@ -65,6 +65,13 @@ struct mountentry {
 #define MNT3_EXPTYPE_VOLUME     1
 #define MNT3_EXPTYPE_DIR        2
 
+/* Structure to hold export-dir AUTH parameter */
+struct host_auth_spec {
+        char                    *host_addr;    /* Allowed IP or host name */
+        int                     routeprefix;   /* Routing prefix */
+        struct host_auth_spec   *next;         /* Pointer to next AUTH struct */
+};
+
 struct mnt3_export {
         struct list_head        explist;
 
@@ -72,6 +79,11 @@ struct mnt3_export {
          * is exported or the subdirectory in the volume.
          */
         char                    *expname;
+        /*
+         * IP address, hostname or subnets who are allowed to connect to expname
+         * subvolume or subdirectory
+         */
+        struct host_auth_spec*  hostspec;
         xlator_t                *vol;
         int                     exptype;
 
@@ -98,9 +110,11 @@ struct mount3_state {
         gf_lock_t               mountlock;
 
         /* Set to 0 if exporting full volumes is disabled. On by default. */
-        int                     export_volumes;
+        gf_boolean_t            export_volumes;
+        gf_boolean_t            export_dirs;
 };
 
+#define gf_mnt3_export_dirs(mst)        ((mst)->export_dirs)
 
 struct mount3_resolve_state {
         struct mnt3_export      *exp;
diff --git a/xlators/nfs/server/src/mount3udp_svc.c b/xlators/nfs/server/src/mount3udp_svc.c
new file mode 100644
index 000000000..fb59e282c
--- /dev/null
+++ b/xlators/nfs/server/src/mount3udp_svc.c
@@ -0,0 +1,189 @@
+/*
+  Copyright (c) 2012 Gluster, Inc. <http://www.gluster.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#include "xdr-nfs3.h"
+#include "logging.h"
+#include "mem-pool.h"
+#include "nfs-mem-types.h"
+#include "mount3.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <rpc/pmap_clnt.h>
+#include <string.h>
+#include <memory.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+
+
+extern struct nfs3_fh* nfs3_rootfh (char *dp);
+extern mountres3 mnt3svc_set_mountres3 (mountstat3 stat, struct nfs3_fh *fh,
+                                        int *authflavor, u_int aflen);
+extern int
+mount3udp_add_mountlist (char *host, dirpath *expname);
+
+extern int
+mount3udp_delete_mountlist (char *host, dirpath *expname);
+
+
+/* only this thread will use this, no locking needed */
+char mnthost[INET_ADDRSTRLEN+1];
+
+mountres3 *
+mountudpproc3_mnt_3_svc(dirpath **dpp, struct svc_req *req)
+{
+        struct mountres3        *res = NULL;
+        int                     *autharr = NULL;
+        struct nfs3_fh          *fh = NULL;
+        char                    *tmp = NULL;
+
+        tmp = (char *)*dpp;
+        while (*tmp == '/')
+                tmp++;
+        fh = nfs3_rootfh (tmp);
+        if (fh == NULL) {
+                gf_log (GF_MNT, GF_LOG_DEBUG, "unable to get fh for %s", tmp);
+                goto err;
+        }
+
+        res = GF_CALLOC (1, sizeof(*res), gf_nfs_mt_mountres3);
+        if (res == NULL) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "unable to allocate memory");
+                goto err;
+        }
+        autharr = GF_CALLOC (1, sizeof(*autharr), gf_nfs_mt_int);
+        if (autharr == NULL) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "unable to allocate memory");
+                goto err;
+        }
+        autharr[0] = AUTH_UNIX;
+        *res = mnt3svc_set_mountres3 (MNT3_OK, fh, autharr, 1);
+        mount3udp_add_mountlist (mnthost, *dpp);
+        return res;
+
+ err:
+        GF_FREE (fh);
+        GF_FREE (res);
+        GF_FREE (autharr);
+        return NULL;
+}
+
+mountstat3 *
+mountudpproc3_umnt_3_svc(dirpath **dp, struct svc_req *req)
+{
+        mountstat3 *stat = NULL;
+
+        stat = GF_CALLOC (1, sizeof(mountstat3), gf_nfs_mt_mountstat3);
+        if (stat == NULL) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "unable to allocate memory");
+                return NULL;
+        }
+        *stat = MNT3_OK;
+        mount3udp_delete_mountlist (mnthost, *dp);
+        return stat;
+}
+
+static void
+mountudp_program_3(struct svc_req *rqstp, register SVCXPRT *transp)
+{
+        union {
+                dirpath mountudpproc3_mnt_3_arg;
+        } argument;
+        char                    *result = NULL;
+        xdrproc_t               _xdr_argument = NULL, _xdr_result = NULL;
+        char *(*local)(char *, struct svc_req *) = NULL;
+        mountres3               *res = NULL;
+        struct sockaddr_in      *sin = NULL;
+
+        sin = svc_getcaller (transp);
+        inet_ntop (AF_INET, &sin->sin_addr, mnthost, INET_ADDRSTRLEN+1);
+
+        switch (rqstp->rq_proc) {
+        case NULLPROC:
+                (void) svc_sendreply (transp, (xdrproc_t) xdr_void,
+                                      (char *)NULL);
+                return;
+
+        case MOUNT3_MNT:
+                _xdr_argument = (xdrproc_t) xdr_dirpath;
+                _xdr_result = (xdrproc_t) xdr_mountres3;
+                local = (char *(*)(char *,
+                                   struct svc_req *)) mountudpproc3_mnt_3_svc;
+                break;
+
+        case MOUNT3_UMNT:
+                _xdr_argument = (xdrproc_t) xdr_dirpath;
+                _xdr_result = (xdrproc_t) xdr_mountstat3;
+                local = (char *(*)(char *,
+                                   struct svc_req *)) mountudpproc3_umnt_3_svc;
+                break;
+
+        default:
+                svcerr_noproc (transp);
+                return;
+        }
+        memset ((char *)&argument, 0, sizeof (argument));
+        if (!svc_getargs (transp, (xdrproc_t) _xdr_argument,
+                          (caddr_t) &argument)) {
+                svcerr_decode (transp);
+                return;
+        }
+        result = (*local)((char *)&argument, rqstp);
+        if (result == NULL) {
+                gf_log (GF_MNT, GF_LOG_DEBUG, "PROC returned error");
+                svcerr_systemerr (transp);
+        }
+        if (result != NULL && !svc_sendreply(transp, (xdrproc_t) _xdr_result,
+                                             result)) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "svc_sendreply returned error");
+                svcerr_systemerr (transp);
+        }
+        if (!svc_freeargs (transp, (xdrproc_t) _xdr_argument,
+                           (caddr_t) &argument)) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "unable to free arguments");
+        }
+        if (result == NULL)
+                return;
+        /* free the result */
+        switch (rqstp->rq_proc) {
+        case MOUNT3_MNT:
+                res = (mountres3 *) result;
+                GF_FREE (res->mountres3_u.mountinfo.fhandle.fhandle3_val);
+                GF_FREE (res->mountres3_u.mountinfo.auth_flavors.auth_flavors_val);
+                GF_FREE (res);
+                break;
+
+        case MOUNT3_UMNT:
+                GF_FREE (result);
+                break;
+        }
+        return;
+}
+
+void *
+mount3udp_thread (void *argv)
+{
+        register SVCXPRT *transp = NULL;
+
+        transp = svcudp_create(RPC_ANYSOCK);
+        if (transp == NULL) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "svcudp_create error");
+                return NULL;
+        }
+        if (!svc_register(transp, MOUNT_PROGRAM, MOUNT_V3,
+                          mountudp_program_3, IPPROTO_UDP)) {
+                gf_log (GF_MNT, GF_LOG_ERROR, "svc_register error");
+                return NULL;
+        }
+
+        svc_run ();
+        gf_log (GF_MNT, GF_LOG_ERROR, "svc_run returned");
+        return NULL;
+}
diff --git a/xlators/nfs/server/src/nfs-common.c b/xlators/nfs/server/src/nfs-common.c
index 369bb1816..f74396ee8 100644
--- a/xlators/nfs/server/src/nfs-common.c
+++ b/xlators/nfs/server/src/nfs-common.c
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -33,7 +24,6 @@
 #include "nfs-mem-types.h"
 #include "rpcsvc.h"
 #include "iatt.h"
-#include "nfs.h"
 
 #include <libgen.h>
 
@@ -89,13 +79,13 @@ nfs_mntpath_to_xlator (xlator_list_t *cl, char *path)
 {
         char            volname[MNTPATHLEN];
         char            *volptr = NULL;
-        int             pathlen = 0;
+        size_t          pathlen;
         xlator_t        *targetxl = NULL;
 
         if ((!cl) || (!path))
                 return NULL;
 
-        strcpy (volname, path);
+        strncpy (volname, path, MNTPATHLEN);
         pathlen = strlen (volname);
         gf_log (GF_NFS, GF_LOG_TRACE, "Subvolume search: %s", path);
         if (volname[0] == '/')
@@ -103,7 +93,7 @@ nfs_mntpath_to_xlator (xlator_list_t *cl, char *path)
         else
                 volptr = &volname[0];
 
-        if (volname[pathlen - 1] == '/')
+        if (pathlen && volname[pathlen - 1] == '/')
                 volname[pathlen - 1] = '\0';
 
         while (cl) {
@@ -132,7 +122,7 @@ nfs_zero_filled_stat (struct iatt *buf)
          * we've already mapped the root ino to 1 so it is not guaranteed to be
          * 0.
          */
-        if ((buf->ia_nlink == 0) && (buf->ia_type == 0))
+        if ((buf->ia_nlink == 0) && (buf->ia_ctime == 0))
                 return 1;
 
         return 0;
@@ -142,53 +132,18 @@ nfs_zero_filled_stat (struct iatt *buf)
 void
 nfs_loc_wipe (loc_t *loc)
 {
-        if (!loc)
-                return;
-
-        if (loc->path) {
-                GF_FREE ((char *)loc->path);
-                loc->path = NULL;
-        }
-
-        if (loc->parent) {
-                inode_unref (loc->parent);
-                loc->parent = NULL;
-	}
-
-        if (loc->inode) {
-                inode_unref (loc->inode);
-                loc->inode = NULL;
-        }
-
-        loc->ino = 0;
+        loc_wipe (loc);
 }
 
 
 int
 nfs_loc_copy (loc_t *dst, loc_t *src)
 {
-	int ret = -1;
-
-	dst->ino = src->ino;
-
-	if (src->inode)
-		dst->inode = inode_ref (src->inode);
+        int  ret = -1;
 
-	if (src->parent)
-		dst->parent = inode_ref (src->parent);
+        ret = loc_copy (dst, src);
 
-	dst->path = gf_strdup (src->path);
-
-	if (!dst->path)
-		goto out;
-
-	dst->name = strrchr (dst->path, '/');
-	if (dst->name)
-		dst->name++;
-
-	ret = 0;
-out:
-	return ret;
+        return ret;
 }
 
 
@@ -202,24 +157,24 @@ nfs_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path)
 
         if (inode) {
                 loc->inode = inode_ref (inode);
-                loc->ino = inode->ino;
+                if (!uuid_is_null (inode->gfid))
+                        uuid_copy (loc->gfid, inode->gfid);
         }
 
         if (parent)
                 loc->parent = inode_ref (parent);
 
-        loc->path = gf_strdup (path);
-        if (!loc->path) {
-                gf_log (GF_NFS, GF_LOG_ERROR, "strdup failed");
-                goto loc_wipe;
+        if (path) {
+                loc->path = gf_strdup (path);
+                if (!loc->path) {
+                        gf_log (GF_NFS, GF_LOG_ERROR, "strdup failed");
+                        goto loc_wipe;
+                }
+                loc->name = strrchr (loc->path, '/');
+                if (loc->name)
+                        loc->name++;
         }
 
-        loc->name = strrchr (loc->path, '/');
-        if (loc->name)
-                loc->name++;
-        else
-                goto loc_wipe;
-
         ret = 0;
 loc_wipe:
         if (ret < 0)
@@ -230,7 +185,7 @@ loc_wipe:
 
 
 int
-nfs_inode_loc_fill (inode_t *inode, loc_t *loc)
+nfs_inode_loc_fill (inode_t *inode, loc_t *loc, int how)
 {
         char            *resolvedpath = NULL;
         inode_t         *parent = NULL;
@@ -239,34 +194,47 @@ nfs_inode_loc_fill (inode_t *inode, loc_t *loc)
         if ((!inode) || (!loc))
                 return ret;
 
-        if ((inode) && (inode->ino == 1))
-                goto ignore_parent;
-
-        parent = inode_parent (inode, 0, NULL);
-        if (!parent)
-                goto err;
+        /* If gfid is not null, then the inode is already linked to
+         * the inode table, and not a newly created one. For newly
+         * created inode, inode_path returns null gfid as the path.
+         */
+        if (!uuid_is_null (inode->gfid)) {
+                ret = inode_path (inode, NULL, &resolvedpath);
+                if (ret < 0) {
+                        gf_log (GF_NFS, GF_LOG_ERROR, "path resolution failed "
+                                "%s", resolvedpath);
+                        goto err;
+                }
+        }
 
-ignore_parent:
-        ret = inode_path (inode, NULL, &resolvedpath);
-        if (ret < 0)
-                goto err;
+        if (resolvedpath == NULL) {
+                char tmp_path[GFID_STR_PFX_LEN + 1] = {0,};
+                snprintf (tmp_path, sizeof (tmp_path), "<gfid:%s>",
+                          uuid_utoa (loc->gfid));
+                resolvedpath = gf_strdup (tmp_path);
+        } else {
+                parent = inode_parent (inode, loc->pargfid, NULL);
+        }
 
         ret = nfs_loc_fill (loc, inode, parent, resolvedpath);
-        if (ret < 0)
+        if (ret < 0) {
+                gf_log (GF_NFS, GF_LOG_ERROR, "loc fill resolution failed %s",
+                                resolvedpath);
                 goto err;
+        }
 
+        ret = 0;
 err:
         if (parent)
                 inode_unref (parent);
 
-        if (resolvedpath)
-                GF_FREE (resolvedpath);
+        GF_FREE (resolvedpath);
 
         return ret;
 }
 
 int
-nfs_gfid_loc_fill (inode_table_t *itable, uuid_t gfid, loc_t *loc)
+nfs_gfid_loc_fill (inode_table_t *itable, uuid_t gfid, loc_t *loc, int how)
 {
         int             ret = -EFAULT;
         inode_t         *inode = NULL;
@@ -276,11 +244,33 @@ nfs_gfid_loc_fill (inode_table_t *itable, uuid_t gfid, loc_t *loc)
 
         inode = inode_find (itable, gfid);
         if (!inode) {
-                ret = -ENOENT;
-                goto err;
-        }
+		gf_log (GF_NFS, GF_LOG_TRACE, "Inode not found in itable, will try to create one.");
+                if (how == NFS_RESOLVE_CREATE) {
+			gf_log (GF_NFS, GF_LOG_TRACE, "Inode needs to be created.");
+                        inode = inode_new (itable);
+                        if (!inode) {
+                                gf_log (GF_NFS, GF_LOG_ERROR, "Failed to "
+                                        "allocate memory");
+                                ret = -ENOMEM;
+                                goto err;
+                        }
+
+                } else {
+			gf_log (GF_NFS, GF_LOG_ERROR, "Inode not found in itable and no creation was requested.");
+                        ret = -ENOENT;
+                        goto err;
+                }
+        } else {
+		gf_log (GF_NFS, GF_LOG_TRACE, "Inode was found in the itable.");
+	}
+
+        uuid_copy (loc->gfid, gfid);
 
-        ret = nfs_inode_loc_fill (inode, loc);
+        ret = nfs_inode_loc_fill (inode, loc, how);
+	if (ret < 0) {
+		gf_log (GF_NFS, GF_LOG_ERROR, "Inode loc filling failed.: %s", strerror (-ret));
+		goto err;
+	}
 
 err:
         if (inode)
@@ -295,7 +285,7 @@ nfs_root_loc_fill (inode_table_t *itable, loc_t *loc)
         uuid_t  rootgfid = {0, };
 
         rootgfid[15] = 1;
-        return nfs_gfid_loc_fill (itable, rootgfid, loc);
+        return nfs_gfid_loc_fill (itable, rootgfid, loc, NFS_RESOLVE_EXIST);
 }
 
 
@@ -311,11 +301,14 @@ nfs_parent_inode_loc_fill (inode_t *parent, inode_t *entryinode, char *entry,
                 return ret;
 
         ret = inode_path (parent, entry, &path);
-        if (ret < 0)
+        if (ret < 0) {
+                gf_log (GF_NFS, GF_LOG_ERROR, "path resolution failed %s",
+                                path);
                 goto err;
+        }
 
         ret = nfs_loc_fill (loc, entryinode, parent, path);
-
+        GF_FREE (path);
 err:
         return ret;
 }
@@ -347,6 +340,8 @@ nfs_entry_loc_fill (inode_table_t *itable, uuid_t pargfid, char *entry,
         if (!parent)
                 goto err;
 
+        uuid_copy (loc->pargfid, pargfid);
+
         ret = -2;
         entryinode = inode_grep (itable, parent, entry);
         if (!entryinode) {
@@ -374,13 +369,18 @@ nfs_entry_loc_fill (inode_table_t *itable, uuid_t pargfid, char *entry,
 
         ret = inode_path (parent, entry, &resolvedpath);
         if (ret < 0) {
+                gf_log (GF_NFS, GF_LOG_ERROR, "path resolution failed %s",
+                                resolvedpath);
                 ret = -3;
                 goto err;
         }
 
         ret = nfs_loc_fill (loc, entryinode, parent, resolvedpath);
-        if (ret < 0)
+        if (ret < 0) {
+                gf_log (GF_NFS, GF_LOG_ERROR, "loc_fill failed %s",
+                                resolvedpath);
                 ret = -3;
+        }
 
 err:
         if (parent)
@@ -389,49 +389,77 @@ err:
         if (entryinode)
                 inode_unref (entryinode);
 
-        if (resolvedpath)
-                GF_FREE (resolvedpath);
+        GF_FREE (resolvedpath);
 
         return ret;
 }
 
 
-int
-__is_nfs_subvolume_disabled (struct nfs_state *nfs, xlator_t *xl)
+uint32_t
+nfs_hash_gfid (uuid_t gfid)
 {
-        int     x = 0;
-        int     notstarted = 0;
+        uint32_t                hash = 0;
+        uint64_t                msb64 = 0;
+        uint64_t                lsb64 = 0;
+        uint32_t                a1 = 0;
+        uint32_t                a2 = 0;
+        uint32_t                a3 = 0;
+        uint32_t                a4 = 0;
+        uint32_t                b1 = 0;
+        uint32_t                b2 = 0;
 
-        if ((!nfs) || (!xl))
-                return 1;
+        if (__is_root_gfid (gfid))
+                return 0x1;
 
-        for (;x < nfs->allsubvols; ++x) {
-                if (!((&nfs->subvols[x])->subvol == xl))
-                        continue;
+        memcpy (&msb64, &gfid[8], 8);
+        memcpy (&lsb64, &gfid[0], 8);
 
-                if (gf_nfs_subvolume_notstarted (&nfs->subvols[x])) {
-                        notstarted = 1;
-                        goto out;
-                }
-        }
-out:
-        return notstarted;
+        a1 = (msb64 << 32);
+        a2 = (msb64 >> 32);
+        a3 = (lsb64 << 32);
+        a4 = (lsb64 >> 32);
+
+        b1 = a1 ^ a4;
+        b2 = a2 ^ a3;
+
+        hash = b1 ^ b2;
+
+        return hash;
 }
 
 
-int
-is_nfs_subvolume_disabled (struct nfs_state *nfs, xlator_t *xl)
+void
+nfs_fix_generation (xlator_t *this, inode_t *inode)
 {
-        int     notstarted = 0;
+        uint64_t                 raw_ctx        = 0;
+        struct nfs_inode_ctx    *ictx           = NULL;
+        struct nfs_state        *priv           = NULL;
+        int                      ret            = -1;
 
-        if ((!nfs) || (!xl))
-                return 1;
-
-        LOCK (&nfs->svinitlock);
-        {
-                notstarted = __is_nfs_subvolume_disabled (nfs, xl);
+        if (!inode) {
+                return;
         }
-        UNLOCK (&nfs->svinitlock);
+        priv = this->private;
 
-        return notstarted;
+        if (inode_ctx_get(inode,this,&raw_ctx) == 0) {
+                ictx = (struct nfs_inode_ctx *)raw_ctx;
+                ictx->generation = priv->generation;
+        }
+        else {
+                ictx = GF_CALLOC (1, sizeof (struct nfs_inode_ctx),
+                                  gf_nfs_mt_inode_ctx);
+                if (!ictx) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "could not allocate nfs inode ctx");
+                        return;
+                }
+                INIT_LIST_HEAD(&ictx->shares);
+                ictx->generation = priv->generation;
+                ret = inode_ctx_put (inode, this, (uint64_t)ictx);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "could not store nfs inode ctx");
+                        return;
+                }
+        }
 }
diff --git a/xlators/nfs/server/src/nfs-common.h b/xlators/nfs/server/src/nfs-common.h
index 3e791a905..2e97f1563 100644
--- a/xlators/nfs/server/src/nfs-common.h
+++ b/xlators/nfs/server/src/nfs-common.h
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _NFS_COMMON_H_
@@ -31,12 +22,13 @@
 #include "rpcsvc.h"
 #include "iatt.h"
 #include "uuid.h"
-#include "nfs.h"
 
-#define NFS_PATH_MAX    PATH_MAX
+//NFS_PATH_MAX hard-coded to 4096 as a work around for bug 2476.
+//nfs server crashes when path received is longer than PATH_MAX
+#define NFS_PATH_MAX    4096
 #define NFS_NAME_MAX    NAME_MAX
 
-#define NFS_DEFAULT_CREATE_MODE 0644
+#define NFS_DEFAULT_CREATE_MODE 0600
 
 extern xlator_t *
 nfs_xlid_to_xlator (xlator_list_t *cl, uint8_t xlid);
@@ -66,7 +58,7 @@ nfs_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path);
 #define NFS_RESOLVE_CREATE      2
 
 extern int
-nfs_inode_loc_fill (inode_t *inode, loc_t *loc);
+nfs_inode_loc_fill (inode_t *inode, loc_t *loc, int how);
 
 extern int
 nfs_ino_loc_fill (inode_table_t *itable, uuid_t gfid, loc_t *l);
@@ -78,6 +70,12 @@ nfs_entry_loc_fill (inode_table_t *itable, uuid_t pargfid, char *entry,
 extern int
 nfs_root_loc_fill (inode_table_t *itable, loc_t *loc);
 
+extern uint32_t
+nfs_hash_gfid (uuid_t gfid);
+
 extern int
-is_nfs_subvolume_disabled (struct nfs_state *nfs, xlator_t *xl);
+nfs_gfid_loc_fill (inode_table_t *itable, uuid_t gfid, loc_t *loc, int how);
+
+void
+nfs_fix_generation (xlator_t *this, inode_t *inode);
 #endif
diff --git a/xlators/nfs/server/src/nfs-fops.c b/xlators/nfs/server/src/nfs-fops.c
index 3f342710b..236b80c76 100644
--- a/xlators/nfs/server/src/nfs-fops.c
+++ b/xlators/nfs/server/src/nfs-fops.c
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -22,6 +13,9 @@
 #include "config.h"
 #endif
 
+#include <grp.h>
+#include <pwd.h>
+
 #include "dict.h"
 #include "xlator.h"
 #include "iobuf.h"
@@ -31,10 +25,80 @@
 #include "nfs-fops.h"
 #include "inode.h"
 #include "nfs-common.h"
-
+#include "nfs3-helpers.h"
+#include "nfs-mem-types.h"
 #include <libgen.h>
 #include <semaphore.h>
 
+void
+nfs_fix_groups (xlator_t *this, call_stack_t *root)
+{
+        struct passwd    mypw;
+        char             mystrs[1024];
+        struct passwd    *result;
+        gid_t            mygroups[GF_MAX_AUX_GROUPS];
+        int              ngroups;
+        int              i;
+        struct nfs_state *priv = this->private;
+        const gid_list_t *agl;
+	gid_list_t gl;
+
+        if (!priv->server_aux_gids) {
+                return;
+        }
+
+	agl = gid_cache_lookup(&priv->gid_cache, root->uid);
+	if (agl) {
+		for (ngroups = 0; ngroups < agl->gl_count; ngroups++) 
+			root->groups[ngroups] = agl->gl_list[ngroups];
+		root->ngrps = ngroups;
+		gid_cache_release(&priv->gid_cache, agl);
+		return;
+	}
+
+	/* No cached list found. */
+        if (getpwuid_r(root->uid,&mypw,mystrs,sizeof(mystrs),&result) != 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "getpwuid_r(%u) failed", root->uid);
+                return;
+        }
+
+        if (!result) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "getpwuid_r(%u) found nothing", root->uid);
+                return;
+        }
+
+        gf_log (this->name, GF_LOG_TRACE, "mapped %u => %s",
+                root->uid, result->pw_name);
+
+        ngroups = GF_MAX_AUX_GROUPS;
+        if (getgrouplist(result->pw_name,root->gid,mygroups,&ngroups) == -1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "could not map %s to group list", result->pw_name);
+                return;
+        }
+
+	/* Add the group data to the cache. */
+	gl.gl_list = GF_CALLOC(ngroups, sizeof(gid_t), gf_nfs_mt_aux_gids);
+	if (gl.gl_list) {
+		/* It's not fatal if the alloc failed. */
+		gl.gl_id = root->uid;
+		gl.gl_count = ngroups;
+		memcpy(gl.gl_list, mygroups, sizeof(gid_t) * ngroups);
+		if (gid_cache_add(&priv->gid_cache, &gl) != 1)
+			GF_FREE(gl.gl_list);
+	}
+
+	/* Copy data to the frame. */
+        for (i = 0; i < ngroups; ++i) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "%s is in group %u", result->pw_name, mygroups[i]);
+                root->groups[i] = mygroups[i];
+        }
+        root->ngrps = ngroups;
+}
+
 struct nfs_fop_local *
 nfs_fop_local_init (xlator_t *nfsx)
 {
@@ -52,13 +116,9 @@ nfs_fop_local_init (xlator_t *nfsx)
 void
 nfs_fop_local_wipe (xlator_t *nfsx, struct nfs_fop_local *l)
 {
-        struct nfs_state        *nfs = NULL;
-
         if ((!nfsx) || (!l))
                 return;
 
-        nfs = nfsx->private;
-
         if (l->iobref)
                 iobref_unref (l->iobref);
 
@@ -74,7 +134,7 @@ nfs_fop_local_wipe (xlator_t *nfsx, struct nfs_fop_local *l)
         if (l->dictgfid)
                 dict_unref (l->dictgfid);
 
-        mem_put (nfs->foppool, l);
+        mem_put (l);
 
         return;
 }
@@ -94,7 +154,7 @@ unsigned int            cval = 1;
 int
 nfs_frame_getctr ()
 {
-        int val = 0;
+        uint64_t val = 0;
 
         pthread_mutex_lock (&ctr);
         {
@@ -122,20 +182,33 @@ nfs_create_frame (xlator_t *xl, nfs_user_t *nfu)
         frame = create_frame (xl, (call_pool_t *)xl->ctx->pool);
         if (!frame)
                 goto err;
+	if (call_stack_alloc_groups (frame->root, nfu->ngrps) != 0) {
+		STACK_DESTROY (frame->root);
+		frame = NULL;
+		goto err;
+	}
+
+        frame->root->pid = NFS_PID;
         frame->root->uid = nfu->uid;
         frame->root->gid = nfu->gids[NFS_PRIMGID_IDX];
-        if (nfu->ngrps == 1)
-                goto err;       /* Done, we only got primary gid */
+        frame->root->lk_owner = nfu->lk_owner;
 
-        frame->root->ngrps = nfu->ngrps - 1;
+        if (nfu->ngrps != 1) {
+                frame->root->ngrps = nfu->ngrps - 1;
 
-        gf_log (GF_NFS, GF_LOG_TRACE,"uid: %d, gid %d, gids: %d",
-                frame->root->uid, frame->root->gid, frame->root->ngrps);
-        for(y = 0, x = 1;  y < frame->root->ngrps; x++,y++) {
-                gf_log (GF_NFS, GF_LOG_TRACE, "gid: %d", nfu->gids[x]);
-                frame->root->groups[y] = nfu->gids[x];
+                gf_log (GF_NFS, GF_LOG_TRACE,"uid: %d, gid %d, gids: %d",
+                        frame->root->uid, frame->root->gid, frame->root->ngrps);
+                for(y = 0, x = 1;  y < frame->root->ngrps; x++,y++) {
+                        gf_log (GF_NFS, GF_LOG_TRACE, "gid: %d", nfu->gids[x]);
+                        frame->root->groups[y] = nfu->gids[x];
+                }
         }
-        frame->root->lk_owner = nfs_frame_getctr ();
+
+        /*
+         * It's tempting to do this *instead* of using nfu above, but we need
+         * to have those values in case nfs_fix_groups doesn't do anything.
+         */
+        nfs_fix_groups(xl,frame->root);
 
 err:
         return frame;
@@ -143,7 +216,7 @@ err:
 
 #define nfs_fop_handle_frame_create(fram, xla, nfuser, retval, errlabel)      \
         do {                                                                  \
-                fram = nfs_create_frame (xla, (nfuser));                      \
+                fram = nfs_create_frame (xla, (nfuser));                \
                 if (!fram) {                                                  \
                         retval = (-ENOMEM);                                   \
                         gf_log (GF_NFS, GF_LOG_ERROR,"Frame creation failed");\
@@ -155,28 +228,31 @@ err:
  * for us to determine in the callback whether to funge the ino in the stat buf
  * with 1 for the parent.
  */
-#define nfs_fop_save_root_ino(locl, loc)                                    \
-        do {                                                                \
-                if ((loc)->ino == 1)                                        \
-                        (locl)->rootinode = 1;                              \
-                else if (((loc)->parent) && ((loc)->parent->ino == 1))      \
-                        (locl)->rootparentinode = 1;                        \
-        } while (0)                                                         \
+#define nfs_fop_save_root_ino(locl, loc)                                \
+        do {                                                            \
+                if (((loc)->inode) &&                                   \
+                    __is_root_gfid ((loc)->inode->gfid))                \
+                        (locl)->rootinode = 1;                          \
+                else if (((loc)->parent) &&                             \
+                         __is_root_gfid ((loc)->parent->gfid))          \
+                        (locl)->rootparentinode = 1;                    \
+        } while (0)
 
 /* Do the same for an fd */
-#define nfs_fop_save_root_fd_ino(locl, fdesc)                               \
-        do {                                                                \
-                if ((fdesc)->inode->ino == 1)                               \
-                        (locl)->rootinode = 1;                              \
-        } while (0)                                                         \
-
+#define nfs_fop_save_root_fd_ino(locl, fdesc)                           \
+        do {                                                            \
+                if (__is_root_gfid ((fdesc)->inode->gfid))              \
+                        (locl)->rootinode = 1;                          \
+        } while (0)
 
 
 /* Use the state saved by the previous macro to funge the ino in the appropriate
  * structure.
  */
-#define nfs_fop_restore_root_ino(locl, preattr, postattr, prepar, postpar)  \
+#define nfs_fop_restore_root_ino(locl, fopret, preattr, postattr, prepar, postpar)  \
         do {                                                                \
+                if (fopret == -1)                                           \
+                        break;                                              \
                 if ((locl)->rootinode) {                                    \
                         if ((preattr)) {                                    \
                                 ((struct iatt *)(preattr))->ia_ino = 1;     \
@@ -201,17 +277,21 @@ err:
 /* If the newly created, inode's parent is root, we'll need to funge the ino
  * in the parent attr when we receive them in the callback.
  */
-#define nfs_fop_newloc_save_root_ino(locl, newloc)                             \
-        do {                                                                   \
-                if ((newloc)->ino == 1)                                        \
-                        (locl)->newrootinode = 1;                              \
-                else if (((newloc)->parent) && ((newloc)->parent->ino == 1))   \
-                        (locl)->newrootparentinode = 1;                        \
-        } while (0)                                                            \
-
-
-#define nfs_fop_newloc_restore_root_ino(locl, preattr, postattr, prepar, postpar)  \
+#define nfs_fop_newloc_save_root_ino(locl, newloc)                      \
+        do {                                                            \
+                if (((newloc)->inode) &&                                \
+                    __is_root_gfid ((newloc)->inode->gfid))             \
+                        (locl)->newrootinode = 1;                       \
+                else if (((newloc)->parent) &&                          \
+                         __is_root_gfid ((newloc)->parent->gfid))       \
+                        (locl)->newrootparentinode = 1;                 \
+        } while (0)
+
+#define nfs_fop_newloc_restore_root_ino(locl, fopret, preattr, postattr, prepar, postpar)  \
         do {                                                                   \
+                if (fopret == -1)                                              \
+                        break;                                                 \
+                                                                               \
                 if ((locl)->newrootinode) {                                    \
                         if ((preattr))                                         \
                                 ((struct iatt *)(preattr))->ia_ino = 1;        \
@@ -226,16 +306,21 @@ err:
         } while (0)                                                            \
 
 dict_t *
-nfs_gfid_dict ()
+nfs_gfid_dict (inode_t *inode)
 {
         uuid_t  newgfid = {0, };
         char    *dyngfid = NULL;
         dict_t  *dictgfid = NULL;
         int     ret = -1;
+        uuid_t  rootgfid = {0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1};
 
         dyngfid = GF_CALLOC (1, sizeof (uuid_t), gf_common_mt_char);
         uuid_generate (newgfid);
-        memcpy (dyngfid, newgfid, sizeof (uuid_t));
+
+        if (uuid_compare (inode->gfid, rootgfid) == 0)
+                memcpy (dyngfid, rootgfid, sizeof (uuid_t));
+        else
+                memcpy (dyngfid, newgfid, sizeof (uuid_t));
 
         dictgfid = dict_new ();
         if (!dictgfid) {
@@ -253,10 +338,10 @@ out:
         return dictgfid;
 }
 
-#define nfs_fop_gfid_setup(nflcl, retval, erlbl)                        \
+#define nfs_fop_gfid_setup(nflcl, inode, retval, erlbl)                 \
         do {                                                            \
                 if (nflcl) {                                            \
-                        (nflcl)->dictgfid = nfs_gfid_dict ();           \
+                        (nflcl)->dictgfid = nfs_gfid_dict (inode);      \
                                                                         \
                         if (!((nflcl)->dictgfid)) {                     \
                                 retval = -EFAULT;                       \
@@ -265,9 +350,6 @@ out:
                 }                                                       \
         } while (0)                                                     \
 
-
-
-
 /* Fops Layer Explained
  * The fops layer has three types of functions. They can all be identified by
  * their names. Here are the three patterns:
@@ -282,7 +364,7 @@ out:
  *
  * nfs_<fopname>
  * Unlike the nfs_fop_<fopname> variety, this is the stateful type of fop, in
- * that it silently performs all the relevant GlusterFS state maintainence
+ * that it silently performs all the relevant GlusterFS state maintenance
  * operations on the data returned to the callbacks, leaving the caller's
  * callback to just use the data returned for whatever it needs to do with that
  * data, for eg. the nfs_lookup, will take care of looking up the inodes,
@@ -301,8 +383,12 @@ nfs_fop_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         struct nfs_fop_local    *local = NULL;
         fop_lookup_cbk_t        progcbk;
 
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
+
         nfl_to_prog_data (local, progcbk, frame);
-        nfs_fop_restore_root_ino (local, buf, NULL, NULL, postparent);
+        nfs_fop_restore_root_ino (local, op_ret, buf, NULL, NULL, postparent);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, inode, buf,
                          xattr, postparent);
@@ -327,7 +413,7 @@ nfs_fop_lookup (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
         nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err);
         nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
         nfs_fop_save_root_ino (nfl, loc);
-        nfs_fop_gfid_setup (nfl, ret, err);
+        nfs_fop_gfid_setup (nfl, loc->inode, ret, err);
 
         STACK_WIND_COOKIE (frame, nfs_fop_lookup_cbk, xl, xl,
                            xl->fops->lookup, loc, nfl->dictgfid);
@@ -342,18 +428,63 @@ err:
         return ret;
 }
 
+int32_t
+nfs_fop_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        struct nfs_fop_local    *nfl = NULL;
+        fop_access_cbk_t        progcbk = NULL;
+
+        nfl_to_prog_data (nfl, progcbk, frame);
+        if (progcbk)
+                progcbk (frame, cookie, this, op_ret, op_errno, xdata);
+
+        nfs_stack_destroy (nfl, frame);
+        return 0;
+}
+
+int
+nfs_fop_access (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
+                int32_t accesstest, fop_access_cbk_t cbk, void *local)
+{
+        call_frame_t            *frame = NULL;
+        int                     ret = -EFAULT;
+        struct nfs_fop_local    *nfl = NULL;
+        uint32_t                accessbits = 0;
+
+        if ((!xl) || (!loc) || (!nfu))
+                return ret;
+
+        gf_log (GF_NFS, GF_LOG_TRACE, "Access: %s", loc->path);
+        nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err);
+        nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
+        nfs_fop_save_root_ino (nfl, loc);
+
+        accessbits = nfs3_request_to_accessbits (accesstest);
+        STACK_WIND_COOKIE (frame, nfs_fop_access_cbk, xl, xl, xl->fops->access,
+                           loc, accessbits, NULL);
+        ret = 0;
+err:
+        if (ret < 0) {
+                if (frame)
+                        nfs_stack_destroy (nfl, frame);
+        }
+
+        return ret;
+}
 
 int32_t
 nfs_fop_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                  int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                  dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_stat_cbk_t          progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, buf, NULL, NULL, NULL);
+        nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, NULL, NULL);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, buf);
+                progcbk (frame, cookie, this, op_ret, op_errno, buf, xdata);
 
         nfs_stack_destroy (nfl, frame);
         return 0;
@@ -377,7 +508,7 @@ nfs_fop_stat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
         nfs_fop_save_root_ino (nfl, loc);
 
         STACK_WIND_COOKIE (frame, nfs_fop_stat_cbk, xl, xl, xl->fops->stat,
-                           loc);
+                           loc, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -391,15 +522,16 @@ err:
 
 int32_t
 nfs_fop_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                   int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                   dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_fstat_cbk_t         progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, buf, NULL, NULL, NULL);
+        nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, NULL, NULL);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, buf);
+                progcbk (frame, cookie, this, op_ret, op_errno, buf, xdata);
 
         nfs_stack_destroy (nfl, frame);
         return 0;
@@ -423,7 +555,7 @@ nfs_fop_fstat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
         nfs_fop_save_root_fd_ino (nfl, fd);
 
         STACK_WIND_COOKIE (frame, nfs_fop_fstat_cbk, xl, xl, xl->fops->fstat,
-                           fd);
+                           fd, NULL);
 
         ret = 0;
 err:
@@ -438,14 +570,14 @@ err:
 
 int32_t
 nfs_fop_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno, fd_t *fd)
+                     int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_opendir_cbk_t       progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, fd);
+                progcbk (frame, cookie, this, op_ret, op_errno, fd, xdata);
         nfs_stack_destroy (nfl, frame);
         return 0;
 }
@@ -467,7 +599,7 @@ nfs_fop_opendir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
         nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
 
         STACK_WIND_COOKIE (frame, nfs_fop_opendir_cbk, xl, xl,
-                           xl->fops->opendir, pathloc, dirfd);
+                           xl->fops->opendir, pathloc, dirfd, NULL);
         ret = 0;
 
 err:
@@ -481,14 +613,14 @@ err:
 
 int
 nfs_fop_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno)
+                   int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_flush_cbk_t         progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno);
+                progcbk (frame, cookie, this, op_ret, op_errno, xdata);
 
         nfs_stack_destroy (nfl, frame);
         return 0;
@@ -510,7 +642,7 @@ nfs_fop_flush (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
         nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
 
         STACK_WIND_COOKIE (frame, nfs_fop_flush_cbk, xl, xl, xl->fops->flush,
-                           fd);
+                           fd, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -524,14 +656,15 @@ err:
 
 int32_t
 nfs_fop_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno, gf_dirent_t *entries)
+                      int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                      dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_readdirp_cbk_t      progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, entries);
+                progcbk (frame, cookie, this, op_ret, op_errno, entries, xdata);
 
         nfs_stack_destroy (nfl, frame);
 
@@ -556,7 +689,7 @@ nfs_fop_readdirp (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *dirfd,
         nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
 
         STACK_WIND_COOKIE (frame, nfs_fop_readdirp_cbk, xl, xl,
-                           xl->fops->readdirp, dirfd, bufsize, offset);
+                           xl->fops->readdirp, dirfd, bufsize, offset, 0);
 
         ret = 0;
 err:
@@ -571,7 +704,8 @@ err:
 
 int32_t
 nfs_fop_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, struct statvfs *buf)
+                    int32_t op_ret, int32_t op_errno, struct statvfs *buf,
+                    dict_t *xdata)
 {
 
         struct nfs_fop_local    *nfl = NULL;
@@ -579,7 +713,7 @@ nfs_fop_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, buf);
+                progcbk (frame, cookie, this, op_ret, op_errno, buf, xdata);
 
         nfs_stack_destroy (nfl, frame);
         return 0;
@@ -602,7 +736,7 @@ nfs_fop_statfs (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
         nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
 
         STACK_WIND_COOKIE (frame, nfs_fop_statfs_cbk, xl, xl,
-                           xl->fops->statfs, pathloc);
+                           xl->fops->statfs, pathloc, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -618,16 +752,21 @@ int32_t
 nfs_fop_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
                     struct iatt *buf, struct iatt *preparent,
-                    struct iatt *postparent)
+                    struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_create_cbk_t        progcbk = NULL;
 
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
+
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, buf, NULL, preparent, postparent);
+        nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, preparent,
+                                  postparent);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, fd, inode, buf,
-                         preparent, postparent);
+                         preparent, postparent, NULL);
 
         nfs_stack_destroy (nfl, frame);
         return 0;
@@ -650,10 +789,10 @@ nfs_fop_create (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
         nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err);
         nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
         nfs_fop_save_root_ino (nfl, pathloc);
-        nfs_fop_gfid_setup (nfl, ret, err);
+        nfs_fop_gfid_setup (nfl, pathloc->inode, ret, err);
 
         STACK_WIND_COOKIE (frame, nfs_fop_create_cbk, xl, xl, xl->fops->create,
-                           pathloc, flags, mode, fd, nfl->dictgfid);
+                           pathloc, flags, mode, 0, fd, nfl->dictgfid);
 
         ret = 0;
 err:
@@ -669,15 +808,16 @@ err:
 int32_t
 nfs_fop_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno, struct iatt *pre,
-                     struct iatt *post)
+                     struct iatt *post, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_setattr_cbk_t       progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, pre, post, NULL, NULL);
+        nfs_fop_restore_root_ino (nfl, op_ret, pre, post, NULL, NULL);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, pre, post);
+                progcbk (frame, cookie, this, op_ret, op_errno, pre, post,
+                         xdata);
         nfs_stack_destroy (nfl, frame);
         return 0;
 }
@@ -702,7 +842,7 @@ nfs_fop_setattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
         nfs_fop_save_root_ino (nfl, pathloc);
 
         STACK_WIND_COOKIE (frame, nfs_fop_setattr_cbk, xl, xl,
-                           xl->fops->setattr, pathloc, buf, valid);
+                           xl->fops->setattr, pathloc, buf, valid, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -718,16 +858,20 @@ int32_t
 nfs_fop_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, inode_t *inode,
                    struct iatt *buf, struct iatt *preparent,
-                   struct iatt *postparent)
+                   struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_mkdir_cbk_t         progcbk = NULL;
 
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
+
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, buf, NULL, preparent, postparent);
+        nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL,preparent, postparent);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, inode, buf,
-                         preparent, postparent);
+                         preparent, postparent, xdata);
         nfs_stack_destroy (nfl, frame);
         return 0;
 }
@@ -748,10 +892,10 @@ nfs_fop_mkdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
         nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err);
         nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
         nfs_fop_save_root_ino (nfl, pathloc);
-        nfs_fop_gfid_setup (nfl, ret, err);
+        nfs_fop_gfid_setup (nfl, pathloc->inode, ret, err);
 
         STACK_WIND_COOKIE  (frame, nfs_fop_mkdir_cbk, xl, xl, xl->fops->mkdir,
-                            pathloc, mode, nfl->dictgfid);
+                            pathloc, mode, 0, nfl->dictgfid);
         ret = 0;
 err:
         if (ret < 0) {
@@ -767,16 +911,20 @@ int32_t
 nfs_fop_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno, inode_t *inode,
                      struct iatt *buf, struct iatt *preparent,
-                     struct iatt *postparent)
+                     struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_symlink_cbk_t       progcbk = NULL;
 
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
+
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, buf, NULL, preparent, postparent);
+        nfs_fop_restore_root_ino (nfl, op_ret,buf, NULL, preparent, postparent);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, inode, buf,
-                         preparent, postparent);
+                         preparent, postparent, xdata);
         nfs_stack_destroy (nfl, frame);
         return 0;
 }
@@ -796,10 +944,11 @@ nfs_fop_symlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, char *target,
         nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err);
         nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
         nfs_fop_save_root_ino (nfl, pathloc);
-        nfs_fop_gfid_setup (nfl, ret, err);
+        nfs_fop_gfid_setup (nfl, pathloc->inode, ret, err);
 
         STACK_WIND_COOKIE  (frame, nfs_fop_symlink_cbk, xl, xl,
-                            xl->fops->symlink, target, pathloc, nfl->dictgfid);
+                            xl->fops->symlink, target, pathloc,
+                            0, nfl->dictgfid);
         ret = 0;
 err:
         if (ret < 0) {
@@ -814,15 +963,16 @@ err:
 int32_t
 nfs_fop_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                       int32_t op_ret, int32_t op_errno, const char *path,
-                      struct iatt *buf)
+                      struct iatt *buf, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_readlink_cbk_t      progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, buf, NULL, NULL, NULL);
+        nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, NULL, NULL);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, path, buf);
+                progcbk (frame, cookie, this, op_ret, op_errno, path, buf,
+                         xdata);
         nfs_stack_destroy (nfl, frame);
         return 0;
 }
@@ -845,7 +995,7 @@ nfs_fop_readlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
         nfs_fop_save_root_ino (nfl, pathloc);
 
         STACK_WIND_COOKIE (frame, nfs_fop_readlink_cbk, xl, xl,
-                           xl->fops->readlink, pathloc, size);
+                           xl->fops->readlink, pathloc, size, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -861,16 +1011,20 @@ int32_t
 nfs_fop_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, inode_t *inode,
                    struct iatt *buf, struct iatt *preparent,
-                   struct iatt *postparent)
+                   struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_mknod_cbk_t         progcbk = NULL;
 
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
+
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, buf, NULL, preparent, postparent);
+        nfs_fop_restore_root_ino (nfl, op_ret,buf, NULL, preparent, postparent);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, inode, buf,
-                         preparent, postparent);
+                         preparent, postparent, xdata);
         nfs_stack_destroy (nfl, frame);
         return 0;
 }
@@ -891,10 +1045,10 @@ nfs_fop_mknod (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
         nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err);
         nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
         nfs_fop_save_root_ino (nfl, pathloc);
-        nfs_fop_gfid_setup (nfl, ret, err);
+        nfs_fop_gfid_setup (nfl, pathloc->inode, ret, err);
 
         STACK_WIND_COOKIE  (frame, nfs_fop_mknod_cbk, xl, xl, xl->fops->mknod,
-                            pathloc, mode, dev, nfl->dictgfid);
+                            pathloc, mode, dev, 0, nfl->dictgfid);
         ret = 0;
 err:
         if (ret < 0) {
@@ -908,16 +1062,17 @@ err:
 int32_t
 nfs_fop_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                   struct iatt *postparent)
+                   struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = frame->local;
         fop_rmdir_cbk_t         progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, NULL, NULL, preparent, postparent);
+        nfs_fop_restore_root_ino (nfl, op_ret, NULL, NULL, preparent,
+                                  postparent);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, preparent,
-                         postparent);
+                         postparent, NULL);
         nfs_stack_destroy (nfl, frame);
         return 0;
 }
@@ -941,7 +1096,7 @@ nfs_fop_rmdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
         nfs_fop_save_root_ino (nfl, pathloc);
 
         STACK_WIND_COOKIE (frame, nfs_fop_rmdir_cbk, xl, xl, xl->fops->rmdir,
-                           pathloc);
+                           pathloc, 0, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -957,16 +1112,17 @@ err:
 int32_t
 nfs_fop_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                    struct iatt *postparent)
+                    struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = frame->local;
         fop_unlink_cbk_t         progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, NULL, NULL, preparent, postparent);
+        nfs_fop_restore_root_ino (nfl, op_ret, NULL, NULL, preparent,
+                                  postparent);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, preparent,
-                         postparent);
+                         postparent, xdata);
         nfs_stack_destroy (nfl, frame);
         return 0;
 }
@@ -989,7 +1145,7 @@ nfs_fop_unlink (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
         nfs_fop_save_root_ino (nfl, pathloc);
 
         STACK_WIND_COOKIE (frame, nfs_fop_unlink_cbk, xl, xl,
-                           xl->fops->unlink, pathloc);
+                           xl->fops->unlink, pathloc, 0, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -1006,16 +1162,21 @@ int32_t
 nfs_fop_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno, inode_t *inode,
                   struct iatt *buf, struct iatt *preparent,
-                  struct iatt *postparent)
+                  struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_link_cbk_t          progcbk = NULL;
 
+        if (op_ret == 0) {
+                nfs_fix_generation(this,inode);
+        }
+
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, buf, NULL, preparent, postparent);
+        nfs_fop_restore_root_ino (nfl, op_ret, buf, NULL, preparent,
+                                  postparent);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, inode, buf,
-                         preparent, postparent);
+                         preparent, postparent, xdata);
 
         nfs_stack_destroy (nfl, frame);
         return 0;
@@ -1040,7 +1201,7 @@ nfs_fop_link (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc,
         nfs_fop_save_root_ino (nfl, newloc);
 
         STACK_WIND_COOKIE (frame, nfs_fop_link_cbk, xl, xl, xl->fops->link,
-                           oldloc, newloc);
+                           oldloc, newloc, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -1056,7 +1217,8 @@ int32_t
 nfs_fop_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno, struct iatt *buf,
                     struct iatt *preoldparent, struct iatt *postoldparent,
-                    struct iatt *prenewparent, struct iatt *postnewparent)
+                    struct iatt *prenewparent, struct iatt *postnewparent,
+                    dict_t *xdata)
 {
 
         struct nfs_fop_local    *nfl = NULL;
@@ -1065,15 +1227,16 @@ nfs_fop_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         nfl_to_prog_data (nfl, progcbk, frame);
         /* The preattr arg needs to be NULL instead of @buf because it is
          * possible that the new parent is not root whereas the source dir
-         * could've been. That is handled in the next macro.
+         * could have been. That is handled in the next macro.
          */
-        nfs_fop_restore_root_ino (nfl, NULL, NULL, preoldparent, postoldparent);
-        nfs_fop_newloc_restore_root_ino (nfl, buf, NULL, prenewparent,
+        nfs_fop_restore_root_ino (nfl, op_ret, NULL, NULL, preoldparent,
+                                  postoldparent);
+        nfs_fop_newloc_restore_root_ino (nfl, op_ret, buf, NULL, prenewparent,
                                          postnewparent);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, buf,
                          preoldparent, postoldparent, prenewparent,
-                         postnewparent);
+                         postnewparent, xdata);
         nfs_stack_destroy (nfl, frame);
         return 0;
 }
@@ -1098,7 +1261,7 @@ nfs_fop_rename (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *oldloc,
         nfs_fop_newloc_save_root_ino (nfl, newloc);
 
         STACK_WIND_COOKIE (frame, nfs_fop_rename_cbk, xl, xl,
-                           xl->fops->rename, oldloc, newloc);
+                           xl->fops->rename, oldloc, newloc, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -1112,14 +1275,14 @@ err:
 
 int32_t
 nfs_fop_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno, fd_t *fd)
+                  int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_open_cbk_t          progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, fd);
+                progcbk (frame, cookie, this, op_ret, op_errno, fd, xdata);
         nfs_stack_destroy (nfl, frame);
 
         return 0;
@@ -1127,7 +1290,7 @@ nfs_fop_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 int
 nfs_fop_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
-              int32_t flags, fd_t *fd, int32_t wbflags, fop_open_cbk_t cbk,
+              int32_t flags, fd_t *fd, fop_open_cbk_t cbk,
               void *local)
 {
         call_frame_t            *frame = NULL;
@@ -1142,7 +1305,7 @@ nfs_fop_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
         nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
 
         STACK_WIND_COOKIE (frame, nfs_fop_open_cbk, xl, xl, xl->fops->open,
-                           loc, flags, fd, wbflags);
+                           loc, flags, fd, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -1157,15 +1320,16 @@ err:
 int32_t
 nfs_fop_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                    struct iatt *postbuf)
+                    struct iatt *postbuf, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_writev_cbk_t        progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, prebuf, postbuf, NULL, NULL);
+        nfs_fop_restore_root_ino (nfl, op_ret, prebuf, postbuf, NULL, NULL);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, prebuf,postbuf);
+                progcbk (frame, cookie, this, op_ret, op_errno, prebuf,
+                         postbuf, xdata);
 
         nfs_stack_destroy (nfl, frame);
 
@@ -1175,20 +1339,20 @@ nfs_fop_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 int
 nfs_fop_write (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
-               struct iobuf *srciob, struct iovec *vector, int32_t count,
+               struct iobref *srciobref, struct iovec *vector, int32_t count,
                off_t offset, fop_writev_cbk_t cbk, void *local)
 {
         call_frame_t            *frame = NULL;
         int                     ret = -EFAULT;
         struct nfs_fop_local    *nfl = NULL;
 
-        if ((!nfsx) || (!xl) || (!fd) || (!vector) || (!nfu) || (!srciob))
+        if ((!nfsx) || (!xl) || (!fd) || (!vector) || (!nfu) || (!srciobref))
                 return ret;
 
         nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err);
         nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
         nfs_fop_save_root_fd_ino (nfl, fd);
-
+/*
         nfl->iobref = iobref_new ();
         if (!nfl->iobref) {
                 gf_log (GF_NFS, GF_LOG_ERROR, "iobref creation failed");
@@ -1197,8 +1361,9 @@ nfs_fop_write (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
         }
 
         iobref_add (nfl->iobref, srciob);
-        STACK_WIND_COOKIE (frame, nfs_fop_writev_cbk, xl, xl,xl->fops->writev
-                           , fd, vector, count, offset, nfl->iobref);
+*/
+        STACK_WIND_COOKIE (frame, nfs_fop_writev_cbk, xl, xl,xl->fops->writev,
+                           fd, vector, count, offset, fd->flags, srciobref, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -1213,15 +1378,17 @@ err:
 int32_t
 nfs_fop_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                   struct iatt *postbuf)
+                   struct iatt *postbuf, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_fsync_cbk_t         progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, prebuf, postbuf, NULL, NULL);
+        nfs_fop_restore_root_ino (nfl, op_ret, prebuf, postbuf, NULL, NULL);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, prebuf,postbuf);
+                progcbk (frame, cookie, this, op_ret, op_errno, prebuf,
+                         postbuf, xdata);
+
         nfs_stack_destroy (nfl, frame);
         return 0;
 }
@@ -1244,7 +1411,7 @@ nfs_fop_fsync (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
         nfs_fop_save_root_fd_ino (nfl, fd);
 
         STACK_WIND_COOKIE (frame, nfs_fop_fsync_cbk, xl, xl,
-                           xl->fops->fsync, fd, datasync);
+                           xl->fops->fsync, fd, datasync, NULL);
         ret = 0;
 err:
         if (ret < 0) {
@@ -1259,16 +1426,17 @@ err:
 int32_t
 nfs_fop_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, struct iovec *vector,
-                   int32_t count, struct iatt *stbuf, struct iobref *iobref)
+                   int32_t count, struct iatt *stbuf, struct iobref *iobref,
+                   dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_readv_cbk_t         progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, stbuf, NULL, NULL, NULL);
+        nfs_fop_restore_root_ino (nfl, op_ret, stbuf, NULL, NULL, NULL);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, vector, count,
-                         stbuf, iobref);
+                         stbuf, iobref, xdata);
 
         nfs_stack_destroy (nfl, frame);
         return 0;
@@ -1291,7 +1459,149 @@ nfs_fop_read (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
         nfs_fop_save_root_fd_ino (nfl, fd);
 
         STACK_WIND_COOKIE (frame, nfs_fop_readv_cbk, xl, xl, xl->fops->readv,
-                           fd, size, offset);
+                           fd, size, offset, 0, NULL);
+        ret = 0;
+err:
+        if (ret < 0) {
+                if (frame)
+                        nfs_stack_destroy (nfl, frame);
+        }
+
+        return ret;
+}
+
+int32_t
+nfs_fop_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, struct gf_flock *flock,
+                dict_t *xdata)
+{
+        struct nfs_fop_local    *nfl = NULL;
+        fop_lk_cbk_t         progcbk = NULL;
+
+        nfl_to_prog_data (nfl, progcbk, frame);
+
+        if (!op_ret)
+                fd_lk_insert_and_merge (nfl->fd, nfl->cmd, &nfl->flock);
+
+        fd_unref (nfl->fd);
+
+        if (progcbk)
+                progcbk (frame, cookie, this, op_ret, op_errno, flock, xdata);
+
+        nfs_stack_destroy (nfl, frame);
+        return 0;
+}
+
+
+int
+nfs_fop_lk (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
+            int cmd, struct gf_flock *flock, fop_lk_cbk_t cbk, void *local)
+{
+        call_frame_t            *frame = NULL;
+        int                     ret = -EFAULT;
+        struct nfs_fop_local    *nfl = NULL;
+
+        if ((!xl) || (!fd) || (!nfu))
+                return ret;
+
+        nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err);
+        nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
+
+        nfl->cmd   = cmd;
+        nfl->fd    = fd_ref (fd);
+        nfl->flock = *flock;
+
+        STACK_WIND_COOKIE (frame, nfs_fop_lk_cbk, xl, xl, xl->fops->lk,
+                           fd, cmd, flock, NULL);
+        ret = 0;
+err:
+        if (ret < 0) {
+                if (frame)
+                        nfs_stack_destroy (nfl, frame);
+        }
+
+        return ret;
+}
+
+int32_t
+nfs_fop_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, dict_t *dict,
+                dict_t *xdata)
+{
+        struct nfs_fop_local    *nfl = NULL;
+        fop_getxattr_cbk_t         progcbk = NULL;
+
+        nfl_to_prog_data (nfl, progcbk, frame);
+
+        if (progcbk)
+                progcbk (frame, cookie, this, op_ret, op_errno, dict, xdata);
+
+        nfs_stack_destroy (nfl, frame);
+        return 0;
+}
+
+
+int
+nfs_fop_getxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
+                  char *name, dict_t *xdata, fop_getxattr_cbk_t cbk, void *local)
+{
+        call_frame_t            *frame = NULL;
+        int                     ret = -EFAULT;
+        struct nfs_fop_local    *nfl = NULL;
+
+        if ((!xl) || (!loc) || (!nfu))
+                return ret;
+
+        nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err);
+        nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
+
+        STACK_WIND_COOKIE (frame, nfs_fop_getxattr_cbk, xl, xl, xl->fops->getxattr,
+                           loc, name, NULL);
+        ret = 0;
+err:
+        if (ret < 0) {
+                if (frame)
+                        nfs_stack_destroy (nfl, frame);
+        }
+
+        return ret;
+}
+
+
+int32_t
+nfs_fop_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        struct nfs_fop_local    *nfl = NULL;
+        fop_setxattr_cbk_t         progcbk = NULL;
+
+        nfl_to_prog_data (nfl, progcbk, frame);
+
+        if (progcbk)
+                progcbk (frame, cookie, this, op_ret, op_errno, xdata);
+
+        nfs_stack_destroy (nfl, frame);
+        return 0;
+}
+
+
+int
+nfs_fop_setxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu,
+                  loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata,
+                  fop_setxattr_cbk_t cbk, void *local)
+{
+        call_frame_t            *frame = NULL;
+        int                     ret = -EFAULT;
+        struct nfs_fop_local    *nfl = NULL;
+
+        if ((!xl) || (!loc) || (!nfu))
+                return ret;
+
+        nfs_fop_handle_frame_create (frame, nfsx, nfu, ret, err);
+        nfs_fop_handle_local_init (frame, nfsx, nfl, cbk, local, ret, err);
+
+        STACK_WIND_COOKIE (frame, nfs_fop_setxattr_cbk, xl, xl, xl->fops->setxattr,
+                           loc, dict, flags, xdata);
         ret = 0;
 err:
         if (ret < 0) {
@@ -1306,15 +1616,16 @@ err:
 int32_t
 nfs_fop_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                       int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                      struct iatt *postbuf)
+                      struct iatt *postbuf, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_truncate_cbk_t      progcbk = NULL;
 
         nfl_to_prog_data (nfl, progcbk, frame);
-        nfs_fop_restore_root_ino (nfl, prebuf, postbuf, NULL, NULL);
+        nfs_fop_restore_root_ino (nfl, op_ret, prebuf, postbuf, NULL, NULL);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, prebuf,postbuf);
+                progcbk (frame, cookie, this, op_ret, op_errno, prebuf,
+                         postbuf, xdata);
 
         nfs_stack_destroy (nfl, frame);
         return 0;
@@ -1337,7 +1648,7 @@ nfs_fop_truncate (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
         nfs_fop_save_root_ino (nfl, loc);
 
         STACK_WIND_COOKIE  (frame, nfs_fop_truncate_cbk, xl, xl,
-                            xl->fops->truncate, loc, offset);
+                            xl->fops->truncate, loc, offset, NULL);
 
         ret = 0;
 err:
diff --git a/xlators/nfs/server/src/nfs-fops.h b/xlators/nfs/server/src/nfs-fops.h
index 3fb157aeb..44e99c66b 100644
--- a/xlators/nfs/server/src/nfs-fops.h
+++ b/xlators/nfs/server/src/nfs-fops.h
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _NFS_FOPS_H_
@@ -96,10 +87,14 @@ struct nfs_fop_local {
          */
         int             rootparentinode;
 
-        char            path[NFS_NAME_MAX];
-        char            newpath[NFS_NAME_MAX];
+        char            path[NFS_NAME_MAX + 1];
+        char            newpath[NFS_NAME_MAX + 1];
         xlator_t        *nfsx;
         dict_t          *dictgfid;
+
+        fd_t            *fd;
+        int             cmd;
+        struct gf_flock flock;
 };
 
 extern struct nfs_fop_local *
@@ -116,7 +111,7 @@ nfs_fop_local_wipe (xlator_t *xl, struct nfs_fop_local *l);
                 nflocal = nfs_fop_local_init (nf);                      \
                 if (nflocal) {                                          \
                         nflocal->proglocal = plocal;                    \
-                        nflocal->progcbk = pcbk;                        \
+                        nflocal->progcbk = *VOID(&pcbk);                        \
                         nflocal->nfsx = nf;                             \
                         if (fram)                                       \
                                 ((call_frame_t *)fram)->local = nflocal;\
@@ -134,7 +129,7 @@ nfs_fop_local_wipe (xlator_t *xl, struct nfs_fop_local *l);
 
 #define nfs_fop_handle_local_init(fram,nfx, nfloc, cbck,prgloc,retval,lab)  \
         do {                                                                \
-                prog_data_to_nfl (nfx, nfloc, fram, cbck, prgloc);          \
+                prog_data_to_nfl (nfx, nfloc, fram, cbck, prgloc);      \
                 if (!nfloc) {                                               \
                         gf_log (GF_NFS,GF_LOG_ERROR,"Failed to init local");\
                         retval = -ENOMEM;                                   \
@@ -180,12 +175,12 @@ nfs_fop_fsync (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
 
 extern int
 nfs_fop_write (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
-               struct iobuf *srciob, struct iovec *vector, int32_t count,
+               struct iobref *srciobref, struct iovec *vector, int32_t count,
                off_t offset, fop_writev_cbk_t cbk, void *local);
 
 extern int
 nfs_fop_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
-              int32_t flags, fd_t *fd, int32_t wbflags, fop_open_cbk_t cbk,
+              int32_t flags, fd_t *fd, fop_open_cbk_t cbk,
               void *local);
 
 extern int
@@ -233,4 +228,21 @@ extern int
 nfs_fop_stat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
               fop_stat_cbk_t cbk, void *local);
 
+extern int
+nfs_fop_access (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
+                int32_t accesstest, fop_access_cbk_t cbk, void *local);
+
+extern int
+nfs_fop_lk (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
+            int cmd, struct gf_flock *flock, fop_lk_cbk_t cbk, void *local);
+
+extern int
+nfs_fop_getxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
+                  char *name, dict_t *xdata, fop_getxattr_cbk_t cbk, void *local);
+
+extern int
+nfs_fop_setxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu,
+                  loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata,
+                  fop_setxattr_cbk_t cbk, void *local);
+
 #endif
diff --git a/xlators/nfs/server/src/nfs-generics.c b/xlators/nfs/server/src/nfs-generics.c
index b248c2875..cb32b7f1b 100644
--- a/xlators/nfs/server/src/nfs-generics.c
+++ b/xlators/nfs/server/src/nfs-generics.c
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -46,6 +37,19 @@ nfs_fstat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
         return ret;
 }
 
+int
+nfs_access (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
+            int32_t accesstest, fop_access_cbk_t cbk, void *local)
+{
+        int             ret = -EFAULT;
+
+        if ((!nfsx) || (!xl) || (!pathloc) || (!nfu))
+                return ret;
+
+        ret = nfs_fop_access (nfsx, xl, nfu, pathloc, accesstest, cbk, local);
+
+        return ret;
+}
 
 int
 nfs_stat (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
@@ -138,7 +142,6 @@ nfs_truncate (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
         return ret;
 }
 
-
 int
 nfs_read (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, size_t size,
           off_t offset, fop_readv_cbk_t cbk, void *local)
@@ -146,6 +149,28 @@ nfs_read (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd, size_t size,
         return nfs_fop_read (nfsx, xl, nfu, fd, size, offset, cbk, local);
 }
 
+int
+nfs_lk (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
+        int cmd, struct gf_flock *flock, fop_lk_cbk_t cbk, void *local)
+{
+        return nfs_fop_lk ( nfsx, xl, nfu, fd, cmd, flock, cbk, local);
+}
+
+int
+nfs_getxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
+              char *name, dict_t *xdata, fop_getxattr_cbk_t cbk, void *local)
+{
+        return nfs_fop_getxattr (nfsx, xl, nfu, loc, name, xdata, cbk, local);
+}
+
+int
+nfs_setxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu,
+              loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata,
+              fop_setxattr_cbk_t cbk, void *local)
+{
+        return nfs_fop_setxattr (nfsx, xl, nfu, loc, dict, flags, xdata, cbk,
+                                local);
+}
 
 int
 nfs_fsync (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
@@ -157,11 +182,11 @@ nfs_fsync (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
 
 int
 nfs_write (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
-           struct iobuf *srciob, struct iovec *vector, int32_t count,
+           struct iobref *srciobref, struct iovec *vector, int32_t count,
            off_t offset, fop_writev_cbk_t cbk, void *local)
 {
-        return nfs_fop_write (nfsx, xl, nfu, fd, srciob, vector, count, offset,
-                              cbk, local);
+        return nfs_fop_write (nfsx, xl, nfu, fd, srciobref, vector, count,
+                              offset, cbk, local);
 }
 
 
@@ -174,7 +199,7 @@ nfs_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
         if ((!nfsx) || (!xl) || (!pathloc) || (!nfu))
                 return ret;
 
-        ret = nfs_inode_open (nfsx, xl, nfu, pathloc, flags, GF_OPEN_NOWB, cbk,
+        ret = nfs_inode_open (nfsx, xl, nfu, pathloc, flags, cbk,
                               local);
         return ret;
 }
@@ -317,3 +342,4 @@ nfs_opendir (xlator_t *nfsx, xlator_t *fopxl, nfs_user_t *nfu, loc_t *pathloc,
 
         return nfs_inode_opendir (nfsx, fopxl, nfu, pathloc, cbk, local);
 }
+
diff --git a/xlators/nfs/server/src/nfs-generics.h b/xlators/nfs/server/src/nfs-generics.h
index 55740f7b2..01876d68e 100644
--- a/xlators/nfs/server/src/nfs-generics.h
+++ b/xlators/nfs/server/src/nfs-generics.h
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _NFS_GENERICS_H_
@@ -43,7 +34,7 @@ struct nfs_direntcache {
  * different NFS versions can simply call a standard interface and have fop
  * interface dependent functions be handled internally.
  * This structure is part of such an  abstraction. The fops layer stores any
- * state is requires in the fd. For eg, the dirent cache for a directory fd_t.
+ * state is requires in the fd. E.g. the dirent cache for a directory fd_t.
  */
 typedef struct nfs_fop_fdcontext {
         pthread_mutex_t         lock;
@@ -92,7 +83,7 @@ nfs_fsync (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
 
 extern int
 nfs_write (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
-           struct iobuf *srciob, struct iovec *vector, int32_t count,
+           struct iobref *srciobref, struct iovec *vector, int32_t count,
            off_t offset, fop_writev_cbk_t cbk, void *local);
 
 extern int
@@ -157,4 +148,21 @@ nfs_read_sync (xlator_t *xl, nfs_user_t *nfu, fd_t *fd, size_t size,
 extern int
 nfs_opendir (xlator_t *nfsx, xlator_t *fopxl, nfs_user_t *nfu, loc_t *pathloc,
              fop_opendir_cbk_t cbk, void *local);
+
+extern int
+nfs_access (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
+            int32_t accesstest, fop_access_cbk_t cbk, void *local);
+extern int
+nfs_lk (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, fd_t *fd,
+        int cmd, struct gf_flock *flock, fop_lk_cbk_t cbk, void *local);
+
+extern int
+nfs_getxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
+              char *name, dict_t *xdata, fop_getxattr_cbk_t cbk, void *local);
+
+extern int
+nfs_setxattr (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu,
+              loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata,
+              fop_setxattr_cbk_t cbk, void *local);
+
 #endif
diff --git a/xlators/nfs/server/src/nfs-inodes.c b/xlators/nfs/server/src/nfs-inodes.c
index 4b8bad717..63d5e8a19 100644
--- a/xlators/nfs/server/src/nfs-inodes.c
+++ b/xlators/nfs/server/src/nfs-inodes.c
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -32,6 +23,14 @@
 
 #include <libgen.h>
 
+#define inodes_nfl_to_prog_data(nflocal, pcbk, fram)                    \
+        do {                                                            \
+                nflocal = fram->local;                                  \
+                fram->local = nflocal->proglocal;                       \
+                *VOID(&pcbk) = nflocal->progcbk;                                \
+                nfs_fop_local_wipe (nflocal->nfsx, nflocal);            \
+        } while (0)                                                     \
+
 void
 nfl_inodes_init (struct nfs_fop_local *nfl, inode_t *inode, inode_t *parent,
                  inode_t *newparent, const char *name, const char *newname)
@@ -49,10 +48,10 @@ nfl_inodes_init (struct nfs_fop_local *nfl, inode_t *inode, inode_t *parent,
                 nfl->newparent = inode_ref (newparent);
 
         if (name)
-                strcpy (nfl->path, name);
+                strncpy (nfl->path, name, NFS_NAME_MAX);
 
         if (newname)
-                strcpy (nfl->newpath, newname);
+                strncpy (nfl->newpath, newname, NFS_NAME_MAX);
 
         return;
 }
@@ -62,15 +61,16 @@ int32_t
 nfs_inode_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                       int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode
                       , struct iatt *buf, struct iatt *preparent,
-                      struct iatt *postparent)
+                      struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = frame->local;
         fop_create_cbk_t        progcbk = NULL;
+        inode_t                 *linked_inode = NULL;
 
         if (op_ret == -1)
                 goto do_not_link;
 
-        inode_link (inode, nfl->parent, nfl->path, buf);
+        linked_inode = inode_link (inode, nfl->parent, nfl->path, buf);
 
 do_not_link:
         /* NFS does not need it, upper layers should not expect the pointer to
@@ -78,10 +78,16 @@ do_not_link:
          */
         fd_unref (fd);
 
-        nfl_to_prog_data (nfl, progcbk, frame);
+        inodes_nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, fd, inode, buf,
-                         preparent, postparent);
+                         preparent, postparent, xdata);
+
+        if (linked_inode) {
+                inode_lookup (linked_inode);
+                inode_unref (linked_inode);
+        }
+
         return 0;
 }
 
@@ -127,21 +133,27 @@ int32_t
 nfs_inode_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno, inode_t *inode,
                      struct iatt *buf, struct iatt *preparent,
-                     struct iatt *postparent)
+                     struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = frame->local;
         fop_mkdir_cbk_t         progcbk = NULL;
+        inode_t                 *linked_inode = NULL;
 
         if (op_ret == -1)
                 goto do_not_link;
 
-        inode_link (inode, nfl->parent, nfl->path, buf);
+        linked_inode = inode_link (inode, nfl->parent, nfl->path, buf);
 
 do_not_link:
-        nfl_to_prog_data (nfl, progcbk, frame);
+        inodes_nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, inode, buf,
-                         preparent, postparent);
+                         preparent, postparent, xdata);
+
+        if (linked_inode) {
+                inode_lookup (linked_inode);
+                inode_unref (linked_inode);
+        }
 
         return 0;
 }
@@ -172,7 +184,7 @@ err:
 
 int32_t
 nfs_inode_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, fd_t *fd)
+                    int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
 
         struct nfs_fop_local    *nfl = NULL;
@@ -186,16 +198,16 @@ nfs_inode_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 /*        else
                 fd_bind (fd);
 */
-        nfl_to_prog_data (nfl, progcbk, frame);
+        inodes_nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, fd);
+                progcbk (frame, cookie, this, op_ret, op_errno, fd, xdata);
         return 0;
 }
 
 
 int
 nfs_inode_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
-                int32_t flags, int32_t wbflags, fop_open_cbk_t cbk, void *local)
+                int32_t flags, fop_open_cbk_t cbk, void *local)
 {
         struct nfs_fop_local    *nfl = NULL;
         fd_t                    *newfd = NULL;
@@ -212,7 +224,7 @@ nfs_inode_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
         }
 
         nfs_fop_handle_local_init (NULL, nfsx, nfl, cbk, local, ret, fd_err);
-        ret = nfs_fop_open (nfsx, xl, nfu, loc, flags, newfd, wbflags,
+        ret = nfs_fop_open (nfsx, xl, nfu, loc, flags, newfd,
                             nfs_inode_open_cbk, nfl);
 
         if (ret < 0)
@@ -234,7 +246,8 @@ int32_t
 nfs_inode_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                       int32_t op_ret, int32_t op_errno, struct iatt *buf,
                       struct iatt *preoldparent, struct iatt *postoldparent,
-                      struct iatt *prenewparent, struct iatt *postnewparent)
+                      struct iatt *prenewparent, struct iatt *postnewparent,
+                      dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_rename_cbk_t        progcbk = NULL;
@@ -247,11 +260,11 @@ nfs_inode_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                       nfl->newpath, nfl->inode, buf);
 
 do_not_link:
-        nfl_to_prog_data (nfl, progcbk, frame);
+        inodes_nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, buf,
                          preoldparent, postoldparent, prenewparent,
-                         postnewparent);
+                         postnewparent, xdata);
         return 0;
 }
 
@@ -284,22 +297,29 @@ int32_t
 nfs_inode_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno, inode_t *inode,
                     struct iatt *buf, struct iatt *preparent,
-                    struct iatt *postparent)
+                    struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_link_cbk_t          progcbk = NULL;
+        inode_t                 *linked_inode = NULL;
 
         if (op_ret == -1)
                 goto do_not_link;
 
         nfl = frame->local;
-        inode_link (inode, nfl->newparent, nfl->path, buf);
+        linked_inode = inode_link (inode, nfl->newparent, nfl->path, buf);
 
 do_not_link:
-        nfl_to_prog_data (nfl, progcbk, frame);
+        inodes_nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, inode, buf,
-                         preparent, postparent);
+                         preparent, postparent, xdata);
+
+        if (linked_inode) {
+                inode_lookup (linked_inode);
+                inode_unref (linked_inode);
+        }
+
         return 0;
 }
 
@@ -330,7 +350,7 @@ err:
 int32_t
 nfs_inode_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                       int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                      struct iatt *postparent)
+                      struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_unlink_cbk_t        progcbk = NULL;
@@ -341,12 +361,13 @@ nfs_inode_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto do_not_unlink;
 
         inode_unlink (nfl->inode, nfl->parent, nfl->path);
+        inode_forget (nfl->inode, 0);
 
 do_not_unlink:
-        nfl_to_prog_data (nfl, progcbk, frame);
+        inodes_nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, preparent,
-                         postparent);
+                         postparent, xdata);
         return 0;
 }
 
@@ -377,7 +398,7 @@ err:
 int32_t
 nfs_inode_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                     struct iatt *postparent)
+                     struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_rmdir_cbk_t         progcbk = NULL;
@@ -388,12 +409,13 @@ nfs_inode_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto do_not_unlink;
 
         inode_unlink (nfl->inode, nfl->parent, nfl->path);
+        inode_forget (nfl->inode, 0);
 
 do_not_unlink:
-        nfl_to_prog_data (nfl, progcbk, frame);
+        inodes_nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, preparent,
-                        postparent);
+                         postparent, xdata);
 
         return 0;
 }
@@ -426,23 +448,30 @@ int32_t
 nfs_inode_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno, inode_t *inode,
                      struct iatt *buf, struct iatt *preparent,
-                     struct iatt *postparent)
+                     struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_mknod_cbk_t         progcbk = NULL;
+        inode_t                 *linked_inode = NULL;
 
         nfl = frame->local;
 
         if (op_ret == -1)
                 goto do_not_link;
 
-        inode_link (inode, nfl->parent, nfl->path, buf);
+        linked_inode = inode_link (inode, nfl->parent, nfl->path, buf);
 
 do_not_link:
-        nfl_to_prog_data (nfl, progcbk, frame);
+        inodes_nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, inode, buf,
-                         preparent, postparent);
+                         preparent, postparent, xdata);
+
+        if (linked_inode) {
+                inode_lookup (linked_inode);
+                inode_unref (linked_inode);
+        }
+
         return 0;
 }
 
@@ -476,22 +505,28 @@ int32_t
 nfs_inode_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                        int32_t op_ret, int32_t op_errno, inode_t *inode,
                        struct iatt *buf, struct iatt *preparent,
-                       struct iatt *postparent)
+                       struct iatt *postparent, dict_t *xdata)
 {
         struct nfs_fop_local    *nfl = NULL;
         fop_symlink_cbk_t       progcbk = NULL;
+        inode_t                 *linked_inode = NULL;
 
         nfl = frame->local;
         if (op_ret == -1)
                 goto do_not_link;
 
-        inode_link (inode, nfl->parent, nfl->path, buf);
+        linked_inode = inode_link (inode, nfl->parent, nfl->path, buf);
 
 do_not_link:
-        nfl_to_prog_data (nfl, progcbk, frame);
+        inodes_nfl_to_prog_data (nfl, progcbk, frame);
         if (progcbk)
                 progcbk (frame, cookie, this, op_ret, op_errno, inode, buf,
-                         preparent, postparent);
+                         preparent, postparent, xdata);
+
+        if (linked_inode) {
+                inode_lookup (linked_inode);
+                inode_unref (linked_inode);
+        }
 
         return 0;
 }
@@ -522,21 +557,20 @@ err:
 
 int32_t
 nfs_inode_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno, fd_t *fd)
+                       int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
 
         struct nfs_fop_local    *nfl = NULL;
         fop_open_cbk_t          progcbk = NULL;
 
-        if ((op_ret == -1) && (fd))
-                fd_unref (fd);
-        else
+        if (op_ret != -1)
                 fd_bind (fd);
 
-        nfl_to_prog_data (nfl, progcbk, frame);
+        inodes_nfl_to_prog_data (nfl, progcbk, frame);
 
         if (progcbk)
-                progcbk (frame, cookie, this, op_ret, op_errno, fd);
+                progcbk (frame, cookie, this, op_ret, op_errno, fd, xdata);
+
         return 0;
 }
 
diff --git a/xlators/nfs/server/src/nfs-inodes.h b/xlators/nfs/server/src/nfs-inodes.h
index 51a23faef..ba7a57124 100644
--- a/xlators/nfs/server/src/nfs-inodes.h
+++ b/xlators/nfs/server/src/nfs-inodes.h
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2010-2011 Gluster, Inc. <http://www.gluster.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _NFS_INODES_H_
@@ -48,7 +39,7 @@ nfs_inode_mkdir (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
 
 extern int
 nfs_inode_open (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *loc,
-                int32_t flags, int32_t wbflags, fop_open_cbk_t cbk,
+                int32_t flags, fop_open_cbk_t cbk,
                 void *local);
 
 extern int
@@ -80,6 +71,6 @@ nfs_inode_mknod (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
                  mode_t mode, dev_t dev, fop_mknod_cbk_t cbk, void *local);
 
 extern int
-nfs_inode_lookup (xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
+nfs_inode_lookup (xlator_t *nfsx, xlator_t *xl, nfs_user_t *nfu, loc_t *pathloc,
                   fop_lookup_cbk_t cbk, void *local);
 #endif
diff --git a/xlators/nfs/server/src/nfs-mem-types.h b/xlators/nfs/server/src/nfs-mem-types.h
index 74499fb08..450b6f2fe 100644
--- a/xlators/nfs/server/src/nfs-mem-types.h
+++ b/xlators/nfs/server/src/nfs-mem-types.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2011 Gluster, Inc. <http://www.gluster.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -37,13 +28,25 @@ enum gf_nfs_mem_types_ {
         gf_nfs_mt_entry3,
         gf_nfs_mt_entryp3,
         gf_nfs_mt_nfs3_fd_entry,
+        gf_nfs_mt_nfs3_fh,
         gf_nfs_mt_nfs_initer_list,
         gf_nfs_mt_xlator_t,
         gf_nfs_mt_list_head,
         gf_nfs_mt_mnt3_resolve,
         gf_nfs_mt_mnt3_export,
+        gf_nfs_mt_int,
+        gf_nfs_mt_mountres3,
+        gf_nfs_mt_mountstat3,
         gf_nfs_mt_inode_q,
-        gf_nfs_mt_subvolumes,
+        gf_nfs_mt_nlm4_state,
+        gf_nfs_mt_nlm4_cm,
+        gf_nfs_mt_nlm4_fde,
+        gf_nfs_mt_nlm4_nlmclnt,
+        gf_nfs_mt_nlm4_share,
+        gf_nfs_mt_aux_gids,
+        gf_nfs_mt_inode_ctx,
+        gf_nfs_mt_auth_spec,
+        gf_nfs_mt_arr,
         gf_nfs_mt_end
 };
 #endif
diff --git a/xlators/nfs/server/src/nfs.c b/xlators/nfs/server/src/nfs.c
index 3fa46662f..75c8fe44e 100644
--- a/xlators/nfs/server/src/nfs.c
+++ b/xlators/nfs/server/src/nfs.c
author	Kaleb S. KEITHLEY <kkeithle@redhat.com>	2013-06-17 13:44:10 -0400
committer	Vijay Bellur <vbellur@redhat.com>	2013-06-26 04:36:47 -0700
commit	f7b1245ca221fd7d4cee626ae358a2ebc933ffdd (patch)
tree	565477a07be9aaf891fbc858f74300d93bc1fdef /extras
parent	183546aa2dbfe3371cf155800e2f70057e95e2bc (diff)